Intelligent Systems in Process Engineering, Part II: Paradigms from Process Operations (eBook)

Chonghun Han, George Stephanopoulos (Herausgeber)

eBook Download: EPUB

1995 | 1. Auflage
314 Seiten
Elsevier Science (Verlag)
978-0-08-056569-9 (ISBN)

Infos

Volumes 21 and 22 of Advances in Chemical Engineering contain ten prototypical paradigms which integrate ideas and methodologies from artificial intelligence with those from operations research, estimation andcontrol theory, and statistics. Each paradigm has been constructed around an engineering problem, e.g. product design, process design, process operations monitoring, planning, scheduling, or control. Along with the engineering problem, each paradigm advances a specific methodological theme from AI, such as: modeling languages, automation in design, symbolic and quantitative reasoning, inductive and deductive reasoning, searching spaces of discrete solutions, non-monotonic reasoning, analogical learning,empirical learning through neural networks, reasoning in time, and logic in numerical computing. Together the ten paradigms of the two volumes indicate how computers can expand the scope, type, and amount of knowledge that can be articulated and used in solving a broad range of engineering problems.

Key Features
* Sets the foundations for the development of computer-aided tools for solving a number of distinct engineering problems
* Exposes the reader to a variety of AI techniques in automatic modeling, searching, reasoning, and learning
* The product of ten-years experience in integrating AI into process engineering
* Offers expanded and realistic formulations of real-world problems

---

Volumes 21 and 22 of Advances in Chemical Engineering contain ten prototypical paradigms which integrate ideas and methodologies from artificial intelligence with those from operations research, estimation andcontrol theory, and statistics. Each paradigm has been constructed around an engineering problem, e.g. product design, process design, process operations monitoring, planning, scheduling, or control. Along with the engineering problem, each paradigm advances a specific methodological theme from AI, such as: modeling languages; automation in design; symbolic and quantitative reasoning; inductive and deductive reasoning; searching spaces of discrete solutions; non-monotonic reasoning; analogical learning;empirical learning through neural networks; reasoning in time; and logic in numerical computing. Together the ten paradigms of the two volumes indicate how computers can expand the scope, type, and amount of knowledge that can be articulated and used in solving a broad range of engineering problems. - Sets the foundations for the development of computer-aided tools for solving a number of distinct engineering problems- Exposes the reader to a variety of AI techniques in automatic modeling, searching, reasoning, and learning- The product of ten-years experience in integrating AI into process engineering- Offers expanded and realistic formulations of real-world problems

Inhalt

Front Cover 1
Advcances in Chemical Engineering, Volume 22 4
Copyright Page 5
Contents 8
Contributors to Volume 22 18
Prologue 20
Chapter 1. Nonmonotonic Reasoning: The Synthesis of Operating Procedures in Chemical Plants 34
I. Introduction 35
II. HierarchicalModeling of Processes and Operations 45
III. Nonmonotonic Planning 55
IV. Illustrations of Modeling and Nonmonotonic Operations Planning 72
V. Revamping Process Designs to Ensure Feasibility of Operating Procedures 89
VI. Summary and Conclusions 95
References 96
Chapter 2. Inductive and Analogic Learning: Data-Driven Improvement of Process Operations 98
I. Introduction 99
II. General Problem Statement and Scope of the Learning Task 102
III. A Generic Framework to Describe Learning Procedures 105
IV. Learning with Categorical Performance Metrics 110
V. Continuous Performance Metrics 117
VI. Systems with Multiple Operational Objectives 129
VII. Complex Systems with Internal Structure 138
VIII. Summary and Conclusions 152
References 153
Chapter 3. Empirical Learning through Neural Networks: The Wave-Net Solution 158
I. Introduction 159
II. Formulation of the Functional Estimation Problem 162
III. Solution to the Functional Estimation Problem 172
IV. Applications of the Learning Algorithm 192
V. Conclusions 200
VI. Appendices 201
References 204
Chapter 4. Reasoning in Time: Modeling, Analysis, and Pattern Recognition of Temporal Process Trends 206
I. Introduction 208
II. Formal Representation of Process Trends 216
III. Wavelet Decomposition: Extraction of Trends at Multiple Scales 228
IV. Compression of Process Data through Feature Extraction and Functional Approximation 248
V. Recognition of Temporal Patterns for Diagnosis and Control 256
VI. Summary and Conclusions 266
References 267
Chapter 5. Intelligence in Numerical Computing: Improving Batch Scheduling Algorithms through Explanation-Based Learning 270
I. Introduction 271
II. Formal Description of Branch-and-Bound Framework 276
III. The Use of Problem-Solving Experience in Synthesizing New Control Knowledge 291
IV. Representation 302
V. Learning 314
VI. Conclusions 328
References 329
Index 332
Contents of Volume in This Serial 342

Textauszug

Prologue

George Stephanopoulos; Chonghun Han

The adjective “intelligent” in the term “intelligent systems” is a misnomer. No one has ever claimed that an intelligent system in an engineering application possesses the kind of intelligence that allows it to induce new knowledge, (1) or “to contemplate its creator, or how it evolved to be the system that it is”. (2) Åström and McAvoy (3) have suggested terms such as “knowledgeable” and “informed” to accentuate the fact that these software systems depend on large amounts of (possibly) fragmented and unstructured knowledge. For the purposes of this book, the term “intelligent system” always implies a computer program, and although the quotation marks around the adjective intelligent may be dropped occasionally, no one should perceive it as a computer program with attributes of human-like intelligence. Instead, the reader should interpret the adjective as characterizing a software artifact that possesses a computational procedure, an algorithm, which attempts to “model and emulate,” and thus automate an engineering task that used to be carried out informally by a human. Whether or not this models the actual cognitive process in a human is beyond the scope of this book.

In the wide spectrum of engineering activities, collectively known as process engineering and encompassing tasks from product and process development through process design and optimization to process operations and control, so-called intelligent systems have played an important role. Ten years ago the broad introduction of knowledge-based expert systems created a pop culture that started affecting many facets of process engineering work. Expert systems were followed by their cousins, fuzzy systems, and the explosion in the use of neural networks. During the same period, the object-oriented programming (OOP) paradigm, one of the most successful “products” of artificial intelligence, has led to a revolutionary rethinking of programming practices, so that today OOP is the paradigm of choice in software engineering. After 10 years of work, 15 books/monographs/edited volumes, over 700 identified papers in archival research and professional journals, 65 reviews/tutorial/industrial survey papers, about 150 Ph.D. theses, and several thousand industrial applications worldwide, (4) the area of what is known as “intelligent systems” has turned from fringe to mainstream in a large number of process engineering activities. These include monitoring and analysis of process operations, fault diagnosis, supervisory control, feedback control, scheduling and planning of process operations, simulation, and process and product design. The early emphasis on tools and methodologies, originated by research in artificial intelligence, has given place to more integrative approaches, which focus more on the engineering problem and its characteristics. So, today, one does not encounter as frequently as 10 years ago conference sessions with titles including terms such as “expert systems,” “knowledge-based systems,” or “artificial intelligence.” Instead one sees many more mature contributions, from both the academic and industrial worlds, in mainstream engineering sessions, with significant components of what one would have earlier termed “intelligent systems.” The evolving complementarity in the use of approaches from artificial intelligence, systems and control theory, mathematical programming, and statistics is a strong indication of the maturity that the area of intelligent systems is reaching.

A THE CURRENT SETTING

The explosive growth of academic research and industrial practice in the synthesis, analysis, development, and deployment of intelligent systems is a natural phase in the saga of the Second Industrial Revolution. If the First Industrial Revolution in 18th century England ushered the world into an era characterized by machines that extended, multiplied, and leveraged human physical capabilities, the Second, currently in progress, is based on machines that extend, multiply, and leverage human mental abilities. (5) The thinking man, Homo sapiens, has returned to its Platonic roots where “all virtue is one thing, knowledge.” Using the power and versatility of moderm computer science and technology, software systems are continuously developed to preserve knowledge for it is perishable, clone it for it is scarce, make it precise for it is often vague, centralize it for it is dispersed, and make it portable for it is difficult to distribute. The implications are staggering and have already manifested themselves, reaching the most remote corners of the earth and the inner sancta of our private lives. In this expanding pervasiveness of computers, intelligent systems can affect and are affecting the way we educate, entertain, and govern ourselves, communicate with each other, overcome physical and mental disabilities, and produce material wealth. Computer-based deployment of “knowledge” has been thrust by modern sociologists into the center of our culture as the force most effective in resolving inequities in the distribution of biological, historical and material inheritance. But what is the tangible evidence? Software systems have been composed to do the following: (5–7) (i) harmonize chorales in the style of Johann Sebastian Bach and automate musical compositions into new territories; (ii) write original stanzas and poems with thematic uniformity, which could pass as human creations for about half of the polled readers; (iii) compose original drawings and “photographs” of nonexistent worlds; (iv) “author” complete books.

Equally impressive are the results in engineering and science. Characterized as “knowledgeable,” “informed,” “expert,” “intelligent,” or any other denotation, software systems have expanded tremendously the scope of automation in scientific and engineering activities. (4–13)

B THE THEORETICAL SCOPE AND LIMITATIONS OF INTELLIGENT SYSTEMS

So what? a skeptic may ask. Are the above examples manifestations of the computer’s long-awaited, human-like intelligence? No one familiar with Gödel’s theorem of incompleteness would ask such a question, (14,15) for this theorem states that it is not possible to create a formal system that is both consistent and complete. As such, you cannot create a software system based on some sort of a formal system, i.e., a consistent set of axioms, which can reflect upon itself and discover (not invent) a new dimension of knowledge. (1)

Indeed, whenever you focus your attention on any of the so-called intelligent systems, and you take the time to learn the mechanisms they use to generate their marvelous and wondrous behavior, you come up with the anti-climactic realization that everything is quite ordinary and perfectly expectable with no surprises or mystical insights. Such reaction reminds us of how Sherlock Holmes reacted when a man questioned the brilliance of his deductive reasoning in solving one of his cases:

Mr. Jabez Wilson laughed heavily. “Well, I never!” said he. “I thought at first that you had done something clever, but I see that there was nothing in it, after all.” “Begin to think, Watson,” said Holmes, “that I made a mistake in explaining. ‘Omneignotum promagnifico,’ you know, and my poor little reputation, such as it is, will suffer shipwreck if I am so candid.”

Similarly, Alan Turing, the father of the digital computer and creator of the Turing Test for checking the “intelligence” of a machine, put it this way:

The extent to which we regard something as behaving in an intelligent manner is determined as much by our own state of mind and training as by the properties of the object under consideration. If we are able to explain and predict its behavior or if there seems to be little underlying plan, we have little temptation to imagine intelligence. With the same object, therefore, it is possible that one man would consider it as intelligent and another would not; the second man would have found out the rules of its behavior.

C THE CHARACTER OF THE TEN PARADIGMS

All the paradigms of intelligent systems in this volume have plans and assume extensive amounts of knowledge. As such they are ordinary computer programs and they emulate a precise computational procedure, which uses a predefined set of data. In the Aristotelian form, “all instruction given or received (by the intelligent systems) by way of argument, proceeds from preexistent knowledge.” Consequently, one should see all cases put forward by the individual chapters as nothing more than paradigms for new uses of the computer. Every one of them carries out deduction from a predefined set of knowledge, using explicit reasoning strategies. The reader should not search for inductive generation of new knowledge, even when the terms “induction” and “inductive reasoning” have been loosely employed in some chapters. Instead, the reader should see each chapter as a computer-based paradigm in capturing, articulating, and utilizing various forms of knowledge. As a result, the reader will notice that the ten chapters of this volume serve...

Produkt-Details

• Erscheint lt. Verlag: 14.11.1995
• Mitarbeit: Herausgeber (Serie): John L. Anderson
• Sprache: englisch
• Themenwelt: Sachbuch/Ratgeber
• Themenwelt: Informatik ► Theorie / Studium ► Künstliche Intelligenz / Robotik
• Themenwelt: Mathematik / Informatik ► Mathematik ► Logik / Mengenlehre
• Themenwelt: Naturwissenschaften ► Chemie ► Physikalische Chemie
• Themenwelt: Naturwissenschaften ► Chemie ► Technische Chemie
• Themenwelt: Technik ► Bauwesen
• Themenwelt: Technik ► Umwelttechnik / Biotechnologie
• Themenwelt: Wirtschaft ► Betriebswirtschaft / Management ► Logistik / Produktion
• ISBN-10: 0-08-056569-7 / 0080565697
• ISBN-13: 978-0-08-056569-9 / 9780080565699