Artificial Intelligence in Industrial Decision Making, Control and - - PDF document

Artificial Intelligence in Industrial Decision Making, Control and Automation

edited by

SPYROS G. TZAFESTAS

Department ofElectrical and Computer Engineering, National Technical University of Athens, Athens, Greece and

HENK B. VERBRUGGEN

Department ofElectrical Engineering, Delft University of Technology, Delft, The Netherlands

KLUWER ACADEMIC PUBLISHERS

DORDRECHT / BOSTON / LONDON

CONTENTS

Preface Contributors

PART 1 GENERAL ISSUES

CHAPTER1 ARTIFICIAL INTELLIGENCE IN INDUSTRIAL DECISION MAKING, CONTROL AND AUTOMATION: AN INTRODUCTION

• S. Tzafestas and H. Verbruggen
• 1. Introduction

^

• 2. Decision Making, Control and Automation

, 2

2.1. Decision Making Theory

2 2.2. Control and Automation 4

• 3. Artificial Intelligence Methodologies

6 3.1 Reasoning under uncertainty 7 3.2 Qualitative reasoning 14 3.3 Neural nets reasoning ^

• 4. Artificial Intelligence in Decision Making

19

• 5. Artificial Intelligence in Control and Supervision

...22

• 6. Artificial Intelligence in Engineering Fault Diagnosis

24

• 7. Artificial Intelligence in Robotic and Manufacturing Systems

26

• 8. Conclusions

™ References TI

VI

CHAPTER 2 CONCEPTUAL INTEGRATION OF QUALITATIVE AND QUANTITATIVE PROCESS MODELS

• E. A. Woods
• 1. Introduction

41

• 2. Qualitative Reasoning

42

2.1. Common Concepts

43 2.2. Qualitative Mathematics 44 2.3. The notion of State 45 2.4. Describing Behaviour 45 2.5. Components of qualitative reasoning 45 2.6. Towards more quantitative modeis 47

• 3. Formal Concepts and Relations in the HPT

48 3.1. Quantities 48 3.2. Physical Objects, process equipment, materials and substances 48 3.3. The input file 49 3.4. Activity conditions 49 3.5. Numerical functions and influences 50 3.6. Logical relations and rules 52

• 4. Defining Views and Phenomena

52

4.1. Individuais and individual conditions

52 4.2. Quantity conditions and preconditions 54 4.3. Relations 55 4.4. Dynamic influences 56 4.5. Instantiating a definition 57 4.6. Activity levels 57

• 5. Deriving and Reasoning with an HPT Model

59 5.1. Extending the topological model 59 5.2. Deriving the phenomenological model 60 5.3. Activity and State space modeis 61

• 6. Discussion and Conclusion

63 References 64

vii

CHAPTER 3 TIMING PROBLEMS AND THEIR HANDLING AT SYSTEM INTEGRATION

• L. Motus

;

• 1. Introduction

57

• 2. Essential Features of Control Systems

68

2.1. Essential (forced) concurrency

70 2.2. Truly asynchronous mode of execution of interacting procsses 70 2.3. Time-selective interprocess communication 71

• 3. Concerning Time-Correct Functioning of Systems

71 3.1. Performance-bound properties 72 3.2. Timewise correctness of events and data 72 3.3. Time correctness of interprocess communication 73

• 4. A Mathematical Model for Quantitative Timing Analysis (Q-Model)

73

4.1. Paradigms used

....... 74 4.2. The Q-model 74

• 5. The Q-Model Based Analytical Study of System Properties

76 5.1. Separate elements of a specification 76 5.2. Pairs of interacting processes 77 5.3. Group of interacting processes .......78

• 6. An example of the Q-Model Application

79

• 7. Conclusions

g5 References g5 CHAPTER 4 ANALYSIS FOR CORRECT REASONING IN INTERACTIVE MAN ROBOT SYSTEMS: DISJUNCTIVE SYLLOGISM WITH MODUS PONENS AND MODUS TOLLENS

• E. C. Koenig
• 1. Introduction

gg

• 2. Valid Command Arguments

90

VU1

• 3. Correct Reasoning: Disjunctive Syllogism

91 3.1. Plausible composite command arguments 92 3.2. Plausible composite commands 92

• 4. Conclusions

96 References 96 PART 2 INTELLIGENT SYSTEMS CHAPTER 5 APPLIED INTELLIGENT CONTROL SYSTEMS

• R. Shoureshi, M. Wheeler and L. Brackney
• 1. Introduction

101

• 2. A Proposed Structure for Intelligent Control Systems (ICS)

102

• 3. Intelligent Automatic Generation Control (IAGC)

105

• 4. Intelligent Comfort Control System

110

• 5. Control System Development

111

• 6. Experimental Results

116

• 7. Conelusion

116 References 119 CHAPTER 6 INTELLIGENT SIMULATION IN DESIGNING COMPLEX DYNAMIC CONTROL SYSTEMS

• F. Zhao
• 1. Introducton

127

• 2. The Control Engineer's Workbench

128

ix

• 3. Automatic Control Synthesis in Phase Space

128 3.1. Overview of the phase space navigator 129 3.2. Intelligent navigation in phase space 129 3.3. Planning control paths with flow pipes 130

• 4. The Phase Space Navigator

131

4.1. Reference trajectory generation

131 4.2. Reference trajectory tracking 133 4.3. The autonomous control synthesis algorithms 135 4.4. Discussion of the synthesis algorithms 137

• 5. An Illustration: Stabilizing a Bückling Column

139 5.1. The column model 140 5.2. Extracting and representing qualitative phase-space structure of the buckling column 141 5.3. Synthesizing control laws for stabilizing the column 143 5.4. The phase-space modeling makes the global navigation possible 148

• 6. An application: Maglev Controller Design

148 6.1. The maglev model 148 6.2. Phase-space control trajectory design 150

• 7. Discussions

155

• 8. Conclusions

155 References 156 CHAPTER 7 MULTIRESOLUTIONAL ARCHITECTURES FOR AUTONOMOUS SYSTEMS WITH INCOMPLETE AND INADEQUATE KNOWLEDGE REPRESENTATION

• A. Meystel
• 1. Introduction

I59

• 2. Architectures for Intelligent Control Systems: Terminology, Issues, and a

Conceptual Framework 161

2.1. Definitions

161 2.2. Issues and problems 165

X

2.3. Conceptual framework for intelligent Systems architecture 170

• 3. Overview of the General Results

171

• 4. Evolution of the Multiresolutional Control Architecture (MCA): Its Active

and Reactive Components 173

4.1. General structure of the Controller

173 4.2. Multiresolutional control architecture (MCA) 175

• 5. Nested Control Strategy: Generation of a Nested Hierarchy for MCA

177 5.1. GFACS triplet: Generation of intelligent behavior 177 5.2. Off-line decision making procedures of planning-control in MCA 178 5.3. Generalised Controller 180 5.4. Universe of the trajectory generator: Second level 181 5.5. Representation of the planning/control problem in MCA 183 5.6. Search as the general control strategy for MCA 185

• 6. Elements of the Theory of Nested Multiresolutional Control for MCA

187 6.1. Commutative diagram for a nested multiresolutional Controller 187 6.2. Tessellated knowledge bases 187 6.3. Generalization 188 6.4. Attention and consecutive refinement 189 6.5. Accuracy and resolution of representation 190 6.6. Complexity and tessellation: e-entropy 194

• 7. MCA in Autonomous Control System

195 7.1. The multiresolutional generalization of System modeis 195 7.2. Perception stratified by resolution 196 7.3. Maps of the world stratified by resolution 197

• 8. Development of Algorithms for MCA

198 8.1. Extensions of the Bellman'soptimality principle 198 8.2. Nested Multiresolutional search in the State space 198

• 9. Complexity of Knowledge Representation and Manipulation

201 9.1. Multiresolutional consecutive refinement: Search in the State space 201 9.2. Multiresolutional consecutive refinement: Multiresolutional search

• f a trajectory in the State space

203 9.3. Evaluation and minimization of the complexity of the MCA 205

• 10. CaseStudies

208 10.1 A pilot for an autonomous robot (two levels of resolution) 208

XI

10.2 PILOT with two agents for control (a case of behavioral duality) 211

• 11. Conclusion

...219 References ...220 CHAPTER 8 DISTRIBUTED INTELLIGENT SYSTEMS IN CELLULAR ROBOTICS

• T. Fukuda, T. Ueyama and K. Sekiyama
• 1. Introduction

225 2.Concept of Cellular Robotic System 226

• 3. Prototypes of CEBOT

227 3.1. Prototype CEBOT Mark IV 229 3.2. Cellular Manipulator 231

• 4. Distributed Genetic Algorithm

234

4.1. Distributed Decision Making

234 4.2. Structure configuration problem 235 4.3. Application of genetic algorithm 236 4.4. Distributed genetic algorithm 239 4.5. Simulation results 241

• 5. Conclusions

245 References 245 CHAPTER 9 DISTRIBUTED ARTIFICIAL INTELLIGENCE IN MANUFACTURING CONTROL

• S. Albayrak and H. Krallmann
• 1. Introduction

247

• 2. Tasks of Manufacturing Control

248

• 3. The State-of-the-Art of the DAI Technique in Manufacturing Control

252 3.1. ISIS/OPIS 9S9

xö

3.2. SOJA/SONIA 254 3.3. YAMS 255

• 4. Distributed Artificial Intelligence

259

4.1. Cooperative problem solving

261 4.2. Phases of cooperating problem solving 261 4.3. Blackboard metaphor, model and frameworks 264 4.4. History of the blackboard model 274 4.5. Advantages of DAI 276

• 5. VerFlex - BB System: Approach and Implementation

277 5.1. Distributed approach to the Solution of the task order execution 277 5.2. Why was the blackboard model used? 281 5.3. The VerFlex - BB System 281 References 292

PART 3 NEURAL NETWORKS IN MODELLING, CONTROL AND SCHEDULING

CHAPTER 10 ARTIFICIAL NEURAL NETWORKS FOR MODELLING A.J. Krijgsman, H.B. Verbruggen and P.M. Bruijn

• 1. Introduction

297

• 2. Description of artificial neurons

298

• 3. Artificial neural networks (ANN)

299

• 4. Nonlinear modeis and ANN

300

• 5. Networks

302 5.1. Mtiltilayered static neural networks 302 5.2. Radial basis function networks 303 5.3. Cerebellum model articulation Controller (CMAC) 304

• 6. Identification of Dynamic Systems Using ANN

306

xiii

6.1. Identification problem defmition 306 6.2. Modeidescription foridentification 308

• 7. Hybrid Modelling

308 Orthogonal least-squares algorithm.; 309

• 8. Model Validation

313

• 9. Experiments and Results Using Neural Identification

314

• 10. Conclusions

323 References 323 CHAPTER11 NEURAL NETWORKS IN ROBOT CONTROL S.G. Tzafestas

• 1. Introduction

327

• 2. Neurocontrol Architectures

328

2.1. General issues

328 2.2. Unsupervised NN control architectures 329 2.3. DIMA II. Neurocontroller for linear Systems 331 2.4. Adaptive learning neurocontrol for CARMA Systems 336

• 3. Robot Neurocontrol

339 3.1. A look at robotics 339 3.2. Neural nets in robotics: General review 341 3.3. Robot control using hierarchical NNs 343 3.4. Minimum torque-change robot neurocontrol 346 3.5. Improved iterative learning robot neurocontroller 349

• 4. Numerical Examples

352

4.1. Example 1: DIMA II Controller for linear Systems

352 4.2. Example 2: Neurocontroller for CARMA Systems 354 4.3. Example 3: Supervised neurocontrol of a broom - balancing System 357 4.4. Example 4: Feedback - error learning robot neurocontrol 361 4.5. Example 5: Iterative robot neurocontrol 366 4.6. Example 6: Unsupervised robot-neurocontroller using hierarchical NN 372

• 5. Conclusions and Discussion

375

XIV

• 6. Appendix: A Brief Look at Neural Networks

376 6.1. Single - layer perceptron (SLP) 377 6.2. Multi - layer perceptron (MLP) 37g 6.3. Hopfield network 3g j References 3g4 CHAPTER 12 CONTROL STRATEGY OF ROBOTIC MANIPULATOR BASED ON FLEXIBLE NEURAL NETWORK STRUCTURE

• M. Teshnehlab and K. Watanabe
• 1. Introduction

3g9

• 2. The Representation of Bipolar Unit Function

..390

• 3. Learning Architecture

391 3.1. The learning of connection weights 392 3.2. The learning of sigmoid unit function parameters 393

• 4. Neural Network - Based Adaptive Controller

394

4.1. The feedback - error learning rule

395 4.2. Adaptation of neural network Controller 396

• 5. Simulation Example

39-7

• 6. Conclusion

4Q2 References 4Q2 CHAPTER 13 NEURO - FUZZY APPROACHES TO ANTICIPATORY CONTROL

• LXI. Tsoukalas, A. Ikonomopoulos and R.E. Uhrig
• 1. Introduction

4Q5

• 2. IssuesofFormalism Anticipatory Systems

407

• 3. Issues of Measurement and Prediction

412

• 4. Conclusions

417 References 410

XV

CHAPTER 14 NEW APPROACHES TO LARGE - SCALE SCHEDULING PROBLEMS: CONSTRAINT DIRECTED PROGRAMMING AND NEURAL NETWORKS

• Y. Kobayashi and H. Nonaka
• 1. Introduction

42i

• 2. Method

.......422

2.1. Problem and method description

422 2.2. Knowledge - based method for lower - level problems 424 2.3. Knowledge - based scheduling method for upper- level problems 431 2.4. Neural networks for upper - level problems 432

• 3. Application Examples

439 3.1. Scheduling Systems .....439 3.2. Problem 439 3.3. Results 439

• 4. Conclusions

444 References 445

PART 4 SYSTEM DIAGNOSTICS

CHAPTER 15 KNOWLEDGE - BASED FAULT DIAGNOSIS OF TECHNOLOGICAL SYSTEMS

• H. Verbruggen, S. Tzafestas and E. Zanni
• 1. Introduction

449

• 2. Knowledge Representation and Acquisition for Fault Diagnosis

451

2.1. Knowledge representation

451

XVI

2.2. Knowledge acquisition 454

• 3. First -and Second - Generation Diagnostic Expert Systems

456 3.1. General issues 456 3.2. First- generation expert Systems 456 3.3. Deep reasoning 457 3.4. Qualitative reasoning 458 3.5. Second - generation expert Systems 462

• 4. A General Look at the FD Methodologies and Second - Generation ES

Architectures 462

4.1. General issues

462 4.2. Diagnostic modelling 463 4.3 Second - generation FD expert System architectures 464

• 5. A Survey of Digital Systems Diagnostic Tools

467 5.1.TheD-algorithm 467 5.2. Davis' diagnostic methodology 468 5.3. Integrated diagnostic model (IDM) 470 5.4. The diagnostic assistance reference tool (DART) 472 5.5 The intelligent diagnostic tool (IDT) 474 5.6. The Lockheed expert System (LES) 476 5.7. Other Systems 476

• 6. A General Methodology for the Development of FD Tools in the Digital

Circuits Domain 477 6.1. Description of the structure 478 6.2. Description of the behaviour 479 6.3. The diagnostic mechanism 480 6.4. The constraint Suspension technique 482 6.5. Advantages of the deviation detection and constraint Suspension technique 485

• 7. A General Methodology for the Development of FD Tools in the

Process Engineering Domain 486

• 8. Implementation of a Digital Circuits Diagnostic Expert System (DICIDEX) 489

8.1.Introduction 489 8.2. Dicidex description 490 8.3. Examples of System - user dialogues 496

xvu

• 9. Conclusions

501 References 502 CHAPTER 16 MODEL - BASED DIAGNOSIS: STATE TRANSITION EVENTS AND CONSTRAINT EQUATIONS K.-E. Arzen, A. Wallen and T.F. Petti

• 1. Introduction

507

• 2. Diagnostic Model Processor Method (DMP)

509

• 3. Model Integrated Diagnosis Analysis System (MIDAS)

512 3.1. MIDAS modeis 512 3.2. MIDAS diagnosis 515

• 4. Steritherm Diagnosis

518

4.1. DMP Steritherm diagnosis

518 4.2. MIDAS Steritherm diagnosis 519

• 5. Comparisons

520

• 6. Conclusions

522 References 523 CHAPTER 17 DIAGNOSIS WITH EXPLICIT MODELS OF GOALS AND FUNCTIONS J.E. Larsson

• 1. Introduction

525

• 2. Basic Ideas in Multilevel Flow Modeling (MFM)

526

• 3. An Example of a Flow Model

526

• 4. Three Diagnostic Methods

528 4.1 Measurement Validation 529 4.2. Alarm analysis 530 4.2. Fault Diagnosis 531

xviii

• 5. Implementation

...531

• 6. Complex Systems

532

• 7. Conclusions

532 References 533

PART 5 INDUSTRIAL ROBOTIC, MANUFACTURING AND ORGANIZATIONAL SYSTEMS

CHAPTER 18 MULTI-SENSOR INTEGRATION FOR MOBILE ROBOT NAVIGATION A.Traca de Almeida, H. Araujo, J. Dias and U. Nunes

• 1. Introduction

537

• 2. Sensor-Based Navigation

537

• 3. Sensory System

53g

• 4. Sensor Integration for Localization : Some Methodologies

540

4.1. Data Integration - Intrinsic sensor level

542 4.2. Data Integration - Extrinsic sensor level 544

• 5. Experimental Setup

547 5.1. Sensors' descriptions 547

• 6. Conclusions

553 References cco

XIX

CHAPTER 19 INCREMENTAL DESIGN OF A FLEXIBLE ROBOTIC ASSEMBLY CELL USING REACTIVE ROBOTS E.S. Tzafestas and S.G. Tzafestas

• 1. Introduction

555

• 2. Description of the Assembly Cell

556

• 3. Basic Architecture of the Robot

559

• 4. Case 1: Theminimal Assembly Cell

561

• 5. Case 2: Extending the Robots Architecture

562

• 6. Case 3: Using More than one Assembly Robots

563

• 7. Case 4: Combining Cases 2 and 3-Interacting Factors

565

• 8. Case 5: The Adaptive Robot - Commitment to Product

567

• 9. Conclusions and Further Work

569 References 570 CHAPTER 20 ON THE COMPARISON OF AI AND DAI BASED PLANNING TECHNIQUES FOR AUTOMATED MANUFACTURING SYSTEMS A.I. Kokkinaki and K.P. Valavanis

• 1. Introduction

573

• 2. Traditional Artificial Intelligence Planning Systems

575

2.1. Theorem proving based planning Systems

577 2.2. Blackboard-based architectures 579 2.3. Assembly planning and assembly sequences representations 582

• 3. Distributed Artificial Intelligence Planning Systems

593 3.1. Coordination in multi-agent planning 594 3.2. Theories of belief 595 3.3. Synchronization of multi-agents 595

• 4. Distributed Planning Systems

596

4.1. Route planning using distributed techniques

596 4.2. Distributed NOAH 600

XX

• 5. Distributed Planning Synchronization examples

601 5.1. CSP influenced synchronization method 601 5.2. Partial plan synchronization 605 5.3. Logic based plan synchronization 606

• 6. Application of Learning to Planning

608

• 7. Conclusions

610 References 612 CHAPTER 21 KNOWLEDGE-BASED SUPERVISION OF FLEXIBLE MANUFACTURING SYSTEMS

• A. K. A. Toguyeni, E. Craye and J.-C. Gentina
• 1. Supervision and AI-Techniques

631

• 2. Piloting Functions

632

2.1. Introduction

632 2.2. Problems met from design to implementation 633 2.3. The knowledge-based System 634 2.4. Conclusion 637

• 3. Manager ofWorking Modes

637 3.1. Introduction 637 3.2. Representation and modelling of the process 638 3.3. The manager framework 642 3.4. Conclusion 648

• 4. A Model-Based Diagnostic System for On-Line Monitoring

650

4.1. Introduction

650 4.2. The modelling method 650 4.3. The causal temporal signature or CTS 651 4.4. The multi-agent framework of diagnostic System 655 4.5. Conclusion 660

• 5. General Conclusion

660 References 661

XXI

CHAPTER 22 A SURVEY OF KNOWLEDGE-BASED INDUSTRIAL SCHEDULING

• K. S. Hindi and M. G. Singh
• 1. Introduction

663

• 2. Knowlegde Acquisition

664

• 3. Knowledge Representation

665 3.1. Logic-based Systems 665 3.2. Rule-based Systems 666 3.3. Frame-based Systems 667 3.4. Multi knowledge representation Systems 668

• 4. Temporal Issues

669

• 5. Control Mechanisms

670 5.1. Forward reasoning Systems 670 5.2. Constraint-directed and opportunistic Systems 671 5.3. Mixed control Systems 673

• 6. Knowledge Based Scheduling Systems (KBSS)

674 6.1. The primary scheduler (PS) 675 6.2. The heuristic scheduler (HS) 676 6.2. The backtracking scheduler (BS) 677

• 7. Reactive and Real-Time Scheduling

678

• 8. Conclusions

679 References 680 CHAPTER 23 REACTIVE BATCH SCHEDULING

• V. J. Terpstra and H. B. Verbruggen
• 1. Introduction

688 l.l.Project 688 1.2. Scheduling 688 1.3. Example case 689 1.4. Definitions 690

XX11

• 2. Scheduling strategy

^ni

2.1. Modelling

go2 2.2. Modularity ^92 2.3. Prediction and cycles ^93 2.4. Reactive behaviour gen 2.5. Robustness z-04

• 3. Modelling

6 9 4

3.1. Theequipment model ^95 3.2. The master reeipe 697 3.3. Master schedule ^08 3.4. The degrees of freedom of the scheduler 699

• 4. Planner

6 9 9

• 5. Integer scheduler

™Q

• 6. Non-integer scheduler

704

6.1. Ganeration of NLP model 704 6.2. Dedicated NLP solver

707

• 7. Reactiveness

7^o

7.1. Horizons 7^o 7.2. Sample Rate 7^9 7.3. Three Control Loops in Scheduler .' 709 7.4 Error Signal

7 1Q

7.5. Timing

7 1 1

7.6. Progressive Reasoning 7^3 7.7. Anticipatory Schedules

7 1 4

7.8. Parallelism

71fi

• 8. Robustness analysis

71fi

• 9. Implementation and Results

7 1 9

• 10. Conclusions

7^^ References

7~^

XXU1

CHAPTER 24 APPLYING GROUPWARE TECHNOLOGIES TO SUPPORT MANAGEMENT IN ORGANIZATIONS

• A. Michailidis, P.-I. Gouma and R. Rada
• 1. Introduction

723

• 2. Groupware

723

2.1. Groups and Computer-supported cooperative work

724 2.2. Groupware taxonomy 724 2.3.Review of groupware Systems 728

• 3. Management

729 3.1. Organizations 730 3.2. Managing organizations 733 3.3. IT Systems for management-support in organizations 735 3.4. Comparing R&D department with organizations 737

• 4. Case Study

738

4.1. Modelling the organizational structure

739 4.2. The activity model environment (AME) model 739 4.3. The modified Version of AME 740

• 5. Implementation- The MUCH System

745

• 6. Conclusion

747 References 748 INDEX 757