Applied Information Technology – serie

It is in the area of Systems Diagnosis. Supervision and Control that Knowledge-Based Techniques have had their most significant impact in recent years. In this volume. Spyros Tzafestas has ably put together the current state of the art of the application of Artificial Intelligence concepts to problems of Systems Diagnosis. All the authors in this edited work are distinguished internationally. recognized experts on various aspects of Artificial Intelligence and its applications. and the coverage of the field that they provide is both readable and authoritative. The sixteen chapters break down in a natural way into three broad categories i.e •• (a) introduction to the applications of Expert Systems in Engineering. (b) Knowledge-based systems architectures. models and techniques for fault diagnosis. supervision and real time control and finally. (c) applications and case studies in three specific 'areas. namely: Manufacturing. Chemical Processes and Communications Networks. The final chapter provides a com­ prehensive survey of the field with an extensive bibliography. The mix of original scientific articles. tutorial and survey papers makes this col­ lection a very timely and valuable addition to the literature in this important field. MADAN G. SINGH Professor of Information Engineering at U.M.I.S.T.

Since the begining of the sixties, control theorists have developed a large body of knowledge concerning complex or large-scale systems theory. Using the state space approach, their purpose was to extend methods to cope with the increasingly sophisticated automation needs of man-made systems. Despite several remarkable contributions, and some successful applications, it can be stated that this theory has not yet become an engineering tool. On the other hand, the emergence of cheap and reliable microprocessors has profoundly transformed industrial instrumentation and control systems. Process control equipment is organized in multilevel distributed structures, closely related to the concepts introduced by complex systems control theory. This similarity should favor a fruitful intersection for practical applications. However, a gap still exists between the literature on control theory and the world of technological achievements. In the many books on complex systems, few have given attention to the technological aspects of a practical control problem. The present book is an attempt to fill this gap. To do this, it consistently reflects the viewpoints that: - Theory and technology are two indivisible facets of the same problem. -On-line implementation for real time applications is the ultimate goal of a control study.

The main characteristics of the real-world decision-making problems facing humans today are multidimensional and have multiple objectives including eco­ nomic, environmental, social, and technical ones. Hence, it seems natural that the consideration of many objectives in the actual decision-making process re­ quires multiobjective approaches rather than single-objective. One ofthe major systems-analytic multiobjective approaches to decision-making under constraints is multiobjective optimization as a generalization of traditional single-objective optimization. Although multiobjective optimization problems differ from single­ objective optimization problems only in the plurality of objective functions, it is significant to realize that multiple objectives are often noncom mensurable and conflict with each other in multiobjective optimization problems. With this ob­ servation, in multiobjective optimization, the notion of Pareto optimality or effi­ ciency has been introduced instead of the optimality concept for single-objective optimization. However, decisions with Pareto optimality or efficiency are not uniquely determined; the final decision must be selected from among the set of Pareto optimal or efficient solutions. Therefore, the question is, how does one find the preferred point as a compromise or satisficing solution with rational pro­ cedure? This is the starting point of multiobjective optimization. To be more specific, the aim is to determine how one derives a compromise or satisficing so­ lution of a decision maker (DM), which well represents the subjective judgments, from a Pareto optimal or an efficient solution set.

A modern mechanical structure must work at high speed and with high precision in space and time, in cooperation with other machines and systems. All this requires accurate dynamic modelling, for instance, recognizing Coriolis and centrifugal forces, strong coupling effects, flexibility of links, large angles articulation. This leads to a motion equation which must be highly nonlinear to describe the reality. r1oreover, work on the manufacturing floor requires coordination between nachines, between each machine and a conveyor, and demands robustness of the controllers against uncertainty in payload, gravity, external perturbations etc. This requires adaptive controllers and system coordination, and perhaps a self organizing structure. The machines become complex, strongly nonlinear and strongly coupled mechanical systems with many degrees of freedom, controlled by sophisticated mathematical programs. The design of such systems needs basic research in Control and System Dynamics, as well as in Decision Making Theory (Dynamic Games), not only in the use of these disciplines, but in their adjustment to the present demand. This in turn generates the need to prepare engineering students for the job by the teaching of more sophisti­ cated techniques in control and Mechanics than those contained in previous curricula. On the other hand, all that was mentioned above regarding the design of machines applies equally well to other presently designed and used mechan­ ical structures or systems.

This volume collects most of the papers presented at the International Workshop on Robustness in Identification and Control, held in Torino (Italy) in 1988. The main focal point of the workshop was Unknown But Bounded uncertainty and associated robustness issues in identification and control. Recent years have seen a growing interest in studying models which include un­ known but bounded uncertainty. The motivation for dealing with such models is derived from robustness considerations. In many applications, some performance specification must be met for all admissible variations of the uncertain parameters. A second motivation for models with this type of uncertainty stems from the fact that the statistical description of uncertain variables may not be well known or even not suitable. For example, in some cases, only a small number of measurements is available and the resulting errors are due to analog-digital conversion, modelling ap­ proximation or round-off, so that a statistical description may actually be unreliable. The interest in unknown but bounded setting is certainly not new. In fact, en­ gineering practice demands for appropriate algorithms in dealing with finite sample properties, finite parameter variations, tolerance analysis, etc. Despite the natural need for such methods, the lack of sufficiently well assessed theoretical results and algorithms prevented a systematic use of these procedures until recent years. How­ ever, in the last few years, important advances have been made both in estimation theory and in stability analysis.

This volume provides the proceedings of the First European Workshop on Parallel Processing Techniques for Simulation which was held at the end of October 1985. The Workshop was organized within the framework of a joint project sponsored by the Commission of the European Communities under the research part of the multiannua1 programme in the field of Data Processing aming at promoting collaborative research work in the Community. The project involved collaborative work between the Complex Systems Group of the Control Systems Centre at UMIST, the Systems Reliability Service of the United Kingdom Atomic Energy Authority and the University of Bergamo, Italy. The aim of this project was to develop decomposition coordination techniques which would be of help in the simulation of complex dynamical systems on parallel processing facilities. One of the major aims of the Workshop was to report on the results produced within the project and to try to relate these to the leading work going on in this field in other centres of excellence. With this in mind, the Proceedings Volume is split up into a number of parts corresponding to the main sessions within the Workshop programme.

Since the begining of the sixties, control theorists have developed a large body of knowledge concerning complex or large-scale systems theory. Using the state space approach, their purpose was to extend methods to cope with the increasingly sophisticated automation needs of man-made systems. Despite several remarkable contributions, and some successful applications, it can be stated that this theory has not yet become an engineering tool. On the other hand, the emergence of cheap and reliable microprocessors has profoundly transformed industrial instrumentation and control systems. Process control equipment is organized in multilevel distributed structures, closely related to the concepts introduced by complex systems control theory. This similarity should favor a fruitful intersection for practical applications. However, a gap still exists between the literature on control theory and the world of technological achievements. In the many books on complex systems, few have given attention to the technological aspects of a practical control problem. The present book is an attempt to fill this gap. To do this, it consistently reflects the viewpoints that: - Theory and technology are two indivisible facets of the same problem. -On-line implementation for real time applications is the ultimate goal of a control study.

The main characteristics of the real-world decision-making problems facing humans today are multidimensional and have multiple objectives including eco­ nomic, environmental, social, and technical ones. Hence, it seems natural that the consideration of many objectives in the actual decision-making process re­ quires multiobjective approaches rather than single-objective. One ofthe major systems-analytic multiobjective approaches to decision-making under constraints is multiobjective optimization as a generalization of traditional single-objective optimization. Although multiobjective optimization problems differ from single­ objective optimization problems only in the plurality of objective functions, it is significant to realize that multiple objectives are often noncom mensurable and conflict with each other in multiobjective optimization problems. With this ob­ servation, in multiobjective optimization, the notion of Pareto optimality or effi­ ciency has been introduced instead of the optimality concept for single-objective optimization. However, decisions with Pareto optimality or efficiency are not uniquely determined; the final decision must be selected from among the set of Pareto optimal or efficient solutions. Therefore, the question is, how does one find the preferred point as a compromise or satisficing solution with rational pro­ cedure? This is the starting point of multiobjective optimization. To be more specific, the aim is to determine how one derives a compromise or satisficing so­ lution of a decision maker (DM), which well represents the subjective judgments, from a Pareto optimal or an efficient solution set.