基于改进一维卷积神经网络和通道注意机制的刀具磨损预测研究

Research on Tool Wear Prediction Based on Improved One-dimensional Convolutional Neural Network and Channel Attention Mechanism

对刀具磨损状态的准确预测有利于提高产品质量和降低生产成本,而现有预测模型多采用人工提取特征,存在费时费力等问题。提出一种结合通道注意机制的改进一维卷积神经网络的刀具磨损预测模型。该模型将卷积神经网络中的全连接层替换为卷积、池化层来增强模型的特征提取能力;最后一层采用1×1卷积和全局平均池化整合全局信息来提高预测精度;同时引入通道注意机制来增强重要特征通道的权重,进一步提高预测精度。实验结果表明,该预测模型的MAE为6.89μm,比预测模型MLP和SVR分别降低了15.54μm,12.27μm,比CNN降低了8.78μm。

Accurate prediction of tool wear plays an important role in improving product quality and reducing production cost.In view of the existing prediction models,most of the features are extracted manually,which is time-consuming and laborious.A tool wear prediction model based on improved one-dimensional convolutional neural network(1D-CANN)combined with channel attention mechanism is proposed.In this model,the full connection layer of convolutional neural network is replaced by convolution and pooling layer to enhance the feature extraction capability of the model.In the last layer,1×1 convolution and global average pooling are used to integrate global information to improve the prediction accuracy.At the same time,channel attention mechanism is introduced to enhance the weight of important feature channels and further improve the prediction accuracy.The results show that MAE of the prediction model is 6.89μm,which is 15.54μm and 12.27μm lower than that of MLP and SVR,and 8.78μm lower than that of CNN.