knowledge-based expert systems; PROLOG; knowledge-base is purely heuristic
The emergence of knowledge-based expert systems provide means with which one can use the computer as an aid to the solution of an ill-structured problem. Expert systems are interactive computer programs based on heuristics, incorporating judgement, rules of thumb, intuition and expertise to provide knowledgeable advice about variety of tasks. Such specialized interactive computer programs can broadly be classified as (1) identifying the relevant design knowledge, (2) providing a formalism for representing and processing the knowledge and (3) implementing the formalism in a computer environment. While the first issue of identifying the relevant knowledge is through knowledge acquisition from various domain experts and verification of the same by other domain experts. The second issue is proper formalism for the varied knowledge requiring the use of logics (prepositional calculus) as a symbolic language; heuristics or rules of thumb wherever necessary and a suitable reasoning methodology, i.e. an inference technique. The third issue is the evaluation of a suitable strategy for successful computer implementation, i.e. the computer problem is subdivided into smaller tasks which have easy solutions; and the capability of interacting the solutions of the smaller tasks into a larger framework. Therefore, this will require theorem proving, search technique and a special purpose computer language such as PROLOG or readily available domain independent shells, i.e. expert system tools. The overall synthesis of all the above is termed as a knowledge-based expert system (KBES). In this paper a KBES is developed for a highly complex building element, ‘beam design’, as part of a larger model involving planning, analysis, design, optimization and cost forecasts along with other allied services such as plumbing and electrical services. In the development of KBES; the knowledge-base is purely heuristic and subjected to alteration by numerical calculations. For an effective search the modified depth technique is adopted. Heuristics are employed to further the search as and when required from the point of view of practical computer implementation. In the overall development, use is made of dependency diagrams, charts, tables and search trees. For computer implementation the necessary tool chosen was the M1 shell developed by Teknowledge Inc., U.S.A. This requires a PC-AT with 640 kB RAM and 40 MB hard disk. A brief overview of the expert system, followed by an example problem, is presented.