基于改进PSO-SVR的连杆机构可靠度敏感性研究  

Research on Reliability Sensitivity of Linkage Mechanism Based on Improved PSO-SVR

作　　者：夏尔冬[1] 王春荣[1,2] 熊昌炯 刘建军[1] XIA Er-dong;WANG Chun-rong;XIONG Chang-jiong;LIU Jian-jun(School of Mechanical&Electronic Eng,Sanming Univ.,Fujian Sanming365004,China;College of Mechanical Engin-eering and Applied Electronics Technology,Beijing University of Technology,Beijing 100124,China)

机构地区：[1]三明学院机电工程学院,福建三明365004 [2]北京工业大学机械工程及应用电子技术学院,北京100124

出　　处：《机械设计与制造》2020年第2期236-239,243,共5页Machinery Design & Manufacture

基　　金：福建省自然科学基金项目（2016J01741,2018J01513）;福建省教育厅科技项目（JK2015046,JT180499）;三明市科技项目（2014-G-6）;福建省引导性项目（2016N0029）;三明学院教育教学改革项目（J1710420)

摘　　要：为提高连杆机构的运动精度可靠性,提出一种利用多体动力学和支持向量回归(Support Vector Regression,SVR)算法构建机构的运动模型并对其进行可靠度敏感性分析的方法。通过引入粒子群优化(Particle Swarm Optimization,PSO)算法对SVR的惩罚参数和核函数参数进行寻优,提高SVR的回归预测精度。为克服PSO容易早熟和搜索精度低等缺点,对惯性权重系数和学习因子进行改进,应用改进算法与标准PSO-SVR算法并结合蒙特卡洛模拟对四杆机构的可靠度敏感性进行分析研究。通过实验对比表明,改进的算法收敛速度更快、回归预测精度更加接近于蒙特卡洛模拟,且计算速度优于蒙特卡洛模拟。In order to improve the kinematical accuracy reliability of linkage mechanism,a method is proposed by establishing the motion model of linkage based on multi-body system dynamics method and Support Vector Regression(SVR),and analyzing the reliability sensitivity of the mechanism.To enhance the regression accuracy of SVR,the Particle Swarm Optimization(PSO)algorithm is used to search the best penalty parameter and kernel function parameter of SVR.The inertial weight and learning coefficients are improved to avoid the premature and low precision of PSO.The improved algorithm and standard PSO-SVR combined with Monte Carlo(MC)algorithms are used for solving the reliability sensitivity problem of linkage mechanism.The comparison shows that the improved PSO-SVR distinctly outperforms PSO-SVR both the rate of convergence and the regression accuracy,particularly the calculating speed better than MC.

关 键 词：连杆机构 可靠度 SVR 粒子群优化 蒙特卡洛模拟 

分 类 号：TH16[机械工程—机械制造及自动化] TH112.1