基于SGPLVM-LSSVM算法的U形折弯件模型参数优化研究  

Research on Parameter Optimization of U-shaped Bending Parts Model Based on SGPLVM-LSSVM Algorithm

作　　者：徐承亮[1] 曹志勇[2] 王大军[3] 胡吉全[4] XU Chengliang;CAO Zhiyong;WANG Dajun;HU Jiquan(Information Engineering School,Guangzhou Vocational College of Technology and Business,Guangzhou Guangdong 511442,China;School of Material Science and Engineering,Hubei University,Wuhan Hubei 430074,China;College of Automation, Chongqing University of Posts and Telecommunications,Chongqing 400065,China;Logistics Engineering College, Wuhan University of Technology,Wuhan Hubei 430072,China)

机构地区：[1]广州科技贸易职业学院信息工程学院,广东广州511442 [2]湖北大学材料科学与工程学院,湖北武汉430074 [3]重庆邮电大学自动化学院,重庆400065 [4]武汉理工大学物流工程学院,湖北武汉430072

出　　处：《机床与液压》2018年第20期29-32,58,共5页Machine Tool & Hydraulics

基　　金：国家自然基金面上项目(51675201);材料成形国家重点实验室开放基金资助项目(P2018-006)

摘　　要：影响高强度U形折弯件回弹的因素众多,比如工件尺寸、力学性能和负载条件等,使得高强度折弯件的弯曲回弹难以控制。把回弹角α和最小弯曲回弹半径R作为双目标函数,首先利用监督学习-高斯过程隐变量模型(SGPLVM)进行变量筛选和降维,构建U形折弯件的最小二乘支持向量机模型(LSSVM);再把SGPLVM-LSSVM实验结果分别与SVM、FEM、实际零件进行比较,验证了此算法模型的可行性。There are many factors influencing springback of high strength U-shaped bending parts,such as workpiece size,mechanical properties and load conditions,which make bending springback of high-strength bending parts be difficult to control.The minimum bending radius R and the springback angleαwere taken as two objective functions.Firstly,supervised gaussian process latent variable model(SGPLVM)was used for variable selection and dimensionality reduction,the least squares support vector machine(LSSVM)model for U-shaped bending part was constructed.The prediction results of SGPLVM-LSSVM were compared with SVM,FEM prediction results and actual engineering parts to verify the feasibility of the proposed model.

关 键 词：U形折弯件 支持向量机模型 监督学习-高斯过程隐变量模型 

分 类 号：TG316[金属学及工艺—金属压力加工]