o U.S. Department of Transportation Federal Railroad Administration Development of Safety Criteria for Evaluating Concrete Tie Track in the Northeast Corridor Office of Research and Development Washington, DC. 20590 Volume I Remedial Projects Assessment DOT/FRA/ORD-86/08.1 June 1986 Final Report This document is available to the U.S. public through the National Technical Information Service, Springfield, Virginia 22161. NOTICE The United States Government does not endorse products or manufacturers. Trade or manufacturers' names appear herein solely because they are con- sidered essential to the object of this report. NOTICE This document is disseminated under the sponsorship of the Department of Transportation in the interest of information exchange. The United States Govern- ment assumes no liability for its contents or use thereof. Technical Report Documentation Page 1. Report Nc FRA/ORD-86/08.1 2. Government Accession No. 3. Recipient's Catalog No. 4. Title and Subtitle 5. Report Date DEVELOPMENT OF SAFETY CRITERIA FOR EVALUATING CONCRETE TIE TRACK IN THE NORTHEAST CORRIDOR VOLUME 1 . REMEDIAL PROJECTS ASSESSMENT June 1986 6. Performing Organization Code 7 Author,) Donald r, Ahlbeck, James M. Tuten, * Jeffrey A. Hadden and Harold D. Harrison 8. Performing Orgoniiation Report No. 9. Performing Organization Name and Address Battel le's Columbus Laboratories 505 King Avenue, Columbus, Ohio 43201 Sa 1 ient Systems, Inc. , Worthington, Ohio 43085 10. Work Unit No. (TRAIS) 11. Contract or Grant No. DTFR53-83-C-00009 12. Sponsoring Agency Name ond Address U. S. Department of Transportation Federal Railroad Administration 400 7th Street, S.W., Washington, D.C. 20590 13. Type of Report ond Period Covered Final Report April 1983-December 1984 14. Sponsoring Agency Code RRS-31 15. Supplementary Notes 16. Abstract Minor problems with the performance of some track components have been noted in Northeast Corridor concrete-tie track. Performance-related events have been caused primarily by short-duration impact loads due to wheel tread or rail running surface roughness. To reduce these occurrences, remedial projects were initiated by Amtrak. These included the development of a wheel impact load detector (WILD) to identify specific impact load-producing wheelsets, field tests of more resil- ient tie pads, use of a different rail clip design, repair of engine burns, etc. This report provides an assessment of these remedial projects. Four test sites were chosen for detailed track walker surveys to define the condition of NEC concrete-tie track. Four surveys on six-month intervals were conducted over each of the sites, a total of 19 miles of track. Component event codes from the surveys were stored in a data management system that allowed anal- ysis of performance issues and correlation with rail anomalies (joints, battered welds, etc.). Little tendency toward clustering of events was found, and the track was generally in excellent condition. The WILD detector system was used both to develop wheel load statistics and to identify passing Amtrak wheelsets developing high impact loads. A wheel truing program was initiated by Amtrak, using the WILD as an inspection tool, that quick- ly eliminated extreme loads and reduced rough tread conditions. Measured loads and wheel profiles were used to develop and validate a vehicle/track dynamics model. Laboratory and in-track tests were conducted to investigate track compo- nent dynamic response to impact loads using a calibrated drop hammer. 17. Key Words Concrete tie track, wheel/rail impact loads, wheel tread profiles, track dynamic response, track performance 19. Security Clossif. (of this report) Unclassified 18. Distribution Statement Document available through National Technical Information Service 5285 Port Royal Road Springfield, VA 22161 20. Security Clossif. (of this page) Unclassified 21. No. of Pages 105 22. Pr Form DOT F 1700.7 (8-72) Reproduction of completed page authorized i/ii Digitized by the Internet Archive in 2012 with funding from University of Illinois Urbana-Champaign http://archive.org/details/developmentofsaf86081ahlb TABLE OF CONTENTS Page PREFACE viii 1.0 SUMMARY 1 2.0 BACKGROUND 6 3.0 TRACK CONDITION ASSESSMENT 8 3.1 Survey Data Processing 8 3.2 Results of Track Surveys 14 3.3 Summary of Track Condition Assessment 32 4.0 REMEDIAL PROJECTS ASSESSMENT 34 4.1 Wheel Impact Load Detector 34 4.1.1 Dynamic Wheel Load Characterization 34 4.1.2 Description of Detector 35 4.1.3 Detector System Development 42 4.1.4 Applications of the Detector ..... 43 4.2 Passenger Wheel Profile Conditions 46 4.2.1 Experiments on the Northeast Corridor 46 4.2.2 Computer Simulation Development 59 4.2.3 Energy Loss Considerations . . . 66 4.3 Rail Running Surface Profiles 70 4.3.1 Rail Profile Measurements 70 4.3.2 Impact Detector Site Tests 74 5.0 TRACK COMPONENT DYNAMIC PERFORMANCE 86 5.1 Track Structure Performance Experiments 88 5.1.1 Impact Simulation With Drop Hammer 88 5.1.2 Laboratory Simulation of Track Impact Dynamics . . 91 5.2 Results of Experiments 93 5.2.1 Tie/Tie Pad Performance 96 5.2.2 Rail Fastener Performance 98 5.2.3 Shoulder/Insert Performance 105 5.3 Conclusions 108 REFERENCES 109 1 1 1 LIST OF TABLES Table 3-1. Description of Detailed Track Component Survey Test Sites Table 3-2. Description of Track Component Survey Event Codes. . . II Table 3-3. Summary of Track Component Performance for Site 1 . . 17 Table 3-4. Summary of Track Component Performance for Site 2 . . ! Table 3-5. Summary of Track Component Performance for Site 3 . . Table 3-6. Summary of Track Component Performance for Site 5 . . 20 Table 3-7. Occurrences of Fastener Incidents in Vicinity of Weld or Rail Joint Anomalies--Sites 1, 2 and 3 . . . . Table 3-8. Occurrences of Fastener Incidents in Vicinity of Engine Burn Anomalies--Site 1 Table 3-9. Comparison of Fastener Safety Evaluation Criteria Site 1, Survey 1 Trackwalker Results 25 Table 3-10. Fastener Event Patterns in Test Sites--Fasteners Out on One Side of Rail Table 3-11. Fastener Event Patterns Within Test Sites- Fasteners Moving or Out on Both Sides of Rail .... Table 3-12. Insulator Events (Broken or Missing) in Test Sites— One Side of Rail 28 Table 4-1. Correlation of Wheel Impact Loads at Edgewood Detector Site With Specific Types of Freight Traffic-- Data From 4-2-84 to 4-30-84, 6-29-84 to 7-6-84 43 Table 4-2. Energy Dissipated by Typical Rough Wheel Profiles 69 Table 4-3. Comparison of Measured and Predicted Peak Vertical Loads Over Edgewood Rail Surface Anomaly 84 LIST OF ILLUSTRATIONS Figure 3-1. Examples of Track Survey Component Events 10 Figure 3-2. Example of Track Diagram for Track Walker Field Notes 12 Figure 3-3. Flow Chart for Detailed Track Survey Data Reduction and Management 13 Figure 3-4. Example of Track Survey Data Storage Format in Basis Data Management Program. . . 15 IV LIST OF ILLUSTRATIONS (Continued) Page Figure 3-5. Example of Fastener Fault Clustering on Rail Opposite a Rail Joint or Battered Weld 22 Figure 3-6. Example of Fastener Fault Clustering on Same Rail as Battered Engine Burn 23 Figure 3-7. Extensive Track Component Fault Pattern Noted in Survey of Test Site #5 30 Figure 4-1. Example Time Histories of Tie Bending Moments and Vertical' Wheel Loads Under Adjacent Trucks of Freight Cars on Concrete Tie Track 36 Figure 4-2. Cumulative Probability Curves of Static and Dynamic Vertical Wheel Loads on Northeast Corridor Concrete Tie Track (All Traffic) 37 Figure 4-3. Examples of Typical Impact Load Detector Circuit Outputs 39 Figure 4-4. Wheel Impact Load Detector Block Diagram 40 Figure 4-5. Three Examples of Wheel Exception Reports 41 Figure 4-6. Effect of Wheel Truing Program on Passenger Wheel Extreme Load Statistics 45 Figure 4-7. Examples of Wheel Vertical Load Statistics: Event Counts in Different Speed and Load Bands in Three Different Ballast Temperature Ranges 47 Figure 4-8. Example of Wheel Vertical Loads Through Impact Detector Site Under Amtrak Test Train 50 Figure 4-9. Comparison of Test Train Wheel Load Statistics With Revenue Freight and Passenger Traffic 52 Figure 4-10. Peak Vertical Wheel Load Measurements Versus Speed for Amcoach Wheels on Impact Test Train 53 Figure 4-11. Peak Vertical Wheel Load Measurements Versus Speed For Heritage Wheelsets on Impact Test Train 54 Figure 4-12. Measured Circumferential Wheel Profiles (Runout) of Amtrak Test Train Wheels 56 Figure 4-13. Peak Vertical Wheel Load Measurements Versus Speed for Heritage Car Wheelset, Axle 22, of Impact Test Train 58 Figure 4-14. Peak Vertical Wheel Load Measurements Versus Speed Under Freshly-Turned Heritable Car Wheels 60 Figure 4-15. Concrete Tie Track Response to Drop-Hammer Impact Load 63 Figure 4-16. Shapes of Concrete Tie Transverse Bending Modes ... 64 LIST OF ILLUSTRATIONS (Continued) Page Figure 4-17. Comparison of Wheel Impact Load Simulation With and Without Concrete Tie Bending Modes 65 Figure 4-18. Comparison of Predicted and Measured Load Tie- Histories for Heritage Car Wheel Tread Anomaly .... 67 Figure 4-19. Comparison of Predicted and Measured Peak Loads for Axle 19 of Impact Test Train Over Speed Range .... 68 Figure 4-20. Rail Longitudinal Running-Surface Profilometer .... 71 Figure 4-21. Measured' Rail Running-Surface Profile Near Engine Burn, M.P. 67.00, Site 1, Section 1, Tie 72, E. Rail 73 Figure 4-22. Measured Rail Running-Surface Profile Near Engine Burn, M.P. 67, Site 1, Section 4, Tie 8, W. Rail . . . 75 Figure 4-23. Measured Rail Running-Surface Profile Near Engine Burn, M.P. 66, Site 1, Section 1, Ties 121/122, W. Rail 76 Figure 4-24. Predicted Dynamic Vertical Wheel Load at Battered Engine Burn Under High-Speed Passenger Car 77 Figure 4-25. Measured Rail Running-Surface Profile at Engine Burn, Edgewood Impact Detector Circuit #3 79 Figure 4-26. Example of Loads From Impact Detector Site Under High-Speed Passenger Train 80 Figure 4-27. Comparison of Vertical Loads Over Smooth Track (Circuit #1) and Engine Burn (Circuit #2)— Amcoach Axle #11, Newly-Turned Wheel Profiles 81 Figure 4-28. Comparison of Vertical Loads Over Smooth Track (Circuit #1) and Engine Burn (Circuit #3)--Heritage Car Axle #17, Newly-Turned Wheels 82 Figure 4-29. Predicted Vertical Wheel Loads for Simulated High- Speed Passenger Car Over Measured Rail Surface Anomaly at Edgewood Impact Detector Circuit #3) ... 83 Figure 5-1. Relationships of Impact Loads Within the Vehicle and Track Structures 87 Figure 5-2. Sketch of Automated Drop Hammer 89 Figure 5-3. Comparison of Impact Load Time-Histories for Passenger Car Wheel and Drop Hammer 90 Figure 5-4. Influence of Wheel Vertical Preload on Rail -Seat Tie Bending Strain 92 Figure 5-5. Comparison of Tie Dynamic Bending Response Beneath Rail Seat, Nee Versus 5-Tie Track 94 Figure 5-6. Influence of Hammer Drop Height on Peak Tie Bending Moment Under Rail Seat for Different Test Conditions . 95 VI LIST OF ILLUSTRATIONS (Continued) Page Figure 5-7. Typical Flexural Cracks Found in Concrete Ties .... 97 Figure 5-8. Attenuation of Tie Bending Response to Impact Loading With Resilient Rail-Seat Tie Pad 99 Figure 5-9. Tie Bending Response to In-Track Drop Tests With Stiff EVA Pads and Resilient (DAYCO) Tie Pads 100 Figure 5-10. Ground Rod Vertical Acceleration Response to In- Track Drop Tests With Stiff EVA Tie Pads and Resilient (DAYCO) Tie Pads 101 Figure 5-11. Comparison of Clip (Toe) Vertical Displacements With Stiff EVA and Resilient Tie Pads 103 Figure 5-12. Comparison of Clip Longitudinal Displacement at Center Leg for Two Clip Designs 103 Figure 5-13. Comparison of Insert and Tie (Shoulder) Vertical Accelerations From In-Track Drop Hammer Tests (Stiff EVA Pads) 107 VI l PREFACE This report is a summary of work performed under Contract No. DTFR53-83-C-00009 by Battel le's Columbus Laboratories (BCL), and was sponsored by the U.S. Department of Transportation, Federal Railroad Administration. The program manager for this study was Mr. Howard 6. Moody of the Federal Railroad Administration, v 0ff ice of Research and Development. The program entitled "Development of Safety Criteria for Evaluation Concrete Tie Track in the Northeast Corridor" had as its basic objective the determination of the "safe capacity" of the track. "Safe capacity" is defined as the ability of the fastener system to retain the rail longitudinally and laterally, to prevent rail or track panel buckling and to maintain track gauge, and the ability of the ties to maintain track gauge and cross level within the limits of track geometry standards. Two aspects of this objective are addressed in the program: • Safe capacity with rail clip fasteners missing or in a weakened condition, • Safe capacity with damaged or weakened concrete ties. An additional program objective was to determine the failure modes of concrete ties and the causes of clip fallout and insert failures. This program consisted of five tasks: (1) Track load/deflection characterization study and track condition assessment, (2) Remedial projects assessment, (3) Clip performance, (4) Tie integrity, and (5) Track strength characterization. This report provides a comprehensive evaluation of Amtrak's remedial programs, primarily the work conducted under the first three tasks of this study. Substantial support was provided by the National Railroad Passenger Corporation (Amtrak) during the test phases of the study on the Northeast Corridor track and during the development of the wheel impact load detector system. Technical liaison was provided by Messrs. Dennis Wilcox and Dan Jerman on behalf of Mr. D. F. Sullivan of Amtrak. Mr. Bud Coffey of De Leuw, Cather/Parsons carefully inspected almost 50,000 ties (200,000 fasteners) four VI n times over the period of the program to provide the track walker survey data for this study. Technical support for the tests conducted by the Association of American Railroads Transportation Test Center (AAR/TTC) at Pueblo, Colorado was provided by Messrs. G. W. Walker, Larry Daniels, Dave Read and others. Coordination with the NECIP was provided by Mr. Ted Ferragut of the FRA Office of Passenger and Freight Services. Battel le's research program was managed by Mr. Harold D. Harrison, who shared in the development of the wheel impact load detector with Mr. James M. Tuten, Principal Research Scientist at BCL. Field and laboratory tests were conducted by James Tuten and Jeffrey Hadden, while computer simulation studies were conducted by Donald Ahlbeck. Technical assistance was provided by Ken Schueller, Gary Conkel, Chris Corogin and Mark Miller of BCL. IX 1.0 SUMMARY An important part of the Northeast Corridor Improvement Project was the installation of more than 400 track-miles of concrete tie track. Since installation, minor problems with track components have been discovered. These problems have included clip fallout, loose inserts, broken insulators and cracked ties. A cooperative investigation was conducted by Amtrak and DOT/FRA, contracting with Battel le's Columbus Laboratories. The primary cause of tie cracking was determined to be high vertical impact loads caused by a few rough wheel profiles on high-speed passenger cars. To improve track performance, remedial projects were initiated by Amtrak. These included the development of a wayside wheel impact load detector to identify specific impact-producing wheel sets. In addition, field tests of more resilient tie pads and a different rail clip design were begun. Research conducted under the program entitled "Development of Safety Criteria for Evaluating Concrete Tie Track in the Northeast Corridor" included an assessment of these remedial projects. At the start of this program, four test sites were chosen for detailed track walker surveys at six-month intervals to determine the extent of component problems. Data from the surveys were analyzed, using an interactive computer data management program. Correlations of component performance with rail surface anomalies (rail joints, engine burns, battered welds), as well as correlations of one performance issue with another, were examined. In particular, the clustering of events (clip fallout, etc.) that could weaken the track was analyzed in detail. Results from the track walker surveys showed that only a very small percentage of components exhibited degraded performance. Track in the four representative test sites was in generally excellent condition. Correlation of clip movement and fallout with rail surface roughness was strongest in the test site with older relay rail, where 32 percent of these events occurred within five ties of a rail anomaly. In the other test sites, less than 11 percent of the clip events occurred within five ties of an anomaly. This confirmed the belief that vibration induced by wheel (as well as rail) roughness must be an important factor in the clip fallout mechanism. Clustering of clip fallout, insert or insulator events did not appear to be a significant problem. Only one location in the rougher-rail site had four clips missing in a row on one side of the rail, and three locations had three clips missing in a row. Very few insert failures were noted in the survey data. Because of the particular NEC fastener design, the rail is still supported laterally by the insert when clips and Insulators are missing. Even a loosened insert (loss of bond) will provide some lateral restraint. Increasing numbers of temporary mechanical rail joints were noted over the course of the study. This may induce greater numbers of insert failures and clip movement. Rewelding the rail at these locations is recommended to reduce the number of running surface anomalies producing impact loads on the rail. Under separate contract with Amtrak, a wayside wheel impact load detector (WILD) system was designed, fabricated and installed by Battel le on the Northeast Corridor near Edgewood, MD. The detector measures four short samples of vertical loads under passing wheels, using strain gage patterns installed on the rail web. A microcomputer system calculates the peak load in each sample for each passing wheel. A brief load exception report, citing specific wheelsets exceeding preset load limits, is prepared by the computer and transmitted to one or more remote terminal printers. These reports include time and date, train speed and direction. The detector responds by commercial telephone line to specific commands to provide optional reports, which include the matrix of all loads (each wheel, each measurement site) for the last train by the detector, and accumulated wheel load statistics as axle counts in speed and peak-load bands. Amtrak first used the WILD system to identify specific wheelsets causing high impact loads. Several of these wheelsets were removed from revenue service and assembled into a test train, which also included a few freshly-turned wheelsets as a control. Tests were then performed over a range of speeds from 25 to 110 mph, and load data were recorded from the detector transducers. Wheelsets were then removed, and circumferential profiles were measured in the Amtrak wheel shop. Wheel treads were examined for damage corresponding to the load-producing profile errors, and in some cases these wheels would pass visual inspection without condemning-limit exceptions, other than minor tread spalling. Some wheelsets showed evidence of incipient bearing failure, such as loss of grease. Test results were used to define the impact load versus speed relationships for the different profile errors. In general, peak load increased with speed, tending to level off at 80 to 90 mph; but high loads were recorded at lower speeds, so that a slow-order to control the loads due to these wheelsets would not be practical. Measured load time-histories and wheel profiles were used to develop and validate a vehicle/track simulation model for predicting loads and dynamic vehicle/track response under different conditions. A wheel truing program was initiated in early 1984 by Amtrak with the installation of a new truing machine in the Ivy City yard. Using the WILD as a maintenance tool, specific wheelsets producing impact loads greater than 60,000 lb were identified and pulled from service. The effect of this program was soon evident in the four-week average of extreme-load statistics collected by the WILD. The percent of total events (load measurements for passing passenger wheelsets) exceeding 60,000 lb fell from 0.14 at the end of February to 0.04 by the end of May. This result was attained by shopping the wheels on 38 cars (about 7 percent of the fleet); and the current statistical level represents about a dozen wheelsets in the car fleet passing the WILD site. With the success of this program, Amtrak has lowered its exception load limit to 55,000 lb. Freight traffic, in the meantime, has remained at a nearly- constant level of 0.14 percent exceeding 60,000 lb. This represents an estimated 1.5 to 2 percent of the passing wheelsets. An estimated 0.5 percent of passing wheelsets generate loads exceeding 75,000 lb, the cracking threshold of the concrete ties. At current traffic levels, roughly 100 wheelsets a week passing the WILD site can initiate a crack. Profiles of rail running surface anomalies (particularly engine burns) were measured on the Northeast Corridor, using a specially-designed profilometer to simulate a 36-inch diameter Amcoach wheel. These profiles were then used with the vehicle/track dynamics model to predict impact loads. Some of the engine burn profiles were found to resemble battered low joints, with a depth of up to 0.09 inch over a 5-foot wavelength. Peak loads on these profiles tended to increase with train speed, while loads over smaller engine burns tended to be independent of speed. In-track and laboratory tests were conducted with an instrumented drop hammer to define the dynamic response mechanism involved in track component failures. The transverse vibrational bending modes of concrete ties, particularly the second and third modes at about 330 and 630 Hz, respectively, were found to be important in the tie cracking phenomenon. Because of the phase relationship of response peaks of these two modes at the rail seat, cracks could be initiated at either the top or bottom surfaces of the tie. In addition, the tie end opposite the impact load (opposite a rail joint, for example) could experience higher bending stresses near the rail seat than at the impacted end. The resilient tie pads recommended from laboratory studies [3] act as a low-pass mechanical filter, attenuating the impact load energy into the tie exciting these higher-frequency vibration modes. These resilient pads therefore reduce the probability of crack initiation under high impact loading. Rail clip movement leading to fallout was not observed either in the field measurements or laboratory tests. However, the current 601A clip exhibited lightly-damped oscillations in the longitudinal direction (along the rail) at about 1000 Hz (a frequency close to the tie fourth bending mode) in response to impact loads. The "e" clip did not show this oscillatory behavior, and to date there have been few (if any) fallout occurrences reported with those "e" clips installed on the NEC. Since the 601A clip is sensitive, particularly, to tie vibrations, the same fallout phenomena would not be expected to occur on more highly damped wood ties. Measurements of 601A clip vertical deflections with stiff EVA pads and the resilient Dayco pads showed about the same peak-to-peak deflection under impact load. In the clip-spreading (rebound) direction, however, smaller deflections were measured with the resilient pads. This implies lower clip peak stresses, rather than higher, with the resilient pads. Similar trends, but with somewhat smaller deflections were noted with the 'e" clip. Tests were performed to explore the insert bond-failure occurrences observed in the NEC track. A total of about 43,000 hammer drops were made on the laboratory track panel, ranging from 52 to 124 percent of the tie cracking threshold, but insert bond failure was not induced. Tests on individual inserts showed that strong compression-wave oscillations in the longitudinal direction (vertical to the tie) can be set up in the 4 to 8 kHz range in response to impact loading. This may cause a microscopic pulverizing action at the insert/tie interface, eventually resulting in loss cf bond. Resilient tie pads are expected to reduce this compression-wave response by reducing tie vibration levels. 2.0 BACKGROUND A significant part of the Northeast Corridor Improvement Project (NECIP) was the extensive use of concrete tie track. Concrete ties with resilient rail clips, stiff pads and insulators were first installed in 1978, primarily in the Boston Division. This section of the corridor has very little freight traffic. Eight miles of concrete ties were installed during December 1978 in Track 4 near Aberdeen, Maryland, an area of relatively heavy freight traffic. Larger scale installations were begun during 1979 in other parts of the southern section of the corridor, also subjected to heavy freight traffic. In June 1980, rail seat flexural cracks from one to six inches in length were discovered during an inspection of concrete ties in Track 4 near Aberdeen, MD. These were generally "hairline" cracks, cracks that required a careful examination of the tie face in the crib to find. The cracks were considered important as a sign of a mechanism that could eventually lead to tie failure in bending at the rail seat. An investigation, sponsored by the Federal Railroad Administration (FRA) and Amtrak, was conducted by Battel le's Columbus Laboratories to iden- tify the cause of the tie crack development and to recommend corrective action to minimize further damage to the track. An analysis of measured vertical wheel loads and tie bending moments from the Northeast Corridor track identi- fied short-duration impact loads from a small percentage of high-speed passen- ger train wheels as the probable cause of the cracks [1,2]. A preliminary examination of these wheels showed a long-wavelength tread irregularity as well as spa! led wheel treads to be the source of high impact loads. Cracked ties were removed from the track and tested for strength using the Amtrak specification as a standard. In general, ties with cracks less than five inches in length showed no loss of static strength and no fail- ure in five million cycles of repeated load. Ties with cracks longer than five inches, although showing some loss in strength, were apparently still functioning adequately in the track. These ties may have reduced life. Other types of tie cracks have also occurred at some locations in the Northeast corridor. Flexural cracks at the tie center and cracks at the insert may also be linked to the impact load events that cause rail seat flexural cracks. Longitudinal and torsional cracks, as well as some insert damage, may be caused by derailments or maintenance equipment, both during construction and normal maintenance. These types of cracks have resulted in several ties being removed from the track. In addition to tie cracking, rail clip fallouts were noted in many locations on the corridor. These fallout locations were at first associated with a rail joint, engine burn or other anomaly on the rail running surface. However, a more random occurrence of clip fallouts was noted that could not be characterized by an obvious rail surface anomaly. In a few instances, the cast iron insert had become loose, with a loss of bond between the insert stem and the tie. The clips could no longer restrain the rail in these cases, and the insert shoulder usually does not provide lateral restraint of the rail when in this loosened condition. Two corrective measures were proposed. The first was to install a more resilient tie pad to attenuate the peak loads between rail and tie. The second was to identify and remove those wheels causing the large loads. Laboratory and field experiments were conducted to study the effects of tie pad stiffness on reducing the impact loading at the tie [3,4]. A fully- automatic wheel impact load detector was designed for Amtrak [5], and was installed on one high-speed track near Edgewood, MD. This detector has been used both as a research tool to develop wheel load statistics under different types of traffic, and by Amtrak as an inspection tool to screen passing wheelsets for profile roughness producing high impact loads. 3.0 TRACK CONDITION ASSESSMENT To establish a statistically valid basis for track condition assess- ment, detailed surveys of fastener and tie conditions were conducted within test zones on the Northeast Corridor concrete tie track. As a result of an Amtrak/FRA inspection in April 1982, five tests sites of four to six miles in length were chosen. Four detailed surveys at approximately six-month inter- vals were eventually made on four of the five sites. (The fifth site was dropped for lack of events and traffic.) Descriptions of the four sites, each containing between 10,475 and 15,450 ties, are given in Table 3-1. 3.1 Survey Data Processing Detailed surveys were conducted by Mr. Bud Coffey of De Leuw, Cather/Parsons, under contract to FRA, who entered specific fastener and tie conditions by an established code on prepared track diagrams. Some of the specific conditions entered in the track walker data base are sketched in Figure 3-1. Survey event codes are given in Table 3-2. Updated track diagrams were prepared by Battel le prior to each survey and bound in looseleaf report form to facilitate data entry under field conditions. An example track diagram is shown in Figure 3-2. In this diagram, a plain code (see Table 3-2) denoted a new activity, code preceded by an asterisk (*) denoted a repeat activity, and code preceded by a dollar sign ($) denoted previous activity (a clip once out, but reinstalled, for example). Codes were entered in one of the seven appropriate locations on a given tie, as shown in the sketch in Table 3-2. Field data from the four surveys were processed at Battelle by the data reduction and management system shown in Figure 3-3. The field data were first entered on a Hewlett Packard HP9845 desktop computer with a user- friendly program that provided prompts of required inputs and instant display of entered data before storage on disk. Data for each site for a given survey were then transferred to a VAX 11-780 mainframe computer and stored. Here a program used the stored data to generate updated track diagrams (see Figure 3- 2) on a line printer. Reduced-size copies of these diagrams were then used for subsequent survey field notes. 8 TABLE 3-1. DESCRIPTION OF DETAILED TRACK COMPONENT SURVEY TEST SITES Test Site 1 - Bush to Aberdeen, Track 2, MP 69-64 (northbound). Curves - MP 66.5 to 66.1, 0°32' LH; MP 65.3 to 64.6, 1°00' RH. Installed July 1980, about 5.4 MGT/yr. older relay rail. Test Site 2 - Davis to Northeast Track 4 (renumbered Track 3, June 1983), MP 44-48 (southbound). Curves - MP 46.8 to 47.3, 1°00' RH; MP 44.0 to 44.7, 0°14' RH. Installed August - September 1980, about 7.5 MGT/yr. Test Site 3 - Davis to Northeast, Track 3, MP 44 - 48 (southbound). Curves - (same as Site 2). Installed August - September 1980, about 16 MGT/yr. Renumbered Track 2, June 1983, switched direction to primarily northbound passenger, about 8 MGT/yr. Test Site 5 - Grundy to Morris, Track 2, MP 65 - 59 (northbound). Installed April 1982, about 3.4 MGT/yr (primarily passenger). New softer (Trelleborg) pads. Curves - MP 64.7 to 64.3, 0°43' LH; MP 62.0 to 61.3, 0°45' RH; MP 60.6 to 60.3, 0°25' RH. Tie Center Cracks Tie Insert Crack Tie Torsional Crack Tie Longitudinal Crack Field Weld Battered ngine Burn Mechanical Damage Fastener Out Or Moving Out -Insulator Missing or Broken FIGURE 3-1 . EXAMPLES OF TRACK SURVEY COMPONENT EVENTS 10 TABLE 3-2. DESCRIPTION OF TRACK COMPONENT SURVEY EVENT COOES Code Description CT1 CT2 CT3 CT4 DT EB EBB FI FM FN FO FW FWB FWC GC IB II IL IM IN J J . JI JIB JIC JM JT PB PD PM S SF SL SM T TB TE LOC Concrete tie center crack Concrete tie insert crack Concrete tie longitudinal crack Concrete tie torsional crack Damaged tie Engine burn Engine burn, break (1n rail) Fastener improperly installed Fastener moving (>£" inside insert) Fastener, new Fastener out Field weld Field weld, battered Field weld, crowned Grade crossing (removed) Insulator broken Insulator improperly installed Insulator loose Insulator missing Insulator, new Joint Joint with depth measured Joint, insulated Joint, insulated, battered Joint, insulated, crowned Joint, mechanical Joint, temporary Pad, broken Pad, displaced Pad missing Skewed tie Shoulder/insert fractured or broken Shoulder/insert loose Shoulder/insert missing Transition, wood to concrete ties Transition, beginning Transition, end Location 1 2 t Increasing Tie Numbers Left Rail Right Rail 11 ♦♦♦♦♦♦a** a ♦•♦♦♦♦a*» ♦ ♦ *a ♦ ♦ ♦ ♦ ♦ »•♦ «■♦♦•♦ ♦♦♦a»**»* ♦ ♦ a ♦- * a a ♦ ♦ a ♦ ♦ ♦ ♦♦♦"*•♦♦*♦♦♦♦*♦•»♦<►♦♦»■»•♦ * * * m ■> ♦ ♦♦♦♦♦♦ i ♦♦♦♦♦♦a******* a a ♦ ♦♦♦♦♦a^»***-** in a • o a • a • a a ■ ♦ ♦ ♦ ♦ ♦ ♦ ♦ ♦ ♦ ♦•♦♦♦♦ ♦ ♦ ♦ ♦ ♦ ♦ ♦ ♦ ♦ ♦ ♦ ♦ ♦ ♦ ♦♦♦♦♦♦ ♦♦♦♦♦♦♦ ♦♦♦♦♦♦♦ a •»•♦♦•» * ♦ ♦■ ♦ ♦ ■ a ♦ ♦ ♦ ♦ ♦»♦♦♦♦ i ♦♦♦♦♦♦♦ ♦ ♦♦♦♦♦•♦••- » ♦ ♦ ♦ ♦♦♦♦♦♦ ♦ ♦ ♦ ♦• a a ■ a m t a a • • a ax i • a ah. 1 a • •> • ♦♦♦♦♦*a****<-»**»a»'»*-»* • * a a • a a • a a a a a a • a z a o • t a » a*. • • a «• a • a a a a a a a at • a t • ♦♦♦♦♦♦•a*«-**-**»»*a«-*»**** a a t a at a at • a o a t t a u. a t a a • a at • a at a a a a t a a • • a at > a at • a at • a at to LU Q _l UJ LU o O o cc X CM I cc CD 12 TRACKWALKER SURVEY DATA E__ DATA ENTRY ON HP 9845 COMPUTER I DISK FILE I BULK DATA TRANSFER TO VAX 11-780 COMPUTER I REFORMATTING PROGRAM FILES, UPDATE PROGRAM I BASIS DATA MANAGEMENT PROGRAM I BASIS INTERACTIVE DATA CORRELATIONS I PROGRAM OUTPUTS REVISED TRACKWALKER DATA DIAGRAMS FIGURE 3-3. FLOW CHART FOR DETAILED TRACK SURVEY DATA REDUCTION AND MANAGEMENT 13 Data stored in the VAX computer were then processed by reformatting and update programs, transferring the data to Battelle's Data Management System, BASIS , which employs a powerful but easy-to-learn query language. The BASIS data base consists of random symbolic keyed files that interrelate via the software system to provide efficient retrieval, manipulation and analysis of the data. Correlation of variables can be performed rapidly with BASIS so that hypotheses on the interrelationships of causes and effects can be tested. An example of the data storage format in BASIS is shown in Figure 3-4 for three successive ties. Each tie has a unique sequence number by which its data are stored in BASIS, as well as a tie number within the site and section. 3.2 Results of Track Surveys The rail running-surface anomalies within the four test sites con- sisted of insulated joints (13 or 14 each site), temporary mechanical joints replacing bad welds or flaws, welds (battered or crowned), and engine burns. Over the course of the four surveys, increasing numbers of temporary rail joints were noted: Temporary Mechanical Joints Site Survey #1 Survey #4 1 2 18 2 2 13 3 2 8 5 8 30 The four test sites had roughly the same number of crowned welds, 40 to 64 each. Site 1, with older relay rail dated 1962, had by far the most engine burns (55) and battered welds (106). Site 5, by contrast, had the next highest number of engine burns (4) and battered welds (75) on rail dating from 1977 to 1981. Smaller numbers of battered welds were found in Site 2 (25) and Site 3 (36), both predominately with newer rail. 14 0. TIE SEQUENCE NR 2. SITE NR 3. HILEPOST NR 4. SECTION NR 5. TIE NR 12. DEFECT LOCATION- -1 13. DEFECT CODE-1 14. SURVEY NR-1 16. DEFECT LOCATION- 4. 17. DEFECT CODE-2 18. SURVEY NR-2 20. DEFECT LOCATION- ■3 21. DEFECT CODE-3 22. SURVEY NR-3 23. DEFECT QUANTITY- 3 9444 1 67.37 11 33 2 FO 1 3 FO 1 3 EB 1 .030 0. TIE SEQUENCE NR 2. SITE NR 3. HILEPOST NR 4. SECTION NR 5. TIE NR 12. DEFECT LOCATION-1 13. DEFECT CODE-1 14. SURVEY NR-1 16. DEFECT LOCATION-2 17. DEFECT CODE-2 18. SURVEY NR-2 9443 1 67 11 34 2 FO 1 3 FO 1 57 o 3 4, 5, 12, 13. 14, 16. 17. 18. TIE SEQUENCE SITE NR HILEPOST NR SECTION NR TIE NR NR DEFECT DEFECT SURVEY DEFECT DEFECT SURVEY LOCATION-1 CODE-1 NR-1 LOCATION-2 CODE-2 NR-2 9446 1 67 11 35 FO 1 2 FH 3 57 FIGURE 3-4. EXAMPLE OF TRACK SURVEY DATA STORAGE FORMAT IN BASIS DATA MANAGEMENT PROGRAM 15 Clip, insulator, pad and tie performance over the four surveys is summarized in Tables 3-3 through 3-6. These tables reflect the revised esti- mates of cumulative tonnage over the test sites which are noted in Table 3-1. In these tables, clip and insulator event numbers represent new occurrences, where maintenance (QC) has been performed between surveys in Sites 1, 2 and 3 to replace clips and broken insulators. Only a partial maintenance was performed within Site 5, between Surveys 2 and 3, so that some cumulative effect is included. Pad, shoulder/insert and tie faults are cumulative in number, since these were not corrected in maintenance. The BASIS data management system was used to explore some of the characteristics of track component faults, particularly clustering of events and correlations between rail surface anomalies (welds, joints, etc.) and component faults, or between one type of fault and another. Examples of this are given in Tables 3-7 and 3-8, where the numbers of fastener events (fasteners moving or out) associated with specific rail surface anomalies are shown. In the case of welds and joints, which occur predominantly on one rail only, not in pairs, there are significantly more fastener events on the oppo- site rail within + 1 tie of the impact load. An example of this is shown in Figure 3-5, where fasteners are moving on three ties beyond the joint on the opposite rail. This effect is gradually lost 3 to 5 ties away from the rail joint or battered weld. In the case of engine burns, which occur more often in pairs opposite one another, roughly the same number of fastener events were found on the same rail as on the opposite rail within + 1 tie. The example in Figure 3-6 shows, in fact, all the fastener action on the same rail as the engine burn. Note that a depth of 0.050 inch indicates some service bending of the rail and subsidence under impact loading at this point. Site 1 (with older 1962 relay rail) showed 32 percent of the fas- tener events occurring within + 5 ties of a rail running-surface anomaly. The other test sites, however, showed less than 11 percent of the fastener events occurring within + 5 ties of an anomaly. This indicates that clip movement and fallout is as much a function of track (particularly tie) vibrations due to passing rough wheels as it is a function of rail running surface roughness. 16 TABLE 3-3. SUMMARY OF TRACK COMPONENT PERFORMANCE FOR SITE 1 Survey Total Tonnage (MGT)* f — (a) Clip Performance % Per Clips Out % MGT Clips Moving % Per MGT 1 2 3 4 10.8 13.1 16.7 20.3 351 0.67 337 0.65 206 0.40 401 0.77 0.29 0.11 0.21 287 1213 1306 1759 0.55 2.33 2.51 3.38 1.01 0.70 0.97 Survey (b) Insulator and Pad Performance Pads Total Insulators % Displaced, Tonnage Broken Per Broken or (MGT) or Missing % MGT Missing % % Insulators Per Installed MGT Incorrectly 1 2 3 4 10.8 13.1 16.7 20.3 238 0.46 235 0.45 0.20 92 0.18 0.05 1 <0.01 <0.01 5 5 12 14 0.02 0.03 0.001 0.05 0.003 0.05 0.003 163 1 Survey Total Tonnage (MGT) (c) Tie Performance Shoulder/ Insert Loose, % Broken or Per Missing % MGT Cracked Ties Damaged Ties Skewed Ties 1 2 3 4 10.8 13.1 16.7 20.3 7 0.013 29 0.056 52 0.100 78 0.150 0.0012 0.0042 0.0060 0.0074 9 11 15 15 6 6 6 6 17 26 39 Total of 13,025 ties in site * Estimated cumulative since track installed. # Maintenance (QC) performed between surveys. 17 TABLE 3-4. SUMMARY OF TRACK COMPONENT PERFORMANCE FOR SITE 2 Survey Total Tonnage (MGT)* (a) Clip Performance % Per Clips Out % MGT Clips Moving % % Per MGT 1 2 3 4 14.1 17.8 23.1 28.4 26 0.06 37 0.09 173 0.41 152 0.36 0.02 0.08 0.07 206 311 467 643 0.49 0.74 1.11 1.53 0.20 0.21 0.29 Survey (b) Insulator and 1 Total Insulators % Tonnage Broken Per (MGT) or Missing % MGT 3 ad Performance Pads Displaced, Broken or Missing % % Insulators Per Installed MGT Incorrectly 1 2 3 4 14.1 17.8 23.1 28.4 40 0.10 59 0.14 0.04 18 0.04 0.01 9 32 38 0.04 0.002 0.15 0.007 0.18 0.006 3 Survey Total Tonnage (MGT) (c) Tie Performance Shoulder/ Insert Loose, % Broken or Per Missing % MGT Cracked Ties Damaged Ties Skewed Ties 1 2 3 4 14.1 17.8 23.1 28.4 8 0.019 12 0.029 14 0.033 20 0.047 0.0013 0.0016 0.0014 0.0017 10 15 16 18 5 8 8 8 2 57 77 81 Total of 10,475 ties in site * Estimated cumulative since track installed. # Maintenance (QC) performed between surveys. 18 TABLE 3-5. SUMMARY OF TRACK COMPONENT PERFORMANCE FOR SITE 3 (a) Clip Performance Total % % Tonnagi e Per Clips Per Survey (MGT)* Clips Out % MGT Moving % MGT 1 31.3 157 0.37 1462 3.44 2 40.7 127 0.30 0.03 703 1.65 0.18 3 47.0 147 0.35 0.05 1518 3.57 0.57 4 51.7 260 0.61 0.13 1300 3.06 0.65 (b) Insulator and ! 3 ad Performance Pads Total [nsulators % Displaced » % Insulators Tonnage (MGT) ( Broken Per Broken or Per In; stalled Survey )r Missing % MGT Missing % MGT Incorrectly 1 31.3 662 1.! 56 112 2 40.7 102 o.; 24 0.03 4 0.02 .0005 3 47.0 59 o.: L4 0.02 15 0.07 .002 4 51.7 21 0.10 .002 (c) Tie Performance Shoulder/ Total Insert Loose, % Tonnage (MGT) Broken or Per Cracked Damaged Skewed Survey Missing % MGT Ties T- ies Ties 1 31.3 7 0.016 0.0005 4 4 2 40.7 13 0.031 0.0008 6 4 23 3 47.0 20 0.047 0.0010 9 4 44 4 51.7 Total of 23 10,625 ties in 0.054 site 0.0010 9 4 55 * Estimated cumulative since track installed. # Maintenance (QC) performed between surveys. 19 TABLE 3-6. SUMMARY OF TRACK COMPONENT PERFORMANCE FOR SITE 5 (a) CI ip Performance Total % % Tonnage Per Clips Per Survey (MGT)* Clips Out % MGT Moving % MGT 1 2.6 399 0.65 994 1.61 2 4.5 535 0.87 -- 1150 1.86 -- 3 6.5 438 0.71 — 1484 2.40 -- 4 7.8 125 0.20 — 603 0.98 -- (b) Insulator # and Pad Performance Pads Total Insulators % Displaced * Insulators Tonnage Broken Per Broken or Per In: stalled Survey (MGT) or Missing % MGT Missing % MGT Incorrectly 1 2.6 888 1.44 132 0.43 0.16 15 2 4.5 286 0.46 — 427 1.38 0.31 3 6.5 133 0.22 -_ 575 1.86 0.29 4 7.8 — 605 1.96 0.25 (c) Ti ie Performance Shoulder/ Total Insert Loose, % Tonnage Broken or Per Cracked Damaged Skewed Survey (MGT) Missing % MGT Ties Ties Ties 2.6 4.5 6.5 7.8 2 3 6 13 0.003 0.005 0.010 0.021 0.0012 0.0011 0.0015 0.0027 5 8 9 10 61 72 90 94 Total of 15,450 ties in site * Estimated cumulative since track installed. # Maintenance (QC) performed between surveys 2 and 3 MP 59-62, only. Percent per MGT cannot be calculated. 20 TABLE 3-7. OCCURRENCES OF FASTENER INCIDENTS* IN VICINITY OF WELD OR RAIL JOINT ANOMALIES- SITES 1, 2 AND 3 Fasteners Moving or Out Welds, j oints Total Total Location No. Location (Rail) No. Ties No. Events Number of Occi jrrences Within (Rail) +0 Ties + 1 Tie +3 Ties +5 Ties Right 197 Right 4149 6827 18 120 381 515 Right 197 Left 4377 6721 74 216 422 555 Left 211 Left 4377 6721 11 111 419 582 Left 211 Right 4149 6827 68 175 395 534 * Fasteners out or moving. Note: Welds, joints at 400 ties, 410 unique values (only 10 occur opposite another at a given tie). TABLE 3-8. OCCURRENCES OF FASTENER INCIDENTS* IN VICINITY OF ENGINE BURN ANOMALIES— SITE 1 Burns No. Fasteners Movinq or Out Enqine Location Location (Rail) Total No. Ties Total No. Events Number of Occurrences W - ithin (Rail) +0 Ties +1 Tie +3 Ties + 5 Tie: Right Right Left Left 22 22 33 33 Right Left Left Right 1717 1917 1917 1717 2846 3013 3013 2846 37 27 46 40 86 117 83 121 103 149 89 113 145 163 177 155 * Fasteners out or moving. Note: Engine burns at 44 ties, 55 unique values (11 occur opposite another at a given tie). 21 DIRECTION OF TRAFFIC "■ 18 J (JOINT) 17 16 M 15 M 14 B. 25 B. B. #21068 #21064 SITE 2, SECTION 8, MP 47.62 SURVEY 4 FIGURE 3-5. EXAMPLE OF FASTENER FAULT CLUSTERING ON RAIL OPPOSITE A RAIL JOINT OR BATTERED WELD 22 37 DIRECTION OF TRAFFIC 9 ^ B. s 22 M S 36 35 34 EB-.050" (ENGINE 33 'FM* B. 32 BURN) IFMI 31 B E 30 29 #9445 #9440 ACTIVE IN LATER SURVEYS SITE 1, SECTION 11, MP 67.57 SURVEY 1 FIGURE 3-6. EXAMPLE OF FASTENER FAULT CLUSTERING ON SAME RAIL AS BATTERED ENGINE BURN 23 Several criteria for evaluating track strength from track walker survey data were examined in this study. Four different criteria can be used, based on fastener fault clustering: (1) Fasteners out, either side of the rail, (2) Fasteners out, one side of the rail only, (3) Fasteners moving or out, either side of the rail, (4) Fasteners moving or out, both sides of the rail. The numbers of locations of clusters, and the number of consecutive ties for a cluster are shown in Table 3-9 for Site 1, Survey 1 data. The most conservative is Criterion 3, which shows the largest number of missing or moving clips. From track strength considerations, however, Criterion 4 is more realistic in its definition, based on a pair of clips missing or moving at a given location. Fastener event patterns by Criterion 2, fasteners out on one side of the rail, are tabulated for the four sites in Table 3-10. Site 1 showed the highest incidence of fault clustering, probably due to the greater number of engine burns on the older relay rail. Similar results are seen in Table 3-11 for Criterion 4, fasteners moving or out on both sides of the rail. Sites 2 and 3 exhibited little evidence of fastener events in clusters, and generally lower percentages per accumulated tonnage (MGT of traffic). Some rail grinding was done within these sites during the two-year period, however. Site 5 had more random fastener movement and fallout, possibly due to instal- lation problems with the resilient pads. It has been noted, however, that rail within Site 5 was not properly straightened at welds, and these kinks may account for higher rail vibration levels. Fastener faults were evenly distri- buted on field and gauge sides of the rail except within Site 1, where roughly 35 percent more faults occurred on the gauge side of the rail. Insulator events (broken or missing) are given in Table 3-12, showing clusters on one side of the rail. Insulator failures occurred in significantly greater numbers on the field side of the rail, possibly because lateral impact loads from passing wheelsets are carried primarily by the field-side insulator into the shoulder. However, installation is more difficult on the field side, particularly in curves in the summer, and 24 TABLE 3-9. COMPARISON OF FASTENER SAFETY EVALUATION CRITERIA SITE 1, SURVEY 1 TRACKWALKER RESULTS Evalution Criteria Total Total Number Number Number Number of Consecutive Ties Fasteners Locations Ties Fasteners out, either side of rail (FO) - left rail 145 158 - right rail 176 193 13 1 Fasteners out, one side of rail (FO) - left rail, field side 86 86 - gage side 72 72 - right rail, field side 104 104 - gage side 89 Fasteners moving or out, either side of rail - left rail 308 (FM or FO) 89 326 18 6 2 right rail 292 312 22 ? i Fasteners moving or out, both sides of rai 1 - left rail - right rail 18 20 36 40 25 TABLE 3-10. FASTENER EVENT PATTERNS IN TEST SITES- FASTENERS OUT ON ONE SIDE OF RAIL Survey Side of Rail Left Rail Right R. lil Site Total Number Ties Number of Locations Number of Consecutive Ties Total Number Ties Number of Locations Numberof Consecutive Ties 1 1 Field 86 6 1 2 4 104 8 2 Gage 72 4 3 2 3 89 4 2 2 1 Field 11 -- 3 -- Gage 1 — 11 1 2 2 4 Field 45 — 39 2 2 Gage 45 2 2 23 3 1 Field 45 2 2 39 1 2 Gage 40 1 2 33 2 1 2 3 3 3 Field 65 2 1 2 3 30 1 2 Gage 29 1 2 23 1 2 5 2 Field 155 5 2 2 3 128 4 2 Gage 126 1 1 2 3 126 1 2 26 TABLE 3-11. FASTENER EVENT PATTERNS WITHIN TEST SITES —FASTENERS MOVING OR OUT ON BOTH SIDES OF RAIL Number of Ties Left Ra ,11 Riqht Rail Coi isecutive Consecutive Site Survey Left Rail Ri ght Rail Locations T" ies Locations Ties 1 1 18 20 1 3 2 2 2 32 16 1 2 -- 3 25 23 1 2 2 2 4 45 41 2 2 4 2 2 1 1 1 2 1 -- -- 3 5 2 -- -- 4 3 17 -- 1 3 3 1 13 18 2 2 5 -- -- 3 20 7 2 2 1 2 4 16 14 -- -- 5 1 26 5 1 2 2 39 10 2 1 2 3 -- 3 40 12 -- -- 4 7 4 — — 27 TABLE 3-12. INSULATOR EVENTS (BROKEN OR MISSING) IN TEST SITES-ONE SIDE OF RAIL Left Rail Right Rail Side Total Number Number of Total Number Numberof of Number of Consecutive Number of Consecutive Site Survey Rail Ties Locations Ties Ties Locations Ties 1 Field 85 103 Gage 22 31 1 Field Gage 23 11 Field 35 Gage 3 -- 8 -- 3 1 Field 142 2 1 2 3 254 12 2 2 3 Gage 99 1 2 167 4 2 3 3 Field 13 -- 35 2 2 Gage 3 -- 8 -- Field 158 83 Gage 29 15 28 failures may be installation-related. Field-side failures ranged from 1.7 to 3.5 times greater than gauge-side failures. Site 3 (with the highest accu- mulated tonnage) and Site 5 (with possible installation problems and improp- erly straightened rail) has the highest number of insulator failures at the time of Survey 1. These numbers dropped significantly in subsequent surveys as the result of maintenance. Few simultaneous occurrences of fastener and insulator faults were noted in the data base. Shoulder/insert failures were found to occur two to three times more often on the field side in Sites 1 and 5, but were evenly distributed on field and gauge sides in Sites 2 and 3. There were relatively few shoulder/insert events, with the highest numbers and percent per MGT found within Site 1, with the older relay rail. Only one cluster of three field-side shoulder/insert failures in a row was found. These three clips were, of course, not effective; but the cluster did not coincide with a more extensive clip fallout cluster. Site 5 (with the more resilient Trelleborg pads) has the highest incidence of pads moving, damaged or missing. Site 5 also had the highest number of skewed ties. One of the busier component fault arrays within Site 5 is shown in Figure 3-7 extending in either direction from a rail joint. The joint, 0.075 inch in depth, was the only apparent cause of all this activity; and the impact loads eventually loosened three of the four inserts in the immediate vicinity of the joint. Two of the skewed ties (47 and 48) were associated with missing clips on both field and gauge sides and broken field- side insulators, all on one rail. Two other ties (56 and 57) that were skewed by a later survey were associated with one clip out, three clips moving on one rail. It has been noted that ties will tend to skew toward the point of impact loading, as the ballast tends to loosen and migrate from under this location. However, no rail surface anomalies were noted at these two points. Comparable (and small) numbers of cracked or damaged ties were noted in all four of the test sites. However, the surveys did not include the careful examination of ties within the cribs necessary to detect the hairline rail seat cracks noted in previous studies. If the rail seat crack extended to the insert, it was counted then as an insert failure. 29 PRIMARY DIRECTION OF TRAFFIC ACTIVE IN LATER SURVEYS w ■ M (pd)pmJ 54 J (RAIL JOINT) FO IBIFO IPM\SLI 53 (FM) M IFOI IBIFO H 52 <3 51 CFMl ^2 -*n; (FOl 50 v/ — 49 © 48 (j) C p g) "CO 46 H #92175 #92170 #92167 SITE 5, SECTION 14, MP 63.74 SURVEY 2 FIGURE 3-7. EXTENSIVE TRACK COMPONENT FAULT PATTERN NOTED IN SURVEY OF TEST SITE #5 30 \, j s ACTIVE IN LATER SURVEYS ^ y PRIMARY DIRECTION OF TRAFFIC (IB, flBtFM? ^— i v ' V H 44H^ ^, *^*4*l^4l'^0•^JlA**m>4*>^• Vwa^vVvWii ■ wv|/p»«» " » " * <* w ^ ,w<, *V>«>« w »w>' 20 40 60 SO Time, m sec 1 00 I20 FIGURE 4-1. EXAMPLE TIME HISTORIES OF TIE BENDING MOMENTS AND VERTICAL WHEEL LOADS UNDER ADJACENT TRUCKS OF FREIGHT CARS ON CONCRETE TIE TRACK 36 STATIC VS. DYNAMIC EXCEEDANCE CURVES 99.9 Q LU Q LU UJ O X LU LU > LU z UJ O o c/> LU D 0- D O D O cc o 1— ra O. 1— rs o i— • — i ZD O Di ►— * C_> cc o i— o UJ LU t— %rr UJ Q *y Q <£. O £ - 1— o w §S5 U_ o 5IP.8 00 UJ 1 O LUS Ll. 3^ O- X . < UJ "lu ° . (/) "* — co o i «3- z UJ cc • 1 ■=> ** ^x zyt co 5 $ 9k ^ O to CD o a: o o O C_> «tf <* as LU CC SI O H- Z3 CD ^ S <° >; o tr -j u y < _i t <_><_| 1- Z 0. 40 WHEEL EXCEPTION REPORT Train Passed at 14:42:16 Speed = 121. Axle Count = 28 Ground Temperature = 83. Levels = 40. 55. kips Axle Number Car Number 84/06/02 Box Temperature =122 Level 1 Exceeded Level 2 Exceeded Yes 62. Train Passed at 14:47:09 84/06/02 Speed = 61. Axle Count = 12 Ground Temperature = 83. Box Temperature =126. Levels = 40. 55. kips Axle Car Level 1 Level 2 Number Number Exceeded Exceeded No Loads Above Limits Train Passed at 15:01:27 Speed = 48. Axle Count Ground Temperature = Levels = 40. 55. kips 84/06/02 272 87. Box Temperature = 126, Axle Number 4 10 90 96 167 168 190 193 194 Car Number 1 3 23 24 42 42 48 49 49 Level 1 Exceeded Yes 43. Yes 40. Yes 41. Yes 42. Yes 52. Yes 47. Yes 40. Yes 43. Level 2 Exceeded Yes 58, FIGURE 4-5. THREE EXAMPLES OF WHEEL EXCEPTION REPORTS 41 the system's integral modem to a commercial telephone junction box located in the Edgewood interlocking tower, about one mile away. Connections to the Amtrak signal system are made at a nearby signal cabinet. 4.1.3 Detector System Development The WILD was installed and calibrated in late Spring of 1983. During the initial verification phase, analog signals were processed both by the detector microcomputer and by manual processing of oscillographic recordings. This allowed the correction of several minor problems. For example, a program modification was made to the front-end processors to prevent false wheel triggering due to the 1200-Hz rail "ringing" caused by nearby wheel impacts. An example of this is shown in Figure 4-3. The current front-end algorithm can eliminate "false wheels" and can accommodate wheels "airborne" part way through the influence zone. Only an exceptionally bad wheel on a long freight train will occasionally confuse the axle count. Several other improvements were made to the system during this development phase. Modifications were made to the chronograph power supply to enable the chronograph to function during low voltage conditions. Some problems were noted when the line voltage dropped below 100 volts. The corrected system should function with line voltage as low as 85 volts. Internal heat exchanger fins were added to the enclosure after a late-June heat wave sent ambient air temperatures up to 100 F, and internal circuit board temperatures rose to 160 F. These internal heat fins reduced the board temperatures 10-15 F to within applicable limits. The wiring and connectors in the tie-mounted junction box were modified to provide better strain relief in this severe shock and vibration environment. This was necessary due to early fatigue failures in solder connections at the connector. Programming changes were made to the system to improve output formats. During August-September 1983, modifications were made to enable the detector to collect and store cumulative load statistics. Each peak load from each circuit for each wheel is treated as a separate event to be indexed and summed according to load and train speed ranges. Load statistics were addi- 42 tionally separated by temperature bands in this first phase: 20 F and under, 21 to 35 F, and over 35 F. Shortly after these changes were made, the system suffered electrical damage during a severe thunderstorm. The detector was removed and shipped to Battel le for repairs, and the following items were found damaged: (1) Three of four front-end microprocessors were damaged and required replacement. An additional level of protection was added to the A/D converter precision voltage reference for each microprocessor on the advice of the manufacturer. (2) The electronic buffer between the trackside relay circuit and the main microprocessor was shorted, probably from a voltage transient on the ground reference side of the relay sensing circuit. Additional current limiting protection was built into this circuit. (3) The modem surge protection circuitry had failed, indicating a major voltage transient on the communications line. Additional overvoltage protection was placed on the communications line. Other "growing pains" associated particularly with the modem and the communications line have been overcome, and the detector system has functioned with minimal problems for the past 6 to 8 months. 4.1.4 Applications of the Detector Amtrak Wheel Truing Program The main objective in developing the WILD was to detect, identify and facilitate the removal of wheel sets from Amtrak equipment causing track- damaging impact loads. Beginning in January 1984, wheels causing loads above 60,000 lb were identified from the wheel exception reports. Tower operators identified the specific train passing the detector, and car inspectors examined the particular wheels at the nearest terminal. Many of these load- producing wheels had no visual profile exceptions by current interchange rules. 43 Shortly after identification had begun, a program to turn tagged wheels was initiated, and a new wheel truing machine was installed at the Ivy City Yard (Washington, D.C.) Wheelshop. An immediate drop was noted in the probability of occurrence and worst-case magnitude of impact loads above the 60-kip maintenance threshold, as shown in Figure 4-6. In this figure, the results of a running four-week average of the statistics for both freight and passenger equipment are shown. A steady decline in the percent of total measurement events exceeding the 60-kip limit under passenger wheels was noted, bottoming at a level near 0.04 percent. A later rise in the curve can be attributed to a few cars with rougher wheels put into service for the Memorial Day weekend rush. It is evident from Figure 4-6 that the impact load statistics for freight traffic has remained about constant, since there is yet no impetus to improve this population of wheels. The relative importance of damage caused by freight traffic is now greater than before. An estimated two percent of all freight wheelsets cause impact loads greater than 60,000 lb, and 0.5 percent cause loads greater than 75,000 lb, the approximate crack-inducing load on a concrete tie. This is equivalent to 100 wheelsets per week through the detector site capable of producing rail seat cracks. Each week, several freight wheel loads are measured which exceed the 102-kip digital saturation limit of the WILD. The program for reducing wheel impact loads under passenger equip- ment has been an extraordinary success because of the accurate identification of these loads by the detector, and the diligence of Amtrak personnel involved in the maintenance program. Extreme loads (over 80,000 lb) under passenger equipment were virtually eliminated in about two months, and the exception limit has since been lowered from 60 to 55 kips, thus progressively improving the fleet wheel conditions. Wheel Load Statistics Programming modifications to the detector system allowed the accumu- lation of wheel load statistics as axle counts in speed and load ranges. In 44 I— I y- I *3- C3 45 addition, the load statistics could be stored separately according to the prevailing ballast temperature in three ranges: less than 20 F, 21 to 35 F, and over 35 F. Examples of load statistics printed at a remote terminal are shown in Figure 4-7. Cold temperatures during January 1984 provided an opportunity to evaluate whether frozen ballast had a significant effect on impact loads. Problems with the mix of traffic during warmer daytime hours and colder nighttime hours were noted. Daytime traffic could have been "salted" with recently reconditioned, dual-brake coaches in Metroliner service. In addition, initial work on older Heritage car wheels on the truing machine began to affect the load statistics as warmer weather approached. Evaluation of the freight traffic for mid-January (mostly freezing temperatures) and traffic for mid-February (above-freezing temperatures) showed nearly identical cumulative distributions of loads. The subsequent conclusion was that frozen ballast has only a small, and probably insignificant, influence on the magnitude of vertical wheel loads on the already-stiff (10,500 lb/in/in modulus) concrete tie track on the Northeast Corridor. Some investigations have been conducted by the NECIP staff to correlate the impact load data from the detector with specific types of freight traffic. Preliminary results of this work are shown in Table 4-1. These results indicate that wheel conditions on unit coal trains produce a greater share of impact load exceedances on a percent per population basis than other types of equipment in general freight traffic. 4.2 Passenger Wheel Profile Conditions 4.2.1 Experiments on the Northeast Corridor A series of tests was conducted on the Northeast Corridor in late November of 1983 using a special Amtrak test train. In these tests the train was run over a wide range of speeds through a wayside test zone, the Edgewood impact detector site. The objectives of these tests were twofold: (1) to determine the influence of speed on wheel/rail vertical loads produced by worn 46 DO YOU WANT CURRENT OR HfcCHlVHL STATISTICS T ENTER "I" FDR CURRENT* "2" FOR ARCHIVAL WHEEL IMPACT STATISTICS COMPLIED AT 15>01H4 84/01^30 Rfly STATISTICS FDR TEMPERATURE RHN6E 1 20 F Or leSS SPEED RAH6E »HPH LDAD»K1PS 0-50 50-65 66-80 81-95 96-110 > 1 1 ALL 0-10 8099 2063 4 9 36 10211 U-15 7714 2088 42 215 385 1116 11560 16-20 3241 660 50 604 720 2058 7333 21-25 2660 471 15 197 284 938 4565 26-30 2293 428 16 115 197 629 3678 31-35 2966 469 6 97 107 284 3929 36-40 970 154 1 *9 52 100 1306 41-45 178 24 1 27 19 69 318 46-50 57 10 10 15 £7 119 51-55 33 4 2 11 16 66 56-6 16 5 3 5 11 40 61-65 8 2 2 9 21 66-70 6 1 4 11 71-75 5 1 1 7 76-6U 3 1 1 5 >60 5 2 5298 7 FILL 28254 6378 132 13U8 1806 43176 PAW STATISTICS FDR TEMPERATURE RHM6E 2 21 to 35 F SPEED RANGE »HPH LDAL>KIPS 0-50 50-65 66-60 81-95 96-110 >110 ALL 0-10 25420 8394 11 46 279 34150 11-15 19101 8474 54 946 2148 6934 37657 16-20 6852 3288 202 2366 3548 8465 24741 21-25 7930 2977 64 759 1358 3234 16342 26-3U 12473 2123 79 452 1039 2426 18592 31-35 21212 3301 117 428 462 933 26473 36-40 5921 1002 45 139 197 423 7727 41-45 971 179 9 67 96 246 1568 46-50 318 64 1 24 52 109 568 51-55 153 24 15 31 68 291 56-60 83 17 4 14 24 142 61-65 68 8 1 4 9 19 109 66-70 33 9 2 2 4 C| 59 71-75 24 3 2 4 33 76-6 U 12 2 1 3 18 >80 37 5 1 1 6 23184 52 ALL 100608 2*d70 594 5dj8 9026 1 68522 RAW STATISTICS FOR TEMPERATURE RANGE 3 Greater than 35 F SPEED RAN6E t MPH LDAD.K1PS 0-50 50-65 66-80 81-95 96-110 > 1 1 ALL 0-10 5371 2434 i 49 7855 11-15 3676 1349 309 561 1679 7574 16-20 1037 135 741 827 1815 4555 21-25 1351 265 227 323 738 2904 26-30 2625 256 163 233 538 3815 31-35 4139 330 101 75 229 4874 36-4 1216 111 30 37 142 1536 41-45 181 11 19 14 59 284 46-50 36 8 14 9 21 66 51-55 23 3 2 4 ' 11 43 56-60 12 3 2 5 22 61-65 11 2 5 18 66-70 4 2 6 71-75 6 2 1 11 76-80 3 3 >60 7 1 8 ALL 19700 4908 16U8 2086 5292 33596 FIGURE 4-7. EXAMPLES OF WHEEL VERTICAL LOAD STATISTICS: EVENT COUNTS IN DIFFERENT SPEED AND LOAD BANDS IN THREE DIFFERENT BALLAST TEMPERATURE RANGES 47 "3- OO (_> I — u> U_ I «— • I — <_> uj o oo- zc co i— i •— • en 3 CM I LU VO OO «*" 00 a: i o o \— en o I LU «*• I— LU O Q -M o co O l 3 CM LU I C3 «* Q LU 21 o i— on oo <: Q t— < «=C O Q _l I t— I - <_> I— I 0k 9s im o <— a o i ** O m o in oo a. on O o 9s 9> WHO' O at o o CO oo at oo o o z oo <: CO a o 9* evrinom (N I — « i-4 ♦ _l o (— <_> «£ Q. z: av cs ea in onsmoo se 1 IP 1/1 >H —4 -« ■A ^ - - 9> <£ 00 »r» o * « o OQ Z a z i ^ « ^h O as o •M LU a. p« u_ o << a. 9\ 10 aom« i-i o >- 1 *r m -* o ac O -H OS tn s«ino I ««1 f» *H -H o C_> X in o ^ a m e in in so ao o u - co to CO CO M Z — z a - o « 2 5 z < < a: X E- X H c_> e- e z « k c_> HKZmK ^^ *■■* « U < Z 06 Lu Z S > 3 O Lu- Z Z > 3 O U_ 3 X ll. 3 X <» in to a o HNn^iflie CJ o 48 wheel profiles on passenger equipment, and (2) to correlate wheel tread conditions with the resulting impact loads. Other experiments were also conducted during this time period to characterize the track dynamic response to impact loads. The test train consisted of an AEM-7 electric locomotive, three Amcoach cars, and two "Heritage" cars (older passenger equipment from pre- Amtrak service). The cars were selected from revenue trains based on high impact loads developed by one or more wheelsets when passing the impact detector site. Several wheelsets newly cut to the standard AAR 1:20 taper were also included in the consist. These cars represent fundamentally different truck designs: The Amcoach with the Budd Pioneer III truck with its elastomeric primary suspension, and the Heritage car with its equalizer beam, coil spring primary suspension and swing-hanger supported secondary suspension. The track structure through the impact detector site is the current standard Northeast Corridor track: concrete ties on 24-inch centers, 140 lb/yd CWR with stiff EVA pads and Pandrol 601A clips. The track has a measured vertical stiffness under a nominal wheel load of 650,000 lb/in (tangent), or a track modulus of 10,500 lb/in/in per rail. During the tests, wheel load measurements were recorded from the impact detector circuits in both analog and peak-load (tabulated) formats. The standard rail web strain gage patterns used as vertical load transducers in four successive cribs at the detector site provided trapezoidal load influence zones, approximately 8 inches long at full amplitude. Therefore, the impact detector in this configuration provides four successive "snap- shots" of passing wheel load, and a 25 to 30 percent probability of capturing a particular impact load from a passing wheel. Repeated runs were therefore required at each test speed to increase the probability of measuring the true maximum under each wheel of the train. An example of vertical wheel loads under a given wheel passing the impact detector site is shown in Figure 4-8a. This wheel had a single 49 50 - 40 - a. 30 - 20 - 1 10 - — 55 KIPS (245 kN) TRAIN SPEED - 65 MPH (105 km/h) SITE #1 SITE #2 20 10 I I I I I I 10 20 30 40 50 60 TIME, MSEC a. VERTICAL LOADS UNDER AMCOACH AXLE #7 -20 CIRCUMFERENCE , Inches b. CIRCUMFERENTIAL PROFILE OF AMCOACH AXLE #7, RIGHT WHEEL FIGURE 4-8. EXAMPLE OF WHEEL VERTICAL LOADS THROUGH IMPACT DETECTOR SITE UNDER AMTRAK TEST TRAIN 50 distinct profile anomaly, as shown in Figure 4-8b, that impacted directly over Site #1 in this particular run. Loads within the influence zones of the other two sites show relatively little dynamic variation about the nominal 16,000-lb vertical wheel load. Vertical Wheel Load Versus Speed Vertical load statistics for the Amtrak test train are compared in Figure 4-9 with the load statistics for one week of revenue traffic, passenger and freight. The higher concentration of impact loads from the collection of rough wheels is evident in the resulting cumulative probability curves. For example, slightly more than two percent of the measured wheel/rail vertical loads exceeded 50 kips for the test train, while only 0.4 and 0.6 percent exceeded 50 kips for freight and passenger traffic, respectively. (Note that these statistics were gathered prior to the start of Amtrak ' s wheel truing program.) Although the test train had a higher than normal population of rough wheels, the impact detector measurement site was capable of detecting about 19 to 23 percent* of a particular wheel's circumference as it passed, or roughly a one-in-four chance of measuring the highest load under the worst spot on a wheel. The random nature of the peak measured loads for repeated runs in different speed bands is seen in the load-versus-speed plots of Figures 4-10 and 4-11. Load-versus-speed plots for the most severely worn wheel on the test train's Amcoach and Heritage cars are given in Figures 4-10 and 4-11, respectively. A sufficient number of repeated runs were made in each speed band to assure that representative load maxima were measured. In each figure two curves have been drawn, one a linear envelope of the largest loads measured at each speed, and the other a linear least-squares fit of the largest loads at each speed. The curves indicate that there is a measurable * Site 3 was "Inoperative" as explained in Section 4.3, Rail Running Surface Profiles. 51 Q oi O UJ UJ O X UJ -J UJ > UJ UJ O oc UJ a. 99.9 99.5 99 98 95 90 80- 70 60 50 40 30 20 10 5 2- 1- 0.50 0.10-1- 0.01 IMPACT TEST TRAIN \^ALL FREIGHT ALL PASSENGER J I L J I L 10 20 30 40 50 60 70 80 90 VERTICAL WHEEL LOAD, kip FIGURE 4-9. COMPARISON OF TEST TRAIN WHEEL LOAD STATISTICS WITH REVENUE FREIGHT AND PASSENGER TRAFFIC 52 PEAK LOADS FOR THREE WORST AMFLEET WHEELS too 80- Q. 2 ■o o o o 60- r 4 °- > 20 Linear Envelope of Largest Loads Mean Curve of Largest Loads at Each Speed oJ- -T- 20 — ?— 50 30 40 60 70 80 Speed, mph 90 100 110 120 FIGURE 4-10. PEAK VERTICAL WHEEL LOAD MEASUREMENTS VERSUS SPEED FOR AMCOACH WHEELS ON IMPACT TEST TRAIN 53 PEAK LOADS FOR FOUR WORST HERITAGE WHEELS too Speed ,mph FIGURE 4-11. PEAK VERTICAL WHEEL LOAD MEASUREMENTS VERSUS SPEED FOR HERITAGE WHEELSETS ON IMPACT TEST TRAIN 54 increase in load with increasing speed. Based on the least-squares curves, the increase in load for a factor of four speed increase (25 to 100 mph) is about 33 percent for Heritage wheelsets and about 60 percent for Amfleet wheelsets. These relatively modest increases in peak load suggest that it would probably be neither effective nor economical to place slow orders on trains with "bad" wheelsets as an alternative to removing the wheelsets from service, since those wheelsets may cause damaging loads even at lower speeds, as shown in these two figures. The differences between loads measured from the Heritage and Amfleet equipment are attributed mainly to the larger population of out-of-round wheels and longer-wavelength profile errors on the Heritage cars. These differences in wheel profile characteristics in turn may be caused by different mileages accumulated by the equipment, and possibly by differences in response due to truck suspension or brake characteristics. Measured Wheel Profiles Immediately following the tests, some of the wheelsets were removed from the cars for measurement of profiles at the Ivy City (Washington, D.C.) wheel shop. A special profilometer was designed to measure the changes in effective radius of wheel rotation around the circumference of the wheel. This profilometer consisted of a piece of rail head guided in the plane of rotation of the wheel, on the desired cant angle. The rail head was spring- loaded on a plunger to move in or out on the wheel radial line. The wheel was then rotated while cradled in is own bearings, and changes in the radial position of the rail head as it followed the contact patch were measured at the plunger with a dial indicator. These measurements provided a direct indication of tread vertical roughness or runout as "seen" by the contact patch. Examples from six of the test train wheels, two Amcoach (Axles 9 and 10) and four Heritage car (Axles 19-22), are shown in Figure 4-12. Vertical loads from several of theses wheel sets were evaluated to correlate wheel profile condition with impact load level. Example cases are discussed below: o Axle 9. The major anomaly on this Amcoach wheel set was a 1 x 1 inch spall indicated in Figure 4-12 by the 0.050 inch dip near 55 o CO o c£ o o UJ Cli o CIRCUMFERENCE . inches FIGURE 4-12. MEASURED CIRCUMFERENTIAL WHEEL PROFILES (RUNOUT) OF AMTRAK TEST TRAIN WHEELS 56 the 45-inch circumference location. The measured loads from this wheel showed peaks in the 65 to 75-kip range at speeds above 70 mph. © Axle 10. A significant characteristic of this Amcoach wheel set was that it was not condemnable by current AAR standards*. The principal anomaly on the wheel set was long, narrow chain of spalls which is shown in Figure 4-12 by the 0.035 inch runout near the 35 inch circumference location. The highest impact loads were measured under this wheel were 45 kips at 30 mph, 61 kips at 108 mph. © Axle 19 . This Heritage wheelset was characterized in Figure 4-12 by two distinct irregularities with runouts of 0.038 and 0.027 inch. (Profiles were not measured outside these two areas.) The load data for this wheelset showed a small speed effect, with peak impact loads of 60 to 75 kips over a wide speed range. © Axle 21. This Heritage car wheelset was another example of a non-condemnable profile which would pass the AAR criteria. As shown in Figure 4-12, small spalls (less than 0.030 inch runout) were present around the circumference. These irregularities were sufficient to cause peak impact loads of up to 70 kips. o Axle 22. This heritage car wheelset was the roughest on the test train. As seen in Figure 4-12, the wheel was visibly out-of- round, with a spread rim and spalls everywhere on the tread except at locations of maximum runout, where spalls were cold- rolled out. The load data plotted in Figure 4-13 for this wheelset indicate a possibly strong speed effect. Peak measured loads ranged from less than 40 kips at 30 mph to nearly 90 kips at over 100 mph. This case might be considered academic, since the wheelset was condemnable, even through not necessarily for the right reasons. Evidence of bearing grease loss and incipient bearing failure had made Amtrak mechanical personnel reluctant to use the wheelset, even on a test train. This case emphasizes the need to detect and remove such a wheelset quickly from the fleet to avoid potentially severe track and equipment damage. AAR Interchange Rule 41Alm which states that a wheel is "condemnable at any time" if the following conditions exist: "Out of round: in excess of 1/32 inch within an arc of 12 inches or less with use of gage as shown", or Rule 41A1 on Slid Flats: "A. Two inches or over in length, b. Two adjoining spots each 1^ inch or over in length. (Rule 41Alm has been dropped in recent editions of the AAR manual.) 57 100 PEAK LOADS FOR WORST WHEEL IN CONSIST y y& 80 Q. !5 o o 60 o o - 40 k. > .y... Linear Envelopes of Largest Loads y y y y y y Aa v v y A A A A V 47 20 V AXLE 22 11-30 A AXLE 22 12-01 T' 20 30 40 50 -T- 60 T" 70 -T— 60 -T- 90 100 110 Speed, mph FIGURE 4-13. PEAK VERTICAL WHEEL LOAD MEASUREMENTS VERSUS SPEED FOR HERITAGE CAR WHEELSET, AXLE 22, OF IMPACT TEST TRAIN 58 Two of the Heritage car wheelsets, Axles 17 and 18, were freshly turned just prior to the tests to provide a "control" case. As expected, the loads from these wheels were extremely consistent. Peak loads versus speed for these wheels are plotted in Figure 4-14. The loads are essentially constant with speed over the range of test speeds from 25 to 110 mph. A histogram, expanded in scale, is also shown in Figure 4-14. The mean load for these axles was 16.5 kips with a standard deviation of 0.9 kips, or 5 percent. When compared with the variation in loads under the worn wheels, the 5 percent variation is indeed small. 4.2.2 Computer Simulation Development The widespread interest in wheel /rail impact loads has led to the development and use of a wide variety of analytical models. These range in complexity from simple two-mass models to complex finite-element models of the track. Sato and Kosuge [7] have recently employed a simple lumped-parameter model consisting of the wheelset (unsprung) and rail effective masses to study rail head surface roughness on the high-speed Shinkasen line of the Japanese National Railways. Newton and Clark of British Railways, on the other hand, have used a much more complex hybrid model [8], which consists of a Discrete Support Model with a simple Euler beam to calculate the wheel/rail contact force, and then a Timoshenko beam model on elastic (Winkler) foundation to calculate rail strains in response to this force. In this discrete support model, a modal analysis is used to calculate the forced motion of the track, using the normal modes associated with the undamped track natural frequencies. A similar approach was employed by Mair [9,10] in his study of rail corrugation. Battelle's vertical wheel/rail impact load model was originally developed to explore the effects of rail joint and flat wheel geometries on wood-tie track structures [11]. This simplified lumped-parameter model consisted of two track masses (the effective rail and tie/ballast masses) and two vehicle masses (an upsprung wheelset and a sprung half-car body). The nonlinear contact stiffness at the wheel/rail interface described by Jenkins, et al [12) was used, and track mass, stiffness and damping parameters were 59 PEAK LOADS FOR FRESHLY TURNED WHEELS 100 .0- Q < Q 6 Ld > SPEED,mph FIGURE 4-14. PEAK VERTICAL WHEEL LOAD MEASUREMENTS VERSUS SPEED UNDER FRESHLY-TURNED HERITAGE CAR WHEELS 60 calculated from the traditional beam-on-elastic-foundation (BOEF) relation- ships. Model-predicted loads compared well with wheel loads measured with an instrumented wheelset on a 100-ton hopper car [13]. In this same study, impact loads were measured by rail strain-gage circuits under passing revenue trains. Although the actual wheel profile geometries were not known, the load magnitudes and time durations compared well with predictions from the model using assumed wheel flat shapes. Efforts to use this simple model to predict impact loads on concrete-tie track, however, were not successful. As predicted by Newton and Clark [8], the simple BOEF model tended to overestimate the peak impact loads. Additional degrees of freedom were added to the model, including the side frame/equalizer beam mass and mass-moment of inertia. Tie and ballast masses were separated and a nonlinear rail/tie (pad) stiffness was added, based on laboratory test results. This seven degree-of-freedom (DOF) model predicted impact loads that compared well with measured loads under a known wheel profile [4]. A more controlled synthesis and validation of the model was possible after the experiments with the Amtrak test train were completed. From these tests, measured wheel load time-histories under measured wheel profiles on a well-defined track structure were at last available to us. A specialized version of the computer program, called IMPWHL, was created to use the measured profile data. The measurements are introduced in tabular form, up to 120 points, on given increments. The program currently uses a simple linear interpolation between points to generate the wheel/rail vertical error position and velocity. Other mathematical methods for providing a smoother input function, such as the cubic spline, have been considered; but the results to date do not justify the use of these more complex algorithms. Initial computer runs with the measured profiles were compared with time-history traces of impact loads "captured" within the influence zones of the impact detector circuits. This provided a short time-history "snapshot" 61 of the passing wheel load -- roughly 8 milliseconds at full amplitude at 60 mph, only 4 milliseconds at 120 mph. If the initial impact occurred at the leading "skirt" of the circuit, the secondary load peaks could be observed. By repeated runs, a fairly complete picture of load response could be reconstructed. These first computer runs showed a strong oscillatory load response at 330 Hz which was not observed in the measured loads. This frequency is prominent in the track response to a drop-hammer impact load with no preload, as shown in Figure 4-15: it is associated with the second (asymmetrical) transverse bending mode of the concrete tie. Tests showed that this response peak is suppressed as the preloading wheelset is moved closer to the point of impact. Upon reflection, it occurred to us that this tie bending mode could be acting as a tuned absorber, and we decided to include the first four tie transverse bending modes in the model. Concrete tie bending modes were defined in laboratory tests on a similar tie, supported by pads under the rail seats. A modal analysis was performed on this tie using a Hewlett-Packard Model 5423A dual-channel analyzer by attaching an accelerometer to one corner and striking locations along the tie with an instrumented hammer. Results for the first three transverse bending modes are shown in Figure 4-16. (Two torsional bending modes at 365 and 406 Hz were also observed.) Based on an average bending rigidity from the first three measured modes, the fourth (asymmetrical) bending mode frequency was calculated to be 1033 Hz. Only the first bending mode of the tie appears to shift significantly in the track from the laboratory-measured value, increasing from 108 Hz to 154 Hz, as seen in Figure 4-15. Measured damping of these modes was small, roughly 0.5 percent of critical: as expected, the concrete tie literally "rings like a bell". A comparison of model -predicted forces for a portion of one of the measured wheel profiles of Figure 4-12, with and without the tie beam-bending modes, is shown in Figure 4-17. A rather dramatic difference in the force time-histories can be seen. An example comparing the measured load from a "direct hit" on a rail load circuit with the predicted load for the same wheel 62 22- INCH OROP, EVA PAD, NO PRELOAD 5.0 _ o IS) I— HA6 0.0 TIE BENDING STRAIN (RAIL SEAT) RAIL ACCELERATION "1 — 1 1 1 1 200 400 600 800 1000 FREQUENCY (HZ) — 2.0 s MAG 2 •— « ►- •x. t r r-o.o 1200 1400 1600 FIGURE 4-15. CONCRETE TIE TRACK RESPONSE TO DROP-HAMMER IMPACT LOAD 63 (A) FIRST BENDING MODE RESPONSE (B) SECOND BENDING MODE RESPONSE FREQ(HZ) £*" ~"~, ^ 633 \___^\ ^ > ^r^5Q (C) THIRD BENDING MODE RESPONSE SAMPLES OF FIRST THREE BENDING MODES FOR CC-244-C CONCRETE TIE FIGURE 4-16. SHAPES OF CONCRETE TIE TRANSVERSE BENDING MODES 64 T 1—1 1 1 1 I 1 1 1 1 1 1 1 1 1 I I • 9 10 11 12 13 14 IS IS 17 !• 19 20 21 22 23 24 25 DISTANCE ALONG RAIL, INCHES a. 7 DEGREE-OF-FREEDOM MODEL (WITHOUT TIE BENDING MODES) AXLE #19, 75 MPH -i — i 1 — i 1 i—i 1 r 7 9 9 10 11 12 13 14 IS 19 17 19 19 20 DISTANCE ALONG RAIL, INCHES b. 11 DEGREE-OF-FREEDOM MODEL (WITH TIE BENDING MODES) FIGURE 4-17. COMPARISON OF WHEEL IMPACT LOAD SIMULATION WITH AND WITHOUT CONCRETE TIE BENDING MODES 65 profile anomaly is shown in Figure 4-18. The predicted load shows an oscillatory load response beyond the peak load, which in the measured time- history is beyond the influence zone of the circuit. The predicted response also shows minor oscillations in the 800-1000 Hz frequency range. This may be due to the one-inch piecewise linear representation of the wheel profile. On the other hand, the rail itself exhibits a transverse bending mode near 800 Hz that is not specifically considered in the model. This mode is influenced by static wheel load and by adjacent wheels. However, it may act as a tuned absorber itself at these frequencies, effectively suppressing any oscillations in the 800 Hz region in the measurements. The predicted load versus speed curve for Axle 19 is plotted in Figure 4-19 along with the measured maximum loads for this wheel profile. Predicted peak loads were based on the larger of two profile anomalies, shown near the 20-inch circumference position in Figure 4-12. Correlation is quite good at speeds below 80 mph. Above 80 mph the predicted values are much higher than the measured values, which indicates that the peak values of impact load were not captured for Axle 19 within the rail circuit influence zones at these higher speeds. 4.2.3 Energy Loss Considerations One of the more interesting aspects of the study was the calculation of the energy dissipated by worn wheel profiles. Since a time-integration solution was used in the computer simulation program, it was a simple matter to calculate the damping forces and relative displacements for each time-step and sum these for the vehicle and the track, separately, over the total solution time. This typically ran 20 milliseconds, or 20 to 30 inches of the wheel circumference. Typical values are given in Table 4-2 for three of the wheel profiles of the test train. Just the larger of the two measured divots on Axle #19 is considered in this table. For a wheel rough around its whole circumference (such as Axle #20, and probably Axle #19), the energy consumed by wheel roughness can easily exceed 20 hp (15 kW) per wheel. In terms of the Davis equation for calculating drawbar resistance, Axle 19 would apply roughly 150 lb of drag at 74 mph, or 2 lb/ton per wheelset. This assumes that the 66 VERTICAL LOADS UNDER HERITAGE CAR AXLE #19 OF TEST TRAIN (74 MPH) a. MEASURED LOAD (RUN 11-30-24. SITE 1) BEYOND CL G < O < o 80 70 60 50 40 30 20 10 10 |_^ CIRCUIT INFLUENCE ZONE b. PREDICTED LOAD (PROGRAM IMPWHL) SOLID = WHEEL/RAIL DOTTED = RAIL/TIE -6-4-2 2 4 TIME (MILLISECONDS) 8 FIGURE 4-18. COMPARISON OF PREDICTED AND MEASURED LOAD TIME- HISTORIES FOR HERITAGE CAR WHEEL TREAD ANOMALY 67 VERTICAL LOAD ,Klpi 100 80- V A Measured Load (11/30/83) Measured Load (12/1/83) Predicted Load (IMPACT) 60- 40- 20- V v * A SO 80 i 20 40 50 60 70 SPEED ,mph 90 100 110 FIGURE 4-19. COMPARISON OF PREDICTED AND MEASURED PEAK LOADS FOR AXLE 19 OF IMPACT TEST TRAIN OVER SPEED RANGE 68 tn O UJ o >- >- CD < I/O or LU CM I CQ 4- o o_ 0) &i "*= 3 r— JX 0) O) .e 3 u QJ > ■■—I* O CL JZ CO VO CM vo S- > o a, a> o 3£ U Id s_ on 0) o O) o. t/o -C a. E X vo i— » • VO CO • 00 CM CM CM CVJ CO CM vo o VO CO CTi VO CO CM ro «3- «3- O VO r-l r*. Lf) CM CM CO *3- in «3" CM IT) vr> CM IT) O vo tn CO cri cr> o T3 a) E 3 (/> CO «J 0) o c (!) J- a> v»- • >>E Cvl f— 3 * — C O 1 O J- «* •r- (- U -a o c i_ >— ro i- CD a> cu ■X) 1 c-5 «tf 1— •»- **- 00 «4- O a> o S- 1- 1- ZJ Q. CU en ■o -i — t- c Li- CU T- CT> »D 0J I 1 ! ! J ▼ u_ o a: Q- < Ll_ ZD I CD a: _j C3 O o CNJ o <* i— o a: 71 be varied from +1:20 (the standard AAR taper) to a minus 1:20 (a hollow-work wheel tread). The actual measurement of the rail surface anomaly, which is centered in the wheel travel, is done by slowly moving the wheel segment from one end to the other. Vertical motions of the pivot point within its guides are translated by the DCDT into Y-axis displacement of the plotter pen with distance along the X-axis. This represents the forced displacement of the wheelset center of rotation by the geometry error, and is a direct input into the dynamic model without the need for geometric translation. The model then provides vertical dynamic deflections of the track in response to load variations. Some minor distortion of the wheel/rail geometry will occur, of course, due to changes in contact patch stresses under the high impact loads. Tests of the profilometer were conducted in the Aberdeen, MD area on April 19, 1984. Profiles of several of the more prominent engine burns within the Track Survey Site 1 test sections were measured. These engine burns generally occur in pairs (one each rail), sometimes under two wheelsets at the same time. Profiles at four different wheel tapers at a typical engine burn are shown in Figure 4-21. The engine burn itself is seen in the center of the profile plot as a sharp dip 0.005 to 0.010 inch in depth. What is striking, however, is the long wavelength depression of the rail on the order of 40 to 60 inches, apparently due to service-bending of the rail under dynamic loading. The engine burn begins to look very much like a dipped rail joint after some time under traffic. In Figure 4-21, the long wavelength profile assumes approximately the shape: e z = D e [1 - sin (f)], -^ < x < -| s where... D e = geometry error depth (under long straight-edge) L s = geometry error total span x = distance along rail. 72 1 i - 4— . 1 ! 1 1 I _1 a: o ■z. o _l «c UJ o «f 1— I/O ►— « o ! 1 ! 1 i ' i _. l I : ' i ,ac __i i I ! ! I • - - ■ 1 UJ a. t— 1 -1 i I , l ! ! 1 ! : ! i i o rlU O " «5t i— o — ! — j — r — -i — 1~ i ' ■ , i CM 1 1 J_ 1 | i i ■ ; 1 1 ! 1 + ■ ■+" 1 i i 1 ; 1 1 . ! 1 i '■■ \ \ \ ■ ■ \ 1 i i 1 —J V V 1 I i ! ! ■ 1 c£ i V ' ' 1 ! ' 1 i UJ ct i VJ I z ! ! 1 I \ i i \ ' ! 1 ! I »-H • 1 i 1 \ ' » i j j ! j 1 O UJ i | 1 i i i ' i ' 2^ UJ •> CNJ - 1 \ | \ | \ j]!:].: \ """ Qi r^ "! ^ i V \ f — — \ : i , ; -1 > ^- ^— — ■ -- , i ! ! ^s; \rr ; • /^ r \ — 1 i*^ t— u. ~ U. _"• 3NIia31N33 "133HM dO NOIlISOd !V3Iiy3A 73 The long wavelength geometry assumes a ramp angle of roughly 2 mi 1 1 iradians within the ten-inch span before and after the anomaly itself, which in itself is within established maintenance limits. For example, Australian National (14] uses a weld misalignment ramp angle limit of 7 mi 11 iradians to avoid rail seat pad damage; and British Rail considers a ramp angle of 10 mi 11 iradians as a maintenance exception level. Both field and gauge-side fasteners at the location in Figure 4-21 had fallen out, been replaced, and were again moving out of the inserts. An example of a more severely battered engine burn is shown in Figure 4-22. Here the ramp angle entering the damaged area (in the normal direction of traffic) is roughly 2 mrad, but the angle out of the engine burn is at least 7 mrad. At this location, track damage (other that the engine burn itself) included the gauge-side clip out, both inserts loosening with a visible field-side insert crack, and extensive spalling of the rail on the gauge side of the running surface. The most severe engine burn profile measured is shown in Figure 4- 23, with a total depth (to the bottom of the burn) of 0.090 inch. The wavelength of this disturbance was longer than the 50-inch span of the profilometer and was estimated to be 60 inches. In the direction of traffic, a 4 mrad ramp angle steepens to 11 mrad in the last few inches before the engine burn, with an included angle of about 24 mrad over a 5-inch span. Track damage included field and gauge-side clips missing, a broken insulator, and a loose insert. To estimate the dynamic vertical forces generated at this site, the profile of Figure 4-23 was used (in place of a wheel profile) in the impact model. Predicted load response is shown in Figure 4-24 for a simulated Amcoach wheelset at 120 mph. A peak load of 66,300 lb was predicted, dissipating 417 ft-lb into the track, 80 ft-lb into the vehicle suspension. 4.3.2 Impact Detector Site Tests Measurements were made of the Edgewood impact detector site "engine burn" (Circuit #3). This anomaly was first noticed (as an impact load- producing site) during October 1983. The profile shape was estimated at the 74 C3 z 00 UJ UJ _J •— • •-> h- u_ - cc «a- 0. z UJ —> et 1— U. O cc: uj => CO _l on t— 1 1 •> — t DC =3 * LO Z ct q: z: co CM CM I O 3NnH31N30 "133HM 30 NOUISOd IVDIlb^A 75 i i i - | ■ i _J t— 1 I j - i l cc i ' ! 1 , I + i i^L LU "3. Ll- " Z ' u_ i c % 1 i — i •> , : j ! i ! ! j : ' \ i | «r ! ! ' : •— < I— }•> \ ! 1 ! ^ s. Ll_ i i cc *— * _l- et ! ; ■ i , : • ! i 9 -c o - 1 ' ! j l ! ' ■' ! i : />s CC r— ! 1 i ! i ' 1 1 1 I ! D- — ! ; ! '■ ! i i —S^. i 1 i i ■ ! ! ! i ^i C_) •— • ; i ! 1 AWT o i &'- CC LU 1 = i ■ < id on : 1 . 1 : f i CD i — ; ::•'!: X 2T 1 - : i ; 1 : : . t— i LU i Z 1— ■ i ff 2: •— < i ■ : • i 1 i v > i i ! : r _J CD i A? I ' •—i vo ■ i i • J/ I i ■ ! ! ■-- - - - *I i l i ■ ff 1 ■ 1 i i C£ • i i i ! . : ! | 1 \/A i ' ! i ; O- Q • i ! i ! : ' i i 1 $ ' 1 ! ,!: : i i Ct ! i ! i 1 i 1 // ' ! I 1 ; rD •> i 1 j ! /7 | j 1 ii i co z i i i i 1 1 : l 1 1 _L ! :.... -*■ C\J 1 i F\ 1 i ! > i ; — ' 1 i ' «d- i ; i i 1 l ~_ 1 1 i ! ! ! ! lo o o O 1 LU : i i ...| i Hr~ -iil : CC | : i — -t 1 ! ~Jh ! ! 1 ZJ ! t ' : ; LJ_ i ": 1 ! 1 1 i l\ 1 \ 1 1 t_ T_ 1 1 i • I | : ! if 1 ! ' "H" i :-£-'v*: 1 i ^ r i _j: r~ ■ ! ! i — - . i ] ! -4 ' i •*•,"-■ 1 I i i 1 i ! _: 1 i 1 ! 1 ! 3NIiy31N3D "I33HM 30 NOUISOd lV3Iiy3A 76 LJ _J l-H z L_ cc O r> oe CD Q_ LJ a Z LJ ^^ tt: CD 3 Z to LJ CE LJ D n LJ cc * LJ _j (-" ^^ t- cc H* 1- LJ OC * o <-4 z * o o Z o o <— » LJ «-• z O LJ z to o * X «— 1 o « (E O LJ O f- to cc TL in a_ oo ji Q o -a: a ■ o «£ o _i o CM _l •-H _i i — i UJ —> Lf> _) o o Q UJ 1— (n «t _J 1— a: ce o O o t-> OO Q 001 Q UJ a: cd ce: <: i— o «t a: o UJ o z _l UJ oo _i on UJ «=C uj a. a: 3 Q «t D- o on •—i i l- n: q: o |_|J 1—4 > IE <_j or '— : UJ - Q Z ct: a id uj cc h- O UJ t-i z Q >-< UJ C3 ct: z Q_ UJ C\J I C3 T 06 t 1 1 1 1 1 r 08 OZ 09 OS Ot 0£ 02 ( sdi» 'auoi nuy/i33HM iuoUcGa 01- 77 time of the impact test train runs using a straight edge and feeler gage. Time histories of vertical loads from Circuit #3 were recorded during the test train runs. The running-surface profile at the Edgewood location measured in April 1984 is shown in Figure 4-25. The ramp angle of the long wavelength dip into the site was quite modest at this time: less than 1 milliradian. The "burn" itself (there is no evidence of distress on the opposite rail) caused the wheel to drop .015 to .018 inch in a span of 6 to 7 inches. A gradual increase in the impact factor under Amcoach wheels -- from 1.71 to 1.83 -- was noted over the first eight months, indicating that the dip was increasing slowly in size as it battered out. Recent checks on passing high-speed passenger trains have shown the impact factor under Amfleet cars to have stabilized at about 1.84. An example print-out from the impact detector is shown in Figure 4- 26 for an AEM-7 locomotive and six Amfleet cars at 118 mph. The increased peak loads under the rail anomaly are seen in the Processor 3 column. Note that a wheel anomaly on Axle 24 impacted at Site 4 with a 38.8-kip peak wheel load. Typical time histories of vertical load under the newly-turned wheels of the impact test train — Axle #11 on an Amcoach, Axle #17 on a Heritage car -- are shown in Figures 4-27 and 4-28. Modestly higher impact loads were recorded under the Heritage car wheelset, possibly due to the different effective wheel taper (and depth of anomaly) for the outside-bearing wheelset, versus the inside-bearing Amfleet wheelset. Predicted loads from the impact model using the measured profile from Figure 4-25 are plotted in Figure 4-29 for the three train speeds of Figure 4-28. Predicted response for the measured 1:20 taper profile matches more closely the response of the Heritage car wheel, as shown in Table 4-3. It must be remembered in comparing Figures 4-28 and 4-29 that the measured load is captured only within the 8-inch influence zone of the rail circuit. Secondary load response oscillations of the "continuous" predicted load are therefore only present in the 20-mph run, where the measured load stays within the circuit influence zone for two complete oscillations. The rail surface anomaly is seen in Figure 4-25 to be battered smoother in the normal direction of traffic, with a sharper dropoff, but easier rise in this direction. 78 en C3 Z o _J D. uj O O I— LO I C3 3NI"ia31N33 133HM JO NOUISOd IVOIiyBA 79 ENTER 0PT1QN VTRAIN PASSED AT 09i 33126 84'03'09 SPEED ■ 118. AXLE CDUhT - £8 GROUND TEMPERATURE 32. BOX TEMPERATURE - 87. AXLE CAR f >ROCESSOR NUMBER NUMBER 1 2 3 4 1 1 30.4 31.2 40.8 22.8 KIPS 2 1 27.6 24.0 37.2 24.8 KIPS 3 1 31.2 26.0 44.0 24.4 KIPS 4 1 27.2 29.6 57.2 24. KIPS 5 2 14.0 16.0 27.2 14.4 KIPS 6 2 13.2 14. 31.6 16.8 KIPS 7 2 14.4 16.8 32.8 16.4 KIPS 8 2 14.4 14.8 34.4 18.8 KIPS 9 3 15.2 14.4 37.2 18.8 KIPS 10 3 24.4 23.6 35.6 24.0 KIPS 11 3 15.2 14.0 27.2 20.4 KIPS 12 3 19.2 16.4 32.4 19.2 KIPS 13 4 16.0 13.6 28.8 16.0 KIPS 14 4 15.2 16.8 29.6 14.8 KIPS 15 4 13.6 14.4 29.2 16.0 KIPS 16 4 14.0 16.4 29.6 18.0 KIPS 17 5 12.8 13.2 28.0 14.4 KIPS 18 5 13.2 14.8 30. 19.6 KIPS 19 5 12.8 14.4 32.8 17.2 KIPS 20 5 24.0 14.0 31.2 17.6 KIPS 21 6 24.0 15.6 30.0 17.2 KIPS 22 6 13.2 15.2 31.6 18.0 KIPS 23 6 14.8 18.8 30.8 16.8 KIPS 24 6 14.4 14.8 31.2 38.6 KIPS 25 7 16.0 28.4 27.2 14.0 KIPS 28 7 16.4 13.2 30.8 30.4 KIPS 27 7 14.8 14.4 29.6 16.0 KIPS 28 7 13.6 15.6 26.4 16.4 KIPS FIGURE 4-26. EXAMPLE OF LOADS FROM IMPACT DETECTOR SITE UNDER HIGH-SPEED PASSENGER TRAIN 80 30 - 25 - 3 20 15 10 o I— | | I ** o- I I I — 30 to a. 5 25 a 2 20 15 - < 10 l_> (— uj 3 ^30 Q. S25 Q 3 20 m 15 - J— 01- O r~ os -3 -S -a 1 31 CD >— i zsz LU ro c_> =*t Q < LU Lu (— 1— cr: t— : =ar =D => _j oo o id Ci 5: _j t 9 »— • ►— « CJ> oo «a cr: cr: cr: o o Q f— u_ LU O cr: LU Ol =i I— Q oo LU <: _i or < LU LU a. UJ > 21 HI o i—i 3 cr: Q _J «3 cf CJ> c (J 3: t— i cr: LU 1— LU CD cr: CD Q LU ■ZL UJ > LU t/> 1— o CO - O _J i—* Q « Q LU 21 UJ LU O a: Q- 2T D_ co ca; Oi CVI <3- cr: Ofr 0C 02 01 SdiM 'ayoi 3iyy/"i33HM nyoiid3A 01- 83 CD S- C i — _*: i a <4- S- CU CD jz Q O C_> c£ UJ < UJ Q- Q UJ h- C_> ) — i Q UJ Od >- O 1 ■=c q z: z: o < Q UJ UJ a: o I/O U_ O 00 CO CO CO cr> CO LD ro *3- CD •-^ (O 00 +J Ql-i- •i- S- -— -nr fO O T3 a> J- 00 x: fO u CD fl3 2: o — - o a> 00 -—» i~ n cr> *3- 00 cu • • • Q. > cr> CO CO •r- CD CM CO •3- ^: a; TD A3 O _l T3 a; +-> •r— r— • TD rt3 O cr> CO oj E • • • S- i- CTi KO on o_ C\J CO CO "D — ^ CD JC CD a. Q- E OO- — OO 00 00 cr> cr> CM o CO o CM o u CL) S- S- o c 1—4 * 84 Similar effects have been noted in the wheel profiles where cars are normally run in a preferred direction. 85 5.0 TRACK COMPONENT DYNAMIC PERFORMANCE Wheel/rail impact loads occur over a brief time interval, typically less than three milliseconds, and contain energy over a wide frequency range. Thus the impacts will excite those structural components in the track which have natural frequencies of oscillation within that frequency range. The structural integrity of a track component can be influenced strongly by this response to wheel/rail impacts. Consequently, a static evaluation of a com- ponent is incomplete, and in many cases misleading. For example, the dynamic response of a concrete tie to impact loading can induce negative bending at the rail seat opposite the point of impact. This reaction is not obvious from static analysis or tests. Further, the fatigue life of a component may be much less than that predicted by considering only static or quasi-static loading conditions, depending on its natural frequencies and modal damping. Recent track research has focused on the processes of track compo- nent deterioration due to wheel/rail impact load. The simplified flow chart shown in Figure 5-1 describes the dynamic nature of these processes. Control of component deterioration relies to a large extent upon detection, inspec- tion, and repair of component damage, and on effective maintenance procedures. Without these controls, impact-induced component deterioration is an unstable feedback process. Of course, the amount of control which is required depends ultimately on cost-benefit trade-offs. This is established in the "feedback" portion of the two maintenance paths shown in Figure 5-1. Extensive dynamic measurements were made in Battel le's mechanical engineering laboratory and on the Northeast Corridor under a wide range of loading conditions and for several combinations of track components. The results of these tests, in conjunction with the analytical models, were used to evaluate the relationships between track component dynamic behavior and tie cracking, rail clip movement and shoulder insert loosening. These experiments and analyses are described in the following sections. 86 Wheel or rail running surface anomaly i' Wheel/rail interaction mechanics Track maintenance and modifications Vehicle maintenance and modifications i , UJ I— a: u_ 1— (-0 >>>>> < T3 a- <_> 1—4 (J rv i — i UJ Q ^ z: UJ _J CQ o o «* U. |— LO >- o <: r— (_> UJ ^ UJ i/i UJ O 1 Z3 2: _i o- 1 1 1 >— i UJ ID e£ cc: _l CC CD • CD ^t CD 1 <— LT) UJ CC ZD CD o CD NIVU1S 9NIQN39 311 4 Wnbl03dS-OinV 92 mechanisms for track damage such as tie cracking, insert loosening and clip fallouts, where a finite span of rail was necessary for proper simulation. The value of the laboratory track as a research tool depends on how well the laboratory results match the field results for similar test condi- tions. Consequently, tests were performed both in the field on NEC track and in the laboratory on the track panel to determine the range of validity and limitations of the simulation. Time histories of tie bending moment under the rail seat are shown in Figure 5-5 for similar impact tests with the drop hammer on NEC track (a) and on the 5-tie track panel (b). The frequency content of the impacted tie response was evaluated for several test conditions and was found to be similar. The largest differences between the field and laboratory data were found at the second tie bending natural fre- quency (about 330 Hz), where the spectral amplitude of response in laboratory tests was about twice that from field tests. These differences may be caused by the tie support conditions, which in the laboratory consisted of neoprene strips between the ties and the building floor, rather than a ballast section. Comparisons of peak amplitudes of tie bending response from the field tests and laboratory tests over a range of drop heights are shown in Figure 5-6. Results from the NEC and 5-tie track tests show close agreement for heights from 10 to 22 inches, using a 3/8-inch neoprene drop hammer shim pad. Results from the 5-tie track panel are also compared with single-tie tests from previous experiments [3], using a 6.5 mm shim pad. Both laboratory simulations appear to be suitable for evaluating certain types of track com- ponent performance. The differences in frequency content of tie bending response, particularly at the second tie bending mode, may be reduced by adjusting the support conditions of the ties as appropriate. 5.2 Results of Experiments The results of the experimental work conducted during this program are discussed below for three track component categories. These are (1) tie/tie pad, (2) shoulder insert, and (3) rail fastener (clip/insulator). The structural performance issues experienced in the NEC track related to each 93 COMPARISON OF TIE DYNAMIC RESPONSE UNDER RAIL SEAT 200 100 100 — Ul (D 300 m 200 — 100 100 FIELD TESTS 14" Drop No Preload 601A Clips EVA Pads Imp^J^^--^ — 10 20 30 10 20 TIME, MILLISECONDS FIGURE 5-5. COMPARISON OF TIE DYNAMIC BENDING RESPONSE BENEATH RAIL SEAT, NEC VERSUS 5-TIE TRACK 94 700 600 & 500 I c O D Field tests (Aberdeen, 3/8" shim) Lob tests (5-tie mock-up, 3/8" shim) Lob tests (5-tie mock-up, 6.5 mm shim) Lob tests* (single tie, 6.5 mm shim) c 0) c "O 400 300 200 100 Reference Approx. tie crocking level K) 12 14 16 Drop Height, inches 18 20 22 FIGURE 5-6. INFLUENCE OF HAMMER DROP HEIGHT ON PEAK TIE BENDING MOMENT UNDER RAIL SEAT FOR DIFFERENT TEST CONDITIONS 95 component are described. Examples of dynamic measurements of the component impact response are presented and evaluated in the context of track structural performance. 5.2.1 Tie/Tie Pad Performance As discussed previously, wheel/rail vertical impact loads have a pronounced effect on concrete tie performance and life. Cracks in the rail seat and insert areas, and loss of insert-to-tie bond have been attributed to impact loads. Typical tie failures are shown in Figure 5-7. Dynamic analyses of the concrete tie have defined the important transverse bending modes of vibration that are strongly excited by impacts. These are shown in Figure 5-8 along with typical frequency spectra for stiff and resilient rail seat pads. The second and third bending modes, at about 333 and 633 Hz, respectively, for the test tie are particularly important because strain amplitudes are near-maximum in the rail seat region. Phase relationship of frequency components of the transient response is critical, since the second vibration mode is asymmetrical, and the third is symmetrical: if response peaks to an impact fall in-phase, high, crack-initiating strains can be produced in the outer fibers of the tie at the bottom or the top surface. A rail-seat crack may consequently occur on the tie end opposite a single impact load, for example at a rail joint or battered weld. The primary load path into the tie for wheel/rail vertical "impact loads is through the rail-seat tie pads. Consequently, the dynamic character- istics of the tie pad are critical to tie performance. The first tie pad used on the NEC was an ethylene vinyl acetate (EVA) pad of high stiffness: about 5,000 kip/in secant stiffness over a typical range of wheel loads. The selection of this pad was made on the assumption that an effectively rigid pad would reduce clip deflections. This assumption did not prove correct. It was evident from the earlier investigation [3,4] that a reduction in tie pad stiffness can reduce significantly the levels of tie strain due to impact loads by reducing the peak force transmitted to the tie. This is not 96 Tie center flexural cracks Shoulder insert and negative bending crocks Longitudinal side cracks Positive rail seat bending cracks FIGURE 5-7. TYPICAL FLEXURAL CRACKS FOUND IN CONCRETE TIES 97 obvious from a static analysis because over the range of existing tie pad stiffness there is no significant effect on the load distribution to adjacent ties. However, the pad acts dynamically as a low-pass filter, and the filter characteristics of the pad are the key to tie performance. A desirable filter characteristic is a "break frequency" low enough to attenuate most of the energy at the second and third tie bending modes. An example of this filter- ing characteristic is evident in Figure 5-8, where the tie bending response with the more resilient 6.5 mm pads (with a secant stiffness typically below 1000 kip/in) is compared with the stiff EVA pad. These results are from the recent experiments on the NEC. Bending-moment time histories for these two pads from in-track drop tests are shown in Figure 5-9, where a marked reduc- tion in the peak bending moment is seen with the resilient pad. A similar reduction in ground rod (ballast vertical) acceleration for the same drop tests is shown in Figure 5-10. Therefore, the use of a properly-designed resilient tie pad will reduce the probability of tie cracking, and will reduce the impact load energy transferred to other track components, such as fasteners and ballast. 5.2.2 Rail Fastener Performance The rail fastening system used on the NEC concrete tie track con- sists of a steel spring clip driven into a shoulder/insert, which is cast into the concrete tie. Plastic insulators center the rail base within the fastener inserts. The spring clip, when new, preloads the rail base to 2000-3000 lb per clip. This "toe load" provides a longitudinal restraint of about 2500 lb per rail, per tie, depending on pad characteristics and friction coefficient. For the current design (the Pandrol 601A clip), progressive clip movement out of the insert is resisted only by the frictional forces at the clip/rail and clip/insert interfaces. Clip relative movement will occur when the net static and dynamic forces on the clip exceed the friction "breakout" levels. Fluctuations in vertical preload occur in response to wheel/rail loads. For wheel/rail impact loading, the dynamic motions of the clip and tie may be sufficient to cause large momentary reductions in preload, which in 98 (A) FIRST BENDING MODE RESPONSE (B) SECOND BENDING MODE RESPONSE (C) THIRD BENDING MODE RESPONSE (A) SAMPLES OF FIRST THREE BENDING MODES FOR CC-244-C CONCRETE TIE STIFF PAD 5 1.0 1.6 2.0 FREQUENCY kHz 2 5 3 (B) TIE BENDING RESPONSE AT RAIL SEAT FIGURE 5-8. ATTENUATION OF TIE BENDING RESPONSE TO IMPACT LOADING WITH RESILIENT RAIL-SEAT TIE PAD 99 22" Drop, No Preload c 250 TIME, msec TIME, msec FIGURE 5-9. TIE BENDING RESPONSE TO IN-TRACK DROP TESTS WITH STIFF EVA PADS AND RESILIENT (DAYCO) TIE PADS 100 s - 10 22" Drop, No Preload TIME, msec a. •10- RESILIENT PAD ~v^- — I- 40 20 TIME, msec FIGURE 5-10. GROUND ROD VERTICAL ACCELERATION RESPONSE TO IN-TRACK DROP TESTS WITH STIFF EVA TIE PADS AND RESILIENT (DAYCO) TIE PADS 101 turn may cause incremental movement of the clip out of the shoulder. Clip fallout therefore could result from repeated impact loads or from the high vibration levels of rail and tie under traffic. A series of laboratory and field experiments was performed to investigate the clip fallout phenomenon. The laboratory tests included modal vibration testing on free clips and repeated impact tests on the 5-tie track section. Field tests on the NEC included hammer impact tests as well as measurements under revenue traffic. The tie pad influences the clip deflection response by its effect on the relative vertical and rocking motions between the rail and tie. Measure- ments of clip deflection response under revenue traffic loading have indicated that rail rocking displacements are smaller with the resilient pads than with the very stiff pads [3]. Similar measurements made under simulated impact loading conditions are shown in Figure 5-11. These results show that the clips are loaded more severely with the effectively rigid EVA pad. Although the initial "peak-to-peak" clip vertical deflections are comparable (about 0.030 to 0.035 inch), the maximum clip spreading deflection due to rail uplift relative to the tie is about 40 percent less with the resilient pad than with the stiff EVA pad. The reduction in preload corresponding to the maximum initial depression of the pad is, however, about 50 percent greater with the resilient pad. From the standpoint of clip performance, the differences in clip spreading deflections (which correspond to the highest clip stresses) for stiff and resilient pads may be more important than the differences in preload reduction. On both the stiff and resilient pads, the maximum measured pad depressions represent a preload reduction of less than 5 percent of the nominal toe load. Thus, the tendency for clip movement longitudinally due to momentary reductions in preload may be similar for both pads. It should be noted, however, that tests on NEC track with 5 mm thick resilient pads (versus the recommended 6.5 mm pad) did produce excessive tie skewing, confirming some loss in toe load. Previous tests on the Type "A" clips have measured static strains of up to 10,000 microstrain in tension after installation. 102 TYPE "A" CLIP 18" IN-TRACK DROP TESTS 2 4 6 8 10 12 14 16 TIME. MILLISECONDS FIGURE 5-11. COMPARISON OF CLIP (TOE) VERTICAL DISPLACE- MENTS WITH STIFF EVA AND RESILIENT TIE PADS STIFF EVA PADS 18" IN-TRACK DROP TESTS 2 5 LU <_> Q. (0 Q FIGURE 5-12. COMPARISON OF CLIP LONGITUDINAL DISPLACE- MENT AT CENTER LEG FOR TWO CLIP DESIGNS 103 Consequently, the larger clip spreading deflections experienced with the stiff pads may result in plastic deformation, loss of toe load and reduced fatigue life. The vertical motions of the Type "B" (Pandrol E) clip were similar to those of the Type "A" clip shown in Figure 5-11. For example, with a resilient pad and an 18-inch hammer drop, the Type "A" clip toe motions were ±0.015 inch, while the Type "B" clip toe motions were +0.013 (unloading), - 0.008 inch (loading). These results reflect the relatively low clip stiffness, either type, when compared to the high pad stiffness. Clip stiffness has a small effect on initial pad compression, but controls rail uplift when the pad is unloaded. Comparison of the laboratory and field tests show that under impact loading, the clips respond strongly to the tie dynamics, particularly at the frequencies near the third and fourth tie bending modes (630 and 1000 Hz, respectively). Further, clip resonant conditions in the 800-2000 Hz range are also excited and influence response in the longitudinal axis of clip fallout movement. Typical time histories of clip longitudinal displacements are shown in Figure 5-12 for two clip designs. Type "A" (the Pandrol 601A) is used predominantly in the current NEC concrete tie track (see Section 3) and has experienced some fallouts and fracture occurrences. Type "B" (the Pandrol E clip) is a somewhat stiffer design, providing about 14 percent more preload. A number of these clips have been installed on the NEC, and no failures have been reported to date. As shown in Figure 5-12, the dynamic behavior of these two clips is strikingly different in the longitudinal (fallout) direction. The longitudinal displacements of Type "A" at the clip-shoulder interface (parallel to the rail) consist of a strong "ringing" oscillation at about 1050 Hz, which is near the fourth tie bending mode and in the range of clip resonant frequencies measured in the laboratory. In contrast, the impact response of Type "B" involves a more highly-damped, broad-band characteristic, with no evidence of the "ringing" measured with Type "A". Modal vibration tests on several Type "A" clips showed that clip installed position (i.e., over-driven versus under-driven) has a strong 104 influence on the clip natural frequencies. Modest variations in the installed position may tune or detune the clip natural frequencies from those of the tie, and consequently increase or decrease the probability of clip fallout motion. From these tests, clip natural frequency was observed to increase by roughly 20 percent as it was driven further into the insert. Test results indicate that the dynamic characteristics of the rail clip may influence its tendency to move out of the insert. Therefore, clip dynamics should be an important consideration in track design, along with the fastener response to nominal wheel loads (particularly on curved track). From the standpoint of clip fallout, the Type "B" clip dynamics are more desirable than the Type "A", since its natural frequencies are more highly damped and not strongly coupled to tie natural frequencies. It should be noted that in none of the field or laboratory tests was clip migration actually measured or observed. Thus, the precise mechanism of clip fallout has not been determined for a repeatable set of conditions. However, it seems reasonable to project that clip fallout results from a combination of several factors, including (1) clip and insert manufacturing tolerances, (2) high track vibration levels from rail surface or passing wheel roughness, and (3) the dynamic response of the clip to tie or rail vibrations at closely-coupled natural frequencies. 5.2.3 Shoulder/Insert Performance A series of tests was performed to explore the cause of bond loss between the cast-in fastener insert and the tie shoulder. Although the tests have not fully identified the failure mechanisms, observations and test results indicate that insert loosening can be caused by vertical impact loads of a magnitude less than that necessary to crack the tie. Observations of field damage and laboratory impact load tests imply that perhaps 100,000 impacts of moderate amplitude are necessary to cause gross insert looseness. The 5-tie laboratory track was used for repeated impact load tests to explore the insert bond loss phenomenon. Drop hammer tests were conducted as summarized below: 105 Pe rcentage Drop Hammer of Tie Cracki ng No. Heiqht, Inc ties Shim Pad Level D r ■ 12 3/8" Neoprene 52 9,500 14 3/8" Neoprene 55 5,900 18 6.5mm Ribbed Tie Pad 124 3,250 20 3/8" Neoprene 61 13,200 22 3/8" Neoprene 64 3,250 24 3/8" Neoprene Total 67 7,900 43,000 drops No visible damage was detected in the laboratory track during the course of this experiment. However, within the first 2000 drops (at 12 inches) some white powder was observed around the perimeter of the insert/tie shoulder interface on both the gauge and field-side inserts at the impacted rail. No additional powder accumulated after this point. Vibration analyses of insert response to rail impact loading have shown that the insert has several modes of axial vibration (vertically oriented in the tie) in the 4 to 8 kHz frequency range. Vertical acceleration spectra for the tie surface (shoulder) and the insert are compared in Fig- ure 5-13, where a strong response is seen at the insert between 5 and 7 Hz, but little corresponding spectral input is seen at the tie surface. These vibrations have a different "signature" for inserts on the same tie, or on different ties. Analyses of the insert design show that longitudinal compression waves will oscillate within this frequency range if the insert is either partially or totally constrained. (There are many locations within the insert that can cause reflections, which complicates the analysis.) A fully unconstrained insert will ring strongly at about 11 kHz: this has been demonstrated both analytically and experimentally. Although insert damage was not produced in the laboratory tests, the test measurements indicate that loss of bond may occur due to the effects of high-frequency longitudinal compression waves in the insert, excited by nearby impact loads. While the amplitudes at the steel -to-concrete interface are quite small, a microscopic pulverizing action may take place that over time 106 A SPEC 1 2.0000. MAS •A: 10 0.0 TIE (SHOULDER) PEC 2 ffc 10 EXPAND .0000— 1 1 INSERT MAC |Lk, A , h IV\ A \jy v v v vv^ vyV 1 i/V /NA ^^^^-^-^__ 0.0 — 1 ■ i i 1 4 6 8 FREQUENCY, kHz 10 12 (a) AUTO-SPECTRA OF VERTICAL ACCELERATIONS -80.000- I AVG 2 «A: 1 400. 00_ K_6750 hz A . INSERT REAL - \MJ\j\r WN fwvwyww- 300.00. \ |P ... ■ ,, ■ i iiif 12 3 4 TIME, MSEC (b) TIME-HISTORIES OF VERTICAL ACCELERATIONS FIGURE 5-13. COMPARISON OF INSERT AND TIE (SHOULDER) VERTICAL ACCELERATIONS FROM IN-TRACK DROP HAMMER TESTS (STIFF EVA PADS) 107 will result in gross loosening of the insert. Again, the more resilient tie pad with its low-pass filter characteristic will reduce the impact energy transferred to the tie, and in turn the excitation to the insert. 5.3 Conclusions Based on the results of these field and laboratory experiments, the following conclusions can be stated on track component dynamic performance: (1) Concrete tie track performance is strongly associated with the ability to withstand impact loads. Proper track design should therefore include dynamic analyses to optimize its performance. (2) Laboratory evaluation of track components can be accomplished with the aid of a drop hammer that simulates the magnitude and frequency content of the impact loads found in revenue service. (3) Component dynamic behavior should be evaluated to assure that no undesirable matching of responses takes place between components. (4) A dynamically optimized track structure must still withstand the abuse of the largest impact loads actually found in ser- vice. An economic trade-off exists between building stronger (and more expensive) track and maintaining wheel and rail running surface profiles in a more ideal geometric condition. (5) Dynamic performance (i.e., impact load response) of vehicle and track components other than concrete ties are not well known. There are ample economic reasons for needing to understand how these other components react to impact loading. 108 REFERENCES 1 Tuten J. M., "Analysis of Dynamic Loads and Concrete Tie Strain From the Northeast Corridor Track", Technical Memo by Battel le's Columbus Laboratories to the Federal Railroad Administration, Improved Track Structures Research Division, Contract DOT-FR-9162, May 1981. 2 Harrison, H. and Moody, H., "Correlation Analysis of Concrete Cross Tie Track Performance", Proceedings, Second International Heavy Haul Railway Conference, Sept. 1982, Paper 82-HH-39, pp. 425-431. 3 Dean F. E., et al, "Investigation of the Effects of Tie Pad Stiffness on the Impact Loading of Concrete Ties 1n the Northeast Corridor", Report No. FRA/ORD-83/05, April 1983. 4 Dean, F. E., et al, "Effect of Tie Pad Stiffness on the Impact Loading of Concrete Ties", Proc, Second International Heavy Haul Railway Conference, Sept. 1982, Paper 82-HH-41, pp. 442-458. 5. Tuten, J. M. and Harrison, H. D., "Design, Validation and Application of a Monitoring Device for Measuring Dynamic Wheel/Rail Loads", ASME Technical Paper, 1984 Winter Annual Meeting. 6. Harrison, H. D., et al , "Correlation of Concrete Tie Track Performance in Revenue Service and at the Facility for Accelerated Service Testing", Volume I A Detailed Summary, Report No. D0T/FRA/0RD-84/02.1 , Final Report, August 1984. 7. Sato, Y. and Kosuge, S., "Evaluation of Rail Head Surface Configuration Viewed from Wheel Load Variation", Quarterly Reports (RTRI), Vol. 24, No. 2, 1983, pp. 68-71. 8. Newton, S. G. and Clark, R. A., "An Investigation into the Dynamic Effects on the Track of Wheelflats on Railway Vehicles", Journal Mechanical Engineering Science, IMechE, Vol. 21, No. 4, 1979, pp. 287-297. 9. Mair, R. I., "Aspects of Railroad Track Dynamics (Part I: Vertical Response)", BHP Melbourne Research Labs, Report MRL 81/3 (BHPMNM/RDC/74/017), Feb. 1974. 10. Mair, R. I., "Natural Frequency of Rail Track and Its Relationship to Rail Corrugation", Civil Engineering Transactions 1977, Inst, of Engineers, Australia, pp. 6-11. 11. Ahlbeck, D. R., "An Investigation of Impact Loads Oue to Wheel Flats and Rail Joints", ASME Paper 80-WA/RT-l, 1980. 12. Jenkins, H. H., et al, "The Effect of Track and Vehicle Parameters on Wheel/Rail Vertical Oynamic Forces", Railway Engineering Journal, Vol. 3, No. 1, Jan. 1974, pp. 2-26. 13. Ahlbeck, D. R., et al "Measurements of Wheel/Rail Loads on Class 5 Track". Report No. FRA/ORD-80/19, Feb. 1980. 14. Mullen, J. D., "Rail Profile Irregularities - a study of their effects on concrete sleepered track and their removal", Proc. 5th Int. Rail Track Conference, Sept. 12-18, 1983, Newcastle, NSW, Australia. 109