基于神经网络的三节伸缩臂参数灵敏度分析  

Parameter sensitivity analysis of three sections telescopic boom based on artificial neural network

作　　者：赵得成[1] 许志沛[1] 唐林 贺田龙 敖维川 Zhao Decheng;Xu Zhipei;Tang Lin;He Tianlong;Ao Weichuan(School of Mechanical Engineering,Southwest Jiaotong University,Chengdu 610031,China)

机构地区：[1]西南交通大学机械工程学院,成都610031

出　　处：《现代制造工程》2018年第9期49-54,共6页Modern Manufacturing Engineering

基　　金：中央高校基本科研业务费专项资金资助项目(2015ZT03)

摘　　要：根据设计要求,设计某型湿喷机机械手机构,用于解决在恶劣施工环境下代替人工作业的问题。三节伸缩臂作为该机械手机构的关键部件,设计较为保守,强度、刚度等均有较大余量,为减轻机械手机构的结构自重,使材料的力学性能得到最大化利用,以三节伸缩臂结构为研究对象,进行参数灵敏度分析,为后期的轻量化设计打下基础。以强度、刚度为约束条件,结构自重为设计指标,探究三节伸缩臂的6个设计变量对设计指标的影响。首先运用MATLAB软件构建了6个设计变量与结构的最大应力(Von mises stress)、最大变形、第4阶模态的固有频率和结构自重的BP神经网络模型;然后集成运用软件ISIGHT和MATLAB实现基于描述性抽样法的蒙特卡洛(Monte Carlo)数值模拟,得到了各个设计变量与设计目标的灵敏度关系。参数灵敏度分析结果对同类结构设计具有重要的借鉴意义和参考价值。According to design requirements, in order to solve the problem of operating concrete wet spraying machinery to substitute for manual tools under an abominable tunnel construction environment, certain manipulator machine had been designed. Three sections telescopic boom are key component of the manipulator machine and it had a conservative design with surplus strength and stiffness. Parameter sensitivity analysis are usually regarded as basis for completing lightweight design, including reducing structure weight and manufacture cost,making good use of materials＇ mechanical properties, etc. Considering strength and stiffness as constraint conditions, structure weight as design objective, researched the effects of 6 variables on design indexes. First of all, utilizing software MATLAB established the BP neural network model with 6 design variables,3 constraint conditions and 1 design objective. After that, integrating the multidisciplinary optimization design software ISIGHT and software MATLAB achieved the Monte Carlo numerical simulation for parameter sensitivity analysis. By this analysis the differences parameters sensitivity are clarified, which will provide great reference for the design of similar structures.

关 键 词：三节伸缩臂 神经网络模型 描述性抽样法 灵敏度分析 

分 类 号：TH21[机械工程—机械制造及自动化]