V** -JH- \S #& XS .'* 
 
 ^ *•• »' v< N V 4 TVV .G* %, '»• • ^ <* * Sg Ti* ,G* o 
 
 0>" ** V^SS*" .^""^ 
 
 «5<^ 
 
 
 <, *'7Vi % .G 
 
 *> 4 
 
 VV 
 
 ,♦" ■ 
 
 * 4* % •' 
 
 ^ 
 
 
 ' . . o - . 
 
 / ^ ^f' 4> * . 
 
 'bV 
 
 ** -^ • 
 
 
 ^ ^ *i 
 
 °o 
 
 ' < y <* *^:?t« g* V 
 
 V,^ :£ta %.^* .^^o ^^ 
 
 vv 
 
 1 • A v *^ 
 
 
 ■» o 
 
 v^V -^iM^» VV 
 
 
 ' O V S »i^, %> 4,0* •' 
 
 **<* 
 
 
 
 
 S.-&&X ***£&kS J?.'&&\ 
 
 si 
 
 
 y - . <^ ' - • 
 
 
 
 ^O*. - 
 
 
 
 « o. 
 
 
 
 V^ .-^^- %,** A'-' %■/ #& %^ •» ^/ -A- % «S it 
 
 ^ •.?^:« ^* ^-» -N85<» .4- v >u • 
 
 
 ^ * 4?' v «>* 
 
 o. 'o.»* .A' ' ^ "*^T»' o v " ^ "*•"• "•'*<* "^ TT«* .G y 
 
 < v . , '-*. " • u A' 
 
 v v #V- 
 
 JO. 
 
 ^. * • • • a° 
 
 ■*V ^ > . • • • . *>\ rv^.« 
 
 ' 4> 
 
 vv !^SK: ^^ 
 
 
 V*V °/^"»a^ r U vP^ 
 
**o* 
 
 «b^ 
 
 
 W 
 
 ^o* 
 
 < 
 
 'bV" 
 
 
 
 **0« 
 
 
 
 ■5°* 
 
 ^<J6 
 
 
 
 
 ^ 
 
 ^ ... "*_ 
 
 W 
 
 
 * «? «£>, ^ 
 
 ^c,- 
 
 4°* 
 
 
 
 
 .0° ^a *•.•,!•• a? 
 
 
 ...» .o* *, '••»• < 
 
 .0' % ^- % \/ C K"% 
 
 
 >, *o . , * a 
 
 
 
 ^ ^. 
 
 %^ ^ 
 
 
 4 °>. 
 
 
 ■^ *^r. 
 
 • <^ 
 
BUREAU OF MINES 
 INFORMATION CIRCULAR/1988 
 
 Recent Developments in Metal and 
 Nonmetal Mine Fire Protection 
 
 Proceedings: Bureau of Mines Technology 
 Transfer Seminars, Denver, CO, October 18-19; 
 Detroit, Ml, October 20-21; Las Vegas, NV, 
 November 1-2; and Spokane, WA, November 
 3-4, 1988 
 
 By Staff, Bureau of Mines 
 
 UNITED STATES DEPARTMENT OF THE INTERIOR 
 
Information Circular 
 
 [ / lirjXtJ &%&*, BMJUmtf \hw) 
 
 Recent Developments in Metal and 
 Nonmetal Mine Fire Protection 
 
 Proceedings: Bureau of Mines Technology 
 Transfer Seminars, Denver, CO, October 18-19; 
 Detroit, Ml, October 20-21; Las Vegas, NV, 
 November 1-2; and Spokane, WA, November 
 3-4, 1988 
 
 By Staff, Bureau of Mines 
 
 UNITED STATES DEPARTMENT OF THE INTERIOR 
 Donald Paul Hodel, Secretary 
 
 BUREAU OF MINES 
 T S Ary, Director 
 

 A 
 
 0* 
 
 <\~z 
 
 0(0 
 
 Library of Congress Cataloging in Publication Data: 
 
 Bureau of Mines Technology Transfer Seminars (1988 : Denver, 
 Colo., etc. 
 
 Recent developments in metal and nonmetal mine fire protections. 
 
 (Bureau of Mines Information circular; 9206) 
 
 Includes bibliographical references. 
 
 Supt. of Docs, no.: I 28.27:9206. 
 
 1. Mine fires-Prevention and control-Congresses. I. United States. Bureau of 
 Mines. II. Title. III. Series: Information circular (United States. Bureau of Mines); 
 9206. 
 
 TN295.U4 
 
 [TN315] 
 
 622 s [622'.8] 
 
 88-600299 
 
PREFACE 
 
 In October and November 1988, the Bureau of Mines held technology transfer seminars on 
 metal-nonmetal mine fire protection at Denver, CO, Detroit, MI, Las Vegas, NV, and Spokane, 
 WA. The papers presented at those seminars are contained in this Information Circular. The papers 
 highlight Bureau research to improve mine fire protection. Areas addressed by this research, and 
 published in this volume, include fire detection and instrumentation, fire warning, fire suppres- 
 sion, diesel equipment, spontaneous combustion, and toxicity analysis of combustion products. Cer- 
 tain of the findings are also applicable to underground coal mining and to surface mining operations. 
 
 The technology transfer seminar used as a forum for the transfer of this research is one of 
 the many mechanisms used by the Bureau of Mines in its efforts to move research developments, 
 technology, and information resulting from its programs into industrial practice and use. To learn 
 more about the Bureau's technology transfer program and how it can be useful to you, please write 
 or telephone: 
 
 Bureau of Mines 
 
 Office of Technology Transfer 
 
 2401 E Street, NW. 
 
 Washington, DC 20241 
 
 202-634-1224 
 
Ill 
 
 CONTENTS 
 
 Page 
 
 Preface i 
 
 Abstract 1 
 
 Introduction 2 
 
 Statistical Analysis of Metal and Nonmetal Mine Fire Incidents in the United States From 1950 to 1984, 
 
 by Shail J. Butani and William H. Pomroy 3 
 
 Computer Models of Underground Mine Ventilation and Fires, by Rudolf E. Greuer 6 
 
 Mine Fire Detection Systems: A Primer, by Charles D. Litton 15 
 
 Fire Detection Systems for Noncoal Underground Mines, by W.H. Pomroy 21 
 
 Diesel-Discriminating Fire Sensor, by Charles D. Litton 28 
 
 Computer-Aided Mine Fire Sensor Data Interpretation in Real Time, by L.W. Laage, 
 
 W.H. Pomroy, and A.M. Bartholomew 33 
 
 Reliability of Underground Mine Fire Detection and Suppression Systems, by Steven G Grannes 42 
 
 Diesel Exhaust Conditioning Systems for Fire and Explosion Control in Gassy Mines, by Kenneth L. Bickel 49 
 
 Spontaneous Combustion Susceptibility of Sulfide Minerals, by G.W. Reimers and W.H. Pomroy 54 
 
 Emission Products From Wood Crib and Transformer Fluid Fires, by Margaret R. Egan 61 
 
 Utilization of Smoke Properties for Predicting Smoke Toxicity, by Maria I. De Rosa and Charles D. Litton 72 
 
 Electromagnetic Fire Warning System for Underground Mines, by Kenneth E. Hjelmstad and William H. Pomroy 78 
 
LIST OF UNIT OF MEASURE ABBREVIATIONS USED IN THIS REPORT 
 
 A 
 
 angstrom 
 
 km/h 
 
 kilometer per hour 
 
 A 
 
 ampere 
 
 kW 
 
 kilowatt 
 
 A/m 
 
 ampere per meter 
 
 lb 
 
 pound 
 
 acre-ft 
 
 acre-foot 
 
 L/min 
 
 liter per minute 
 
 Btu/min 
 
 British thermal unit per minute 
 
 m 
 
 meter 
 
 cm 
 
 centimeter 
 
 mCi 
 
 millicurie 
 
 cm 2 /p 
 
 square centimeter per particle 
 
 mg 
 
 milligram 
 
 cm 2 /(p«g) 
 
 square centimeter per particle per gram 
 
 mg/m 3 
 
 milligram per cubic meter 
 
 cm 3 /min 
 
 cubic centimeter per minute 
 
 mho/m 
 
 mho per meter 
 
 °C 
 
 degree Celsius 
 
 min 
 
 minute 
 
 °C/min 
 
 degree Celsius per minute 
 
 mm 
 
 millimeter 
 
 dB/m 
 
 decibel per meter 
 
 m/s 
 
 meter per second 
 
 ft 
 
 foot 
 
 m 2 
 
 square meter 
 
 ft/min 
 
 foot per minute 
 
 m 3 
 
 cubic meter 
 
 °F 
 
 degree Fahrenheit 
 
 m 3 /s 
 
 cubic meter per second 
 
 g 
 
 gram 
 
 /iCi 
 
 microcurie 
 
 gal 
 
 gallon 
 
 Mg/m 3 
 
 microgram per cubic meter 
 
 g/cm 3 
 
 gram per cubic centimeter 
 
 ^m 
 
 micrometer 
 
 g/g 
 
 gram per gram 
 
 p/cm 3 
 
 particle per cubic meter 
 
 g/kJ 
 
 gram per kilojoule 
 
 pet 
 
 percent 
 
 g/(m 3 «ppm) 
 
 gram per cubic meter per part per 
 
 p/kJ 
 
 particle per kilojoule 
 
 
 million 
 
 psi 
 
 pound per square inch 
 
 g/min 
 
 gram per minute 
 
 ppm 
 
 part per million 
 
 g/s 
 
 gram per second 
 
 ppm/g 
 
 part per million per gram 
 
 h 
 
 hour 
 
 ppm/min 
 
 part per million per minute 
 
 hp 
 
 horsepower 
 
 s 
 
 second 
 
 Hz 
 
 hertz 
 
 St 
 
 short ton 
 
 in 
 
 inch 
 
 V 
 
 volt 
 
 kg 
 
 kilogram 
 
 V dc 
 
 volt, direct current 
 
 kHz 
 
 kilohertz 
 
 W 
 
 watt 
 
 kJ/g 
 
 kilojoule per gram 
 
 yr 
 
 year 
 
RECENT DEVELOPMENTS IN METAL AND NONMETAL 
 MINE FIRE PROTECTION 
 
 Proceedings: Bureau of Mines Technology Transfer Seminars, 
 
 Denver, CO, October 18-19; Detroit, Ml, October 20-21; 
 
 Las Vegas, NV, November 1-2; and Spokane, WA, November 3-4, 1988 
 
 By Staff, Bureau of Mines 
 
 ABSTRACT 
 
 Great strides have been made in recent years to reduce the disaster potential of underground 
 mine fires. However, mines can still be caught unprepared for a fire emergency. Fires can grow 
 too large before they are detected, warning systems can be too slow and uncertain to reliably signal 
 the danger, and suppression systems can be inadequate to extinguish the flames. New mining systems 
 and equipment may create unanticipated fire hazards, and new materials may generate highly toxic 
 combustion products. This report contains papers that summarize recent significant developments 
 from the Bureau of Mines mine fire protection research program relating to these problems. Cer- 
 tain of these findings are also applicable to surface mining operations. The papers fall into the 
 general categories of fire detection and instrumentation, fire warning, fire suppression, diesel equip- 
 ment, spontaneous combustion, and toxicity analysis of combustion products. 
 
INTRODUCTION 
 
 No peril is more feared by miners than an underground fire. 
 Fresh air is limited and the workings can rapidly fill with choking 
 smoke and fire gases. Careful attention to fire prevention is the first 
 priority . Special precautions must be taken to limit ignition and fuel 
 sources underground. In the unlikely event that a fire does occur, 
 quick action is essential — miners must be warned and evacuated, 
 and fire-fighting operations must be initiated. 
 
 Great strides have been made in recent years to reduce the 
 disaster potential of underground fires, as evidenced by the steady 
 decline in the number of fire incidents and the number of related 
 injuries and fatalities recorded over the past four decades. However, 
 mines can still be caught unprepared for a fire emergency. 
 
 Often, fires grow too large before they are detected, warning 
 systems are too slow and uncertain to reliably signal the danger, 
 and suppression systems are lacking or are inadequate to extinguish 
 the flames. In addition, new mining systems and equipment may 
 create unanticipated fire hazards and new materials may generate 
 highly toxic combustion products. The result is loss of life, damage 
 to mining facilities and equipment, and loss of valuable minerals 
 resources. 
 
 Since 1978, about 10 underground metal and nonmetal mine 
 
 fires have been reported annually to the Mine Safety and Health 
 
 Administration (MSHA). Each year another estimated 100 to 150 
 
 fires occur that are legally "nonreportable" because they last less 
 
 . than 30 min or do not cause an injury. With nearly 20,000 mine 
 
 accidents occurring annually in the United States, mine fires are, 
 by comparison, quite rare events. But unlike most mine accidents, 
 every fire has the potential to develop into a mine disaster. It is 
 also significant to note that the fire incidence rate, or the number 
 of fires per worker-hour of exposure, is actually higher in 
 underground mines than for aboveground industrial occupancies. 
 
 Since its creation in 1910, the Bureau of Mines has been com- 
 mitted to developing the necessary mining and safety technology 
 to reduce or eliminate fire hazards in underground mines. Indeed, 
 much of the improvement in mine fire safety performance during 
 this century can be traced to pioneering research by the Bureau. 
 This tradition continues to the present with a vigorous program of 
 basic and applied research aimed at devising practical solutions to 
 the industry's most pressing mine fire problems. This effort 
 represents a mutually beneficial partnership of the Bureau, MSHA, 
 and the mining community. 
 
 The papers contained in this Information Circular summarize 
 recent significant developments from the Bureau's mine fire pro- 
 tection research program that are relevant to underground metal 
 and nonmetal mining; certain of these findings are also applicable 
 to underground coal mining and to surface mining operations. The 
 papers fall into the general categories of fire detection and instrumen- 
 tation, fire warning, fire suppression, diesel equipment, spontaneous 
 combustion, and toxicity analysis of combustion products. 
 
STATISTICAL ANALYSIS OF METAL AND NONMETAL MINE FIRE INCIDENTS 
 IN THE UNITED STATES FROM 1950 TO 1984 
 
 By Shail J. Butani 1 and William H. Pomroy 2 
 
 ABSTRACT 
 
 This paper presents the results of a Bureau of Mines analysis of Mine Safety and Health Ad- 
 ministration (MSHA) mine fire reports for the 1950-84 period. The analysis also includes non- 
 reportable fires (less than 30-min duration and no injury) to show the magnitude of the problem. 
 The most frequent ignition sources, burning substances, equipment types, fire locations, and suc- 
 cessful extinguishing agents are discussed. Fire incidence rates in both surface and underground 
 metal and nonmetal mines are not decreasing. New mining technology introduces new fire hazards 
 into the workplace; fire safety emphasis must focus on newly emerging mining technologies in 
 order to reduce incidence rates. 
 
 INTRODUCTION 
 
 In support of the Bureau's program of mine fire protection 
 research, two separate baseline studies of fire incidents have been 
 conducted. The first study addressed coal mine fires 3 and the sec- 
 ond addressed metal and nonmetal mine fires. 4 Together, these two 
 reports provide a comprehensive factual summary of the mining 
 industry's fire experience. 
 
 The purpose of this paper is to summarize the most significant 
 findings of the second study, which analyzed separately all official 
 MSHA mine fire reports prepared during the 1950-84 period, plus 
 accounts of selected nonreportable fires (i.e., less than 30 min and 
 no injury) and mine safety director hazard opinion data. Nonreport- 
 able fires were included to show the true magnitude of the fire 
 problem. 
 
 Because metal and nonmetal mines have been legally required 
 to report fires to MSHA only since 1968, MSHA files prior to that 
 date are incomplete. Although some fires were reported prior to 
 
 'Mathematical statistician (now with Bureau of Labor Statistics, Washington. DC) 
 
 2 Group supervisor. 
 Twin Cities Research Center, Bureau of Mines. Minneapolis. MN. 
 
 'McDonald. L. B., and W. H. Pomroy. A Statistical Analysis of Coal Mine Fire 
 Incidents in the United States From 1950 to 1977. BuMines IC 8830, 1980, 42 pp. 
 
 4 Butani, S. J., and W. H. Pomroy. A Statistical Analysis of Metal and Nonmetal 
 Mine Fire Incidents in the United States From 1950 to 1984. BuMines IC 9132, 1987. 
 41 pp. 
 
 1968, doubtless a great many were not. Also, the reporting regula- 
 tions that took effect in 1968 specify that only fires lasting 30 min 
 or longer or involving an injury need to be reported. However, fires 
 lasting less than 30 min and involving no injury pose a significant 
 hazard, and much can be learned from such incidents. MSHA fire 
 reports are thus limited in scope by MSHA's legal authority. To 
 provide a more comprehensive data base, it was necessary to gather 
 and analyze mine company records of nonreportable fires. Opin- 
 ion data from mine safety directors were collected and separately 
 analyzed in an effort to broaden the discussion and to characterize 
 and rank mine fire hazards in general. 
 
 Because mine fires are relatively rare events, it is desirable 
 not only to analyze the fires themselves, but also the near-misses 
 (which occur much more frequently) and the unsafe conditions that 
 could give rise to future fires. The opinion data give the necessary 
 insight into near-misses and unsafe conditions. 
 
 Where possible, both reported and nonreportable fires were 
 analyzed by time trends (1950-67, 1968-77, 1978-84), ore type, 
 ignition source, burning substance, location in mine, equipment in- 
 volved, means of detection, duration, number of injuries, number 
 of fatalities, mining method, and successful extinguishing agent. 
 The analysis was performed separately for underground fire inci- 
 dents and for surface fire incidents occurring at surface mines or 
 at surface locations of an underground mine. 
 
RESULTS 
 
 Major findings of the study appear in tables 1 and 2. The most 
 frequent ignition sources, burning substances, equipment types, 
 locations, and successful extinguishing agents of reported and non- 
 reportable fires are discussed in the following sections. 
 
 IGNITION SOURCE 
 
 The most frequent ignition source in underground mine fires 
 is electricity. This is true for both reported and nonreportable fires 
 
 Table 1.— Major Study Findings of Reported Fires 
 
 Category 
 
 1950-77 
 
 1 978-84 
 
 Overall 
 
 UNDERGROUND 
 
 Ore type Iron, copper Lead-zinc, salt Iron, copper, lead-zinc, salt. 
 
 Ignition source' Electrical, welding Electrical, engine heat Electrical, welding, engine heat. 
 
 Burning substance 1 Timber, insulation, combustible 
 
 liquids. 
 
 Location 1 Haulageway-drift, shaft-raise- 
 winze. 
 
 Combustible liquids, timber, 
 insulation. 
 
 Haulage-drift, shaft-raise- 
 winze. 
 
 Timber, combustible liquids, 
 insulation. 
 
 Haulageway-drift, shaft-raise- 
 winze. 
 
 Equipment involved 1 Mobile, electrical Mobile, electrical Mobile, electrical. 
 
 Means of detection 1 Workers (not immediate), operators- Operators-workers (immediate), 
 
 workers (immediate). workers (not immediate). 
 
 Operators-workers (immediate), 
 workers (not immediate). 
 
 Duration h. . 24+, 1+ to 4 1+ to 4, 24+ 1+ to 4, 24+. 
 
 Successful extinguishing agent 1 ... Water-dry chemical Water-dry chemical Water-dry chemical. 
 
 SURFACE 
 
 Ore type Iron, crushed limestone Crushed limestone, iron Crushed limestone, iron. 
 
 Ignition source 1 Electrical, welding, engine heat. . . . Engine heat, welding, electrical. . . . Engine heat, electrical, welding. 
 
 Burning substance 1 Combustible liquids, construction 
 
 material. 
 
 Combustible liquids, construction 
 material. 
 
 Combustible liquids, construction 
 material. 
 
 Location 1 Surface building Surface, surface building Surface building, surface. 
 
 Equipment involved 1 Mobile, electrical Mobile, conveyor Mobile, conveyor, electrical. 
 
 Means of detection 1 Operators-workers (immediate), 
 
 workers (not immediate). 
 
 Operators-workers (immediate), 
 workers (not immediate). 
 
 Operators-workers (immediate), 
 workers (not immediate). 
 
 Duration h . 
 
 Successful extinguishing agent 1 . . . 
 
 to 0.5 to 0.5 to 0.5. 
 
 Water, burned out Water, dry chemical Water, burned out. 
 
 1 Factors listed in sequence of significance. 
 
 Table 2.— Major study findings of nonreportable fires 
 
 (Factors listed in sequence of significance) 
 
 Category 
 
 Underground 
 
 Surface 
 
 NONREPORTABLE FIRES 
 
 Ignition source 
 
 Burning substance 
 
 Location 
 
 Equipment involved 
 
 Successful extinguishing agent. 
 
 Electrical, welding, friction Welding, electrical, engine heat. 
 
 Combustible liquids, wiring insulation, timber .... Combustible liquids, insulation. 
 
 Haulageway-drift, substation Surface, surface building. 
 
 Mobile, electrical, maintenance-shop Mobile. 
 
 Dry chemical, cut off electrical power, water Dry chemical. 
 
 OPINION DATA 
 
 Ignition source 
 
 Burning substance 
 
 Successful extinguishing agent. 
 
 Welding, electrical Welding, engine heat. 
 
 Combustible liquids, wiring insulation Combustible liquids, rubber (hose or belt). 
 
 Dry chemical, water Dry chemical, water. 
 
and for all three major time periods. It is also the primary cause 
 of underground injury fires. In recent years, however, engine heat 
 has become the leading cause of injury fires. Also significant is 
 the increasing number of electrical fires that occur on diesel- 
 powered mobile vehicles. Engine heat is the leading ignition source 
 for surface fires that were reported, while for nonreportable sur- 
 face fires the ignition source was welding. 
 
 BURNING SUBSTANCE 
 
 The most frequent burning substance in reported underground 
 fires is timber, followed by combustible liquids and insulation. In 
 nonreportable fires, combustible liquids, wiring insulation, and 
 timber are involved with about equal frequency. For reported sur- 
 face fires, the most frequent burning substance is combustible 
 liquids, followed by construction material. In nonreportable sur- 
 face fires, combustible liquids are most frequently involved. 
 
 LOCATION 
 
 Reported underground fires occur along haulageways or in 
 drifts where electrical or diesel equipment is concentrated. Non- 
 reportable fires also occur in these locations more frequently than 
 at any other. 
 
 Reported surface fires occur primarily in mill buildings, and 
 nonreportable surface fires occur primarily on mobile equipment 
 
 along haulage roads or in the pit area. In recent times, however, 
 the reported surface fires are about evenly split between surface 
 buildings and surface areas other than buildings. 
 
 EQUIPMENT INVOLVED 
 
 The most frequent equipment involved in reported and non- 
 reportable underground and surface fires is mobile type such as 
 load-haul-dump vehicles. In underground fires (reported and non- 
 reportable), the second most frequently involved equipment is of 
 electrical type. For the reported surface fires, it is conveyors in 
 the most recent time period (1978-84), and electrical in the 1950-77 
 period. 
 
 SUCCESSFUL EXTINGUISHING AGENT 
 
 The most frequently successful extinguishing agent for reported 
 fires is water. For nonreportable fires, dry chemical hand-portable 
 fire extinguishers are used most often. This is consistent with the 
 duration of reportable fires. First attack on a fire is generally with 
 a hand-portable extinguisher. If the attempt is successful, then the 
 fire is most likely extinguished at the nonreportable stage. If the 
 fire has grown in size or initial extinguishing attempts prove un- 
 successful, then the fire will probably become reportable and water 
 is used as an extinguishing agent. 
 
 CONCLUSIONS 
 
 Several important conclusions can be drawn from the data of 
 the metal-nonmetal mine fire study. First and foremost, fire inci- 
 dence rates in both surface and underground mines are not declin- 
 ing. Despite the considerable efforts of mine safety personnel and 
 focused regulatory action, little progress toward reducing the in- 
 cidence of fire is apparent. While uncertain reporting during the 
 earlier time periods could be blamed for this apparent lack of prog- 
 ress, the latest time period (1978-84), for which fire reports are 
 believed to be quite complete, is not so easily rationalized, and is 
 therefore quite disturbing. One possible explanation, which is sup- 
 ported to some extent by the data on ignition sources, burning 
 substance, and equipment involved, is that fire hazards are chang- 
 ing as mining methods, materials, and equipment evolve. As specific 
 
 fire hazards are recognized and corrected, new mining technology 
 introduces other hazards into the workplace. This explanation sug- 
 gests that the present level of fire safety effort may not succeed 
 in reducing fire incidence rates, and that an accelerated pace of ac- 
 tivity with particular focus on newly emerging mining technolo- 
 gies is required if incidence rates are to be reduced. 
 
 Another observation also relates to the changing patterns of 
 data evident over the three time periods analyzed (1950-67, 1968-77, 
 1978-84). Conclusions regarding the relative importance of a given 
 fire hazard for one period do not necessarily hold for subsequent 
 periods, suggesting the value of regular updates to the fire incident 
 data base. Timely collection, analysis, and publication of such data 
 will help ensure that fire safety efforts address the greatest needs. 
 
COMPUTER MODELS OF UNDERGROUND MINE VENTILATION AND FIRES 
 
 By Rudolf E. Greuer 1 
 
 ABSTRACT 
 
 The design of mine fire emergency plans requires that the contamination of mines by fumes, 
 and the mutual interaction of fires and ventilation systems be precalculated. Several computer models 
 were developed by the Bureau of Mines for this purpose during the last decade. The models advanced 
 in sophistication from the transient-state simulation of fume concentrations for the early stages of 
 a fire to the steady-state simulation of airflow rates, concentrations, and temperatures for fully 
 developed fires, and finally to the complete transient-state simulation of fires and ventilation systems. 
 The principal features of these models are described. 
 
 INTRODUCTION 
 
 To calculate the airflow distribution in mine ventilation systems 
 as a result of fans, thermal forces, and flow resistances is a for- 
 midable mathematical problem. It comprises the solution of twice 
 as many equations as there are airways and half of these equations 
 are square equations. This sort of problem led to the design of special 
 analog computers in the fifties and sixties and, from the early six- 
 ties on, to the increasing use of electronic digital computers. With 
 the rapidly increasing availability and capacity of electronic digital 
 computers, airflow rate and pressure loss distribution calculations, 
 commonly called ventilation network calculations, have become 
 routine, and a great number of computer programs exist for this 
 purpose. Practically all the programs are capable of performing the 
 required calculations, although differences exist in how the square 
 equations are linearized, the mass conservation law is introduced 
 and observed, the fan characteristics are simulated, and the ther- 
 mal drafts are considered. All of the programs are based on steady- 
 state conditions. A few have rather rudimentary sections for con- 
 centration and temperature calculations, as far as the literature allows 
 such a judgment. 
 
 Mine ventilation control and mine fire detection and fighting 
 are inseparable. Mine fires produce gases and heat, which the ven- 
 tilation systems transport through the mines. The gases can be 
 poisonous or explosive. The heat can cause ventilation disturbances, 
 which take the gases along unexpected routes or affect the forma- 
 tion of explosive methane mixtures. 
 
 Of greatest concern in the past were the fire-generated ven- 
 tilation disturbances. Ventilation engineers developed a large number 
 of methods, by manual calculation, to detect potentially unstable 
 airways with airflow reversals in case of a fire. When the analog 
 and electronic digital computers became available for ventilation 
 planning, they were almost immediately used for this purpose also. 
 The expected fire-generated ventilating pressures were manually 
 inserted into the network simulations, with their values usually 
 obtained from experience or from rough calculations. The mutual 
 influence of fire intensities and ventilation conditions were not taken 
 into account. If gas concentrations were calculated at all, then only 
 for the case that no recirculation existed. All calculations were, as 
 in conventional network calculations, based on steady-state condi- 
 tions or based on the assumption that no changes with time occur. 
 If changes with time happen, one would have transient-state 
 conditions. The full potential of computers was for a long time not 
 utilized, mainly because the task seemed to be too demanding. 
 
 This attitude changed gradually. During the last 15 yr the 
 Bureau supported the development of a number of pertinent 
 computer models for the interaction of mine fires and ventilation 
 systems. Of these, this paper will describe the models considered 
 as particular benchmarks. 
 
 TRANSIENT-STATE CONCENTRATION SIMULATION 
 
 JUSTIFICATION FOR PERFORMING THIS WORK 
 
 The assumption of steady-state conditions may be good enough 
 for the determination of airflow rate and temperature distributions, 
 because airflow rate changes are caused by temperature changes, 
 
 'Mining engineer. Twin Cities Research Center, Bureau of Mines, Minneapolis, 
 MN; professor of mining engineering, MI Technological Univ., Houghton, MI. 
 
 and temperature changes are observed in the immediate vicinity 
 of the fire only. Changed airflow rate and temperature distribu- 
 tions are, therefore, reached almost instantaneously and time is a 
 minor factor. 
 
 Time is definitely an important factor with concentration 
 distributions. Gases travel with the ventilation currents, which 
 normally means speeds of less than 3 to 4 km/h. Many airways 
 downwind of a fire can, for a considerable time after the start of 
 
the fire, remain clear of gases. In many cases, the increase of airway 
 gas concentration happens only gradually before steady-state con- 
 ditions are reached. Assuming steady-state concentration distribu- 
 tions in fire emergency planning frequently means needlessly 
 excluding airways as escapeways, although they are perfectly safe 
 for a long time. 
 
 RESULTING PROGRAM VERSIONS 
 
 be helpful for the design of fire escape plans if an early fire warning 
 system allows for the evacuation of the mine before the fire has 
 progressed beyond its initial stage. 
 
 In the second case, the determination of the airflow rates has 
 to precede the concentration calculations. This requires detailed 
 information on the ventilation system and some ventilation exper- 
 tise by the program user. The determination of airflow rates will 
 be described in a subsequent section of this paper. 
 
 A pertinent program for concentration distributions was com- 
 pleted in 1981 (7-2). 2 Program sections for the exposure simula- 
 tion of escaping miners were completed in 1983 (3). Provisions 
 for the cooperation with computer programs for the simulation of 
 escape movements of miners, based on warning times and travel 
 speeds, were added in 1983 also (4), and mobile contaminant sources 
 were included in 1984 (3). 
 
 The program is based on the assumption of constant airflow 
 rates. These can be the airflow rates prevailing in the early stages 
 of a fire or the airflow rates resulting from the equilibrium of fire- 
 generated (thermal) and other (fan, airway resistances) ventilating 
 forces. In the first case, very little input data and expertise on the 
 part of the program user are required. The only ventilation infor- 
 mation needed is data on the network configuration (airway and 
 node identification numbers), airflow rates, and airway lengths and 
 cross-sectional areas. The program should be of use for all such 
 cases where fire-generated ventilation disturbances do not yet occur, 
 in particular for the layout of fire detection systems. It should also 
 
 2 Italic numbers in parentheses refer to items in the list of references at the end of 
 this paper. 
 
 PROGRAM DESCRIPTIONS 
 
 The simulation of concentration distributions does not involve 
 physical principles other than the law of mass conservation. Addi- 
 tionally, the assumptions are made that perfect mixing of air cur- 
 rents in nodes and no longitudinal diffusion in airways exists and 
 that flow velocities in planes perpendicular to flow directions are 
 equal. 
 
 The mine atmosphere is divided into control volumes of 
 homogeneous concentrations, which advance with the flow through 
 the ventilation system. When control volumes meet at airway junc- 
 tions, they are extinguished and new control volumes are formed. 
 Because many junctions can be reached via different paths with dif- 
 ferent amounts of dilution having occurred, the number of control 
 volumes can become quite large, depending on the type of ventila- 
 tion system. When recirculation occurs, the number can virtually 
 become infinity with smaller and smaller concentration differences 
 between the newly formed control volumes. Figure 1 gives the 
 example of a simple, idealized system with recirculation, which 
 indicates the complexity of the problem. 
 
 Contaminant 
 source 
 
 After 15 min 
 
 Contaminant 
 source 
 
 After 30 min 
 
 Total 
 
 1st 
 
 rong 
 
 Key 
 Medium 1®®@©1 Weak 
 
 Absent 
 
 Figure 1 .—Concentration distribution with recirculation contamination. 
 
The problem of simulation is not one of mathematics but of 
 sorting. All control volumes that arrive at a junction have to be 
 detected, and the sequence of their arrival has to be determined. 
 The new control volumes, which are generated in the junctions, 
 have to be advanced into the outgoing airways together with checks 
 to see if they affect any other junctions. If they do, part of the com- 
 putations have to be repeated. 
 
 The main difficulty in writing this program was, therefore, not 
 one of mathematics but one of keeping the computing time in 
 tolerable limits. Experience showed that the choice of the time in- 
 crement in which the control volumes are advanced is of great im- 
 portance. If the increment is too large, control volumes may have 
 to be passed through more than one junction, which lengthens and 
 complicates calculations. If the increment is too short, a large 
 number of advancements has to be performed. The solution was 
 to let the program select the optimal length of the time interval as 
 a function of the network type. 
 
 The basic organization of the program is shown in figure 2. 
 Its core is a triple-nested DO-loop with a joint starting point for 
 all three loops. The innermost loop updates state and location of 
 the control volumes, an intermediate loop initiates the updating in 
 the chosen time intervals, and the outermost loop provides for out- 
 put in user-specified time intervals. 
 
 By far the largest part of the program is occupied by the 
 innermost loop. It is here that the control volumes are advanced, 
 extended, shortened, deleted, or generated (fig. 3). The deletion 
 and generation of control volume happens mainly by mixing in net- 
 
 work junctions. It is, for this purpose, necessary to find the proper 
 sequence of control volumes arrivals injunctions, which leads to 
 a continual process of cross-referencing among junctions. 
 
 Besides initiating the updating of control volumes, the inter- 
 mediate time loop has to fulfill a large number of other functions. 
 It checks, for instance, if airways have, at least temporarily, reached 
 steady -state conditions. It keeps track of available storage for con- 
 trol volumes to prevent loss of data, and it organizes the output 
 in such a way that repetitious production of results is avoided as 
 far as possible. 
 
 Program sections added in 1983 (J) allow one to calculate the 
 fume or other contaminant exposure of miners. This looks, in prin- 
 ciple, like a simple task that only requires determining the miners' 
 locations and the pertinent contaminant concentration. The reality 
 is that one deals with moving miners in a moving atmosphere. Dur- 
 ing the time intervals that determine the updating of the ventilation 
 system, miners may stay in junctions with or without concentra- 
 tion changes occurring. They may travel within one or several air- 
 ways. In each airway they may travel in the same or the opposite 
 direction of the air and with the same speed or faster or slower 
 than the air. Their original travel or escape plans will need modifica- 
 tion depending on the state of the ventilation system. Miners will, 
 in smoke obscured airways and under apparatus, travel with dif- 
 ferent speeds than unimpeded miners. This all adds up to quite 
 entwined calculations, and the added program sections for exposure 
 calculations are almost as lengthy as the original program. 
 
 Airway loop 
 
 Input 
 
 Determine 
 
 optional time 
 
 increment 
 
 Update 
 
 location of 
 
 control volumes 
 
 Airway 
 loop 
 
 Time 
 loop 
 
 Output 
 loop 
 
 Increase time 
 
 Output 
 
 concentration 
 
 distribution 
 
 No 
 
 
 < 
 
 < 
 
 
 
 
 Update airways 
 
 (delete, generate, 
 
 advance control 
 
 volumes) 
 
 
 
 1 
 
 
 
 Has airway end 
 been reached? 
 
 
 Go to next 
 
 No 
 
 nirway 
 
 
 
 Yes 
 
 
 
 ' 
 
 
 
 Find all control 
 
 volumes affecting 
 
 same node 
 
 
 
 
 1 
 
 ' 
 
 
 Delete, generate, 
 
 advance control volumes 
 
 in proper sequence 
 
 
 
 I 
 
 
 
 
 Has a 
 
 airwa 
 
 been re 
 
 nother 
 y end 
 cached? 
 
 —Yes — 
 
 Go to next 
 
 airway 
 
 Figure 2.— Basic principle of transient-state concentration 
 simulation. 
 
 Figure 3.— Update of control volumes in airway loop. 
 
STEADY-STATE FIRE SIMULATIONS 
 
 It was stated in the ' 'Introduction' ' section of this paper that 
 possible ventilation disturbances caused by fires have always been 
 of the greatest concern in fire emergency planning. This may be 
 less the case where shallow deposits are mined and fans provide 
 brisk air currents with significant pressure differences between net- 
 work parts. Where mine workings extend, however, over larger 
 elevation differences and where the ventilation is weak, such dis- 
 turbances have been the cause of large mine disasters. The 1928 
 mine fire at Roche-la-Moliere (France), which killed 48 miners, 
 is a classical and often quoted example. The first computer model 
 (5) on the interaction of mine fires and mine ventilation systems 
 focused, therefore, on the assessment of these disturbances. 
 
 To be able to make use of the existing programs for ventila- 
 tion network calculations and to obtain a workable model in com- 
 paratively short time, steady-state conditions for the ventilation as 
 well as for the fire were assumed. This may come close to reality, 
 it may in other cases mean a gross simplification. The resulting 
 program turned out to be useful nevertheless and became popular 
 because of its simplicity. For this reason and since it is part of more 
 advanced transient-state computer models, it shall be described here 
 in some detail. 
 
 UNDERLYING PHYSICAL PRINCIPLES OF 
 STEADY-STATE FIRE SIMULATIONS 
 
 If the terminology of ventilation engineers is used, one can say 
 that fires have two effects on ventilation systems: a throttling effect 
 caused by the volume increase of the air passing through the fire 
 zone; and a natural draft effect caused by the conversion of heat 
 into mechanical energy. 
 
 The throttling effect is easy to assess. The energy requirement 
 per unit mass to overcome friction losses is proportional to the square 
 of the flow velocity. The latter is in turn proportional to the specific 
 volume or the absolute temperature. The energy requirements to 
 transport air through an airway are, therefore, proportional to the 
 square of the absolute temperature. If a fire lets the temperature 
 increase, the same amount of energy will transport less air. 
 
 Natural draft effects, or the conversion of heat into mechanical 
 energy, can occur where the air changes its pressure and its specific 
 volume and where it goes through a cyclic process. Every loop of 
 the ventilation system can be considered as a cyclic process and 
 can, therefore, develop natural draft effects. The amount of heat 
 converted into mechanical energy is, according to the first law of 
 thermodynamics, equal to the integral of the products of specific 
 volume times pressure change. 
 
 The determination of numerical values of this integral is 
 cumbersome because of the continual pressure changes in ventila- 
 tion systems and owing to the fact that the specific volume cannot 
 be directly measured. A good approximation is possible in most 
 cases by establishing a functional relationship between this integral 
 and an analogous integral of the products of temperature times eleva- 
 tion change, which is easier to determine. 
 
 Precalculations of the throttling and natural draft effect require, 
 therefore, the precalculation of temperatures. These are functions 
 of the airflow distribution, which determines the oxygen supply of 
 the fire, the heat transfer to the airway walls, and the mixing of 
 air currents injunctions. The airflow distribution, on the other hand, 
 is a function of the temperature-induced throttling and natural draft 
 effects. It was this interaction between fires and ventilation systems, 
 together with the occurrence of recirculation, which was for many 
 years considered to be too complicated for computer simulations. 
 
 RESULTING PROGRAM 
 
 This program combines conventional network (airflow rate, 
 pressure loss) calculations with concentration and temperature 
 calculations and takes the throttling and natural draft effects 
 automatically into account. It tries to establish the equilibrium at 
 which fire-generated thermal forces (throttling and natural draft) 
 are balanced by the other ventilation forces (fans, resistances). 
 
 Figure 4 shows the program organization. The beginning is 
 a network calculation for the prefire state, to ascertain that the input 
 data are correct. Based on the type of fuel, fuel loading, and air 
 supply, the heat and contaminant production is calculated next. This 
 is followed by a calculation of the temperature and concentration 
 distribution, based on the original airflow rates. The fire-generated 
 thermal forces (throttling and natural draft effect) are then deter- 
 mined and inserted into a new network calculation. This sequence 
 of calculations is repeated until the calculated thermal forces remain 
 constant. This means that the equilibrium has been found and, 
 temporarily, steady-state conditions for the airflow rate, pressure 
 loss, and temperature distribution have been established. 
 
 This approach can be justified because changes of the 
 temperature distribution are observed in the vicinity of the fire and 
 short distances downwind from it only. Fire-caused thermal forces 
 are, therefore, generated almost instantaneously, and changed 
 airflow rate and temperature distributions also establish themselves 
 
 Input 
 
 Network calculation 
 (airflow rate and pressure loss) 
 
 Heat and contaminant 
 production calculation 
 
 Temperature and 
 concentration calculation 
 
 Determination of thermal forces 
 
 Are thermal forces different 
 from those used in 
 network calculation? 
 
 — Yes J 
 
 No 
 
 Output 
 
 Figure 4.— Basic principle of steady-state simulation of mine 
 fire-ventilation interaction. 
 
10 
 
 almost instantaneously. This does not apply to the concentration 
 distributions, however, and only in rare cases to fire intensities. 
 Even if the fire intensity does not change, the temperature 
 distribution will change with time, since the airway walls will be 
 heated and provide less cooling. This is taken into account by con- 
 sidering the transient-state nature of the heat transfer, although with 
 the simplifying assumption that the equilibrium airflow did prevail 
 from the very beginning of the fire. 
 
 PROGRAM DESCRIPTION 
 
 A more detailed plan of the program in the form of a flowchart 
 is given in figure 5. It shows that the program has been divided 
 into two main parts, network and concentration and temperature. 
 To make the program as flexible as possible, network, temperature, 
 and concentration calculations can be performed separately or com- 
 bined. Methane concentrations are, however, always determined 
 when a change in the airflow distribution takes place, since this 
 is indispensable for coal mines. 
 
 The network part basically contains a well-proven program for 
 conventional network calculations. The term network calculation 
 
 means airflow rate and pressure loss of calculations. The program 
 uses the Hardy Cross method, since studies, which were recently 
 updated (5), showed that for networks of the size of mine ventila- 
 tion systems, this method still gives the shortest calculating times 
 when accompanied by a convergency promoting method of mesh 
 selection. This can be done by letting airways with large products 
 of airway resistances and airflow rates appear in as few meshes 
 as possible. The pertinent mesh assembly process occupies a large 
 portion of the program. 
 
 So-called fixed quantity airways or regulators, which main- 
 tain a constant airflow rate and which are a valuable planning aid, 
 have to be converted into regular airways if more than a conven- 
 tional network calculation is demanded. In case of an emergency, 
 there will be no time to pay attention to the adjustment of regulators 
 that keep the airflow rate constant. 
 
 Concentration and temperature calculations are performed 
 jointly in the second program part because they have many com- 
 mon features. One can think of heat as being a contaminant also, 
 and temperature changes caused by heat influx or by mixing are 
 calculated in the same way as concentration changes. 
 
 The flowscheme program section uses airway and node iden- 
 tification numbers to establish which air currents go into the same 
 
 Network 
 
 Start 
 
 Network calculation demanded? 
 
 No 
 
 Yes 
 
 Output of 
 
 network Input 
 
 Read and try to 
 complete input 
 
 Incomplete 
 
 Assemble meshes 
 
 Satisfy junction equation 
 
 Caluclate natural ventilation pressure from 
 temperature and evaluation of junctions 
 
 Apply cross correction 
 simulate fan characteristics 
 
 Is this the first 
 network calculation? 
 
 t 
 
 Yes 
 
 Convert regulators 
 
 Print results of first 
 network calculations 
 
 No 
 
 No 
 
 Concentration and Temperature 
 
 Temperature or concentration 
 calculation demanded? 
 
 No 
 
 Yes 
 
 Read and try to 
 complete input 
 
 Incomplete 
 
 Calculate CH4 production 
 
 Print Input 
 
 _L 
 
 Establish flowscheme 
 
 ■e 
 
 _L 
 
 Perform roadway calculations 
 
 Estimate properties 
 of recirculated air 
 
 Perform junction calculations 
 
 No 
 
 4 
 
 Do estimates for 
 recirculated air satisfy? 
 
 Yes 
 
 Calculate fire— generated 
 thermal forces 
 
 Equilibrium reached? 
 
 Yes 
 
 _L 
 
 Print results 
 
 Stop 
 
 Figure 5.— Flowchart of steady-state simulation. 
 
11 
 
 junction to be mixed and which air currents leave from the same 
 junction, and, therefore, have the same properties at their 
 beginnings. 
 
 Concentration and temperature changes in roadways can be 
 caused by entering methane, by fire-produced contaminants and 
 heat, and by heat entering or leaving the airway. Injunctions they 
 can occur because of mixing of air currents with different concen- 
 trations and temperatures. Both processes are fundamentally dif- 
 ferent, and two separate program sections have been provided. 
 
 The calculations start at a node with known temperatures and 
 concentrations. Normally this will be the surface or some place in 
 the intake airways. With the conditions at their beginnings thus 
 known, roadway calculations are conducted for all airways leaving 
 this node. Next, a check for junctions where the conditions of all 
 entering airways are known is performed. If found, the entering 
 air currents are mixed and roadway calculations are performed for 
 all airways leaving this junction. 
 
 This process is interrupted when recirculated air enters a junc- 
 tion. In this case the alternating roadway and junction calculations 
 cannot be continued, since concentrations and temperatures of the 
 recirculated, entering air are not known. The difficulty is overcome 
 by using an iteration method. Starting out with estimated values 
 for the properties of the recirculated air, the roadway and junction 
 calculations are continued as if they were known. With this assump- 
 tion, the values, which are then obtained for the recirculated air 
 by the successive roadway and junction calculations, are next used 
 as new, better estimates. The process is repeated until estimated 
 and calculated values agree. Airways with recirculation are placed 
 in a special list in the output to draw attention to the fact that they 
 carry potentially contaminated air into intake airways. It is surprising 
 how much recirculation exists in many mines without being 
 recognized as such. 
 
 Heat and combustion products developed by fires can either 
 be estimated by the user and extend into the program, or can be 
 determined by the program as functions of the oxygen supply of 
 the fire. The heat exchange with the airway walls is calculated with 
 the help of Fourier's equation of thermal conduction, for which 
 solutions have been built into the program. 
 
 A crucial role in these calculations is played by the rock 
 temperature, which is a function of virgin rock temperature and 
 
 the airway history. It is possible, in principle, to determine the 
 temperature distribution in the rock and to provide an accurate 
 solution of Fourier's equation, if the airway history is known. 
 Normally this will not be the case. 
 
 Because fire emergency plans deal with short time spans, com- 
 pared with the age of the airway at least, and since only thin layers 
 of rock surrounding the airways are affected by the temperature 
 changes, it seems to be accurate enough to work with effective rock 
 temperatures for this layer. These are close to the normal air 
 temperatures, modified by the temperature difference caused by 
 convection, and are determined by the program. If better informa- 
 tion exists, it can, of course, be fed into the computer and used. 
 
 The fire-generated thermal forces are the throttling and natural 
 draft effects. They are determined by making use of the calculated 
 temperature distributions. 
 
 PROGRAM USE 
 
 The program was written for routine applications by practical 
 ventilation engineers but without violating physical laws or simplify- 
 ing facts of practical importance. The amount of necessary input 
 data has been kept small and checks are performed for their com- 
 pleteness and for such errors as occur most frequently. Where they 
 do not influence the results significantly, user-stated average values 
 can be used to reduce the input. 
 
 The output provides a listing of airflow rates and pressure losses 
 in airways, and of methane and contaminant concentrations and 
 temperatures at airway ends. Concentrations and temperatures are 
 also listed for junctions. Recirculation paths and airways with airflow 
 reversals, as well as roadways and junctions with critical condi- 
 tions, are additionally listed to alert the user to potential danger 
 zones. 
 
 Because of its capabilities to calculate concentrations and 
 temperatures and to take thermal forces into account, the program 
 is not only used for fire emergencies but also for ordinary ventila- 
 tion planning purposes. There exist several versions of it, developed 
 by program users. The program has found substantial use for fire 
 simulations in high-rise buildings during the last few years. 
 
 COMPLETE TRANSIENT-STATE SIMULATIONS 
 OF VENTILATION SYSTEMS 
 
 The transient concentration distribution programs are useful 
 in many ways. The assumption of time-constant airflow rates is 
 justified for the early stages of a fire when a weak fire does not 
 influence the airflow distribution. When, at a later stage, more in- 
 tense fires may interact noticeably with the ventilation system, with 
 the fire intensity controlled by the air supply to the fire, and with 
 the airflow distribution affected by the fire intensity, the interac- 
 tion should be taken into account. Correspondingly, a program was 
 written for this condition in which, for the sake of simplicity, steady- 
 state conditions were assumed. 
 
 The shortcoming of this program is that it determines the state 
 of a ventilation system at the end of a specified time interval only, 
 with the assumption that the airflow distribution at the end of this 
 interval prevailed from the beginning of the fire throughout the time 
 interval. This may be true, but it may be a simplification of reality. 
 
 There are several different approaches for a complete transient- 
 state simulation of ventilation systems. The natural one to use seems 
 to be a finite difference method based on Newton's second law. 
 One starts with the momentum balance of small elements of air 
 masses to set up the governing equations, and ends up with a set 
 
 of simultaneous equations that have to be solved for the whole 
 system in each time increment. This method was discarded after 
 a lengthy trial for the following two reasons: 
 
 1 . With the involvement of simultaneous equations, the calcula- 
 tion load increases rapidly when a ventilation network becomes 
 large. 
 
 2. For most cases in mine ventilation, the airflow can assume 
 steady state in a time period short enough to justify the application 
 of steady-state theories. 
 
 The chosen approach was then to consider the transient proc- 
 ess as a sequence of short-time steady-state processes during each 
 of which an equilibrium between ventilating pressures (fan and 
 thermal pressures) and airflow-rate-produced pressure losses exists. 
 If the temperature distribution changes, the equilibrium will be 
 disturbed and a new airflow distribution will result. The new 
 equilibrium for the end of the time interval is found by a sequence 
 of alternating temperature, thermal pressure, and network calcula- 
 tions. The results of the last calculation serve as the input for the 
 following calculation. 
 
12 
 
 Neither the airflow rates at the beginning of the time interval 
 nor at its end are representative for the whole time interval. With 
 the assumption that the airflow rate changes during the time inter- 
 val follow a linear pattern, the arithmetic average of the two values 
 is used to represent the average airflow rate. 
 
 PROGRAM ORGANIZATION 
 
 An interval-oriented simulation technique, which updates its 
 data base in every prefixed time interval, was adopted (6). An event- 
 oriented simulation would be more efficient, but with a transient- 
 state simulation of ventilation systems conditions change constantly, 
 which makes events undistinguishable. The control volume 
 approach, with control volumes being blocks of air of uniform com- 
 position, was retained. In other words, the time is divided into a 
 series of time intervals, the airflow into air segments. 
 
 The control volume approach causes difficulties with 
 temperatures and water vapor concentrations, which are not 
 uniform. Their distributions are calculated, and discontinuities be- 
 tween the front and rear ends of adjoining control volumes are 
 averaged out. 
 
 If data records are set up for every airway in a system, a ma- 
 jority of the efforts would be unnecessary because the variation of 
 ventilation parameters in most airways that were not directly af- 
 fected by the hot fumes are at most of a secondary significance in 
 shaping the airflow distribution in the system. It is assumed that 
 the mean air temperature and airflow resistance of an airway will 
 remain unchanged if no significant changes in ventilation condi- 
 tions occur. The criteria to judge a significant change include 
 whether fumes ever got into the airway and whether drastic condi- 
 tion changes (air temperature change larger than 1° F, or fume con- 
 centration change larger than 0.1 pet) happened in the beginning 
 junction of the airway. When no significant change happens in an 
 airway, its original resistance and mean air temperature are taken 
 as the values in the present time interval. 
 
 Figure 6 shows a very simplified flowchart of the transient- 
 state program. The real program is quite lengthy, with approxi- 
 mately 4,500 lines of FORTRAN 77 statements and with 42 
 subroutines. Since it is in many respects a combination of the 
 transient-state concentration program and the steady-state network 
 and temperature calculation model, it encounters the same dif- 
 ficulties that these two models have to face. Two important im- 
 provements were made: a model for the fire zone or the fire 
 behavior; the consideration of mass transfer in the form of water 
 and water vapor. 
 
 COMPUTER MODEL OF FIRE ZONE 
 
 Great difficulties are still encountered in simulating the fire 
 behavior. The steady-state program uses highly simplified, em- 
 pirically obtained functional relations between contaminant and heat 
 production of a fire and the air passing through it. This is essen- 
 tially just a way to define the potential strength of a fire rather than 
 a simulation of fire behavior. 
 
 Although many data and observations on mine fires have been 
 collected, there does not yet exist a satisfactory mathematical model 
 for them. Different researchers focused their attention on different 
 parameters. Some came up with empirical or semiempirical models, 
 which may be useful for a limited range, but all of which suffer 
 from lack of generality. 
 
 Combustion processes are not easy to describe. They are self- 
 sustaining exothermic reactions, which provide heat energy and 
 combustion products at a rate depending on fuel, prevailing 
 temperatures, pressures, and reactant concentrations. Physical proc- 
 
 
 Start 
 
 
 1 
 
 Input 
 
 
 
 
 
 
 
 
 Mesh formation 
 
 
 1 
 
 
 
 
 Temperature distribution 
 before the mine fire 
 
 
 
 Parameter 
 updating 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 Intermediate 
 data transfer 
 
 
 
 
 Airflow reversal? 
 
 
 
 
 
 i 
 
 
 
 
 
 
 
 Fire Source 
 parometer updating 
 
 
 Change fan 
 operation? 
 
 
 
 
 1 
 
 
 
 
 ZE 
 
 in each airway 
 
 
 Change 
 resistance? 
 
 1 
 
 
 
 1 
 
 
 
 Updating parometer record 
 for airway ending 
 
 
 
 
 
 
 
 
 1 
 
 
 
 
 Update junction 
 conditions 
 
 
 
 
 1 
 
 
 
 
 Advance new control 
 segments into airways 
 
 
 
 
 1 
 
 
 
 
 Calculate natural drafts 
 and throttling effects 
 
 
 
 
 1 
 
 
 
 Balance network 
 
 
 I 
 
 
 
 Detect fume— free 
 evacuation routes 
 
 
 
 
 
 1 
 
 
 
 
 Output 
 
 
 e out 
 
 
 Tin- 
 
 
 End 
 
 
 Figure 6. — Flowchart of transient-state ventilation simulation 
 program. 
 
 esses exert a significant, sometimes dominant, influence on the fire 
 also through the transport of matter and energy. 
 
 Reasonable models must take reaction kinetic, mass, and heat 
 transfer in their various forms into account. Such models are pres- 
 ently developed (7) and it is hoped that one day they can be helpful 
 in the design of realistic emergency plans. Because of the large 
 amount of calculation work required by the digital simulation, they 
 should not yet be used for routine emergency planning, when 
 simplified methods can do almost as well. 
 
 Using the example of timber, which seems to be the by far 
 most researched fuel for mine fires, the organization of the com- 
 puter models shall be described. The fire zone is divided along the 
 airway into segments of a few inches width. The segments are 
 radially, or perpendicular to the airway axis, divided into sections. 
 The boundaries between segments and sections are marked by nodes. 
 Updating of conditions in segments, sections, and nodes is done 
 in time increments of 1 s or less. Since a fire zone can be tens or 
 even hundreds of feet long, and every update demands the solution 
 of several equations per segment, section, and node, the calcula- 
 tion effort is quite large. 
 
 In a typical segment as shown in figure 7, section 1 represents 
 the control volume in the bulk flow of air current. Section 2 stands 
 for the control volume in the wall layer where the products of wood 
 pyrolysis burn. Its thickness is taken as a twentieth of the airway 
 hydraulic radius, roughly comprising the eddying sublayer and the 
 buffer zone, passing by a small percentage of the total mass flow 
 of air. 
 
13 
 
 Sections 
 
 Nodes 
 
 3 ft 
 
 h 
 
 T 
 
 H 
 
 r Rock 
 
 y Wood 
 layer 
 
 y Wall 
 layer 
 
 y Bulk 
 flow 
 
 Figure 7.— Location of sections and nodes in fire zone 
 segment. 
 
 Sections 3 and 4 are the wood layer close to the surface, sec- 
 tion 5 is the rest of the wood. The total thickness, X, of the wood 
 support is calculated from the wood loading per foot of airway 
 length. Recommended values for the thickness of sections 3 and 
 4 are 0.1 *X for X < 76.2 mm (0.25 ft) or 7.62 mm (0.025 ft) for 
 X > 76.2 mm (0.25 ft). 
 
 Sections 6 and 7 are in the surrounding rock. A total thickness 
 of 914 mm (3 ft) is recommended for both sections. This is well 
 above the thickness of the rock layer in which the temperature is 
 affected by a fire. 
 
 The pyrolysis of wood is expected to occur in the region be- 
 tween nodes 3 anad 5 according to Arrhenius' equation. As the 
 pyrolysis is mainly and sensitively controlled by the local 
 temperature, denser subgrids of 20 strips are put into this region 
 with the local temperatures in strips being determined through ex- 
 ponential interpolation between the adjacent node temperatures. 
 
 It is necessary to divide the airflow into segments also, although 
 these segments do not need to be as short as the wall segments. 
 
 The basic logic of the program can be outlined as follows. Sets 
 of equations are developed to describe local energy balances. They 
 contain endothermic heat of pyrolysis, exothermic heat of combus- 
 tion of volatiles and char, conductive and radiative heat transfer, 
 enthalpy flux accompanying mass movement, and external ignition 
 heat input. The pyrolysis is regarded as a function of local 
 temperature using Arrhenius' equation. If thermal decomposition 
 
 has taken place, it is a function of its pyrolysis history also. The 
 exchange of volatiles, oxygen, and combustion products between 
 bulk flow and wall layer is calculated using mass transfer theories. 
 The oxygen supply is compared with the volatile emission. If surplus 
 oxygen exists, it is used for the combustion of char residue. 
 
 Many of the equations for node temperatures are nonlinear. 
 They are linearized through Taylor's expansion with temperature 
 corrections as new unknowns. These are determined using the Gauss 
 elimination method and node temperatures are obtained through an 
 iterative approach. 
 
 CONSIDERATION OF WATER VAPOR TRANSFER 
 
 There exist by now a large number of approaches to predict 
 the combined dry heat transfer by conduction within the rock and 
 by convection from the rock to the air. None of these approaches, 
 however, take the parallel mass transfer of water and its temperature 
 effect satisfactorily into account. 
 
 This is regrettable because it is recognized that even seemingly 
 dry airways are to a considerable extent influenced by water migra- 
 tion, evaporation, or condensation. Small water quantities can, due 
 to the large latent heat of water, have great effects. 
 
 The fact that no convincing mathematical descriptions of the 
 simultaneous heat-mass transfer exist indicates the complexity of 
 the problem. 
 
 Past attempts to describe the effects of water on mine air 
 temperatures fall roughly into two categories. One of them uses 
 statistical tools to interpret systematical field-measured data for em- 
 pirical relationships among temperature, humidity, and other ven- 
 tilation parameters. It does not attempt to explain the nature of their 
 dependence or to draw generally valid conclusions. Rather it offers 
 useful equations of localized significance. 
 
 The other is a semiempirical technique that tries to derive func- 
 tional relationships between temperature and other ventilation 
 parameters. A number of loosely defined factors, coefficients, and 
 more or less justified relations are introduced for this purpose to 
 provide for general applicability. There exist, however, so many 
 assumptions, which are neither theoretically sound nor universally 
 tenable, that the usefulness of this approach has to be limited in 
 range or a lack in accuracy has to be expected. 
 
 The computer model reported here tries to follow a rigorous 
 analytical approach and a rigorous analytical solution can be obtained 
 (8). Since the calculation of numerical solutions turns out to be a 
 very time-consuming task, approximations for short time periods, 
 with which one is concerned in emergency planning, were developed 
 also. 
 
 Additional difficulties arise from boundary conditions. Water 
 can evaporate or condense on airway walls with pertinent changes 
 in wall temperatures. It can condense as fog in the bulk flow of 
 the air with a pertinent change in the temperature gradient between 
 airway walls and air. To be properly considered, these wall 
 temperature changes have to be recorded. To keep the amount of 
 record keeping small, so-called stations in intervals of a few hun- 
 dred feet are established, for which the correct data are recorded. 
 For the space between stations, interpolations are used. 
 
 It is very much desirable to find a more simplified approach 
 for considering the influence of water transfer on mine temperatures. 
 Work in this direction continues. But to simplify sensibly, the 
 unsimplified facts have to be known first. 
 
14 
 
 SUMMARY 
 
 A transient-state computer model for ventilation systems has 
 been the ultimate goal toward which all program development 
 strove. The transient-state concentration model and the steady-state 
 fire and ventilation interaction model were essentially compromises 
 between what was desirable and what was feasible at the time. 
 
 The lack of a transient-state ventilation model, which allows 
 the quantitative prediction of ventilation patterns in thermally 
 disturbed mine ventilation systems, has been overcome. Morever, 
 the program described in this paper has been made user friendly 
 and a personal computer version has been developed, which should 
 make it helpful to every engineer dealing with fire or ventilation 
 problems. 
 
 The simulation of the transient processes in fire zones and the 
 simulation of water transport is still cumbersome, because it was 
 attempted to take all significant factors, like chemical kinetics, heat 
 and mass transfer, unabridged and unsimplified into account. 
 Extensive testing of these newly developed program sections with 
 experimental data may show that justified simplifications are 
 possible. 
 
 Since only the well-established laws and principles of thermo- 
 dynamics, fluid mechanics, heat transfer, and mass transfer have 
 been used, the transient-state computer model should be theoretically 
 sound. It has given plausible results so far. 
 
 REFERENCES 
 
 1. Bastow, K.R. Real-Time Simulation of Contaminant Flow Through 
 Mine Ventilation Networks Under the Influence of Mine Fires. M.S. Thesis, 
 MI Technol. Univ., Houghton, MI, 1979, 306 pp. 
 
 2. Greuer, R.E. Real-Time Precalculation of the Distribution of Com- 
 bustion Products and Other Contaminants in the Ventilation System of Mines 
 (contract JO285002, MI Technol. Univ.). BuMines OFR 22-82, 1981, 263 
 pp.; NTIS PB 82-183104. 
 
 3 . . A Study of Precalculation of Effect of Fires on Ventilation 
 
 System of Mines (contract JO285002, MI Technol. Univ.). BuMines OFR 
 19-84, 1983, 293 pp.; NTIS PB 84-159979. 
 
 4. Sheng. J. Determination of the Cumulative Exhaust Effects of Diesel 
 Powered Equipment Underground. M.S. Thesis, MI Technol. Univ., 
 Houghton, MI, 1984, 159 pp. 
 
 5. Greuer, R.E. Study of Mine Fires and Mine Ventilation; Part I. Com- 
 puter Simulation of Ventilation Systems Under the Influence of Mine Fires 
 (contract SO241032, MI Technol. Univ.). BuMines OFR 115(l)-78, 1977, 
 165 pp.; NTIS PB 288 231. 
 
 6. Chang, X. Digital Simulation of Transient Mine Ventilation. Ph.D. 
 Thesis, MI Technol. Univ., Houghton, MI, 1987, 162 pp. 
 
 7. Chang, X., and R.E. Greuer. A Mathematical Model for Mine Fires. 
 Ch. in Proceedings of the 3d U.S. Mine Ventilation Symposium. Soc. Min. 
 Eng., AIME, Littleton, CO, 1987, pp. 453^62. 
 
 8. . Simplified Method To Calculate the Heat Transfer Between 
 
 Mine Air and Mine Rock. Paper in Proceedings of the 2d U.S. Mine Ven- 
 tilation Symposium, ed. by P. Moussett- Jones. A. A. Balkema, 1985, pp. 
 429-438. 
 
15 
 
 MINE FIRE DETECTION SYSTEMS: A PRIMER 
 
 By Charles D. Litton 1 
 
 ABSTRACT 
 
 This Bureau of Mines paper discusses a simple approach to the design and implemen- 
 tation of mine fire detection systems. Topics include state-of-the-art technology in fire sensors 
 and systems, a discussion of the hazards that different types of fires can produce, and some 
 simple guidelines for determining the level of protection required for different mining areas 
 and types of combustibles used or stored. 
 
 INTRODUCTION 
 
 The successful performance of a fire detection system 
 depends upon its ability to detect the presence of a fire rapidly 
 and reliably. It should be sufficiently rapid so that there 
 remains enough time to either safely evacuate a mine or 
 successfully extinguish and control the fire, or both. It should 
 be reliable so that false alarms are minimized without sacri- 
 ficing sensitivity of time response. Systems should also be 
 durable enough to withstand the mine environment over long 
 periods of time while maintaining a high degree of reliability. 
 Maintenance and calibration of the system should be simple 
 and not require the expenditure of excessive time and labor to 
 keep it operational. 
 
 Many factors influence the design of a fire detection 
 system, such as the potential sources and modes of ignition, 
 the types of combustibles involved, and the quantities of 
 combustibles available for fire growth and flame spread. Mine 
 air ventilation influences the growth and spread of a fire and 
 also serves to transport the products of combustion to other 
 areas of a mine, remote from the fire. Fire sensor alarm 
 thresholds place limits on the sizes of fires that can be 
 detected. The time to respond to a detected fire, either in terms 
 
 of evacuation or control and extinguishment, place additional 
 time constraints on the design of fire detection systems. 
 
 These factors interact with each other, often in complex 
 ways, and it is the definition and understanding of these 
 factors and their interactions that hold the key to the design of 
 adequate and reliable fire detection systems. The level of 
 understanding of the problem has increased dramatically 
 within recent years and it is the intent of this paper to discuss 
 the progress that has been made and how the information can 
 best be used to implement fire detection systems in under- 
 ground mines. 
 
 To address the potential for fire detection systems, a 
 framework must be created with which the design of these 
 systems can evolve. This framework should address not only 
 the detection of fires, but also the levels of detection that may 
 be required for different areas of mines. The different levels of 
 detection are based upon the types of fires that may be 
 expected, and their potential hazards. To begin to understand 
 the problem, it is necessary to have knowledge about the fires 
 that are most likely to occur and the sensors and systems 
 available for their detection. 
 
 MINE FIRE SENSORS AND SYSTEMS 
 
 Most mine operators and mine safety personnel are 
 familiar with minewide monitoring systems that are available 
 from a number of different manufacturers. Mine fire detection 
 is a major component of these systems, which now use CO fire 
 sensors. Currently, the Bureau-developed submicrometer par- 
 ticle smoke detector is being developed commercially for use in 
 such systems and should be available in the near future. Both 
 CO and smoke detectors are product-of-combustion fire sen- 
 sors. These sensors are designed primarily to detect the CO 
 and smoke liberated from either a smoldering or flaming fire 
 into the mine ventilation airflow where their concentrations 
 
 1 Supervisory physical scientist, Pittsburgh Research Center, Bureau of Mines, 
 Pittsburgh, PA 
 
 are diluted and carried along with the airflow. The diluted 
 concentrations of CO and smoke that are present within the 
 ventilation airflow depend upon the rates at which they are 
 produced relative to the quantity of air that mixes with these 
 combustion products. 
 
 For fixed sensor alarm thresholds, it becomes apparent 
 that very small fires may be detectable at low ventilation 
 airflows, while at higher airflows larger fires are needed to 
 reach the same alarm level. Sensor sensitivity, then, is an 
 important parameter, but setting of alarm thresholds most 
 often depends upon the mine background level of either CO or 
 submicrometer particles. Alarm thresholds should be signifi- 
 cantly higher than these background levels to protect against 
 false alarms of the sensors. High false alarm rates can destroy 
 
16 
 
 confidence in a fire detection system, rendering the system 
 useless. At the other extreme are sensors that are too insensi- 
 tive, with high alarm thresholds incapable of detecting fires in 
 their earlier stages of development. Figure 1 provides some 
 insight into this type of phenomena. At low values of the 
 alarm threshold level to background level ratio, false alarms 
 significantly reduce the confidence in the system. As this ratio 
 increases, so does confidence in the system, until, at some 
 point, the sensor becomes much less sensitive to small fires and 
 confidence in the system again decreases. 
 
 For CO fire sensors, confidence greater than 90 pet is 
 usually achieved when this ratio is in the range of about 3 to 
 15, assuming that the background CO levels typically are 1 to 
 5 ppm and an alarm threshold level of 15 ppm is used. 
 Assuming a minimum background level of 1 ppm, increasing 
 the alarm threshold to greater than 15 ppm decreases the 
 sensitivity of the system and at much larger values, the system 
 becomes incapable of achieving its desired function. A similar 
 analysis can be made for smoke particles. 
 
 The electrical and electronic components necessary for 
 the system to operate must also exhibit a high degree of 
 reliability. Their function is to provide power to the sensor, 
 accept sensor signals, supervise the sensors to assure proper 
 operation, and record and display information and alarms. 
 This is usually done in two stages. Outstations, to which a 
 number of fire sensors and other transducers (such as veloci- 
 meters) are electrically connected, are installed underground. 
 Sensor power is supplied by these outstations and sensor 
 signals accepted. This information is then transmitted to a 
 central processing unit, usually located aboveground at a 
 permanently staffed location. The central station accepts the 
 data from undergound, records it, and if desired, displays it 
 on a video screen. Many levels of sophistication are possible 
 for the total system. What level is used depends upon the 
 particular mine and its needs, as well as the cost of the system. 
 
 At the outstation level, most commercial systems are de- 
 signed to accept signals from a variety of transducers. This is 
 important from a fire detection standpoint when sensors, other 
 than CO or smoke, such as optical sensors are used. Optical 
 sensors, such as ultraviolet or infrared sensors, are fast respond- 
 ing detectors (approximately milliseconds or faster) that sense the 
 radiant energy emitted by a fire. They are recommended for use 
 in applications where instantaneous response is critical, such as 
 fuel storage areas and power centers. Such sensors may not be 
 available from the manufacturer of the monitoring system, but 
 can be purchased separately and are easily interfaced with most 
 monitoring systems. It is usually wise to determine in advance if 
 a monitoring system can use a variety of transducers, or if 
 interfacing them to the system is a simple procedure which is 
 relatively inexpensive. 
 
 Mine monitoring systems and, in particular, those systems 
 used primarily for fire detection have shown tremendous 
 growth in use within the past 2 to 3 yr. At the same time, users 
 of such systems are beginning to express some dissatisfaction 
 with the calibration and maintenance requirements of the 
 sensors. This is particularly true for systems installed in large 
 mines where the number of sensors required can be quite high. 
 Past research by the Bureau has shown that an alternative 
 approach to that of fixed-point fire sensors is a pneumatic 
 monitoring system. 
 
 For this type of system, individual tubes extend from a 
 central pumping and detection station to the desired monitor- 
 ing locations, as indicated in figure 2. This central station may 
 be located aboveground or at some convenient location under- 
 ground. When located underground, this central station is 
 analogous to the outstation of the fixed-point mine monitoring 
 
 A/B 
 
 Figure 1 .—Variation in the percent confidence in a mine fire 
 detection system as a function of the ratio of alarm threshold 
 level to background level (A/B) for product-of-combustion fire 
 sensors. 
 
 Tubes from 
 monitoring locations 
 
 Sample 
 pump 
 
 .Solenoid valves 
 or equivalent 
 
 (sensor Ij - 
 ^Sensor 2 J- 
 
 ( Sensor nj- 
 
 ► Exhaust 
 
 Scavenger 
 pump 
 
 Exhaust 
 
 Figure 2.— Schematic of central detection and pumping sta- 
 tion for use with a pneumatic fire detection system. 
 
17 
 
 system. At the central station (fig. 2), all tubes except for the 
 one being sampled are continuously purged by a scavenger 
 pump. Each tube has its flow diverted through a three-way 
 solenoid valve at fixed time intervals to a smaller sample pump, 
 which presents the tube contents to one or more sensors for 
 measurement of CO, smoke, or other gases of importance. 
 
 This type of system is attractive for a number of reasons. 
 First, the cost of the system is less than that of a fixed-point 
 monitoring system. Second, because all components, includ- 
 ing sensors, are located at the same place, calibration and 
 maintenance requirements are significantly less. Third, mea- 
 
 surement of any additional gas at all locations is easily 
 accomplished by adding a single gas analyzer to the system. 
 This system is ideally suited for continuous monitoring of 
 gobs, mined-out areas, return airways, etc. Through proper 
 choice of tube diameters and pump capacities, the system can 
 be designed for rapid response. 
 
 The use of this system suffers from the fact that it is not 
 commercially available as a system. However, all components 
 necessary are readily available, off-the-shelf items and design 
 of these systems is easily accomplished. 
 
 FIRE CHARACTERISTICS 
 
 Fires produce heat, light, and combustion products (gases 
 and smoke). These represent measurable quantities that, when 
 detected at low levels, signal the presence of a developing fire. 
 Fires develop through three distinct stages. The first stage is 
 the preflaming, smoldering stage where fuel is pyrolyzed and 
 produces products of incomplete combustion and fuel vapors 
 that are eventually ignited, resulting in open, flaming combus- 
 tion. Smoldering fires produce little heat or light, but copious 
 amounts of smoke, CO, and often other gases. Further, the 
 smoldering stage of combustion may last for prolonged peri- 
 ods of time, hours or days, or it may be of much shorter 
 duration depending upon the source of ignition. 
 
 The transition to open, flaming combustion signals the 
 beginning of the second fire stage. During this stage, detect- 
 
 able levels of heat, light, and products of combustion are 
 produced as the fire grows in intensity. The rate of growth may 
 also vary dramatically. Liquid pool fires reach a stage of 
 steady-state burning in a matter of several seconds, while fires 
 of coal rubble or other solid combustibles grow in size at a 
 much reduced rate. In general, this stage of the fire remains 
 localized, consuming the combustibles present in the general 
 vicinity until, at some point, the fire intensity is sufficient to 
 produce propagation or rapid flame spread down an entry. 
 
 The propagating fire represents the final stage of a 
 developing fire and represents the most severe hazard. Once 
 this stage is reached, entire mines can be devastated. Little, if 
 any, opportunity exists for either escape or control of the fire. 
 
 FIRE HAZARDS 
 
 Smoldering fires, or fires that develop to the point of 
 smoldering due to spontaneous combustion or other modes of 
 self-heating, produce little heat or light, but are more appro- 
 priately characterized by the liberation of smoke, CO, and 
 other potentially toxic gases. External sources, such as over- 
 heated equipment or electrical shorts, may produce smoldering 
 periods that last for minutes or hours, while smoldering fires 
 of spontaneous origin may have much longer periods of 
 prolonged smoldering before flaming combustion is reached. 
 
 The primary hazard associated with smoldering combus- 
 tion is the toxicity of the smoke and gases that are produced. 
 Knowledge of this toxicity is lacking and represents an area of 
 intensive research. However, both CO and smoke are major 
 products of smoldering fires and either CO or smoke detectors 
 are capable of providing for early warning of this stage of fire 
 development. Research into the relative toxicities of smolder- 
 ing fires will provide the information necessary to evaluate the 
 response of CO and smoke sensors relative to the potential 
 toxic hazard. It is sufficient to say, for the moment, that CO 
 and smoke sensors provide for early warning of most smolder- 
 ing, incipient fires, but the relationships between their alarm 
 level thresholds and the potential toxic hazard are not well 
 understood. 
 
 The hazards that result from flaming fires are functions of 
 the ratio of the fire size to the ventilation rate, expressed as 
 Q f /v f A, where Q f is the heat release rate of the fire (kW), v f is 
 the ventilation velocity (m/s), and A is the average entry 
 cross-sectional area (m 2 ). Three major hazards result from 
 fires: 
 
 1. Toxicity associated with the gases and smoke products 
 of combustion. 
 
 2. Reduced visibility (smoke obscuration) from the smoke 
 particles. 
 
 3. Fire propagation or rapid flame spread along the 
 combustible surfaces of an entry. 
 
 The relative toxicities of products produced from a variety 
 of mine combustibles is an area of research currently being 
 addressed. It is known that CO is a major toxic gas component 
 of the combustion products. Others have been found, such as 
 HC1 from materials containing chlorine. For purposes of this 
 paper, it will be assumed that CO represents the primary toxic 
 hazard reaching an incapacitating stage at a concentration of 
 approximately 1,500 ppm (0.15 pet) or greater. The fire size 
 necessary to produce this concentration is given by 
 
 (Q r /v f A) T > 300.0, 
 
 (1) 
 
 where the subscript T denotes the toxic hazard. 
 
 The production of copious amounts of smoke severely 
 reduces mine visibility and can significantly reduce the poten- 
 tial for escape. It is generally recognized that when the smoke 
 level reaches a concentration sufficient to reduce visibility by 
 approximately 84 pet over a 12-ft path, an extreme hazard 
 exists. To produce this reduced visibility, a smoke particle 
 concentration of approximately 225 mg/m 3 is required. The 
 fire size necessary to produce this concentration is given by 
 
 (Q f /v f A) s > 150.0, 
 
 (2) 
 
 where the subscript S denotes the smoke hazard. 
 
 Fully developed, sustained fire propagation occurs when 
 the hot gas temperatures from a fire are sufficient to ignite 
 
18 
 
 exposed surfaces of combustible materials along an entry. 
 Both theory and data indicate that this condition is reached 
 when the fire size is defined by 
 
 (Q r /v f A) P > 450.0, (3) 
 
 where the subscript P denotes the propagation hazard. 
 
 The times that are required for fires to reach these critical, 
 hazardous levels can vary dramatically with the type of fuel 
 involved, the amount of fuel available, and its distribution. 
 Fires that involve spills of flammable liquids or liquid pools 
 develop within seconds to these levels. For instance, a spill of 
 liquid heptane, if ignited, will reach the critical smoke hazard 
 level in 5 to 10 s, and the critical propagation hazard in 
 approximately 40 s. Fire detection systems for these cases 
 should also contain means for automatic suppression and 
 extinguishment. 
 
 At the other extreme are fires that are preceded by 
 prolonged periods of smoldering. Such fires generally occur in 
 areas of the mine that are free from sources of external heat 
 and energy, and are due to spontaneous, self-heating. The 
 detection of these fires need not be so rapid. Several minutes or 
 hours will generally suffice, but if left undetected, the conse- 
 quences can be just as dramatic as those of the liquid pool fire. 
 
 Between these two extremes lie fires, usually of solid 
 combustibles, which are ignited by external sources of heat and 
 energy. Periods of smoldering may be of short duration or 
 nonexistent. Once flaming occurs, the fire intensity begins to 
 increase, although generally at a much slower rate than for a 
 liquid pool fire. Clearly, once flaming occurs, there exists only 
 a finite amount of time available to detect the fire, evacuate 
 personnel, and extinguish the fire before the situation becomes 
 critical, lives are lost, and property devastated. 
 
 Figure 3 illustrates the times that are involved during the 
 growth stages of a fire involving solid combustibles (solid 
 line). At the origin of the fire, alarm levels of smoke or CO are 
 produced within a minute or so, perhaps even before flaming if 
 there is enough smoldering combustion. Once these alarm 
 levels are produced, the ventilation airflow carries them away 
 from the fire to a fire sensor that must respond and issue an 
 alarm. All of this must occur in approximately 15 min, or less. 
 This leaves about 50 min before the fire reaches a stage where 
 smoke obscuration becomes critical. In about another 60 to 70 
 min, the CO becomes life threatening, and within another 60 
 min the fire is spreading rapidly throughout the mine. 
 
 For comparison, the dashed line shows the growth rate of 
 a liquid pool fire with a surface area of 1.5 m 2 . In less than a 
 minute, the fire is large enough to begin propagating. 
 
 10.000 
 
 ■a 1,000 
 ui 
 
 CO 
 
 < 
 
 100 
 
 10 
 0.01 
 
 ~-<^ 
 <&** 
 
 _Propagation y 
 
 CO 
 
 critical 
 Smoke « 
 critical 
 
 Alarm 
 
 + 
 
 Detectable products 
 (smoke, CO) 
 
 0.10 
 
 1.0 10.0 
 
 TIME, min 
 
 100.0 
 
 1,000 
 
 Figure 3.— Graphical representation of the sequence of events 
 common to development of many underground mine fires. 
 
 From this discussion, it is apparent that fires of differing 
 origins require different levels (or times) of detection. For 
 smoldering fires, detection times of a few to several hours will 
 suffice. For fires of solid combustibles driven by external 
 sources of heat and energy, detection times of a few minutes 
 are required. And for fires of flammable liquids or fires in 
 areas where significant combustible materials (solid, liquid, or 
 gas) are stored or used, detection must be within a few seconds 
 and often must contain a means for automatic suppression of 
 the fire. This analysis points to a need for defining those areas 
 of a mine applicable to the three levels of detection. 
 
 It is proposed that a mine be divided into three distinct 
 classes when considering the design of minewide fire detection 
 systems. The first class (class I) areas are those areas where the 
 risk of fire is high and also where the consequences are most 
 dramatic. Typical class I areas would include fuel storage 
 areas, fuel transfer areas, maintenance areas, and other areas 
 where significant quantities of combustibles or flammable 
 materials are stored or used. 
 
 The second class (class II) areas are those areas where the 
 risk of fire and its consequences are not so great, but where 
 there exist sufficient external ignition sources. Examples of 
 class II areas would include conveyor belt entries, track entries, 
 or other entries within which potential external sources of 
 ignition exist and on which routine mining operations depend. 
 
 The third class (class III) areas are those areas where the 
 risk of fire is the least. In general, these areas do not contain 
 sources of external ignition, but rather, fires develop via 
 spontaneous heatings or self-heatings independent of external 
 sources. Examples would include return airways, mined-out 
 areas that are feebly ventilated, and other areas that in general 
 are remote and not crucial to the routine operation of the 
 mine. 
 
 CLASS I AREAS 
 
 In general, these areas are relatively small, localized areas 
 where the risk of fire and its consequences are greatest. Rapid 
 detection, often coupled with automatic suppression capabil- 
 ity, within seconds is generally warranted. To achieve this level 
 of detection, optical sensors, or a well-defined array of 
 thermal sensors, should be used as the primary detection 
 system. 
 
 While these areas may be well-defined, localized areas, 
 they are also under the influence of some degree of ventilating 
 airflow. Consequently, the possibility exists that these fires 
 may contaminate the ventilating airstream before detection 
 and suppression can be achieved. To provide some degree of 
 protection against this possibility, it is also recommended that 
 one or more product-of-combustion sensors be located down- 
 stream of these areas in the primary ventilating airflow to 
 detect any leakage of combustion products. Further, fires 
 within these areas may, in some instances, be preceded by a 
 prolonged smoldering stage for which no heat or light is 
 produced. To provide protection for this possibility, a second- 
 ary product-of-combustion sensing system should be used in 
 conjunction with the primary optical or thermal sensing 
 system. 
 
 If the ventilation flow within a class I area is well defined, 
 then the product-of-combustion sensor should be located 
 within the area and immediately upstream of the point at 
 which the ventilation flow from this area and the primary 
 ventilating air mix. If the ventilation pattern within the area is 
 poorly defined, then a system of two or more sensors spaced 
 according to National Fire Protection Association standards 
 (or equivalent standards) should be used. 
 
19 
 
 To summarize, class I areas represent extremely high risk 
 areas for which the potential consequences of fire are the most 
 severe. Rapid detection of the initial flaming stages dictates the 
 use of an optical or thermal sensing system. In many areas, the 
 system should also contain automatic fire suppression capa- 
 bilities. An additional secondary product-of-combustion fire 
 sensing system should be used in parallel with the primary 
 system to protect against the possibility of smoldering fires. 
 One or more additional product-of-combustion sensors should 
 be installed within the primary ventilating air immediately 
 (approximately 50 to 100 ft) downstream of a class I area to 
 protect against potential contamination of the primary air. 
 
 CLASS III AREAS 
 
 At the other extreme lie the class III areas for which the 
 risk of fire is the least. However, if these areas are left 
 unprotected, then the consequences of fire can be just as 
 dramatic. In general, these areas are characterized by the 
 absence of external ignition sources. As a result, fires that 
 develop in these areas can be expected to develop over longer 
 periods of time, most probably because of spontaneous com- 
 bustion or self-heating resulting from other causes. 
 
 Since no heat or light is produced during smoldering, the 
 fire sensor of choice is a product-of-combustion sensor. The 
 choice of which product-of-combustion sensor to use (CO or 
 smoke sensor) depends upon the most probable fire scenario. 
 If a mine has a history of spontaneous combustion or if the 
 area to be protected consists of caved-in areas or gobs, then a 
 CO sensor is the logical choice. This arises from the fact that 
 CO is the best indicator of spontaneous combustion fires. 
 Smoke is usually generated at a later point in time relative to 
 CO. If smoldering fires other than those of spontaneous origin 
 can be expected within a class III area, then smoke sensors are 
 the obvious choice to provide an earlier warning. For smolder- 
 ing fires of unknown origin, a combination of CO and smoke 
 sensors may be warranted. 
 
 But where are the sensors to be located? Excluding for the 
 moment return entries as class III areas, the bulk of class III 
 areas can be expected to be remote areas, normally weakly 
 ventilated and for which ventilation patterns may be ill de- 
 fined. However, it is generally known that ventilating air goes 
 into an area and comes out. To determine if combustion 
 products are produced within this area, it is prudent to make a 
 differential measurement. This means that a sensor should be 
 located within the airsteam entering a class III area and one 
 out by the area to measure what comes out. The difference is 
 what is produced within that area. 
 
 For primary return airways, sensors should be located 
 immediately upstream of each junction of airways where flows 
 mix. If, for instance, three splits of return air mix or merge at 
 some point, then a sensor should be located within each split 
 entry just prior to that point. An alternative to this approach 
 would be to locate a single sensor immediately downstream of 
 the mixing point. However, the concentration of CO or smoke 
 measured at that point will be reduced because of dilution by 
 clean air from the unaffected splits. This implies that a more 
 sensitive detector may be warranted if only one is used. The 
 single-sensor approach also does not provide the information 
 of which split entry contains the fire, and significant time may 
 be lost in locating the fire. 
 
 Optimum location of mine fire sensors for the protection 
 of class III areas is an area of research that needs to be 
 addressed more explicitly. For these areas, the primary concern 
 is the detection of fires in their spontaneous or smoldering 
 stages for which function either CO or smoke sensors are the 
 primary sensors to be considered. These products are carried 
 from point to point by the ventilating airflow. As a result, the 
 optimum placement of sensors depends critically upon accu- 
 rate definition of the airflow patterns within these areas. 
 
 A tool that is ideally suited for this application is the 
 Michigan Technological University mine (fire) ventilation 
 code. Other ventilation codes exist, but this code is designed to 
 provide information that is more relevant during a mine fire 
 emergency. First, the code is capable of predicting ventilation 
 patterns from a minimum of information. Second, the code is 
 capable of predicting the rates of contamination of all areas of 
 the mine by fire combustion products. The first capability can 
 provide the information necessary to optimize fire sensor 
 locations. The second capability can provide information 
 relevant to the most prudent escape route to be followed during 
 a fire emergency. 
 
 The code is available for use by the mining industry. Its 
 potential for use as a tool in the design of fire detection 
 systems is a subject of current research. 
 
 CLASS II AREAS 
 
 Firmly entrenched between class I and class III areas of a 
 mine are the class II areas. For these areas, there exist many 
 potential sources of ignition — overheated equipment, electri- 
 cal arcs and shorts — to name two of the most common. These 
 areas are characterized by the presence of equipment necessary 
 to the normal mining activity — conveyor belt haulage systems, 
 electrical trolley entries, track haulage entries, etc. They are 
 also entries, some of which may be long or short, that are 
 frequently used and which are critical to the mine operation. 
 Class II areas may contain class I areas, but these are 
 separately protected. 
 
 The times that are available before critical hazards are 
 reached in class II areas were previously discussed. Unlike class 
 
 I or class III areas, the time to respond to a fire within a class 
 
 II area is a significant factor, which must be addressed during 
 the design stages of the system. 
 
 Ventilation is usually well defined in class II areas, and it 
 is the ventilation which transports the combustion products 
 from a fire to a fire sensor. A good estimate of the spacing 
 requirements between sensors can be made by multiplying the 
 desired maximum response time by the velocity of the air 
 within an entry. For instance, if a response time of 15 min or 
 less is desired in an entry where the velocity is 100 ft/min, then 
 the sensor spacing should be no more than 1,500 ft. 
 
 The detection of fires based on product-of-combustion 
 sensors is one approach where the number of required sensors 
 is relatively low. The sensors also provide a capability for the 
 detection of any smoldering combustion, which would precede 
 the onset of flaming. Thermal or optical sensors may also be 
 used, but their numbers increase dramatically in order to 
 provide the same level of detection. Such sensors would also be 
 insensitive to any smoldering combustion that might occur. 
 
20 
 
 DISCUSSION 
 
 There exist many areas of mine fire detection research that 
 need to be addressed. The toxicity associated with products of 
 combustion is not well understood. Advances in detector 
 technology need to be made, especially for applications in 
 diesel-operated mines where the diesel combustion products 
 often overwhelm the fire products and render detection sys- 
 tems to a low level of reliability. The effects of ventilation 
 patterns on developing fires and the converse effects of a fire 
 on the ventilation system need further study in order to 
 quantify these effects. Mine ventilation codes offer significant 
 potential for use as tools in defining optimum sensor locations 
 and in the preplanning of evacuation and escapeways. 
 
 Much remains to be done before the understanding of 
 mine fire problems and detection systems is complete. Yet, the 
 
 level of information and resources available at the current time 
 indicate that there exists the potential for significant improve- 
 ment in mine fire detection systems. As the level of under- 
 standing of these problems increases, so will the levels of 
 detection that can be provided increase. 
 
 The design and implementation of fire detection systems 
 need not be overwhelming. It can begin by identifying those 
 areas of mines that are most in need of protection and building 
 from there. The framework presented here serves as a guide as 
 to what is practical and possible with the present level of 
 technology. 
 
21 
 
 FIRE DETECTION SYSTEMS FOR NONCOAL UNDERGROUND MINES 
 
 By W. H. Pomroy 1 
 
 ABSTRACT 
 
 Early fire detection can be a critical element in a mine's overall strategy for dealing with fire 
 emergencies. This paper describes results of Bureau of Mines research to evaluate, through long- 
 term in-mine tests, the performance of fire detection devices in all four fire detection categories: 
 heat, flame, smoke, and fire gas. The basis for detector selection, as well as installation and 
 maintenance practices to help insure reliable detector performance, are also provided. 
 
 INTRODUCTION 
 
 Early detection of an underground mine fire can significantly 
 increase the likelihood of survival for underground workers and 
 minimize the time required for a mine to return to production after 
 such an emergency. Early fire detection (or ideally, detection dur- 
 ing the incipient stage), may enable evacuation before smoke and 
 toxic fire gases grow to life -threatening concentrations or block 
 visibility. Also, since smaller fires are more easily extinguished 
 and less hazardous to fight, early fire detection contributes to faster 
 and more effective fire control. Improved fire control minimizes 
 fire damage, thereby permitting production activities to resume more 
 quickly. 
 
 The National Fire Protection Association classifies fire detec- 
 tion devices into four general categories, as follows (7): 2 
 
 Heat detector Detects abnormally high tempera- 
 tures or rate of temperature rise. 
 
 Flame detector Detects the infrared (IR), ultra- 
 violet (UV), or visible radiation 
 produced by a fire. 
 
 Smoke detector 
 
 Fire gas detector 
 
 Detects the visible or invisible par- 
 ticles of combustion. 
 Detects gases produced by a fire. 
 
 All four fire detector types have application in underground 
 metal and nonmetal mines. The purpose of this paper is to describe 
 recent Bureau research to field test each of the fire detector types 
 in typical underground metal and nonmetal mine applications. The 
 intent is not to provide an exhaustive listing of all detector varia- 
 tions and potential uses, but rather to illustrate with case examples 
 typical uses of each type of detector. Site-specific conditions such 
 as mining method, airflows, combustible materials present, provi- 
 sion for fire warning and evacuation, and mining equipment used, 
 would determine the suitability of a particular detector for a given 
 application. 
 
 HEAT DETECTION 
 
 Industrial-grade thermal fire detection devices are generally 
 characterized by high reliability and durability but low maintenance 
 requirements, even when used under the harshest conditions. 
 Clearly, these attributes are desirable for mining applications. 
 
 One limiting feature of thermal detectors is that they rely on 
 convected thermal energy for response. The distance between the 
 detector and the fire, the relative spatial orientation and placement 
 of the detector relative to the fire, and local air currents profoundly 
 affect detector performance. Heat detectors are, as a result, the 
 preferred choice for small, well-defined fire hazards, especially if 
 they are enclosed, such as electrical boxes. In order to provide for 
 
 ■Group supervisor. Twin Cities Research Center. Bureau of Mines, Minneapolis, 
 MN. 
 
 2 Italic numbers in parentheses refer to items in the list of references at the end of 
 this paper. 
 
 large area coverage, numerous closely spaced detectors are required. 
 A common example of large-area thermal detection in the mining 
 industry is the typical conveyor belt fire detection system mandated 
 for underground coal mines. Spot-type, thermal detectors, spaced 
 at 125-ft intervals along the entry provide early warning of a belt 
 fire. 
 
 A string of spot-type detectors arrayed in a similar manner in 
 a mine shaft is a feasible approach to shaft fire detection, however, 
 the use of a line-type device would offer superior performance. A 
 line-type device senses the heat from a fire at any point along its 
 length. It can be thought of as spot-type detection in the limiting 
 case where the distance between adjoining detectors equals zero. 
 
 A prototype line-type thermistor strip fire detection system for 
 mine shafts was developed by the Bureau and installed along the 
 entire length of the 1 ,200-ft main production shaft of a salt mine 
 
22 
 
 in Detroit. MI (2). The thermistor strip detection system selected 
 was the Alison Control A888-M106 fire detection system. 3 
 
 The control unit was housed in a National Electrical Manufac- 
 turers Association (NEMA) 12 enclosure. A separate annunciator 
 was provided in a NEMA 9 enclosure. The system provided two 
 independent, adjustable levels of alarm (prealarm and alarm) that 
 were annunciated at both the control unit and annunciator. The lo- 
 cation of the overheated area was also indicated in feet above or 
 below ground level on a digital display at the annunciator. 
 
 The sensor was completely supervised. An abnormal condi- 
 tion was indicated at the control unit and annunciator if an open 
 circuit or short occurred anywhere along the entire length of the 
 sensor. All interconnections between the control unit and annun- 
 ciator were also supervised. 
 
 Should the system lose ac input power, the power supply was 
 automatically disconnected and standby batteries (located at the 
 bottom of the control unit) were switched in automatically. The 
 system contained a battery charger that automatically maintained 
 the batteries fully charged when ac power was present. 
 
 The sensor cable consisted of stainless steel tubing containing 
 a specially formulated ceramic thermistor core. A center wire 
 imbedded in the core ran the entire length of the sensor. 
 
 The sensor center-wire-to-case resistance exhibited a negative 
 temperature coefficient. This meant that as the temperature in- 
 creased, the resistance of the sensor decreased exponentially. 
 
 'Reference to specific brand names does not imply endorsement by the Bureau of 
 Mines. 
 
 It was this decrease in resistance that was sensed by the alarm 
 instrumentation. 
 
 The 40-ft sensor sections were connected in series to form 
 two sensor circuits, each 600 ft (15 sections) in length. The entire 
 sensor length and all three junction boxes were coated with a 
 heavy polymer jacket for further protection from the corrosive 
 atmosphere. Figure 1 illustrates the layout of the control panel, an- 
 nunciator, and detection cable. 
 
 The system was completely installed by a three-person crew 
 over a 4-day period. Because this shaft is the main mine exhaust, 
 the air is laden with salt. This highly corrosive atmosphere is 
 detrimental to the operation of electrical systems, necessitating 
 great care in hermetically sealing each detector segment intercon- 
 nection with a silicone adhesive-sealant. All external parts of the 
 detector wire and connections were stainless steel, which was fur- 
 ther protected with a corrosion-resistant fluorocarbon polymer 
 jacket. Following installation, the system was functionally tested. 
 At a known elevation in the shaft, a propane torch was used to 
 heat a section of the detection cable. The prealarm and alarm func- 
 tions operated properly and the hotspot indicator displayed the cor- 
 rect elevation. 
 
 The system was operated continuously for 14 months without 
 hardware failure. Once during that period, a lightning strike at the 
 headframe structure caused a momentary alarm, however, the 
 system returned to normal operating mode without further incident. 
 These test results are significant because they indicate that the hard- 
 ware and installation precautions are suitable for this worst case 
 corrosive environment. 
 
 Headframe and crusher building 
 
 System control panel 
 
 Thermister strip 
 lower detection zone 
 
 Mining level 
 
 Figure 1.— Schematic showing mine shaft linear thermal fire detection system control panel, annunciator, and detection cable. 
 
23 
 
 FLAME DETECTION 
 
 Flame detectors, either UV or IR, are preferred where ex- 
 tremely fast response to a fire is essential because of the likelihood 
 of very rapid fire growth or explosion, as with class B liquid fuels. 
 Typically, flame detectors are connected directly to fast-acting 
 automatic fire or explosion suppression systems in such high-hazard 
 areas as fueling platforms, petrochemical plants, and hyperbaric 
 chambers. An example of the use of flame detectors in underground 
 mines is the Bureau-developed fueling area fire protection system, 
 which utilizes UV detection to trigger the release of twin agent (dry 
 chemical and aqueous film-forming foam) fire suppressants (3). UV 
 detection was selected over IR to minimize false alarms. Sources 
 of false alarms include arc welding for UV detectors and hot sur- 
 faces or gases for IR detectors. It was determined that the system 
 could be disabled during welding operations, but that vehicular 
 traffic, a probable source of hot surfaces and gases, could not be 
 avoided in the fueling area. 
 
 The Detronics U7602 UV detector was selected for the sys- 
 tem (fig. 2). This detector responds to the wavelengths of light in 
 the UV range of 1,850 to 2,450 A. The electronics are housed in 
 an explosion-proof enclosure constructed of two screw-together 
 coaxial cylinders that are heavily plated for corrosion resistance. 
 The UV viewing area is a 90° cone. Two digital alarm modes are 
 provided: one closed relay for fire alarm and one open relay for 
 dirty lens alert. The detector operates on 24-V dc power. The detec- 
 tor utilizes a Geiger-Mueller type tube to sense UV radiation (2, 4). 
 
 The detector is also equipped with an UV test lamp that mon- 
 itors the integrity of the optical lens and deenergizes a relay when 
 
 the surfaces become obstructed with oil, dirt, or dust. The UV test 
 lamp emits UV radiation that passes through the lens, reflects off 
 a beveled reflecting ring mirror, passes back through the lens and 
 into the tube. 
 
 No operational problems were encountered with the detector 
 during extensive laboratory tests, or during in-mine fire tests and 
 long-term in-mine endurance tests at two mines. In each in-mine 
 installation, two detectors were cross-zoned within the control unit 
 to minimize false alarms. Cross-zoning requires that both detec- 
 tors respond to a fire at the same time to actuate the suppression 
 subsystem. The control units were provided with internal adjusta- 
 ble time delay circuits to enable attending personnel to abort the 
 discharge of suppressant if necessary (for example, a false alarm 
 due to welding in the fueling area). 
 
 Fire testing involved igniting diesel fuel contained in a 2- by 
 3-ft pan placed under a load-haul-dump vehicle mockup. Because 
 one of the goals of the in-mine fire tests was to evaluate the ex- 
 tinguishing effectiveness of the twin agent suppressant system, the 
 control unit was operated in the abort mode during these tests. This 
 was necessary because operation in the normal mode would have 
 resulted in almost instantaneous detection and suppressant release. 
 In the abort mode, the fires achieved full fuel involvement and thus 
 represented a more severe test for the suppression system. How- 
 ever, the detectors did respond almost immediately to the test fire 
 flames, and no deterioration in detector performance was observed 
 during the long-term endurance test period. 
 
 SMOKE DETECTION 
 
 One of the earliest products of incipient combustion is sub- 
 micrometer sized particulates, or smoke. Smoke detection systems 
 that are capable of reliably detecting these particulates are extremely 
 valuable because fires can be detected before they reach the flam- 
 ing combustion stage. With the aid of such systems, emergency 
 
 Figure 2.— UV flame detector used in fueling area fire protec- 
 tion system. 
 
 procedures, such as personnel evacuation and fire-fighting efforts, 
 can be undertaken at the earliest opportunity, often before the fire 
 poses a direct threat. A complete prototype smoke detection sys- 
 tem was designed, fabricated, and installed in an underground cop- 
 per mine in Arizona for prolonged testing and evaluation. 
 
 The submicrometer particulate detector selected was the Anglo 
 American Electronics Laboratory Becon MK IV ionization type 
 combustion particle detector (fig. 3). The cylindrical outer casing 
 of the Becon detector is made from nylon-dipped stainless steel to 
 ensure detector longevity in the highly humid and corrosive under- 
 ground mine environment, and to provide a radiation shield. 
 
 Towards the lower end of the Becon detector are vertical rec- 
 tangular ports, which allow mine air to enter the ionization chamber. 
 The ports in the stainless steel shield are internally overlapped by 
 a nylon-dipped stainless steel baffle plate, which shields the areas 
 outside the detector from direct radiation, reduces the effects of 
 high velocity airflow in the ionization chamber, and causes a mix- 
 ing of the mine air inside the ionization chamber. 
 
 Internally, the Becon MK IV particle detector contains a 
 shielded single ionization chamber, a radioactive source, an ion 
 collecting electrode (grid), and a current amplifier. Because of the 
 inherent corrosive nature of the underground mine atmosphere, all 
 internal components of the detector are made of plastic or are 
 hermetically sealed. 
 
 The radiation source, which ionizes the air within the ioniza- 
 tion chamber, is a sealed glass vial containing 5 mCi of krypton 
 85 gas. The design and operating principal of the Becon MK IV 
 is described in reference 2. 
 
 The location for mounting the Becon MK IV detector should 
 be near or on the downwind side of a potential fire hazard area, 
 however the ventilation air velocity in the chosen area should not 
 exceed 1,200 ft/min. 
 
24 
 
 Figure 3.— lonization-type combustion particle detector install- 
 ed in underground mine. 
 
 Because the Becon MK IV detector has no moving parts, very 
 little maintenance is necessary. Periodic examination of the elec- 
 trical cable for breaks or frays and calibration are all that is required. 
 
 The system in the mine consisted of 10 detection instruments. 
 Each detector was equipped with a digital telemetry module 
 (mounted in the detector cap) to convert the detector analog output 
 to a digital word for transmission of the detector value along with 
 a unique address and verification words to the system control unit. 
 A microcomputer system control, a line interface control module 
 
 for communication to the computer through an industry standard 
 protocol (RS232C), a disk drive to store the control program and 
 detector output records, a color video display with graphics to 
 highlight alarms, and a printer to provide hard copy of alarm and 
 fault messages were also provided. 
 
 The detectors were linked to the system control by a single- 
 pair closed-loop telemetry circuit. Connecting outstations in a closed 
 loop minimizes cable costs and installation time and provides a 
 redundant signal path for uninterrupted signal transmission in case 
 of a broken telemetry line. 
 
 The system was completely installed by a four-person crew 
 over a 1-week period. Minor debugging was required following 
 installation because of problems with several telemetry modules, 
 however, the necessary repairs were effected on site during the week 
 following the installation. 
 
 The system operated for approximately 1 yr with a simplified 
 control program while the final version of the control software 
 was developed and debugged. During this period, system opera- 
 tion was limited to a video display of real-time detector outputs 
 and an audible alarm and printout whenever any detector output 
 exceeded its individually programmed alarm threshold. 
 
 The system control software provided video graphics of the 
 detector locations on color mine maps, simple instructions, and 
 three-key coded function commands and alarm, fault, and trouble- 
 shooting messages. Following the on-screen prompts and using the 
 simple three-key commands, operators could display one of the three 
 mine maps covering the system, change any sensor alarm threshold, 
 display 72-h sensor history in tabular or graphic form, and manually 
 control the printer. 
 
 During the first 3 months of system operation, numerous false 
 alarms were issued. The detector-mounted digital telemetry mod- 
 ules, which are susceptible to low voltage conditions, were found 
 to be the cause. Boosting line voltage slightly corrected the prob- 
 lem. This detector has been used by the Bureau in several other 
 research installations and performance in every instance has been 
 excellent (2, 5-5). 
 
 Following this initial "burn-in" period, the system operated 
 for approximately 24 months. During this period, three abnormal 
 events (smoking rubber drive pulleys on two pumps and a smok- 
 ing electrical controls enclosure) were detected by the system. 
 
 FIRE GAS DETECTION 
 
 Like smoke detection, gas detection is particularly useful for 
 large-area coverage in underground mines where ventilation air- 
 flow can transport airborne products of combustion great distances 
 from the source of a fire. 
 
 Bureau mine fire detection research utilizing fire gas detec- 
 tors has included two basic fire detection system configurations: 
 the pneumatic tube bundle approach and the fully electronic tele- 
 metry approach. Pneumatic detection involves sampling the mine 
 atmosphere through a network of plastic tubes that terminate at a 
 central analytic station equipped for gas monitoring. The electronic 
 felemetry approach involves the placement of detection devices at 
 each underground location to be monitored. Detector outputs are 
 transmitted to a central control point over electronic telemetry lines. 
 Such systems may consist of any number of detection devices, with 
 one or more detectors installed at each monitoring point. An ex- 
 ample of each approach is described in the following sections. 
 
 PNEUMATIC DETECTION APPROACH 
 
 Although pneumatic detection systems have been used in 
 aboveground occupancies for many years (factories, ocean vessels, 
 
 etc.), their use in underground mines is fairly recent. Practical, 
 mineworthy pneumatic detection systems were developed about 20 
 yr ago in the United Kingdom for the detection of slowly develop- 
 ing spontaneous combustion fires in coal mines. Considerable 
 research effort has been directed toward application of this approach 
 in North American coal mines (9-1 J), however, it was never widely 
 accepted by industry outside the United Kingdom. The Bureau has 
 recently completed a research program to design and in-mine test 
 a rapid response pneumatic fire detection system tailored to the 
 unique requirements of multilevel metal mines. The prototype 
 pneumatic detection system consisted of three primary subsystems: 
 air sampling, detection, and control. Each subsystem is described 
 in the following sections. 
 
 Air Sampling Subsystem 
 
 The air sampling subsystem was required to draw samples of 
 the mine atmosphere from various underground locations through 
 plastic tubes to an analytic station where the presence and level of 
 combustion gases could be determined. Vacuum pumps were pro- 
 vided in the analytic station for this purpose. All electrical and 
 
25 
 
 mechanical equipment would thus be centralized for ease of main- 
 tenance and removed from the harsh underground environment for 
 improved performance. Polyethylene tubing was selected for its 
 durability, flexibility, light weight, and low cost. A main bundle, 
 consisting of the sample tubes surrounded by Vi in of thermal in- 
 sulation and a tough outer neoprene jacket, was installed in the shaft, 
 with individual tubes branching off on various levels to specific 
 monitoring locations. 
 
 Water traps were required to prevent the accumulation of water 
 in the sample lines. Such accumulations could seriously impair 
 sample flows. This effect would be particularly acute where warm, 
 moist mine air is drawn through sample tubes that are routed in 
 intake air near a mine opening or other area where the air temper- 
 ature is below the dewpoint of the sample. Accumulated water could 
 also freeze, further compounding the problem. 
 
 Standard water traps were modified for this purpose using a 
 specially designed two-way check valve. During normal operation 
 (i.e., under vacuum), a vacuum check valve prevented air leakage 
 into the water trap and hence, dilution of the sample. To empty 
 the traps, the system was periodically cycled into a pressure mode, 
 wherein the entire tubing network was pressurized with compressed 
 air. The water traps were equipped with float-type check valves 
 that permitted accumulated water to be blown out by compressed 
 air pressure but then sealed against pressure loss once the trap was 
 empty. 
 
 Two vacuum pumps were required for system operation: a 
 purge pump and a sample pump. The purge pump maintained a 
 constant flow in the lines, exhausting to the atmosphere. The sam- 
 ple pump drew air samples from each line in sequence, exhausting 
 to the detection instruments. 
 
 Three-way, solenoid-operated valves having low flow resistance 
 were installed in each line. The valves were sequentially cycled 
 by the system control to direct sample gas from one line at a time 
 to the sample pump and gas analyzers, while the remaining flows 
 from the other 1 1 lines passed through the purge pump and were 
 exhausted. 
 
 Detection Subsystem 
 
 It was determined that both CO and C0 2 detection should be 
 incorporated into the prototype detection system. Although both 
 gases are formed in most fires, one or the other gas would likely 
 predominate, depending on the type of fire. Thus, analysis of the 
 ratio of the two gases, along with other data such as the known 
 fire zone, would enable a characterization of any fire that might 
 occur. A system incorporating two detectors would also be in- 
 herently more reliable than one utilizing a single detector, as the 
 two detectors would provide a degree of redundancy. 
 
 Detectors based on the operating principle of nondispersive IR 
 (NDIR) absorption were selected for the prototype pneumatic detec- 
 tion system. Radiation from an IR source is passed through a cell 
 containing the sample of gas to be analyzed, and is absorbed by 
 the gas present. A filtered IR detector responds to this change in 
 radiation and its output is compared with a reference cell, condi- 
 tioned by suitable electronics, and read out on an appropriately 
 marked meter. A stable reading is generally obtained in 2 to 5 s, 
 followed by a rezero in 3 to 4 s. 
 
 NDIR detectors are quick and accurate, sensitivity to 1 pet of 
 full scale can be achieved. For CO, a detection range of to 100 
 ppm was utilized, with the resulting sensitivity being 1 ppm (2 pet 
 of the 50-ppm threshold limit value (TLV)). For C0 2 , a detection 
 range of to 2,500 ppm was utilized, with the resulting sensitivity 
 being 25 ppm (0.5 pet of the 5,000-ppm TLV). Until recently, 
 NDIR detectors were confined to laboratory use only, as they were 
 too delicate to withstand even moderate temperature and humidity 
 variations. Present models are more robust and are designed for 
 
 limited field exposure. Because the detection instruments in a 
 pneumatic system are centralized in a relatively clean environment, 
 this was judged to be an acceptable application for NDIR detection. 
 
 Control Subsystem 
 
 The detection system was controlled by a 64-K RAM micro- 
 computer and associated hardware. The computer operated in a 
 process control mode to monitor the operational status of the vari- 
 ous detection system components (pumps, analyzers, etc.), cycle 
 the solenoid valves in the proper time sequence, initiate gas analyzer 
 calibration and water trap blowout routines, and issue alarm and 
 trouble warnings. The computer also stored system data and pro- 
 vided the user with several menu-selectable video display, system 
 output, and system control options. 
 
 To insure accuracy in measuring gas concentrations, the system 
 also controlled a pair of automatic gas analyzer calibrators. When 
 activated by the computer, these calibrators automatically standard- 
 ized the gas analyzers using calibration gases contained in high- 
 pressure gas storage cylinders. 
 
 Figure 4 shows the system control room containing the pumps, 
 valves, gas analyzers, autocalibrators, and computer. 
 
 In-Mine Test Results 
 
 The prototype pneumatic detection system was installed and 
 functionally tested at a multilevel underground zinc mine in Og- 
 densburg, NJ. The system was installed by a four-person crew over 
 a 4-week period. The system monitored 18 locations through 12 
 tubes. 
 
 A total of 39,900 ft of tubing was used in the prototype system. 
 About 80 pet of the total was contained in the main bundle in the 
 shaft, with the remaining 20 pet being single tubes that branched 
 off on various levels and led to specific monitoring locations. Tubes 
 contained within the main bundle were color coded and numbered 
 to facilitate installation and subsequent system layout changes. 
 
 Overall system performance was found to be satisfactory dur- 
 ing the 3-yr in-mine evaluation period, in that the capability to 
 monitor and record elevated CO and C0 2 levels at the desired 
 underground locations, and warn mine personnel when these levels 
 exceeded specified thresholds, was demonstrated. 
 
 System operations were closely monitored throughout the 3-yr 
 evaluation period. Late afternoon excursions in both the CO and 
 C0 2 levels, corresponding to end-of-shift blasting operations, were 
 a daily qualitative check on system performance. 
 
 The provision for blowing water from the sample tubes proved 
 to be quite effective. Inspection of the tubes shortly after the sys- 
 tem was commissioned, but prior to the installation of the water 
 traps, revealed accumulations of water at low points where the 
 tubes sagged. Following installation of the traps, such accumula- 
 tions were not completely eliminated, however enough of the water 
 was removed by the traps that system performance was not im- 
 paired by accumulated water. Purging the tubes with high-pressure 
 (80 psi) compressed air cleared both the traps and other water 
 accumulations. 
 
 Leaks in the tubing system were noted throughout the test 
 period. These leaks, along with longitudinal mixing of gases within 
 the tubes, made the direct determination of gas concentration at 
 the sampling point impossible. However, each sampling point could 
 be calibrated with a test gas of a known concentration and subse- 
 quent measurements adjusted accordingly in the control software 
 if a precise quantitative measure of concentration is needed. 
 Significantly, the leaks that did occur did not affect the transit time 
 of the samples in the tubes, meaning that a timely warning would 
 be received even if a system was prone to leaks. All tubing system 
 
26 
 
 Figure 4.— Control room for pneumatic detection system showing pumps, valves, gas analyzers, autocalibrators, and computer. 
 
 leaks were traced to faulty connections between individual lengths 
 of tubing. Prevention of leaks could be accomplished by using longer 
 lengths of tubing (therefore requiring fewer connections) and us- 
 ing better connectors. 
 
 Overall system maintenance requirements were minimal. Mine 
 personnel made a visual check of the system about once per week, 
 and calibration gas tanks required replacement every 2 months. At 
 that time, the analyzers were manually calibrated and the vacuum 
 pumps were checked for dirt and debris in their in-line filters. 
 
 Reliability of the gas analyzers exceeded the project's design 
 goals. After 6 months of operation, the analyzer cells were removed 
 and cleaned; however, no subsequent maintenance was required 
 or performed. Autocalibrator reliability also exceeded design goals, 
 with the only problem being a faulty control valve, which was 
 discovered and replaced during installation. 
 
 The computer control functioned continuously throughout the 
 test period without difficulty. Power outages at the mine site oc- 
 curred occasionally, however, the self-booting feature of the con- 
 trol software functioned as designed, and the system automatically 
 returned to a state of full operational readiness when power was 
 restored. 
 
 ELECTRONIC TELEMETRY APPROACH 
 
 The electronic telemetry approach is by far the more common 
 type of fire gas detection. Indeed, with respect to current industry 
 practice, the pneumatic approach is limited to deep coal mines in 
 the United Kingdom, whereas the electronic telemetry approach 
 
 is widely used in the major mining districts, both coal and non- 
 coal, worldwide (2, 12-13). The electronic telemetry approach 
 enables the monitoring of any gas, particulate, or condition (air 
 velocity, direction, temperature, humidity, etc.) for which a detec- 
 tion instrument exists. Equally important are the communication 
 and control functions, such as telephones, pump switches, fans, 
 and doors, that can utilize the same telemetry lines. Where telephone 
 lines are already installed, fire detectors can usually be added at 
 a nominal cost without affecting voice communication. 
 
 Telemetry -type fire gas detection systems, especially CO 
 systems, are becoming increasingly popular in both coal and non- 
 coal mines in the United States. An example of this technology is 
 a fire gas detection system installed by the Bureau at an Idaho silver 
 mine for long-term evaluations. The detectors were installed on the 
 1900 level of the mine in the main exhaust. In this location, the 
 detectors were subjected to quite severe environmental exposures: 
 air velocity of 800 ft/min, air temperature of 85 ° F, and saturated 
 humidity. 
 
 The system included both NDIR analyzers (of the type used 
 in the pneumatic detection system) and electrochemical detectors 
 for the continuous monitoring of CO. The units were placed side 
 by side at the same location for comparison purposes. 
 
 Because the NDIR unit is designed for, at most, limited field 
 exposures, it was necessary to enclose it in an environmental hous- 
 ing. Because of the extreme humidity at the test site, a stainless 
 steel NEMA 12 box was selected. The analyzer was placed inside, 
 along with a sample pump and calibration gas tanks. The finished 
 assembly was 2 by 3 by 4 ft and weighed over 160 lb. 
 
27 
 
 Three electrochemical cell units were included in the system: 
 Ecolyzer 5000, Ecolyzer 4000, and MSA 571 . All three detectors 
 utilize the electrochemical properties of a fuel cell to sense CO. 
 The electrochemical sensor is constructed of three electrodes— the 
 sensing electrode, the reference electrode, and the counter 
 electrode — all suspended in an acid solution. The materials to be 
 chemically reacted are CO and O gases from the mine ambient air. 
 These gases diffuse into the acid (or in the case of the Ecolyzer 
 4000. are pumped into the fuel cell by an air pump) solution and 
 ionize. 
 
 The CO is electrochemically oxidized at the sensing electrode 
 while O reduction occurs at the counter electrode. The ion concen- 
 tration in the acid solution, because of the dissolved gases, is pro- 
 portional to the concentration of CO in the air; likewise, the cur- 
 rent flow through the cell is proportional to the ion concentration 
 in the solution. Therefore, the current flow through the cell is pro- 
 portional to the CO content of the air. This current flow is then 
 amplified and compensated for temperature before it is sent to the 
 sensor control. All three units are supplied by the manufacturer 
 in rugged environmental housings and are intended for installation 
 where harsh environmental exposures are expected. Figure 5 shows 
 an Ecolyzer 4000 installed underground. 
 
 All the detectors were linked to the surface through a digital 
 telemetry system similar to that used for the smoke detection system 
 described earlier. 
 
 Throughout the 1-yr test period, all units functioned properly. 
 Calibration was performed approximately 24 h after powerup, as 
 previous research had shown that considerable drift could be ex- 
 pected during the first 24 h of operation. All units tracked closely, 
 with end-of-shift blasting a daily qualitative performance indicator. 
 Based strictly on performance, no basis could be established for 
 choosing one detector over another. However, cost considerations 
 clearly favored the electrochemical units. 
 
 Figure 5.— Electrochemical cell-type CO detector installed in 
 underground mine. 
 
 SUMMARY AND CONCLUSIONS 
 
 Early fire detection through the use of specialized fire detec- 
 tion instruments can play an important part in a mine's overall mine 
 fire emergency plan. It provides sufficient advance warning of an 
 emergency to enable safe evacuation and effective fire fighting. 
 
 Numerous commercially available detection devices have been tested 
 in a variety of mine settings with such consistent success that early 
 fire detection and warning must be considered a proven technology 
 whose application can be recommended industrywide. 
 
 REFERENCES 
 
 1. National Fire Protection Association. Standard on Automatic Fire 
 Detectors. National Fire Code NFPA 72E-1984, 1984, 48 pp. 
 
 2. Pomroy, W. H., and R. E. Helmbrecht. Design and Operation of 
 Four Prototype Fire Detection Systems in Noncoal Underground Mines. 
 BuMines IC 9030, 1985, 25 pp. 
 
 3. The Ansul Co. Improved Fire Protection System for Underground 
 Fueling Areas. Volumes 1 and 2 (contract H0262023). BuMines OFR 
 120-78, 1977, 325 pp., NTIS PB 288298; BuMines OFR 160-82, 1981, 
 111 pp., NTIS PB 83-114744. 
 
 4. Baumeister, T., and L. S. Marks (eds.). Standard Handbook for 
 Mechanical Engineers. McGraw-Hill, 7th ed., 1967, pp. 16-29. 
 
 5. Pomroy, W. H. Spontaneous Combustion Fire Detection for Deep 
 Metal Mines. BuMines IC 9144, 1987, 25 pp. 
 
 6. Johnson, G. A., and D. R. Forshey. Inmine Fire Tests of Mine Shaft 
 Fire and Smoke Protection Systems. BuMines IC 8783, 1978, 17 pp. 
 
 7. Stevens, R. B. Demonstration of a Mine Shaft Fire and Smoke Pro- 
 tection System for Coal Mines (contract H0100017. ESD Corp.). BuMines 
 OFR 116-85, 1985, 294 pp.; NTIS PB 86-146933. 
 
 8. . Mine Shaft Fire and Smoke Protection System. (Final 
 
 Report.) Volume II— Validation Testing and Cost-Effectiveness Evalua- 
 tion (contract H0242016, FMC Corp.). BuMines OFR 73(l)-78, 1987, 210 
 pp.; NTIS PB 284 166. 
 
 9. Litton, C. D. Design Criteria for Rapid-Response Pneumatic Monitor- 
 ing Systems. BuMines IC 8912, 1983, 23 pp. 
 
 10. Hertzberg, M., and C. D. Litton. Pneumatic Fire Detection With 
 Tube Bundles, J. Fire and Flammability. v. 9, Apr. 1978, pp. 199-216. 
 
 11. Chakravorty, R. N., and R. L. Woolf. Evaluation of Systems for 
 Early Detection of Spontaneous Combustion in Coal Mines. Paper in the 
 Proceedings of the 2d International Mine Ventilation Congress (Reno, NV, 
 Nov. 4-8, 1979), ed. by P. Mousset-Jones. Soc. Min. Eng. AIME (Little- 
 ton, CO), 1980, pp. 429-436. 
 
 12. Burrows, J. (ed.) Environmental Engineering in South African Mines. 
 Mine Vent. Soc. South Africa, Cape and Transvaal Printers (Pty) Ltd., Cape 
 Town, South Africa, 1982, p. 884. 
 
2S 
 
 DIESEL-DISCRIMINATING FIRE SENSOR 
 
 By Charles D. Litton 1 
 
 ABSTRACT 
 
 This Bureau of Mines paper describes a novel fire detector that can be used to 
 discriminate between smoke produced by a fire and smoke produced by a diesel engine. The 
 detector uses a pyrolysis technique whereby a sample of smoke-laden gas passes through a 
 short, heated tube within which fire smoke particles pyrolyze and increase their number 
 concentration and decrease their average size, while diesel smoke particles are unaffected. 
 The detector is designed for use in mines that use diesel-powered equipment where the 
 detection of fires is complicated because of the diesel emissions background levels of smoke 
 and other products of combustion. 
 
 BACKGROUND 
 
 Diesels produce smoke, CO, CO z , and other combustion 
 products. In an underground mine that uses diesel-powered 
 equipment, these diesel exhaust products can mix with the 
 ventilation airflow resulting in concentration levels sufficient 
 to produce frequent false alarms of product-of-combustion 
 fire sensors, such as smoke and CO sensors. Further, if a fire 
 were to occur in the presence of these elevated product levels, 
 its detection could go unheeded, reaching a substantial size 
 before it is finally detected. As a result, early warning capa- 
 bilities (crucial to life safety) are seriously compromised. The 
 problems of false alarms and degradation of early warning 
 capabilities in diesel-operated mines require that some type of 
 fire sensor be used that essentially ignores the diesel back- 
 ground levels yet remains very sensitive to combustion prod- 
 ucts produced by fires. 
 
 Previous contract research by the Bureau of Mines 2 
 identified the ability of fire smoke particles to pyrolyze upon 
 passage through a short, heated tube whose surface tempera- 
 ture was held constant between 300° and 350° C. Upon 
 passage through this pyrolysis tube, the number concentra- 
 tions, n , of fire smoke particles increased dramatically. At the 
 same time there was a corresponding reduction in the number 
 mean particle diameter, d . Diesel smoke particles were unaf- 
 fected upon passage through the same pyrolysis tube. 
 
 1 Supervisory physical scientist, Pittsburgh Research Center, Bureau of Mines, 
 Pittsburgh, PA 
 
 2 Skala, G. F., and F. W. Vanluik, Jr. Development of Selective Submicrometer 
 Particulate Fire Detectors for Underground Metal Mines (contract H0387O25, 
 Environment/One Corp.). BuMines OFR 58(1 )-83, 1979, 129 pp.; NT1S PB 
 83-178947. 
 
 A prototype detector was subsequently built based upon 
 this pyrolysis principle. The incoming smoke-laden gas was 
 split into two parallel paths, one path containing the pyrolysis 
 tube, the other path being a straight section of tubing. The 
 concentration of smoke along each path was measured by a 
 cloud condensation nuclei counter and an alarm threshold for 
 fire set at a ratio of pyrolyzed to unpyrolyzed concentrations 
 equal to 1.5. Even though the prototype functioned as ex- 
 pected, it was not well suited to the mine environment. 
 
 At about this same time, Bureau researchers had devel- 
 oped and patented a sensitive smoke detector for use under- 
 ground. This sensor uses an ionization chamber to efficiently 
 charge smoke particles and measure their concentration. In 
 August 1983, the U.S. Patent Office granted an exclusive 
 license to a major manufacturer of mine monitoring equip- 
 ment to develop and market this detector commercially. This 
 detector's response is a function of the size and concentration 
 of the smoke particles and its principles of operation have been 
 discussed in detail elsewhere. 3 Basically, the response increases 
 with increasing smoke particle diameter and concentration. 
 Because the pyrolysis tube increases the fire smoke concentra- 
 tion but decreases the average particle diameter, could this 
 sensor be used to measure both pyrolyzed and unpyrolyzed 
 smoke particles and if so, would it be sensitive enough to use 
 as a discriminating, early warning fire sensor in diesel- 
 operated mines. A series of laboratory experiments were then 
 conducted to answer this question. 
 
 3 Litton, C. D., L. Graybeal, and M. Hertzberg. Submicrometer Particle 
 Detector and Size Analyzer. Rev. Sci. Instrum., v. 50, No. 7, July 1979, pp. 
 817-823. 
 
29 
 
 LABORATORY EXPERIMENTS 
 
 The system used to conduct the initial laboratory experi- 
 ments is shown in figure 1 . It consisted of a cubical chamber, 
 45 cm on each edge, and constructed of plexiglass. Exhaust 
 from a diesel-operated generator could be directed into the 
 chamber via a short tube. Inside the chamber, a hotplate was 
 used to heat samples of mine combustibles. Directly above the 
 hotplate, a sample tube continuously pulled gas samples from 
 the chamber for analysis by the Bureau smoke detector. 
 Outside the chamber, this sample tube split into two parallel 
 paths; with one path of flow containing the pyrolysis tube and 
 the second path, a plain, unheated section of tubing. Flow was 
 diverted along one path or the other by turning valve A or B. 
 In this manner, the smoke detector could sample the unpyro- 
 lyzed smoke particles and the pyrolyzed smoke particles. The 
 
 pyrolysis tube was a 2.8-cm-long piece of 0.64-cm-diam stain- 
 less steel tubing heated resistively to a surface temperature of 
 350° C. 
 
 Combustible samples of coal, wood, conveyor belt, and 
 plastic line brattice were tested both in the presence of diesel 
 smoke and without diesel smoke. Measurements of diesel 
 smoke alone were also made during these series of experi- 
 ments. In a typical test, the combustible material was heated to 
 a stage of sustained smoldering and then the unpyrolyzed and 
 pyrolyzed signals measured alternately by switching valves A 
 and B. With tests in the presence of diesel smoke, a steady- 
 state diesel smoke level was established and the combustible 
 material was then heated in the same manner and sampled 
 alternately along each path. 
 
 LABORATORY STUDIES 
 
 These tests verified that diesel smoke showed no effects of 
 pyrolysis. Various diesel smoke levels through the pyrolysis 
 path and through the unpyrolyzed path were constant, indi- 
 cating that diesel smoke showed no pyrolysis behavior. 
 
 In the experiments, the particle sizing capabilities of the 
 smoke detector were used to determine the average diameters 
 of the smoke particles. For diesel smoke, the average diameter 
 was found to be 0.23 ^jm. For the fire smoke, the average 
 diameters varied from -0.05 to -0.70 /xm. 
 
 Fire smoke particles produced in the absence of diesel 
 smoke were found to pyrolyze linearly with their average 
 diameter, d . A plot of the data obtained for the four 
 combustible materials tested is shown in figure 2, where G n 
 represents the ratio of number concentrations of the particle 
 leaving the pyrolysis tube, n' Q , to the number concentrations 
 entering the pyrolysis tube, n . That is, G„ o = n' /n . A linear 
 curve fit of this data yielded 
 
 Unpyrolyzed flow 
 
 n'/m = G„ = 55-d„ 
 
 (1) 
 
 where d = initial unpyrolyzed fire smoke particle diameter, 
 nm. Further, the average particle diameter, d' , of the repyro- 
 lyzed smoke particles was found to vary according to 
 
 d'„ = 0.27 d 2/3 , 
 
 (2) 
 
 with d' Q and d in micrometers. 
 
 To summarize, the larger the average fire smoke particle 
 diameter, d , the greater is the increase in number concentra- 
 tion upon passage through the pyrolysis tube. The pyrolyzed 
 smoke particles have a smaller average particle diameter, but 
 the increase in concentration is sufficiently great that the 
 effects of particle size are negligible. 
 
 Now, when a combination of fire smoke particles and 
 diesel smoke particles are present, only the fire smoke particles 
 pyrolyze. If the concentration of fire smoke particles is low 
 relative to the concentration of diesel smoke particles, the ratio 
 of pyrolyzed to unpyrolyzed signals will be low. At the other 
 extreme, if the concentration of fire smoke particles is high 
 relative to the concentration of diesel smoke particles, this 
 ratio will be higher and will approach the value obtained if no 
 diesel smoke is present. However, the difference between 
 pyrolyzed and unpyrolyzed signals is not significantly affected 
 by the concentration of diesel smoke particles since it is the fire 
 smoke particles only that pyrolyze. 
 
 Sample tube 
 
 Exhaust 
 
 Pyrolyzed & 
 
 Diesel 
 generator 
 
 \t{— — Hotplate 
 
 9-m 3 with sam P |e 
 chamber 
 
 Pyrolysis 
 tube 
 
 Figure 1 .—Experimental setup for measuring response char- 
 acteristics of Bureau's detector to smoke particles passing 
 through a small pyrolysis chamber. 
 
 'no 
 
 60 
 40 
 
 20 
 
 IO 
 8 
 
 6 
 4 
 
 i — < — i — ' i ' i 1 1 1 1 — i — i — i— r 
 
 j i 
 
 KEY 
 o Wood smoke 
 • Coal smoke 
 ■ SBR conveyor 
 
 belt smoke 
 a PVC line brattice 
 
 smoke 
 
 _l I I I I L 
 
 0O2 
 
 004 006 008 0.10 
 
 0.20 
 
 0.40 0.60 0.80 1.00 
 
 do, /Am 
 
 Figure 2. — Ratio of number concentration of smoke particles 
 pyrolyzed to number concentration entering the repyrolysis 
 element as a function of the average diameter of the particles 
 entering the pyrolysis tube. 
 
30 
 
 These laboratory results indicate that a pyrolysis detector 
 using the Bureau smoke detector as the primary sensor and 
 utilizing the difference between pyrolyzed and unpyrolyzed 
 
 signals to distinguish between fire smoke and diesel smoke has 
 the potential to reliably detect fires in the presence of signifi- 
 cant diesel background levels. 
 
 PYROLYSIS DETECTOR 
 
 The use of the pyrolysis concept requires that both a 
 pyrolyzed and an unpyrolyzed signal be measured. One option is 
 to devote a separate detector to each sampling line. This option, 
 however, requires not only two sensors, but two sensors whose 
 response characteristics are very similar (ideally, identical). 
 
 Now, the conventional Bureau smoke detector utilizes an 
 ionization chamber consisting of a set of parallel plate elec- 
 trodes with one electrode having 5 piCi of americium 241 
 deposited uniformly on its surface. To create two identical 
 chambers, a piece of fluorocarbon polymer, 1/8-in-thick, was 
 used to divide the original chamber into two separate, distinct 
 chambers. Each of these new chambers has its own electrically 
 isolated negative electrode, while the positive, radioactive 
 electrode is shared by the two chambers so that only one 
 common voltage source is needed to power both chambers. 
 The result is two identical, independent ionization chambers, 
 one chamber continuously measuring the pyrolyzed smoke 
 particles while the other measures the unpyrolyzed smoke 
 particles. 
 
 The original pyrolysis tube required ~ 45 A at a voltage of 
 0.27 V to maintain the 300° C temperatures necessary for 
 operation. To reduce the current requirements for the pyrolysis 
 tube, a new tube was fabricated which utilizes a 2.0-in-long, 
 1/32-in-diam ceramic rod. Nichrome wire of 0.008 in diameter 
 and ~9.0 in long is wound around the ceramic rod and the 
 ends of the nichrome electrically connected by a 6-V power 
 supply which consumes -0.70 A. This rod is then inserted 
 into the air space between two swagelock fittings. At the gas 
 inlet port to the new sensor, a T-connection allows one-half of 
 the flow to pass through this new pyrolysis tube to its 
 ionization chamber while the other half of the flow passes 
 through a plain section of tubing to the other ionization 
 chamber. 
 
 Figure 3fi shows the new pyrolysis tube connected to the 
 inlet of one of the ionization chambers. Figure 3A shows the 
 dual ionization chamber with an attached pyrolysis tube 
 mounted in a housing with the electronics, internal pump, and 
 other necessary components. 
 
 PROTOTYPE OPERATION 
 
 When no smoke particles are present in the air being 
 sampled, the output signals of both chambers are amplified 
 and electronically adjusted to identical levels of ~ 7.0 V. When 
 smoke enters either chamber, the signal levels decrease accord- 
 ing to the expression 
 
 where 
 V c 
 
 d 
 "o 
 
 V = (V /k d n )(l -exp(-k d nj), 
 
 (3) 
 
 steady state chamber signal in the absence of 
 
 smoke, ~7.0 V; 
 
 smoke particle diameter, cm; 
 
 smoke particle concentration, p/cm 3 , 
 
 chamber constant, 0.0025 cm 2 /p. 
 
 and 
 
 For example, at an average smoke particle diameter of 
 d = 0.20 pim (2 X 10" 5 cm), a 1-pct reduction in unpyrolyzed 
 signal occurs when the smoke concentration is 4 x 10 5 p/cm 3 . 
 However, from equations 1 and 2, the value of n' when d is 
 0.2 ftm is n' D = 4.4 x 10 6 and d' Q is 0.092 ^m (9.2 X 10" 6 cm) 
 so that d' -n' is 40.6 and the corresponding signal reduction 
 in the pyrolysis chamber is — 5 pet. This means that the 
 unpyrolyzed signal minus the pyrolysis signal is -0.28 V. 
 Subsequent data acquired with the prototype indicates that an 
 alarm threshold around 0.3 to 0.5 V appears reasonable. 
 
 * # 
 
 Figure 3.— Photographs of the major components of the pyrolysis fire detector. A, Dual ionization chamber; 8, new 
 pyrolysis tube connected to inlet of one ionization chamber. 
 
31 
 
 PROTOTYPE EVALUATION 
 
 The prototype pyrolysis fire detector was subsequently 
 evaluated in a series of intermediate-scale tests. For these tests, 
 a diesel exhaust was diverted, via a 4-in-diam flexible hose, 
 into a ventilated intermediate-scale fire tunnel. 4 The resultant 
 difference signal due to diesel smoke levels was measured with 
 the detector prototype. 
 
 After a steady-state diesel level has been established, 
 approximately 9 to 10 kg of coal was heated with an imbedded 
 electrical strip heater, producing both smoldering and eventual 
 flaming combustion of the coal. Figure 4 is typical of the 
 relative increases in both pyrolysis difference voltage and part 
 per million CO, as the coal mass begins to smolder. 
 
 It is worth noting that the difference in signal due to diesel 
 operations only is less than the zero level signal established due 
 to only ambient air while the diesel-only CO level is 115 ppm. 
 When the CO has increased to 15 ppm above this ambient 
 level, the pyrolysis detector signal is 1.60 V. 
 
 In another test, the coal fire was allowed to develop 
 without any diesel background. The data are shown in figure 
 5. The results of this test are essentially the same. At 19 min 
 into the test, the diesel was turned on and the CO level 
 increased from 75 ppm to 230 ppm while the pyrolysis detector 
 was virtually unaffected. 
 
 Several subsequent intermediate-scale tests have been con- 
 ducted to determine the response of the pyrolysis detector to 
 
 4 Egan, M. R. Transformer Fluid Fires in a Ventilated Tunnel. BuMines IC 
 9117, 1986, 13 pp. 
 
 10 15 
 TIME, min 
 
 Figure 4.— Response of a CO detector (A) and the pyrolysis 
 fire detector (B) to a developing fire in the presence of diesel 
 exhaust combustion products. 
 
 smoldering combustibles. Table 1 shows the data obtained 
 during steady-state smoldering at combustible surface temper- 
 atures of -300° C. 
 
 Table 1 . — Average pyrolysis difference voltages, 
 
 smoke diameters, and concentrations obtained for 
 
 smoldering combustion 
 
 Sample v"V, V d , nm n c , p/cm 3 
 
 Styrene butadiene rubber belt 0.45 0.14 2.0 x 10 6 
 
 Neoprene belt 2.52 .52 8.5 x 10 s 
 
 PVC belt 1.05 .31 7.5 x 10 5 
 
 Plywood 3.30 .30 2.2 x 10 6 
 
 Coal 2.50 .35 1.3 x10 6 
 
 225 
 
 i i 
 _ A 
 
 ■ i ■ i ■ i 
 
 1 
 
 200 
 
 - 
 
 
 ' - 
 
 I75 
 
 - 
 
 
 - 
 
 E I50 
 
 — 
 
 Diesel | 
 
 - 
 
 Q. 
 
 t I25 
 o 
 
 
 an 
 flan- 
 First 
 
 i 
 ie / 
 
 - 
 
 ° I00 
 
 
 visible 
 
 smoke name 
 
 
 
 75 
 
 
 
 15-ppm 
 
 
 
 
 50 
 
 
 
 CO alarm 
 
 
 
 ' 
 
 25 
 
 
 
 
 
 
 
 - 
 
 
 2.6 
 
 ■ 1 1 i 
 
 
 
 i i 
 
 i 
 
 i 
 
 i 
 
 
 
 
 
 
 
 1 
 
 1 
 
 1 
 
 i i 
 
 i 
 
 fJ \ . 
 
 i 
 
 2.3 
 
 _ B 
 
 
 
 
 J 
 
 V 
 
 V*- 
 
 > 
 
 
 
 
 / 
 
 u/V 
 
 
 
 - 2.0 
 
 - 
 
 
 
 / 
 
 rV r 
 
 
 — 
 
 _i 
 
 
 
 
 r 
 
 
 
 
 < 1 .7 
 
 — 
 
 
 
 J 
 
 
 
 — 
 
 z 
 
 
 
 
 •r 
 
 
 
 
 <2 |. 4 
 
 — 
 
 
 
 / 
 
 
 
 - 
 
 (/) 
 
 
 
 
 ' 
 
 
 
 
 UJ I.I 
 
 - 
 
 
 
 
 
 
 - 
 
 o 
 
 
 
 
 
 
 
 
 2 .8 
 
 
 
 
 
 
 
 — 
 
 UJ 
 
 
 
 
 
 
 
 
 UJ 3 
 
 - 
 
 
 
 
 
 
 - 
 
 U_ 
 
 
 
 
 
 
 
 
 u- .2 
 
 - 
 
 
 
 
 
 
 — 
 
 -.1 
 
 ^^ 
 
 
 
 
 
 
 _ 
 
 — a 
 
 i 
 
 i i 
 
 n i 
 
 i i \ 
 
 N , 
 
 M 
 
 i 
 
 
 
 10 15 
 
 TIME, min 
 
 20 25 
 
 Figure 5.— Response of a CO detector (A) and the pyrolysis 
 fire detector (B) to a developing fire when no diesel combustion 
 products are present. 
 
32 
 
 CONCLUSIONS AND DISCUSSION 
 
 The concept of repyrolysis of fire smoke particles appears thus reducing, if not totally eliminating, the problems of false 
 
 to be a valid concept for rapid and reliable detection of fires in alarms of fire sensors due to diesel-produced combustion 
 
 mines using diesel-powered equipment. The low pyrolysis products. Initial tests to evaluate the performance of a proto- 
 
 temperatures have no positive effect on diesel smoke particles, type pyrolysis fire detector have been very encouraging. 
 
33 
 
 COMPUTER-AIDED MINE FIRE SENSOR DATA 
 INTERPRETATION IN REAL TIME 
 
 By L.W. Laage, 1 W.H. Pomroy, 2 and A.M. Bartholomew 3 
 
 ABSTRACT 
 
 Throughout recorded history, underground fires have plagued mining operations. Compared 
 with other hazardous situations in underground mining, a fire can become a global problem by 
 swiftly spreading deadly carbon monoxide and other products of combustion (POC's) throughout 
 the whole mine, often without warning. Experience has shown that when fires are detected and 
 located in their early stages, they are much easier to control and proper escape routes can be more 
 intelligently selected. Recent advances in sensor and data communication technology have made 
 reliable mine fire detection system installations possible. Unlike building construction, in a mine, 
 it is impractical to install detectors at every desired location. Abandoned workings and unsafe loca- 
 tions preclude sensor installations from both safety and economic standpoints. The net effect is 
 that some fires are detected and located early while the location of others, even if detected early, 
 remains unknown too long for effective evacuation and fire fighting. 
 
 This paper discusses recent research by the Bureau of Mines to develop a strategy to locate 
 mine fires in real time using a minimum of selectively placed sensors coupled with computer-aided 
 data interpretation. 
 
 INTRODUCTION 
 
 Underground mining continues to rank among the most hazard- 
 ous of all industrial occupations (I). 4 This disparity is particularly 
 evident in the case of fires. During 1984 the incidence rate for in- 
 dustrial fires in the United States, expressed as the number of fires 
 per 200,000 worker-hours, was 0.095 (2). During the same year, 
 
 'Mining engineer. 
 
 Hjroup supervisor. 
 
 'Statistical assistant. 
 Twin Cities Research Center, Bureau of Mines, Minneapolis, MN. 
 
 'Italic numbers in parentheses refer to items in the list of references at the end of 
 this paper. 
 
 the fire incidence rate for underground metal and nonmetal mines 
 was 0. 15 1 ; 60 pet higher than for general industry (3). The poten- 
 tial for disaster is compounded by the limited number of possible 
 escape routes and limited fresh air supplies available underground. 
 In the event of a fire, fresh air supplies and safe travelways are 
 actually decreasing. 
 
 This paper demonstrates a computer simulation technique for 
 locating underground mine fires utilizing an array of strategically 
 placed underground fire detectors. This technique provides an 
 automated means by which the network branch, or set of network 
 branches terminating in a single junction, in which a mine fire occurs 
 can be determined quickly and with reasonable confidence. 
 
34 
 
 ACKNOWLEDGMENT 
 
 The authors wish to acknowledge the assistance of John R. 
 Marks, chief ventilation and health engineer, Homestake Mining 
 Co., for providing detailed ventilation data for the Homestake Mine 
 
 in Lead, SD, which was the subject of the case study experiment 
 discussed in this paper. 
 
 BACKGROUND 
 
 Rapid determination of the location of an underground mine 
 fire has long been the dream of mining personnel. If the location 
 is known, preferred escape routes can be designated, rescue teams 
 can concentrate searches in areas of greatest need, and firefighters 
 can select the most efficient route to the fire and the most effective 
 fire-fighting strategy. Also, once the location is known, other related 
 aspects of the fire such as intensity, fuel, and growth rate can be 
 inferred. 
 
 Occasionally, fires occur at an active working face, shop, or 
 other area that is under direct observation. In these cases, immediate 
 action can be taken to notify appropriate mine officials regarding 
 the fire's location and other relevant information. However over 
 one-half of the fires in underground metal and nonmetal mines 
 reported to the Mine Safety and Health Administration occur in 
 inactive or otherwise unoccupied areas (i). Miners eventually see 
 or smell the smoke, however it is impossible for them to know, 
 without further investigation, its origin. 
 
 Fire bossing (the systematic inspection of mine workings for 
 fire) is the only means available for determining the location of an 
 underground mine fire. The shortcomings of fire bossing are 
 threefold: it is slow (especially in large mines); inspectors are subject 
 to severe fire, smoke, and toxic gas hazards; and, ironically, it is 
 often unsuccessful in determining the exact location of a fire. The 
 latter occurs because progress of the fire boss is stopped if heavy 
 smoke is encountered. 
 
 Fire bossing is more effective in mines equipped with 
 strategically located electronic fire detection devices, as the search 
 for fire can be restricted to the general area of the detector(s) in 
 alarm. Over the past 15 yr, use of electronic fire detection devices 
 in underground mines has expanded from a few research installa- 
 tions to become an accepted industry practice (4). Systems employ- 
 ing 40 or more monitoring points have been installed in some mines, 
 with coverage ratios (ratio of number of detectors to number of 
 network branches in a mine) typically ranging from 1:50 to 1:10. 
 
 However, even in mines equipped with fire detectors, personnel 
 must still be relied upon to locate the fire. The shortcomings of 
 fire bossing are somewhat mitigated by the use of fire detection 
 devices; however, the overall effectiveness of fire bossing is 
 inherently limited. 
 
 The simplest solution to the problem of quickly and safely 
 locating a fire's source is to install a fire detector in every network 
 branch in the mine. However as most mines comprise hundreds 
 of branches, installation of a detector in every branch would be 
 cost prohibitive. 
 
 A more practical approach involves integration of a fire detec- 
 tion system with a ventilation network analysis computer model. 
 In the event of fire, POC's flow along predictable courses in a mine, 
 following the ventilation. It is hypothesized that by carefully noting 
 the pattern and timing of alarms from a few strategically placed 
 detectors with reference to known ventilation flows, it should be 
 possible to determine the network branch in which a fire is located. 
 Although fires can themselves alter ventilation flow conditions, thus 
 interfering with such a determination, it is further hypothesized that 
 
 during the early stages of a fire the initial distribution of POC's 
 would precede ventilation disturbances. 
 
 Because fires in their initial stages typically produce copious 
 amounts of smoke and CO but generate little heat to disturb the 
 ventilation flow, this second hypothesis was considered reasonable. 
 However, for further support, a study of the data base of reported 
 metal and nonmetal underground mine fires from 1950 to 1984 was 
 conducted to determine, where possible, whether this slow grow- 
 ing fire scenario was valid (3). From a practical standpoint, only 
 fires not immediately detected would be located with this strategy. 
 This strategy would apply to slow growing fires with modest flame 
 spread rates, such as burning timber, insulation, or conveyor belting. 
 Sites with a potential for fast growing fires with high flame spread 
 rates, such as fuel storage locations, would require individual 
 detectors. 
 
 The method of fire detection was divided into three categories: 
 (1) Worker not in immediate area, incoming shift workers, and shift 
 boss, foreman; (2) operator or worker in immediate area and 
 welding crew; and (3) other and unspecified. In category 1 there 
 . were 104 reported underground fires, category 2 had 81, and there 
 were 44 in category 3. Category 1, where the fire was not im- 
 mediately detected, was then analyzed with respect to equipment, 
 ignition source, burning substance, and location. 
 
 Equipment: Excluding the category not specified, electrical 
 equipment incidents made up the largest group, with 23, followed 
 by mobile equipment with 12, and conveyors with 9 incidents. Of 
 the 23 electrical equipment incidents, 18 had electrical as the igni- 
 tion source. 
 
 Ignition source: Fire incidents caused by electrical ignition 
 sources accounted for 41 of the reported underground fires; spon- 
 taneous combustion followed with 17 and welding with 14. Not 
 surprising is that spontaneous combustion was the cause of 10 out 
 of 17 fires in the mined-out and/or waste location. 
 
 Burning substance: The burning substance involved in the 
 majority (55 pet) of the reported underground fires was timber. 
 There were 24 fire incidents that involved two burning substances, 
 mainly insulation and the category other. Three of the fire incidents 
 involved three burning substances: insulation, rubber, and other. 
 When comparing burning substance with location, of the 20 shaft- 
 raise- winze location fires, 18 of them reportedly had timber as the 
 burning substance. 
 
 Location: The haulage-drift area had the highest number of fire 
 incidents with 34. The majority of the welding fires occurred in 
 the haulage-drift and shaft-raise-winze locations. Fourteen of the 
 eighteen fires in the substation-shop-storage-pump location were 
 caused by electrical equipment. 
 
 Eighty percent of the fires involved timber, insulation, or rub- 
 ber while twelve percent involved unknown materials. Only 8 pet 
 of the fires involved combustible liquids. 
 
 From this study it appears that most of the fires that were not 
 immediately detected in metal and nonmetal mines started small 
 and/or had been slow growing in nature with little potential for ven- 
 tilation disturbances in the early stages of the fire. 
 
35 
 
 DESCRIPTION OF FIRE LOCATION ALGORITHM 
 
 Based on the two hypotheses, an algorithm was developed to 
 utilize the real-time outputs from a system of strategically placed 
 underground detectors as inputs to a computer model for deteimining 
 the location of an underground mine fire. The algorithm involves 
 three steps, as follows: 
 
 1. Instrumentation.— Electronic instrumentation is installed at 
 strategic underground locations. Environmental parameters to be 
 monitored include POC's such as CO and C0 2 , air temperature 
 velocity and direction, and barometric pressure. The number and 
 placement of POC detectors is critical and is determined iteratively. 
 The detectors are linked through appropriate telemetry to a master 
 control. 
 
 2. Computer Simulation.— In the event of fire underground, 
 the current state of the mine ventilation is simulated using mine 
 ventilation network analysis software (5). The ventilation monitoring 
 instruments (temperature, velocity, direction, and pressure) pro- 
 vide real-time input data to the network analysis program, thereby 
 insuring a high degree of accuracy. Potential fire locations are deter- 
 mined through an analytic procedure that identifies all upstream 
 network branches common to the detectors in alarm. As POC waves 
 reach additional detectors, potential fire locations are redefined 
 through a "tree walk" procedure (i.e., impossible locations are 
 eliminated). Ventilation travel times are then calculated from every 
 network branch that is a potential fire location to every network 
 branch in which a fire detector is located. The travel times are deter- 
 mined using average air velocities and branch lengths. Next, 
 predicted combustion product arrival times (CPAT's) are calculated 
 from every potential fire location to each detector. CPAT is the 
 elapsed time beginning with the arrival of POC's at the first alarm- 
 ing detector (defined as arrival time=0 for that detector) until subse- 
 quent detectors begin to alarm. For example, if POC's reach detector 
 A first, detector B 3.5 min later and detector C 2.6 min after that, 
 the respective CPAT's would be for A, 3.5 for B, and 6.1 for 
 C. The CPAT's are then stored in a data array. 
 
 3 . Fire Location Determination. — The computer engages in an 
 array scanning routine wherein the real-time pattern of incoming 
 detector alarms is compared with the CPAT's stored in the data 
 array. The data array fire location corresponding to the combus- 
 tion product arrival time pattern that best matches the pattern of 
 real-time incoming detector alarms is the most probable location 
 of the actual fire. 
 
 The iterative process for determining proper detector number 
 and placement involves three steps. First, initial detector locations 
 are chosen. These initial locations can be based on somewhat 
 arbitrary criteria, such as the availability of power or site accessi- 
 bility. Next, the data array is created through computer modeling. 
 Finally, this data array is scanned to identify entries having similar 
 arrival time patterns. Adjustments to the number and placement 
 of detectors are made and the last two steps repeated until no similar 
 arrival time patterns are produced. Through this process, potential 
 fire locations can be associated with a recognizable pattern of 
 CPAT's at the various detector locations. 
 
 At this juncture, it is important to note a characteristic of the 
 fire location algorithm which, under certain circumstances, may 
 make it impossible to distinguish between two or more network 
 branches that have equal possibility of being the fire's true loca- 
 tion. A necessary condition for the proper function of the algorithm 
 is that the branch in which the fire is located terminate at a junc- 
 tion that splits the air into two independent paths leading to respec- 
 tive detectors. At junctions where separate splits of air merge, each 
 split loses its individual identity (perfect mixing at the junction is 
 assumed). A downstream detector can not determine from which 
 split the combustion products originated. 
 
 As a result of this characteristic, the algorithm is unable to 
 resolve fire location in two situations. The first situation is where 
 two or more branches enter a junction and two or more branches 
 exit the same junction. If detectors are appropriately located in 
 independent paths downstream of the junction, the algorithm would 
 indicate equal probability of the fire occurring in each branch enter- 
 ing the junction. This outcome still represents a significant improve- 
 ment over fire bossing, however, as the search could be limited 
 to those branches entering the junction. 
 
 The second situation occurs where two or more branches com- 
 bine at a junction to produce a single branch. In this situation, the 
 condition requiring two independent paths to respective detectors 
 cannot be satisfied, hence the fire's location remains uncertain. 
 However this outcome would be limited to a relatively small number 
 of branches on the return side of the ventilation system and would 
 therefore not significantly diminish the overall value of the 
 algorithm. 
 
 FIRE LOCATION CASE STUDY EXPERIMENT 
 
 Using the fire location algorithm, a hypothetical case study 
 experiment was performed. The objective of the experiment was 
 to determine whether detector locations could be selected that would 
 produce a recognizable pattern of CPAT for every potential fire 
 location and to determine if a slow growing fire would adversely 
 affect the performance of the differential arrival time algorithm. 
 In order to test the algorithm under as realistic circumstances as 
 possible, the subject of the experiment was a portion of the ven- 
 tilation network from the Homestake gold mine in Lead, SD. 
 
 The Homestake Mine is a complex multilevel mine having in 
 excess of 2,000 airways. By combining parallel airways, a simplified 
 ventilation network consisting of 404 branches and 237 junctions 
 was created. The case study experiment was performed on a sec- 
 tion of the mine comprising 83 branches and 48 junctions. A 
 
 simplified illustration of the overall ventilation network and the case 
 study test section are shown in figure 1. 
 
 The experiment began by applying conventional nework 
 analysis techniques to determine airflows, velocities, and resulting 
 branch times (time required for air to travel the entire length of 
 a branch). Next, locations for fire detectors were specified. The 
 hypothetical detection system initially comprised seven detectors, 
 resulting in a coverage ratio of 1:11.9. The test section, airflow 
 directions, branch times, and initial detector locations are shown 
 in figure 2. CPAT's were then computed from every branch in the 
 test section (i.e., potential fire location) to each detector location. 
 
 This procedure produced three distinct zones within the test 
 section (fig. 3). In zone I, fire location was resolved to a single 
 branch or pair or adjoining branches, indicating satisfactory detector 
 
36 
 
 placement. In zone II, three groups of 8 to 10 branches each had 
 the same arrival time patterns, indicating the need to either relocate 
 detectors, add detectors, or both. In zone III, the independent paths 
 to respective detector locations, which are a necessary condition 
 for proper operation of the fire location algorithm, led outside the 
 test section and hence, fire location could not be resolved to a single 
 branch or pair of adjacent branches. However, appropriately located 
 detectors outside the test section would provide satisfactory resolu- 
 tion of fire location in these branches. 
 
 Zone II was subject to further analysis to improve resolution 
 of fire location in that zone. Examination of arrival times indicated 
 that two of the detector locations, B and F, did not uniquely identify 
 any fire location and could thus be eliminated from the detection 
 system with no adverse effect. Detectors B and F were relocated 
 to positions H and I. An eighth detector, J, was also added. The 
 detection system now comprised eight detectors, resulting in a 
 coverage ratio of 1:10.4. 
 
 Using the new detector locations, CPAT's were again com- 
 puted. Satisfactory resolution was achieved for 22 of the 27 branches 
 composing zone II. Fire location uncertainty remained for the five 
 branches terminating at junctions 69, 70, and 71. However the 
 condition requiring at least two independent paths to respective 
 detector locations could not be satisfied for these branches, hence 
 
 the detector location study was concluded. If resolution of fire loca- 
 tion in these branches was desired, additional detectors could be 
 installed as needed. Initial and final detector locations, as well as 
 the remaining area of fire location uncertainty are shown in figure 4. 
 
 A fire was then added at the start of the branch between junc- 
 tions 46 and 65 to evaluate the effect of a slow growing fire on 
 the differential arrival time for that airway. To further complicate 
 the problem, the environmental conditions in the mine were changed 
 from completely saturated air at 60 ° F to 9 pet relative humidity 
 air at 75° F. 
 
 The arrival times were recalculated with a ventilation simulator 
 MFIRE (6) capable of dynamic simulation of transient state ven- 
 tilation under the influence of mine fires. Figure 5 shows the new 
 POC travel times in the case study area under the changed condi- 
 tions before initiation of the fire. POC travel time to detector H 
 was 10.65 min and travel time to detector I was 13.88 min under 
 these initial conditions. 
 
 After the initiation of a 1 ,000 Btu/min fire (the size of a trash 
 can fire), with dynamic updates to the state of the ventilation system 
 every minute, the POC spread was again tracked to detectors H 
 and I. New arrival times were 10.70 min to detector H and 13.93 
 min to detector I, a very close agreement. 
 
 Case study 
 test section 
 
 Figure 1 .—Mine ventilation network with case study test section indicated. 
 
37 
 
 Junction 
 
 00.0 Branch time, min 
 — ^ Airflow direction 
 Network branch 
 
 ^ Branch exiting test section 
 Initial detector location 
 
 Figure 2.— Case study test section showing airflow directions, branch times, and initial detector locations. 
 
38 
 
 1.3 
 
 1.8 
 
 1.0 
 
 1.3 
 
 2.0 
 
 2.6 
 
 3.7 
 
 2.2 
 
 1.8 
 
 4.8 
 
 1.6 
 
 ^^ 
 
 .T^ 
 
 0.4 
 
 (109)- 
 O.2] 
 (l08> 
 
 0.2 
 (l07V 
 
 0.3 
 
 0.4 
 — (l05> 
 
 0.6 
 — (104> 
 
 0.8 
 fl03> 
 
 1.5 
 
 102} 
 
 KEY 
 Junction 
 
 00.0 Branch time, min 
 wm^m Branch comprising zone 
 
 .•.•.•.•.■.v.-. Branch comprising zone 
 Branch comprising zone 
 
 Figure 3. — Case study test section showing zones of varying fire location capability with detectors in initial locations. 
 
39 
 
 00.0 Branch time, mln 
 Ml New detector locations 
 
 Initial detector locations 
 
 Remaining area of fire 
 location uncertainty 
 
 Network branch 
 
 Figure 4.— Case study test section with initial and final detector locations and remaining area of fire location uncertainty. 
 
40 
 
 A 
 
 /\ 
 
 .20 
 
 .22 
 
 .22 
 
 .24 
 
 .29 
 
 .29 
 
 .33 
 
 .36 
 
 .85 
 
 14.30 
 
 6.30 
 
 11.52 
 
 11.54 
 
 9.45 
 
 150.00 
 
 8.52 
 
 9.27 
 
 7.73 
 
 56 
 
 i.16 
 
 —(57 
 
 4.15 
 
 5fT 
 
 4.49 
 
 59 
 
 [41.67 
 [60 
 
 H LJ 1. 
 61 
 
 1.32 
 
 62 
 
 0.58 
 
 63 
 
 0.55 
 
 64 
 
 0.47 
 
 65 
 
 9.84 
 
 6.83 
 
 7.82 
 
 9.11 
 
 4.79 
 
 2.78 
 
 3.85 
 
 45.61 
 
 0.90 
 
 1.93 
 
 2.48 
 
 3.46 
 
 2.16 
 
 1.77 
 
 4.63 
 
 1.50 
 
 /\ 
 
 0.42 
 
 (109 
 
 0.18 
 {108) — 
 
 E h 0.21 
 -0) 
 
 [0.30 
 {106V 
 
 t X" 42 
 
 — (fosV- 
 
 [0.57 
 -(104) 
 
 J 0.80 
 
 {uSV - 
 
 1.28 
 {102) := 
 
 1.90 
 
 KEY 
 
 00.0 Travel times, min 
 
 I— I Detector location 
 
 00 ' Junction 
 
 ► Airflow direction 
 
 ^. Branch exiting test section 
 
 _ Network branch 
 
 ■ Product of combustion path 
 
 (1 Fire 
 
 Figure 5.— Case study test section showing new conditions and POC paths to detectors. 
 
41 
 
 DISCUSSION OF EXPERIMENT RESULTS 
 
 The case study experiment demonstrates the operation of the 
 fire location algorithm. Several observations relating to the results 
 of this case study experiment are discussed in the following 
 paragraphs. 
 
 This test illustrates the need to maintain a high degree of control 
 over mine ventilation parameters for the fire location system to 
 perform properly. In some instances, the basis for specifying one 
 network branch over another as the probable fire location was a 
 difference in CPAT's of less than 0.5 min at a single detector. If 
 knowledge of ventilation flow parameters is only approximate, the 
 resulting inaccuracies would preclude a proper determination of fire 
 location. 
 
 This test illustrates the importance of proper detector place- 
 ment. With detectors in the initial seven locations, a large blind 
 area was produced where the location of the fire could not be 
 specified. Only after detectors B and F were moved (to locations 
 H and I) and detector J added, was the system capable of performing 
 properly. In larger and more complex networks, several detector 
 relocation iterations could be necessary. It should also be noted that 
 if mine conditions (no power, inaccessibility, etc.) prevented place- 
 ment of detectors in the required locations, a system performance 
 degradation could be expected. 
 
 This test illustrates the need for a high level of detection system 
 reliability for the fire location function to perform properly. The 
 loss of one or two detectors (power outage, telemetry failure, zero 
 or span drift, fuel cell failure, etc.) would seriously compromise 
 the ability of the system to determine the correct fire location. In 
 addition to systematic detector and telemetry inspection and 
 
 maintenance, consideration should be given to installing multiple 
 detectors at each detector outstation to provide a backup detection 
 capability in case of detector failure. 
 
 The algorithm can perform well for small fires in their initial 
 stage. 
 
 The need for mine ventilation parameter monitoring is essential 
 to obtain an accurate version of the current state of the ventilation 
 system. This is illustrated by casual inspection of the variations of 
 travel times between figure 4 and figure 5. 
 
 Certain enhancements to the system are possible that could 
 potentially improve its speed and accuracy. One obvious enhance- 
 ment is the inclusion of POC concentrations in the fire location 
 scheme. If the pattern of POC concentration was tracked, the ar- 
 rival of second and third waves of POC's (air that had originated 
 at the fire site and traveled to the detector over well-defined but 
 slower paths than that accounting for the first arrival) could be noted. 
 The pattern of second and third arrivals, observed at a single detector 
 or at multiple detector locations, could help define a unique fire 
 location where the same detectors, using first arrival times alone, 
 would leave blind areas. 
 
 The authors also wish to emphasize that a single experiment 
 consisting of simulations performed on data from a portion of a 
 single mine hardly constitutes a definitive treatment of this sub- 
 ject. Additional simulations to exercise the algorithm over a broad 
 range of mine types and ventilation conditions are planned. Because 
 the fire location system incorporates certain assumptions regarding 
 both fire and ventilation system behavior, full-scale in-mine valida- 
 tion tests are also planned. 
 
 CONCLUSIONS 
 
 A technique for determining the location of an underground 
 mine fire using a system of underground fire detectors and a spe- 
 cially designed ventilation network analysis computer model has 
 been presented. The technique is simple and straightforward and 
 can be implemented using off-the-shelf, mineworthy detection, 
 telemetry, and computer hardware. The technique has certain limita- 
 tions relating to ventilation network configuration; however, the 
 limitations affect only small parts of a network, and thus do not 
 detract significantly from its many potential benefits. 
 
 Although the findings of the case study were favorable, the 
 reader should be cautioned that this approach to mine ventilation 
 
 and fire safety analysis is meant to supplement, and not supplant, 
 the traditional decisionmaking processes at a mine. This system 
 should be regarded as a source of heretofore unavailable informa- 
 tion which, when integrated with other relevant data, can form the 
 basis for important ventilation decisions. It should not be used as 
 the sole data source for an automated closed-loop feedback system 
 of ventilation controls. The complexity of modern mine ventilation 
 systems and the compounding effect of the sometimes unpredictable 
 human element dictate that such controls be exercised only by teams 
 of trained and experienced experts. The value of the fire location 
 system is in the information it supplies to human decisionmakers. 
 
 
 REFERENCES 
 
 1. Stout-Wiegand, N. National Traumatic Occupational Fatalities, 
 1980-1984. NIOSH, Div. Saf. Res., Morgantown, WV, June 11, 1987, 
 12 pp. 
 
 2. Cote, A.E. (ed.). Fire Protection Handbook. National Fire Protec- 
 tion Association, Quincy, MA, 16th ed., 1986, p. 1-9.3. 
 
 3. Butani, S.J., and W.H. Pomroy. A Statistical Analysis of Metal and 
 Nonmetal Mine Fire Incidents in the United States From 1950 to 1984. 
 BuMines IC 9132, 1987, 41 pp. 
 
 4. Marks, J.R. Carbon Monoxide Monitoring at the Homestake Gold 
 Mine — Lead, South Dakota. Pres. at Am. Min. Congr. Min. Convention, 
 
 San Francisco, CA, Sept. 13-16, 1987, 12 pp.; available upon request from 
 L.W. Laage, BuMines, Minneapolis, MN. 
 
 5. Edwards, J.C., and R.E. Greuer. Real-Time Calculation of Product- 
 of-Combustion Spread in a Multilevel Mine. BuMines IC 8901, 1982, 
 117 pp. 
 
 6. Chang, X. The Transient-State Simulation of Mine Ventilation Systems. 
 Ph. D. Thesis, MI Technol. Univ., Houghton MI, 1987, 162 pp. 
 
42 
 
 RELIABILITY OF UNDERGROUND MINE FIRE DETECTION 
 AND SUPPRESSION SYSTEMS 
 
 By Steven G. Grannes 1 
 
 ABSTRACT 
 
 The Bureau of Mines has investigated the reliability of mine fire detection and suppression 
 systems and the effectiveness of inspection and maintenance practices. Interviews with Mine Safety 
 and Health Administration (MSHA) inspectors and field data indicate that reliability of mine fire 
 suppression systems could be improved. Limitations of current inspection and maintenance prac- 
 tices are discussed. Predictive diagnostics methods were developed and were tested in the field. 
 The predictive diagnostics method employs functional parameter measurement to predict wear- 
 out-related failures. Using these techniques, an intermittent system electrical relay failure was 
 diagnosed and corrected. An impending actuator failure was also noted. Limited field data indicate 
 a point estimate reliability of 75 pet for water deluge type systems. Reliability can be improved 
 by careful adherence to standard maintenance and testing procedures, and by applying preventive 
 maintenance techniques. 
 
 INTRODUCTION 
 
 Parts 57 and 70 of title 30 of the Code of Federal Regulations 
 (CFR) requires the installation and maintenance of mine fire sup- 
 pression systems in various locations in metal and nonmetal, and 
 coal mines. The CFR testing and maintenance requirements 
 reference various National fire codes, which in turn, reference 
 manufacturers recommendations. Manufacturer testing and 
 maintenance recommendations vary from product to product, and 
 may not allow for system degradation due to the severe mine 
 environment. Recent MSHA concern over the apparently low 
 reliability of mine fire suppression systems, and maintenance in- 
 consistencies, has led to the Bureau of Mines initiating a study of 
 maintenance and testing for mine fire suppression systems. This 
 paper presents a general summary of this research. 
 
 The objective of this research program was to improve the 
 reliability of underground fire suppression systems by developing 
 systematic inspection and testing procedures and by developing 
 predictive diagnostic techniques. Predictive diagnostics would allow 
 correction of impending reliability failures before they occur. The 
 combination of systematic testing procedures with predictive 
 diagnostics should virtually eliminate system unreliability. The 
 research program consisted of three parts: (1) system and failure 
 mode identification, (2) system reliability testing design, and (3) 
 in-mine evaluation of test concepts. 
 
 REVIEW OF FIRE SUPPRESSION SYSTEM FUNCTION 
 
 There are two basic types of fire suppression systems, the 
 automatic sprinkler type and the fire sensor actuated types. The 
 automatic sprinkler system type uses heat activated sprinkler heads, 
 each opening individually in response to fire. These systems are 
 designed to contain fires, and to minimize damage, but may not 
 toally extinguish fires. Automatic sprinklers have the advantage of 
 
 'General engineer. Twin Cities Research Center, Bureau of Mines, Minneapolis, 
 MN. 
 
 efficiently putting water where heat occurs, but may lag behind fires 
 where rapid flame spread occurs. 
 
 Fire sensor actuated systems use fire sensors that activate a 
 separate fire suppression system. Typical suppression systems 
 include water deluge systems (directed open water nozzles), high 
 expansion foam generators, and dry chemical systems. Fire sensing 
 devices include rate of rise heat detectors, bimetallic heat switches, 
 and optical flame detectors. The suppression agent systems are ac- 
 tivated by an actuation valve controlled by electrical or mechanical 
 
43 
 
 circuits. These systems are designed to protect a defined area, and 
 to prevent rapid fire spread by applying extinguishant ahead of the 
 fire. Water deluge systems are the most common suppression 
 systems of this type. Dry chemical systems are usually used in areas 
 where water freezing may be a problem, such as surface belt drive 
 locations. High expansion foam systems are usually used where 
 water availability is limited, such as parts of the western United 
 States, or as part of a multiple agent fire fighting plan. 
 
 Because of the added complexity of the separate fire sensing 
 and suppression systems, the reliability of sensor actuated type 
 systems is generally lower than automatic sprinklers. Because of 
 this complexity, the primary focus of this work was on these 
 
 systems, although the general concepts can be applied to all fire 
 suppression system types. 
 
 Figure 1 shows a typical water deluge system designed for pro- 
 tection of a conveyor drive. Notice the separate heat sensing and 
 water valve actuation circuits. The heat sensors normally open and 
 close at the actuation temperature of 180° F. The closing sensor 
 circuit activates a latching relay that (1) closes an audible and visible 
 alarm circuit, (2) activates the suppression agent discharge system 
 through an electronic servovalve (motor or solenoid), and (3) deac- 
 tivates the belt drive motor. The distribution lines have a manually 
 operated valve to bypass the sensor based actuation circuit. 
 
 Water deluge 
 control box 
 
 Electric 
 water 
 valve 
 
 Heat 
 sensors 
 
 Figure 1.— Water deluge system. 
 
44 
 
 SYSTEM AND FAILURE MODE IDENTIFICATION 
 
 In order to assess system types and potential failure modes, 
 MSHA electrical inspectors were interviewed as part of a 1986 elec- 
 trical retraining session. The purpose of this interview was to identify 
 sensor activated system types and common failure modes. The 
 system types identified in the course of this interview were primarily 
 water deluge (90 pet) with some dry chemical and high expansion 
 foam systems. This information was used to focus the research on 
 the most common system types. 
 
 Table 1 summarizes the results of the MSHA interviews, 
 describing common failure types for water deluge, dry chemical, 
 and high expansion foam system types. Common problems included 
 dead batteries, open contact points, and clogged nozzles. The general 
 consensus was that fire suppression system reliability could be im- 
 proved through better system maintenance. 
 
 Preliminary field system data were collected in four mines in 
 the Pikeville, KY, area in July 1986. This area was selected as a 
 representative eastern U.S. underground coal mining area. Data 
 collected included system types in place, system installation prac- 
 tices, temperature and humidity conditions, and potential fire sup- 
 pression failure modes. High humidity was observed at each mine. 
 System access problems were noted, in particular, heat sensors and 
 distribution nozzles were difficult to access because of proximity 
 to the moving belt and the presence of guards. Water deluge nozzle 
 caps were not observed on any of the systems. Electronic control 
 wires and electronic control box installation was often improper. 
 Lines were often strung either too tautly or unsecured, and control 
 boxes were found loosely hanging from control lines. These obser- 
 vations and subsequent discussions with safety personnel regarding 
 reliability problems were consistent with the previous MSHA in- 
 spectors survey. 
 
 Contacts with several system manufacturers were made to deter- 
 mine the current maintenance and inspection procedures used for 
 various system types. Schematics were also obtained. Functional 
 
 testing was the most commonly recommended system testing 
 method. The manufacturers suggested procedures were a useful 
 starting point, 2 but did not fully meet the test method design objec- 
 tives for inspection procedures, particularly the aspect of predict- 
 ing impending failures. 
 
 Two commercially available representative water deluge 
 systems were acquired to evaluate normal operating conditions and 
 functional parameters. This laboratory evaluation was the basis for 
 further test system development. The two systems had functional 
 similarities with each other and with other systems in the field. Close 
 examination of functional components indicated that the most prob- 
 able failure types would be either caused by electrical discontinuity, 
 mechanical seizure, or battery deterioration. Each of these failure 
 types would be accentuated by the wet and dust-contaminated con- 
 ditions in mines. 
 
 Table 1 .—Survey of common failure fire suppression failure types 
 
 System type 
 Water deluge systems. 
 
 Dry chemical systems . 
 
 High expansion foam 
 systems. 
 
 Failure modes 
 
 Relays inoperative. 
 
 Dead batteries. 
 
 Clogged water lines. 
 
 Spray nozzles clogged. 
 
 Moisture in control box. 
 
 Battery corrosion. 
 
 Sticky solenoid water valve. 
 
 Insufficient or no water supply. 
 
 Heat sensor failure. 
 
 Broken wiring. 
 
 Burned out belt shutdown switch. 
 
 Trigger device seizure. 
 
 Moisture and corrosion in control box. 
 
 Blocked or broken distribution lines. 
 
 Generally do not function correctly. 
 
 SYSTEM RELIABILITY TEST DESIGN 
 
 The survey and field observations were essential for the 
 development of a practical inspection procedure that would have 
 a high probability of identifying critical system failures and which 
 would predict impending failures. Testing that predicts impending 
 failures is important because preventative maintenance can then be 
 used to correct problems prior to system failure. It is also desirable 
 for the inspection procedures not interfere with conveyor belt pro- 
 duction except when verifying shutdown relay operation. The 
 approach should not be labor or cost intensive, or require destructive 
 testing (i.e., dry chemical system discharge), and be generic, for 
 applicability to various system types. 
 
 Reliability is the probability that an item will perform its 
 intended function for a specified period under stated use conditions. 
 There are three general types of product failures: Infant mortality 
 failures, random product life failures, and wear-out failures. These 
 failure types are common for populations of product and are shown 
 in the general bathtub curve in figure 2. The dependent variable 
 is the average probability of failure for product populations as a 
 function of time. This conventional perspective is useful for product 
 populations but not for discrete product units. 
 
 A given product unit under specified conditions will have only 
 one discrete failure time. The failure will either be due to normal 
 wear out or due to a discrete random occurrence. Normal wear out 
 is caused such gradual factors as mechanical wear, electronic com- 
 ponent corrosion, or general mechanical or thermal fatigue. Ran- 
 dom failures occur all at once and are typified by mechanical 
 breakage, electrical burnout, or some unforeseen environmental 
 event. Wear-out-related failures can be predicted by quantifying 
 the degree of wear. The stated objective of this work was to iden- 
 tify possible precursors to system failures. 
 
 There are three methods of determining product functionality: 
 Subjective observation, discrete testing, and continuous functional 
 variable measurement. These three concepts can be illustrated us- 
 ing the simple example of carbon-zinc batteries. Functionality can 
 be subjectively inferred from the appearance, i.e., presence of cor- 
 rosion on the case would indicate bad batteries. Functionality can 
 be tested by determining the ability of the batteries to light a bulb. 
 
 Manufacturer recommendations should be followed in any inspection and 
 maintenance program. The ultimate object of this work is to point out methods for 
 manufacturers to improve recommended maintenance and inspection techniques. 
 
45 
 
 -Useful life 
 
 TIME 
 
 
 / |Time 
 
 / lOf 
 
 Electric / [system 
 water valve / failure 
 voltage-x / 
 
 Warning \^s — level 
 
 
 I time 
 
 Figure 2.— Failure probability over product population life. 
 
 Figure 3.— Wear-out trending to predict failure time. 
 
 This type of testing is called go, no-go testing or discrete testing. 
 Most functional tests are discrete by nature (the system either works 
 or it does not). Functionality can be determined by continuous 
 variable measurement. This type of testing would involve the 
 measurement of battery power, and would indicate a quantitative 
 level of performance. 
 
 Continuous variable measurement is useful to determine safety 
 margins, as well as quantifying system wear-out trends. The analysis 
 of system wear is also called functional (reliability) trending. 
 Random failure events, as previously defined, cannot be predicted. 
 Frequent inspections are therefore necessary to minimize the effects 
 of these failures. 
 
 Many sensor actuated fire suppression system failure modes 
 are wear-out related, and therefore predictable. Battery internal 
 resistance will often gradually increase, relay and sensor point con- 
 tacts and wire connectors will gradually corrode, moving parts in 
 relays and water valves will gradually become more sticky. These 
 wear-out modes constitute a large percentage of fire suppression 
 system failures. 
 
 Figure 3 illustrates the concept of failure prediction. In the 
 figure, battery voltage and electric water valve actuation voltage 
 are trended over time. If the battery voltage drops below the solenoid 
 actuation voltage, the system will not work. Battery voltage 3 will 
 decrease over time because of wear out, while electric water valve 
 actuation voltage will increase because of corrosion and water 
 chemical deposition. By tracking these performance parameters it 
 should be possible to predict the time to failure. 
 
 Failure prediction is not possible for random failure events, 
 such as control wires being cut by roof falls or machinery, or by 
 rock dust application clogging deluge nozzles. Visual observation 
 and functional testing are the best ways to assess these conditions. 
 The inspection interval for random failures types should be at least 
 weekly. 4 Specific items to check include battery and auxiliary power, 
 
 electric continuity, and suppression agent distribution lines. 
 Manufacturer's inspection recommendations outline visual and func- 
 tional testing methods, such as the following. 
 
 Water Deluge Inspection 
 Weekly Test 
 
 1 . Test batteries by pushing battery test switch. Batteries are good 
 if light emitting diode battery indicator light comes on. 
 
 2. Test heat sensor and solenoid valve circuit continuity by 
 pushing circuit test switch. Circuits are good if sensor and solenoid 
 circuit light emitting diode indicator lights turn on. 
 
 3. Check circuit and water hose runs for looseness, breakage, 
 or possible abrasion. 
 
 4. Check to see that all nozzle blowoff dust covers are on. 
 
 5. Open strainer flush out valve to check for good water flow 
 and low sedimentation. 
 
 Table 2 illustrates the various methods of reliability assessment, 
 and lists advantages and disadvantages. MSHA inspection guidelines 
 
 Table 2.— Comparison of reliability assessment techniques 
 
 Assessment method Advantages 
 
 Disadvantages 
 
 'Actual measurements are of battery internal resistance, which correlates to decreased 
 battery energy potential. 
 
 'This is the current visual inspection interval for which written visual inspection 
 records are required. 
 
 Subjective observation 
 problems (i.e., visual 
 examination). 
 
 Discrete testing (i.e., 
 go, no-go testing or 
 functional test). 
 
 Continuous functional 
 parameter measure- 
 ment (i.e., measure- 
 ment of battery 
 power). 
 
 Fast decisionmaking. 
 Quickly find faults. 
 
 Systems can be tested 
 functionally. Accurate 
 assessment of current 
 condition. 
 
 Important components 
 can be individually 
 tested. Wear-out 
 trends can be 
 assessed. Statistically 
 efficient. 
 
 Not precise. Easy to 
 miss failures. 
 
 Cannot isolate in- 
 dividual bad parts. Im- 
 pending failures not 
 detectable. May be 
 destructive to system 
 tested. Chance that 
 intermittent failures 
 will be missed. 
 
 Need to separate com- 
 ponents. Time con- 
 suming. Need to 
 understand compo- 
 nent interrelationships. 
 
46 
 
 call for weekly visual (subjective) inspections, with annual func- 
 tional (discrete go, no-go testing) of water deluge systems. Func- 
 tional testing is useful, but will not guarantee future reliability. 
 Visual observations are effective for obvious defects such as broken 
 wires or cut distribution lines, but will not detect loose connectors 
 or poor electrical continuity. A good maintenance program should 
 include a combination of subjective observation, functional testing, 
 and quantitative function measurement. 
 
 In order to investigate the effectiveness of predictive diagnostics 
 techniques for mine fire suppression systems, several test devices 
 were designed and constructed. The purpose of these devices was 
 to quantitatively measure those functional parameters that may show 
 signs of wear out. The devices constructed included a portable 
 battery powered oven for heat detector actuation temperature 
 measurements; a variable power supply for measuring electric water 
 valve actuation energy; a variable voltage supply for measuring relay 
 actuation voltages; a constant current supply for determining con- 
 tact point line resistance and control line continuity; and a 1-s pulsed 
 load circuit for measuring battery internal resistance. Figure 4 shows 
 the portable sensor oven; figure 5 shows the combination power- 
 voltage-current control box and the battery tester. Readouts were 
 obtained with intrinsically safe multimeters. 
 
 The devices were all built using intrinsically safe design prin- 
 ciples, although this was not required for the locations covered. 
 These circuits were relatively simple in design and construction. 
 
 Figure 4.— Portable sensor oven. 
 
 m*#!&®mj#**?< 
 
 Figure 5.— Power-voltage-current control box and battery tester. 
 
47 
 
 IN-MINE EVALUATION OF TEST CONCEPTS 
 
 In order to test the effectiveness and practicality of the test 
 devices and techniques, data were collected from four mines in 
 Colorado and Utah. A test procedure had been outlined in the 
 laboratory, but it was necessary to obtain field reliability data, as 
 well as to develop a practical field approach to system testing and 
 maintenance. MSHA assisted in the selection of four representative 
 western U.S. mines, to complement the data from the eastern U.S. 
 mines. 
 
 Several fire suppression system installation types were observed 
 including automatic sprinklers, dry chemical systems, high expan- 
 sion foam systems, and water deluge systems. Equipment protected 
 included electrical transformers, motor control stations, rock dust 
 application machines, and belt drives. The condition of the fire 
 suppression systems observed were similar to those in the eastern 
 U.S. mines. 
 
 The procedure used to measure functional parameters was as 
 follows: 
 
 1. Disconnect and remove batteries. Measure battery internal 
 resistance. 
 
 2. Place heat sensor oven on accessible sensors, measure 
 actuation temperatures. 
 
 3. Energize relay with variable voltage device, measure actuation 
 voltage, and contact point resistance. 
 
 4. Measure water valve actuation power requirements using 
 variable power source. 
 
 5. Carefully reassemble system. Perform functional test to assure 
 correct assembly. 
 
 Additional observations made included the condition of the 
 sensor and control wires, the condition of the suppression distribu- 
 tion lines, the water flow rates, evidence of nozzle clogging, the 
 presence of nozzle covers, general control box condition, and obser- 
 vation of contamination or corrosion. 
 
 Four similar systems in two mines were tested using the 
 procedures outlined. The first test took 1 h; the final test took only 
 10 min because of improved organization. Data for the parameters 
 tested are presented in table 3. The systems generally showed little 
 variability even after over 1 yr of field installation. 
 
 Functional parameter measurement successfully diagnosed an 
 intermittent relay closure failure apparently caused by rock dust 
 contamination on the final system tested. The contact point resistance 
 varied between 0. 1 and 100 ohms. The readings were so high in 
 some cases that the test meter initially appeared to be broken. The 
 functional test confirmed this condition, in that the system functioned 
 on the first test, not on the second test, and about every other time 
 for subsequent tests. Given these results the system would have 
 passed the annual functional test, but would not have worked in 
 the event of a real fire. This intermittent condition was pointed out 
 to mine personnel for correction. 
 
 Analysis of test 3 revealed an alarming near-failure condition. 
 Test 3 was conducted on a dry chemical system. The battery voltage 
 was 12.3 V under a no-load condition, with a measured dual battery 
 internal resistance of 1.0 ohm. The battery voltage would be at 
 1 1 .6 V under the 0.71 -A load of the dry chemical actuation plunger. 
 Since the actuation plunger requires a minimum of 1 1 .4 V to fire, 
 the system appeared very close to failure (intersecting lines in 
 figure 3). 
 
 Table 3.— Summary of field test data 
 
 Parameter Test 1 Test 2 Test 3 1 Test 4 
 
 Battery internal resistance: 
 A: 
 
 Resistance ohms.. 1.4 1.6 2.1 1.8 
 
 Voltage V.. 13.3 12.3 12.3 12.3 
 
 B: 
 
 Resistance ohms.. 1.5 1.8 1.8 1.8 
 
 Voltage V.. 13.0 12.3 12.3 12.4 
 
 Heat sensor actuation temperature . °F . . 1 82 ( 258 263 273 
 
 [ 260 273 266 
 
 / 263 272 273 
 
 Relay closure voltage V. . NM NM 7.5 7.2 
 
 Relay contact point resistance . .ohms. . NM NM 0.024 ( 4.5 
 
 I 0.01 
 
 | 100 
 
 Water valve actuation power: ' 
 
 Voltage V. . 8.23 7.39 2 11.4 7.8 
 
 Strength A.. 0.61 0.53 *0.71 0.55 
 
 Time to complete min . . 60 60 30 10 
 
 NM Not measured. 1 Dry chemical. Actuation plunge. 
 
 RESULTS AND CONCLUSIONS 
 
 The results from the field work were mixed. It could be inferred 
 that one in four systems is not reliable, but given the limited sample 
 size (four) it is likely that the actual percentage of unreliable systems 
 is somewhat greater or less than 25 pet. The correct diagnosis of 
 the contact point failure and the near-failure condition was signifi- 
 cant, since these confirmed the effectiveness of functional parameter 
 measurement. 
 
 The test procedures developed effectively supplement the func- 
 tional testing method, but the procedures may be too complicated 
 and time consuming to be followed. Because functional parameter 
 measurement requires system disassembly, careful circuit 
 reassembly is critical; assembly errors could potentially introduce 
 more failures than averted through the techniques. It would be possi- 
 ble to incorporate the test circuit concepts in the design of the fire 
 suppression system control circuits, but this may increase system 
 costs. Some manufacturers include some effective test circuits at 
 this time; the battery test indicator as shown in figure 6 is one 
 example. A suggested incorporation would be to use a series resistor 
 
 Figure 6.— Typical commercial battery test circuit. 
 
48 
 
 in the battery activated water valve circuit during functional testing. 
 This resistor would provide a margin of safety in the event of a 
 real system actuation (without test resistor). 
 
 Predictive diagnostics will provide the most benefit when 
 careful manufacturing, installation, and maintenance procedures are 
 
 
 followed. Predictive diagnostics has the potential of being an 
 effective failure prevention technique. The responsibility for system 
 reliability is shared by manufacturers, end users, and regulatory 
 agencies. A better understanding of mine equipment reliability 
 failure modes will result in a safer mining industry. 
 
 
49 
 
 DIESEL EXHAUST CONDITIONING SYSTEMS FOR FIRE 
 AND EXPLOSION CONTROL IN GASSY MINES 
 
 By Kenneth L. Bickel 1 
 
 ABSTRACT 
 
 Diesel-powered mining equipment operating in underground gassy mines must be equipped 
 with control devices to lower surface and exhaust temperatures and prevent flames and sparks from 
 being emitted to the mine atmosphere. The primary control device used to meet these requirements 
 is the water scrubber. Water scrubbers have performed well over a number of years, but they do 
 have disadvantages, which include frequent maintenance, large size, and high water consumption. 
 
 The Bureau of Mines, as part of its program in diesel exhaust control technology, is conducting 
 research in cooperation with mining companies, equipment manufacturers, and equipment sup- 
 pliers on a promising alternative to water scrubbers; the dry exhaust conditioning system. The dry 
 exhaust conditioning system will cool the exhaust and suppress flames and sparks without direct 
 contact between the exhaust gas and water. 
 
 This paper describes the dry exhaust conditioning system, and discusses an ongoing program 
 to evaluate the system for use on a large engine in a gassy noncoal mine, and for a small engine 
 operating in a coal mine. 
 
 INTRODUCTION 
 
 Diesel-powered mining equipment offers a number of ad- 
 vantages over other types of materials handling equipment. The 
 mobility, versatility, fuel economy, and long service life of diesel - 
 powered equipment have allowed it to gain wide acceptance in 
 underground nongassy mines. 
 
 Methane and combustible dusts present in gassy mines pose 
 fire and explosion hazards that must be considered in the design 
 of diesel-powered equipment. Hot engine and exhaust system sur- 
 faces, as well as hot exhaust gases, must be cooled to prevent fires 
 and explosions. Provisions must be made to prevent the discharge 
 of flame or sparks to the mine atmosphere. 
 
 The use of diesel equipment in underground mines is governed 
 by regulations in the U.S. Code of Federal Regulations (CFR), title 
 30. Part 36 outlines the requirements diesel equipment must meet 
 to be approved and certified as permissible for use in gassy non- 
 coal mines. Part 36 considers toxic or objectionable gases, the 
 ignition of flammable gas mixtures by the engine or electrical equip- 
 ment, fire hazards presented by combustible materials coming in 
 contact with the equipment, and mechanical hazards (I). 2 In October 
 1987, new standards for classifying gassy noncoal mines were 
 enacted (2). A chapter specifically for diesel equipment in 
 underground coal mines has not yet been established, but equip- 
 ment to be used in gassy areas of underground coal mines is cur- 
 rently being tested by the Mine Safety and Health Administration 
 (MSHA) in accordance with part 36, except the maximum allowable 
 surface temperature is reduced from 400° to 302° F. 
 
 'Mining engineer, Twin Cities Research Center, Bureau of Mines, Minneapolis, 
 MN. 
 
 2 Italic numbers in parentheses refer to items in the list of references at the end of 
 this paper. 
 
 Part 36, subpart B, gives construction and design requirements 
 for diesel-powered equipment in gassy noncoal mines. Section 36.25 
 outlines requirements for the exhaust system. These include an 
 exhaust flame arrestor, surface temperature requirements, tight 
 joints, exhaust gas dilution, the ability of the exhaust system to with- 
 stand an explosion, and an exhaust cooling system. Cooling must 
 be obtained by passing the exhaust through a conditioner that con- 
 tains water or a dilute aqueous chemical solution, or a spray of water 
 or aqueous solution. These conditioners, referred to as water scrub- 
 bers, are used in those areas of gassy noncoal mines and coal mines 
 where permissible equipment is required. 
 
 All equipment presently certified under part 36 use diesel 
 exhaust gas water scrubbers. While water scrubbers have proven 
 to be effective in cooling exhaust and acting as flame traps, they 
 do have a number of problems that have been described in a Bureau 
 report (3). These problems include sludge and mineral deposit 
 buildup on internal baffles and passages, premature failure at mount- 
 ing points and welds, severe corrosion of mild steel welds and com- 
 ponents, and pitting corrosion of stainless steel components. 
 
 The use of water scrubbers has other disadvantages. Scrub- 
 bers consume large amounts of water, and scrubber solution must 
 be added frequently. Entrained water in the exhaust can condense 
 when discharged to the atmosphere, obstructing visibility in the 
 mine. Vehicle design and operator field of vision may be affected 
 by the large size of scrubbers, and the back pressure induced on 
 the engine by the scrubber may affect its performance. 
 
 The amount of water consumed in a water scrubber is directly 
 proportional to the horsepower rating of the engine. Water scrub- 
 bers are designed to operate for one shift before more scrubber solu- 
 tion must be added. Engines used in underground coal production 
 
50 
 
 equipment generally do not exceed 150 hp, but the frequent addi- 
 tion of scrubber solution, flushing of accumulated sludge, and repair 
 and replacement of corroded parts is costly. Several coal mine 
 operators have estimated scrubber maintenance costs to range from 
 $500 to $1,500 per month for each scrubber (4). 
 
 Engines in some gassy mines, such as domal salt mines, may 
 have ratings in the 600- to 750-hp range. Engines used in oil shale 
 mines are of comparable size. Oil shale mines are currently not 
 declared gassy, but they may be declared gassy sometime in the 
 future if methane is found as mining progresses deeper into the oil 
 shale formation. Water scrubbers required for the engines used in 
 these mines would be very large, and may obstruct operator field 
 of vision. The water consumption would also be very high. It has 
 been estimated that in an oil shale mine using 40 vehicles with 
 engines in the 750-hp range, the annual water consumption would 
 
 be in excess of 25 million gal, which is equivalent to 77 acre-ft 
 of water (5). 
 
 The need for an alternative to water scrubbers for large engines 
 was recognized, and the Bureau initiated a project to find an alter- 
 native method of cooling diesel exhaust (5). The most promising 
 alternative identified was the dry exhaust conditioning system. The 
 dry exhaust conditioning system was chosen because it does not 
 consume water, does not require extensive development, and prom- 
 ises to require less maintenance than a water scrubber. 
 
 The following discussion describes two different dry exhaust 
 conditioning systems for underground gassy mine applications. One 
 system was designed for use with large engines for gassy metal or 
 nonmetal mine application, such as oil shale or salt mines. The other 
 system is for small engines used in coal mines. 
 
 DRY EXHAUST CONDITIONING SYSTEM 
 
 Any explosion-proofing system must have the capability to con- 
 trol surface and exhaust temperatures, prevent sparks and flame 
 from being emitted to the mine atmosphere, and must maintain struc- 
 tural integrity in the event of an internal explosion. After identifying 
 and evaluating six different concepts for cooling exhaust gas, the 
 dry exhaust conditioning system was chosen as the most promising 
 concept for further development (5). 
 
 The dry exhaust conditioning system does not require direct 
 contact between the exhaust gas and water. Dry system technology 
 
 was originally developed for use in other industries where com- 
 bustible gases or dusts may be present, such as munitions factories 
 or offshore drilling platforms. These systems have recently been 
 adopted for use in European underground coal mines, but have not 
 been approved for use in U.S. mines. Two companies presently 
 manufacture dry exhaust conditioning systems that may be adapted 
 for use in the United States (6-7). 
 
 Figure 1 illustrates a dry exhaust conditioning system with a 
 certified diesel engine. The exhaust from the engine may pass 
 
 Certified 
 diesel 
 engine 
 
 Heat exchanger 
 
 Cool exhaust 
 
 Flame trap 
 
 Figure 1.— Schematic of dry exhaust conditioning system. 
 
51 
 
 through a water-cooled manifold and piping to a heat exchanger, 
 or the exhaust manifold and heat exchanger can be incorporated 
 into one component. One or more heat exchangers cool the exhaust 
 gas. The heat from the exhaust gas is dissipated using either the 
 engine radiator and coolant, or a completely separate cooling system 
 using a second radiator and additional coolant. The water jacketing 
 of exhaust components and waste heat from the heat exchanger add 
 a significant heat load. If the engine's cooling system is used, it 
 will need to be larger than standard by 70 pet or more. Flame 
 arrestors downstream of the heat exchangers prevent any flames 
 from being emitted to the mine atmosphere. 
 
 The dry exhaust conditioning system has the potential for 
 needing much less maintenance than water scrubbers. It consumes 
 no water, does not require water level floats and control valves, 
 and its components are smaller than a water scrubber, allowing more 
 flexibility in placement of the system on the vehicle. 
 
 DRY EXHAUST CONDITIONING SYSTEM FOR 
 LARGE ENGINES IN GASSY NONCOAL MINES 
 
 In 1986, the Bureau entered into a cooperative research program 
 with the Colorado Mining Association, the Caterpillar Equipment 
 Co. , Wagner Equipment Co. , Union Oil Co. , and MSHA to design, 
 fabricate, and laboratory and in-mine test a dry exhaust condition- 
 ing system for a 50-st-capacity haul truck used in an underground 
 oil shale mine. The 31 -month project is being performed under 
 contract to J.F.T. Agapito and Associates, and is scheduled to be 
 completed in 1988 (8). 
 
 The system was designed for a Caterpillar 3412 turbocharged 
 engine rated at 650 hp installed in a Caterpillar 773B haul truck. 
 It was designed to meet the 400° F surface temperature require- 
 ment, and to cool the exhaust gas to the same temperature. 
 
 After passing through the engine's exhaust manifold and turbo- 
 charger, which are water jacketed, the exhaust enters another 
 manifold where it is directed to three shell-and-tube heat exchangers. 
 Three heat exchangers are required to cool the high exhaust flow 
 at rated speed and load conditions. The exhaust gas passes through 
 the tubes of the heat exchangers, with the cooling liquid circulating 
 through the shell. The coolant for the entire dry system is circulated 
 through a radiator, and is kept completely separate from the engine 
 cooling circuit. The cooled exhaust leaves the heat exchangers and 
 passes through a device that functions as both a flame and spark 
 arrestor. Finally, the exhaust is dumped in front of the dry system 
 radiator, where it is blown upwards and away from the truck. 
 
 The system has completed a laboratory evaluation at MSHA's 
 Approval and Certification Center, and has been installed on a haul 
 truck at Union Oil's Long Ridge oil shale mine. The laboratory 
 evaluation included a series of explosion tests, a surface temperature 
 test, and an endurance test. The system performed well during the 
 laboratory test, although a soot buildup problem in the heat 
 exchanger tubes resulted in the exhaust gas exceeding 400° F, 
 indicating that more heat rejection capacity may be required. 
 
 Figure 2 shows the Caterpillar 3412 engine and dry system on 
 the laboratory test stand. Currendy, the system is undergoing testing 
 in the mine. After completion of the test, the system will be 
 removed, and a complete evaluation of each component will be made 
 to determine any damage that may have occurred during the in- 
 mine evaluation. 
 
 Figure 2.— Large dry exhaust conditioning system and Caterpillar 3412 engine undergoing laboratory testing at MSHA's Approval 
 and Certification Center. 
 
52 
 
 DRY EXHAUST CONDITIONING SYSTEM FOR 
 SMALL ENGINES IN COAL MINES 
 
 In 1987, a cooperative project to test a dry exhaust condition- 
 ing system for use on small engines (up to 150 hp) in coal mines 
 was initiated. The project was established with the assistance and 
 cooperation of manufacturers of diesel engines, mining machines, 
 exhaust control devices, and coal mine operators. The Colorado 
 Mining Association represents these participants, and coordinates 
 activities among them, the Bureau, and MSHA. The objectives of 
 the program are to laboratory and in-mine test a dry system with 
 and without a diesel particle filter (DPF) installed downstream of 
 the dry system. The DPF is a device designed to remove approx- 
 imately 90 pet of the diesel particulate matter (DPM) from the 
 exhaust before its discharged into the mine air (9). 
 
 The dry exhaust system was designed for a Caterpillar 3306 
 engine installed in a Jeffrey 4114 ramcar. The engine's exhaust 
 manifold is replaced by a fin-and-tube heat exchanger, where 
 exhaust gas cooling takes place. The engine coolant passes through 
 the water-jacketed components and the heat exchanger. Cooling the 
 exhaust and engine and exhaust system surfaces to below 302 ° F 
 requires an additional 70 pet in heat rejection capacity over the 
 standard engine cooling system. A spaced-plate flame arrestor is 
 located downstream of the heat exchanger. 
 
 Diesel engines emit large amounts of DPM. Tailpipe emissions 
 may exceed 400 mg/m 3 . DPM has been shown to make up 40 to 
 80 pet of the respirable dust found in coal mines using diesels, and 
 therefore is indirectly regulated under the 2-mg/m 3 respirable dust 
 coal mine standard. There is also a potential health risk associated 
 with exposure to DPM (4). For these reasons, the dry system is 
 being tested with a DPF to determine the feasibility of having an 
 exhaust system that not only controls fires and explosions, but also 
 removes DPM. 
 
 The DPF consists of a cellular ceramic substrate in a stainless 
 steel housing. It is sized so that it will operate one to two shifts 
 before it requires regeneration. Regeneration is the cleaning proc- 
 ess where the DPM collected in the substrate ignites and burns, 
 leaving very little ash. Regeneration can occur on a vehicle, if the 
 vehicle has a heavy duty cycle where its exhaust temperature exceeds 
 950° F for sustained periods of time. However, for this applica- 
 tion where the exhaust is cooled, regeneration must be done off 
 of the vehicle; that is, the DPF must be removed from the vehicle 
 for regeneration. 
 
 The research program is divided into two parts consisting of 
 a laboratory test and in-mine evaluation. The laboratory test is being 
 conducted by both the Bureau and MSHA. The Bureau is currendy 
 testing the system, with and without the DPF. Figure 3 shows the 
 dry system and DPF installed in the Bureau's engine testing facility. 
 
 
 I ' ~S , 
 
 fwfj 
 
 Figure 3.— Small dry exhaust conditioning system and Caterpillar 3304 engine undergoing Bureau laboratory testing. 
 
53 
 
 During the Bureau's evaluation, surface and exhaust temperatures, 
 and exhaust pressure buildup across components are being 
 measured. The major emphasis is on determining the efficiency of 
 the DPF in removing particulate, and on characterizing the effect 
 of the dry system and DPF on the size distribution and chemical 
 composition of the diesel particulate. MSHA's test program will 
 evaluate the dry system to determine if it will meet the requirements 
 of CFR part 36. It will consist of a series of tests that include sur- 
 
 face temperature, cooling efficiency, endurance, induced faults, 
 safety shutdown, and spark arrestor tests (4). 
 
 After satisfactory completion of the laboratory evaluation, the 
 dry system and DPF will be installed on a Jeffrey 41 14 ramcar in 
 a coal mine in the western United States, for a period of 6 months. 
 The system and DPF will be evaluated for its durability and per- 
 formance under operating mine conditions (4). 
 
 SUMMARY 
 
 Water scrubbers are presently being used to cool exhaust and 
 act as flame and spark arrestors on diesel-powered equipment used 
 in gassy mines. The problems associated with their use led the 
 Bureau to initiate a program to develop an alternative to water scrub- 
 bers that could be used in coal mines and gassy noncoal mines. 
 The dry exhaust conditioning system was selected as a promising 
 alternative because it offers the potential for explosion-proofing a 
 diesel' s exhaust system without consuming water, and with much 
 less maintenance than that required of a water scrubber. Two 
 
 cooperative research projects are under way to evaluate dry systems 
 for different applications. One project is evaluating the dry exhaust 
 system for use with a 650-hp engine on a haul truck operating in 
 an underground oil shale mine. The other is evaluating the system 
 for use on a 150-hp engine on a haulage vehicle operating in a coal 
 mine. It is being evaluated with a DPF to determine if an exhaust 
 control system can be developed that will control fires and explo- 
 sions, and remove diesel particulate matter before it is emitted into 
 the mine air. 
 
 REFERENCES 
 
 1. U.S. Code of Federal Regulations. Title 30— Mineral Resources; 
 Chapter 1— Mine Safety and Health Administration, Department of Labor; 
 Subchapter E— Mechanical Equipment for Mines; Part 36— Mobile Diesel- 
 Powered Transportation Equipment for Gassy Noncoal Mines and Tunnels; 
 July 1, 1984. 
 
 2. Federal Register. U.S. Mine Safety and Health Administration (Dep. 
 Labor). Safety Standards for Methane in Metal and Nonmetal Mines: Final 
 Rule. V. 52, No. 126, July 1, 1987, pp. 24942-24951. 
 
 3. Waytulonis, R.W., S.D. Smith, and L.C. Mejia. Failure Analysis of 
 Diesel Exhaust-Gas Water Scrubbers. BuMines RI 8682, 1982, 19 pp. 
 
 4. Waytulonis, R.W., and G.J. Dvorznak. New Control Technology for 
 Diesel Engines Used in Underground Coal Mines. Paper in Proceedings 
 of the 3d Mine Ventilation Symposium (Penn State Univ., University Park, 
 PA). Soc. Min. Eng. AIME, 1987, pp. 279-285. 
 
 5. Paas, N. Explosion-Proofing of Large Vehicles (contract JO 113070, 
 Foster-Miller, Inc.). BuMines OFR 205-84, 1984, 275 pp.; NTIS PB 
 85-145803. 
 
 6. Minecraft, Inc. Product brochure, 1986; available on request from 
 K.L. Bickel, BuMines, Minneapolis, MN. 
 
 7. Pyroban Corp. Product brochure, 1986; available on request from K.L. 
 Bickel, BuMines, Minneapolis, MN. 
 
 8. J.F.T. Agapito & Associates. Development of a Dry Exhaust Condi- 
 tioner for Large Diesel Engines. Ongoing BuMines contract HO267001; 
 for inf. contact K.L. Bickel, TPO, BuMines, Minneapolis, MN. 
 
 9. Baumgard, K.J., and K.L. Bickel. Development and Effectiveness 
 of Ceramic Diesel Particle Filters. Paper in Diesels in Underground Mines. 
 Proceedings: Bureau of Mines Technology Transfer Seminar, Louisville, 
 KY, April 21, 1987, and Denver, CO, April 23, 1987. BuMines IC 9141, 
 1987, pp. 94-102. 
 
54 
 
 SPONTANEOUS COMBUSTION SUSCEPTIBILITY 
 OF SULFIDE MINERALS 
 
 By G. W. Reimers 1 and W. H. Pomroy 2 
 
 ABSTRACT 
 
 Under certain conditions present in the underground mine environment, sulfide minerals can 
 self-heat as a result of exothermic oxidation reactions. This self-heating can lead to spontaneous 
 combustion of carbonaceous materials such as mine timbers or even the sulfide minerals themselves. 
 This Bureau of Mines paper describes thermal gravimetric analysis (TGA) tests conducted to measure 
 the reactivity of 10 sulfide minerals in terms of their ignition point. TGA results also indicated 
 evidence of exothermic reaction at temperatures below the sulfide ignition point. An isothermal 
 oxidation procedure was also used to obtain additional quantitative information on the reactive 
 behavior of six pyrite samples during low-temperature oxidation to ferrous sulfate. 
 
 Samples of pyrite and marcasite were found to be the most reactive sulfides and ignition point 
 values below 300° C were measured by TGA. The two other iron sulfides, pyrrhotite and 
 arsenopyrite, and the copper sulfides, chalcopyrite and chalcocite, were less reactive. Lead sulfide 
 (galena) and the zinc sulfides (sphalerite and wurtzite) failed to ignite below 500° C. Molybdenite 
 had an ignition point of about 375 ° C but failed to indicate any self-heating tendencies. Isothermal 
 oxidation tests conducted on pyrite samples confirmed that this sulfide was reactive in moist at- 
 mospheres. Conversion of up to 30 pet of the sulfide to sulfate was measured after 24 h at 100° C. 
 
 INTRODUCTION 
 
 When the proper conditions are present in the mine environ- 
 ment, sulfide minerals can be susceptible to self-heating behavior 
 that leads to spontaneous combustion and ultimately a mine fire. 
 The Bureau of Mines conducted laboratory experiments to measure 
 the reactivity of 10 sulfide minerals to identify the sulfides that would 
 tend to pose the greatest hazard. These tests are part of a broader 
 research effort to develop methods of controlling the oxidation of 
 sulfides in underground mines and thereby reduce the possibility 
 of spontaneous combustion mine fires. 
 
 Preceding the current research effort, Ninteman (7) 3 conducted 
 an extensive literature review of the nature and problems associated 
 with the oxidation and combustion of sulfide minerals in 
 underground metal mines. Sulfide oxidation is an exothermic proc- 
 ess that produces heat, which, if not dissipated, causes a temperature 
 rise within the mineral mass. If the temperature begins to rise, a 
 thermal runaway situation may develop that will cause the sulfide 
 and other combustibles to burn. An example of this thermal runaway 
 situation was found in the Sullivan Mine (2-3) where, following 
 the blasting of large blocks of reactive sulfide ore, self-heating was 
 observed. If the mining activity was not of a sufficient rate to 
 dissipate the heat, combustion of the ore resulted and fire-fighting 
 measures were required. 
 
 The prolonged incipient stage (weeks or months) and seem- 
 ingly spontaneous appearance of these fires make their occurrence 
 
 'Physical scientist. 
 
 'Group supervisor. 
 Twin Cities Research Center, Bureau of Mines. Minneapolis. MN. 
 
 'Italic numbers in parentheses refer to items in the list of references at the end of 
 this paper. 
 
 difficult to predict and detect. Furthermore, these fires often start 
 in abandoned, backfilled, and/or caved areas where abundant fuel 
 is present but where access for fire fighting is difficult or in some 
 cases impossible. Since detection and suppression are so difficult, 
 a high priority is placed on fire prevention. However, the reac- 
 tions and mechanisms giving rise to spontaneous combustion in 
 sulfide ores are not well understood. Thus, current efforts by mine 
 personnel to control spontaneous combustion problems, regardless 
 of how well intended, are not always based on sound engineering 
 and technical principles. 
 
 One facet of the Bureau research involves monitoring the 
 behavior of various sulfides exposed to oxidizing condition while 
 being heated to their ignition point. This will provide a background 
 of data to aid in the selection of fire prevention efforts having the 
 maximum effectiveness. In this paper, the susceptibility of the 
 various sulfide samples to spontaneous combustion was assessed 
 by determining their respective ignition points. Using this criteria, 
 samples with the lowest ignition points would be considered the 
 most reactive. 
 
 Data are excerpted from a Bureau study (4) on the analysis 
 of the oxidation of pyrrhotite, marcasite, arsenopyrite, chalcopyrite, 
 chalcocite, and galena. Additional data are included on the igni- 
 tion points of pyrite, sphalerite, and molybdenite as determined by 
 TGA. TGA analyses also suggest that sulfate formation could have 
 a role in sulfide self-heating and the results of a quantitative method 
 for following the oxidation of pyrite by measuring the conversion 
 of insoluble iron sulfide to soluble iron sulfate are summarized. 
 In this procedure, reactivity was gauged by the extent of oxidation 
 during the defined test interval. 
 
55 
 
 EQUIPMENT AND TEST PROCEDURES 
 
 Ignition points were obtained using thermal analysis equipment 
 consisting of the following Perkins Elmer 4 components: a TGS-2 
 thermogravimetric analyzer in conjunction with a system 7/4 ther- 
 mal analysis controller. Data from the analyzer were continuously 
 fed to a thermal analysis data station (TADS), and the test results 
 were displayed on paper by a TADS-1 plotter. 
 
 In the TGA tests, a flow rate of 500 cmVmin was used for 
 purging the system with nitrogen and for supplying air during the 
 oxidation tests. To add moisture to the system, the oxidizing gas 
 was bubbled into water contained in a heated flask and routed to 
 the TGA furnace. The sample, approximately 40 mg in size, was 
 loaded into the balance pan, and the system was closed by raising 
 the furnace assembly into position. The assembly was then purged 
 with nitrogen while the sample was heated to 100 ° C. At this point, 
 the nitrogen was replaced with the oxidizing gas, and the sample 
 was heated at 25° C/min while sample weight was monitored. 
 
 Isothermal oxidation experiments were run with Lindburg 
 model 55035, horizontal-hinged, tube furnaces having a bore of 
 2.5 cm and a length of 33 cm, fitted with auxiliary temperature 
 controllers. A single sample of 5 g was loaded into a 97- by 16- 
 
 by 10-mm porcelain boat that was positioned at the center of the 
 furnace inside a 2.2-cm-ID Pyrex glass tube. Nitrogen was used 
 to purge the furnace tube while the sample was heated to the test 
 temperature. The test atmosphere was then introduced at a flow 
 rate of 500 cm 3 /min and timing of the test was initiated. For the 
 tests run with air containing 5 pet water vapor, air was bubbled 
 through water contained in a heated flask. When air containing 60 
 pet water vapor was used, water was metered into a furnace 
 assembly ahead of the test furnace that vaporized the water so that 
 it could be carried along with the airflow to the oxidizing furnace. 
 At the completion of each test, nitrogen flow resumed and the sample 
 was moved to the cool end of the furnace tube. When cool, the 
 sample was weighed and transferred to a bottle and sealed. 
 
 The extent of pyrite oxidation for each test was determined by 
 leaching with a known volume of water the soluble iron sulfate that 
 is formed at that temperature. Solutions were collected by filtra- 
 tion and then analyzed for their iron content. Knowing the iron con- 
 tent of the starting sample and the amount of soluble iron formed 
 during an oxidation test allowed calculation of the degree of ox- 
 idation of the sulfide to the sulfate. 
 
 SAMPLE PREPARATION 
 
 The sulfides studied in this research effort were obtained from 
 various sources including chemical suppliers, mineral specimen sup- 
 pliers, and from a commercial mill. When necessary the sulfides 
 were crushed in stages to minus 6 mm and the visible gangue was 
 sorted out. Further size reduction was done in a Bleuler mill by 
 milling 20-g batches for set time intervals ranging from 15 to 120 
 min. To decrease the possibility of oxidation during this size- 
 reduction step, the samples were milled and stored under heptane. 
 Just prior to testing, a portion of the sample slurry was dried in 
 
 air at room temperature, and then loaded into the test apparatus. 
 Size analysis measurements were conducted on several of the 
 sulfides, and the average particle size of these samples was in the 
 range of 5.8 to 8.7 /urn and 3.4 to 4.8 ^m for the 15- and 120-min 
 grinding times, respectively. Tables 1 and 2 list the partial chemical 
 analysis of the various samples. X-ray diffraction (XRD) analysis 
 was also conducted on the samples and the major constituents were 
 verified. 
 
 RESULTS OF THERMAL GRAVIMETRIC ANALYSES 
 
 PYRITE 
 
 Pyrite (FeS 2 ) samples collected from several sources were ex- 
 amined by TGA in atmospheres of dry air or air containing water 
 vapor. Figure 1 illustrates the results of tests conducted on pyrite 
 sample A that was milled for 15 min. The Y-axis of the figure in- 
 dicates the percentage of the original sample weight, measured over 
 the temperature range of the test. The ignition point of pyrite was 
 taken as the point on the curve where rapid weight loss occurred. 
 Examination of the sample after TGA confirmed the conversion 
 of the iron sulfide to ferric oxide (Fe 2 3 ) and was used as evidence 
 to confirm ignition. 
 
 Curve A indicates the results when the sample was oxidized 
 with dry air, and under this condition an ignition point of 335° C 
 was measured. In moist air (curve B) the ignition point increased 
 slightly to 345° C. Curve B also indicates that the sample began 
 to gain weight at about 290° C. This weight gain prior to ignition 
 was observed for all the pyrite samples when they were oxidized 
 in the presence of water vapor. Chemical and XRD analysis of pyrite 
 samples that were oxidized in moist air for several hours at 
 temperatures lower than the ignition point confirmed that the sulfide 
 was partially converted to sulfate and its formation was responsi- 
 ble for the weight gain. Increasing the milling time produces finer 
 sulfide particles and the increased surface area would be expected 
 to also increase the reactivity of the sulfide. 
 
 Table 1. — Partial chemical analysis of iron sulfide samples, 
 weight percent 
 
 Eleme 
 
 Jnt Pyrite 
 
 
 
 Pyrrhotite 
 
 Marca- 
 site 
 
 Arseno- 
 
 A B C D 
 
 E 
 
 F 
 
 A B 
 
 pyrite 
 
 Fe .. 
 
 . . 43.1 38.6 40.2 45.4 
 
 45.8 
 
 45.8 
 
 45.6 44.0 
 
 45.6 
 
 28.5 
 
 S . . . 
 
 . . 46.9 45.2 44.6 51.5 
 
 52.2 
 
 52.5 
 
 30.2 24.2 
 
 50.7 
 
 13.4 
 
 Si. . . 
 
 2.2 1.7 3.8 .21 
 
 1.2 
 
 .85 
 
 .77 8.5 
 
 ND 
 
 9.7 
 
 Ca . . 
 
 .61 1.5 .79 <.3 
 
 <.3 
 
 <.3 
 
 1.8 1.9 
 
 ND 
 
 <.3 
 
 Al. . . 
 
 .54 .26 .46 <.2 
 
 1.0 
 
 .22 
 
 <.2 2.4 
 
 ND 
 
 .31 
 
 Zn 
 
 .47 3.2 <.1 1.2 
 
 <.1 
 
 .14 
 
 ND ND 
 
 ND 
 
 ND 
 
 Pb .. 
 
 .3 <.1 <.1 1.1 
 
 <.1 
 
 .22 
 
 ND ND 
 
 ND 
 
 ND 
 
 Cu . . 
 
 .04 .1 .34 .02 
 
 .1 
 
 <04 
 
 .1 .14 
 
 ND 
 
 ND 
 
 Ni . . 
 
 ND ND ND ND 
 
 ND 
 
 ND 
 
 .0 3.8 
 
 ND 
 
 ND 
 
 As .. 
 
 ND ND ND ND 
 
 ND 
 
 ND 
 
 ND ND 
 
 ND 
 
 27.6 
 
 ND 
 
 Not determined. 
 
 
 
 
 
 
 Table 2.— Partial chemical analysis of copper, lead, zinc, and 
 molybdenum sulfides, weight percent 
 
 Elemt 
 
 mt Chalcopyrite 
 
 Chalcocite 
 A B 
 
 Galena 
 
 Sphal- 
 erite 
 
 Wurt- 
 zite 
 
 Molyb- 
 
 A 
 
 B 
 
 C 
 
 denite 
 
 Fe . . 
 
 . . 37.4 
 
 33.3 
 
 30.8 
 
 5.3 
 
 5.6 
 
 0.26 
 
 1.1 
 
 <0.02 
 
 0.19 
 
 Cu . . 
 
 .. 21.2 
 
 30.7 
 
 33.3 
 
 71.4 
 
 68.0 
 
 .24 
 
 ND 
 
 ND 
 
 .035 
 
 Pb . . 
 
 ND 
 
 ND 
 
 ND 
 
 ND 
 
 ND 
 
 82.2 
 
 ND 
 
 ND 
 
 ND 
 
 S . . . 
 
 . . 31.8 
 
 33.0 
 
 31.6 
 
 20.7 
 
 19.0 
 
 12.1 
 
 32.2 
 
 32.4 
 
 37.7 
 
 Al... 
 
 .68 
 
 <.20 
 
 .31 
 
 .37 
 
 <.2 
 
 ND 
 
 ND 
 
 ND 
 
 ND 
 
 Ca .. 
 
 <.5 
 
 <.5 
 
 <5 
 
 <.5 
 
 <.5 
 
 ND 
 
 ND 
 
 ND 
 
 ND 
 
 Si... 
 
 1.7 
 
 .48 
 
 1.1 
 
 .34 
 
 2.2 
 
 <.1 
 
 .51 
 
 ND 
 
 .42 
 
 Zn . . 
 
 ND 
 
 ND 
 
 ND 
 
 ND 
 
 ND 
 
 <.01 
 
 60.8 
 
 64.4 
 
 ND 
 
 Mo. . 
 
 ND 
 
 ND 
 
 ND 
 
 ND 
 
 ND 
 
 ND 
 
 ND 
 
 ND 
 
 54.0 
 
 ■■Reference to specific products does not imply endorsement by the Bureau of Mines. 
 
 ND Not determined. 
 
56 
 
 Figure 2 shows the results of TG A tests condueted on pyrite 
 sample E. milled for 15, 30, 60. and 120 min in the temperature 
 range of 100 to 400"' C in dry air. The ignition points for the 
 respective samples were 330°, 325°, 315°. and 270° C. These 
 results illustrate the effect increasing the milling time (decreased 
 particle size) had on lowering the ignition point. TGA tests of the 
 remaining pyrite samples produced curves similar to those in figure 
 1. however, there was a measurable difference in ignition points. 
 Tabic 3 lists the ignition point values measured for milled pyrite 
 samples in various oxidizing atmospheres. 
 
 PYRRHOTITE 
 
 Pyrrhotite (Fe„.,S„, with n ranging from about 5 to 16) was 
 the second iron sulfide examined by TGA. Results of tests con- 
 ducted on pyrrhotite A, milled for 15 min and oxidized in air and 
 air containing 40 pet water vapor, are shown in figure 1A and in- 
 dicate ignition point values of 415° C. For this pyrrhotite sample, 
 the presence of water vapor failed to have a significant effect on 
 the ignition point of the sample or the sample weight change at 
 temperatures below 300° C. Increasing the milling time to 120 min 
 
 Table 3.— Ignition point values of milled iron sulfides 
 determined in selected oxidizing atmospheres 
 
 „ , Milling 
 
 Sample ,. a . 
 
 r time, mm 
 
 Pyrite: 
 
 A 15 
 
 B 15 
 
 C 15 
 
 D 15 
 
 E 15 
 
 30 
 
 60 
 
 120 
 
 F 15 
 
 Pyrrhotite: 
 
 A 15 
 
 120 
 B 120 
 
 Marcasite 15 
 
 120 
 
 Arsenopyrite 15 
 
 120 
 
 ND Not determined. 
 
 
 Ignition 
 
 poi 
 
 nt, °C 
 
 
 Dry 
 
 
 Air 
 
 with— 
 
 
 air 
 
 5 pet water 
 
 40 pet water 
 
 335 
 
 345 
 
 
 
 ND 
 
 340 
 
 380 
 
 
 
 ND 
 
 375 
 
 380 
 
 
 
 ND 
 
 325 
 
 340 
 
 
 
 ND 
 
 330 
 
 350 
 
 
 
 ND 
 
 325 
 
 ND 
 
 
 
 ND 
 
 315 
 
 ND 
 
 
 
 ND 
 
 270 
 
 ND 
 
 
 
 310 
 
 370 
 
 345 
 
 
 
 ND 
 
 415 
 
 410 
 
 
 
 415 
 
 365 
 
 365 
 
 
 
 360 
 
 420 
 
 410 
 
 
 
 410 
 
 ND 
 
 ND 
 
 
 
 380 
 
 320 
 
 ND 
 
 
 
 255 
 
 ND 
 
 ND 
 
 
 
 445 
 
 390 
 
 340 
 
 
 
 330 
 
 h r- 
 
 ~i r 
 
 KEY 
 
 A Air 
 
 B Air containing 5 pet woter 
 
 a- I05 
 
 100 150 
 
 200 250 300 
 
 TEMPERATURE, °C 
 
 Figure 1.— Thermal gravimetric analysis of pyrite sample A 
 milled for 15 min. 
 
 I- 
 
 o 
 
 I 
 
 CC 
 
 rr 
 
 > 
 a. 
 
 IOO 
 
 95 
 
 no 
 
 IOO 
 
 £ 90 
 
 < 
 o 
 
 1 80 
 
 70 
 
 1 1 1 
 
 A Pyrrhotite A 
 
 1 1 
 
 1 1 
 A 
 
 - 
 
 1 1 1 
 
 B 
 
 \ 
 
 — I 1 — 
 
 B Marcasite 
 
 -i 1 1 r 
 
 Pyrite E 
 
 
 " 
 
 
 D 
 
 C 
 
 KEY 
 
 
 
 A 15 min 
 
 
 
 8 30 min 
 
 
 
 C 60 min 
 
 
 
 D 120 min 
 
 i 
 
 i i 
 
 250 
 
 TEMPERATURE, 
 
 Figure 2.— Thermal gravimetric analysis illustrating effect of 
 milling time on oxidation of pyrite sample E in air. 
 
 105 
 
 - IOO 
 
 95 
 
 90 
 
 85 
 
 80 
 
 — i 1 r 
 
 C Arsenopyrite 
 
 KEY 
 
 A Air 
 
 B Air containing 40 pet 
 water 
 
 IOO 150 200 250 300 350 400 450 500 
 TEMPERATURE, °C 
 
 Figure 3.— Thermal gravimetric analysis of (A) pyrrhotite sam- 
 ple A milled for 15 min, (B) marcasite milled for 120 min, and 
 (C) arsenopyrite milled for 120 min. 
 
57 
 
 lowered the ignition point by about 50° C (table 3). Again, the ad- 
 dition of water vapor had little effect on the ignition point, and no 
 effect was observed on oxidation in the lower temperature regions. 
 Results of TGA tests on pyrrhotite B milled for 120 min yielded 
 an ignition point in dry air of 420° C, indicating less reactivity than 
 the first sample. While water vapor had little effect on the ignition 
 point of this sample it did, however, promote the weight gain of 
 this sample at lower temperatures. XRD analysis of this sample 
 after isothermal oxidation in moist air confirmed that ferrous sulfate 
 was formed. Table 3 lists the ignition points obtained for the pyr- 
 rohotite samples under various test conditions. 
 
 MARCASITE 
 
 Marcasite has the same chemical formula as pyrite but has a 
 different crystal structure. A TGA test conducted on the marcasite 
 sample indicated that it was the most reactive in terms of the lowest 
 ignition point of the sulfides tested in this report. Figure 3fi illustrates 
 the TGA results for the sample milled for 120 min and oxidized 
 in dry air and air containing 40 pet water vapor. For this sulfide 
 the presence of water vapor had a strong influence on the ignition 
 point and was found to lower the ignition point to 255° C, which 
 was 65 ° C lower than the value obtained in dry air. XRD analysis 
 of the samples after testing indicated conversion of the sulfide to 
 ferric oxide. Marcasite ignition point values determined for various 
 conditions are listed in table 3. 
 
 ARSENOPYRITE 
 
 The final iron sulfide examined in this series was arsenopyrite. 
 Arsenic replaces a sulfur in pyrite, making this mineral a 
 sulfarsenide of iron (FeAsS). TGA results for a sample milled for 
 120 min are illustrated in figure 3C. For this mineral 40 pet water 
 vapor had the effect of lowering the ignition point from 390° to 
 330° C and also caused the sample to gain weight at lower 
 temperatures. Figure 3 C also illustrates an unusual double weight 
 loss when water vapor was present. It was assumed that iron sulfide 
 ignited first and was followed by the ignition of iron arsenide. XRD 
 analysis of a sample oxidized to the completion of the first step 
 indicated the presence of ferric oxide, but it failed to confirm if 
 the sulfide or arsenide component of the mineral had oxidized. Table 
 3 lists the ignition points measured for the arsenopyrite sample. 
 
 Table 4.— Ignition point values of milled copper and 
 
 molybdenum sulfides determined in 
 
 selected oxidizing atmospheres 
 
 Sample 
 
 Milling 
 time, min 
 
 Ignition point, °C 
 
 Dry 
 air 
 
 Air with— 
 
 5 pet water 40 pet water 
 
 Chalcopyrite: 
 
 A 15 
 
 30 
 
 60 
 
 120 
 
 B 15 
 
 120 
 
 C 15 
 
 120 
 
 Chalcocite: 
 
 A 15 
 
 120 
 B 15 
 
 Molybdenite 15 
 
 120 
 
 365 
 345 
 335 
 330 
 385 
 305 
 365 
 300 
 
 (360) 
 (335) 
 (370) 
 
 375 
 375 
 
 ND 
 ND 
 ND 
 ND 
 ND 
 ND 
 ND 
 ND 
 
 ND 
 ND 
 ND 
 
 ND 
 375 
 
 360 
 ND 
 ND 
 335 
 385 
 300 
 355 
 320 
 
 (360) 
 (365) 
 (420) 
 
 ND 
 370 
 
 ND Not determined. 
 
 NOTE. — Values in parentheses represent conversion to sulfate. 
 
 
 
 
 D 
 
 C 
 
 
 IUU 
 
 KEY 
 
 
 
 
 
 A I5n 
 
 in 
 
 
 
 
 9b 
 
 B 30 n 
 
 ,in 
 
 
 
 
 
 C 60 n 
 
 lln 
 
 
 
 8 
 
 
 CI20n 
 i 
 
 1 1 
 
 
 
 
 200 250 
 
 TEMPERATURE, 
 
 Figure 4.— Thermal gravimetric analysis illustrating effect of 
 milling time on oxidation of chalcopyrite sample A in air. 
 
 CHALCOPYRITE 
 
 Chalcopyrite, an important ore of copper, is a sulfide of cop- 
 per and iron and has a typical formula of CuFeS 2 . Results of TGA 
 tests conducted in dry air, on chalcopyrite sample A, milled for 
 15, 30, 60, and 120 min, are shown in figure 4. These tests again 
 illustrate that the ignition point decreases with decreasing particle 
 size. Two additional chalcopyrites, samples B and C, were tested 
 and their ignition points are also listed in table 4. When this sulfide 
 was oxidized in moist air, the samples gained weight prior to igni- 
 tion. Water vapor did not change the ignition point for samples A 
 and B but did raise the ignition point of sample C. XRD analysis 
 of chalcopyrite samples oxidized to a final temperature of 450° C 
 indicated the presence of ferric oxide, cupric sulfate, and cupric 
 oxide. 
 
 CHALCOCITE 
 
 Chalcocite is also an important ore of copper and it has the 
 chemical formula Cu 2 S. Figure 5 shows TGA results of chalcocite 
 sample A milled for 15 and 120 min and oxidized in dry air. In- 
 stead of the typical weight loss, this figure illustrates a sample weight 
 
 o 
 o 
 <j 
 
 < 
 
 X 
 
 120 
 
 no 
 
 100 
 
 90 
 
 
 1 1 1 
 
 Chalcocite A 
 
 1 
 
 1 f 
 Bl 
 
 - 
 
 KEY 
 
 
 / ; 
 
 
 A 15 min 
 
 
 1*1 
 
 
 B 120 min 
 
 
 U 
 
 1 1 1 1 1 
 
 100 
 
 150 
 
 200 250 300 
 
 TEMPERATURE, °C 
 
 350 
 
 400 
 
 Figure 5.— Thermal gravimetric analysis of chalcocite sample 
 A milled for 15 and 20 min and oxidized with air. 
 
58 
 
 gain. Sulfate formation would cause a weight gain, and the presence 
 of cupric sulfate in the oxidized sample was confirmed by XRD 
 analysis. As cupric sulfate formation is an exothermic reaction, and 
 as figure 5 indicates a weight change rate similar to that noted in 
 previous tests, the energy produced would be similar to that of ig- 
 nition to an oxide. 
 
 The standard heats of reaction of chalcocite and other com- 
 mon sulfides to form oxides and sulfates are listed in reference 4. 
 Table 4 lists the temperature at which rapid weight gain was in- 
 itiated for the two chalcocite samples. In this group chalcocite A 
 milled for 120 min reacted at the lowest temperature (335 ° C). The 
 addition of water vapor to the oxidizing gas increased the sample 
 weight slightly in the early stages of oxidation and it tended to raise 
 the temperatre at which rapid weight gain occurred. 
 
 GALENA 
 
 Galena (PbS), a lead sulfide, is the most common ore of lead. 
 TGA of samples milled for 120 min failed to exhibit the rapid weight 
 change observed with previous sulfides and no ignition was noted 
 at a final test temperature of 500° C. The addition of water vapor 
 caused the sulfide to undergo a slight weight gain during heating 
 and XRD analysis of samples oxidized isothermally at 300 ° C con- 
 firmed that lead sulfate was formed. 
 
 SPHALERITE AND WURTZITE 
 
 Sphalerite and wurtzite are both zinc sulfides having the same 
 chemical formula (ZnS), but with different crystal structures. TGA 
 tests conducted on the zinc sulfides to a final temperature of 500 ° 
 C gave no indication of a rapid weight change reaction (ignition). 
 The finest milled wurtzite underwent about a 1-pct weight gain dur- 
 ing oxidation with air containing 40 pet water vapor. This weight 
 gain was initiated at about 450° C. 
 
 MOLYBDENITE 
 
 The final sulfide examined in this series was molybdenite, which 
 has the chemical formula MoS 2 . The results of TGA tests conducted 
 
 UJ 
 
 o 
 
 CD 
 
 105 
 
 100 
 
 95 
 
 90 - 
 
 85 
 
 100 150 200 250 300 350 400 450 500 
 
 TEMPERATURE, °C 
 
 Figure 6.— Thermal gravimetric analysis of molybdenite milled 
 for 120 min. 
 
 on this sulfide, milled for 120 min and oxidized with air and with 
 air containing 40 pet water vapor, are shown in figure 6. The 
 presence of water vapor failed to alter the shape of the weight-change 
 curve prior to ignition and only lowered the ignition point slightly 
 from 375° to 370° C. Additional ignition point values are listed 
 in table 4 and indicate little effect due to particle size or oxidizing 
 atmosphere. An interesting sidelight of the TGA tests was the for- 
 mation of needlelike crystals of molybdenum oxide in the cool zone 
 of the furnace tube. Molybdenum disulfide begins to sublime at 
 450 ° C (5) and conducting tests to a final temperature of 500 ° C 
 apparently produces sulfide vapors that are oxidized to M0O3, which 
 in turn condenses to form crystals. 
 
 1 1 1 1 1 
 Molybdenite 
 
 i 
 
 i 
 
 
 
 \B 
 
 KEY 
 
 A\ 
 
 
 A Air 
 
 
 
 3 Air containing 40 pet 
 
 
 
 water 
 
 i i i i i 
 
 
 i 
 
 RESULTS OF ISOTHERMAL OXIDATION ANALYSIS 
 
 PYRITE 
 
 Several of the sulfides that were tested by TGA underwent a 
 weight gain prior to ignition. This weight gain was caused by the 
 formation of sulfates, which is an exothermic reaction that could 
 lead to spontaneous combustion fires. In an effort to obtain addi- 
 tional information on this reaction at lower temperatures, the six 
 pyrite samples examined by TGA were also oxidized isothermally 
 and then subjected to a chemical analysis procedure to obtain a quan- 
 titative measure of sulfide oxidation to sulfate. This method was 
 found to be more convenient than gravimetric methods for conduct- 
 ing long-term experiments. It also avoided the need to reconcile 
 sulfide weight changes that are due to the volatilization of sulfur 
 and the formation of sulfate and sulfur dioxide. 
 
 Figure 1A illustrates the conversion of the pyrite to a soluble 
 iron salt after oxidation at 200° C for 24 h in dry air and in air 
 containing water vapor. These results illustrate the effect a small 
 percentage of water vapor has on promoting the formation of water- 
 soluble ferrous sulfate at this temperature. It should be noted that 
 pyrite sample D ignited during the test because the self-heating due 
 to sulfate formation raised the temperature of the sample. Because 
 this sample burned to ferric oxide, it was not analyzed. A similar 
 test series was run at 100° C and these results are illustrated in 
 figure IB. Comparing the results, oxidation with air containing 5 
 pet water vapor produced slightly more sulfate than dry air. Rais- 
 ing the water vapor content to 60 pet promoted the oxidation of 
 sulfide to sulfate with sample D undergoing a 30-pct conversion 
 in 24 h. 
 
59 
 
 O 
 CO 
 
 CO 
 
 cr 
 
 UJ 
 
 > 
 z 
 o 
 o 
 
 Dry ai r 
 
 Air with 5 pet 
 water vapor 
 
 Air with 60 pet 
 water vapor 
 
 Figure 7.— Isothermal oxidation of pyrite samples (A-F) to soluble iron at (A) 200° C and (B) 100° C. 
 
60 
 
 CONCLUSIONS 
 
 TGA was used to measure the reactivity of 10 types of metal 
 sulfides in terms of the temperature at which the samples under- 
 went the rapid weight change associated with ignition or sulfate 
 formation. For the six pyrite samples there was a 40° to 50° C 
 range in the ignition points. Attempts at associating the variability 
 in ignition point with minor differences in the chemical or XRD 
 analyses was not conclusive. Although each pyrite sample was 
 milled in the same manner, it is possible that differences in the grind- 
 ability of the pyrites could result in slightly different particle size 
 distributions that could influence reactivity. Research to resolve the 
 factors that cause one sample to exhibit a greater reactivity than 
 another is continuing. 
 
 Adding water vapor to the oxidizing atmosphere was found to 
 raise the ignition point for all the pyrites. The addition of water 
 vapor also caused the pyrites to gain weight prior to ignition. 
 Chemical and XRD analysis of partially oxidized samples confirmed 
 that the weight gain was due to the formation of ferrous sulfate. 
 As this is an exothermic reaction it is quite probable that sulfate 
 formation can lead to the self-heating that causes sulfide ignition. 
 
 Pyrrhotite is often suspected as the mineral responsible for 
 sulfide fires, however, the pyrrhotite samples examined in this report 
 had fairly high ignition points. This was not unexpected as Good 
 (J) reported a wide variation in ignition point values for samples 
 taken from an ore pillar with known self-heating tendencies. Water 
 vapor in the oxidizing atmosphere had a small effect on the pyr- 
 rhotite ignition points determined by TGA and had a mixed effect 
 on sulfate formation. 
 
 The ignition point measured for the finely milled marcasite, 
 when oxidized in moist air, was the lowest of all the sulfides tested. 
 While water vapor had the effect of raising the ignition point of 
 pyrite, it lowered the value measured for marcasite. Water vapor 
 also promoted the formation of sulfate during the oxidation of mar- 
 casite. Arsenopyrite ignited at about the same temperature as pyr- 
 rhotite and like marcasite its ignition point was lowered when water 
 vapor was present. Water vapor also promoted the weight gain of 
 this sulfide at lower temperatures. 
 
 The ignition points of the chalcopyrite samples were in- 
 termediate to the common iron sulfides of pyrite and pyrrhotite. 
 The presence of water vapor tended to promote sulfate formation 
 but it had a mixed effect on the ignition point. Chalcocite failed 
 
 to ignite during TGA testing, instead it underwent a rapid weight 
 gain that was caused by the formation of cupric sulfate. As the heat 
 produced from this exothermic reaction would be similar to igni- 
 tion, the temperature at which it occurred was noted. These values 
 were similar to the ignition points measured for chalcopyrite. Water 
 vapor in the oxidizing atmosphere had a slight effect in promoting 
 sulfate formation in the early stages of oxidation of chalcocite but 
 tended to raise the temperature at which rapid weight gain occurred. 
 The lead and zinc sulfide samples failed to ignite below 500° C 
 and because of their low reactivity little sample weight gain was 
 noted at lower temperatures. Molybdenite ignited at about the same 
 temperature as the copper sulfides but did not exhibit the weight 
 gain prior to ignition that was noted with many of the other sulfides. 
 While the ignition point of the iron and copper sulfides decreased 
 as the particle size decreased, the ignition point of molybdenite was 
 not affected by particle size. 
 
 Based on ignition point, several of the pyrites and the mar- 
 casite sample were found quite reactive. The remaining iron and 
 copper sulfides were found to have slightly higher ignition point 
 values but like pyrite and marcasite may form sulfates at lower 
 temperatures that could lead to self-heating. The four iron sulfides 
 (pyrite, marcasite, pyrrhotite, and arsenopyrite) have all been men- 
 tioned as being responsible for initiating spontaneous combustion 
 in underground mines (7) and in certain mines, sulfide samples could 
 be found that are much more reactive than those studied in this 
 report. The lead, zinc, and molybdenum sulfides tested were less 
 reactive and by themselves should not present a spontaneous com- 
 bustion hazard. 
 
 Isothermal oxidation followed by chemical analysis was found 
 to be a useful method to quantitatively measure the conversion of 
 pyrite to sulfate at lower temperatures. This method illustrated the 
 effect that water vapor had on promoting this exothermic reaction 
 when oxidizing pyrite at 100° and 200° C. One of the more reac- 
 tive pyrite samples ignited during testing at 200° C and this in- 
 dicated self-heating to the ignition point when sulfate was formed. 
 TGA testing uses a very small sample and the heat produced dur- 
 ing sulfate formation is dissipated, however, when a larger sample 
 was used the heat produced exceeded the heat that was dissipated 
 and the sulfide in the combustion boat reached the ignition point. 
 
 REFERENCES 
 
 1. Ninteman, D. J. Spontaneous Oxidation and Combustion of Sulfide 
 Ores in Underground Mines: A Literature Survey. BuMines IC 8775, 1978, 
 36 pp. 
 
 2. Farnsworth. D. J. M. Introduction to and Background of Sulphide 
 Fires in Pillar Mining at the Sullivan Mine. CIM Bull., v. 70, No. 782, 
 June 1977. pp. 65-71. 
 
 3. Good, B. H. The Oxidation of Sulphide Minerals in the Sullivan Mine. 
 CIM Bull., v. 70, No. 782, June 1977, pp. 83-88. 
 
 4. Reimers, G. W., and K. E. Hjelmstad. Analysis of the Oxidation 
 of Chalcopyrite, Chalcocite, Galena, Pyrrhotite, Marcasite, and 
 Arsenopyrite. BuMines RI 9118, 1987, 16 pp. 
 
 5. Killeffer, D. H., and A. Linz. Molybdenum Compounds— Their 
 Chemistry and Technology. Interscience Publ., 1952, 407 pp. 
 
61 
 
 EMISSION PRODUCTS FROM WOOD CRIB AND 
 TRANSFORMER FLUID FIRES 
 
 By Margaret R. Egan 1 
 
 ABSTRACT 
 
 The Bureau of Mines investigated the characteristics of the combustion products of 
 wood cribs and transformer fluid. The products from these two diverse fuels were analyzed 
 for gas production and smoke characteristics. This included the combustion of four wood 
 crib configurations at several ventilation rates and three commercially available brands of 
 transformer fluid. 
 
 Each fuel was studied independently and the results compared. These studies indicate 
 that wood crib fires may be more difficult to detect because they are cleaner burning (as 
 measured by the smoke obscuration level). By comparison, transformer fluid fires produce a 
 thick smoke that reduces visibility, making escape and rescue more difficult. 
 
 INTRODUCTION 
 
 The Bureau of Mines conducts research to improve health 
 and safety conditions in the mining industry. Exceptional 
 circumstances, such as an underground fire, create life- 
 threatening situations. Escape is dependent upon early detec- 
 tion of a fire, and is hampered by reduced visibility due to 
 smoke and the toxicity of the combustion products. In order to 
 design more efficient detection and rescue equipment, the 
 Bureau has investigated emission products of combustible 
 materials found in underground mines. Similar analyses could 
 also be beneficial in determining the existence, stage, or extent 
 of an underground fire. 
 
 These two fuels investigated were chosen either because of 
 the quantities used or their potential hazard. Wood is the most 
 
 abundant material found in mines. It has many uses especially 
 as supports for mine workings. Transformer fluid is the 
 coolant used in electrical transformers. Fires and explosions 
 are potential risks whenever petroleum products are part of 
 electrical equipment. 
 
 The objectives of this study were to analyze and compare 
 the combustion emission products of these two fuels. Com- 
 parisons of heat-release rates, smoke particle sizes, smoke 
 obscuration levels, and smoke and gas production constants 
 are included. These measurements form a data base with 
 which results of future studies of other mine combustibles can 
 be compared. 
 
 EXPERIMENTAL EQUIPMENT 
 
 INTERMEDIATE-SCALE FIRE TUNNEL 
 
 The Bureau's intermediate-scale fire tunnel was used to 
 simulate a mine environment. This tunnel has been shown to 
 successfully predict full-scale mine conditions. 2 A schematic 
 of the tunnel with its data acquisition system is shown in figure 
 1. The tunnel is 0.8 m wide by 0.8 m high by 10 m long and is 
 divided into several sections. The first horizontal section is 1.5 
 
 1 Research chemist, Pittsburgh Research Center, Bureau of Mines, Pittsburgh, 
 PA 
 
 2 Lee, C. K., R. F. Chaiken, J. M. Singer, and M. E. Harris. Behavior of 
 Wood Fires in Model Tunnels Under Forced Ventilation Flow. Tests with 
 Untreated Wood. BuMines Ri 8450, 1980, 58 pp. 
 
 m long and cone shaped. It is hinged and can be lifted to allow 
 entrance for the placement of the fuel. It begins with an 
 air-intake cylinder that is 0.25 m long by 0.3 m in diameter and 
 gradually enlarges until it matches the tunnel dimensions at the 
 hinged area. Next is the fire zone where a gas burner is located 
 and the fuel pan is balanced on a load cell. The fire zone and 
 the remaining horizontal section are lined with firebrick and 
 contain thermocouples, flow probes, and gas sampling ports. 
 The diffusing grid begins the vertical exhaust section of the 
 tunnel. Located in this section is an orifice plate that can be 
 manually adjusted to attain the desired airflow. The final 
 section of the exhaust section is horizontal and ends with an 
 exterior fan. 
 
62 
 
 22 m 
 
 10- m length 
 
 0.8 -m square duct 
 
 intake 
 
 Load eel 
 
 I2 _ m length 
 0.6l - m diam duct 
 
 -0.305-nrdiam 
 entrance duct 
 hinged and movable) 
 
 Manually 
 adjustable 
 orifice plate 
 
 ' — Diffusing grid 
 
 Air 
 
 TP 
 
 'mi 
 
 1 exhaust 
 Ventilation 
 
 fan 
 ( 2-speed ) 
 
 TEOM 
 
 CNM 
 
 CO meter 
 
 CO2 meter 
 
 Pressure transducers 
 
 LLLJ 
 
 48" 
 channel 
 data- 
 acqui- 
 sition 
 system 
 
 3X 
 - detector 
 
 -Load cell 
 
 -Digital input 
 
 for CNM 
 
 range 
 
 DECNETn 
 
 PDP 
 11/44 
 
 Control 
 terminal 
 
 \ . 
 
 VAX 
 
 11/780 
 
 Printer 
 
 VAX 
 
 terminal 
 
 ■•• • 2o • ••• 
 thermocouples 
 
 A Pressure transducer (flow probe) 
 
 x Differential pressure transducer 
 
 * 3\ detector 
 
 • Thermocouples 
 ■ Sampling ports 
 
 CALCOMP 
 plotter 
 
 KEY 
 
 CALCOMP California computer products 
 
 CNM Condensation nuclei monitor 
 
 DECNET Digital equipment networking 
 
 PDP Programmed data processor 
 
 VAX 
 
 Virtual address extension 
 
 Figure 1 .—Schematic of intermediate-scale fire tunnel (top) and data-acquisition system (bottom). 
 
 DATA LOGGING 
 
 The tunnel was equipped with a 48-channel data collec- 
 tion system. As the experiments were in progress, all channels 
 were scanned, recorded, and calculated. 
 
 WOOD CRIBS 
 
 Four configurations of wood cribs were tested in dupli- 
 cate. The height and spacing of the sticks were chosen to study 
 the growth and propagation of the fire. The cribs were 
 
 constructed of Douglas fir using a small amount of carpenter's 
 glue for joining. The configurations and their dimensions are 
 shown in figure 2. 
 
 TRANSFORMER FLUID 
 
 Ten experiments were completed using three commercially 
 available brands of transformer fluid — four with brand A and 
 three with brands B and C. A 25-cm-diam pan filled with 
 transformer fluid to a depth of 2.5 cm was used for each 
 experiment. 
 
63 
 
 3.01625 cm 
 
 29.21cm 
 
 29.21 cm 
 
 Standard high 
 
 Quadratic 
 
 29.21 cm 
 
 Standard low Linear 
 
 Figure 2.— Crib configurations. Air flows from left to right. 
 
 INSTRUMENTATION 
 
 All instruments were periodically cleaned and calibrated 
 according to manufacturers' instructions for the quantity of 
 smoke or gas and the amount of use each had received. 
 
 GAS MONITORS 
 
 The CO analyzer measures accurately within 1 pet of full 
 range or ±5 ppm. The C0 2 analyzer measures accurately 
 within 1 pet of full range or ±250 ppm. These analyzers were 
 calibrated at the beginning of each experiment. In addition, 
 the concentrations of the span gases were checked at the 
 beginning of each series of experiments. 
 
 SMOKE MONITORS 
 
 The particle number concentrations (N ) of the smoke 
 particles were obtained by a condensation nuclei monitor, 
 manufactured by Environment One Corp. 3 In the range of 
 
 these experiments, the calculated error including the dilution 
 factor was ±22 pet. 
 
 The particle mass concentration (M ) of the smoke was 
 obtained by a tapered-element oscillating microbalance, devel- 
 oped by Rupprecht and Patashnick Co., Inc. It is capable of 
 measuring dust concentrations with a better than 10 pet 
 accuracy at the 250-ftg/m 3 level. 
 
 A three-wavelength light transmission technique was used 
 to measure particle size and light obscuration. This technique 
 was developed by the Bureau. 4 
 
 WEIGHT-LOSS MONITOR 
 
 Continuous weight loss information was obtained by a 
 strain-gauge conditioner and a load cell with a range up to 
 22.68 kg. Their combined accuracy is stated as 0.05 pet of full 
 scale or ±11.3 g. 
 
 3 Reference to specific products does not imply endorsement by the Bureau of 
 Mines. 
 
 Cashdollar, K. L., C. K. Lee, and J. M. Singer. Three-Wavelength Light 
 Transmission Technique To Measure Smoke Particle Size and Concentration. 
 Appl. Opt., v. 18, No. 11, 1979, pp. 1763-1769. 
 
64 
 
 TYPICAL TEST PROCEDURE 
 
 The pan containing the fuel was positioned inside the 
 tunnel. The shaft of the pan extended through a hole in the 
 tunnel floor and was supported on the load cell. Prior to each 
 experiment, background readings were obtained after the fuel 
 was positioned and the exhaust fan started. All instruments 
 were continuously scanned and data were recorded throughout 
 the experiment. 
 
 The fuel was ignited with a natural gas burner located 
 immediately upstream from the pan. Once ignited, the fuel 
 rapidly reached a steady-state, flaming stage. Once the flames 
 are no longer visible, the decaying or smoldering stage begins. 
 
 At a ventilation rate of approximately 0.65 m 3 /s, the 
 transformer fluid took less than 1 min to ignite. The ventila- 
 
 tion rate was then lowered to approximately 0.47 m 3 /s during 
 the steady-state burning. The flames engulfed the entire pan 
 instantaneously. As the transformer fluid was consumed, the 
 flames began to die down. The experiments were concluded 
 when the flames were no longer visible. 
 
 The ignition time of the wood was longer (at least 3 min) 
 depending on the crib configuration. Most cribs were burned 
 at a ventilation rate of approximately 1 m 3 /s. They burned 
 rapidly for at least 7 min and smoldered for an additional 30 
 min. The experiments were concluded when the CO concen- 
 tration returned to background level. 
 
 CALCULATIONS 
 
 It is necessary to measure certain parameters in order to 
 compare the steady-state combustion products and ultimately 
 the hazards of various fuels. Among these are gas concentra- 
 tions, smoke particle mass and number concentrations, venti- 
 lation rate, and mass loss rate. Other combustion properties 
 can be calculated once these values are known. 
 
 PRODUCT GENERATION RATES 
 
 The generation rate (G x ) of a gas is related to its density, 
 ventilation rate, and concentration by the expressions 
 
 and 
 
 where M c 
 
 M r 
 
 x ACO, 
 
 x V A x ACO, 
 
 (1) 
 (2) 
 
 and 
 
 V A 
 AX 
 
 = 1.97 x 10- 3 g/(m 3 ppm), 
 
 = 1.25 x 10- 3 g/(m 3 ppm), 
 
 = ventilation rate, m 3 /s, 
 
 = measured change in a given gas, ppm. 
 
 Substituting these values in equation 3 yields 
 Q A = V A [(0.01875)(ACO 2 ) + 6.06 X 10" 3 (ACO)] (4) 
 
 for wood and 
 
 Qa = V A [(0.025 l)(ACO z ) + 7.07 x 10" 3 (ACO)] (5) 
 
 for transformer fluid. Because measurements of V A , AC0 2 , 
 and ACO were made continuously, the actual heat release rates 
 and gas generation rates could be calculated, using equations 1 
 through 5. 
 
 A typical fire rarely attains the state of complete combus- 
 tion. Therefore, the actual heat of combustion (H A ) is usually 
 less than the total heat of combustion. By calculating both the 
 actual heat release rate, using equations 4 and 5, and measur- 
 ing the fuel mass loss rate, M f , the actual heat of combustion 
 can be calculated from the expression 
 
 H A = Q A /M f . 
 
 (6) 
 
 HEAT RELEASE RATES 
 
 It has been shown 5 that the total heat release rate realized 
 during a fire can be calculated from the expression. 
 
 
 H, 
 
 Hco (K C o) 
 
 K, 
 
 (3) 
 
 where Q A = actual heat release, kW, 
 
 H c = net heat of complete combustion of the fuel, 
 16.4 kJ/g for wood and 40.7 kJ/g for trans- 
 former fluid, 
 K co = theoretical yield of C0 2 , 1.723 g/g for wood 
 
 and 3.19 g/g for transformer fluid, 
 H co = heat of combustion of CO, 10.1 kJ/g, 
 and K co = theoretical yield of CO, 1 .097 g/g for wood and 
 2.03 g/g for transformer fluid. 
 
 5 Tewarson, A. Heat Release Rate in Fires. Fire and Mater, v. 4, No. 4, 1980, 
 pp. 185-191. 
 
 In a flaming fire, the actual heat release rate can be used to 
 estimate the fire hazards. However, in an actual mine fire, it is 
 difficult, if not impossible, to measure the fuel mass loss. 
 Therefore, the actual heat of combustion cannot be calculated. 
 Since the true yield of a combustion product depends upon 
 this information, significant errors can result in predicting the 
 resultant concentration increases. 
 
 PRODUCTION CONSTANTS 
 
 The generation and heat release rates can be used to 
 calculate production constants, or beta values (/3 X ), by the 
 expression 
 
 0x = Gx/Q, 
 
 (7) 
 
 Once beta values have been determined for a given fuel, the 
 resultant gas and smoke concentration can then be calculated 
 as a function of the ratio of fire size to ventilation rates. 
 
65 
 
 SMOKE PARTICLE DIAMETERS 
 
 Measurements of both number and mass concentrations 
 of the smoke can be used to calculate the average size of the 
 smoke particles by the expression 
 
 7rd r 
 
 N„ = 1 x 10 j IvL 
 
 (8) 
 
 where 
 
 p p = individual particle density, g/cm 3 , 
 d m = diameter of a particle of average mass, fim, 
 and 1 x 10 3 = the appropriate unit's conversion factor. 
 Assuming a value of p p = 1.4 g/cm 3 , then the diameter of 
 average mass can be calculated from 
 
 d m = 11-09 (£ 
 
 (9) 
 
 when the particle diameter is expressed in micrometers. 
 
 Another method was used to determine the size of the 
 smoke particles. It uses the three-wavelength smoke detector. 
 
 Light was passed through the smoke and transmission of light 
 (T) was calculated for each wavelength. An extinction- 
 coefficient ratio can be calculated for each pair of wavelengths 
 (X) by the following log-transmission ratios: 
 
 InT(M.OO) lnT(Al.QO) 
 lnT(\0.63) ' lnT(X0.45) 
 
 or 
 
 lnT(A0.63) 
 lnT(\0.45) 
 
 Using these extinction coefficients and the Cashdollar curve, 6 
 the mean particle size (d 32 ) can be determined. (Calculation of 
 the extinction-coefficient curves assumes spherical particles 
 with an estimated refractive index.) 
 
 Smoke obscuration is the percentage of light absorbed by 
 the smoke or 100 pet of the light minus the percent transmis- 
 sion. It is calculated using the following equation: 
 
 Obscuration = 100(1 - T). 
 
 (10) 
 
 The obscuration rate is an average of the attenuation of the 
 beam of light at the two wavelengths in the visible range, 0.45 
 and 0.63 /xm. 
 
 RESULTS OF COMBUSTION OF WOOD CRIBS 
 
 The wood cribs burned in three distinctive stages: igni- 
 tion, flaming, and decaying. Table 1 lists the length of the 
 ignition stage and the steady-state burning, the mass loss rates 
 for each stage, and the ventilation rates for each experiment. 
 Steady-state burning refers to the rapidly burning period. The 
 mass loss rate is the rate at which the wood was consumed. The 
 length of the ignition stage indicates the ease with which each 
 configuration was ignited. It was dependent upon the place- 
 ment of the sticks for heat transfer and their accessibility to the 
 burner flames. 
 
 All the values listed in this paper are an average of the 
 steady-state flaming stage. It was in this stage that the greatest 
 mass loss, heat, and C0 2 were produced. Following steady- 
 state burning is the decaying or smoldering stage in which most 
 of the CO was produced. At this time, 84 pet of the available 
 wood had been burned and only a low-level postflame smol- 
 dering remained. 
 
 Graphs in figure 3 compare CO versus C0 2 concentra- 
 tions, heat-release versus total mass loss, particle mass versus 
 number concentrations, and show the diameter of average 
 mass as a function of time for test 3. In this experiment, the 
 standard high crib ignited at 4 min at which time the burner 
 was turned off. 
 
 GAS CONCENTRATIONS AND HEAT 
 PRODUCTION 
 
 Table 2 lists the gas concentrations, generation rates, and 
 heat production for each crib configuration. The high produc- 
 tion of C0 2 indicated that these wood cribs were burning in an 
 oxygen-rich environment. The reaction favored complete com- 
 bustion. An average concentration of 6,760 ppm was produced 
 during the flaming stage when the wood was being consumed 
 at its maximum rate. The average generation rate for C0 2 was 
 calculated to be 11.5 g/s. 
 
 The CO concentration averaged 145 ppm during the 
 flaming stage. Because CO is a product of incomplete com- 
 bustion, it reaches its highest concentration in the decaying 
 
 Table 1. — Ignition times, steady-state burning times, 
 ventilation rates, and mass loss rates for wood cribs 
 
 Crib type ] f mn Steady-state Venti | ation , Mass loss rate, g/min 
 
 and test time ' burmn 9 time . m 3 /s 
 
 min min Ignition Flaming Decaying 
 
 Standard low: 
 
 Test 2 8.1 15.3 1.0525 28.9 448.86 13.96 
 
 Test 4 14.68 15.62 1.0363 17.65 412.96 26.65 
 
 Quadratic: 
 
 Test 1 6.42 10.08 1.0279 19.9 647.5 17.3 
 
 Test 8 9.23 15.7 .8211 15.41 529.87 5.75 
 
 Standard high: 
 
 Test 3 3.6 10.8 1.0471 38.31 725.35 5.7 
 
 Test 7 4.6 12.0 .4859 63.51 668.14 16.72 
 
 Linear: 
 
 Test 5 9.75 8.92 1.0018 35.38 778.62 5.59 
 
 Test 6 6.0 6.7 .6167 48.99 769.68 13.43 
 
 Table 2. — Gas concentrations, generation rates, 
 
 and heat production for wood cribs 
 
 Crib type C0 2 , G C02 , CO, G co , Q A , 
 
 and test ppm g/s ppm 10" 2 g/s kW 
 
 Standard low: 
 
 Test 2 4,948 10.26 102 13.42 98.2 
 
 Test 4 4,321 8.82 119 15.41 84.6 
 
 Quadratic: 
 
 Test 1 ND ND 143 18.37 ND 
 
 Test 8 5,921 9.58 162 16.63 91.9 
 
 Standard high: 
 
 Test 3 7,215 14.88 204 26.70 142.8 
 
 Test 7 3,446 3.30 120 7.29 31.7 
 
 Linear: 
 
 Test 5 10,158 20.05 143 17.91 191.3 
 
 Test 6 11,303 13.73 163 12.57 131.1 
 
 ND Not determined. 
 
 H A . 
 kJ/g 
 
 13.1 
 12.3 
 
 ND 
 10.4 
 
 11.8 
 
 2.9 
 
 14.7 
 10.2 
 
 Work cited in footnote 4. 
 
66 
 
 400 
 
 14 
 
 90 
 
 12 
 
 80 
 
 
 70 
 
 10 
 
 
 E 
 
 CL 
 
 ■° E 60 
 
 o 
 
 •v 
 
 I 1 50 
 
 6 c\j 
 
 £ 40 
 
 o 
 
 < 
 
 u 
 
 s 30 
 
 4 
 
 
 
 20 
 
 2 
 
 
 
 10 
 
 
 
 
 
 8 
 
 0.35 
 
 o 
 
 co 
 
 CO 
 O 
 
 CO 
 CO 
 < 
 
 
 1 
 
 1 
 
 1 ' ' c 
 
 - 
 
 
 
 — 
 
 - 
 
 1 If 
 A i ' 
 
 
 KEY 
 Mass 
 
 /'l 
 1 1 
 
 1 1 ' V 
 
 \ V V 
 
 
 Number 
 
 1 1 
 1 
 — \f* 
 
 
 \\ '\ 
 
 A 
 
 
 , / 
 
 V' i 
 
 
 
 , 1 
 I f 
 
 1 ' 
 
 \ 1 
 
 
 AI'V 
 \ \ 
 
 \ *■ 
 / \ N 
 i \ \ 
 
 f 
 
 \l 
 
 
 \ 1 \ v 
 
 *\ 
 
 
 
 \ / \ N 
 
 1 
 
 
 
 \J ^s~\ \ _ 
 
 1 
 
 1 
 
 1 
 
 i V 
 
 20 
 18 
 
 14 | 
 
 a. 
 
 12^ 
 o 
 
 10 ~ 
 cr 
 
 o UJ 
 O CD 
 
 6 3 
 
 z 
 4 
 
 2 
 
 
 
 Figure 3 
 C, particle 
 
 TIME, mm 
 
 —Results of test 3, standard high crib. A, Heat-release rate and mass loss rate; B, CO and C0 2 concentrations; 
 mass and number concentrations; and D, diameter of average mass. 
 
 
 stage. Approximately 21 min after the crib ignited, the average 
 maximum concentration of 378 ppm was achieved. Generated 
 CO levels, at this time, averaged 0.39 g/s. Figure 3A shows the 
 actual gas concentrations for test 3. A shift of the peak 
 concentration for each gas is clearly visible. 
 
 The concentrations of CO and C0 2 measured during the 
 flaming Fire can be used to calculate the fire size, provided the 
 ventilation in the affected entry is known. Remote sensing of 
 these two gases could contribute significant information about 
 the intensity and stage of a mine fire. In these experiments, the 
 average fire size or heat release was 110.2 kW. 
 
 Figure 35 shows the relationship of the heat-release rate to 
 the mass loss rate. As the majority of the wood is consumed, 
 the greatest heat is released. 
 
 SMOKE CHARACTERISTICS 
 
 The average number of smoke particles (N ) produced 
 during the flaming stage was 6.0 x 10 6 p/cm 3 . The average 
 particle mass concentration (M ) produced during the flaming 
 stage was 49 mg/m 3 . In most experiments, the production of 
 smoke dropped as the flaming stage was ending but increased 
 slightly during the decaying stage. Figure 3C shows the actual 
 data for test 3 and table 3 lists the smoke characteristics for 
 each experiment. 
 
 The size of the smoke particles showed the same variation 
 as the smoke production. The largest particles (averaging 0.22 
 /*m) were produced in the flaming stage during their highest 
 production rate. Figure 3D shows the data for test 3. The 
 
67 
 
 Table 3.— Smoke characteristics for wood cribs 
 
 Crib type N , M , d m , Obscuration rate, 
 
 and test 10 6 p/cm 3 mg/m 3 ^m pet 
 
 Standard low: 
 
 Test 2 5.26 82.84 0.278 2.1 
 
 Test 4 7.20 31.40 .181 13.0 
 
 Quadratic: 
 
 Test 1 1.13 ND ND 12.0 
 
 Test 8 8.09 84.99 .243 5.2 
 
 Standard high: 
 
 Test 3 9.06 40.90 .183 6.0 
 
 Test 7 3.24 33.38 .241 9.8 
 
 Linear: 
 
 Test 5 9.02 38.62 .180 13.1 
 
 Test 6 5.31 31.70 .201 7.9 
 
 ND Not determined. 
 
 Table 4.— Production constants for wood cribs 
 
 Crib type /3 CC , 2 , /3 co , /3 No , Mo , 
 
 and test 10^ 2 g/kJ 10' 3 g/kJ 10'° p/kj 10~ 4 g/kJ 
 
 Standard low: 
 
 Test 2 10.45 1.37 5.64 8.88 
 
 Test 4 10.42 1.82 8.82 3.85 
 
 Quadratic: 
 
 Test 1 ND ND ND ND 
 
 Test 8 10.43 1.80 7.23 7.60 
 
 Standard high: 
 
 Test 3 10.42 1.87 6.64 3.00 
 
 Test 7 10.40 2.29 4.96 5.11 
 
 Linear: 
 
 Test 5 10.48 .94 4.72 2.02 
 
 Test 6 10.48 .96 2.50 1.49 
 
 ND Not determined. 
 
 quantity, mass, and size showed a slight increase during the 
 decaying stage. The average diameter of mass during the 
 decaying stage was 0.16 ^m. 
 
 PRODUCTION CONSTANTS 
 
 Table 4 lists the production constants for all wood exper- 
 iments. Crib configuration seemed to be of more importance 
 than the ventilation rate. The standard high and linear cribs 
 tended to have lower smoke production constants than the 
 other configurations. They were also the fastest burning cribs. 
 The linear crib also had the lowest CO production constant 
 because more complete combustion occurred during its short 
 burning time. The C0 2 production constants are all approxi- 
 mately the same indicating the even production regardless of 
 the fire size. 
 
 CRIB CONFIGURATION 
 
 As can be seen in table 1 , the configuration of the crib was 
 a significant factor in the ignition time, the length of the 
 burning time, and the mass loss rate. The rate at which any fire 
 propagates largely depends upon the heat transfer. When the 
 arrangement of the sticks and the ventilation rate was varied, 
 a noticeable difference in the heat transfer occurred. The best 
 configuration for fire propagation was the linear crib. Its stick 
 arrangement permitted the most efficient heat transfer, which 
 produced the greatest mass loss and heat release rates and 
 shortest burning time. Its flaming stage lasted only 8 min. 
 
 Lowering the profile of the crib decreased the mass loss 
 rate and lengthened the time of ignition. The standard low crib 
 had the least efficient heat transfer, resulting in the longest 
 steady-state stage and lowest mass loss rate. 
 
 Decreasing the angle at which the sticks were positioned 
 increased the mass loss rate but did not affect the time of 
 ignition. The quadratic and linear cribs ignited in about the 
 same time, but the linear crib, with a lower angle and a more 
 efficient heat transfer, burned faster. 
 
 VENTILATION RATE 
 
 Lowering the ventilation rate reduced the mass loss rate. 
 With less available oxygen, the wood was consumed at a much 
 slower rate. In addition, the efficiency of the heat transfer was 
 reduced. In underground fires, it may be beneficial to decrease 
 or stop the airflow, if possible. This would slow the progress of 
 the fire by reducing the available oxygen and by lowering the 
 heat transfer. The same relationship of burning rates to 
 airflows has been reported. 7 However, the burning rates 
 reached a maximum at airflows between 2 and 3 m 3 /s. Beyond 
 this point, convective cooling competes with the heat transfer 
 and slows the burning rate. 
 
 Larger sized particles were produced at lower ventilation 
 rates. The slower transport time may have permitted the 
 particles to coagulate. Incomplete combustion may also have 
 contributed to the production of larger sized particles. 
 
 A representative sample of the combustion gases may be 
 difficult to obtain with low ventilation rates. Stratification of the 
 gas layers may have accounted for the low C0 2 concentration. 
 
 BURNING RATE 
 
 The burning rate affects the smoke particle characteris- 
 tics. The rapid oxidation of the linear crib yielded a clean 
 burning fire with less smoke and toxic gas production. Much 
 higher production constants are generated from the slower 
 burning standard low crib. The other smoke characteristics 
 show little variation among the different configurations. The 
 smoke obscuration must be at least 15 pet before the mean 
 particle diameter (d m ) can be calculated. 
 
 7 Tewarson, A. Analysis of Full-Scale Timber Fire Sets in a Simulated Mine 
 Gallery. Factory Mutual Res. Corp., Norwood, MA, Tech. Rep. J. I. OEON1.RA 
 and J. I. OFON3.RA, June 1982, 55 pp. 
 
bS 
 
 RESULTS OF COMBUSTION OF TRANSFORMER FLUID 
 
 All three brands showed similar results for gas production, 
 heat release, and heat of combustion. However, the tested brands 
 showed somewhat different results for smoke characteristics. 
 
 GAS CONCENTRATIONS AND HEAT 
 PRODUCTION 
 
 The CO concentration remained fairly constant, rising 
 slightly as the flames died. The C0 2 concentration gradually 
 rose at an average rate of 22 ppm/min as the fuel was 
 consumed. The CO and C0 2 concentrations for all brands are 
 found in figures 44 and 4B. 
 
 The heat release rate remained fairly constant for most of 
 the experiments, increasing slightly just before the flames died. 
 The average total mass loss was 2,000 g, at a rate of 0.87 g/s. 
 The actual heat of combustion also remained fairly constant 
 throughout the steady-state burning but rose sharply just 
 before the fuel was completely consumed. The average values 
 for each brand are listed in table 5. 
 
 SMOKE CHARACTERISTICS 
 
 The particle number concentration (N ) slowly increased 
 until the fuel was almost completely consumed and then it 
 started to drop. The particle mass concentration (M ) varied 
 throughout the experiments. An average was taken during the 
 steady-state burning when it was the most stable. Using these 
 values, the diameter of average mass (d m ) was calculated. The 
 average values for each brand are listed in table 6. The mass 
 and number concentrations and the diameter of average mass 
 for each brand are found in figure 5. 
 
 Using the three-wavelength smoke detector, the average 
 mean particle size (d 32 ) can be calculated for each wavelength 
 These averages and the obscuration rate for each brand are 
 listed in table 7. 
 
 Table 5. — Gas concentrations and heat production 
 for transformer fluid 
 
 Brand 
 
 CO, 
 
 ppm 
 
 C0 2 , 
 
 ppm 
 
 Qa. 
 
 kW 
 
 A 
 
 B 
 
 C 
 
 Average . 
 
 112 
 
 1,779 
 
 21.5 
 
 Ha, 
 
 kJ/g 
 
 120 
 
 1,682 
 
 20.8 
 
 24.1 
 
 97 
 
 1,871 
 
 22.5 
 
 25.5 
 
 120 
 
 1,784 
 
 21.3 
 
 24.9 
 
 24.8 
 
 Table 6. — Smoke characteristics for transformer fluid 
 
 Brand 
 
 N , 
 
 10 6 p/cm 3 
 
 M , 
 
 mg/m 3 
 
 d m , 
 
 A 
 
 B 
 
 C 
 
 Average . 
 
 1.57 
 
 40.6 
 
 0.309 
 
 1.35 
 
 9.0 
 
 .211 
 
 .70 
 
 58.1 
 
 .501 
 
 1.21 
 
 35.9 
 
 .340 
 
 Table 7. — Mean particle sizes and obscuration rates 
 for transformer fluid 
 
 Brand 
 
 lnT(X0.63), 
 lnT(X0.45) 
 
 /i(T! 
 
 InT(XLOO), 
 lnT(X0.45) 
 
 lnT(X1.00), 
 
 lnT(X0.63) 
 /xm 
 
 Average 
 d 32 , 
 
 Obscuration 
 rate, 
 pet 
 
 A 
 
 B 
 
 C 
 
 0.302 
 .339 
 .336 
 
 0.351 
 .395 
 .431 
 
 0.391 
 
 .440 
 .538 
 
 0.348 
 .391 
 .435 
 
 39.9 
 
 40.8 
 59.6 
 
 Average 
 
 326 
 
 .392 
 
 .456 
 
 .391 
 
 46.8 
 
 1 60 
 
 I40 
 
 I20 
 
 EIOO - 
 
 o 
 ° 80 
 
 60 - 
 
 40 
 
 20 
 
 KEY 
 
 Brand A 
 
 Brand B 
 
 — Brand C 
 
 \ 
 
 w 
 
 I 
 
 20 
 
 30 
 
 40 
 
 50 
 
 TIME, mm 
 Figure 4.— Comparison of gas production for three brands of transformer fluid. A, CO concentrations; B, C0 2 concentrations. 
 
69 
 
 70 
 
 60 
 
 50 
 
 40 
 
 30 
 
 20 
 
 10 
 
 0.7 
 
 'i 
 
 KEY 
 
 - Brand A 
 
 - Brand B 
 
 - Brand C 
 
 _ I I 1 L 
 
 L^ 
 
 
 2.0 
 
 
 1.8 
 
 
 1.6 
 
 ■o 
 
 1 4 
 
 fc 
 
 
 
 
 V 
 
 
 Q. 
 
 1.2 
 
 o 
 
 
 c r 
 
 1.0 
 
 y 
 
 
 o 
 
 
 ( ) 
 
 
 ir 
 
 08 
 
 UJ 
 
 
 CD 
 
 
 5 
 
 0.6 
 
 ■z. 
 
 
 
 0.4 
 
 
 0.2 
 
 rv\ 
 
 v y V.' V V\ A< 
 
 - \ 
 
 10 20 30 
 
 TIME, min 
 
 40 
 
 50 
 
 Figure 5.— Comparison of smoke characteristics for 
 three brands of transformer fluid. A, particle mass 
 concentration; S, particle number concentration; and 
 C, diameter of average mass. 
 
 20 30 
 
 TIME, min 
 
 PRODUCTION CONSTANTS 
 
 The production constants or beta values are calculated as 
 a function of the fire size. For the tested brands of transformer 
 fluid, the fire sizes are very similar. Therefore, it is expected 
 that the beta values reflect the same variability as the gas and 
 smoke concentrations. Table 8 lists the average production 
 constants for each brand. 
 
 Table 8.— Production constants for transformer fluid 
 
 Brand 
 
 0CO' /3c0 2 ' 0N O . 0M o i 
 
 10~ 3 g/kJ 1CT 2 g/kJ 10 1o p/kJ 10" 4 g/kJ 
 
 A 
 
 B 
 
 C 
 
 Average 
 
 3.43 
 
 7.57 
 
 3.59 
 
 9.28 
 
 2.51 
 
 7.64 
 
 2.80 
 
 1.87 
 
 3.24 
 
 7.59 
 
 1.51 
 
 12.54 
 
 3.06 
 
 7.60 
 
 2.63 
 
 7.89 
 
 COMPARISON OF TRANSFORMER FLUID BRANDS 
 
 Brand B generated a slightly higher C0 2 concentration 
 and lower CO concentration resulting in a larger fire size (Q A ) 
 and actual heat of combustion (H A ). These analyses were 
 based on relatively few experiments and may reflect only the 
 range of gas production. Considering this, the gas concentra- 
 tions generated by the tested brands of transformer fluid were 
 similar. 
 
 The differences between the brands are more evident in 
 their smoke characteristics. The most noticeable variations are 
 the 75 pet lower mass concentration of brand B and the 42 pet 
 lower number concentration of brand C. These low values are 
 also reflected in the reduced production constants. 
 
 Because the particle size can be calculated by two inde- 
 pendent methods, the diameters obtained by one method 
 
70 
 
 should confirm those obtained by the other. The calculations 
 indicate good agreement between the average d m and d 32 . The 
 low number and high mass concentrations for brand C resulted 
 in the largest calculated d m . By both methods, brand C 
 produced the largest particles. This is corroborated by the high 
 obscuration rate of brand C. 
 
 However, differences are apparent in comparing the par- 
 ticle sizes for brand B. The small mass concentration has 
 
 lowered the d m , while the d 32 approximates the average. Here 
 the agreement between the two methods is not very good. 
 
 The results of these experiments showed little variation 
 between the transformer fluid brands for CO and C0 2 pro- 
 duction, heat release, and heat of combustion. However, the 
 smoke characteristics data indicate that brand C produced the 
 heaviest and thickest smoke, while brand B generally produced 
 the lowest CO concentration. 
 
 COMPARISON OF EMISSION PRODUCTS OF FUELS 
 
 When the smoke characteristics from wood fires are 
 compared with the smoke characteristics of transformer fluid 
 fires, the data indicate that a wood fire must be five times 
 larger to produce the same obscuration rate as that obtained 
 for transformer fluid. Transformer fluid generates a denser, 
 more hazardous, but more detectable fire than wood. 
 
 When 84 pet of the light is obscured, reduced visibility 
 lessens the possibility of escape. Using a 25-cm-diam pan, a 
 46-pct obscuration rate was attained. This pool size is compa- 
 rable to a very small spill. By comparison, the wood burns 
 with very little light obscured. The particle sizes and obscura- 
 tion rates for wood and transformer fluid fires are given in 
 table 9. 
 
 In addition to producing denser smoke, transformer fluid 
 fires generate more toxic products than do comparably sized 
 wood fires. Table 10 gives the emissions for fires at the same 
 ventilation rates, when the transformer fluid fires equal the 
 intensity of wood. Although transformer fluid smoke is more 
 hazardous, its denser smoke may make it more detectable. A 
 small transformer fluid fire may be detected before a well- 
 
 developed wood fire. In addition, the speed with which liquid 
 pool fires propagate makes early detection essential if life and 
 property are to be protected. 
 
 Table 9.— Wood and transformer fluid fire particle 
 and obscuration rates 
 
 sizes 
 
 Fuel 
 
 d m , 
 
 d 32 , Obscuration 
 /xm rate, pet 
 
 Wood crib 
 
 0.22 
 
 ND 
 0.39 
 
 86 
 
 Transformer fluid 
 
 .34 
 
 46.8 
 
 ND Not determined. 
 
 Table 10.— 
 
 I/Vood and transformer fluid fire emissions 
 
 Fuel 
 
 CO, 
 
 ppm 
 
 co 2 , 
 
 ppm 
 
 No. 
 
 10 6 p/cm 3 
 
 M , 
 
 mg/m 3 
 
 Wood crib 
 
 145 
 
 6,759 
 4,992 
 
 6.04 
 2.79 
 
 49.1 
 
 Transformer fluid... 
 
 284 
 
 85.2 
 
 CONSIDERATIONS 
 
 In any mine fire, all the burning materials combine to 
 produce a wide variety of smoke particles and volatile gases 
 that are transported by the ventilating system. A comparison 
 
 of emission products is included as a means of demonstrating 
 their hazardous characteristics and to assess the potential 
 impact on the detectibility of the smoke produced. 
 
71 
 
 APPENDIX.— SYMBOLS USED IN THIS PAPER 
 
 u 32 
 
 G x 
 H A 
 H c 
 
 H co 
 
 K x 
 
 In 
 
 M f 
 
 conversion factor of a combustion product 
 
 diameter of a particle of average mass, ^m 
 
 mean particle size, /im 
 
 generation rate of a given combustion product, g/s 
 
 actual heat of combustion, kJ/g 
 
 net heat of combustion of the fuel, kJ/g 
 
 heat of combustion of CO, kJ/g 
 
 theoretical yield of a given gas, g/g 
 
 logarithm, natural 
 
 fuel mass loss rate, g/s 
 
 particle mass concentration, mg/cm 3 
 
 M x 
 
 N 
 
 P 
 
 Qa 
 T 
 
 AX 
 X 
 
 density of a given gas, g/(m 3 -ppm) 
 
 particle number concentration, p/cm 3 
 
 particle 
 
 actual heat release, kW 
 
 transmission of light 
 
 ventilation rate, m 3 /s 
 
 production constant of a given combustion product, 
 
 g/kJ or p/kJ 
 
 measured change in a given gas, ppm 
 
 wavelength, /xm 
 
 individual particle density, g/cm 3 
 
71' 
 
 UTILIZATION OF SMOKE PROPERTIES 
 FOR PREDICTING SMOKE TOXICITY 
 
 By Maria I. De Rosa 1 and Charles D. Litton 2 
 
 ABSTRACT 
 
 The Bureau of Mines has conducted two series of experiments to determine if a smoke 
 particle characteristic, such as the smoke particle of an average diameter, d g , average number 
 concentration, n , and their product, d g -n , could be correlated with the smoke relative 
 toxicity of smoldering combustibles. A previous set of experiments had shown that the smoke 
 parameter d g -n differed among combustibles. 
 
 In the first series of tests, the inverse of the smoke particle diameter-concentration 
 product, l/d g n , yielded during the combustion of various combustibles, was found to 
 correlate with smoke toxicity data found in the literature for similar materials. 
 
 In a second, more detailed series of experiments, tests were conducted using samples of 
 mine conveyor belts. For these belts, the main toxicity is HCl. The results of this second series 
 of tests showed that the smoke parameter, l/d g n load, inverse of total d g n values per gram 
 of sample weight loss, correlated directly with the HCl load, HCl concentration per gram of 
 sample weight loss. 
 
 INTRODUCTION 
 
 In 1984, the Bureau of Mines initiated a series of experi- 
 ments to study smoke particle characteristics, and to determine 
 whether they differ among combustibles. In these tests, a wide 
 range of combustibles including coal, wood, burlap, rubber, 
 and polyvinyl chloride (PVC) brattice materials were tested. 
 The results showed that the smoke particle of an average 
 diameter, d g , smoke particle of an average number concentra- 
 tion, n , and their product, d g -n , differed among combusti- 
 bles (1-2). 3 The question posed was: does d g n differ accord- 
 ing to the smoke relative toxicity of the combustion products? 
 
 In the first series of tests of this study, attempts have been 
 made to correlate the inverse of the smoke particle diameter- 
 concentration product, l/d g n , yielded during the combustion 
 of various materials, with relative toxicity data obtained 
 during the combustion of similar materials, by Alarie (3) 
 (animal toxicity data) and Paciorek (4) (chemical analyses), 
 and reported in the literature. 
 
 In the second series of tests, attempts have been made to 
 correlate the smoke particle parameter, l/d g n load (inverse of 
 
 ' Industrial hygienist. 
 
 2 Supervisory physical scientist. 
 Pittsburgh Research Center, Bureau of Mines, Pittsburgh, PA. 
 
 'Italic numbers in parentheses refer to items in the list of references at the end 
 of this paper. 
 
 the total d g n values per gram of sample weight loss) with 
 smoke relative toxic loads (toxic concentrations per gram of 
 sample weight loss) evolved during the combustion of PVC, 
 neoprene, and styrene-butadiene mine conveyor belts. 
 
 Hydrogen chloride (HCl — threshold limit value, 5 ppm; 
 short-term exposure limit, 25 ppm; immediately dangerous to 
 life, 100 ppm) was found by Paciorek (4), and confirmed by 
 the Bureau, to be the main toxic gas evolved at temperatures of 
 <350° C (early combustion conditions, under which miners 
 must plan and prepare for escape), during the combustion of 
 all types of mine conveyor belts. The HCl release begins early, 
 and proceeds rapidly at rates that depend on the chlorine 
 content of the material, and on the ease with which materials 
 undergo thermal decomposition (5). 
 
 Paciorek (4), and confirmed by the Bureau, also found 
 that PVC and neoprene belts released the highest concentra- 
 tions, and styrene-butadiene the lowest. Other major toxic 
 products, such as carbon monoxide (CO) was found (4) to 
 evolve at much higher temperatures. 
 
 Therefore, the study reported in this paper correlates the 
 inverse of the smoke particle diameter-concentration product, 
 l/d g n load, with the HCl loads yielded during the thermal 
 degradation of three types of mine conveyor belts. 
 
73 
 
 BACKGROUND 
 
 Conveyor belts are complex mixtures of a variety of 
 components. However, although the number of individual belt 
 components is very large, the number of major pure compo- 
 nents is comparatively small (6). The belting most widely used 
 in deep mines contain either a halogenated base polymer (PVC 
 and neoprene belts) or a halogenated additive (styrene- 
 butadiene belts) added to impart fire resistance (7). Because of 
 economic considerations, PVC resin probably is the most 
 widely used of the halogenated polymers. However, this poly- 
 mer releases HC1 at temperatures as low as 180° C, and the 
 reaction is accelerated in oxidizing atmospheres; for pure resin, 
 HC1 is the only product (90 pet) evolved at temperatures below 
 200° C (8). Because PVC is a resin, it requires large quantities 
 of plasticizers (phthalate esters) which co-evolve as phthalic 
 anhydride, CO, and unsaturated hydrocarbons (9). Neoprene, 
 because of its chlorine content, exhibits flame-retardant char- 
 
 acteristics; however, although considered thermally stable up 
 to 300° C, it has been found to degrade at 180° C; the onset 
 temperature for cured compositions could be as low as 125° C 
 (70). For neoprene, as for PVC, HC1 is the major degradation 
 product at 250° C; it accounts for 80 pet of the chlorine 
 present, with the rest liberated in the form of chlorinated 
 materials (11). Styrene-butadiene raw gum is flammable, and 
 requires large quantities of chlorinated additives, in addition 
 to fillers, plasticizers, and sulfur organic activators (12); its 
 final degradation products derive from all these constituents. 
 The major toxic component is again HC1, although in much 
 smaller quantities (13). Therefore, in the Bureau study, corre- 
 lations have been established between HC1 loads and the 
 inverse of the smoke particle diameter-concentration product, 
 l/d g n load, evolved during the thermal degradation of mine 
 conveyor belts. 
 
 OXIDATIVE THERMAL DEGRADATION AND SUBMICROMETER 
 PARTICLE DETECTOR-ANALYZER SYSTEMS 
 
 The degradation system (fig. 1) consists of a furnace (9 by 
 9 by 14 in) with a range of 100° to 1,200° C, set a priori at 
 250° C for this study, which allows the temperature to rise 
 automatically from ambient at a rate of 30° C/min to 40° 
 
 C/min (fig. 2). During the experiments, the temperature is 
 monitored continuously with type K thermocouples attached 
 to a strip-chart recorder. 
 
 Exhaust 
 
 Lindberg furnace, 
 9by9byl4in 
 
 X 
 
 Hood 
 
 Exhaust 
 Gas 
 sampling 
 port 
 
 Thermo- 
 couples 
 
 Valve Quartz tubing, 
 
 / 5 /£Hn-diam by 3~ft long 
 
 Temperature /\. 
 
 Gas 
 sampling 
 port 
 
 Quartz pedestal 
 sample 
 
 u u u 
 
 Sampling ports 
 
 control 
 
 Load cell 
 
 and 
 accessory 
 
 Bridge 
 amplifier 
 
 ( Strip-chart recorders \ 
 Weight loss Temperature 
 
 Figure 1 .—Oxidative thermal degradation and submicrometer particle detector-analyzer systems. 
 
74 
 
 
 350 
 
 
 300 
 
 o 
 
 
 
 
 250 
 
 «. 
 
 
 Q. 
 
 
 ^ 
 
 
 III 
 
 200 
 
 K 
 
 
 rr 
 
 
 hi 
 
 IbO 
 
 00 
 
 
 ^ 
 
 
 < 
 
 100 
 
 o 
 
 
 1 
 
 Set furnace temp, 
 250° C . 
 
 50- 
 
 1 r 
 
 J L 
 
 J L 
 
 J L 
 
 12 3456 789 10 
 TIME, min 
 
 Figure 2.— Furnace temperature versus time at 250° C. 
 
 A universal load cell, located under the furnace floor and 
 contacted by the furnace quartz sample-cup pedestal, trans- 
 mits the sample weight loss to another strip-chart recorder. A 
 pump draws ambient air continuously into the furnace (10 
 L/min) via an opening at the center of the furnace door, and 
 sends combustion air to the gas analyzers through a quartz 
 tube (40 by 1 in) inserted in the upper right rear of the furnace. 
 A flowmeter is installed between the pump outlet and the 
 infrared gas analyzers for continuous visual flow indication. 
 The furnace air is monitored continuously by two analyzers: 
 one for CO (0 to 100 ppm and to 1 ,000 ppm) and the other 
 for 2 (0 to 25 pet). 
 
 The submicrometer particle detector-analyzer system (fig. 
 1) consists of a strip-chart recorder and a submicrometer 
 particle detector-analyzer (SPDA) through which a small 
 quantity (1.6 L/min) of furnace airflow is directed. The SPDA 
 is a prototype instrument developed by the Bureau; its basic 
 operating principles are described elsewhere (14). The present 
 version has been modified so that real-time data can be 
 acquired for the simultaneous determination of the average 
 smoke particle diameter and the average smoke particle con- 
 centration, without need for the more time-consuming deter- 
 mination of the actual distribution of particle diameters. 
 
 DETERMINATION OF SMOKE RELATIVE TOXICITY AND CHARACTERIZATION 
 OF SMOKE PARTICLE DIAMETER-CONCENTRATION 
 
 In the first series of tests, the smoke toxicity data of 
 various combustibles is derived from studies performed by 
 Alarie (3) and by Paciorek (4). For Alarie, the toxic load of a 
 specific combustible (LTC50) is derived from the quantity of 
 material, and the time necessary to cause a 50-pct animal 
 mortality. For Paciorek, the yields of toxic gases are deter- 
 mined analytically in terms of quantity of toxic gas (toxic 
 loads, parts per million per gram) produced per mass of 
 sample consumed. The toxic load (TL) of a specific combus- 
 tible is the sum of each toxic gas yield per 1 g of weight loss 
 divided by its short-term exposure limit. To equate the data 
 from these independent tests, the average TL value for wood 
 reported by Alarie was normalized to the average TL for wood 
 reported by Paciorek according to A/(LTC50) = (TL) (A = 
 constant); for wood, A had a value of 4,600. All other LTC50 
 values were scaled to toxic load according to TL = 
 4,600/(LTC50). 
 
 The TL values derived from Alarie data, via the preceding 
 relationship, were in excellent agreement with the TL values 
 derived analytically by Paciorek. 
 
 The characterization of smoke particles during the com- 
 bustion of similar materials (table 1), was done by obtaining 
 the time-average value of the product d g n during the combus- 
 tion, at 250° C set-furnace temperature, of materials similar to 
 those used by Alarie and Paciorek. The inverse of this average 
 value, l/d g n , was correlated with the relative toxicity (TL) 
 obtained from references 3 and 4. 
 
 Table 1.— Materials investigated during the first series of tests 
 
 Material Description 
 
 PVC brattice PVC yellow plastic, reinforced with nylon fabric. 
 
 PVC resin PVC pellets. 
 
 Wood Pine wood, untreated. 
 
 Rubber Natural rubber mats. 
 
 Cotton-polyester 65-pct-cotton, 35-pct-polyester fabric. 
 
 In the second series of tests, the smoke relative toxicity 
 (HC1) and the smoke particle diameter-concentration product, 
 d g n , evolved during the combustion of mine conveyor belts 
 (table 2) were derived in the following manner: sets of exper- 
 iments (eight experiments in the set; each experiment repeated 
 
 Table 2. — Materials investigated during the second series 
 of tests 
 
 Conveyor belt _ . .. „. , 
 
 / . i Description CU, pet 
 
 material r d r 
 
 PVC: 
 
 P1 Polymer component is PVC resin with fillers. 18 
 
 P2 do 16 
 
 Neoprene: 
 
 N1 Polymer component is neoprene rubber with 20 
 
 fillers. 
 N2 do 13.2 
 
 Styrene- 
 butadiene: 
 
 S1 Polystyrene rubber, with chlorinated additives 2.3 
 
 for fire retardancy. Contains carbon black, 
 antiozonant, phosphorous, tackifying 
 resins. 
 
 S2 Polystyrene rubber, not treated for fire 1.25 
 
 retardancy. Contains carbon black, 
 antioxidant, antiozonant, phosphorous, 
 tackifying resins. 
 
 S3 Polystyrene rubber, with chlorinated 7.25 
 
 additives. Contains carbon black, 
 antioxidant, antiozonant, phosphorous, 
 tackifying resins. 
 
 S4 Polystyrene rubber, with chlorinated additives 15 
 
 reinforced with nylon thread. Contains 
 carbon black, antioxidant, tackifying 
 resins. 
 
75 
 
 three times) were performed at furnace set-temperature of 250° 
 C for a duration of 14 min (these are conditions under which 
 most of the HC1 evolves, with no gross combustion or 
 oxidation of the sample) with 1 g sample of PVC (PI and P2), 
 neoprene (Nl and N2), and styrene-butadiene (SI, S2, S3, and 
 S4) conveyor belts. The variables measured as a function of 
 time were the HC1 concentrations (parts per million) from 
 which the toxic loads (HC1 concentrations measured at a 0.1 -g 
 sample weight loss (= 10th minute) per 1 g of weight loss, 
 parts per million per gram) were derived. 4 Other variables 
 include the furnace temperature, the sample weight loss, and 
 the SPDA voltages; the ratio of the experimental and initial 
 SPDA voltage outputs (I e /I ) was use d to determine the 
 product, d g -n (particles per square centimeter), of the smoke 
 particle of an average diameter (d g , centimeters) and number 
 concentration (n , particles per cubic centimeter), from which 
 its inverse, l/d g n (square centimeters per particle), was 
 
 derived, following the relationship in equation 1 (see also 
 figure 3). 
 
 1.. 
 
 f = l/(Kd g n )(l-exp(-Kd g n )), 
 
 (1) 
 
 where K = charging constant (K = 0.012, cm 2 /p). 
 The d g n load (total d g n values per 1 g of weight loss, particles 
 per square centimeter per gram, was used to derive its inverse, 
 l/d g n load, square centimeter per particle per gram). 
 
 After each experiment, the furnace was turned off and the 
 char was weighed. Measurements of HC1 concentrations were 
 made at 2-min intervals at the sampling port nearest the 
 furnace air exit using Draeger short-term exposure tubes. 
 Analyses were performed to determine the chlorine content 
 (percent) of each conveyor belt using the method of oxygen 
 bomb combustion-ion selective electrode. 
 
 RESULTS AND DISCUSSION 
 
 First Series of Tests; Set-Furnace Temperature of 
 250° C 
 
 From figure 4, it is apparent that the parameter l/d g n 
 correlates very well with the toxicity of the various combusti- 
 bles. For each combustible material, the symbols represent the 
 average of the data for l/d g n at the average value of the toxic 
 load obtained from references 3 and 4. 
 
 The rectangles around each symbol, along the abscissa, 
 correspond to the total range of toxic loads derived from both 
 studies, and along the ordinate, to the range of values of 
 l/d g n obtained in the Bureau's series of tests for each type of 
 combustible; the symbols represent the average values. A 
 simple power curve fit to the data is defined by the solid line 
 and is given by the expression (also shown in figure 4) RTH = 
 2.303 x 10 5 (l/d g n ) 2 - 192 . The units of l/d g n represent some 
 measure of the effective cross section for ionization of the 
 different experimental smokes. This cross section also scales 
 with the relative toxicity hazard or RTH. 
 
 It could be argued that a high value of d g n signifies that 
 a smaller percentage of the consumed mass is available for 
 toxic gas production, while a lower value of d g n signifies that 
 a higher percentage of the consumed mass is realized as toxic 
 gas, rather than smoke. Although this argument agrees qual- 
 itatively with the test results, no rigorous theoretical explana- 
 tion is offered at this time. 
 
 Second Series of Tests; Set-Furnace Temperature 
 of 250° C 
 
 In the second series of tests, the inverse of the smoke 
 particle diameter-concentration, l/d g n load, correlated di- 
 rectly and significantly (r = = 0.80) with the HC1 toxic loads 
 for all types and kinds of conveyor belts (fig. 5). As expected, 
 
 KEY 
 
 * Rubber 
 a Wood 
 
 • Polyester 
 
 a PVC brattice 
 o PVC resin 
 — 8est fit 
 RTH = 2.303" 
 
 J 12 192 
 
 RELATIVE TOXICITY, RTH- 
 
 4 Bureau-derived results are in agreement with experimental results obtained by 
 Paciorek (4) with a different methodology. 
 
 Figure 4.— Correlation of the inverse of smoke particle 
 diameter-concentration product, 1/d g n ol with toxicity data from 
 literature. 
 
 Figure 3.— Smoke particle diameter-concentration relation- 
 ship curve. 
 
 io- 
 
 5 1 0-3 
 
 I0-4 
 
 I0-5 
 
 
 I 
 
 I 
 
 
 D _ J -^ < 
 
 — 
 
 Set-furnace temp, 250°C 
 
 
 
 -Cr*. 
 
 ~ 
 
 A ^^* 
 
 O 
 
 I 
 
 • PI 
 ■ P2 
 
 DN1 
 
 KEY 
 
 ♦ N2 OS3 
 ASl <>S4 
 AS2 
 
 10 100 
 
 HC^ LOAD, ppm/g 
 
 1,000 
 
 Figure 5.— Correlation of the inverse of smoke particle diam- 
 eter-concentration product, 1/d g n Q load, with HCI toxic loads 
 during oxidative thermal degradation of conveyor belts, at 
 250° C 
 
76 
 
 PVC (PI and P2) and neoprene (Nl and N2) belts yielded the 
 highest HC1 loads (maximum loads: 1,000 and 500 ppm/g), 
 due to their high chlorine content, and the highest l/d g n load 
 (0.0023 cm 2 /(pg)) (table 3), due perhaps to the fact that the 
 majority of particles, being in the liquid state as HC1, are not 
 being registered by the SPDA. By contrast, most of the 
 styrene-butadience belts (SI, S2, and S3) yielded the lowest 
 
 HC1 loads (20, 30, and 60 ppm/g) and the lowest l/d g n loads 
 (0.000065 cm 2 /(pg)) (table 3), due to their low content of 
 chlorinated additives. Exceptions were observed for belt S4, 
 which yielded a much higher HC1 load (300 ppm/g) and a 
 higher l/d g n load (0.00071 cm 2 /(p-g)), due to a higher 
 percentage of chlorinated additives (table 3). 
 
 Table 3.— Oxidative thermal degradation data at 250° C 
 
 Time, min 
 
 Conveyor belt 10th 14th 
 material 1 
 
 HCI, WL, 1/d % n c , HCI, WL, 1/d % n , 
 
 ppm g cm /p ppm g cm /p 
 
 PVC: 
 
 P1, 18 pet Cl 2 220 0.2 0.016 370 0.4 0.0057 
 
 P2, 16 pet Cl 2 363 .5 .007 500 .8 .0019 
 
 Neoprene: 
 
 N1,20pctCI 2 162 .3 .01 315 .7 .0034 
 
 N2, 13.2 pet Cl 2 42 .1 .007 120 .36 .0012 
 
 Styrene-butadiene: 
 
 S1,2.3pctCI 2 .0013 2 .1 .00065 
 
 52, 1.25 pet Cl 2 10 .4 .0003 18 .8 .0007 
 
 53, 7.25 pet Cl 2 .0025 6 .1 .00096 
 
 54, 15pctCI 2 28 .1 .019 70 .35 .0025 
 
 WL Weight loss. 
 
 1 1-g sample 
 
 2 Derived from the 10th minute HCI concentrations per 1 gram of sample weight loss. 
 
 3 Inverse of d g n loads. Each d g n load is the sum of all d g n values per gram of weight loss 
 
 
 Derived loads 
 
 
 2 HCI, 
 ppm/g 
 
 3 1/d g n ol 
 cm 2 /(pg) 
 
 d g n , 
 p/(cm 2 -g) 
 
 1,000 
 
 0.0023 
 
 440 
 
 720 
 
 .0015 
 
 650 
 
 540 
 
 .0023 
 
 430 
 
 418 
 
 .0005 
 
 2,135 
 
 20 
 
 .000065 
 
 15,500 
 
 30 
 
 .000053 
 
 19,000 
 
 60 
 
 .000096 
 
 10,400 
 
 300 
 
 .00071 
 
 1,415 
 
 
 CONCLUSIONS 
 
 The results indicate that the smoke particle characteristic, 
 d g n , is indicative of smoke toxicity by discriminating not only 
 among materials whose main toxic loads are CO or HCI (first 
 series of tests), but also among materials whose HCI loads 
 varies from material to material (second series of tests). 
 According to the findings, the inverse of the smoke particle 
 diameter-concentration product, l/d g n 0> correlates directly 
 with the toxic loads of various natural and synthetic combus- 
 tibles derived through smoke chemical analyses and laboratory 
 
 animal exposure. It also correlates directly and significantly 
 with the main toxic loads (HCI) of three types and various 
 kinds of conveyor belts. Although more correlations are 
 needed for a variety of combustibles whose main toxic loads 
 are other than CO or HCI, the excellent correlation established 
 between l/d g n load with smoke toxic loads, the reliability of 
 the SPDA, and the simplicity of the methodologies suggest 
 their possible use as standards for determining the toxic hazard 
 of combustibles during fire. 
 
 
77 
 
 REFERENCES 
 
 1. De Rosa, M. I., and C. D. Litton. Oxidative Thermal Degrada- 
 tion of PVC-Derived, Fiberglass, Cotton, and Jute Brattices, and 
 Other Mine Materials. A Comparison of Toxic Gas and Liquid 
 Concentrations and Smoke-Particle Characterization. BuMines RI 
 9058, 1986, 14 pp. 
 
 2. . Determining the Relative Toxicity and Smoke Obscuration 
 
 of Combustion Products. Pres. at Symposium on Mining Rescue in 
 the Service of Mines Workmen, Bytom, Poland, Sept. 28-30, 1987, 10 
 pp.; available from M. I. De Rosa, BuMines, Pittsburgh, PA. 
 
 3. Alarie, Y., and R. C. Anderson. Toxicological Classification of 
 Thermal Decomposition Products of Synthetic and Natural Polymers. 
 Toxicol, and Appl. Pharmacol, v. 57, 1981, pp. 181-188. 
 
 4. Paciorek, K. L., R. H. Kratzer, J. Kaufman, and J. H. Nakahara. 
 Coal Mine Combustion Products — Identification and Analysis Proce- 
 dures and Summary (contract H0133004, Ultrasystems Inc.). BuMines 
 OFR 109-79, 1978, 140 pp.; NTIS PB 299 559. 
 
 5. De Rosa, M. I. Correlation of Combustion Products From 
 Smoldering Conveyor Belts. Pres. at AIHA Conference, Montreal, 
 Canada, June 2, 1987, 15 pp., available upon request from M. I. De 
 Rosa, BuMines, Pittsburgh, PA. 
 
 6. Hartstein, A. M., and D. R. Forshey. Coal Mine Combustion 
 Products: Identification and Analysis. BuMines RI 7872, 1974, 12 pp. 
 
 7. Paciorek, K. L., R. H. Kratzer, J. Kaufman, and J. H. Nakahara. 
 Coal Mine Combustion Products — Identification and Analysis. Ultra- 
 systems, Inc., Tech. Rep. SN8220-A2, 1974, pp.1-149. 
 
 8. Wagner, J. P. Survey of Toxic Species Evolved in the Pyrolysis of 
 Combustion of Polymers. Fire Res. Abs. and Rev., v. 7, 1973, pp. 
 1-23. 
 
 9. Barrow, C. S., H. Lucia, and Y. C. Alarie. A Comparison of the 
 Acute Inhalation Toxicity of Hydrogen Chloride Versus the Thermal 
 Decomposition Products of Polyvinyl Chloride. J. Combust. Toxicol., 
 v. 6, 1979, pp. 3-12. 
 
 10. Boettner, E. A. Combustion Products From Incineration of 
 Plastics. Div. Res. and Dev., Univ. MI, Final Rep. EPA Contract 
 032050, Feb. 1973, 200 pp. 
 
 11. Cullis, C. F, and M. Hirsehler. The Combustion of Organic 
 Polymers. Clavendom Press, Oxford, Int. Ser. of Monographs on 
 Chemistry, 1981, pp. 12-13. 
 
 12. Paciorek, K. L., R. H. Kratzer, J. H. Nakahara, and D. H. 
 Harris. Determination of Products of the Oxidative Thermal Degra- 
 dation of Variously Treated Woods and Mine Materials (contract 
 J0395008, Ultrasystems). BuMines OFR 4-82, 1980, 182 pp., NTIS 
 82-146275. 
 
 13. Levin, B. C, M. Paabo, J. L. Guermean, and F. E. Harrif. 
 Effects of Exposure to Single or Multiple Combinations of the 
 Predominant Toxic Gases and Low Oxygen Atmospheres Produced in 
 Fires. Fund, and Appl. Toxicol., v. 9, 1987, pp. 236-250. 
 
 14. Litton, C. D., L. Graybeal, and M. Hertzberg. A Submicrome- 
 ter Particle Detector and Size Analyzer. Rev. Sci. Instrum., v. 50, No. 
 7, July 1979, pp. 817-823. 
 
78 
 
 ELECTROMAGNETIC FIRE WARNING SYSTEM 
 FOR UNDERGROUND MINES 
 
 By Kenneth E. Hjelmstad 1 and William H. Pomroy 2 
 
 ABSTRACT 
 
 This Bureau of Mines paper describes a fire warning system for underground mines. The sys- 
 tem utilizes the transmission of an electromagnetic signal through mine rock to underground work- 
 ings where miners equipped with miniature radio-type receivers are made aware of the presence 
 of a mine fire. The microreceiver can be mounted within and powered by a cap lamp battery, or 
 mounted on and powered by batteries of mobile equipment. 
 
 Tests of the system at two underground metal mines indicate that the electromagnetic signal 
 is capable of penetrating over 762 m of rock with high attenuation characteristics. With the trans- 
 mitting antenna located either on the surface or underground, the electromagnetic fire warning 
 alarm system has the potential of serving as a fire warning system for many metal and nonmetal 
 mines, capable of alerting miners in remote parts of a mine to the existence of a mine fire. 
 
 INTRODUCTION 
 
 The principal safety hazard associated with underground mine 
 fires is the rapid spread of smoke and toxic fire gases throughout 
 the mine workings. Even miners who are quite far removed from 
 the fire itself (up to several miles) can be exposed to life-threaten- 
 ing concentrations of these combustion products within minutes. 
 This occurs because the mine's ventilation system, which contin- 
 uously supplies fresh air to the mine, can circulate the combustion 
 products from a mine fire with equal efficiency. In underground 
 fires the recommended course of action is to evacuate to the sur- 
 face, a refuge station, or other safe location as quickly as possible. 
 A fire warning system that is capable of alerting the miners quickly 
 would thus save precious time and help insure a successful 
 evacuation. 
 
 Evacuation of personnel from underground mine fires can re- 
 quire considerable time. A survey of 50 underground noncoal mines 
 shows an average evacuation time of 27 min. The range in evacua- 
 tion times was from 5 to 85 min and was strongly correlated with 
 depth of the shaft. In deeper mines, the length of time is greater 
 and therefore can exceed the rated capacity of the presently used 
 self-rescuer, which is 60 min operating time in a mine environ- 
 ment of 1 pet carbon monoxide. Any delay in warning miners of 
 the necessity of donning their self-rescuers can be disastrous (7)? 
 thus the benefit of an electromagnetic system capable of sending 
 a signal to miners in seconds of time is obvious. 
 
 DEFICIENCIES OF PRESENT MINE FIRE WARNING SYSTEMS 
 
 Given the need for rapid evacuation, it is clear that reliable 
 and timely fire warning systems are essential. In typical aboveground 
 occupancies (factories, apartment buildings, hospitals, commercial 
 buildings, etc.), conventional fire alarms such as bells, gongs, lights, 
 whistles, public address announcements, and even word-of-mouth 
 are sufficient. However, in underground mines, these methods are 
 generally not suitable, and are therefore seldom used. 
 
 Underground mines are characterized by workers who are 
 widely scattered over very large areas with little or no means of 
 
 'Geophysicist. 
 2 Group supervisor. 
 Twin Cities Research Center, Bureau of Mines. Minneapolis, MN. 
 
 communication between groups or individuals. Even a short sep- 
 aration between the worker and an audible or visual alarm would 
 render the alarm useless, especially if the worker was not in direct 
 line of sight of the alarm or the worker was using noisy equipment. 
 In many mines, working areas are completely isolated, without links 
 to any other part of the mine by telephone, power cable, compressed 
 air line, conveyors, rail, or any other continuous or semicontin- 
 uous conductor over which a warning signal could be transmitted. 
 Most mines are so large that the cost of installing a conventional 
 signaling system (bells, lights, etc., and the associated wiring) would 
 
 3 Italic numbers in parentheses refer to items in the list of references at the end of 
 this paper. 
 
79 
 
 be prohibitive. The principal disadvantages of the prior art of 
 underground fire warning systems are inherent slowness, vulner- 
 ability to damage, and limited mine coverage. 
 
 The most common fire warning system in hard-rock mines 
 utilizes the ventilation system to transport the fire warning signal. 
 Known as the stench system, it operates by releasing an odorifer- 
 ous chemical (the same chemical used to odorize natural gas) into 
 the mine's ventilation and/or compressed air streams. When the 
 miners detect the odor, they immediately begin to evacuate accord- 
 ing to a prearranged evacuation plan (2). The principal disadvan- 
 tages of the stench system are the time required for the odor to reach 
 the remotest work places, and the tendency for some parts of a mine 
 to be consistently missed completely (3). These problems are par- 
 ticularly acute in mines with openings having large cross-sectional 
 areas, and therefore, extremely slow ventilation velocity. Under 
 certain conditions, a fire can even generate its own ventilation forces 
 that are counter to the mine's ventilation and further slow down 
 or reverse the stench flow. 
 
 The deficiencies of the stench system are well known. How- 
 ever because of the lack of a superior alternative, it is still the most 
 commonly used system. Considerable research effort has been 
 directed toward the development of wireless, radio frequency sig- 
 naling systems; however, each has inherent disadvantages that have 
 precluded their widespread use in mines. Ultrahigh frequency (UHF) 
 systems, because of their negligible through-the-rock transmission 
 capabilities, are limited to line-of-sight applications. Once a miner 
 travels around a corner such that the pocket pager type receiving 
 antenna is not in a direct line of sight with the transmitting antenna, 
 the wireless communication link is broken. In order to achieve 
 minewide coverage for the warning system, it becomes necessary 
 
 to install transmitting antennas throughout virtually the entire mine. 
 In large mines that might comprise several hundred miles of work- 
 ings, the cost of such an installation would be prohibitive. 
 
 A second radio frequency system operates in the medium fre- 
 quency (MF) spectrum (4). Although it too has limited through- 
 the-rock transmission capabilities, it has an advantage over UHF 
 systems because specialized transmitting antennas are not required. 
 Transmission signals parasitically couple into any continuous or 
 semicontinuous metallic conductors present. Thus, a receiver need 
 only be within line of sight of any such conductor (power line, rail, 
 compressed air pipe, etc.) for the system to operate. The disad- 
 vantages of the MF system are that the receiving antenna is quite 
 large and cumbersome (worn like a vest with large batteries in the 
 pockets), and that many modern mines which utilize diesel -powered 
 mobile equipment do not have continuous or semicontinuous metal- 
 lic conductors installed throughout the mine. They may be present 
 in certain locations, but too many areas would be left unprotected, 
 and those miners working in remote parts of the mine may not be 
 made aware of the existence of a mine fire. 
 
 In summary, the use of stench in the ventilation system in high- 
 back, room-and-pillar mines where air movement is slow can result 
 in excessive time delay in sending a fire warning to underground 
 miners. A fire warning system using wire for warning signal trans- 
 mission can be disabled when rockfalls or explosions break the wire. 
 Conventional fire warning systems are usually expensive to install 
 in a mine; therefore, the mine company may only install them where 
 the majority of miners are working. Those miners working in remote 
 parts of the mine may not be close enough to a fire warning device 
 and as a result, would fail to be alerted to the existence of a mine fire. 
 
 ULTRALOW FREQUENCY (ULF) ELECTROMAGNETIC FIRE WARNING SYSTEM DESIGN 
 
 The electromagnetic fire warning system described in this paper 
 combines the transmission and reception of ULF electromagnetic 
 through-the-earth fields as a means of sending a fire warning sig- 
 nal to underground miners. The magnetic field generated about the 
 energized transmitting antenna together with its accompanying elec- 
 tric field is the means of signal transmission. The fields emanate 
 from the transmitting antenna in a somewhat spherical manner, 
 described by the following equation (5): 
 
 for reception of the through-the-earth signal. The small (15-in-long) 
 antenna allows for mobility of the receiver wearer and increases 
 the likelihood of survivability of the warning system, since it has 
 no long connecting wire that would be exposed to damage from 
 fire, rockfalls, or explosions (fig. 1). The transmitting unit is of 
 conventional design, without significant size limitation on antenna 
 configuration or transmitter power (fig. 2). 
 
 H = 
 
 INAfG], 
 
 2nZ 3 
 
 H 
 
 I 
 
 N 
 
 A 
 
 G 
 
 n 
 
 Z 
 
 magnetic field strength, A/m, 
 
 anntenna current, A, 
 
 number of turns, 
 
 area of the antenna loop, m 2 , 
 
 an attenuation constant, 
 
 the constant 3.14 (pi), 
 
 distance through the transmitting medium, m. 
 
 The transmitter consists of a small signal generator, a 1 ,000-W 
 audiofrequency range amplifier, and a transmitting antenna; all 
 located either at the surface or underground in the mine. If the mine 
 is very deep, the transmitter could be located at a midpoint in the 
 mine and the signal then transmitted in all directions. The trans- 
 mitting antenna used in tests of the system was made up of 10 turns 
 of No. 10 or No. 12 insulated copper wire formed into a 100-ft- 
 diam loop. (Line configuration antennas might also be used for 
 transmitting.) 
 
 The uniqueness of the system is the mobility of the receiving 
 unit, which utilizes a high-permeability wound ferrite core antenna 
 
 Figure 1.— Fire warning system receiver. 
 
80 
 
 Figure 2.— Fire warning system transmitter. 
 
 Tuning to resonance of both transmitter and receiver antenna 
 allows for maximizing transmitting antenna current (and power), 
 while maintaining a small-sized receiver antenna for the conven- 
 ience of the wearer of the receiver. Tuning of both antennas to a 
 common resonant frequency creates a system that discriminates 
 against electromagnetic noise, while still accepting the signal in the 
 warning frequency range. By choosing frequencies that avoid har- 
 monics of power frequency (60 Hz) and using pass band filters in 
 the receiver, electromagnetic noise effects can be eliminated. 
 
 The microcircuit receiver is responsive to frequencies from 300 
 Hz to 10 kHz. Input to the receiver is from a high-permeability 
 ferrite core wound with No. 30 enameled copper wire to form an 
 antenna that can be tuned to the carrier wave frequency of the 
 transmitting antenna, i.e., 630 or 1,950 Hz. Tuning is accomplished 
 by placing capacitance of appropriate size in series with the an- 
 tenna to achieve a resonant frequency of choice. The receiver is 
 powered by a small battery, therefore it is very portable and con- 
 venient to carry. 
 
 For use as a fire warning system for underground mines, the 
 receiver would in all probability be powered by the cap lamp bat- 
 
 tery usually worn by an underground miner, or by a vehicle bat- 
 tery if mounted on a vehicle. The use of the cap lamp battery as 
 a power source for the receiver insures that the receiver would 
 always have an adequate power supply since the cap lamp batteries 
 are recharged each 24-h period and checked daily. To insure that 
 the receiver functions properly, a routine check of it could be made 
 at the same time by exposing it to a ULF pulse emitted by a test 
 fixture. 
 
 The high-magnetic-permeability ferrite core receiving antenna 
 has exceptional magnetic flux gathering capabilities. This makes 
 it possible for the antenna to be very sensitive and capable of cap- 
 turing even the weakest electromagnetic signal. Through amplifica- 
 tion, the generated antenna voltage can be used to initiate a blink- 
 ing of the miner's cap lamp in a manner that is recognizable to the 
 miner as a fire warning signal. When the miner is certain of the 
 warning, he or she can acknowledge the signal by pushing a but- 
 ton on the battery to eliminate further blinking of the cap lamp. 
 
 Research has determined that the risk of ULF initiation of elec- 
 tric blasting caps is negligible. 
 
 
81 
 
 INITIAL FIELD TEST RESULTS 
 
 Initial field tests of the fire warning system were made at a 
 local Minneapolis sandstone mine which had 15 m of overburden 
 made up of sandstone, limestone, and glacial till. The transmitter 
 and a 30-m-diam six-turn loop transmitting antenna were located 
 on the surface and the receiver unit was underground. 
 
 For the initial tests, an 8-ohm speaker was used in conjunction 
 with the receiver to allow the operator to hear signal reception. The 
 first tests were made at low power levels (2 W). The results of these 
 tests are shown in figure 3. The area of reception is about 10 times 
 the area of the transmitting antenna. The signal was received through 
 about 15 m of overburden. The results of this first test were en- 
 couraging enough to justify additional field tests of the system. 
 
 A second series of tests was made in the Tower-Soudan 
 underground iron mine located in northern Minnesota. Unlike the 
 first test, the transmitting antenna was placed underground in a stope 
 above the 27th level drift, located 762 m from the main shaft. The 
 transmitting antenna was made up of 10 turns of No. 10 copper 
 wire formed into a 30-m-diam loop. The receiver (fig. 1) and 
 transmitter (fig. 2) were the same as used previously. 
 
 It is possible to establish the voltage generated in a ferrite core 
 receiving antenna by measuring across it leads. This was done while 
 the receiving unit was placed at the center of the transmitting loop 
 antenna. It was established that the ratio of the power in the receiving 
 antenna to the power in the transmitting antenna, in decibels, was 
 a constant value regardless of the transmitting power levels, pro- 
 vided distance between antennas was constant. 
 
 Receiving antenna power levels were then established at various 
 points throughout the mine at various transmission power levels. 
 The ratio of the power level in the receiving antenna to the power 
 level in the transmitting antenna, in decibel loss, is the basis for 
 establishing power loss at various distances. The results of these 
 measurements are shown in figure 4 and indicate that the through- 
 the-rock signal decays with distance in a manner similar to an in- 
 verse cubic function. However, computer analysis of the data pro- 
 duced a best fit for the data to be a hyperbolic function. The graph 
 indicates reception through rock to be in excess of 762 m, the max- 
 imum distance of signal reception from the transmitting antenna. 
 Maximum transmitter power level was 53.3 W. 
 
 Another in-mine test of the fire warning system was conducted 
 at the New Jersey Zinc Co. Sterling Mine near Ogdensburg, NJ. 
 A 10-turn 30-m-diam transmitting antenna was positioned at two 
 different locations on the surface. At one site, the antenna was posi- 
 tioned on overburden of water-saturated saprolite (a claylike mate- 
 rial). At the other site, the antenna was positioned on marble (fig. 
 5). For this test a 30-cm-diam, 500-turn loop antenna was used to 
 measure the magnetic field strength at depth. The field strength at 
 depth was compared with the established value of magnetic field 
 strength generated by the surface antenna. The ratio of the two 
 values of field strength permitted evaluation of attenuation. When 
 compared, it was determined that the marble had lower attenua- 
 tion characteristics than saprolite and was correspondingly more 
 transparent to electromagnetic waves. 
 
 Tests at the Sterling Mine and the two sites in Minnesota were 
 done at 2.000-Hz carrier wave frequency. If lower values of fre- 
 quency are used, the signal attenuation will be less. Skin depth of 
 the signal (the depth at which the signal loses 1/e or 36 pet of its 
 
 original value) is inversely proportional to both frequency and con- 
 ductivity as shown by the following equation. 
 
 Skin depth = 6 = (^—) Vi , 
 
 where wja is frequency and a is conductivity. 
 
 Areo of '///'777?7?f7^. 
 
 reception in — r&////////'''' ■ - 
 cove indicated jy////////////^ 
 by shaded 
 area 
 
 
 3 
 
 I 
 
 !te=y,£=J*=4il=Si3* 
 
 i\SZ^£^iS2= 
 
 Figure 3. — Map showing reception area of first field test. 
 
 14 
 
 
 1 1 1 
 
 12 
 
 
 - 
 
 10 
 
 -1 
 
 - 
 
 8 
 
 -\ 
 
 - 
 
 6 
 
 
 - 
 
 4 
 
 - \ 
 
 - 
 
 2 
 
 ov 
 
 - 
 
 
 1 1 1 1 1 1 u 
 
 100 200 300 400 500 600 
 
 DISTANCE FROM TRANSMITTER, m 
 
 700 
 
 800 
 
 Figure 4.— Graph of signal attenuation at various distances. 
 
yj 
 
 Figure 5.— Transmitting antenna on surface. 
 
 If n is assumed to be 4 rr times 10~ 7 H/M for a nonmagnetic 
 material, this equation can be reduced to the following: 
 
 503 3 
 Skin depth = — ■ 
 
 (fa)V2 
 
 Through further calculations the value of o can be established. 
 The two types of overburden at the Sterling mine had values of 
 
 conductivity greater than 10" moh/m, which is a value greater than 
 the value of conductivity for most mine rock. A high-conductivity 
 material retards penetration by a magnetic field, but since the signal 
 was still detectable on the 563-m level, the tests suggest that through- 
 the-rock signaling is a viable means of warning miners of the 
 presence of a mine fire, and that the system is likely to function 
 well in a great many mines. 
 
 CONCLUSIONS 
 
 Successful tests of the electromagnetic fire warning system have 
 been completed at two underground metal mines. The maximum 
 through-the-rock signal transmission distance measured was 765 
 m through rock with high attenuation characteristics. Because most 
 metal and nonmetal mines in the United States are less than 914 
 m deep, it is believed that the electromagnetic fire warning system 
 
 described has potential for widespread use in the domestic mining 
 industry. With improved equipment designs and proper transmit- 
 ting antenna placement, it is reasonable to assume that the signal 
 could be made to reach miners in remote parts of the mine at great 
 distance from the transmitting antenna and serve to alert them to 
 the existence of a mine fire. 
 
 REFERENCES 
 
 1. Ontario Provincial Government. Improving Ground Control and Mine 
 Rescue; The Report of the Provincial Inquiry Into Ground Control and 
 Emergency Preparedness in Ontario Mines. ISBNO-7729-1064-2, 1986, 
 108 pp. 
 
 2. Pomroy, W. H., and T. L. Muldoon. Improved Stench Fire Warning 
 for Underground Mines. BuMines IC 9016, 1985. 33 pp. 
 
 3. Muldoon. T. L.. T. Lewtas, and T. E. Gore. Upgrade Stench Fire 
 Warning System — System Development and Prototype Tests (contract 
 
 H0292002, Foster-Miller Assoc, Inc.). BuMines OFR 136-81, 1981, 142 
 pp.; NTIS PB 82-122128. 
 
 4. Stolarczyk, L. G. A Medium Frequency Wireless Communication 
 System for Underground Mines (contract H0308004, A.R.F. Products, Inc.). 
 BuMines OFR 115-85, 1984, 221 pp.; NTIS PB 86-134103. 
 
 5. Sacks, H. K. Trapped Miner Location and Communication System. 
 Ch. in Underground Mine Communication (In Four Parts). 4. Section-to- 
 Place Communication. BuMines IC 8745, 1977, pp. 31-43. 
 
 INT.-BU.OF MINES,PGH.,PA. 28819 
 
QC 
 
 LU 
 
 >- 
 
 o 
 
 LU 
 
 z 
 cc 
 
 o 
 
 Q. 
 
 Q. 
 
 o 
 
 -I 
 < 
 
 o 
 
 LU 
 
 89 
 

 
 
 ^ 
 
 
 
 
 ECKMAN 
 
 NDERY INC. 
 
 A OCT 89 
 
 N. MANCHESTER, 
 INDIANA 46962 
 
 
 /■-isu>* "</.^<° y-^^s 
 
 
 
 
 ^ AC? ^ 
 
 
 W 
 
 
 «** ^ ' • « • " a! 
 
 iON 
 
 
 ^'