Engineering Tools for Building Knowledge-Based Systems on Microsystems M. zyxw QPON R. WlGAN zyxw MLKJ Australian Road Research Board nowledge-based systems tools are increas- zyxwv LKJIH K ingly available on microcomputer systems. The range and characteristics of some of these tools are briefly reviewed including zyxwvut HGFED moioc and L ~ S P sys- tems and Expert System Shells for both text- and rule-based approaches. While the tools available up to about 1984 were reasonably adequate for self and academic teaching purposes and for prototyp- ing systems, production quality and capability tools have subsequently become widely available. Dif- ferent paradigms for problem reduction and analy- sis are covered, followed by the matching of prob- lem type, approach, and tool to both the learning phase and the production process. There is a bias towards the construction, design, and transport en- gineering areas of civil engineering. zyxwvu LKJIHG INTRODUCTION There is a wide variety of interpretations of the term “expert system.” These differing ideas reflect a lack of systematic discussion of the problems towards which expert systems are directed, and the means by which they are implemented. The basic terms with which this report is involved are the following: 1. explanation, 2. deduction, zyxwvutsr JIHGFED 3. advice, and 4. inference. Most useful expert systems offer the user explanations if they ask .to have the reasons for a particular question Address correspondence to: M.R. Wigan, Australian Road Research Board, zyxwvuts 500 Burwood Highway, Vermont, Victoria. 3 1 33. Australia. being posed, and will explain the chain of deduction that led to this point. Treating problems in this way leads to a series of spe- cial requirements. Ideally the complexities of the infer- ence system, the painful coding of the user interface and the articulation of the explanation mechanisms would all be provided to a practical analyst wishing to start the task of building an expert system for a specific problem are with which he was concerned. This series of tasks are the parallels to compilers and problem generators (“4th generation languages”) for conventional computer systems, and the days are long gone when an applications analyst had to consider writ- ing a compiler before he started his application coding. The family of tools now available to cocoon expert sys- tem builders from the internals and mechanisms of an expert system are usually termed ”shells,” by analogy to the UNIX shell which does a similar task for users of that operating system. Like UNIX, many of the available shells are still objects of considerable complexity, flex- ibility, and programming power, and-again like UNIX- are not yet always easily recognized as a “simple”in- terface. This article is based on the supporting documentation for an engineering expert systems seminar (751, and is concerned primarily with disentangling the languages and expert system shells now available on small systems (280 with CP/M 80 up to-typically-680 x 0 and PDP- 11 systems running UNIX). This exploration will take us through the conceptual chain of developments which led to the appearance of these tools to the specific tools which are now available on small computers to exercise expert system building and operation. Initial engineering applications of expert systems are proving most effective in areas where diagnosis, clas- sification, fault detection, control, or explanation are needed-especially in areas like road and other forms 52 Microcomputers in Civil Engineering 1. 52-68 119861 0 1986 Elsevier Science Publishing Co.. Inc. S.? Vanderbilt Avenue 0885-9607;86/$3.50