A knowledge-based expert system as a pre-post processor in engineering optimization

Abstract

Although conventional computer programs use efficient and precise optimization algorithms, they cannot emulate the problem-solving capabilities of human experts. A design optimization process involves a number of tasks which require human expertise and experience. Traditional optimization systems have concentrated on numerical aspects of a design process and have not been successful in integrating the numerical parts with human expertise. Numerical processes such as analysis, simulation and optimization are frequently used in design. In this paper a design system which has the capabilities of knowledge processing and numerical computation by integrating the multiobjective optimization method and the knowledge-based system was developed. To generate the Pareto optimal set efficiently based on the constraint method, a hybrid optimization method is developed by coupling the genetic algorithm and the direct search method. The knowledge-based system for symbolic processing is also developed. Rules for knowledge representation and the inference mechanism of the system are written in LISP. There are 147 rules altogether in the rule bare. The knowledge-based multiobjective optimum design system is finally developed by integrating the multiobjective optimization method and the knowledge-based system. The system is applied to an optimum design of a ship in the preliminary design stage. It is found that the system well simulates design variables and objective functions of the design model. Copyright (C) 1996 Elsevier Science Ltd