基于PSO算法改进BP神经网络的氟金云母点磨削工艺参数优化

摘要

Through a high speed point grinding experiment,the surface hardness and roughness of the finished surface was tested,and the variations of the surface hardness and roughness with the process parameters were analyzed.Single factor experimental values were predicted with PSO-BP.A se-ries of one-dimensional models of surface hardness and roughness process parameters of fluorophlogo-pite were built by least-squares fitting.Correlation coefficient test was used to verify the models'high reliability.Multivariate models about surface hardness and roughness process parameters were pro-posed by analyzing one-dimensional models.The multivariate numerical models were optimized accord-ing to the results of orthogonal experiments and PSO and were proved to have high reliability by ex-periment.A dual obj ective optimization of two multivariate models was carried out by PSO algorithm, and a set of reasonable process parameters was obtained.%通过氟金云母的高速点磨削试验,测试了加工表面硬度和表面粗糙度,分析了表面硬度和表面粗糙度随工艺参数的变化规律.基于单因素试验值和PSO算法改进的BP神经网络,利用最小二乘拟合,建立了氟金云母点磨削表面硬度和表面粗糙度关于各工艺参数的一元模型,以相关系数检验模型,证明模型具有较高可靠性.分析一元模型,提出表面硬度和表面粗糙度分别关于工艺参数的多元模型.基于正交试验值和PSO算法,对模型进行优化求解,并通过试验证明了模型具有较高可靠性.利用PSO算法对两个多元模型进行双目标优化,求解得到一组较为合理的工艺参数值.