超声挤压轴类零件表面粗糙度预测模型分析

Analysis of surface roughness prediction model for ultrasonic extrusion shaft parts

为实现对超声挤压轴类零件表面粗糙度的合理预测,采用多元回归法、MLP神经网络法以及支持向量机回归法进行预测分析研究。将模型预估时所得数值与实际检测值进行比较分析,结果表明:相比于其他学者借助径向基神经网络构建的模型,其预测的相对误差为5.2%;借助多元回归法构建的模型具有更高的预测准确度,其预测的平均相对误差最小,最小值为4.92%。该预测模型可以进行不同参数下零件表面粗糙度的预测。

Multiple regression,MLP neural network and support vector machine regression were used to predict the surface roughness of ultrasonic extrusion shaft parts.The estimated value of the model and the actual detection value are compared,and the analysis shows that the relative mistake of the model is 5.2%compared with that of the model developed by other scholars with the help of RBF neural network,and the model constructed by multiple regression method has higher prediction accuracy,and its average relative mistake is small,the minimum value is 4.92%.The model can be used to predict the surface roughness of parts with different parameters.