Abstract: To improve the accuracy and generalization performance of online tool wear monitoring,an online tool wear monitoring method was proposed based on image coding technology and convolutional neural network.Using Gramian Angle Field(GAF)image coding technology to image the time-series signal data collected during the milling process,which not only retained the original characteristic information of the signal but also enhanced time series feature information.To avoid the complexity and limitations caused by artificial feature extraction,the deep convolutional neural network was used to adaptively extract image features.Under the premise of ensuring the accuracy and generalization of the model,the small features in the image signal data were mined by deepening the number of network layers.Experiments with data sets used in similar researches had verified the effectiveness and feasibility of this method in online tool wear monitoring,and its accuracy had been greatly improved compared with other methods under multiple evaluation standards.

Key words: tool wear, online monitoring, Gramian angle field, image coding, deep convolutional neural network, feature extraction

摘要： 为提高刀具磨损在线监测的精度及泛化性能,提出一种基于格拉姆角场(GAF)编码技术和卷积神经网络(CNN)的刀具磨损值在线监测方法,利用GAF图像编码技术将铣削加工过程中采集的时间序列信号数据图像化,既保留了信号的原始特征信息,又增强了时间序列特征信息。采用深度CNN自适应的提取图像特征,避免人工特征提取带来的复杂性和局限性。使用同类研究所用的数据集进行实验,验证了该方法在刀具磨损在线监测中的有效性和可行性,在多项评价标准下其精度较其他几种方法有了较大提高。

关键词: 刀具磨损, 在线监测, 格拉姆角场, 图像编码, 深度卷积神经网络, 特征提取