基于卷积门控循环神经网络的刀具磨损状态监测

Tool Wear Condition Monitoring Based on Convolution-gated Recurrent Neural Network

针对刀具磨损状态在线监测需求,提出一种基于卷积门控循环神经网络的刀具磨损状态在线监测方法。综合卷积神经网络和门控循环神经网络的优点,构建了卷积门控循环神经网络;以切削力为输入信号,通过小波变换滤除噪声;利用卷积神经网络提取表征刀具磨损状态关键信息的高维特征;通过门控循环神经单元使模型在时间尺度上的累积效应得到充分表达,体现磨损的时序特性。实验表明,在有限的刀具磨损数据样本条件下,通过卷积门控循环神经网络进行刀具磨损状态监测具有较好的效果,其准确率达到97%。

Aiming at the demand of online monitoring tool wear status,a method for online monitoring tool wear status based on convolution gated recurrent neural network was proposed.Based on the advantages of convolutional neural network and gated recurrent unit neural network,a convolutional gated recurrent neural network was constructed;the cutting force was used as the input signal and the noise was removed by wavelet noise reduction;the convolutional neural network was used to extract the high-dimensional features that represented the key information of the tool wear status;the cumulative effect of the model on the time scale was fully expressed through the gated recurrent neural unit,which reflected the wear time-series characteristics.Experiments showed that under the condition of limited tool wear data sample,the tool wear state monitoring through convolution-gated recurrent neural network had a good effect,and the accuracy rate was 97%.