Estimation of Tool Wear and Surface Roughness Development Using Deep Learning and Sensors Fusion

Abstract

This paper proposes an estimation approach for tool wear and surface roughness using deep learning and sensor fusion. The one-dimensional convolutional neural network (1D-CNN) is utilized as the estimation model with X- and Y-coordinate vibration signals and sound signal fusion using sensor influence analysis. First, machining experiments with computer numerical control (CNC) parameters are designed using a uniform experimental design (UED) method to guarantee the variety of collected data. The vibration, sound, and spindle current signals are collected and labeled according to the machining parameters. To speed up the degree of tool wear, an accelerated experiment is designed, and the corresponding tool wear and surface roughness are measured. An influential sensor selection analysis is proposed to preserve the estimation accuracy and to minimize the number of sensors. After sensor selection analysis, the sensor signals with better estimation capability are selected and combined using the sensor fusion method. The proposed estimation system combined with sensor selection analysis performs well in terms of accuracy and computational effort. Finally, the proposed approach is applied for on-line monitoring of tool wear with an alarm, which demonstrates the effectiveness of our approach.

Monitoring and Predicting the Surface Generation and Surface Roughness in Ultraprecision Machining: A Critical Review

“…DL, widely regarded as the future of modern-day manufacturing, addresses bottlenecks in areas such as intelligent fault diagnosis, tool condition monitoring, and bearing fault diagnosis. There are several types of DL architectures discussed by various researchers in the area of manufacturing [109]. The most widely used architectures in DL includes convolutional neural networks [110], auto encoder [111], and recurrent neural network [112]; all these DL techniques have the advantage of automatic feature learning.…”

A Novel Multi-Task Learning Model with PSAE Network for Simultaneous Estimation of Surface Quality and Tool Wear in Milling of Nickel-Based Superalloy Haynes 230

“…Rao et al established an optimized gray model (OGM) (1,N) to predict surface roughness and tool wear simultaneously using vibration signals [20]. Huang et al estimated tool wear and surface roughness simultaneously with one-dimensional convolutional neural network (1D CNN) and multiple sensors fusion method [21]. Similarly, Chen et al studied the applications of 1D CNN and vibration signals in tool wear detection and surface roughness estimation [22].…”

Neural Architecture Search for 1D CNNs—Different Approaches Tests and Measurements

“…These layers make use of several components/techniques, such as convolutions, activation functions, pooling, dropout, batch normalisation, fully connected blocks, among others, that can be combined in multiple ways. A CNN structure can be built according to the following aspects [ 11 , 13 , 14 ]: The input data can have a 1D, 2D, or 3D format. The source of this data can be, for instance, from sensors, audio, video, and 3D images.…”

“…Despite the scarcity of 1D CNN research, some very recent research works published on MDPI Sensors Journal can be found, e.g., the work developed by R. A. Osman et al, on interference avoidance on device communications using 5G communication; the advanced diagnosis of soft faults on motor power cables by H. Kim et al; the research “Estimation of Tool Wear and Surface Roughness Development Using Deep Learning and Sensors Fusion” presented by P.M. Huang and C.H. Lee; the monitoring of the spatter behaviours on acoustic signals developed by S. Luo et al Although all the papers cited before applied 1D CNNs on their research, neither of these very recent works used Neural Architecture Search (NAS) techniques (some do not even discuss the optimisation), resorting instead to the “traditional” grid search or trial-and-error approaches, which is the research gap that led to the need for this research [ 14 , 27 , 28 , 29 , 30 ].…”