基于遗传算法和BP神经网络的花盘结构优化设计  

Optimal design of faceplate structure based on BP neural network and genetic algorithm

作　　者：唐军[1] 黄筱调[1] 方成刚[1] 

机构地区：[1]南京工业大学工业装备数字制造及控制技术重点实验室,南京210009

出　　处：《机械设计与制造》2011年第7期27-29,共3页Machinery Design & Manufacture

基　　金：江苏省科技计划项目(BC2009012);江苏省自然科学基金(BK2008374)

摘　　要：综合利用有限元法、正交试验法、BP神经网络以及遗传算法对大重型数控转台的花盘结构系统进行优化研究。首先对花盘结构系统进行谐响应动力学分析,找出对结构动态特性影响最大的模态频率,并确定BP神经网络的输入变量,然后利用正交试验法和有限元分析法确定出BP神经网络样本点数据,建立反映花盘结构特性的BP神经网络模型,最后利用遗传算法对建立的BP神经网络优化。仿真结果表明,花盘第一阶固有频率提高15.5%,其自重降低9.8%。With the application of finite element method,orthogonal experiment method,BP neural network and genetic algorithm,faceplate structure system of large,heavy duty NC rotary table is studied in optimizing.First of all,a kinetics analysis of the harmonic response is processed in the Faceplate structure system to find out the frequency mode which effects dynamic behavior of the system significantly and confirm the input variables of the BP neural network.Then by applying an orthogonal experiment and finite element analysis,sample point data for BP neural net is set up to establish BP neutral net model which reflects the features of Faceplate structure.At last,the BP neural network model will be optimized through the genetic algorithm.Simulation results show that the first order inherent frequency increases by 15.5 percent with 9.8 percent weight lost.

关 键 词：BP神经网络 遗传算法 有限元法 正交试验 结构优化 

分 类 号：TH122[机械工程—机械设计及理论] TP183[自动化与计算机技术—控制理论与控制工程]