An expert facility layout system: an object–oriented approach

Abstract

A good layout design is important for a cellular manufacturing system (CMS) to operate efficiently. Traditional algorithms for layout design are confining and have limitations with regard to decision factors and knowledge representation. Knowledge–based expert systems have opened up a new way for modelling and evaluation of the ill–structured CMS layout problem. In addition to the advantages of integrating the vast amount of available knowledge and making it accessible to ordinary users, the knowledge–based approach is also useful for dealing with ambiguous, qualitative information and empirical knowledge that are often part of the CMS layout design process. The objective of this research is to develop a prototype for CMS Layout Design Expert System (CLADES) by using the expert system shell KEE (Knowledge Engineering Environment). This expert system would provide a powerful and eficient decision support tool for users in CMS layout design. CLADES solves the CMS layout problem in two stages. In the first stage, the machine cell layout is determined. In the second stage, the machine layout within each cell is determined. To obtain a reasonable layout design for a CMS, several factors which affect the decision are used as criteria in the CMS layout design problem. They are production quantity of the system, variation of part process routing, material handling system, unit load of an item, and dimensions of the floor plan. The design options considered by CLADES include three types of layout for machine cells, four types of layout for machines within a cell, and five types of material handling systems. In its current form, CLADES contains 58 rules, three models and three algorithms, and one algorithm is directly coded in CLADES in the LISP language. CLADES uses a tandem architecture, which connects the knowledge base to the model and algorithm base. This structure takes advantage of not only the expert system approach, but also the optimization approach. CLADES is coded using KEE/LISP in an object–oriented programming (OOP) fashion. It also has a user–friendly interface which makes the system easy to use. The object–oriented programming approach allows new rules, models and algorithms to be easily added to CLADES without changing the developed system structure. This built–in flexibility and convenience is very important for the modification and expansion of the system. This research has proved the feasibility of applying the knowledge–based approach to solve the CMS layout design problem. And the framework of CLADES shows it is a promising expert system to be developed into an industrial application.