基于遗传算法与BP神经网络的微晶玻璃点磨削工艺参数优化  

Process Parameter Optimization Based on BP Neural Networks and GA in Point Grinding Low Expansion Glass

作　　者：马廉洁[1,2] 曹小兵[1] 巩亚东[2] 陈小辉[1] 

机构地区：[1]东北大学秦皇岛分校,秦皇岛066004 [2]东北大学,沈阳110819

出　　处：《中国机械工程》2015年第1期102-106,共5页China Mechanical Engineering

基　　金：国家自然科学基金资助项目(51275083)

摘　　要：通过低膨胀微晶玻璃点磨削实验,测试了加工表面粗糙度、表面硬度,分析了实验数据变化趋势。通过最小二乘拟合,建立了关于粗糙度、表面硬度的一元数值模型,并将模型预测值与实验值进行了比较,以验证模型的精确性,结果表明模型具有较高的精度。根据正交实验结果,基于BP神经网络算法和遗传算法,建立了粗糙度、表面硬度的多元数值模型并以此作为目标函数,以表面硬度最大和表面粗糙度最小作为优化目标,基于遗传算法进行了工艺参数的双目标优化,获得了一组点磨削工艺参数的最优解范围,实验验证结果表明优化结果是合理的。The trends of experimental data were analyzed, the surface roughness and surface hard- ness were tested in point-grinding low expansion glass ceramics. The numerical models of surface roughness and hardness were established by the least square fitting. The accuracy of the model was tested by coefficient of determination, and the model predictions were compared with experimental da- ta to validate the accuracy of the model. The results indicated that the model has high accuracy. Based on BP neural networks and GA, the multivariate numerical models were built on surface roughness and hardness according to the results of orthogonal experiments. And both of the models were select- ed as the objective function. Optimization goal was the minimum of surface roughness and the maxi- mum surface hardness, dual objectives optimization was carried out based on GA. A range of the opti- mal solution was obtained about point grinding process parameters. Experimental validation results indicate that optimal results are reasonable.

关 键 词：BP神经网络 参数优化 点磨削 微晶玻璃 

分 类 号：TH161[机械工程—机械制造及自动化]