A new tool wear condition monitoring method based on deep learning under small samples. (15th February 2022)

• Record Type: Journal Article
• Authors: Zhou, Yuqing; Zhi, Gaofeng; Chen, Wei; Qian, Qijia; He, Dedao; Sun, Bintao; Sun, Weifang
• Is Part Of: Measurement. Volume 189(2022)
• Publication Date: 2022-02-15
• Subjects: Small samples -- Tool condition monitoring -- Recurrence plot -- Multi-scale edge-labeling graph neural network; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2021.110622
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: A new tool wear condition detection method based on deep learning with multi- cutting force time series signal is proposed under small samples. Each cutting force sensor signal is expanded and encoded into a two- dimensional gray recurrence plot (RP), and then aggregated into a color RP. A multi-scale edge-labeling graph neural network is proposed to extract features from aggregated color RP to establishing a fully connected graph, in which the values of edge labels are obtained by updating the nodes and edge features. The proposed method outperforms three state-of-the-art methods with time series signal under small samples. Abstract: Tool wear condition monitoring (TCM) is an important part of machining automation. In recent years, deep learning (DL) based TCM methods have been widely researched. However, almost DL-based methods need sufficient learning samples to obtain good accuracy, which is hard for TCM in terms of cost and time. In order to enhance the recognition accuracy of DL-based TCM under small samples, this paper proposed a new improved multi- scale edge-labeling graph neural network (MEGNN). Each channel signal of a cutting force sensor is expanded to multi- dimensional data through phase space reconstruction. Then, these multi- dimensional data are encoded into a gray recurrence plot (RP), and aggregated into a color RP, which is input to MEGNN to extract features for establishing a fully connected graph. Finally, the tool wear condition is estimated through the updated edge labels using a weighted voting method. Applications of the proposed MEGNN- based method to PHM 2010 milling TCM dataset and our experiments demonstrate it outperforms three DL-based methods (CNN, AlexNet, ResNet) under small samples.