结合时空特征的多传感器刀具磨损监测

Multi-Sensor Tool Tear Yonitoring Combined with Temporal and Spatial Characteristics

针对传统深度学习方法监测刀具磨损状况时,相关特征提取繁琐,数据隐含信息提取不全面导致识别精度较低等问题,提出了结合时空特征的多传感器刀具磨损监测模型。首先,将不同传感器采集的波形信号经简单预处理后作为输入,再使用多通道1D卷积神经网络(MC-1DCNN)提取输入数据的空间特征;然后,利用双向长短时记忆网络(BiLSTM)提取时序特征;最终,由全连接层和Softmax层对特征进行分类。仿真结果表明,监测模型流程简单、识别准确率高,具备较强的可适用性。

In order to solve the problem that the traditional depth learning method for monitoring tool wear is tedious in extracting relevant features,and the incomplete extraction of data hidden information leads to low recognition accuracy,a multi-sensor tool wear monitoring model combined with spatio-temporal features is proposed.First,the waveform signals collected by different sensors are simply preprocessed as input,then the multi-channel 1D convolutional neural network(MC-1DCNN) is used to extract the spatial features of the input data,and then the bidirectional long short memory network(BiLSTM) is used to extract the temporal features.Finally,the features are classified by the full connection layer and the Softmax layer.The simulation results show that the monitoring model has a simple process,high recognition accuracy,and strong applicability.