Multiple Solutions for Plant Design Analyses through a Genetic Algorithm with Tabu Lists

In this paper, we explore the problem to efficiently design a series of similar plants. The target plants consist of complex mechano-electronical systems. It will take much time to design them from the initial phases. The problem requires solving complex nonlinear differential equations with multiple objectives. We apply a genetic algorithm with tabu lists, which is able to solve multi-modal and/or multi-objective problems. The paper presents the techniques we have applied to, especially focuses on the landscape search among feasible good solutions. The results are summarized as follows: (1) The GA method equipped with tabu search, minimal gap generation (MGG), and ordinary two-point crossover work well to obtain multiple solutions of the task, and (2) Classification of candidate solutions and neighborhood parameter search enable us to estimate the quality of solutions. The intensive experiments have suggested that the proposed method is effective for the tasks.