Research on Optimization of Multi-objective Function Problems

• Main Authors: LIOU, PIN-JYUN, 劉品均
• Other Authors: TSAI, JINN-TSONG
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/twmjm8

碩士 === 國立屏東大學 === 資訊科學系碩士班 === 106 === 　　Industrial optimization often has multiple-objective requirements. How to experiment under unknown models, reduce costs and obtain available parameters has always been a problem for researchers. In this study, the genetic algorithm and the uniform design algorithm are used to optimize the multi-objective function. Under different solution spatial ranges, the parameters that meet the multiple-objective requirements are searched, and compared to the parameter values and the computational cost. Genetic algorithms use selection, crossover, and mutation to evolve good parameter values. The uniform design algorithm adopts uniform distribution characteristics, and uniformly distributes the experiment in the solution space. By the optimal experimental point and the current solution space range, the new solution space range is determined, and then converge solution space range, so that the parameter values converge toward the optimal solution. 　　Among the eleven multi-objective functions used, the genetic algorithm can stably obtain excellent parameters, but a large number of experiments may lead to excessive cost and system wear problems. In contrast, the uniform design algorithm can select specific retention rate parameters, reduce costs through a small number of experiments, and obtain good parameters.