基于卷积神经网络的制孔出口毛刺预测方法

受深度学习理论的启发,对使用卷积神经网络预测航空装配制孔质量进行研究。以工艺参数(制孔转速、进给、每转进给)与主轴电流信号为网络输入,以制孔出口毛刺高度为预测目标,在实验基础上,分析了激活函数、目标优化算法、卷积核个数、卷积层层数、卷积窗口大小和学习率等对制孔出口毛刺卷积神经网络预测模型的影响,并通过启发式算法确定了最优的网络设置。研究结果表明,制孔出口毛刺预测平均相对误差为9.34%,实验集测试预测相对误差在15%以内,优于传统理论建模30%左右的预测相对误差。

Method for Predicting Burr of Hole Made Based on Convolutional Neural Network

Inspired by the theory of deep learning,this paper studies the use of convolutional neural networks,which is used to predict the quality of aerospace assembly holes.The process parameters(hole rotation speed,feed,feed per revolution)and spindle current signal are input to the network,and the burr height of the hole outlet is used as the prediction target.Based on the experiment,the influence of the activation function,target optimization algorithm,the number of cores and convolution layers,the size of the convolution window and the learning rate on the prediction model of the burr convolutional neural network for the hole outlet are analyzed and the optimal network settings are determined by the heuristic algorithm.The results show that the average relative error of the burr prediction of the hole outlet is 9.34%,and the relative error of the experimental set test is less than 15%,which is better than the prediction relative error of about 30%of the traditional theoretical modeling.