基于神经网络算法的GH4169铣削温度研究

Research on milling temperature of GH4169 based on neural network

为了研究GH4169铣削加工中铣削参数对铣削温度的影响规律,设计了铣削参数与铣削温度之间的正交试验。基于正交试验的数据,分别利用径向基函数神经网络、反向传播神经网络和广义回归神经网络算法建立了GH4169铣削温度的预测模型,并对模型进行了训练、测试与验证。另外,采用极差分析法分析了工艺参数对铣削温度的影响规律。结果发现,基于径向基函数神经网络、广义回归神经网络、反向传播神经网络建立的GH4169铣削温度预测模型的平均预测误差分别为:3.27%、4.24%、5.05%,平均预测精度分别为:96.73%、95.76%、94.95%。基于径向基函数神经网络建立的模型预测精度最高,其次是广义回归神经网络模型和反向传播神经网络模型。GH4169铣削温度随着铣削速度、每齿进给量、铣削深度的增加而增加;影响铣削温度最主要的因素是铣削速度,其次是每齿进给量和铣削深度。

To study the influence of milling parameters on milling temperature in milling GH4169,an orthogonal test between milling parameters and milling temperature was designed.Based on the data obtained from the orthogonal test,the prediction models of GH4169 milling temperature were established by using radial basis function neural network,back propagation neural network and generalized regression neural network algorithms respectively,and the models were trained,tested and verified.In addition,the influence laws of milling parameters on milling temperature was analyzed by the range analysis method.The results showed that the average prediction error of GH4169 milling temperature prediction models based on RBF neural network、GRNN neural network、BP neural network were 3.27%、4.24%、5.05%,and the the average prediction accuracy were 96.73%、95.76%、94.95%.The model based on radial basis function neural network had the highest prediction accuracy,followed by back-propagation neural network model and generalized regression neural network model.The milling temperature of GH4169 increased with the increase of milling speed,feed per tooth and milling depth;The most important factor affecting milling temperature was milling speed,followed by feed rate per tooth and milling depth.