Knowledge-B ased Process Planning for Construction and Manufacturing Carlos Zozaya-Gorostiza Chris Hendrickson Daniel R. Rehak Department of Civil Engineering and Engineering Design Research Center Carnegie Mellon University Pittsburgh, Pennsylvania ACADEMIC PRESS, INC. Harcourt Brace Jovanovich, Publishers Boston San Diego New York Berkeley London Sydney Tokyo Toronto
Copyright 1989 by Academic Press, Inc. All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. ACADEMIC PRESS, INC. 1250 Sixth Avenue, San Diego, CA 92101 United Kingdom Edition published by ACADEMIC PRESS INC. (LONDON) LTD. 24-28 Oval Road, London NW1 7DX IBM, PC/AT, and RT PC are registered trademarks of International Business Machines. KNOWLEDGE CRAFT is a trademark of Carnegie Group, Inc. LOTUS and 1-2-3 are registered trademarks of Lotus Development Corporation. MacDraw is a trademark of Claris, Inc. SCRIBE is a registered trademark of Scribe Systems. Sun Workstation is a registered trademark of Sun Microsystems, Inc. EXPLORER is a trademark of Texas Instruments. Library of Congress Cataloging-in-Publication Data Zozaya-Gorostiza, Carlos, Date- Knowledge-based planning for construction and manufacturing/ Carlos Zozaya-Gorostiza, Chris Hendrickson, Daniel R. Rehak. p. cm. Bibliography: p. Includes index. ISBN 0-12-781900-2 1. Industrial project management Data processing. 2. Production planning Data processing. 3. PLANEX (Computer program) I. Hendrickson, Chris. II. Rehak, Daniel R. III. Title. TA190.Z69 1989 89-6789 658.5'03O285-dc20 CIP Printed in the United States of America 89 90 91 92 9 8 7 6 5 4 3 2 1
Preface Construction and manufacturing process planning is a crucial and challenging management task. A good plan is essential to project success. Plans are the basis for a project budget and production schedule. Despite its importance and complexity, process planning relies on manual formulation of plans and is usually performed in an intuitive and unstructured fashion with considerable reliance on engineering judgment. There are few computer-based process planning aids which address the needs and complexities of the construction and manufacturing processes. Those aids which do exist are primarily algorithmic analysis or graphical display tools. Knowledge-based program development methodologies provide a new technological basis for the development of process planning tools. This approach provides the means to represent and utilize process planning knowledge and judgment which is lacking in current tools. This work describes a knowledgebased system architecture used to develop process planning systems PLANEX. PLANEX is a domain-independent, knowledge-based process planning system architecture. Starting from a description of the physical artifact to be constructed or manufactured, PLANEX generates the set of activities used to create the artifact. These activities, with their required resources, are linked into a process planning network which can be used in project scheduling or management. This work presents the concepts, requirements and the resulting architecture of PLANEX, and includes detailed descriptions of applications of the system in construction and manufacturing. This work originated as an investigation of the application of knowledgebased systems technology in construction management. When the work was conceived in 1984, there were no significant prototype knowledge-based systems in the construction domain; only a few small-scale exploratory applications had been developed. The second and third authors had done extensive work in the application of expert systems in other areas of Civil Engineering, and it was vii
viii Preface evident that the technology could be successfully applied to problems which were intractable using conventional programming methodologies. Creating a project schedule is knowledge-intensive and a prerequisite to the application of other computer tools in project management. Due to the importance of this task, the lack of suitable aids and the promise offered by a knowledge-based systems approach, construction project planning was selected as an appropriate domain to explore in detail. The initial goals were two-fold: (1) to demonstrate the applicability of the knowledge-based approach to the problem; and (2) to investigate the problems with the technology which need to be addressed for it to be successfully applied in this domain. Although the general domain-independent nature of the process planning problem was considered from the outset, the initial goal was to develop a system limited to construction project planning. In the initial stages, a variety of domains were considered: buildings, bridges and highways. Building construction was selected as a focus and the work commenced. Initial planning explorations were made in the perspicuous blocks-world domain, and as the work evolved, it moved from a system targeted at construction to the more general process planning architecture presented herein. While this work originated in the construction management domain, the ideas and concepts presented have wider application. As presented, the emphasis is on process planning as opposed to project planning. Many of the examples and illustrations are derived from construction planning. This should not be interpreted as a bias in the concepts, but rather as a result of the authors' experience and knowledge. This work is directed at two audiences: (1) builders and developers of practical process planning aids; and (2) researchers of planning and scheduling systems. It is particularly relevant to those whose research is in the areas of artificial intelligence planning systems; operations research and management science optimization methods for planning; management information and decision support systems; and knowledge-based planning systems. The development of process planning tools could substantially benefit from generalization and from cross-fertilization among the different planning and decision support approaches. Researchers familiar with only one approach should broaden their horizons and thereby improve their own contributions. Therefore, a major objective of this work was to foster just this sort of crossfertilization. Because of the diverse backgrounds and interests of different readers, it is not expected that the individual reader will devote the same attention to each section of this book. For example, operations researchers might omit Section 2.2 reviewing plan scheduling methods. As another example, researchers seeking only a conceptual understanding of knowledge-based process planning might omit the implementation details of the construction and manufacturing applications in Sections 6.2 and 7.2.
Preface ix Organization This monograph begins with an overview of the process planning problem and the motivation for this work, presented in Chapter 1. This introduction includes a discussion of the knowledge-based approach, a brief overview of PLANEX and a review of its development. Chapter 2 is a review of background material relevant to this work. It includes a review of ΑΙ-based models for process planning and plan formulation; models for deterministic project scheduling, including critical path, resource allocation and resource leveling algorithms; and prior work in the practice of construction and manufacturing planning, scheduling and monitoring. This background material is used to develop a formal model for process planning. The characteristics of this model are described in terms of a conceptual process planning model, and the resulting requirements for a knowledge-based process planning model are presented in Chapter 3. Based on these requirements, the architecture of PLANEX is described in Chapter 4. This description includes the overall structure, knowledge representation, control and user interface components of the architecture. Use of this system architecture to develop process planning models is presented in Chapter 5. In addition to the basic features of the PLANEX architecture and the process of developing a system with PLANEX, the chapter includes an overview of four different applications (construction planning, excavation planning, manufacturing planning, and blocks-world planning) and an evaluation of the architecture. Chapter 6 provides a detailed description of CONSTRUCTION PLANEX, a system used to plan the construction of mid-rise concrete and steel-frame office buildings. This chapter includes a discussion of the construction planning models used in CONSTRUCTION PLANEX, along with details of the representation, knowledge, problem-solving and user interface components of the system. It concludes with the presentation of an example problem. HARNESS PLANEX, a system which plans the manufacturing operations for automobile electrical wire harnesses, is presented in Chapter 7. The organization and structure of this chapter parallels that of Chapter 6. These two final chapters are intended to provide sufficient detail for implementation of knowledge-based process planning systems so that builders of new tools can use this information as a starting point. More casual readers can skip some sections of these chapters. PLANEX Software PLANEX has gone through several cycles of development and refinement. The initial implementation was in LISP. The CONSTRUCTION PLANEX and HARNESS PLANEX prototypes described in Chapter 6 and 7 were implemented in COMMON LISP and KNOWLEDGE CRAFT on a TEXAS INSTRUMENTS EXPLORER. Associated subsystems, such as the ANIMATOR, were developed in C on a SILICON GRAPHICS IRIS Workstation. The examples presented herein were developed with this version of PLANEX.
χ Preface Recently a version of PLANEX was implemented entirely in COMMON LISP, and has been successfully ported to a number of hardware platforms including an IBM RT PC, an IBM PC/AT and a Sun Workstation. This COMMON LISP version does not include graphical schedule displays or the animation subsystem, but generates activity plans for input to commercial scheduling packages. The COMMON LISP version of PLANEX is available to researchers, educators and institutions who wish to experiment, to refine or to extend the system. Anyone who wishes to obtain PLANEX should contact the second author (CH) at Carnegie Mellon. Acknowledgments This monograph is based on the Ph.D. dissertation of the first author [115]. The work was supervised by the second author, and was conducted, in part, under a research grant awarded to the second and third authors. In preparing this book, the authors have reorganized, refined and extended the material from the dissertation. Development of PLANEX was supported in part by the National Science Foundation, Grant MSM-8503400; by the Engineering Design Research Center of Carnegie Mellon University, an NSF Engineering Research Center; by the U.S. Army Construction Engineering Research Laboratory; and by the Department of Civil Engineering of Carnegie Mellon University. Additional support to prepare this monograph was provided by the Engineering Design Research Center. Computer facilities used to develop PLANEX and to prepare this book were provided by the Engineering Design Research Center, the Department of Civil Engineering and the Robotics Institute of Carnegie Mellon. The authors gratefully acknowledge the funding and financial support which made this work possible. A number of individuals have made contributions to the development of PLANEX and the applications described herein. Much of the initial construction planning knowledge came from Eduardo Baracco-Miller. Throughout much of the project, Peter S. Lim worked as an undergraduate programmer and was responsible for developing many of the tools and utilities incorporated in the PLANEX architecture. The expertise and skill of these individuals was invaluable in the development effort. The authors would like to thank two of our colleagues, Prof. Clive L. Dym of the University of Massachusetts and Prof. Raymond E. Levitt of Stanford University for their critical comments in reviewing our manuscript and their continued support and interest throughout the development of PLANEX. We would also like to thank numerous industrial colleagues for the comments reviewing the various prototype systems.
Preface xi This monograph was produced electronically. Pages were laserset using the SCRIBE document production system and figures were produced with MacDraw. Copyediting was done by Heather Walls. The production editor was Joni Hopkins of Academic Press. Sponsoring editor was Sari Kalin of Academic Press. Carlos Zozaya-Gorostiza Chris Hendrickson Daniel R. Re hak
Introduction This study describes the development of knowledge-based computer tools that assist an engineer in process planning (i.e., formulating a process plan) for construction and manufacturing. The result of the investigation is a comprehensive knowledge-based system architecture for process planning, called PLANEX, which has been used to develop a number of prototype process planning systems in construction and manufacturing domains. The issues and problems in process planning can be illustrated through a simple, everyday example: preparing a meal. Meal preparation starts with a menu of the dishes which comprise the meal. The menu corresponds to the specification of the product to be prepared. Recipes provide the descriptions of how to create each dish. Depending on the source, the recipes may provide a detailed description of the tasks or activities used to create the dish, or they may be an abstract description of the preparation process. No matter how much detail is provided in a recipe, all of the steps for all of the activities are not specified (e.g., recipes often call for clarified butter but do not explain how to prepare it). The cook must determine, based on experience, all of the steps needed to prepare the dish (a process denoted activity formulation). Individual steps from various recipes can often be combined into larger-scale project activities (e.g., chopping all vegetables for a dish at one time in a food processor rather than individually). Such aggregation reduces the size of the planning problem. Associated with each recipe is the normal yield and the list of ingredients. These material resources have to be scaled to provide the desired yield. This is a complex process as some ingredients (e.g., spices, eggs) do not scale linearly, the quantity needed depends upon the form of the ingredient (e.g., fresh versus dried herbs) and complex substitutions for ingredients may be made. 1