基于改进BP算法的模具零件表面抛光质量预测研究  

Research on quality prediction of surface polishing on die and mould parts based on improved BP algorithm

作　　者：刘守河 易建业 谢晖 LIU Shouhe;YI Jianye;XIE Hui(Ji Hua Laboratory,Foshan 528200,China;Agile Intelligent Technology(Guangdong),Co.,Ltd.,Foshan 528225,China)

机构地区：[1]季华实验室,广东佛山528200 [2]大捷智能科技(广东)有限公司,广东佛山528225

出　　处：《模具工业》2023年第10期9-19,共11页Die & Mould Industry

基　　金：季华实验室项目(X210181TB210);佛山市科技创新项目(1920001000041)。

摘　　要：针对模具零件表面自动化抛光的工艺参数现场调试困难、抛光后模具零件表面质量不一致等问题,提出了基于改进BP算法的模具零件表面抛光质量预测模型。通过采集模具零件表面抛光试验样本参数,构建预测模型输入参数集,将混沌理论、动态权重、动态学习因子和高斯变异策略引入粒子群优化算法(PSO),利用改进后的粒子群优化算法(IPSO)对BP算法中权值和阈值的更新策略进行优化,并构建了基于IPSO-BP算法的模具零件表面抛光质量预测模型,结合快速非支配排序遗传算法(NS-GA-II)建立多目标优化模型,实现对模具零件表面抛光质量的高精度预测以及抛光工艺参数的优化,对比5种常规预测模型,结果表明基于IPSO-BP算法的预测模型具有较高的预测精度。Aiming at the problem of debugging difficulties of automatic polishing process parame⁃ters on site and inconsistent surface quality of die and mould parts after polishing,a prediction model for surface polishing quality of die and mould parts based on improved BP algorithm was put forward.By collecting the sample parameters of the surface polishing test of die and mould parts,the input parameter set of the prediction model was constructed.The chaos theory,dynamic weights,dynamic learning factor,and gauss mutation strategy were introduced into the particle swarm optimization(PSO),which was used to optimize the update strategy of weights and thresh⁃olds.The prediction model of the surface polishing quality of die and mould parts was constructed based on the IPSO-BP algorithm,combined with the quick non-dominated sorting genetic algo⁃rithm(NSGA-II),a multi-objective optimization model was established to realize the high-precision prediction and the optimization of polishing process parameters.Compared five conventional pre⁃diction models,the results showed that the model based on IPSO-BP had higher accuracy.

关 键 词：模具零件抛光 粒子群优化算法 神经网络 非支配排序遗传算法 多目标优化 表面粗糙度 

分 类 号：TG580.692[金属学及工艺—金属切削加工及机床]