模糊稳健优化设计中的神经网络技术  

Application of Neural Networks in Fuzzy Robust Optimal Design

作　　者：郭惠昕[1] 

机构地区：[1]湖南文理学院机械工程系,湖南常德415003

出　　处：《湖南文理学院学报（自然科学版）》2004年第3期52-54,67,共4页Journal of Hunan University of Arts and Science(Science and Technology)

基　　金：湖南省教育厅科学研究重点项目(03A031)

摘　　要：由于模糊稳健设计中的准则函数和约束函数是随机性可控因素与不可控因素的函数,其概率密度函数的表达式难以确定,故在优化设计中采用随机模拟方法.但采用遗传算法进行优化计算时,随机模拟消耗的机时很大,优化计算效率极低.为此,采用BP神经网络对模糊随机函数进行逼近,探讨了神经网络的构造与程序实现方法;用插装式溢流阀调压弹簧的模糊稳健优化设计问题进行了数值仿真与验证.实例表明,BP神经网络的仿真精度可以达到工程优化问题的要求,把遗传算法与神经网络技术相结合,可以有效地求解模糊稳健设计等复杂优化设计问题.The criterion functions and restriction functions contained random controllable and incontrollable factors in fuzzy robust optmal design. Since it was difficult to determine the expressions of prooability density functions, the Monte Carlo method was often used in optimal calculation. When genetic algorithm was used, the wasted time by Monte Carlo simulation was very long, and the efficiency of optmal calculation was very low. Then it was proposed to use BP neural network for approaching fuzzy random functions, and the me hods for constructing neural networks and programming were discussed. As an example of numerical simulation, the fuzzy robust optimal design of an adjusting spring in an inserted overflow valve was completed, and it showed that the BP neural network was satisfactory in simulation precision. The hybrid optimal method of genetic algoritbm combined with neural networks can effectively solve complex preblems of fuzzy robust optimal design.

关 键 词：约束函数 逼近 随机模拟 求解 概率密度函数 表达式 模糊随机 程序实现 遗传算法 BP神经网络 优化设计 模糊稳健设计 

分 类 号：TH164[机械工程—机械制造及自动化] TP391.7[自动化与计算机技术—计算机应用技术]