Production control and information systems for
component-manufacturing shops
Citation for published version (APA):
Bertrand, J. W. M., & Wortmann, J. C. (1981). Production control and information systems for component-manufacturing shops. Elsevier Scientific Publishing Company. https://doi.org/10.6100/IR123728
DOI:
10.6100/IR123728
Document status and date: Published: 01/01/1981 Document Version:
Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers) Please check the document version of this publication:
• A submitted manuscript is the version of the article upon submission and before peer-review. There can be important differences between the submitted version and the official published version of record. People interested in the research are advised to contact the author for the final version of the publication, or visit the DOI to the publisher's website.
• The final author version and the galley proof are versions of the publication after peer review.
• The final published version features the final layout of the paper including the volume, issue and page numbers.
Link to publication
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain
• You may freely distribute the URL identifying the publication in the public portal.
If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please follow below link for the End User Agreement:
www.tue.nl/taverne Take down policy
If you believe that this document breaches copyright please contact us at: openaccess@tue.nl
providing details and we will investigate your claim.
-COMPONENT-MANUFACTURING SHOPS
Proefschrift
teJt veJtlvUjging vrut de. g!taad van doctoJt in de. te.chni4che weten6chappen aan de Te.chnibche. Hoge6chool Eindhoven, op gezag van de
Jte.ctOJt magni6icw.,, P1to6.ilt. J. 5'tkeleM voOJt
ee.m com~~ie. aangewezen dooJt het colle.ge van dekanen in he;t openba.aJt te. veJtdedige.n op
v!Ujdag 10 ap!til te. 14. 00 uuJt
dootL
JAN WILLEM MARIE BERTRAND
door de promotoren
Prof.ir. W.M.J. Geraerds Prof.ir. W. Monhemius
COMPONENT-MANUFACTURING SHOPS
Proefschrift
t~ venkkijging van de gkaad van docto~ in de teeh~che weten6ehappen aan de Tec~ehe
Hoge6ehoot Eindhoven, op gezag van de
!LectM magn-tM-c.!Lb, PJW6.~. J. E~kden6 voo!L
een comm~~Ie aangewezen doo~ het c.ottege va.n deka.nen In het openba.M te v~dedigen op
v!Lijda.g 10 a.p~ te 16.00 UU!L
doo~
JOHAN CASPER WORTMANN
door de promotoren
Prof.ir. W.M.J. Geraerds Prof.dr.ir. E.A. Koldenhof
PRODUCTION CONTROL AND
INFORMATION SYSTEMS
FOR
COMPONENT-MANUFACTURING SHOPS
J.W.M. BERTRAND and J.C. WORTMANN
Department of Industrial Engineering,
Eindhoven University of Technology, Eindhoven, The Netherlands
ELSEVIER SCIENTIFIC PUBLISHING COMPANY
Molenwerf 1, 1014 AG Amsterdam P.O. Box 211, Amsterdam, The Netherlands
Distributors for the United States and Canada:
ELSEVIER/NORTH-HOLLAND Inc. 52, Vanderbilt Avenue
New York, N.Y. 10017
ISBN 044441964-0 (Vol. 1)
ISBN 044441963~2 (Series)
©Elsevier Scientific Publishing Company, 1981
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or l:>y any means, electronic, mechanical, photocopying, recording or other-wise, without the prior written permission of the publisliler, Elsevier Scientific Publishing Company, P.O. Box 330, Amsterdam, The Netherlands.
PREFACE
This book is concerned with four issues in the field of production control and information system design. Firstly, the text presents a specific approach - the
i~ approach - to the design of production control and information systems as a project. It reports on the application of the approach to a specific situation in practice. Secondly, the text discusses the decompo~ition of complex production control problems into relatively simple subproblems that can easily be coordinated. For a specific class of production situations - component manufacturing shops with functional layout - such a decomposition is presented. For a theoretical production
control situation belonging to this class, the decomposition is investigated by means of systematic computer simulation; and for a specific situation in practice, it is used as a basis for production control system design.
Thirdly, the text contributes to the problem of controlling the adequac~ of an existing production control system with time. Basic concepts from general control theory are used to derive the types of variables to be monitored over time. These types of variables are related to each other and to the control performance in a coherent monitoring scheme. The use of this scher.le is demonstrated ,in the design of a production control system in practice. Fourthly, the text investigates problems related to the changeabilit~ and ~gidit~ of information systems.1Changeability may
be realized as flexibility or as adaptability. Changeability and rigidity may refer to the form, to content and to the use of information. The text develops prescrip-tions for the employment of these types of changeability, both for the design of a production and information system and for the operational use of the system. The text furthermore reports on the use of these prescriptions in an actual design project in practice.
The book is intended for professional workers in the field of production control and information system design as well as for scientific workers in these disciplines.
The book consists of four parts. Part I presents basic theoretical considerations with respect to the four issues mentioned above. Part II gives the results of a theoretical investigation of the problem decomposition approach to control systems design. Part III reports on the use of all four issues in designing an actual pro-duction control and information system in practice. Part IV evaluates currently available software packages for production control, with respect to the four issues treated in this text, and finally presents concluding remarks. All four Parts are confined to component manufacturing shops with functional layout.
The material in this text is the result of two strongly related research projects at the Department of Industrial Engineering of the Eindhoven University of Technolog~
The Ph.D. project on production control sys tern design, carried out by J. \U1. Bertrand is reported on in the Chapters 2, 4, 6, 8, 11, 14, 15, 18 and in Section 20.3. The Ph.D. project on the design of information systems for production control, carried out by J.C. Wortmann; is reported on in the Chapters 3, 5, 9, 10, 12, 13, 16, 17, 19 and in Section 20.4. The Chapters 1 and 7 and Sections 20.1 and 20.2 are the shared responsibility of the authors.
Aclm.owledg eme;t:t6
It has become customary for authors to express their gratitude to their families for substantial support, which remains unobserved by the outside world. In our case such gratitude is more than justified - so we are happy to do so.
The projects were carried out under the supervision of Prof. W.M.J. Geraerds. We owe much to his inspiring views on production control and many other aspects of production organizations. In particular, we are indebted to him for the confidence he showed in the outcome of the projects during the earlier years, for his criticism during the later years, and for his general support of the project. Prof. E.A. Kolden-hof provided many valuable suggestions on the design of production control information systems. We experienced his confidence in our approach as a powerful stimulus.
Furthermore, he advised us to publish the results of both research projects in one text. Prof. W. Monhemius have many detailed and thoughtful comments on earlier drafts of this text and thus contributed much to its improvement. He always managed to balance encouragement and criticism.
An important part of the research project, the design and implementation of an actual system in practice, was carried out in the Diffusion Department IC-bipolair (Department 36) of the semiconductor manufacturing plant of Philips' Industries in Nijmegen. We thank the staff of the plant, and especially Department 36, for their hospitality and cooperation, and we were very happy to find that the final results of the design project justified the resources they spent on the project. While it would be impossible to mention all those who have supported the project, we would
like to sing)e out some people for special mention. Mr. A. Frederiks, former Ma-terials Manager, and Dr. F.J. Stommels provided organizational support which gave the project an unique opportunity to start. Mr. T. t1engelberg, chief of Department 36, was the main driving force behind the design project in practice. Finally, Mr. I. Derksen implemented the software part of the information system on a minicomputer system. His criticism was of much help in the final design phase.
Furthermore, we thank Philibert Beekman, who developed the job-shop s imul ati on software package, Eduard van der Weegen, who contributed in the earlier stage of the simulation experiments, and Jan Henselmans', who did much work in developing the models for multi -task capacity types. We are grateful to severa 1 other students who helped in gathering and processing data for statistical analysis, and in the
execution of simulation experiments: Andre Bergmans, Paul Durlinger, Ed Kessels, Huub Koninkx and Kees Timmer. t~any others were helpful in preparing this text. The majority of the drawings were produced by Eric Rensen at an amazing speed, and the remainder by Piet van der Kamp and Thomas Klarecki. We express our appreciation furthermore to Mrs. Lenie Schrage, who typed the first version of this text from a scarcely legible manuscript, and last but not least to Miss Lieske van den Boezem, who typed the final version, and who often succeeded in typing the text we had in mind rather than the text we had written down.
February 1981 Eindhoven
J .fJ.M. Bertrand
CONTENTS
PREFACE CONTENTS
PART I · PRINCIPLES
1. I NT RODUCTI ON 1.1. Pre 1 imi nary 1. 2. The field 1.3. Organizations 1.4. The design approach 1.5. Methodological problems 1.6. An outline of the book
vii xi 3 3 3 5 6 12 14
2. PRODUCTION CONTROL: BASIC CONCEPTS 16
2.1. Introduction 16
2.2. The production control function 16
2.3. A framework of concepts 17
2.4. Models 21
2.5. Structuring the control problem 29
2.6. Additional considerations for the design of a control system 31 2.7. Further limitation of the problem to be considered 32
3. PRODUCTION CONTROL INFORMATION SYSTEr1S: A FRAt~EWORK OF CONCEPTS 34
3.1. Introduction 34
3.2. Outline of the production-control information system 35
3.3. Contributions to information systems theory 49
3.4. The syntactical, semantical and practical aspect of information 52
3.5. Usability of information systems 62
3.6. Changeability of information systems 68
3.7. Dynamic aspects of the PCI model 76
4. THE STRUCTURING OF THE PRODUCTION CONTROL PROBLEM 4.1. Introduction
4.2. The process to be controlled 4.3. Structuring the control problem
78 78 78
5. DEVELOPr,1ENT OF PRODUCTION-CONTROL IN FORMAT! ON SYSTEMS 5.1. Introduction
5.2. The conceptual design phase
5.3. The design of the experimental system 5.4. The experimentation phase
5.5. The consolidation phase
PART II CONTRIBUTION TO PRODUCTION CONTROL THEORY
6. PRODUCTION CONTROL FOR THE THEORETICAL JOB SHOP
94 94 96 101 106 109 6.1. Introduction 115
6.2. The sequencing function 118
6.3. The scheduling and due-date assignment procedure 126
6.4. Experimental strategy 130
6.5. Performance of the sequencing function 132
6.6. Performance of the scheduling and due-date assignment function 141
6.7. Interactions 149
6.8. Selection of parameter values 155
6.9. Comparisons with the performance of the job-shop control system
i nves ti gated by Conway et a 1 . 157
6.10. Conclusions 159
PART III A DESIGN PROJECT IN PRACTICE
7. THE PRODUCTION SYSTEM 7.1. Introduction
7.2. The manufacturing process
7.4. Organizational environment of the production department 7.5. The organization
7.6. Computer systems supporting internal decision-making
8. STRUCTURING OF THE PRODUCTION CONTROL PROBLEt~
8.1. lntroducti on
8.2. The production control problem
8.3. Analysis of the control relations of the'separate decision functions
8.4. Analysis of the benefits
165 165 165 174 178 181 183 183 183 190 197
9. THE CONCEPTUAL PRODUCTION CONTROL INFORMATION MODEL 9.1. Introduction
9.2. General considerations 9.3. The PCI model
9.4. Changeability and rigidity of the conceptual PCI model
10. CAPACITY MODELS FOR nUAL CONSTRAINT PRODUCTION SYSTEMS 10.1. Introduction
10.2. Finite loading
10.3. Capacity constraints for multi-task capacities 10.4. An alternative model for multi-task capacities
11. DESIGN OF THE EXPERIMENTAL PRODUCTION CONTROL SYSTEM 11.1. Introduction
11.2. The production level evaluation function 11.3. The batch release function
11.4. The batch scheduling function 11.5. The operator allocation function 11.6. The batch selection function 11.7. The decision coordination function
12. DESIGN OF THE EXPERIMENTAL PRODUCTION CONTROL INFO~N\TION
202 202 202 205 218 223 223 224 228 231 234 234 234 237 257 259 263 265 SYSTEM 269 12.1. Introduction 269
12.2. Measuring production progress 269
12.3. The PC! model: transaction processing and quality monitoring 272 12.4. The OS system for production level evaluation 276
12.5. Generation of batch schedules 281
12.6. The OS system for batch release 283
12.7. Information for batch selection 289
12.8. The DS system for operator allocation 292
12.9. The OS system for decision coordination 300
13. EXPERIENCES FROM THE EXPERIMENTAL PRODUCTION CONTROL INFORMATION
SYSTEM: USABILITY AND CORRECTNESS 301
13.1. Introduction
13.2. The PCI model in the first experimentation phase 13.3. OS systems in the first experimentation phase
301 302 304
14. EXPERIENCES FROf1 THE EXPERH1ENTAL PRODUCTION CONTROL SYSTEt1 310
14.1. Introduction 310
14.2. The batch scheduling and the batch selection function 310
14.3. The operator allocation function 310
14.4. The batch release function 312
14.5. The production level evaluation function 316
14.6. The decision coordination function 317
15. REDESIGN OF THE PRODUCTION CONTROL SYSTHI 15.1. Introduction
15.2. Main experiences
15.3. The operator capacity decision function 15.4. The production level evaluation function 15.5. The batch release function
15.6. The operator allocation function 15.7. The decision coordination function
16. REDESIGN OF THE PRODUCTION CONTROL INFORMATION SYSTEM:
321 321 321 322 325 325 331 333 EMPLOYING CHANGEABILITY 335 16.1. Introduction 335
16.2. The OS system for the operator capacitydecisions and the
production level evaluations 336
16.3. The DS system for batch release 337
16.4. The DS system for operator allocation 342
16.5. Redesign of the PCI model 344
17. EXPERIENCES WITH THE REDESIGNED PRODUCTION CONTROL INFORI1ATION SYSTEM: CONSOLIDATION
17.1. Introduction
17.2. OS information for capacity control, production level and workload control
17.3. OS information for operator allocation 17.4. Responsibility accounting
18. EXPERIENCES FROII THE REDESIGNED PRODUCTION CONTROL SYSTEM:
348 348 350 354 354 CONTROL PERFOR~1ANCE 356 18.1. Introduction 356
18.2. The batch scheduling function and the batch selection function 356
18.3. The operator allocation function 356
18.5. The production level evaluation function 18.6. The decision coordination function
PART IV EVALUATION AND CONCLUSIONS
19. EVALUATING STANDARD SOFTWARE PACKAGES 19.1. Introduction
19.2. Analysis of production control
19.3. Standard software packages from an information systems point of view
20. CONCLUSIONS 20.1. Introduction
20.2. The learning approach 20.3. Production control design 20.4. Information systems design
LIST OF IMPORTANT SYMBOLS REFERENCES SUBJECT INDEX 366 366 371 371 374 379 383 383 383 385 387 390 395 400
PART I
PRINCIPLES
TieL~ PCU!:t pltel.>e;~& p>U'.nc.ipte.o ofi de6~gn-ing ptwduc.:tLan c.onvwt and -in6oJtma:tian 6tJ.O:temo. A de/.\~fln app!Wac.h c.ai..ted the "ieaJtn-ing appJtoac.h" L& advoc.a:ted in Chap:te!t 1. The Cfu:tp.te.M 2 and 3 deM2Jt.tbe ·bMic.
c.on-c.ep:t.o fiaJt p!toduwon c.an.t;wt and in60Jt-ma:tion .6tj.O:tem.o. Chap:teJt 4 deman.otlta:te/.1· haw .the ave.!tai..t p!toduc.:tLan c.onvwt p!tob.tem c.an be dec.ompo.oed -in:to manageable.
t>ttb-i :tlviA dec.ampMW.o n -it> c.ai..te.d ".o:t!tuc.:twcLng" o6 :the ove.Mil. pJtobiem. F-inai..ty, 5 fuc.uMe6 .-in detail how
p!toduc.Uon eon:t-tol and MJ~:te.ml.>
duig n ptw jew p!tO c.e.e.d pJuX.c.uc.e., when :the. "ieaJtn-ing app!tOac.h" -it> emp.f.o ye.d.
1. INTRODUCTION
1.1. ~reliminary considerations
Control of production at reasonable cost in the face of uncertainties is still one of the main problems in manufacturing organizations. The increase in data-processing capabilities of modern computers has not eliminated the production con-trol problem but has largely improved the problem-solving tools. Technical develop-ments in data processing do not relieve a decision-maker of his responsibilities to decide, to understand the process he is controlling, and to account for his decisions. Advanced tools may improve decision-making performance, but they do not necessarily simplify the decision-making process. Furthermore, designing and organizing information flows and data-processing procedures (computerized or manual) has itself become a problem area in many organizations. This is not merely a problem of data storage, data manipulation, or data distribution by electronic data pro-cessing (EDP) but also a problem of information collection, information interpre-tation and information use by human beings. These human activities should be organ-ized in a framework of organizational procedures, and they should be part of the tasks of employees. The growing complexity of this flow of information has resulted in many cases in the creation of a new organi zati ona l function, -Ln{,oJtmcction !:,tJ~.:,.tenM management. Our view of the relation between production control and information systems is depicted in Fig. 1.1.
The design of production control systems and the design of information systems can be distinguished but cannot be separated. For example, consider a single de-cision function within a' production control system. The designer of
eantlwt ~.:,y1:dem~:, may provide the decision maker with, say, either a rule-of-thumb or a complex model of the process to be controlled. This choice will certainly be influenced by the information system which has to be designed in either case. On the other hand, in order to determine the information requirements of such a system the designer of ,Ln6oJtma.tion MJ4:tenM may choose between the use of standard software packages for production control and the use of special purpose software. This choice will certainly be influenced by the extent to which production control requirements can be met in either case.
1.2. The field
In this text, the field of p~duction eon:tnot is concerned with all decisions on the vo.tumv., and :tljpe-6 of products to be produced, and on the time. Mpec;t of
production control system
r( __________
m_a_n.u_f_a_c_t_ur_i_n~g~p_ro_c_e_s_s ______________ ~production. Attenti-on is focussed on the control of job flow times, capacity util-izations and job due date performance. The main decision areas considered are:
job loading and scheduling, including due-date assignment job releasing
capacity allocation and job sequencing.
The field of production control will be described in more detail in Chapter 2, where production control is distinguished from other areas; concepts for the analysis of production control systems will be given, and a method for the design of produc-tion control systems will be presented. Chapter 4 presents an outline of a general production control structure for the class of control problems considered.
Furthermore, in this text, an in6onmatlon ~y~tem is considered to be a set of organizational procedures, together with tools such as hardware and software, direct-ed towards the presentation of information for decision making. In the design of information systems, two aspects can be distinguished. First, the question of w!Uch
information is to be presented (or stored); this matter is sometimes referred to as: the in6ologica£ duign (Lundeberg et al., 1979a). Second the question of how
to process and to store information; this matter is sometimes referred to as: the datafogical duign. In this text, the emphasis is on infological design. In Chapter 3, the field of information systems will be described more fully and a framework of concepts will be presented. Chapter 5, finally, gives an approach to information system design for production control.
1.3. Organizations
Both production control systems and information systems are designed in order to perform a certain function within an organization. This means that decision-making for production control should form part of the tasks of employees in the organization. Similarly, the collection, storage, transformation and presentation of information should be part of the tasks of employees. Therefore, a (re)design of a production control system or of an information system may involve a restruc-turing of organizational functions, tasks and responsibilities; in consequence, such a (re)design generally implies organizational
This text does not aim at increasing our knowledge of the psychological, socio-logical or organizational aspects of the design and implementation process. The jiscvssion focusses instead on production control theory and information systems theory; if required, insights gained from behavioural sciences are used. Thus, our jesign and implementation approach is partly based on a number of empirical obser-vations reported in the literature on organizational change, as will be discussed in Section 1.4. Part III of this book reports on experience gained from an appli-:ation of our approach in practice. The successive stages of the design and imple-nentation process are described systematically in order to reveal the interactions
between production control design and information systems design; furthermore, the development of the interaction between these formal design activities and the induced organizational changes at various stages is described. In this description. aspects of interest for the behavioural sciences are presented as fas as we considered these relevant for understanding the actual design and i~plementation process.
1.4.
Most textbooks on production control theory pay attention to the process of designing and implementing production control systems; the same holds for textbooks on information systems. Furthermore the approaches advocated in both fields resemble each other remarkably.
The approach advocated up to 1970. roughly speaking, can be illustrated by Fig. 1.2. The approach is essentially ;...tJt<Ught6M.waJtd, in the sense that no feedback
feasibility
-+
gross-+
detailed l...+i111plemen-study design design tat ion
is assumed to exist from later design stages to induce some redesign. Blumenthal (1969) contributed essentially to this idea by requiring that each project should be part of a company-wide framework; this framework was called the maD~~ by many later writers on information systems.
The importance of user participation and top management support for the design process is stressed in a vast amount of literature from the behavioural sciences (cf. Greiner, 1967); for further references, see Edstrom (1977).
After 1970, a growing awareness of design as an iterative process emerged. In the first place, it was stressed that findings 'of the detailed design could induce changes in the gross design; then that experiences in the implementation stage could require modifications of the detailed design, and so on. This resulted in the idea that the design process could be cancelled and restarted at any point; the approach is depicted in Fig. 1.3, derived from Davis (1975).
The growing consensus on the iterative nature of the design process stems from the fact that most actual design projects actually are iterative. However, merely including feedback loops in the design approach is unsatisfactory because many questions remain, such as: To what extend can the design process be planned? What are the causes of the failures in a previous design stage? etc. As we shall see,
definition
..
physical...
implemen-stage
-
design-
tat ionstage stage
J•
t
I
t
'
I
Hg. 1. 3. The. -Ue.tw .. tive. app!toac.h
1n answer to this question may be given by the behavioural sciences.
In design situations, the relevant knowledge is distributed over two (generally lisjoint) groups of people. On the one hand, there are the people whom we shall call :he exp~. Members of this group have knowledge about production control and information systems theory and are experienced in design. On the other hand, there 1re the people whom we shall call the cUent I.>IJ!.dem. t1embers of this group know i lot about the actual problem situation; ideally some of them may have knowledge 1bout production control or information systems as well. The client system generally ~ill not be a homogeneous group; it may consist of a number of subgroups, with :onflicting interests, different values, different languages, different concep-tualizations of the problem situation, and operating on different levels of authoriy. ~ow let us suppose that all these groups are represented in a design team. Let us ;uppose, furthermore, that the design team is able to develop a common language ind succeeds in transferring this language to all relevant subgroups. Let us suppose, idditionally, that similar agreements can be reached with respect to the initial
~onceptualization of the problem situation. Let us assume, finally, that conflicting interests can be dealt with satisfactorily. However, even if all these conditions are fulfilled, we canno-t MMM pJteci.oef.y how the c.onc.eptua.Uzatio;u, o6 the. p!tob.tem
de.6~nLtion witt change. due to the. ~p.te.me.ntatian a6 a p!topoJ.,e.d J.,o.tution ~n the
c.i{e.Kt oJtganization. This message can be drawn from behavioural-science literature on organizational change. For example, Chin and Benne(1969), summarizing strategies of change, emphasize that in so-called normative re-educative strategy, Jte~.>eaitc.h,
~ng and actLon cannot be seperated; furthermore, this strategy considers or-ganizational change as a .te.aJtning p!toce~.>~.>; both for the experts and for the client system.
Because of these insights, we advocate the te.Mrvi.ng app!Loac.h as depicted in Fig. 1.4. Defining the problem, designing the solution and organizational implemen-tation are viewed as an interactive, evolving, and iterative process, where experi-mentation, communication, and learning are central themes. Similar design approaches found in practice are described by Alter (1980, p. 170) as: " ... conscious design strategies by which irnplementers attempt to shorten the feedback loops between themselves and their clients and between their intentions and their products". The ideas behind our "learning" approach are also compatible with the ISAC approach described by Lundeberg et al. (1979a and 1979b). However, these authors do not provide guidelines for avoiding the danger of ,"redesigning the designs you have just redesigned". The learning approach presented here aims at avoiding this danger.
For a more detailed treatment of the learning approach, consider again Fig. 1.4. (A comprehensive discussion follows 'in Chapter 5). Four phases are dis-tinguished. Ideally, a project passes through the first and the last phases only once, whereas i t may pass through the second and third phase several times.
It goes without saying thata project such as that considered here should be started only after a thorough analysis of the existing tools has shown that these tools contain serious flaws, and that there is ample reason to expect major improve--ments from redesign. In the. description of our approach we will assume that an
entirely new control and information system is to be designed. In many cases this assumption will not be valid, but it is a convenient assumption for the present purpose of clarifying the learning approach.
The first phase, the de4~gn
o6
~he caneep~ual hYh~em. serves several purposes. First, it aims at creating a common language to be used by all parties involved in production control. Second, it tries to achieve agreement on the ultimate goals of production control activities and on the means available for attaining these goals. An important purpose of the first design phase is to specify which partsof the system may be subject to change during the subsequent design phase and which parts are fixed. Thus, concepts such as &!e!UbW.ty, adapta.b-U.Uy and tUg-UiLty
will play an important role in .the following chapters. Furthermore, in this phase a number of basic assumptions regarding the control problem at hand are developed. These assumptions serve as a basis for determining a proper decision procedure. If agreement has been reached on these points, the common language can serve as a basis for a definition of the type of data sets to be created and maintained before the ~xperimentation can start.
It should be noted that even the conceptual design phase requires considerable or-ganizational activities. These include training of the client system, or subsystem (transfer of production control theory and - to a minor extent - information system theory) as well as gathering production control information (transfer of knowledge to the experts). Once agreements about language' have been achieved, i t wi 11 be
I
"'
N .,... <:::: II)"'
s... 0 ...., s... II) ...., (/) ~"'
<:::: 0 .,... ...., conceptual design ·basic assump-tions ·language a lication experimental design observation ·eliminate obso-lete parts ·design solution control systemFig. 1.4. The teaAning appnoach
design of conceptual system design of experimental system experimentation phase consolidation phase
very difficult to change this common language. The same holds a fortiori for agree-ments on the conceptualization of the control problem. For this reason, we did not include a loop back to the conceptual design phase (in Fig. 1.4): £n auJt opbuan
the. p!toje.ct !.hould be" coMidM.ed ct majOJt Jte.vLi>ian ol) the. conce.ptuat de.'->ign p!tove.~ to be. ne.c..e.5Mlt!f cU: ct lctteA '->tage..
The second phase is the de.~ig>t the. expe.Jtime.ntat '->if'->·tern. This phase wi 11 often proceed in a piecemeal manner, because restructuring all decision functions simul-taneously would require too much effort. For each decision function, a decision-making strategy is outlined, and decision support models are designed. An information subsystem is designed to support decision-making according to the decision procedure chosen. Of course frequent feedback between Production Control System (PCS)-design and Production Control Information System (PCIS)-design will exist. Finally, a plan for implementation and experimentation must b~ developed. This plan includes model validation, performance measurement, user acceptance, and so on.
In this phase organizational activities are directed in the first place towards "creating an awareness of the need for change'and a climate of receptivity to change" (Alter, 1980, p. 144).
The third phase is the e.xpeJ!bne.nta:Uon phcu;e.. During this phase, a lot of infor-mation will be gained by the experts about the actual production control problem, and a 1 ot of specific production-contra 1 knowl;edge will be transferred to the c 1 ient system. Apart from being a tentative solution to the control problem, the experimen-tal system should also be a good vehicle for these processes.
Four effects of applying the experimental system can be distinguished: The definition of the production control problem may change. This is a point which is often overlooked in production-control system design. It means that information gained during experimentation may reveal that one or more of the assumptions underlying the experimental design is invalid. Furthermore. owing to new insights produced by the experimentation, the client system may wish to change the definition of the problem.
Some concepts or some pieces of theory may prove to be very difficult or even impossible to transfer to the client system. This will interfere with good inter-pretation of production control information based on these concepts.
The level of detail of the models used and of the definition of variables may be inadequate. For example, measuring the workload in terms of the number of jobs on the floor may yield the same control performance as measuring it in terms of operation time required, whereas the first measurement is much easier.
• Technical deficiencies of the PCS or the PCIS may be revealed.
At the end of the experimentation phase. information on all these aspects will be available simultaneously. Necessary changes in the organization, in the PCS, or in the PCIS, may lead to re-entry of the experimental design phase for the deci-sion point involved. However, before such a redesign is started. it is often more
practical to wait for the results of experimentation with other, related decision functions.
Finally, if a decision function is performing satisfactorily, the con~olidation
p~e can be started. In this phase, the task of monitoning the ~olution is trans-ferred to the client system. The system is simplified as much as possible, and various performance indicators are designed for the client organization. The expert activities are directed at documentation and withdrawal.
Finally, a short discussion of possible pitfalls in the learning approach is necessary.
In the first place, it has been mentioned earlier that top management app~oval is essential for the type of projects described here. The role of top management has both a promoting and a restraining aspect. The promoting aspect has been thor-oughly described in the literature (for example in Putnam, 1977). Less attention has been paid to the restraining aspect; a notable exception is the work of Plossl and Welch (1979), who give the following advice: "Beware of sophisticated techniques; better work is done with simple tools in skilled hands" (Plossl and Welch, 1979, p. 13). Indeed, the purpose of the learning approach is to develop simple, but effective tools and to support the decision-makers in becoming skilled. From a management point of view, the situation resembles the testing phase in product innovation or process innovation projects (Benjamin, 1971, and Batter, 1980, Chapter 9). In order to avoid endless experimentation, top management should demand a clear presentation of experimental results before a new experimentation cycle is entered.
In the second place, the clients (most often middle-management executives) should really paJ!.;i:icipate. A deadly enemy of the experimental approach is the "wait and see what the experts do" attitude on the part of the clients. If this occurs, the project will always remain in the conceptual design phase, because the conceptuali-zations of the problem remain different for the different parties involved. Further-more, the transfer of knowledge from the client system to the experts and vice versa will be hampered. Finally, without real involvement the client organization will certainly not be inclined to "ready adoption of a tool they have asked for", whereas such a favourable mood is required if the experimentation is to test the tools thoroughly.
Last but not least, the exp~ have a dual responsibility. On the one hand, they should build a system which meets the user's requirements, and modify this system in interaction with the user. On the other hand, the system should not be
the following risks:
(1) The system must be redesigned for each new person who becomes a user; (2) There will be no uniformit~ if more than one user is making the same type of decision;
(3) The system may not suit the executive who reviews the user's performance; (4) The system that is tailored to the individual style may reinforce unsuitable approaches to management and decision-making.
These dangers can only be avoided if the experts in the design team work as professionals following the guidelines described above. Our requirements for flexi-bility and adaptaflexi-bility are partly meant to deal with the first two problems whereas the professionalism of the experts should prevent the occurrence of the latter two risks.
Finally, the best proof of the experts' lack of professionalism is the presenta-tion to the client organizapresenta-tion of software which is not fully debugged; the learning approach should never be used as an excuse for such deficiencies.
1.5. Methodological problems
1 • 5. 1 • The emp..f/uc.at eyc£e.
We consider that empirical scientific knowledge develops in a number of steps (De Groot, 1961):
first, a set of concepts are developed for the description of an empirical field; (during the last decades the term ~digm has often been used for such a set of concepts);
second, a number of relations between these concepts are developed which leads to a "theory" or a set of related hypotheses; this process is sometimes called "induction";
third, these theories are expressed in such a form that they can be tested; this requires (logical) deduction and (empirical) specification of the theories and of the concepts developed;
fourth, the actual testing takes place;
fifth, the test results are evaluated, leading to the disproving or corroboration of the theory;
finally, if the above steps have been repeated successfully a number of times, the paradigm may be found usable for the empirical field; i f not, it will have to be reconsidered.
For the field of research of this book, the pattern of growth of empirical knowledge is more complicated; complications stem from the fact that the field of research is concerned with OJtgaruz~ovtJ.J and with du~gvt.
The problems of empirical investigations within organizations can be illustrated by a simple example. Consider the following formula describing a linear relation between the mean order flow-time, MFTA, and the mean utilization U, of a production unit:
MFTA a + sU
Suppose that an empirical corroboration of this formula for a certain production unit is to be examined. This can be done by deliberate variation of U, the independent variable, and by measuring MFTA, the dependent variable, in order to estimate the coefficients a and
e.
Thus, during the first period, U is to be kept at some value U'; in the next period, U is to be kept at another value, say 1.2U' and so on.Even such a very simple example shows the two problems encountered:
(1) In a real-life situation, it is seldom possible to ask a decision-maker to change his policies in a systematic manner for the purpose of scientific investiga-tion. If, in our example, the value of U really matters, it is highly improbable that researchers can systematically vary this factor. In other words, the independent variables are not subject to control by the researchers.
(2) The independent and dependent variables of the model should be measured. For example, perhaps the flow times and utilizations cannot be measured peA plwdue-~on unLt. Consequently, a separate, experimental information system should be designed and implemented. Such a system would incur costs, but no direct benefits for the organization may be expected. Even if such variables have already been measured, their exact definition may differ from the definition required for the experiment. Such a situation would also require a change of the information sys~ tern.
This means that empirical testing of even such a simple hypothesis often requires a change of the existing production control and information system. In other words, the empirical testing of hypotheses may require redesign and implementation of a control and information system.
1. 5. 3. Ve.s-i.gn
From the above discussion it follows that empirical testing of models and theories for the fields of research considered in this text, will often not be easy. On the other hand, design should be based on models and theories. This creates a vicious circle, from which one can escape by embodying small pieces of descriptive theoreti cal research into the design process. Nevertheless, the matter is rather complicated
because of the different goals for the two activities involved (viz. research and design). Empirical knowledge can be considered to grow by disproving theories (Popper, 1959). This requires that a designer applies a specific design in extreme situations. However, it is unrealistic to expect a designer in a field such as production control to implement a control system which might fail: the prime interest of a design activity is to construct a system which works.
Another complication lies in the fact that design not only relies on explicit theories; it also depends on a number of less formalized considerations which are sometimes called "know-how", and on a number of considerations applying to the spe-cific situation. Therefore, design failure not necessarily implies refutation of a well formulated theory; often such a failure may occur because one of these other factors have not been taken into account properly.
Finally, in a design situation, it is difficult to predict in advance which
hypothesis is going to be corroborated or refuted, if any. This makes a systematic approach to the growth of knowledge difficult; Even a well planned design project is often unpredictable with respect to its results from a scientific point of view (in comparison with situations where systematic experimentation is possible}. Infor-mation regarding the validity of assumptions underlying the design may become avail-able at random times, and in an indirect form.
Part of this book describes the development and application of a specific design approach, the learning approach. This approach has been used in a specific practical situation considered difficult from the point of view of production control and information systems, but which is not considered too difficult from the point of view of the other disciplines (such as social psychology, which is involved because of organizational change). During the design process the researchers explicitly formulate. the assumptions on which the design is based, as well as other assumptions of interest, which may be corroborated or refuted by the situation at hand. During implementation, budgets are available for data collection and testing of these assumptions, which may lead to a change of the design. The entire process gives rise, in turn, to an evaluation of the learning approach.
1.6. An outline of the book
From the above discussion, it will be clear that there should be a close inter-action between production-control systems {PCS) design and production-control infor-mation systems (PCIS) design; these aspects can be distinguished, but can rarely be separated. The interaction itself is one of the main topics of this book. Therefore the authors are happy to find themselves in a position where both aspects can be treated in one text.
The remainder of Part I (up to Chapter 5) is devoted to a description of relevant theories and design principles in the two fields - production control systems (PCS)
and production-control information systems {PCIS).
Part II (Chapter 6) reports on the results of theoretical research into the design of a production control system for the theoretical job-shop model. The func-tioning and performance of the designed decision procedures have been investigated by means of systematic computer simulations.
Part III gives an account of a research project in which the approaches described in Part I have been applied in practice. Chapter 7 describes the production system selected for this project. The conceptual design of the PCS (Chapter 8) and of the PCIS (Chapter 9) are presented next. For both the PCS and the PCIS the first experimental systems and the first experiences gained are described in the Chapters 10 to 14. The ultimate systems and the final experiences are presented in Chapters 15 to 18.
Finally, Part IV presents general conclusions. First, Chapter 19 discusses standard software packages for the production control situations considered (cf. Chapter 2). An approach is presented for evaluating such packages based on the theoretical concepts discussed in Part I. The control systems developed in Part II and Part III are used as an example in the description of the approach. The last Chapter, Chapter 20, evaluates the design principles treated in Part I on the basis of the results reported in Parts II and III.
2. PRODUCTION CONTROL: BASIC CONCEPTS 2.1. Introduction
In general, actual production systems can be considered as dynamic, stochastic and complex systems. These characteristics have implications for the structure of adequate production control systems. In this text we concentrate on these implica-tions for the structuring of production control problems and for the design of ad-equate production contra l systems.
We wi 11 build on concepts and results from: gevteJLai c,or1k'Wt theolty, which de a 1 s with the study of dynamic, stochastic control problems; in this theory complexity is often dealt with by decomposition of the problem. In Section 2.3 we will present our view in the form of some basic concepts, ~nd in Sections 2.4~ 2.5, and 2.6 the implications of this view are discussed; special attention is given to the impact of the characteristics of an organizational environment on the design of control systems.,
But first, in Section 2.2, we will isolate the production control problem con-sidered in this text from the general area of,production control. Some further re-strictions on the scope of the problem under consideration are presented in Sec-tion 2.7.
2.2. The production control function
According to Greene (1970), the production 'control function is defined as the set of activities in a production organization that are directed at the control of the volume and types of products produced at specific places as a function of time (Greene, 1970, pp. 1-4).
In this scope, production control includes ,long-range planning, product develop-ment, manufacturing process developdevelop-ment, customer service control, factory-layout planning, transportation and physical distribution, manpower planning, materials supply control and materials handling, capacity planning, scheduling, loading, dis-patching and expediting, and inventory control.
In this text attention is focussed only on a small part of the area described above. The following restrictions have been mape on the field:
• The a 11 ocati on of production facilities and planning of p 1 ant 1 ayout, as we 11 as the development, design and implementation of new: products, are not considered. These activities are considered as given; they spec~fy the constraints and goals for the more restricted production-control problem.
·The manufacturing process is considered as a black box, requiring a specific se-quence of operations to be applied to raw materials. Manufacturing operations are characterised by the specified. state of the input materia 1 s, by the sea 1 e and type of capacity required, and by the duration of the operation. Control problems related to the manufacturing technology are not considered.
·The quality control system is not considered. Within our scope, its functioning results in a probabilistic relation between the materials entering a specific pro-ductian stage and the products resulting from that stage - the production yield. ·We consider the control of volume of products only M a {Ju,nc.Uon o6 UJ1W,; the
spa-tial distribution of the goods produced and the allocation of production orders to production plants at different places are not considered.
·The control of materials supply is not considered; the availability of materials is assumed to follow some probability-density function.
·The technical and organizational aspects of inventory control and work-in-process control are not considered in detail.
2.3. A framework of concepts
2.3.1. IntAoducUon
Production control problems can be considered as part of the wider class of gen-eral control problems. The theory of gengen-eral control problems is well documented. On the one hand, this general control theory deals with the analysis of the behaviour of processes and controlled systems- attention is focussed on aspects such as stability, noise amplification and steady state behaviour. On the other hand, control theory deals ~ith the design of optimal control rules for specific situations, using tech-niqes such as dynamic programming or the Wiener filter theory.
Techniques and knowledge from general control theory have often been succesfully used in the study of specific pnototqpe pnoblern~ related to production-inventory control (e.g. Simon, 1952, Forrester, 1961, Adelson, 1966, Porter and Crossley, 1972, Schneeweisz, 1974, Gaalman, 1976, Bradshaw and Porter, 1978, and Bertrand, 1980). However the available techniques from general control theory can rarely be applied in a straightforward way to the design of production control systems Lit fYootco"'"'"''c·
In our opinion this is due to the wide gap existing between the relatively simple problems subjected to analytic treatment and the complex nature of actual production control problems (Tocher, 1970}. Nevertheless the basic concepts that are generally used for conceptualizing control problems, such as input, output, state variable, feedback loop, feed forward loop and disturbance, are generally used in mo.dern text-, books on production control (e.g. Mize, White and Brooks, 1971, Van Hees and Mon-hemius, 1972, Johnson, Newell and Vergin, 1972, and Wild, 1979}.
and thereafter they are primarily used as basic terms in discussing the nature of specific control problems and for reference to structural analogies with general control problems. Used in this way the concepts can help greatly in clarifying the nature of the control problems. However, the basic concepts are rarely used to exam-ine the special character of production control problems, or to highlight the differ-ence with control problems in the technical area nor are they used to establish the impact of the special character of production control problems on the design of a PCS.
In our opinion, the basic concepts of general control theory certainly are rel evant for understanding production-control systems design. Whether or not a specific production control problem is too complex for application of techniques from general control theory depends on the complexity of the process to be controlled. Thus, we should first eonc.eptual.ize the problem situation in terms of control theory and identify the processes to be controlled; then we should try to design usable madet6 of the process, i.e. models that can predict, to a reasonable extent, the behaviour of the process. Even if such a model is too complex for direct application of ana-lytic techniques from general control theory~ then we can still apply some basic principles from control theory for the design of good control systems.
In this section we will present a set of basic control-theory concepts; these concepts are relatively common in all modern literature in the field of general con-trol theory (e.g. Tou, 1964, Elgerd, 1967, and Sage, 1968).
In the next section we will investigate a number of implications that follow from the use of these concepts for the design of production control systems in practice.
2. 3. 2. B<t6)_e c.oneep:U
In the conceptual framework to be presented here, it is assumed that a complex production control situation can be decomposed into a number of more or less dis-tinct, simpler control problems. These simpler problems may require coordination This coordination will be discussed in Section 2.5; for the moment we will concen-trate on a distinct single control situation.
A control situation can be described as follows: exists,
·the decision function has noMn~, N(t), for a set of goal v~able~, D(t), ·the decision function can manipulate a set o!f eon:tJwiiabie. v~abiM, I ( t), ·the goal variables are influenced by the controllable variables,
·there exists a set of envVr.onmen.tal v<t'lhLbiM, E ( t) , which influence the goa 1 vari ables, and which cannot be manipulated by the decision function, but can be observed by the decision function,
·the relation between the controllable variables and the environmental variables, on the one hand, and the goal variables, on the other hand, is called the p~ae~6~ to
be controlled. This definition reflects our view that the nature of the process
6oliow~ from the definition of the goal variables and the controllable variables. The process is relative to these earlier definitions,
·if the process is dynamic, there exists a set of pltoc.u~.> ~.>£:ate va!Uab.tu, S ( t); state variables are defined so that the behaviour of the goal variables during some time interval (t,t+dt) is completely determined by the state at timet and by the behaviour of the inputs I(t) and E(t), during the time interval (t,t+dt),
·the decision function may receive information about the behaviour of the goal vari-ables, the state variables and the environmental variables. The decision function may use, either implicitly or explicitly, a mode£ of, the envL'Lonmenta£. ptwc.UJ.., Me' in order to produce predictions of future values of the environmental variables.
Ex.-p.tanatolty envL'Lonmenta£. va!Uablu, H{t), which are used as external inputs in the prediction model, may then be defined. If the environmental process is dynamic in nature, then env .. uwnmen:ta£. ,state va!UabLu, G(t), can be defined; in that case, M
8
operates on H{t) and G{t). The relation between H(t) and G(t), on the one hand, and E( t), Qn the other hand, is called the ertv..Uwnmentat y.vwc.U-6,
·the decision point uses, implicitly or explicitly, a made£ o6 the .. c..ontJwlied p.!Lo-ce..~.>-6, Mp' to evaluate possible decisions in light of the norms, the actual values of the state and goal variables and, possibly, predictions about the environmental variables.
Figure 2.1 presents the structure of a decision function according to these basic concepts. In the figure, I'(t) represents possible decisions that are investigated with respect to their expected result D'(t), given the state of the process S(t) and given the expected values of the environmental variables E'(t). Of course, the "what-if" character of the decision selection routine is used purely for illustration pur-poses. The decision function could also operate with a simple decision rule acting on E'(t) and S' (t), in which case the model is implicit- thus the models Me and Mp need not be explicitly used in the decision procedure. However, because in our view the models behind the procedures are the essence of the control, their presence is stressed in Fig. 2.1.
Generally, the models Me and Mp will not be perfect: not all variables influencing O(t) or E(t} in reality will be incorporated; their relations may really be much more complex; the observations of the decision function with respect to H(t), S(t) and G{t) may contain errors; and finally the decision that follows from the appli-cation of the formal decision procedure may not always be completely implemented, for various reasons. In our view on control processes, each of these different sources of uncertainty constitutes a d..Lotu.Jr..barwe... The concept of disturbance com-prises all factors that affect the predictive quality of the models Mp and Me. Furhtermore, it includes errors in the information available for decision making and deviations that occur in the implementation of decisions.
observe environmental variables G(t) observe environmental state variables
r---1 I I N(t} mapping Me E(t) implement decision D(t) observe process state variables-
-
-- -
- ---
----,
select > - - - t - . t a 1 tern at i ve decision I'(t) mapping M p l I l L __ _ - - - - _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ J decl6ion 6unction Idetermined partly by the quality of the models used.
2.4. Models
2.4.1. lntnoduction
From the previous section it will be clear that our view of production-control system design emphasizes the modelling of the process to be controlled. The relevance of explicit process models for the control of complex processes has frequently been stressed in control literature during the last decades (e.g. Ashby, 1956, Conant and Ashby, 1970, Beer, 1966, and Tocher, 1970). We fully agree with Tocher (1970) who, discussing the predictability of the results of alternative control actions, states that: " .... This can be achieved only by the use of a model of the controlled process. At the heart of every control system there is a model. This may be very rudimentary,
its existence may not impinge on the consciousness, but without it control is im-possible". Because production control is an aspect of the functioning of an organiza-tion that may be related to other control aspects, we feel that the concept of process modelling needs some further discussion.
Production control decisions are made by people, and people have me~ mode£6
of the process. Landeweerd (1978) describes the mental process model of a decision maker as the knowledge he has regarding:
·the relations between changes in inputs to the process and the resulting changes in the outputs of the process,
·the relations between the undesired outputs and the control actions that should lead to desired outputs.
These mental models often are richer than the formal process models on which, for instance, a computerized decision routine or a rule of thumb is based (see Fig. 2.2). Besides, it may not be expected that production processes remain stationary; thus provisions should be made for the monitoring and the adaptation of the control rules.
In this section we will discuss modelling as a part of the design activity; in the next section we will discuss the difference between the me~ model. used by the decision maker and the t)of1m~Lt decL~;_on p!Wc.edUJte that can be used to support the decision-making process.
For convenience we will discuss here only the modelling of the controlled process; a similar reasoning can be applied to the modelling of the environmental process.
Consider the position of a designer of a production control system. After identi fication of the controllable variables, the environmental variables and the goal variables, the next step to take is to decide how sophisticated the formal decision
formal decision procedure decision function goal variables
Fig. 2. 2. Symbolic. ne.p~te..oe.nt:atJ.an a6 the. menta£ made.! and the. 6oJLma! de.c.-U...ian p!Lac.e.dwr.e. u.oe.d bt the. de.c.L~-Zon-mai<J.ng p!Lac.u.o.
procedure should be for the proper control of the process. For this decision, the designer may consider various process models having different levels of sophistica-tion and detail; he will have to evaluate these process models on two aspects: .quality. This refers to the level of disturbance implied by the use of a process
model, which is closely related to the ques~ion of the control performance that can be obtained by using the process model;
.ca.ot~. This refers to the organizational effort required for the development and operation a 1 use of a decision procedure deri,ved from this process mode 1.
The performance to be obtained should, of ,course, justify the efforts. The results of this evaluation process may be that no process model can be found with sufficient predictive quality to justify the design of a' formal decision procedure. In that case the best thing we can do is to design an information system which presents informa-tion about the relevant environmental and state variables to the decision maker. The decision maker uses his mental model of the process to make adequate decisions based on this formal information and perhaps on additional informal information. However, even the design of an information system that reports on state variables only assumes
