项目名称: 基于集合偏好关系的高效多目标优化理论与算法研究
项目编号: No.11202073
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 数理科学和化学
项目作者: 刘桂萍
作者单位: 湖南大学
项目金额: 25万元
中文摘要: 进化算法因其基于种群搜索的特性及与问题具体特征无关的进化本质而成为多目标优化领域研究的热点,但目前进化算法在多数实际工程多目标问题中并未能实现真正地应用,主要研究难点在于设计者最终满意方案的获取和算法求解效率的提高。本项目拟基于集合偏好关系,将设计者的模糊偏好信息加入优化过程,引导算法向设计者感兴趣的非支配解区域搜索,使其最终能更便捷地选择到满意的优化方案,此外还将结合代理模型技术,针对目标函数值计算耗时的问题,解决算法求解效率低下的难题。主要研究集合偏好关系的表达方法,包括设计者的模糊偏好信息的表达方法,及其与非支配关系、非支配解间距等集合偏好的联合表达方法;研究基于集合的快速优化搜索策略,包括集合个体变异策略和局部搜索策略;研究结合代理模型技术的方法,具体针对高档数控磨床静压电主轴系统的多目标优化问题。通过本项目研究,将有望真正实现多目标进化算法在实际工程优化问题中的应用。
中文关键词: 多目标优化设计;集合偏好关系;进化算法;代理模型;电主轴
英文摘要: Research on evolutionary algorithms has become a hot one in the multi-objective optimization field for their natural ability of finding multiple optimal solutions in one single simulation run and evolutional nature of solving problems without knowig their features. However, in many cases, these algorithms are not acturally used to solve engineering multi-objective optimization problems. There are two reasons, one is that designers have difficultis to select a satisfied design, and the other is the low efficiency of the algorithms. This project will add designers fuzzy preference informations in optimization process based on set preference relations, and guard the search to the area where designer is interest.It is a convenient way for the designers choosing a satisfied design.For the problems with expensive objective functions, the surrogate models will be used to improve the efficiency. There are three research contents in this project. First, the express methods of set preference relations are studied, including the express method of designer's fuzzy preference information and a combined express method of designer's preference, nondominated relation and distance between nondominated solutions. Second, several fast optimization strategies based on set are studied, including mutation strategies and local search
英文关键词: multi-objective optimization design;set preference relations;evolutionary algorithm;surrogate model;electrical spindle