Vous êtes sur la page 1sur 24

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/242928913

A review of scheduling rules in flexible


manufacturing systems

Article in International Journal of Computer Integrated Manufacturing · November 1989


DOI: 10.1080/09511928908944424

CITATIONS READS

61 24

3 authors, including:

Mahesh C. Gupta Cee Bector


University of Louisville University of Manitoba
73 PUBLICATIONS 2,030 CITATIONS 154 PUBLICATIONS 1,973 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Role of entrepreneurship in hospitals View project

All content following this page was uploaded by Mahesh C. Gupta on 19 May 2016.

The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
International Journal of Computer Integrated
Manufacturing

ISSN: 0951-192X (Print) 1362-3052 (Online) Journal homepage: http://www.tandfonline.com/loi/tcim20

A review of scheduling rules in flexible


manufacturing systems

YASH P. GUPTA , MAHESH C. GUPTA & C. R. BECTOR

To cite this article: YASH P. GUPTA , MAHESH C. GUPTA & C. R. BECTOR (1989) A review
of scheduling rules in flexible manufacturing systems, International Journal of Computer
Integrated Manufacturing, 2:6, 356-377, DOI: 10.1080/09511928908944424

To link to this article: http://dx.doi.org/10.1080/09511928908944424

Published online: 24 Oct 2007.

Submit your article to this journal

Article views: 42

View related articles

Citing articles: 27 View citing articles

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=tcim20

Download by: [University of Wisconsin Platteville] Date: 19 May 2016, At: 06:25
INT. J. COMPUTER INTEGRATED MANUFACTURING, VOL. 2, NO.6, 356-377

A review of scheduling rules In flexible


manufacturing systems
YASH P, GUPTA, MAHESH C. GUPTA and C. R. BECTOR

Abstract. In this paper, the literature dealing with the parts and Stecke (1987) further categorized the flexible manu-
scheduling problem in flexible manufacturing systems (FMS) facturing systems into dedicated FMS and random FMS,
has been reviewed. Although the parts scheduling problem is
and flexible flow systems into flexible assembly systems
Downloaded by [University of Wisconsin Platteville] at 06:25 19 May 2016

only part of a much larger decision-making process, it has


fundamental implications on the overall performance of the and flexible transfer lines.
system: internally by affecting the utilization of expensive Generally, when an FMS is being planned, the objec-
resources (c.g, machine tools and fixtures), and externally by tive is to design a system which will be most efficient in the
affecting its responsiveness to meet the changing customer production of the entire range of parts. The effective
demands. This paper has two objectives: (1) to develop a
installation requires the solution to design, production
framework within which to discuss the current literature on
dispatching rules; and (2) to compare the developed list of planning, scheduling and the actual control problems of
dispatching rules and performance criteria from the surveyed an FMS. Carrie and Pestopoulos (1985) recognized the
literature with that desired by the FMS operators. The research important management decisions that have to be made
findings arc critically reviewed and. analysed according to prior to its design and implementation. Stecke (1981)
various two-dimcnsicnal classifications. The results from real-
developed a hierarchical scheme comprising five produc-
life case studies of FMS. are also discussed. Finally, the results
arc summarized to present suggestions for future research tion planning problems which must be solved prior to an
efforts. FMS's operation. At the subsequent level, on-time sched-
uling problems have to be solved such as input sequence
determination and alternate route selection. Lastly, the
solutions to the control problems such as system main-
1. Introduction tenance, machine breakdowns, quality control and
output control need to be obtained (Nof et at. 1979, Stecke
The concept of flexible manufacturing systems (FMS) 1985). Figure 2 shows these problems and decisions in a
has brought about signifIcant advantages over traditional systematic framework.
batch manufacturing in the areas of work-in-progress A variety of analytical as well as simulation models
(WIP) levels of inventory and utilization of resources, have been proposed to deal with some of the above
specifically in the machine tools industry (Groover 1985, problems in FMSs (Suri 1985). Many excellent attempts
Hatvany 1985, Ranky 1983, Stecke and Suri 1986). have been made in consolidating research efforts dealing
Jaikumar (1986) stated that the average utilization of with these models. For example, see Buzacott (1985),
resources in such systems varied between 52% and 84% Buzacott and Yao (1986), Heragu and Kusiak (1988),
and manufacturing overheads declined by 92 %. Kalkunte et al. (1986), Stecke (1985) and van Looveren
Flexible manufacturing systems are bcing implemented et al. (1986). A number of comprehensive surveys on
in various forms. These systems can be classified across a installations of FMSs in various nations have been writ-
number of dimensions such as mode of operation, level of ten by several authors including: Dupont-Gatelmand
automation and degree of flexibility (see Fig. 1). From the (1982), Hatvany (1983), and Hutchinson (1979).
viewpoint of the scheduling procedures, Rachamadugu Browne et al. (1984) reviewed various types of flexibility
inherent in FMS such as: machine flexibility, route
flexibility, product flexibility, process flexibility, operation
Authors: Yash P. Gupru. Department of Management, and Mahesh C. flexibility, volume flexibility, expansion flexibility and
Gupur. Department of lndustrinl Engineering, University of Louisville, production flexibility. ]aikumar (1986) surveyed about
Louisville, KY 40292, USA; and C. R. Bccror. Department of
Actuarial and Management Sciences, Faculty of Management, Uni- 70 % of worldwide installed FMSs and observed that the
versity of Manitoba, wtnnipcg. Manitoba, R3T 2N2, Canada. terms flexibility and the management of flexibility 'were

0951.192X/89 13.00 © 1989 Tayler & FrancisLrd.


Scheduling rules in FMSs 357

HIGH

JOB SHOP
PART
VARIETY NC MACHINING CENTRE

FLEXIBLE MANUFAcruRING CElL

FLEXIBLE MANUFACTURING SYSTEMS

FLEXIBLE FLOW SYSTIMS

LOW L- :>

IJJW HIGH
PART VOLUME
Downloaded by [University of Wisconsin Platteville] at 06:25 19 May 2016

LEVEL OF AUTOMATION
--------------------:>
Figure 1. Classification of advanced manufacturing systems.

FLEXIBLE MANUFACTOlUNG SYSTEMS

DESIGN PLANNING SCHEDULING: CONTROL


PROBLEMS PROBLEMS PROBLEMS PROBLENS

" PRODUcr RANGE " PARTTYPESELECTION " ON-UNESCHEDUUNG " MAINTENANCE


PROBLEM PROBLEM PROBLEM PROBLEM

" MACHINE CAPACITY " MACHINE GROUPING - DISPATCHING RULES " QUALITY CONTROL
PROBLEM PROBLEM - ALTERNATE ROUTES PROBLEM
- M.<: BEAKDOWN
"TRANSPORTATION "PRODUCTIONRATIO "OUTPUTCONTROL
PROBLEM PROBLEM PROBLEM

" FlXTURING " RESOURCE ALLOCATION


PROBLEM PROBLEM

"PAUEI'SElECTION " LOADING PROBLEM


PROBLEM

" PROCESS PLANNING


PROBLEM
Figure 2. Hierarchy of FMS problems.

poorly understood. Slack (1989) studied how managers Stecke and Solberg (1981) provided an excellent review
viewed the flexibility of their manufacturing systems. He of part loading models in computerized manufacturing
found that flexibility was the least understood manufac- systems. A variety of plausible loading objectives have
turing objective. In the past few years, a number of been suggested and optimization models have been
successful attempts have been made to quantify the types formulated.
of flexibility. In a recent paper, Gupta and Goyal (1989) Kusiak (1985 b) presented a taxonomy to study mater-
synthesized the literature on manufacturing flexibility. ial handling problems related to FMSs and' Automated
Stecke (1985) suggested frameworks to study planning Storage and Retrieval Systems' (AS{RSs). Stecke and
and scheduling problems of FMSs. Buzacott (1985), Browne (1985) discussed various types of material hand-
Buzacott and Yao (1986) and van Looveren et al. (1986) ling systems (MHS) that are relevant to material handling
reviewed the planning models and discussed planning in FMS and, further, categorized such MHSs with
techniques (queueing networks, mathematical program- respect to their relative flexibility of various types.
ming, and simulation) within the FMS framework. Heragu and Kusiak (1988) suggested four basic types of
:J58 Y. P. Gupta et al.

layouts of machines and the algorithms to solve them. performance measures used and the FMS configuration
Gcrshwin et al. (1986) established a framework for study- considered. The literature is critically reviewed to draw
illg manufacturing systems issues from a control theorist's some useful inferences and suggestions for future research
viewpoint and showed how ongoing research could be in context of industrial scheduling practices are provided.
described within this framework. Gershwin (1989) pro-
vidcd an in-depth survey of the optimal control theory
approach to the control problems of FMS. 2. FMS scheduling problem
Kaner and Adelsberger (1987) reviewed the expert
systems (BS) scheduling research and commented on the Glossen and Malstrom (1982) stated that hundreds of
most comprehensively reported applications of ES in this robots and millions of dollars worth of computer-
area. Further, Steffen (1986) surveyed the literature on controlled equipment are worthless if they are under-
Al-bascd scheduling systems on the basis of historical, utilized or if they spend their time working on the wrong
methodological, application and implementation perspec- part because of poor planning and scheduling.
tives. As shown in Fig. 3, scheduling in automated manufac-
Among the problems stated above and shown in Fig. 2, turing systems can be described by a hierarchical struc-
Downloaded by [University of Wisconsin Platteville] at 06:25 19 May 2016

the FMS scheduling problems have received relatively ture ranging from top level decision making to detailed
more attention among researchers and practitioners. level scheduling decisions. The hierarchical structure of
These problems arc important because of their own FMS planning and scheduling problems has been
intrinsic value as well as their role as a crucial step in the addressed in a number of different ways such as top
implementation of such manufacturing systems. The level/detailed level scheduling (Vaithianathan 1982), and
literature on scheduling has matured to the point that a medium-term production planning/short-term produc-
number of survey and review articles have been written tion planning (Menon 1985).
(for instance, Smith et al. (1986) and Racharnadugu and The former type (i.e. medium-term planning or top
Stucke (1987». Smith et al. investigated the various desir- level scheduling) emphasizes planning for production and
able scheduling criteria preferred by the FMS users and plant operations over extended periods of time which may
the level of satisfaction of the FMS users with the include, for example, part type selection, resource (such
scheduling software in use. They found that most com- as tools and fixtures) planning and generation of sequen-
panies consider meeting due dates, maximization of ces of operations. The objective at this level is the
system utilization and machine utilization, and minimi- coordination of activities for multiple function areas. The
zation of in-process inventory being the most important output of this scheduling function is a tentative plan or a
criteria. They argued that the scheduling procedure master schedule. It sets the production targets and serves
developed by the researchers does not seem to adequately as a basis to estimate, plan for and procure resources
satisfy the needs of industry. Rachamadugu and Stecke (Walker and Miller 1986, Vaithianathan and McRoberts
(1987) defined various classes of advanced manufacturing 1981).
systems, and distinguished the employed scheduling pro- At the detailed level, scheduling (or short-term produc-
cedures with those of conventional job shop systems. tion planning) controls production over the course of each
They also provided an extensive literature review of day and provides 'a means to achieve the production
analytical and algorithmic approaches to the scheduling targets. It attempts to find the optimal routeing of jobs
problems of automated systems. Among other con- combined with making efficient use of expensive resources
clusions, they suggested that loading decisions should which are subject to environmental and procedural con-
now be considered as a part of scheduling decisions and straints. These objectives are accomplished by allocating
that system performance be evaluated when these two resources such as machine tools, fixtures and raw mater-
decisions are made simultaneously. Because of the broad ials according to time (due dates) and place (routeing)
scope of their paper, Rachamadugu and Stecke (1987) did constraints (Walker and Miller 1986). Jobs are assigned
1101 review the scheduling literature with respect to the to specific work stations (WS) on a weekly, daily or hourly
dispatching rules in detail. It is, therefore, timely and basis taking into account the type, quantity and place-
appropriate to develop a framework to synthesize the ment of resources as well as any associated time values
literature on dispatching rules. and processing priorities (Harrington 1984, Groover
In this paper, the published FMS scheduling literature 1985). At this level, the scheduling problem poses a
is reviewed from two perspectives-the performance cri- unique and challenging problem in several ways. For
reria opted for and the type of on-line dispatching tech- example (Chang et al. 1984): (1) after a work-part is
niques applied. Various studies with respect to the loaded, which machines should it visit and at what time
dispatching rules have been systematically categorized should it be processed on specific machines?; (2) if a
with respect to the type of dispatching rules applied, the machine fails, how should the schedule be adjusted for
Scheduling rules in FMSs 359

FMS SCHEDULING
I

TOP LEVEL SCHEDULING DETAILED LEVEL SCHEDULING


I

LOADING PROBLEMS ON-LINE SCHEDULING


I

HEURISTICS/ALGORITHMS SCHEDULING RULES


I[
Downloaded by [University of Wisconsin Platteville] at 06:25 19 May 2016

STATIC DYNAMIC
I I

LOCAL GLOBAL LOCAL GLOBAL


Figure 3. FMS scheduling taxonomy.

each work-part in the system?; and (3) if a high priority problem in an attempt to identify the dispatching rules.
work-part is loaded, then how should the schedule be The most general approach involves a set of related
adjusted for each work-part to accommodate the high decisions which are classified as: (1) pre-release decisions;
priority job? (2) part-releasing decisions; and (3) part-routeing
While the top level production planning problems have decisions (Fig. 4-). Several pre-releasing decisions have to be
already been extensively reviewed (Stecke 1983, Stecke made at the arrival of an order, such as (a) setting the due
and Solberg 1981b, Sarin and Wilhelm 1983, Kusiak date, (b) qualifying the order, (c) determining the priori-
1986, Menon 1985), the literature on dispatching rules is ties, (d) determining the lot sizes: and (e) setting the part
reviewed in the present paper. routeing. Part release decisions comprise two basic
elements: (a) which part to select next; and (b) when
should it be released. This is usually termed a part
3. Dispatching problems dispatching decision, i.e. selecting the part to release
next. Part routeing decisions can be made in the pre-release
Dispatching rules are used primarily to help the pro- phase or upon completion of each operation. When no
duction manager on the shop floor to make decisions alternative routes exist, the decision is normally made in
(described earlier). If such rul~s are to be of any use to the the pre-release phase whereas if alternative routeings
manager, the performance of such rules must be demon- exist, it is usually better to delay the routeing decisions as
strated against the criterion or criteria which the manager much as possible (see, for example, Nof et al. (1978),
perceives as being the most relevant. In the literature, Buzacott and Shantikumar (1980), Denzler and Boe
terms such as scheduling rules, priority rules or dispatch- (1987c».
ing rules are often used synonymously. Cere (1966) has Another approach most commonly adopted by resear-
made an attempt to distinguish between priority rules and chers to study dispatching rules involves two steps: first,
dispatching rules and also between static and dynamic either formulate a loading model using one or more of the
priority rules and dispatching rules and also between objective functions and a set of appropriate constraints
static and dynamic priority rules. Table 1 provides the and solve it, or develop some loading strategies; and
definition of these terms. second, evaluate the performance of such loading strat-
A review of published studies reveals that at least two egies using appropriate dispatching rules (see, for
approaches have been applied to the FMS dispatching example, Lin and Lu (1986), Bell and Billalis (1982),
360 Y. P. Gupta et a1.

Table I. Basic definitions (adapted from Gere (1966». Shanker and Tzen (1987), and Stecke and Solberg
(1981 a). This approach constitutes a part of the decision-
I. A priority 111le or priority function is that function which
making framework developed in the former approach
assigns to each waiting job a scalar value, the minimum of
which, among jobs waiting at a machine, determines the jobs (Fig. 4). Many decisions are assumed to be made
to be selected over all other jobs for scheduling. beforehand while solving loading problems, see Stecke
2. A scheduling rule dictates which job among those waiting for (1983), so that the impact of various dispatching rules
service is to be scheduled in preference to the others. could be analysed based on the performance criterion
'Scheduling a job' means scheduling the next operation of selected.
the job.
3. The class of priority rules we shall call random or simple
include any rule whose priority function is not a function of 4. Scheduling rules
the job file. The data derived job file may not be used, e.g.
number of operations, operation times and due dates. On In this section, an attempt is made to classify the
the other hand. there is a class of priority rules known as
surveyed dispatching rules into simple priority rules, a
complex rules which employs the data from job file in one way
or the other. . combination of simple priority rules, and other rules and
heuristics.
Downloaded by [University of Wisconsin Platteville] at 06:25 19 May 2016

4. A static priority rule or priority function is that function


which assigns to each job a scalar value before scheduling
begins and then either remains unchanged or varies in some
way independent of the schedule and of the job file. 4.1. Classification of scheduling rules
5. A dynamic priority rule or priority function is that function
We now present the scheduling rules following the
which assigns to each job a dynamic scalar value which
reflects the status of jobs from time to time as the schedule schema proposed by Panwalker and Iskander (1977). An
progresses, attempt is made to use the exact code used by them
6. A local priority rule only requires information about the job
wherever possible. The code names used for different
competing for service at a particular machine centre. rules mayor may not coincide with the names used by
other researchers. Moreover, the categorization pre-
7. A gtobal priority rule requires additional information about
jobs or machine status at other centres or waiting lines. sented here is not unique and some rules can be put into
different categories.

Part
Routetng Decisions
• Alternate Process Plans

9 Part Release' Decisions


Pre Release
Decisions
. Loading Scheduling
• Setting the Due Date Models Rilles and
• Qualifying the Order and - Algorithms
• Deterrnlnlng the Lot Size Strategies
• Assigning the Prlorities

• Setting the part Routeing

Figure 4. On-line dispatching taxonomy.


Scheduling rules in FMSs 361

4.1.1. Simple priority rules (SPR) related to time'. Select the job with highest priority
(Vaithianathan 1982).
Processing time (PT):
SI Select the job with the' shortest imminent Random Arrival Times (A TR):
operation time'. (Bell and Bilalis 1982, FIFO Select the job with 'first in, first out' (Bell
Carrie and Petsopoulos 1985, Chan and and Bilalis 1982, Carrie and Petsopoulos
Pak 1986, Chang et al. 1986, Co et al. 1985, Chan and Pak 1986, Chang et al.
1981, Denzler and Boe 1987 c, Elmaraghy 1986, Co et al. 1981, Elmaraghy 1981,
1981, Lin and Lu 1986, Schriber and Lin and Lu 1986, Nof et al. 1979,
Stecke 1986, Shanker and Tzen 1985, Schriber and Stecke 1987, Wang 1986).
Slomp et al, 1988, Stecke and Solberg LIFO Select the job with 'last in, first out'
1981 a, Vaithianathan 1982). (Denzler and Boe 1987 c, Kanet 1987,
LI Select the job with the 'largest imminent Kimemia and Gershwin 1983).
operation time' (Bell and Bilalis 1982, RDM Select in 'random order' (Choi and Mal-
Chang et al. 1986, Han et al. 1989, Shan- strom 1988, Elmaraghy 1981).
EAT
Downloaded by [University of Wisconsin Platteville] at 06:25 19 May 2016

ker and Tzen 1985, Stecke and Solberg Assign the highest priority to the job with
1981 a). earliest available time and select it (Pour-
SR Select the job with the' shortest remaining babai 1986).
o

processing time' (Co et al. 1981, Han et al.


1989, Nof et al. 1979, Stecke and Solberg Slack rules (SLK):
1981 a). S-1 Select the job with the 'least amount of
LR Select the job with the 'largest remaining slack' (available time before due date time
processing time' (Nof et al. 1979, Stecke for remaining operations) (Chan and Pak
and Solberg 1981 a). 1986).
S-2 Select the job with the least 'static' slack
Due dala (DD): (difference between due date and arrival
EDD Select the job with the 'earliest due date'. time) (Vaithianathan 1982).
OPNDD Select the job with 'earliest operation due RSTPI Select the job with the 'least ratio of the
date' (Denzler and Boe 1987 c, Pourbabai job slack time to the remaining processing
1986). time' (Vaithianathan 1982).
MDD Select the job with the 'earliest modified
due date'. (Raman et al. 1986). Machine (M):
MOD Select the job with the 'earliest modified WINQ Select the job that will go on to its next
operation due date' (Raman et al. 1986). operation at a machine that has the least
work load (Choi and Malstrom 1988,
Number of operations (NOP): Dar-El and Sarin 1986, Denzler and Boe
FOPNR Select the job that has 'fewest operations 1987 a, Lin and Lu 1986).
remaining' (Chang et al. 1986, Stecke and SMCPT Select the job that has shortest machining
Solberg 1981 a). centre processing time at the next centre.
MOPNR Select the job that has 'most operations NINQ Select the job whose direct successor
remaining' (Chang et at. 1986, Shanker machine has the shortest queue (Choi and
and Tzen 1985, Stecke and Solberg Malstrom 1988, Co el al, 1981).
1981 a).
Tool (T):
Cost rules (CR) MTM Select the job which includes 'least
Cost Select the job with 'minimum tardiness number of tool movement' (Han el al.
and processing cost' (Dar-El and Sarin 1989).
1986).
4.1.2. Combination oj simple priority rules (CSPR)
Setup lime rules (STR):
HGST Assign highest priority to 'highest setup SPT.TOT Select the job with the' highest product of
time'. Select the job with highest priority its shortest imminent processing time and
(Vaithianathan 1982). total processing time' (Slomp et al. 1988,
SGST Assign lowest priority to 'highest setup Stecke and Solberg 1981 a).
362 Y. P. Gupta et al.

SPT/TOT Select the job with the 'highest ratio of its ESTA Select the job with 'earliest starting time
shortest imminent processing time and with alternatives considered' (Iwata et al.
total processing time' (Slomp et al. 1988, 1982).
Stecke and Solberg 1981 a). EFTA Select the job with 'earliest finishing time
LPT. TOT Select the job with the highest product of with alternatives considered' (Iwata et al.
its largest imminent processing time and 1982, Siomp et al. 1988).
total processing time' (Stecke and Solberg
1981 a).
LPT/TOT Select the job with the' highest ratio of its 4.1.3. Heuristics and algorithms (H&A)
shortest imminent processing time and HMS Select the job with 'highest mean sim-
total processing time' (Stecke and Solberg ilarity' (Vaithianathan 1982).
1981 a). ALGOR'M To select the next job 'use algorithm'
MOD/MOD Select the job with the 'highest ratio of (Chang and Sullivan 1984, Chang et al.
modified operation due date to modified 1986, Dar-EI and Sarin 1986, Kanet
due date' (Raman et al. 1986). 1987, Kimemia and Gershwin 1983,
Downloaded by [University of Wisconsin Platteville] at 06:25 19 May 2016

PR/TR Select the job if the' ratio of its remaining Pourbabai 1986).
production requirement to original
requirement is larger than the ratio of
remaining production time to day's pro-
duction time' (Nof et al. 1979). 5. Information on surveyed articles
CRITICAL Select the job with the 'highest critical
ratio' (Denzler and Boe 1987 c, Raman In this section, we provide classified information on the
etal.1986). articles surveyed. It is categorized in three parts: (a)
SOTA Select the job with 'shortest operation performance measure considered; (b) dispatching rule(s)
time with alternatives considered' (Iwata used; and (c) the system configuration.
et al. 1982). Table 2 provides a summary of the classification

Table 2. Classification of literature.

SPR

Criteria PT DD NOP CR STR ATR SLK M T CSPR H&A

I. Production rate [ 1-7] [3,7) [1-5,19,20J [2,19) [4) [5-7) [23)


2. Machine/system [ 1,2,5,6,8,9) [3,10] [ 18] [ 1,3,5,8-10,20] [ 18,21] [5,6,22] [ 18,23]
utilization [9]
3. Throughput/fiowtime [1,3,8,9,11-15] [12] [3,11] [15] [1,3,8,9,11,12] [15] [9] [ 12-14,22) [ 11,24]
[ 15J
4. Meeting the due date
Mean tardiness [ 12] [ 12,13] [12] [ 12,13] [25]
Maximum tardiness [ 17] [18] [ 17] [ 18] [ 17] [18]
Others [9,16] [ 13J [9,16) [16] [9) [13 )
5. Other criteria
In-process inventory [ 15) [15] [15] [15]
Flowtime variance [12] [12] [ 12] [ 12]
Lead time [16] [ 16] [16]
Setup changeover [15] [ 15] [ IS] [15]
Output capacity [8] [8]
MHS utilization [22)

SPR. simple priority rules; PT, processing time; DO, due date; NOP. number of operations; CR, cost rules; STR, setup rules; ATR, random arrivals;
SLK, slack rules; M, machine; T, 1001; CSPR, combination of SPR; H&A, heuristics and algorithms.
References: II] Carrie and Petsopoulos (1985), [2J CO et al. (1981), [3] Elmaraghy (1981), [fJ Han a at. (1989), [5) Nor rial. (1979), (6]
Schriberand Stccke (1986, 1987), [7] Steeke and Solberg (1981 a), [8J Bell and Bilalis (1982), [9J Lin and Lu (1986), [10) Shankerand Tzen (1985),
1111 Chang rial. (1986), [121 Denzler and Boe(1987e), [13] Raman rial. (1986), [lfJ Slomprlal.(l988), [IS) Vailhianathan(1982), [16] Chan
and Pak (1986), [17) Pourbabai (1986), [18] Dar-EI and Sarin (1986), [19) Choi and Malstrom (1988), [20J Kanct et al. (1987), [21 J Denzlerand
Boe (1987a), [221 Iwata r/al. (1982), [23] Kanet (1987), [2f) Chang and Sullivan (198f), [251 Kiran and Alptekin (1985).
Scheduling rules in FMSs 363

between the performance criterion and the type of dis- 5.4. Chang and Sullivan ( 1984 b)
patching rules employed by the various studies. The
system configuration is summarized as part (c). An (a) Total flow time.
attempt is made to capture the major factors being (b) Algorithms.
considered in a particular paper in order to enhance the (c) A heuristic algorithm to generate active schedules
conceptualization of the results reported later in this for a general dynamic shop (e.g. FMS) consisting of
paper. Based on this information, conclusions are derived several groups of similar or parallel work stations is
with respect to the types of manufacturing systems proposed. Significant transfer times exist between work
studied (for example, dedicated and random FMS) and stations. Also, a particular operation can be performed by
the various factors considered while performing a simula- different work stations with different operation times.
tion study (for example, secondary resources and job
characteristics) .
5.5. Chang et al. (1986)

5 1. Bell and Bilalis (1982) (a) Mean flow time, throughput.


Downloaded by [University of Wisconsin Platteville] at 06:25 19 May 2016

(b) SI, LI, FOPNR, MOPNR, FIFO, Algorithm.


(a) Machine utilization, throughput times, and the
(c) A combined network and discrete-event simulation
output capacity. model of a hypothetical FMS is developed. The system
(b) SI, LI, FIFO.
consists of four work stations, three part types, one
(c) Digital simulation of a hypothetical FMS is used. transfer line, and 20 pallets.
The system consists of two lathes, one drill and I
conveyer line. Handling devices dedicated to only one
machine tool are used. Each pallet can carry one to five
parts, from One batch only. The average batch size is ten 5.6. Choi and Malstrom (1988)
and its maximum value over a hundred. A new batch
starts when all the parts from the previous batch have (a) Actual system effectivity, total travel time of parts,
been machined. production output, throughput time, work-in-process,
and production lateness.
(b) Random, FIFO, WINQ, NINQ.
5.2. Carrie and Petsopoulos (1985) (c) An actual FMS was modelled using actual data. All
process sequences and times are based on actual data
(a) Production rate, system utilization. obtained from an FMS recently installed by a midwestern
(b) SI, FIFO. manufacturing facility. Seven part families are con-
(c) An FMS commissioned by a Scottish engineering sidered. The physical simulator is used to evaluate work
company consisting of five similar CNC horizontal scheduling rules. The simulator consisted of the following
machining centres and one special horizontal machining major facilities: an AS/RS with seven parallel storage
centre is simulated. All the machines have a tool maga- areas for raw materials and one area for semi-finished
zine with a capacity of a hundred tools. Castings are fixed parts, an AS/RS cart, six parallel and identical NC
and moved on pallets by an AGV following a wire guide machining centres, a turning cell (including a robot, two
in the floor. At the load/unload area and at each machine vertical NC lathes, and a washing station), overhead
there are two pallet stands acting as buffers. Currently, conveyors and many other necessary facilities.
seven part types are involved, with operation times
varying between 14 and 286 minutes, and the total
processing time varying between 832 and 1314 minutes. 5.7. Co et al. (1988)

(a) Throughput rate.


5.3. Chan and Fak (1986) (b) SI, SR, FIFO, NINQ.
(c) The computer simulation is used to evaluate the
(a) Cost of tardiness, makespan, average lead time. performance of sequencing rules on the throughput of a
(b) SI, FIFO, S-1. FMS when the queue-length is small. The closed network
(c) Digital simulation in Fortran of a hypothetical FMS modelling approach is used to simulate the FMS. The
is used. The system consists of four machining centres hypothesis tested by them is that in closed systems of short
connected through a fixed conveyor system. Each queue-lengths, the mean flow times of jobs processed is
machining centre has a local storage buffer. insensitive with respect to the sequencing rules.
364 Y. P. Gupta et al.

5.8. Dar-El and Sarin (1986) 5.12. Han et al. (1989)

(a) Machine utilization, minimum job tardiness. (a) Throughput rate.


(b) Cost, WINQ, Algorithm. (b) LI, MTM, SR, RAN.
(c) Digital simulation in PLI of Ingersol Rand's FMS (c) The FMS configuration consists of several machin-
in Roanoke, Virginia, is used. The system consists of two ing centres, each of which has a tool magazine, a tool crib
milling, two boring, two drilling machines connected with which stores not assigned tools, and a tool transfer device.
each other through a fixed conveyor system. Each No material handling is required because each part visits
machine has one pallet storage space associated with it. only one of the machining centres for its complete
There are 25 part types to be processed and each part type processing. Therefore, the major considerations involve
may have alternative routing. the initial tool loading and subsequently, the transfer of
tools either from one machine to another or from tool crib
to another machine.

5.9. Denzler and Boe (1987a)


Downloaded by [University of Wisconsin Platteville] at 06:25 19 May 2016

5.13. Iwata et al. (1982)


(a) Machine utilization rate.
(b) WINQ. (a) Makespan, mean flow time, machine utilization,
(c) A dedicated FMS which manufactures eight types MHS utilization.
of transmission housings is simulated. The system is (b) SOTA, ESTA, EFTA.
comprised of five head indexes, II module machining (c) An FMS is modelled as a system consisnng of
centres, two load stations and two unload stations. The machine tools, a loading/unloading station, buffer
system has 37 carts and 51 pallet fixtures. Each pallet storages, and material and cutting tool transportation
fixture is designed to position one of the eight parts. Each system. The system consists of three machine tools which
machine has the capacity to store two parts before and can perform several different kinds of machining opera-
two parts after processing on its machine shuttle. The tions, two types of buffer storages installed in front of each
automatic transmission cases to be machined are attached machine tool, two carts which can move along a track
to pallet fixtures and then moved by carts to the various in both directions, a loading station and an unloading
machines on a dual loop fixed-track tow line system. station, and a cutting tool transportation system. Each
buffer capacity is equal to a unit. Ten different kinds of
part are manufactured. Each part is processed through
ten processing stages according to a specified sequence,
and two alternative machine tools are permitted for each
5. 10. Denzler and Boe (198 7 c)
processing stage of a part.

(a) Mean part flow time, part flow time variance,


percentage tardy jobs, mean tardiness. 5.14. Kimemia and Gershwin (1983, 1985)
(b) SI, OPNDD, LIFO, CRITICAL.
(c) A hypothetical random FMS with a relatively (a) System utilization, production rate.
smaller production batch is considered. The actual pro- (b) LIFO, Algorithm.
cessing times are randomly generated upon arrival. (c) While a four-level control structure to compensate
for work station failures and changes in part requirements
is proposed, a simulation model is developed for a flexible
transfer line. It consists of two work stations and down
5.11. Elmaraghy (1981) stream buffer. Each station has an initial buffer with a
capacity for five pieces.
(a) System utilization, throughput time, production
rate. '
(b) SI, FOPNR, FIFO, RANDOM. 5.15. Kiran and Alptekin (1985)
(c) A FORTRAN based simulator FMSSIM is
developed. The simulator is a modular, user-oriented (a) Mean tardiness.
package which allows the designer to evaluate a wide (b) Algorithm.
range of systems with varied design parameters and select (c) A scheduling system for an hypothetical FMS is
an efficient configuration. considered for which its actual state, and a set of parts
Scheduling rules in FMSs 365

with known processing time requirements and due dates 5.20. Schriber and Stecke (1986)
are known and given. It determines the start and comple-
tion time of operations of each part to be produced. (a) Machine utilization, production rate.
(b) SI, FIFO.
(c) The hypothetical FMS considered consists of one
milling machine, two drilling machines, and two vertical
5.16. Lin and Lu (1986)
turret lathes. The AGVs are used as a transportation
device. It has a centralized buffer which can be used by
(a) Machine utilization, mean waiting time, mean flow
any part type. The model does not take into account the
time, and number of late jobs.
geometric layout of the FMS. All part types visit
(b) SI, FIFO, WINQ.
machines in a mill-drill-lathe sequence and no machine
(c) A hypothetical FMS is simulated which consists of
type can perform substitute functions.
four machining centres, a line transportation cart and a
load/unloading area.

5.21. Shanker and Tzen (1985)


Downloaded by [University of Wisconsin Platteville] at 06:25 19 May 2016

5.17. No! et al. (1979)


(a) Machine utilization, number of late jobs.
(b) SI, LT, MOPNR, FIFO.
(a) Machine utilization, production rate. (c) A hypothetical FMS is simulated. It has m machines
(b) SR, LR, FIFO, PR/TR. and each tool magazine has a known tool slot capacity.
(c) A hypothetical versatile manufacturing system (a The jobs arrive in random order. For each job the
prototype of an actual facility) is simulated. The system processing time, tool slot requirement and due dates are
consists of three pairs of machining centres intercon- known.
nected with a loop conveyor. A large number of parts can
be produced, each of which may have up to four alternat-
ive process plans. In the analysis different sub-sets of part
5.22. Siomp et al. (1988)
types are considered.
(a) Makespan, mean flow time.
(b) SI, SPT.TOT, SPT/TOT, EFTA.
5.18. Pourbabai (1986) (c) The FMS consists of a load station, an unload
station and two machining centres. The machining
(a) Maximum tardiness. centres can perform the same processes. However, one
(b) OPNDD, EAT, Algorithm. machining centre is more efficient than the other and
(c) A mixed binary LP model for a hypothetical flexible consequently the processing times are smaller for one
batch manufacturing system consisting of a set of work machining centre. There are local buffers in front of each
stations; loading and unloading stations linked by a MHS machining centre with the capacity of two parts. One
is suggested. Each job is assumed to be a batch of transport device and two operators are assumed. The
homogeneous parts, which may require more than one moving times of the secondary resources are also con-
stage of operation at each work station. The due date and sidered.
availability time of each job can be specified.

5.23. Stecke and Solberg (1981 a)


5. 19. Raman et al. (1986)
(a) Production rate.
(a) Mean tardiness, proportion of tardy jobs, mean (b) SI, LT, SR, LR, FOPNR, MOPNR, SPT.TOT,
flow time, standard deviation of tardiness. SPT/TOT, LPT.TOT, LPT/TOT.
(b) MDD, MOD, MOD/MOD, CRITICAL. (c) The Caterpillar FMS is simulated. The system
(c) The digital simulation of a hypothetical FMS IS consists of four large five-axis machining centres, three
developed which is comprised of five machining centres four-axis machining centres, two vertical turret lathes and
and a cart. Twenty part types are considered, arriving an inspection station. Each machine has a limited tool
at the system randomly. There is a load/unload station magazine capacity. The transporters run on a straight
where incoming jobs are received and to which the track and carry parts from machine to machine. Sixteen
finished jobs are routed. stations and a load/unload station is provided.
366 Y. P. Gupta et al.

5.24. Vaithianathan (1982) research on dispatching rules in an FMS environment is


still at an early stage. In spite of the fact that the on-line
(a) Throughput time, setup change over, in-process scheduling function has a fundamental impact on the
inventory. overall system utilization, no fr<{mework has been pro-
(b) SI, HGST, SGST, S-2, RSTPI, HMS. posed in the literature to study this problem. Recognizing
(c) Digital simulation of a hypothetical static FMS is the abundance of information available on the scheduling
used. The system consists of n part types and M machin- rules and frameworks in conventional production systems
ing centres. Each part type requires a number of opera- (for example, Panwalker and Iskander 1977, Blackstone
tions to be performed on different machining centres. et al. 1982), we anticipate that there will be an increasing
number of studies conducted in this area of FMS in the
near future. Therefore, it will be helpful to provide a
5.25. Wang (1986) framework to direct the future studies in this area. The
need for elaborate studies in FMS scheduling has recently
(a) Production rate/output rate. been recognized and many ongoing research projects
(b) FIFO. have been reported in the recent conferences on FMSs
Downloaded by [University of Wisconsin Platteville] at 06:25 19 May 2016

(c) Three hypothetical FMS configurations are sim- (e.g. Denzler and Boe 1987 c, Chang et al. 1986, Morton
ulated. Each consists of one machining centre, two and Smunt 1984).
machining centres, and three machining centres respect- The analysis of the performance measures addressed in
ively. Each machining centre has its buffer storage. Five the literature is summarized in Table 3. It can be seen
different part types are produced in each system configur- that among the most intensely studied criteria are produc-
ation and each part type has a unique routeing associated tion rate, machine/system utilization and throughput
with it. time or flow time. A number of attempts have been made
addressing the due date related evaluation criteria, how-
ever, there is only a limited number of studies dealing
6. Discussion with a specific criterion within this category (Table 2). In
the following section, the results obtained by several
From the review of the literature, it IS clear that researchers are discussed.

Table 3. Scheduling criteria (ranking from literature).

Criteria
Ranking (referred in References) Articles'[ COUnl

I. Maximizing machine/system utilization [1-11] 12


2. Minimizing throughput/mean flow time [1,2,5,6,8,12-17] 11
:1. Maximizing production rate [2,5,7,9,10,18-21] 10
4. Lateness related criteria: 7
Minimize mean tardiness [15,22,23]
Minimize maximum tardiness [3,24]
Others [8,25]
5. Othn- criteria: 6
Minimizing in-process inventory [17J
Minimize flow time variance [14]
Minimize average lead time [25J
Minimize setup changeover [17J
Maximizing output capacity [1]
Maximizing MHS utilization [6)

til] Bell and Bilalis (1982). [2] Carrie and Petsopoulos (1985), [3] Dar·EI and Sarin(1986), [4] Denzler
and Boe (1987 a), [5} Elmaraghy(1981), [6} Iwata er al. (1982). [7] Kaner (1987), [8] Linand Lu(1986),
[9J Noret al. (1979). [10] Schriber and Stocke (1986. 1987), [11] Shanker and Tzen (1985), [12J Chang
and Sullivan (1984), [13] Chang et al, (1986), [14] Denzler and Boe (1987 e), [15] Raman et al. (1986),
[16J Siomp et oi. (1988), [17} Vaithianathan (1982). [18J Choi and Malstrom (1988), [19J CO et ai.
(1981), [20] Han el al. (1989), [21] Stecke and Solberg (1981 a), (22) Diesch and Malstrom (1985), [23]
Kiran and Alptckin (1985). [24] Pourbabai (1986), 125] Chan and Pak(1986).
Scheduling rules in FMSs 367

6.1. Results related to the production rale criterion studied individually for developing efficient loading strat-
egies and dispatching rules.
One of the earliest studies on the dispatching problem In another study, Carrie and Petsopoulos (1985) con-
is by Nof el al. (1979). They explained the signif- ducted a number of simulation experirncnts to assess the
cance/impact of various planning and scheduling issues effects of job sequencing and work balancing or unbalan-
such as part type selection, part mix ratio and alternate cing in an existing FMS. Job sequencing was done in two
routeing. Their simulation study assumed that the part steps. Several launching (i.e. part-releasing) rules were
type selection and part mix problems have been solved proposed and dispatching rules (FIFO and SI) were used
a priori, and examined the effect of two aspects of parts to select a part from the internal queue. The results led to
scheduling (i.e. entry rules and dispatching rules) on the conclusion that while there is potentially a significant
system performance in a breakdown free FMS. The entry gain in system performance by increasing the flexibility
of parts in the system was divided into two types, initial due to the allocation of operations to machines, none of
entry and general entry. Several entry rules were invest- the other variables, including launching rules and queue
igated but none was shown to dominate the others. discipline, appear to have any significant effect on the
However, in a previous study, Nof el al. (1978) concluded performance of the system of this FMS.
that the initial entry rules do affect the system's perform- Schriber and Stecke (1986) provided an excellent simu-
Downloaded by [University of Wisconsin Platteville] at 06:25 19 May 2016

ance in the long run. In allocating work to machines, the lation study of a dedicated FMS. The paper documented
SR, LR, FIFO and PR/TR dispatching rules were tested. various modelling parameters pertinent to a FMS simula-
The ratio rule (defined earlier) was found to be superior tion study, These parameters were further categorized by
than SR, LR and FIFO when the performance was Schriber and Stecke (1987) (see Table 4). The authors
measured in terms of production rate and machine successfully demonstrated the validity of the theoretical
utilization. However, it may be noted here that the results and further, conducted a variety of experiments to
classical statistical significance tests were not conducted to
validate the conclusions. Moreover, it is not clear how
their model handled the tool loading problem. Since four
alternative routes were assumed for each part type, it may Table 4. Factors for FMS simulation models.
be safe to infer that all machines were loaded with the
t. FMS configuration
required tools. In other words, the tooling did not present
Dedicated FMS
a bottleneck. Finally, no part entry rules or dispatching Random FMS
rules emphasizing due date related criterion were con-
2. Secondary FMS resources
sidered.
Number of:
A subsequent study dealing with the effectiveness of Buffers spaces-internal/or central
dispatching rules in an FMS environment was published Pallets and fixtures
by Stecke and Solberg (1981 a). In their scheduling Loading/unloading station
procedure, each operation was assigned to an individual 3. Geometric considerations
machine tooled to perform that task, thereby creating a Location of:
fixed routeing for each part. In their testing of loading Loadingjunloading stations
Machining centres
strategies and dispatching rules, they created alternate
Alternative routcings
routeings by assuming increased tooling flexibility. In
4. Secondary time requirements
total, five loading strategies were tested against 16 dis-
Transfer times
patching rules. For all loading strategies the dispatching Palletizingjdepalletizing
rule, SPT!TOT, performed best. For the specific Fixturing, dcfixturingjrefixruring
dedicated FMS, SI alone performed worse than the
5. Operating procedures
average rule. A loading strategy where all machines of a Buffer capacities
kind were pooled together, in combination with the Sequencing rules
dispatching rule SPT!TOT produced the overall Part input sequence
maximum output in the 80 simulation combinations 6. Operating discontinuties
compared. Although the classical statistical significance Machine/tool breakdowns
tests were not conducted to validate the conclusions, the Scheduled maintenance
results were compared with those of real system and Machine substitution
significant improvements were observed. These results 7. Secondary job characteristics
were 8-24% better than those obtained with the plant'S Due dates
Precedence constraints
existing scheduling rules. One important observation
Fisturingfrefixturing requirements
made by the authors was that each system should be
368 Y. P. Gupta et al.

test the sensitivity of parameters such as the number of machine group configuration had a significant impact on
AGVs, the number of buffers and the level of WIP. The system performance and that MHS had a moderate
dispatching rule employed was SPT. The study provided impact, but that the demand fluctuations had no impact
a complete and detailed explanation of the statistical on the system's output. However, no consideration was
aspects of the results and thus, confirmed the validity of given to the selection of appropriate dispatching rules.
the results. Thus it seems that further investigations should be car-
Based on the analysis of results, some of the conclusions ried out to derive some conclusions.
derived by Schriber and Stecke (1986) are as follows: (a) Choi and Malstrom (1988) analysed an actual FMS
when balanced FMS production ratios are used, machine using physical simulation. The performance of seven part
utilizations, production rates, and system residence times selection rules and four machine centre selection rules was
vary signilicantly with the level ofWIP and the number of investigated. The combination of SPT and WINQ (SPT
AGVs used to transport WIP; (b) machine utilization and rule as the part selection rule and WINQ as the machine
production rates increase more rapidly with increasing selection rule) performed well. Each part type was
levels of WIP than with increasing numbers of AGVs; assumed to require two operations. Each part type
and (c) mean residence time and its viability tend to required to visit the lathe cell for the first operation and
decrease somewhat as the number of AGVs increases, the second operation could be performed by any of the six
Downloaded by [University of Wisconsin Platteville] at 06:25 19 May 2016

and increase noticeably as the level of WIP increases. machine centres. Several machine selection rules were
The problem of tool allocation was not addressed by tested. It is not clear whether the results could be
Schriber and Stecke (1986), it may be assumed that all the generalized to cases where more than two operations
machines in a particular group (for example, two drills) needed to be performed on the part type (a more realistic
were loaded with the tools required to operate on all part situation). However, this paper provided a new modelling
types (i.e. tooling flexibility). An analysis of the results approach using a physical simulator (discussed later).
obtained for the two major performance measures
(machine utilization and production rate) seemed to have
a linear relationship with each other in all the experiments 6.2. Results related to the machine/or system utilization criteria
conducted. The number of other dispatching rules
addressing due date related performance measures may Bell and Bilalis (1982) suggested five order release rules
be further investigated. and tested their performance using FIFO, SI and RI
Schriber and Stecke (1987) further extended their dispatching rules against five different configurations of
previous study and provided a comparison of FIFO and the systems. No machine breakdown was assumed. Their
SPT scheduling rules. However, the major focus of this conclusion was that the combination of SI with all the
study was to demonstrate the intimate relationship order release rules showed a considerable improvement in
between mathematical programming and simulation system utilization. The results presented in the study
approaches to the production planning and scheduling appeared to correspond to a single simulation run for
problems of FMSs. It was shown that since the math- each combination. They did not provide any statistical
crnatical programming approach provided a solution in validation of the results. The authors suggested that
an aggregate sense (i.e. secondary resources, job char- though the conclusions were drawn using a simple hypo-
acteristics and palleting and fixturing considerations are thetical FMS, similar trends were observed on a larger
ignorcd), the simulation modelling should be used as a and more complex system. "
second step while analysing the machine utilizations at Kimemia and Gershwin (1983) developed a multilevel
the disaggregate level. The authors provided a com- hierarchical control algorithm for the control of produc-
prehensive analysis with respect to machine utilization tion in an automated manufacturing system with unreli-
and production rate performance measures while using able machines. They divided the control into a hierarchy
simple scheduling rules. Again, it would be a worthwhile consisting of a number of different levels. Each level was
attempt to extend their study to incorporate due date characterized by the length of the planning horizon and
related performance measures and dispatching rules. the kind of data required for the decision-making process.
Wang (1986) examined a complex FMS where several In this study, an FMS control policy was proposed
factors (such as machine group configurations, material whereby long-term production objectives were met by
handling systems (M H S) and demand fluctuations) sim- loading the parts in a way that would not overload the
ultaneously affect the system's performance. The experi- system or cause congestion. The proposed algorithm
ments with different combinations of settings were involved a stochastic optimal control problem at the first
analysed using the various statistical models (one-way level. The computational scheme was verified using simu-
analysis of variance, multi-factor analysis of variance and lation. The simulation model consisted of two stations
multiple regression analysis). It was concluded that the with two machines in each station and an internal buffer
Scheduling rules in FMSs 369

with a capacity of five pieces. At each station's internal of the analysis based on the results obtained for only one
buffer the LIFO dispatching rule was employed. High shift still needs to be examined carefully. It may be
machine utilization rates were recorded for station A but interesting to extend the proposed loading algorithms to
only medium machine utilization rates at station B. The take the subsequent periods into consideration. Also, the
explanation for such differences between the two stations implications of incorporating machine breakdowns in the
was that the system tended to develop a bottleneck at heuristic provides another possible direction for future
station A. No efforts were directed in testing some other research efforts.
dispatching rules because the purpose of their present Shanker and Tzen (1985) formulated five loading
study was to propose a control methodology for the strategies: two strategies based on the results of the
short-term production planning which explicitly takes optimization model, two based on heuristics and one
repair and failure into account. We shall elaborate on this based on random loading. However, they tested only two
hierarchical approach in the next section on industrial strategies, i.e. random loading and balancing the work
scheduling practices. load (a heuristic algorithm) against four dispatching rules
Dar-El and Sarin (1986) and Sarin and Dar EI (1986) FIFO, SI, LI and MOPNR to obtain the best system
developed a heuristic procedure for scheduling parts performance for a random FMS. It may be noted here
Downloaded by [University of Wisconsin Platteville] at 06:25 19 May 2016

using WINQ dispatching rule. Their heuristic resulted in that their heuristic algorithms are strikingly similar to the
fairly high levels of machine utilization. The authors ones proposed by Lin and Lu (1984). The results of the
suggested that sophisticated modifications of the simulation model revealed that balancing the work load
algorithm would result in even higher machine utilization provided better average machine utilization, no matter
levels. The study did not address how the tool allocation which dispatching rule was followed. Because of the
problem was solved (i.e. whether both machines in each limited number of simulation runs, no final results about
of the three groups were allocated the same tools). the efficiency of the dispatching rules were provided.
Machine breakdowns were not considered. The proposed However, based on the initial runs, it was concluded that
approach seems to become more cumbersome in finding a the LI priority rule worked least well. The MOPNR rule
schedule, when machine breakdowns were considered (a performed best for L3, i.e. when the system had a better
more realistic situation). Rachamadugu and Stecke workload balance. However, the SI rule performed best
(1988) provided a review of this study in more detail. on average. Again, the statistical validation of the results
Lin and Lu (1984) proposed two heuristics for machine was not provided and a number of possible extensions are
loading with the following objective functions: (a) bal- discussed in the critical study of Lin and Lu (for more
ance workload (BWL), and (b) balance workload and detailed comments on this study, see Rachamadugu and
minimum number of late jobs (BWL-MNLJ). The load- Stecke (1988».
ing algorithms were solved for only one time period (say, Denzler and Boe (1987 a,b) tested the effectiveness of
one shift of 8 h). The loaded jobs were scheduled on the scheduling heuristics developed for an existing dedicated
system using various scheduling rules. The authors col- FMS. The scheduling of parts, carts, and machines was
lected the statistics for the selected time periods. These controlled by software package which consisted of three
loading algorithms along with other commonly used decision rules: (a) part load rule; (b) part launch rule; (c)
strategies such as 'earliest due date' (EDD) and 'first part routeing rule. Their heuristic scheduling approach
come first served' (FCFS) were tested against SI, FIFO, proposed two types of part loading rules (three order file
and a global dispatching rule, WINQ. Using the variance oriented rules and three system/machine oriented rules),
of machine utilization as the performance measure, the one launching rule (36 pallets availability), and one
BWL policy performed better than the FCFS policy. routeing rule, WINQ.
When due dates were employed, the BWL-MNLJ policy The performance was tested using various experimen-
performed better than the EDD policy when FIFO tal factors such as the number of pallets and machine
dispatching rule was employed. In the case of SI and flexibility. The objective was to investigate which of the
WINQrules, the effect of the BWL-MNLJ policy was not part loading rules would perform best in scheduling the
obvious. It was concluded that the system's performance FMS when it is subjected to different conditions of pallet
with respect to any of the loading policies had improved availability and machine flexibility. The statistical anal-
significantly by using the global rule WINQ. It may be ysis of the simulation results was performed. Some of the
noticed that the paper did not provide any details with conclusions derived by the authors were as follows: (a)
respect to the statistical validity of the results obtained two part loading rules ((i) SPJL, i.e. load the part which
except that ten production runs were simulated. From the has the smallest proportion of job launched, and (ii) NEP,
proposed approach, it was not clear how the situation was i.e. as each pallet is unloaded, reload it with a like
handled after the selected time period was over (i.e. product, if possible) gave the highest utilization rates; (b)
loading in the next time period). Moreover, the validity machine flexibility was a significant experimental factor;
:170 Y. P. Gupta et al.

(e) the number of pallets was not a significant experimen- examined, i.e. makespan, mean flow time, mean utiliza-
tal factor; and (d) the order file oriented loading rules did tion of machine tools and carts. It was concluded that the
not appear better or worse than the system oriented rules. combination of the ESTA rule for the selection of
The authors pointed out that the part loading rule that machine tools and transport devices yielded the
Stucke and Solberg (1981 a) reported as performing best maximum values of mean utilization of the machine tools
in their study did not perform as well as NEP and SPJL. and transport devices, and that the combination of the
However, it may be noted that the performance measures EFTA rule for the selection of machine tool and transport
in these two studies were different. Stecke and Solberg devices yielded the minimum values of the makes pan as
considered the production rate where as in this study, well as mean Oow time. The paper did not provide details
the performance measure was machine utilization. In of the computational approach or statistical validation of
general, however, there seems to be a linear relationship the results, and the machine breakdowns were also not
between these two measures (see, for example, Schriber considered. Overall, the results from the proposed model
and Stocke (1986)), and if this is true then we may provided a useful insight into FMS scheduling. Although
compare the above two studies. a large number of studies have been conducted in the
The above analysis was performed under the assump- recent past, it is surprising that no further studies were
Downloaded by [University of Wisconsin Platteville] at 06:25 19 May 2016

tion of certainty, i.c. no machine breakdowns or failures reported on the same line.
or carts etc. was allowed at the shop floor level. In Denzler Chang et al. (1984) proposed a two-phase approximate
and Hoc (1987 b), the former study was extended to method that incorporated global information for on-line
include uncertainty factors. A complete statistical test was scheduling of FMSs. The proposed algorithm was com-
performed by thc authors. Again, NEP and SPJL main- pared with SI, LT, FIFO, MOPNR and FOPNR. From
tained their superiority over other part loading rules. The the experiments it was observed that: (a) the two-phase
results did not show a significant decrease in the machine approximate method outperformed the tested dispatching
utilization rate other than that caused by the direct rules in terms of throughput and mean Oow times (in fact,
reduction in net system capacity caused by machine it resulted in a 39% increase in throughput over the next
breakdowns. Although the above study provides an best rule); and (b) the rule SI was slightly better than the
extensive and complete analysis of the scheduling pro- other rules in terms of throughput. However, the disper-
cedures employed in an existing FMS, some issues that sion of mean flow time for various part types was high
were not addressed by the study but which could be of for Sf. FIFO yielded the most uniform throughput
interest in general arc: (a) it does not explain how the among part types (for more detailed comments, see
part type selection, tool allocation and part mix ratio Rachamadugu and Stecke (1988)).
problems were addressed, and how analysis of the effects Vaithianathan (1982) developed loading scheme based
of the solutions to these problems on proposed scheduling on the concepts of group technology. Similarity measures
procedures would be of interest; and (b) the machine and clustering were used to produce subgroups of jobs
grouping problem was not addressed (specifically, in the that resulted in level loads on all the machines (taking
second study where machine breakdown was considered) capacity constraints into consideration). Once the sub-
and it would be interesting to analyse the effects of groups were obtained, several rules such as SI, HGST,
grouping heuristics on system performance. Since the SGST, S-2, RSTPI and HMS were employed to test the
above study was performed on an existing FMS, we may performance of the loading scheme. The scheme when
usc the idea of logical machine groupings as suggested by tested for lower throughput times against SI, HGST,
Srcckc (1983). SGST, S-2, RSTPI and HMS, showed that all dispatch-
ing rules performed better than Sf. However, a detailed
analysis with respect to ranking the remaining rules was
not provided. Also, no statistical tests were reported in
6.3. Results related to throughput/or }low times criteria the study. As mentioned by the author, in order success-
fully to implement this algorithm to obtain a balanced
Iwata et al, (1982) considered the problem of determin- load, the original set of orders to be scheduled must be
ing the schedules for machining and transporting parts, large and diverse. The proposed approach gave no
and for transporting cutting tools simultaneously so as to consideration to due date related measures.
minimize the makespan of production. The model cap- Recently, Chan and Pak (1986) proposed two heuristic
lures most of the features pertaining to an FMS. A algorithms (HI and H2) for the FMS scheduling problem.
heuristic procedure based on three decision rules (i.e. For the due date related problem, the heuristics were used
SOTA, ESTA and EFTA) was developed. To investigate to obtain a schedule such that the respective due dates
the effects of these rules, computational experiments were were met or, failing this, the cost of tardiness was
carried out. Four kinds of scheduling performance were minimized. For the makespan and average lead time
Scheduling rules in FMSs 371

related problems, the algorithms found an optimal due proposed heuristic to incorporate both of the conflicting
date for each job. The authors showed the effectiveness of objectives, i.e. minimize the tool movement and balance
the algorithms using a numerical example of job shop the workload. Further, rigorous statistical tests (such as
adapted from Nicholson and Pullen (1971). factorial designs) should be performed to validate the
The simulation results showed that the proposed heur- results.
istics outperformed the other scheduling rules, namely Slomp et al. (1988) proposed three quasi on-line sched-
FIFO and SI. In the case of a dynamically loaded FMS, uling procedures for FMSs consisting of machining
the proposed heuristics were accordingly modified to centres (WS), transport devices (TD) and operators. The
adjust for due date. The conclusion was that the perform- procedures are based on an activity-model and are imple-
ance of HI, when it is fully or partially adjusted for due mented by executing four distinguished functions: (a)
dates, ranked highest among all scheduling rules tested. choosing a WS upon which one or more processing stages
The study neither mentioned how the tool allocation might be executed; (b) selecting a TD for the transport
problem was addressed nor was a statistical analysis activities of an operation; (c) assigning an operator to one
performed to validate the application of heuristics in FMS or more activities of an operation; and (d) determining
environment. Moreover, no computational experience which operation must be placed in the scheduling
Downloaded by [University of Wisconsin Platteville] at 06:25 19 May 2016

was reported in the paper. Thus, for practical purposes, it sequence next. The three procedures called Function
was not clear how the results could be used in an Sequential Scheduling (FSS), Function Integrated Sched-
integrated production and scheduling problem. uling (FIS), and Function Phased Scheduling (FPS),
Han et al. (1989) presented a different but related tool differed in the way the four abovementioned functions
loading and job dispatching problem. In this problem, a were handled hierarchically.
part visits only one of the machining centres for its The performance of the procedures were compared
complete processing. The initial assignment of the tools against a number of scheduling rules (such as SPT,
was made optimally and then, as a real-time control, the SPT.TOT, SPT!TOT and EFTA) using computer simu-
tools which were required by a machine but not present, lation. The system performance is presented by means of
were borrowed either from another machine if not in use makespan and mean flow time. They concluded that the
or from the tool crib if available. The authors argued FPS procedure was inferior as compared to others. While
about the practicality of such an approach in the defence the SPT!TOT performed better on makespan measure, it
industry in which small-size, high-precision metal parts performed worst on the mean flow time measure of
are produced. The throughput performance measure was performance. This paper presented a different approach
to minimize the tool movement at prescribed workload to the FMS scheduling problem. The case study consisted
imbalances at the machining centres. of a rather small FMS and it is not clear how the part type
A non-linear integer programming model was selection problem was addressed. For example, no justif-
developed to load the tools and parts simultaneously. In cation was given for selecting the particular part input
order to solve the computationally difficult problem, they ratio and no simulation was performed to test its effect
proposed an optimization-based approximate solution on the system performance. Further, the paper did not
method which resulted in two problems, i.e. part loading address the tool loading problem. For example, while
and tool loading. Next, these two problems were solved implementing the scheduling procedures on more realistic
iteratively. A greedy heuristic rule was also proposed and systems, is it practical to assume that all the machining
the performance was compared. Based on the analysis of centres have the tools required to perform all the required
two problem sets, the optimization-based method reduced operations? The conclusions derived did not appear to be
the number of times tools were borrowed. However, the justified. For example, the paper did not explain why the
computation time also increased sharply as the problem SPT!TOT provided opposite results for the two measures
size increased. Using the results from the tool loading considered in the paper. Also, comparison of the results
problem, a computer simulation was performed to ana- with the study of Stecke and Solberg (1981) does not seem
lyse various other issues such as the impact of: (a) various appropriate since the measures of performance con-
factors of workload imbalance; (b) central queue versus sidered are different.
individual queues; (c) various queue disciplines LPT,
MTM, SPT, RAN; (d) returning borrowed tools to the
tool crib immediately after their use or when there is a 6. 4. Results related to due dale
request for them in another machine. Based on the
experimental results, the heuristic tool-loading method
with the 'no tool-return unless requested' policy was 6. 4.1. Mean tardiness
recommended, together with any dispatching rule. As a
possible extension, it would be interesting to modify the Raman el al. (1986) concluded that an increase In the
372 Y. P. Gupta et al.

relative imbalance of machine workloads led to a deterior- uling practices. They concluded that the literature did not
ation in mean tardiness when the MOD dispatching rule correspond to the needs of the practioners.
was employed. They also formulated two more rules, (a) In this section, we will summarize some of the con-
MOD/M DO; and (b) CRITICAL and tested their per- clusions and findings from case studies of real life FMSs.
formance. The preliminary investigations indicated that The choice of applicable loading and scheduling strat-
both rules outperformed the MOD rule for mean tar- egies depends on many system-specific variables (Stecke
diness. They also concluded that an increase in the and Solberg 1981 a). Which combiantion of loading and
relative imbalance of machine workloads leads to a dispatching strategies might be best is highly system
deterioration in the mean flow time when the MOD rule dependent. Denzler and Boe (1987 a) observed that the
was employed. part loading rule that Stecke and Solberg (1981 a)
Kiran and Alptekin (1985) proposed an algorithm in an reported as performing best did not perform as well in
attempt to minimize tardiness. However, no experimen- the system they studied. Since the system simulated by
tal results were reported in the study. It was claimed that Denzler and Boe was significantly different from the one
the algorithm was flexible enough to incorporate other simulated by Stecke and Solberg, the differing results
scheduling criteria and more general cases. imply that the loading strategies and the appropriate
Downloaded by [University of Wisconsin Platteville] at 06:25 19 May 2016

dispatching rules should be tailored for each type of


flexible manufacturing system. Moreover not all
6.4.2. Maximum tardiness researchers have tested all the rules on a given FMS.
Thus the literature does not provide baseline data to
Pourbabai (1986) proposed a branch and bound
compare the performance of various scheduling rules
algorithm to minimize the maximum tardiness such that
even if the type of system can be controlled. In conclusion
the production of each job type equalled its demand. For
it can be said that the literature on scheduling rules in a
loading the batch of jobs in appropriate work stations,
FMS is in a state of confusion because of the lack of
pre-specified dispatching rules such as DD and EAT were
uniform assumptions, structure and other system char-
incorporated in the model. By assuming that all the jobs
acteristics. It would be extremely useful to test all the
arc available at zero time and that the set up times could
rules discussed in this paper on a given system and then
be neglected, a mixed binary linear programming form-
study how the performance of these rules vary with
ulation was suggested and a number of algorithms were
respect to various factors such as systems characteristics
provided to solve the model. However, no illustrative
and loading strategies.
example was solved and no computational experience
A relatively new analysis method, a physical simulator,
with the approach was reported.
has been proposed to assess the performance of an FMS
under various scheduling rules (Diesch and Malstrom,
Choi and Malstrom 1988, Choi et al. 1987). To quote
7. Industrial scheduling practices Diesch and Malstrom:

McKay et al. (1988) argued that the theoretical A physical simulator is an operational scale model of the
approach of operations research to scheduling often was actual system. When controlled by a computer, a physical
simulator can be used to obtain operational data from the
not appl icable to the dynamic characteristics of an actual system. These data may· then be used as an aid in the
situation. From an informal survey on scheduling prac- design, installation and operation of the actual system.
tices of 40 schedulers and meetings with several manufac-
turing consultants, the authors concluded that the A detailed physical simulator, constructed in 1982 at
theoretical formulations of the problem may be irrelevant Iowa State University, was designed to simulate the
as they did not capture the essence of the scheduling operation of an automated flexible manufacturing system
problem faced by schedulers, and the results had little currently being installed by a midwestern manufacturing
applied value. facility (described earlier). Actual data from this facility
Although the above criticisms were related to job-shop were used both in the construction of the model and in
scheduling, the conclusions so derived appear equally specifying its simulation parameters. The authors invest-
valid for FMS scheduling practices. In the present study, igated seven part type selection rules (i.e. part entry rule)
we observed that most of the research studies were and four machine centre selection rules against six
conducted using a hypothetical FMS under various different performance criteria. As mentioned earlier, the
assumptions, thereby, making it almost impossible to FMS configuration assumed that there are two operations
derive fruitful conclusions which could be used for prac- needed to be performed. The first operation must be
tical purposes. The situation was supported by a survey performed on a lathe cell and the second operation can be
conducted by Smith et al. (1986) of actual FMS sched- performed by anyone of the six machining centres. For
Scheduling rules in FMSs 373

this configuration, the SLACK/WINQ and SPT/WINQ environment (particularly due to machine failures and
rule sets dominated all the decision rule sets. The uncertainty and variability in production requirements),
large data requirements, multiple level hierarchies, and
SLACK/WINQ rule set showed superior performance other issues ..
when WIP inventory was the criteria and the
SPT/WINQ rule set showed superior performance when The authors argued that the optimization of such large
throughput time and actual production output were scale systems is impossible and suboptimal strategies for
considered. The authors conjectured that the results planning based on hierarchical decomposition may be
obtained using the physical simulator may be extended to developed and used for practical purposes (Gershwin et al.
other FMS configurations since the SPT, DDATE and 1984 a, b, 1985). They proposed a new hierarchical pro-
SLACK rules have been found to perform well for a duction scheduling policy for failure-prone FMSs. The
variety of different job shop configurations. It would be approach incorporated both machine status and demand
interesting to validate the aforementioned conjecture. deviation feedback. The scheduling decisions would be
Avonts et al. (1989) addressed an interesting real-life determined on-line based on the current status of each
production scheduling problem. Most of the literature on machine in the system and the current difference between
FMS production planning and scheduling problems production and demand. The parts would be loaded into
Downloaded by [University of Wisconsin Platteville] at 06:25 19 May 2016

assumed FMS as the only production system in a given the system at rates that are constrained by the current
facility and provided scheduling techniques for that capacity (for details, refer to Gershwin et al.
system. In practice, however, the FMS may be competing (1984a, b, 1985). The approach was validated using a
with a conventional system (CS), i.e. when a firm is in the simulation model of the real-life system at IBM's General
process of installing an FMS. Avonts et al. argued that Productions Division at Tucson, an automated card
there may be a subset of the product types which can be assembly line. The authors claimed that the policy can be
manufactured in different way (such as FMS or CS). The very effective in scheduling FMS.
problem is then to decide which product types and what
quantities ought to be produced by the FMS.
The authors used linear programming (LP) to solve the 8. Future research
problem of allocating the products to FMS and CS. The
LP models were proposed and solved while considering Tables 2, 3 and 5 show that researchers have given
various objective functions such as inventory costs, pro- more importance to machine utilization and production
duction costs, setup costs, total costs and system utiliza- rate criteria in determining optimal solutions for FMS
tion. The LP models contained only essential trade-offs scheduling. On the other hand, due date related criteria
and the fixture and pallet constraints, tool magazine seem to be more important to practitioners. Therefore, it
constraints, etc. were relaxed. The authors claimed that would be useful to direct more research efforts to the
the LP models were accepted by the company as a development of scheduling strategies using due date
planning tool because of their simplicity. However, in related criteria.
order to test the validity of the models, detailed simula- Table 6 shows that the majority of the dispatching rules
tions were conducted by including all the constraints. The used in flexible manufacturing systems are local dispatch-
LP results when compared with simulations, provided ing rules in both static and dynamic categories. Very little
good approximation. They observed that the differences work has been done in developing and testing global
in the results were due to the constraints as well as the dispatching rules in either of the two categories. It would
operating rules (e.g. the sequence in which the parts were be useful to undertake studies in this direction.
entered into the FMS, the priority rules at each machine, A recent study of a job shop (Russel et al. 1987)
the operating rules for carriers, etc.). Thus, the simul-
ation model may be used to investigate the feasibility of
the solution derived from the LP. Overall, the authors Table 5. Scheduling criteria (ranking of FMS operators)
claimed that the company was satisfied with the results (adapted from Smith tl al. (1986)).
from the simple LP models. However, for complex cases,
more sophisticated models such as multi-objective linear Rank Criteria Points

programming models could be used. 1 Meeting the due date 57


According to Gershwin et al. (1984): 2 Maximizing system/machine utilization 44
3. Minimizing in-process inventory 23
while the technology of manufacturing is improving 4 Maximizing-production rate 13
rapidly, a basic understanding of the systems issue remain 5 Minimizing setup time and tool change 13
incomplete. These issues include production planning, 6 Minimizing mean flow time 8
scheduling and control of work-in-process. They are 7 Balancing machine usage 3
complicated by randomness in the manufacturing
374 Y. P. Gupta et al.

Table 6. Relationship among dispatching rules.

Static Dynamic

Local
Shortest imminent processing time Shortest remaining processing time
Longest imminent processing time Largest remaining processing time
First in first out Fewest operations remaining
First in last our Most operations remaining
Random selection Greatest setup time
Earliest due date Modified due date
Static slack Modified operation due date
Shortest processing time/total operation time
Shortest processing time x total operation time
Largest processing time/total operation time
Largest processing time x total operation time
Earliest available time
Least amount of slack
Least ratio of job slack to remaining
Downloaded by [University of Wisconsin Platteville] at 06:25 19 May 2016

processing time
PR!TR
CRITICAL

Global
WINQ
MOD!MDD
SMCPT
NINQ

concluded that the performance of the COVERT rule at work centres influences shop performance; and (c)
(i.c, prioritize jobs in a queue according to the largest information about work centre congestion along a job's
rat io of expected job tardiness to operation processing routeing is more useful information than general shop
times) for a variety of tardiness measures excels when conditions. Some of the best due date assignment pro-
compared to S-I, S-2, MOD and other dispatching rules. cedures are summarized in Table 4. It may be worthwhile
Since COVERT is a global rule, it might be useful to test
its performance in an FMS environment. Table 7. Due dale assignment rules (for details refer to Ragatz
and Mabert (1984)).
Denzler and Boe (1987 c) argued that a number of the
job shop results can be applied in studying random I. Exogenous
FMSs. Panwalker and Iskander (1977) and Blackstone The due dates are fixed and given an attribute of a job.
et al. (1982) provided an extensive review of dispatching Two methods under this category are:
rules used in a job shop. Therefore, it may be worthwhile CON: all jobs are given exactly the same Oow allowance
to test some of these rules in FMSs and validate the claims RAN: the flow allowance for a job is randomly assigned
made by Denzler and Boe. 2. Endogenous
Conway (1965) found that the performance of all rules The due dates are set internally as the job arrives on the
related to mean lateness and number of tardy jobs were basis of:
somewhat sensitive to the method of due date establish- Job characteristics (such as arrival times, and processing
ment. An analysis of due date management problems in a times)
FM S scheduling environment reveals that while a variety TWK: due dates are based on total work content
SLK: jobs are given flow allowance that reflects equal
of sophisticated decision rules have been suggested to
slack or waiting time
assign due date (sec Table 7), not much attention has Nap: due dates are determined on the basis of
been paid to the selection of such rules. In the literature, number of operations to be performed on each job
the most widely used due date assignment rule is TWK Shop status information (such as queue length)
(c.g. Shanker and Tzen 1985, Choi et at. 1987). JIQ: due dates are determined based on current
An analysis of the recent simulation studies (e.g. queue lengths in system
Ragatz and Mabert 1984) of due date assignment rules JIS: due dates are determined based on information
reveals that: (a) job characteristic and shop status on number of jobs in system
PPW: due dates are determined based on information
information should be used to develop due date assign- on waiting time in system
ment rules; (b) the dispatching rule used to sequence jobs
Scheduling rules in FMSs 375

investigating the impact of the above fmdings on the BELL, R. and BILALIS, N. 1982, Loading and control strategies
performance of a flexible manufacturing system. for an FMS for rotational parts. Proceedings oj the First
International on FMS, pp. 77-87.
Finally, we observed that most of the simulation studies
BLACKSTONE,J. H., PHtLLIPS, D. T. and HOGG, G. L. 1982,
performed by researchers were on hypothetical FMSs. A state-of the-art survey of dispatching rules for manufactur-
To make it worse, no two papers considered the same ing job shop operations. International Journal of Production
configurations. Thus, it would be highly desirable to Research, 20, 27-45.
perform simulation studies in a way that further investi- BROWNE, J., DUBOIS, D., RATHMtLL, K. SETHI, S. P. and
STECKE, K. E. 1984, Classification of flexible manufacturing
gations could be performed successfully in the future. For
systems. FMS Magazine, 2,114-117.
this purpose, we suggest that researchers should consider Buzxcorr, J. A. 1985, Modelling manufacturing systems
various factors as suggested by Schriber and Stecke and Robotics and Computer Integrated ManuJacturing, 2, 25-32.
are tabulated in Table 4. The use of this table as a BUZACOTT,J. A. and SHANTHIKUMAR,j. G. 1980, Models for
checklist would be helpful to derive the conclusions and understanding flexible manufacturing systems, AlIE Transue-
tions, 12, 339-350.
use the results in real life situations.
BUZACOTT, J. A. and YAO, D. 1986, Flexible manufacturing
systems: a review of analytical models. Management Science, .32,
890-905.
9. Summary CARRIE, A.S. and PETSOPOULOS, A. C. 1985, Operation
Downloaded by [University of Wisconsin Platteville] at 06:25 19 May 2016

sequencing in a flexible manufacturing system. Rohotica, 3.


259-264.
In this paper, we have classified the dispatching rules
CHAN, T. S. and PAK, H. A. 1986, Heuristical job allocation in
employed in FMS scheduling and attempted to explain a flexible manufacturing system. Journal of ManuJacturing
the general idea behind different rules and develop a Systems, 1, 69-90.
framework to guide future research efforts in this direc- CHANG, Y. and SULLIVAN, R. 1984, Schedule generation in a
tion (Tables 2,3,6). We tried to analyse the results of dynamic job shop, work paper, University of Texas at
Austin.
various researchers and found that the task was quite a
CHANG, Y. L. and SULLIVAN, R. S. 1984, Real-time schedul-
formidable one, for several reasons: (a) a variety of ing of flexible manufacturing systems. Presented at the
hypothetical as well as real system configurations had to TIMS!ORSA San Francisco Meeting, May.
be considered; and (b) the combination of various loading CHANG, Y., SULLIVAN, R. and BAGCHt, U. 1986, Experimen-
strategies and dispatching rules applied in the studies tal investigation of quasi real lime scheduling in FMS.
Proceedings oj the 2nd ORSA!TIMS ConJerence on FMS,
gave system specific results. Thus the literature does not
pp.307-312.
provide baseline data to compare the performance of CHOt, R. H., KNIGHT, N. and MALSTROM, E. 1987, A
various scheduling rules. It also appears that the choice of comparison of digital vs. physical simulation for evaluation of
scheduling rules is dependent upon the operating char- FMS scheduling rules. Proceedings of the Spring Annual Con-
acteristics of the system such as variations in processing ference. Institute of Industrial Engineers, Washington, D.C.,
May.
times. It is concluded that the literature on scheduling
CHOI, R. H. and MALSTROM, E. M. 1988, Evaluation of work
rules in FMS is in a state of confusion because of the lack scheduling rules in a flexible manufacturing system using
of uniform assumptions, structure and other system a physical simulator. Journal oj ManuJacturing Systems, 7,
characteristics. It would be extremely useful to test all the 33-45.
rules discussed in this paper on a given system and then CLOSSEN, R. J. and MALSTROM, E. M. 1982, Elfeclive
capacity planning for automated factories requires workable
study how the performance of these rules vary with
simulation tools and responsive shop 0001' controls. Industrial
respect to various factors such as systems characteristics Engineering, April, 15, 73-79.
and loading strategies. CO, H. C., JAW, T. J. and CHEN, S. K. 1981, Sequencing in
flexible manufacturing systems and other short queue-length
systems. Journal of ManuJacturing Systems, 7, 1-8.
CONWAY, R. W. 1965, Priority dispatching and job lateness in
Acknowledgement
a job shop. Journal of Industrial Engineering, 16, 123.
DAR-EL, E. and SARIN, S. 1984, Scheduling parts in FMS to
We acknowledge the constructive comments made by achieve maximum machine utilization. Proceedings of the First
the referees on the earlier version of this manuscript. ORSA!TIMS Conference on FMS, pp. 300-306.
DENZLER, D. and BOE, W. 1987 a, Experimental investigation
of FMS Scheduling decision rules. International Journal oj
Production Research, 25, 979-994.
References DE:-JZLER, D. R., BOE, W. J. and DUPLAGA, E. 1987b, An
experimental investigation of FMS scheduling rules under
uncertainty. Journal of Operations Management, 7, 139-151.
AVONTS, L. H., GELDERS.L. F. and WASSENHOVE, L. N. V. DE:-JZLER, D. and BOE, W. 1987 c. An evaluation of part
1988, Allocating work between an FMS and a conventional scheduling rules in a random FMS. Proceedings of the Decision
job shop: a case study. European Journal of Operations Research, Science Institute, November, pp. 787-789.
.3.3, 245-256. DIESCH, K. H. and MALSTRO~1, E. M. 1984, Physical modcll-
376 Y. P. Gupta et al.

ing of flexible manufacturing system. Proceedings of the Fall KUSIAK, A. 1985 b, Material handling in flexible manufactur-
Annual Conference. Institute of Industrial Engineers, Atlanta, ing systems. International journal oj Material Flow, 2,
October. 79-95.
DUI'ONT-GATEL~1AND,C. 1982, A survey of flexible manufac- KUSIAK, A. 1986, Loading models in FMS. 'Flexible Manufactur-
turing systems. journal of Manujacturing Systems, 1, 1-16. ing, edited by A. Rouf and S. Ahmed (Elsevier Science
ELMARAGHI', H. A. 1981, Simulation and graphical animation Publishers, Amsterdam).
of advanced manufacturing systems. Journal of Manufacturing LIN, L. S. and Lu, S. 1984, The scheduling problem in random
SyJtWIJ, I, 53-63. FMS. Proceedings oj the First ORSA/T1MS Conjerence,
GERE, W. S. 1966, Heuristics in job shop. Management Science, 1'1'.278-283.
13,167-190. MCKAY, K. N., SAFAYENI, F. R. and BUZACOTT,J. A. 1988,
GERSHWIN, G. 1:1., HILDERBRANT, R. R., SURI, R. and Job-shop scheduling theory: what is relevant? Interfaces, 18,
MnTER, S. K., 1984, A control theorist's perspective on 84-90.
recent trends in manufacturing systems. Proceedings of 23rd MENON, U. 1985, A multi-objectioe production planning framnoork
Conference on Decision and Control, Las Vegas, December, for automated manufacturing system, Ph.D. dissertation, Uni-
1'1'.209-225. versity of Nottingham.
GERSHWIN, G. 1:1., AKELLA, R. and CHOONG, Y. 1984, A MORTON, T. E. and SMUNT, T. L. 1984, Schedule generation
hierarchical scheduling policy applied to printed circuit board in a dynamic job shop. Proceedings oj the First ORSA/T1MS
assembly. Robotics & Computer l ntegraud Manufacturing, I, Conference.
299-305. NICHOLSON, T. A. and PULLEN, R. D. 1971, A practical
Downloaded by [University of Wisconsin Platteville] at 06:25 19 May 2016

GERSHWIN, G. 1:1., AKELLA, R. and CHOONG, Y. 1985, control for optimizing production schedules. Internationaljour-
Short-term production scheduling of an automated rnanufac- nal oj Production Research, 9, 219-227.
turing facility. IBM journal oj Research Development, 29, NOF, S., BARASH, M. M. and HERALD, M.J. 1978, Analysis of
392-400. operating rules in a computerized manufacturing system,
GROOVER, M. P. 1985, Automation, Production Systems, and ASME Publication # 78-WA/Prod-38.
Computer-A,'ded Manujacturing, Prentice-Hall, Englewood NOF, S., BARASH, M. M. and SOLBERG, J. 1979, Operational
Clil],s, NJ. control of item flows in versatile manufacturing systems.
GUPTA, Y. P. and GO\'AL, S. K. 1989, Flexibility ofmanufac- International journal oj Production Research, 17, 479-489.
turing systems: concepts and measurements. European journal PANWALKER, S. and ISKANDER, W. 1977, A survey of schedul-
of Operational Research, 43, 1-17. ing rules. Operations Research, 25, 45-61.
I-IAN, M., NA, Y. K. and HOGG, G. L. 1989, Real-time tool POURBABAI, B. 1986, A production planning and scheduling
control and job dispatching in flexible manufacturing sys- model for FMS. Proceedingsof the Second ORSA/TIMS Conjerence
tems. International Journal of Production Research, 27, on FMS, pp. 533-543.
1257-1267. RACHAMADUGU, R. and STECKE, K. E. 1988, Classification
HATVAN\', J. (editor), World Survey on CAM (Butterworths, and review of FMS scheduling procedures, Working paper
Borough Green). #481 c, University of Michigan, Ann Arbor, MI.
HARRINGTON, J., Understanding the Manujacturing Process (Marcel RAGATZ, G. and MABERT, V. 1984, A simulation analysis of
Dekker, New York). due date assignment rules. journal oj Operations Management, 5,
H"RAGU, S. and KUSIAK, A. 1988, Machine layout problems in 27 -39.
flexible manufacturing systems. Operations Research, 36, RAMAN, N., TALBOT, F. B. and RACHAMADUGU, R. V. 1986,
258-268. Simulation scheduling of machines and material handling
HUTCHINSON, G. K. 1979, Flexible manufacturing systems in devices in automated manufacturing. Proceedings oj the Second
the United States. Management Research Centre, University ORSA/TIMS Conjerence on FMS.
of Wisconsin-Milwaukee, Milwaukee, WI. RANK\', P. 1983. The Design and Operation oj Flexible Manujactur-
IWATA, K. MUROTSU, A. and OBA, F. 1982, Production ing Systems (North-Holland, Amsterdam).
scheduling of flexible manufacturing system. Annals oj CIRP, RUSSEL, R. S., DAR-EL, E. M. and TA\'LOR, B. 1987, A
31,319-322. comparative analysis of the COVERT job sequencing rule
JAIKUMAR, J. 1986, Post industrial manufacturing. Harvard using various shop performance measures. International Jour·
Business Review, 6, 69-76. nal oj Production Research, 25, 1523-1540.
KALKUNTE, M. V., SARtN, S. C. and WILHELM, U. E. 1986, SARIN, S. C. and DAR-EL, E. M. 1984, Approaches to the
FMS: A review of modelling approaches for design, justifica- scheduling problem in FMS. Proceedings oj the Institute oj
tion, and operations. FMS: Methods and Studies, edited by A. Industrial Engineers Fall Conjerence, pp. 76-86.
Kusiak (North-Holland, Amsterdam). SARIN, S. and WILHELM, E. 1983, FMS: a review of modeling
KA:-IlcT, J. J. 1987, Expert systems in production scheduling. approaches for design, justification and operation, Working
European journal oj Operational Research, 26, 51-59. paper, Ohio State University.
KIMEMIA, J. and GERSHWIN, S. 1983, An algorithm for the SCHRIBER, T. J. and STECKE, K. E. 1986, Machine utilizations
computer control of a FMS. llE Transactions, 15, 353-362. and production rates achieved by using balanced aggregate
KIMEMIA, J. and GlmsHwlN, S. 1985, Flow optimization in FMS production ratios in a simulated selling. Proceedingsojthe
flexible manufacturing systems. International Journal of Produc- Second ORSA/TiMS Conference on FMS.
tion Research, 23, 81-96. SCHRIBER, T. J. and STECKE, K. E. 1987, Using mathematical
KtRAN, A. and ALPTEKIN, S. 1985, Scheduling algorithms for programming and simulation to study FMS machine utiliza-
flexible manufacturing systems, Working paper 85-18, Uni- tions. Proceedings oj the 1987 Winter Simulation Conference.
versity of Southern California. SHANKER, K. and TZEN, J. 1985, A loading and dispatching
KUSIAK, A. 1985 a, Flexible manufacturing systems: a struc- problem in a random FMS. International journal oj Production
t ural approach. International Journal of Production Research, 23, Research, 23, 579-595.
1057-1073. SLACK, N. 1989, The flexibility of manufacturing systems
Scheduling rules in FMSs 377

International Journal of Operations and Production Management; 7, STECKE, K. E. and SURI, R. (editor) 1986, Flexible manufacturing
35-45. systems. Proceedings of the l1MS/ORSA Conference on FMS (Else-
SLOMP, J., GAALMAN, G. J. L. and NAWIjN, W. M. 1988, vier Science Publishers B. V., Amsterdam).
Quasi on-line schedulers for flexible manufacturing system. STEFFEN, M. S. 1986, A survey of artificial intelligence based
International fournal of Production Research, 26, 585-598. scheduling systems. Fall Industrial Engineering Conference Pro-
SMITH, M. L., RAMESH, R., DUDEK, R. and BLAIR, E. 1986, ceedings, 395-405.
Characteristics of US flexible manufacturing systems-survey. SURI, R. 1985, An overview of evaluative models for flexible
Proceedings of the Second ORSA/TIMS Conference on FMS, manufacturing systems. Annals of Operations Research, 3,
pp.477-486. 13-21.
STECKE, K. E. 1981, Production Planning Problems for FMS, VAITHIANATHAN, R. 1982, Scheduling in flexible manufactur-
unpublished Ph.D. Dissertation, Purdue University. ing systems. Fall Institute of Industrial Engineers Proceedings,
STECKE, K. E. 1983, Formulation and solution of non-linear pp.421-425.
integer production planning problems of FMS. Management VAITHIANATHAN, R. and McROBERTS, K. L. 1981, On
Science, 29, 273-288. scheduling in a GT environment, SME Technical Paper
STECKE, K. E. 1985, Design, planning, scheduling and control MS81-492S.
of flexible manufacturing systems. Annals of Optrations VAN LOOVEREN, A.]., GELDERS, L. F. and VAN WASSEN·
Research, 3, 3-12. HOVE, L. N. 1986, A review of flexible manufacturing
STECKE, K. E. and BROWNE, J. 1985, Variations in flexible systems planning models. Modelling and Design of Flexible
manufacturing systems according to the relevant types of Manufacturing Systems, edited by A. Kusiak (North-Holland,
Downloaded by [University of Wisconsin Platteville] at 06:25 19 May 2016

automated materials handling. Material Flow, 2, 179-185. Amsterdam), pp. 3-32.


STECKE, K. E. and SOLBERG, J. 1981 a, Loading and control WALKER, T. C. and MILLER, R. K. 1986, Expert systems 1986:
policies for FMS. International Journal of Production Research, 9, An Assessment of Technology and Applications (Madison, GA:
481-490. SEAl Technical Publications), pp. 121-128.
STECKE, K. E. and SOLBERG, J. 1981 b, The CMS loading WANG, H., 1986, An experimental analysis of the FMS. FMS:
problems. The Optimal Planning of Computerized Manufacturing Methods and Studies, edited by A. Kusiak (North-Holland,
Systems, Report 20, Feb. 1981 (School oflndustrial Engineer- Amsterdam).
ing, Purdue University).

View publication stats

Vous aimerez peut-être aussi