Maintenance of belt conveyors using an expert system based on fuzzy logic


Original Research Article

Maintenance of belt conveyors using an expert
system based on fuzzy logic

D. Mazurkiewicz *

Lublin University of Technology, Nadbystrzycka 36, 20-618 Lublin, Poland

1. Introduction

New technological possibilities result in an extremely dynamic
industrial development. Due to ever-increasing competition
on the market, higher customer demands and pursuit of
innovative products, manufacturers are compelled to imple-
ment more and more advanced engineering solutions,
particularly ones that lead to higher efficiency and reduced

operational costs. In effect, machinery and technological
devices are becoming more and more complex, which imposes
higher demands on their users, including the necessity of
applying suitable methods and techniques to ensure durability
and reliability of frequently complex and elaborate production
systems [1,2].

In recent years, conveyor belt transport systems have taken
on a new significance due to numerous research studies on
innovative design solutions for conveyor belts [3–10]. The

a r c h i v e s o f c i v i l a n d m e c h a n i c a l e n g i n e e r i n g 1 5 ( 2 0 1 5 ) 4 1 2 – 4 1 8

* Tel.: +48 815384229.
E-mail address: d.mazurkiewicz@pollub.pl

a r t i c l e i n f o

Article history:

Received 13 November 2014

Accepted 19 December 2014

Available online 13 January 2015

Keywords:

Computer-aided maintenance

management systems

Belt conveyor

Fuzzy logic

a b s t r a c t

In recent years, conveyor belt transport systems have taken on a new significance due to

numerous research studies on innovative design solutions. The application of these new

developed solutions leads to considerable reduction in operational costs of transport

systems, while ensuring their high reliability and service life at the same time. Nonetheless,

there are still areas that pose challenge to both research and development. Typical chal-

lenges are analyzed in this paper. The solution to the problems of conveyor transport

maintenance can be the implementation of a system for estimation of technical condition of

conveyor belt joints. It serves as a second level safety diagnostic system for transport.

Besides real-time measurements, the system enables a long-term analysis of historic data

for every single joint that makes up the conveyor belt loop, from the moment of its

manufacture to the final operation. The effectiveness of a conveyor belt diagnostic system

primarily depends on the use of a decision supporting system. With adequate inference

rules applied, this system would increase the effectiveness and shorten the time of decision-

making as well as verify generated signals. The above tasks can be performed by a suitable

expert system that predicts values of the analyzed time series, using the predicted values

and inference rules to verify any potential false alarm signals at the same time. The idea and

algorithm of such an expert system were presented in this article as well.

# 2015 Politechnika Wrocławska. Published by Elsevier Urban & Partner Sp. z o.o. All
rights reserved.

Available online at www.sciencedirect.com

ScienceDirect

journal homepage: http://www.elsevier.com/locate/acme

http://dx.doi.org/10.1016/j.acme.2014.12.009
1644-9665/# 2015 Politechnika Wrocławska. Published by Elsevier Urban & Partner Sp. z o.o. All rights reserved.

http://crossmark.crossref.org/dialog/?doi=10.1016/j.acme.2014.12.009&domain=pdf
http://crossmark.crossref.org/dialog/?doi=10.1016/j.acme.2014.12.009&domain=pdf
http://dx.doi.org/10.1016/j.acme.2014.12.009
mailto:d.mazurkiewicz@pollub.pl
http://www.sciencedirect.com/science/journal/16449665
http://www.elsevier.com/locate/acme
http://dx.doi.org/10.1016/j.acme.2014.12.009


application of these new developed solutions leads, among
others, to considerable reduction in operational costs of
transport systems, while ensuring their high reliability and
service life at the same time. Nonetheless, there are still areas
that pose challenge to both research and development. As
demonstrated in the studies [3,4,11], it is absolutely vital from
a practical point of view that the service life of conveyor belt
joints be made longer owing to the fact that these joints are
crucial elements of the entire conveyor transport system. For
instance, vulcanized and adhesive-bonded joints of textile-
rubber belts are made up of lap joints, their theoretical
durability for three separators amounting to only 67% of the
belt's rated durability [4]. Moreover, the forces that affect the
belt load and durability of both the belt and its joints vary
during operation. For this reason, methods applied to this end,
e.g. ones for determining non-stationary states, are often
inaccurate, which results from lack of precise data and
insufficient determination of boundary conditions. On the
other hand, it is vital to optimize the belt tension force for
various conveyor load conditions, as this should result in
higher durability of both the belt and its joints. The application
of a solution based on standard control algorithms seems,
however, inadequate due to the nature of the problem as well
as lack of both sufficient insight into the problem and
measurement data.

2. Operation of an intelligent advisory system
in considerable uncertainty conditions

The solution to the above problems of conveyor transport
maintenance can be the implementation of a system for
estimation of technical condition of conveyor belt joints [11]
that serves as a second level safety diagnostic system for
transport. Besides real-time measurements, the system
enables a long-term analysis of historic data for every single
joint that makes up the conveyor belt loop, from the moment
of its manufacture to the final operation. The monitoring
device can also be converted into an autonomously function-
ing instrument that can independently react to variable
operational conditions and thus eliminate states that pose
threat to the system's reliability, e.g. it can counteract belt
failure by predicting the occurrence of operational parameters
and their consequences. However, the vast number of data and
information generated by the measuring system (which is the
case when the monitoring covers a great deal of joints that are
usually located on up to even several conveyors in a complex
transport system) leads to a series of complications regarding
interpretation of the collected data, which consequently
hinders effective decision-making by the monitoring staff.
In addition, given the specific nature of operation of conveyor
belts (e.g. their exposure to high momentary loads), the
measured discrete values of variations in length of an
adhesive-bonded joint can generate false alarms that can
result in unjustified stoppage of the transport system,
unnecessary inspection of the joint or complicated and
time-consuming examination of the generated signal by the
operator. The system operator is then continuously required to
take decisions, analyze a vast data set and observe changes
therein, as well as predict the consequences of further changes

in functional properties. This means taking decisions in
uncertainty conditions [12,13], when we have no information
about the probable occurrence of particular states (safe/
emergency/alarm) and can only predict values of the analyzed
signals that have the form of a discrete time series.

The conveyor transport system is an object that is affected
by a number of factors. These factors are either imprecisely
determined or difficult, if not impossible, to measure because
of their random and incidental nature. This object is hard to
describe owing to its dependence on numerous unpredicted
variables that are often additionally difficult to be precisely
determined. Failure-signaling symptoms can occur either
once or many times; they can also have different intensity
and nature. Another common problem here is lack of
unambiguous information about the analyzed object and
the required short time of a potential reaction. When
equipping a comprehensive conveyor transport system with
efficient and advanced monitoring and diagnostic systems, it
is therefore necessary that sophisticated and intelligent tools
be used to support the above. More specifically, the intelligent
advisory system will be a dynamic system for the supervision
of complex processes that are characterized by variable
operating conditions and two time scales: micro- and
macro-time scales [14], which is essential with regard to the
entire life cycle of a single adhesive joint or a belt section that
usually undergoes frequent relocations.

3. Inference supporting system based on fuzzy
residuum evaluation in the diagnostics of
conveyor belts

The effectiveness of a conveyor belt diagnostic system
primarily depends on the use of a decision supporting system.
With adequate inference rules applied, this system would
increase the effectiveness and shorten the time of decision-
making as well as verify generated signals. The above tasks
can be performed by a suitable expert system that predicts
values of the analyzed time series, using the predicted values
and inference rules to verify any potential false alarm signals
at the same time. The prediction of variations in length of an
adhesive-bonded joint involves classification and estimation
of time series. Although all standard classification and
estimation methods can, under certain conditions, be also
employed for prediction, the most effective method for
predicting time series involves the use of artificial neural
networks [15–20]. A supervision system for adhesive-bonded
joints of conveyor belts should therefore also include an
intelligent expert system based on the popular knowledge
representation and rule-based systems (Fig. 1).

When measuring variations in length of belt joints by a
computer-aided measurement system, it can be assumed that
the observations are made at discrete and equal intervals of
time t. For every joint we have data from the evaluation of its
length DL in time moments t � 1, t � 2, . . ., t � n, which
corresponds to the values of DLt�1; DLt�2; DLt�3; . . . ; DLt�n. The
aim of the analysis is to determine values of elongation of the
monitored conveyor belt joint DLtðwÞ, denoting the prediction in
a moment t + w, i.e. in advance w, with the lowest possible mean
deviations DLtþw � DLtðwÞ. When evaluating the potential

a r c h i v e s o f c i v i l a n d m e c h a n i c a l e n g i n e e r i n g 1 5 ( 2 0 1 5 ) 4 1 2 – 4 1 8 413



https://isiarticles.com/article/46362