Professional Documents
Culture Documents
Expert System in PP A State of The Art Survey
Expert System in PP A State of The Art Survey
Intelligent solutions, based on expert systems, to solve problems in the ®eld of production planning
and scheduling are becoming more and more widespread nowadays. Especially the last decade has
witnessed a growing number of manufacturing companies, including glass, oil, aerospace,
computers, electronics, metal and chemical industriesÐto name just a fewÐinterested in the
applications of expert systems (ESs) in manufacturing. This paper is a state-of-the-art review of the
use of ESs in the ®eld of production planning and scheduling. The paper presents famous expert
systems known in the literature and current applications, analyzes the relative bene®ts and concludes
by sharing thoughts and estimations on ESs future prospects in this area.
Keywords: Expert systems, arti®cial intelligence, production planning, scheduling, intelligent
manufacturing
can make a signi®cant contribution to improving contains the knowledge needed for solving a speci®c
control and manufacturing systems (Wilson, 1985; problem. The knowledge may be in the form of facts,
Shaw and Whinston, 1986). This paper comes to review heuristics (e.g. experiences, opinions, judgments,
the use of expert systems in the area of production predictions, algorithms) and relationships usually
planning and scheduling. In the following sections, we gleaned from the mind of experts in the relevant
present concepts and characteristics of famous expert domain. Knowledge can be represented using a
systems in this area, the bene®ts gained through their variety of representation techniques (e.g. semantic
utilization and estimations on the future trends. nets, frames, predicate logic) (Badiru, 1992; Ignizio,
1991; Mital and Anand, 1994), but the most
commonly used technique is ``If-Then'' rules, also
known as production rules.
2. The expert systems technology The inference engine is employed during a
consultation session, examines the status of the
ESs are one of the most commercially successful knowledge base, handles the content of the knowledge
branches of arti®cial intelligence (AI). Welbank base and determines the order in which inferences are
(1983) de®nes an expert system as follows: made. It may use various inference methods (Badiru,
1992; Ignizio, 1991; Mital and Anand, 1994).
An expert system is a program which has a wide
The user interface part enables interaction of the
base of knowledge in a restricted domain, and uses
system with the user. It mainly includes screen
complex inferential reasoning to perform tasks
displays, a consultation/advice dialogue and an
which a human expert could do.
explanation component. In addition, expert systems
In other words, an ES is a computer system containing provide interfaces for communication with external
a well-organized body of knowledge which emulates programs including databases and spreadsheets.
expert problem solving skills in a bounded domain of ESs have some signi®cant advantages in compar-
expertise. The system is able to achieve expert levels ison with the traditional computer systems. These
of problem solving performance, which would advantages are presented in the following Table 1.
normally be achieved by a skilled human when Concerning the development of ESs, there are
confronted with signi®cant problems in the domain mainly two groups of development tools (Baker,
(BCS, Expert Systems Specialist Group). As illu- 1988; Huntington, 1985; Jackson, 1986):
strated in Fig. 1, an ES consists of three main * High level programming languages (C ,
components, which include the knowledge base, the
PROLOG, LISP, etc.). Using these languages,
inference engine and the user interface.
the system designer has a great deal of freedom
The knowledge base is the heart of the system and
in his choice of knowledge representation
techniques and control strategies. However,
use of these languages requires a high degree
of expertise and skill.
* Expert system shells. They attempt to combine
the ¯exibility of AI languages with the cost-
effectiveness and provide more general devel-
opment facilities. There are a number of
commercial shells available in the market with
varying features (Nexpert Object, XpertRule,
KnowledgePro, CLIPS, ReSolver, EXSYS, VP-
Expert, ACQUIRE, etc.). Most of them are
relatively low priced and provide a rule-based
knowledge representation mechanism.
In general, the success of an ES depends on how much
knowledge it has and how qualitative that knowledge
Fig. 1. Expert system's architecture. is.
Expert systems in production planning 255
Availability
Experts are not born. They have to be trained and then practiced. It generally takes over ®ve years for someone to acquire
expertise in a particular area. In contrast to the human, expert system has all the expertise inside, it never gets tired or
dies. The included knowledge is often more readily available to trainee experts or users.
Consistency
Even the best experts can make mistakes or may forget an important point. Once an expert system is programmed to ask
for and use certain inputs, it is not prone to forgetfulness. If a line of reasoning is acceptable, it will remain so in different
consultations.
Comprehensiveness
An expert can only draw upon his own knowledge and experience. In some domains an expert systems could encapsulate
the knowledge of more than one expert and consequently offer several options.
3. Expert systems towards production planning hand, Wong et al. (1994) implemented a study in 1993
and scheduling examining the current utilization of ESs and their
bene®ts in manufacturing among the 500 largest
Planning and scheduling are forms of decision- industrial companies in the USA. They invited all
making, which play a crucial role in manufacturing Fortune 500 industrial corporations (based on the
as well as in service industries. Planning is the process 1990 ranking) to participate in a mail survey. The
of selecting and sequencing activities such that they mailing procedure produced 98 usable responses in
achieve one or more goals and satisfy a set of domain total, which meant a usable response rate of 19.6%.
constraints. Scheduling is the process of selecting Among the 98 responding companies, the mean
among alternative plans and assigning resources and number of employees was 19,000, while gross
times to the set of activities in the plan. These annual sales averaged 6.2 billion dollars. In this
assignments must obey a set of rules or constraints study, scheduling emerged as the most common
that re¯ect the temporal relationships between application area of ESs, as shown in the Table 2.
activities and the capacity limitations of a set of In an other study implemented by Byrd (1995), who
shared resources (Artiba, 1997; Brucker, 1998; interviewed 74 knowledge engineers (KEs) of 28
Pinedo, 1995). organizations, production scheduling appeared to be
In the current competitive environment, effective the second most common type of ES in general, the
planning and scheduling has become a necessity for ®rst being diagnosis. Table 3 gives an indication of
survival in the marketplace. Companies have to meet how many ESÐrelated to production managementÐ
shipping dates committed to the customers, as failure was in use and being developed in the 28 organiza-
to do so may result in a signi®cant loss of good will tions of the KEs.
and reliability. They also should schedule activities in Concerning the bene®ts reported from the use of
such a way as to use the resources available in an this technology by the KEs, the interviewees said they
ef®cient way. In this framework manufacturing received from their ESs:
companies decided to adopt intelligent solutions,
* Better customer service
since the traditional manufacturing planning and
* Reduction in time to complete tasks
scheduling mechanisms were found insuf®ciently
* Organizational learning
¯exible to respond to changing production styles
* Increases in production
and highly dynamic variations in product require-
* More effective use of resources
ments (Meredith et al., 1994; Kusiak, 1990).
* Reduction in staff
A mid-1990s survey reported by Durkin (1996) has
revealed manufacturing industry to be one of the most Moreover, many researchers have regularly written
widely applied area for expert systems. On the other about the use of ESs in production planning and
256 Metaxiotis, Askounis and Psarras
Table 2. Areas of ES applicationsÐSurvey of Wong et al. 4. Application of ESs in production planning and
(1994) scheduling
System domain Frequency Percentage1
A number of applications of ESs to the area of
Scheduling 18 35.3 production planning and scheduling have been
Process design 16 31.4 developed and documented. The intelligent sche-
Maintenance and repair 15 29.4 duling and information system (ISIS) was the ®rst
Process selection 13 25.5 application of ES to job-shop scheduling (Fox and
Facility layout 11 21.6 Smith, 1984). ISIS used hierarchical planning to
Material selection 6 11.8
decompose complex problems into manageable
Production planning and control 4 7.8
Capacity planning 4 7.8 pieces. The research with ISIS led to work on the
Facility location 3 5.9 development of the opportunistic scheduler (OPIS)
Project management 3 5.9 (Ow and Smith, 1986), a knowledge-based factory
Tool selection 2 3.9 scheduling system which uses problem decomposi-
Data selection 2 3.9 tions to generate constraint-satisfying shop schedules.
Quality control 2 3.9 The prototype expert priority scheduler (PEPS)
Forecasting 2 3.9 (Robbins, 1985) is a rule-based ES which solves
Storeroom design 1 2.0 problems in shop ¯oor control level, although its
Vendor selection 1 2.0 drawback is the fact that it is not able to recognize
1
Percentages do not add up to 100 because the respond- uncertainty and downstream data dependency.
ents could choose more than one area of application. PATRIARCH (Morton et al., 1984) is a multilevel
planning, scheduling and control system that was
scheduling and the potential bene®ts of them (Coursey developed at Carnegie Mellon University for manu-
and Shangraw, 1989; Ehner and Bax, 1983; Jayaraman facturing. The four levels of the PATRIARCH system
and Srivastava, 1996; Kanet and Adelsberger, 1987; include: (1) strategic planning, (2) capacity planning,
Kusiak and Chen, 1988; Liu, 1985; Mertens and Kanet, (3) scheduling, (4) dispatching. The OPT scheduling
1986; Meziane et al., 2000; Pereira, 1996; Pham and system was reported by Jacobs (1983).
Pham, 1999; Zhang and Chen, 1999). According to A hybrid expert system HESS (Deal et al., 1992)
these researchers, ES can help organizations to cut was developed at the University of Houston in support
costs by reducing the need for some personnel, of product scheduling at a major petrochemical ®rm's
preserve and disseminate scarce expertise throughout re®nery. The knowledge base in HESS was developed
the organization, give better consistency to decision to determine what products to produce at what time,
making, improve quality of products. and through which processors. HESS was developed
using the EXSYS expert system shell and consists of
Table 3. Total number of ES used and being developedÐ approximately 400 production rules.
Study of Byrd (1995) A management analysis resource scheduler MARS
(Marsh, 1985) has been developed to schedule
Status Number of ESs
resources for the space transportation system.
Currently used Chiodini (1980) developed an expert system for
0 1 dynamic manufacturing rescheduling, while Biegel
1±5 15 and Wink (1989) proposed an expert system for
6±10 6 industrial job-shop scheduling.
11±40 3 A knowledge-based simulation system for manu-
4 40 2
facturing scheduling was proposed by Palaniswami
Being developed
0 0 and Jenicke (1992), while Alexander (1987) devel-
1±5 13 oped an expert system for the selection of scheduling
6±10 7 rules in a job shop.
11±40 1 A knowledge-based simulation model for job shop
4 40 4 scheduling was also proposed by Abdallah (1995).
The knowledge base of the model was built using the
Expert systems in production planning 257
simulation technique by studying the effect of based rescheduling expert system which was adapted
different technological factors on the selection of to the ¯exible manufacturing environment, while
scheduling decisions. Tayanlthi et al. (1992) proposed a knowledge-based
De Toni et al. (1996) proposed an intelligence- simulation system to analyze and handle the dis-
based production scheduler, which utilizes a hybrid turbances (including machine breakdowns and rush
push/pull approach to schedule. This scheduler uses orders) in a ¯exible manufacturing environment.
some blackboard techniques of the type hypothesized Recently a production rescheduling expert simula-
by Hayes-Roth (1985). The production-scheduling tion system was also proposed by Li et al. (2000). This
blackboard consists of frames, lists and production system integrates different techniques and methods,
rules, plus a blackboard controller with a shop ¯oor including simulation technique, arti®cial neural net-
control system interface and codes/routings archives. work, expert knowledge and dispatching rules and
Custodio et al. (1994) discussed the issue of deals with four sources of production disturbances: (a)
production planning and scheduling using a fuzzy incorrect work, (b) machine breakdowns, (c) rework
decision system, while several outlines concerning the due to quality problems; and (d) rush orders.
development of a rule-base for the speci®cation of During this survey's literature research, we found
manufacturing planning and control systems were some research projects using expert systems tech-
recently made by Howard et al. (2000). nology for manufacturing planning, scheduling and
A fuzzy rule-based scheduler was proposed by execution control. Table 4 presents a summary of
Subramaniam et al. (2000), which dynamically these projects.
selects, from several candidate dispatching rules, the
most appropriate dispatching rule to employ, based on
the prevailing job shop conditions. An expert system 5. Conclusions/recommendations
named KDPAG was built by Chen et al. (1998)
applied to materials design and manufacture. It is sure that the manufacturing enterprises of the 21st
In addition, particular attention is also dedicated to century will be in an environment where markets are
the issue of effective rescheduling (Brown, 1989; frequently shifting, new technologies are continu-
Sarin and Salgame, 1989; Szelke and Kerr, 1994). ously emerging and competitors are multiplying
Yamamoto and Nof (1985) suggested a Regeneration globally. Manufacturing strategies should therefore
Method when they exploited production schedule shift to support global competitiveness, new product
expert system. Driscoll (1993) studied a knowledge- innovation and rapid market responsiveness. The next
258 Metaxiotis, Askounis and Psarras
generation manufacturing systems will thus be more Operations & Production Management, 15(10), 89±
strongly time-oriented, while still focusing on cost 102.
and quality. Alexander, S. M. (1987) An expert system for the selection
The ®ndings of this survey show that ESs are of scheduling rules in a job shop. Computers &
generally perceived to be very useful in production Industrial Engineering, 12(3), 167±171.
Artiba A. (1997) Planning and Scheduling of Production
planning and scheduling (Kumura et al., 1986;
SystemsÐMethodologies and Applications, Chapman
Liebowitz and Lightfoot, 1987). ES are becoming & Hall.
more and more common decision-making tools in Badiru, A. B. (1992) Expert Systems Applications in
many organizations. The bene®ts reported from the Engineering and Manufacturing, Prentice Hall, New
use of ESs in this area include more accurate Jersey.
decisions, time gains, improved quality and more Baker, S. (1988) Nexpert object: mainstreaming AI
ef®cient use of resources. It is our belief that the applications. IEEE Expert, Winter 1988.
usefulness of ESs in production planning and Biegel, J. E. and Wink, L. J. (1989) Expert systems can do
scheduling will gain more recognition, if they are job shop scheduling: an exploration and a proposal.
properly integrated with operations research (OR) Computers & Industrial Engineering, 17(1), 347±352.
techniques, especially simulation. Sabuncuoglu and Brown, M. C. (1989) The dynamic reproduction scheduler:
Hommertzheim (1989) have observed major bene®ts conquering the changing production environment, in
Pau, L. F., Motiwalla, J., Pao, Y. H. and Theh, H. H.
provided by expert simulation systems.
(ed.), Expert Systems in Economics, Banking and
Most of the ESs that have been discussed and
Management, North-Holland, Amsterdam.
developed are essentially stand-alone systems. Brucker, P. (1998) Scheduling Algorithms, Springer Verlag.
However, it is very likely that in the near future a Byrd, T. A. (1995) Expert systems implementation:
large portion of the ESs developed will be embedded interviews with knowledge engineers. Industrial
systems, that is, systems which form a part of the Management & Data Systems, 95(10), 3±7.
overall software package. Hybrid ESs are one Chen, N., Li, C. and Qin, P. (1998) KDPAG expert system
example of such an approach. Since most operations applied to materials design and manufacture.
management problemsÐin generalÐare not isolated Engineering Applications of Arti®cial Intelligence,
problems by their nature, isolated ESs can not solve 11, 669±674.
exactly the problem of the manufacturing manager. Chiodini, V. (1980) An expert system for dynamic
On the other hand, it must be pointed out that top manufacturing rescheduling, Symposium on Real
Time Optimisation in Automated Manufacturing
management plays a pivotal role in the productive
Facilities, National Bureau of Standards,
implementation of ESs. Grindley (1988) reported that
Gaithersburg, MD.
some executives have never even heard of the term Coursey, D. H and Shangraw, R. F. (1989) Expert system
``expert system'', while others have different percep- technology for managerial applications: a typology.
tions of what an ES is or does and to what extent their Public Productivity Review, 12(3), 237±262.
expectations can be realized. There is no doubt that Custodio, L. M. M., Sentieiro, J. J. S. and Bispo, C. F. G.
lack of knowledge about ESs often results in a lack of (1994) Production planning and scheduling using a
interest in utilization. In some cases top management fuzzy decision system. IEEE Transactions on Robotics
do not support ES projects because they see the and Automation, 10(2), 160±168.
computer as a tool solely for ®nancial and accounting De Toni, A., Nassimbeni, G. and Tonchia, S. (1996) An
applications. So, in order to ensure success, top arti®cial, intelligent-based production scheduler.
management should show a genuine interest in Integrated Manufacturing Systems, 7(3), 17±25.
exploring the applications of ESs in manufacturing Deal, D. E., Chen, J. G., Ignizio, J. P. and Jayakamer, V.
(1992) An expert system scheduler: some re¯ections on
planning and scheduling.
expert system development. Journal of Computers and
Operations Research, 5(3), 10±17.
Dornan, B. (1987) A status report: arti®cial intelligence.
Production, 46±50.
References Driscoll, W. C. (1993) A microcomputer-based scheduling
assist system. Computers and Industrial Engineering,
Abdallah, M. (1995) A knowledge-based simulation model 25(4), 223±226.
for job shop scheduling. International Journal of Durkin, J. (1996) Expert systems: A view of the ®eld, in
Expert systems in production planning 259
IEE: Expert, Intelligent Systems with Applications, Meredith, J. R., McCutcheon, D. M. and Hartley, J. (1994)
April, 56±63. Enhancing competitiveness through the new market
Ehner, W. and Bax, F. R. (1983) Factory of the futureÐa value equation. International Journal of Operations &
manufacturing viewpoint. Production, 91(4), 38±48. Production Management, 14(11), 7±22.
Fox, M. S. and Smith, S. F. (1984) ISIS: a knowledge-based Mertens, P. and Kanet, J. (1986) Expert systems in
system for factory scheduling. Expert Systems, 1(1), production management: an assessment. Journal of
25±49. Operations Management, 6(4), 393±404.
Grindley, K. (1988) End users are key to success. Meziane, F., Vadera, S., Kobbacy, K. and Proudlove, N.
Datamation, 34(7), 100±111. (2000) Intelligent systems in manufacturing: current
Hayes-Roth, B. (1985) A blackboard architecture for developments and future prospects. Integrated
control. Arti®cial Intelligence, 26, 251±321. Manufacturing Systems, 11(4), 218±238.
Howard, A., Kochhar, A. and Dilworth, J. (2000) Case Mital, A. and Anand, S. (1994) Handbook of Expert
studies based development of a rule-base for the Systems, Chapman & Hall.
speci®cation of manufacturing planning and control Morton, T. E, Fox, M. and Sathi, A. (1984) PATRIARCH: A
systems. International Journal of Production multilevel system for cost accounting, planning,
Research, 38(12), 2591±2606. scheduling, Partial working document, Graduate
Huntington, D. (1985) EXSYS Expert Systems Development School of Industrial Administration, Carnegie Mellon
Package, EXSYS Manual, Albuquerque, New Mexico. University.
Ignizio, J. P. (1991) Introduction to Expert Systems, Ow, P. S. and Smith, S. F. (1986) Viewing scheduling as an
McGraw-Hill Inc., USA. opportunistic problem-solving process, Working paper,
Jackson, P. (1986) Introduction to Expert Systems, Addison- Carnegie-Mellon University, Pittsburg, PA.
Wesley, Wokingham, England. Palaniswami, S. and Jenicke, L. (1992) A knowledge-based
Jacobs, R. D. (1983) The OPT scheduling system: A review simulation system for manufacturing scheduling.
of a new production scheduling system. Production International Journal of Operations & Production
and Inventory Management, 24(3). Management, 12(11), 4±14.
Jayaraman, V. and Srivastava, R. (1996) Expert systems in Pereira, M. G. (1996) Expert systemsÐaspects we must
production and operations management. International consider for use in production programming. Com-
Journal of Operations & Production Management, puters & Industrial Engineering, 31(1±2), 413±425.
16(12), 27±44. Pham, D. T. and Pham, P. T. N. (1999) Arti®cial intelligence
Kanet, J. J. and Adelsberger, H. H. (1987) Expert systems in in engineering. International Journal of Machine Tools
production scheduling. European Journal of & Manufacture, 39, 937±949.
Operational Research, 29, 51±59. Pinedo, M. (1995) Scheduling: Theory, Algorithms and
Kumura, S., Joshi, S., Kashyap, R., Moodie, C. and Chang, Systems, Prentice Hall.
T. (1986) Expert systems in industrial engineering. Robbins, J. H. (1985) PEPS: The Prototype Expert Priority
International Journal of Production Research, 24, Scheduler, Technical paper MM-1093, Society of
1107±1125. Manufacturing Engineers, Dearborn, MI.
Kusiak, A. (1990) Intelligent Manufacturing Systems, Sabuncuoglu, I. and Hommertzheim, L. (1989) Expert
Prentice-Hall, Englewood Cliffs, NJ. simulation systemsÐrecent developments and applica-
Kusiak, A. and Chen, M. (1988) Expert systems for planning tions in ¯exible manufacturing systems. Computers &
and scheduling manufacturing systems. European Industrial Engineering, 16(4), 575±585.
Journal of Operational Research, 34(2), 113±130. Sarin, S. C. and Salgame, R. (1989) A knowledge-based
Li, H., Li, Z., Li, L. and Hu, B. (2000) A production system approach to dynamic scheduling, in Kusiak, A.
rescheduling expert simulation system. European (ed.) Knowledge-based Systems in Manufacturing,
Journal of Operational Research, 124(2), 283±293. Taylor & Francis, Philadelphia, PA.
Liebowitz, J. and Lightfoot, P. (1987) Expert systems for Shaw, P. and Whinston, B. (1986) Application of arti®cial
scheduling: a survey and preliminary design concepts. intelligence to planning and scheduling in ¯exible
Applied Arti®cial Intelligence Journal, 1(3). manufacturing, Flexible Manufacturing Systems:
Liu, D. (1985) Intelligent manufacturing planning systems, Methods and Studies, Kusiak, A. (ed.), North-
Technical Paper MS85-1070, Society of Manufacturing Holland, 223±242.
Engineers. Subramaniam, V., Ramesh, T., Lee, G. K., Wong, Y. S. and
Marsh, C. A. (1985) MARSÐan expert system using the Hong, G. S. (2000) Job shop scheduling with dynamic
automated reasoning reasoning tool to schedule fuzzy selection of dispatching rules. International
resources. Robotics and Expert SystemsÐProceedings Journal of Advanced Manufacturing Technology, 16,
of Robex 85, Instrument Society of America, 123±125. 759±764.
260 Metaxiotis, Askounis and Psarras
Szelke, E. and Kerr, R. M. (1994) Knowledge-based Wong, B. K., Chong, J. K. S. and Park, J. (1994) Utilization
reactive scheduling. Production Planning & Control, and bene®ts of expert systems in manufacturing: a
5(2), 124±45 study of large American industrial corporations.
Tayanlthi, P., Manivannan, S. and Banks, J. (1992) A International Journal of Operations & Production
knowledge-based simulation architecture to analyze Management, 14(1), 38±49.
interruptions in a ¯exible manufacturing system. Yamamoto, M. and Nof, S. (1985) Scheduling/rescheduling
Journal of Manufacturing Systems, 11(3), 195±214. in the manufacturing operating system environment.
Welbank, M. (1983) A Review of Knowledge Acquisition International Journal of Production Research, 23(4),
Techniques for Expert Systems, British Telecom- 705±722.
munications Research Laboratories Technical Report, Zhang, Y. and Chen, H. (1999) A knowledge-based dynamic
Ipswich, England. job-scheduling in low-volume/high-variety manufac-
Wilson, J. (1985) Expert systems capture manufacturing turing. Arti®cial Intelligence in Engineering, 13(3),
brainpower. Production, 96(5), 60±64. 241±249.