Multi-objective Optimization Based On Parallel Mul-families Genetic Algorithm And Visualization

Optimal problem has always been a hotspot and focus in the study of science and engineering. For years, researches in some of the traditional optimization problems such as single objective optimization problem are already very mature. But optimization problems of industry process often involved multiple targets, so multi-objective optimization problems are more practical than single objective optimization problem. In recent years, with the development of computer technology, various kinds of advanced modern optimization algorithms appeared, solution of the traditional multi-objective optimization problems find new ways with modern optimization algorithms, and achieved well. But it is always unsatisfied when these algorithms were used to solve practical problems in industry, for it is hard to find a strict and effective math mode to describe the actual industrial process, and the efficiency is not quite well.Based on the problems above, the multi-objective optimization algorithm, the parallel computing technology, visualization technology, multi-objective optimization design and multi-objective parameter optimization problem are studied in this paper, and proposed the parallel mul-families genetic algorithm which combined with process simulator to slove optimization design and parameter optimization problem, and use visualization technology to analysis the optimization results. This paper includes the following contents:(1) Studied and summarized the domestic and foreign researches in the multi-objective optimization problems. Compared a variety of multi-objective genetic algorithm, and selected the NSGA-â…¡ as the basic algorithm used in this paper.Summarized the domestic and overseas studies on solving multi-objective problem of process industry using multi-objective genetic algorithm including the NSGA-â…¡ algorithm, which presented a time-consuming problem.(2) Studied the parallel computing technology, and proposed a parallel mul-family genetic algorithms, which makes the multi-objective genetic algorithm can be easily operated on multiple computers, and discussed the algorithm parameters-setting problem, at last it improved the computational efficiency, and got better results. Combined PMOGA and process simulator, and studied how to deal with the constraint conditions.(3) Proposed the nonlinear dimensionality reduction mapping and inverse mapping method based on the principle of "adapt locally-verifica globally", which is used to extract optimization space of Pareto solution.(4) Solve the multi-objective optimization design and multi-objective parameter optimization problems in chemical process using method proposed in this article, good results were obtained. It shows that the presented calculation method is effective.(5) At last I have summarized study comprehensively, and put forward the prospects for the next step.