Case-based evolutionary design approach for satellite module layout

Introduction Various attempts have been made to apply evolutionary algorithms (EAs) to complex engineering layout designs. EAs are powerful optimisation techniques inspired by genetics and natural selection while lacking some important characteristics (prior knowledge) that make other classical and numerical search algorithms very effective. Case-based reasoning (CBR) technique is to solve the problem by retrieving previously solved cases, which are similar to the current problem. CBR has been used in various optimization problems, such as, layout design problems and nesting problems. In CBR process, retrieved cases should be adapted and then reused as solutions for new problems. Therefore, key to achieving the best performance from system is to combine retrieved cases and algorithms with the adaptation process. Pearce analyzed effect of injecting cases to the population of EAs. Ramscy proposed a method of using cases as initial population of EAs. Rosenman used cases not only in the initial population, but also in the process of EAs, and applied the method to building design problems. Sun proposed a modified genetic algorithm to generate an initial layout scheme for layout design. A humanalgorithm-knowledge based layout design method (HAKD) has been proposed to afford prior knowledge solutions for EA. This study aims to improve convergence time of EAs and discusses how CBR techniques can be used to extract reusable cases, which will be injected into evolutionary process of design circuits. Taking layout design of a satellite module (Fig. 1), this paper presents a similarity algorithm for layout case retrieval. A prototype of a 2-D layout design of satellite module is used to test the efficacy and efficiency of this approach.