Tong / Sriram Artificial Intelligence in Engineering Design

Volume III: Knowledge Acquisition, Commercial Systems, And Integrated Environments
1. Auflage 2012
ISBN: 978-0-08-092602-5
Verlag: Elsevier Science & Techn.
Format: EPUB
Kopierschutz: 6 - ePub Watermark

E-Book, Englisch, 388 Seiten

Artificial Intelligence in Engineering Design is a three volume edited collection of key papers from the field of artificial intelligence and design, aimed at providing a description of the field, and focusing on how ideas and methods from artifical intelligence can help engineers in the design of physical artifacts and processes. The book surveys a wide variety of applications in the areas of civil, mechanical, chemical, VLSI, electrical, and computer engineering. The contributors are from leading academic computer-aided design centers as well as from industry.

Autoren/Hrsg.

Tong, Christopher

Sriram, Duvvuru

Leseproben

Chapter 1

INTRODUCTION

Chris Tong and Duvvuru Sriram

1.1. WHAT THIS BOOK IS ABOUT

What is ? Design is the process of constructing a description of an artifact that satisfies a (possibly informal) functional specification, meets certain performance criteria and resource limitations, is realizable in a given target technology, and satisfies criteria such as simplicity, testability, manufacturability, reusability, etc.; the design process itself may also be subject to certain restrictions such as time, human power, cost, etc.

Design problems arise everywhere, and come in many varieties. Some are born spontaneously amidst the circumstances of ordinary human lives: design a dish for dinner that uses last night’s leftovers; design some kind of hook-like artifact that will enable me to retrieve a small object that fell down a crack; design a “nice-looking” arrangement of the flowers in a vase. Other design problems are small but commercial in nature: design a paper clip-like device that doesn’t leave a mark on the paper; design a lamp whose light can be turned to aim in any particular direction; design an artifact for storing up to twenty pens and pencils, in an easily accessible fashion. Still other design problems are formidable, and their solutions can require the efforts and coordination of hundreds of people: design a space shuttle; design a marketable electric car; design an international trade agreement; etc.

Because design is so ubiquitous, anything generic we can say about the – the activities involved in actually solving a design problem – can have great impact. Even better would be to provide active help to designers.

This book is all about how ideas and methods from Artificial Intelligence can help engineering designers. By “engineering design”, we primarily mean the design of or of various kinds. In this book, we will see the design of a wide variety of artifacts exemplified, including: circuits and chips (Volume I, Chapters 2, 8, 12 and Volume II, 2, 8, 9, 10), swinging doors (Volume I, Chapter 6), copying machines (Volume I, Chapter 9 and Volume III, Chapter 6), cantilever beams (Volume I, Chapter 3), space telescopes (Volume II, Chapter 5), air cylinders (Volume I, Chapter 7), protein purifaction processes (Volume I, Chapter 10), fluid-mechanical devices (Volume II, Chapters 4 and 6), new alloys (Volume II, Chapter 7), graphics interfaces (Volume I, Chapter 14), automobile transmissions (Volume I, Chapter 4), spatial layouts (Volume I, Chapter 13), elevators (Volume I, Chapter 11), light-weight load-bearing structures (Volume II, Chapter 11), mechanical linkages (Volume II, Chapter 12), buildings (Volume III, Chapter 12), etc.

What you will not find in this book is anything on AI-assisted software design. On this point, our motivation is twofold: no book can (or should try to) cover everything; and AI and software engineering has already been treated in a number of edited collections (including [15,30]).

This book is an edited collection of key papers from the field of AI and design. We have aimed at providing a state of the art description of the field that has coverage and depth. Thus, this book should be of use to engineering designers, design tool builders and marketeers, and researchers in AI and design. While a small number of other books have surveyed research on AI and design at a particular institution (e.g., [12,31]), this book fills a hole in the existing literature because of its breadth.

The book is divided into three volumes, and a number of parts. This first chapter provides a conceptual framework that integrates a number of themes that run through all of the papers. It appears at the beginning of each of the three volumes. Volume I contains Parts I and II, Volume II contains Parts III, IV, and V, and Volume III contains Parts VI through IX.

Part I discusses issues arising in designs and design information. Parts II and III discuss a variety of models of the design process; Part II discusses models of routine design, while Part III discusses innovative design models. We felt that creative design models, such as they are in 1991, are still at too preliminary a stage to be included here. However, [11] contains an interesting collection of workshop papers on this subject. Parts IV and V talk about the formalization of common sense knowledge (in engineering) that is useful in many design tasks, and the reasoning techniques that accompany this knowledge; Part IV discusses knowledge about physical systems, while Part V gives several examples of formalized geometry knowledge. Part VI discusses techniques for acquiring knowledge to extend or improve a knowledge-based system. Part VII touches on the issue of building a knowledge-based design system; in particular, it presents a number of commercially available tools that may serve as modules within a larger knowledge-based system. Part VIII contains several articles on integrating design with the larger engineering process of which it is a part; in particular, some articles focus on designing for manufacturability. Finally, Part IX contains a report on a workshop in which leaders of the field discussed the state of the art in AI and Design.

1.2. WHAT DOES AI HAVE TO OFFER TO ENGINEERING DESIGN?

In order to answer this question, we will first examine the nature of engineering design a little more formally. Then we will briefly summarize some of the major results in AI by viewing AI as a software engineering methodology. Next we will look at what non-AI computer assistance is currently available, and thus what gaps are left that represent opportunities for AI technology. Finally, we outline how the AI software engineering methodology can be applied to the construction of knowledge-based design tools.

1.2.1. Engineering Design: Product and Process

Engineering design involves mapping a specified onto a (description of a) realizable physical – the designed artifact. The desired function of the artifact is what it is supposed to do. The artifact’s physical structure is the actual physical parts out of which it is made, and the part-whole relationships among them. In order to be realizable, the described physical structure must be capable of being assembled or fabricated. Due to restrictions on the available assembly or fabrication process, the physical structure of the artifact is often required to be expressed in some which delimits the kinds of parts from which it is built. A is one whose physical structure correctly implements the specified function.

Why is design usually not a classification task [6], that is, a matter of simply looking up the right structure for a given function in (say) a parts catalog? The main reason is that the mapping between function and structure is not simple. For one thing, the connection between the function and the structure of an artifact may be an indirect one, that involves determining specified behavior (from the specified function), determining actual behavior (of the physical structure), and ensuring that these match. For another, specified functions are often very complex and must be realized using complex organizations of a large number of physical parts; these organizations often are not hierarchical, for the sake of design quality. Finally, additional non-functional constraints or criteria further complicate matters. We will now elaborate on these complications.

Some kinds of artifacts – for example, napkin holders, coat hangers, and bookcases – are relatively “inactive” in the sense that nothing is “moving” inside them. In contrast, the design of a involves additionally reasoning about the artifact’s both external and internal. The external behavior of a system is what it does from the viewpoint of an outside observer. Thus, an (analog) clock has hands that turn regularly. The internal behavior is based on observing what the of the system do. Thus, in a clock, we may see gears turning. Behavior need not be so visible: electrical flow, heat transmission, or force transfer are also forms of behavior.

In a physical system, behavior function and structure. The is achieved by the in a certain way. If we just possessed the physical structure of a clock, but had no idea of how it (or its parts) behaved, we would have no way of telling that it achieves the function of telling time.

Not only in a physical system but also in a physical system, behavior tends to act as intermediary between function and structure. Associated with a specified function is a ; we would be able to tell time if the angle of some physical structure changed in such a way that it was a function of the time. Associated with a physical structure is its ; for example, a gear will provided that some rotational force is applied to it. In rough terms then, designing a physical system involves...