基于VGG-19卷积神经网络的刀具磨损监测方法

Tool wear monitoring method based on VGG-19 convolution neural network

有刀具状态监测的加工生产既能提高加工效率又能降低生产成本,是智能制造生产的关键。近几年深度学习成为研究刀具磨损问题的主流算法。提出一种基于VGG-19卷积神经网络的刀具磨损监测方法,该方法应用小波包变换对振动信号进行处理并提取能量图,应用VGG-19卷积神经网络预测刀具磨损状态。结果表明,适当增加网络层数,可以学习更多数据特征并得到更好的预测表现;与其他卷积神经网络相比,VGG-19层数适合,预测准确率稳定,损失函数值最小,该方法对刀具磨损类型的预测表现最好。

Processing and production with tool condition monitoring can not only improve processing efficiency,but also reduce production costs,which is the key to intelligent manufacturing production.In recent years,deep learning becomes the mainstream algorithm for studying tool wear.A tool wear monitoring method based on VGG-19 convolution neural network is proposed.This method uses wavelet packet transform to process vibration signal and extract energy maps,and uses VGG-19 convolution neural network model to predict tool wear status.The results show that more data features can be learned and better prediction performance can be obtained by properly increasing the number of network layers.Comparing with other convolutional neural networks,VGG-19 convolution neural network has suitable number of layers,stable prediction accuracy and minimum loss function value.This method performs best in predicting tool wear types.