平面并联五连杆机构的改进PSO-SVR可靠性分析  

Reliability Analysis of Improved PSO-SVR for Planar Parallel Five-Bar Linkage Mechanism

作　　者：户燕会 任燕[2] 户东亮[3] 姜奎 HU Yanhui;REN Yan;HU Dongliang;JIANG Kui(Sino-German School of Engineering,Tongji University,Shanghai 201804,China;School of Mechanical Engineering,He’nan Polytechnic Institute,He’nan Nanyang 473000,China;SDIC Nanyang Thermal Power Co.,Ltd.,He’nan Nanyang 473000,China;Nanyang Changbow Machinery Technology Co.,Ltd.,He’nan Nanyang 473000,China)

机构地区：[1]同济大学中德工程学院,上海201804 [2]河南工业职业技术学院机械工程学院,河南南阳473000 [3]国电投南阳热电有限责任公司,河南南阳473000 [4]南阳长弓机械科技有限公司,河南南阳473000

出　　处：《机械设计与制造》2025年第2期179-182,共4页Machinery Design & Manufacture

基　　金：河南省高等学校重点科研项目(21B46002)。

摘　　要：为获得更高的SVM精度,采用改进粒子群优化(PSO)算法来实现SVM惩罚参数以及核函数的寻优过程,进一步改进了PSO权系数并对学习因子进行了优化,获得了更强的全局和局部搜索性能,有效避免算法产生早熟以及搜索精度不高的情况,显著增强了SVM回归预测能力。以五连杆机构为例,开展可靠性灵敏度分析。研究结果表明:改进PSO-SVR模型获得了更高的可靠性,这里算法可以达到更高预测精度。改进PSOSVR获得了更快收敛速度,同时RMSE也更小。采用改进PSO方法寻求参数可以使SVR获得更高预测精度。采用改进PSO-SVR模型只需少量样本便能够达到理想拟合精度,从而实现对参数可靠性灵敏度的准确分析。提高L1均值后,可以使五连杆机构达到更高可靠性,而L2、L3、L4均值和各参数方差提高后,则会引起系统可靠性的下降。In order to obtain higher accuracy of the SVM,an improved particle swarm optimization(PSO)algorithm to implement punish parameters and kernel function of SVM optimization process,and further improved PSO weight coefficient and the learning factor is optimized,the more global and local search performance,effectively avoid premature and search algorithm accuracy is not high,The regression prediction ability of SVM is significantly enhanced.Taking a four-bar linkage as an example,the reliability sensitivity analysis was carried out.The results show that the improved PSO-SVR model has higher reliability,and the proposed al⁃gorithm can achieve higher prediction accuracy.The improved PSOSVR results in a faster convergence rate and a smaller RMSE.Using the improved PSO method to find parameters can make the SVR obtain higher prediction accuracy.The improved PSO-SVR model needs only a small number of samples to achieve ideal fitting accuracy,so as to achieve accurate analysis of parameter reli⁃ability sensitivity.Increasing the mean value of L1 can make the four-bar linkage achieve higher reliability,while increasing the mean value of L2,L3 and L4 and the variance of each parameter will lead to the decrease of system reliability.

关 键 词：连杆机构 可靠性 支持向量机 粒子群优化 蒙特卡洛 预测精度 

分 类 号：TH16[机械工程—机械制造及自动化]