(109c) Regression Model for Tool Wear Monitoring in Precision Machining
AIChE Annual Meeting
Applied Artificial Intelligence, Big Data, and Data Analytics Methods for Next-Gen Manufacturing Efficiency I
Monday, November 8, 2021 - 1:10pm to 1:30pm
Modern tool wear monitoring systems leverage data-driven models applying machine learning (ML) to predict tool wear and detect machine abnormalities . Many studies have involved ML algorithms that use techniques such as K-nearest neighbors , support vector machine , and convolutional neural networks . Although these methods have shown good performance, the fundamental methodology and complete toolchain that is suitable to predict tool wear still need to be investigated. Moreover, regression models that predict tool wear as a continuous process state/variable are seldom. The often-applied classification aims to predict discrete class labels, but tool wear is a continuously progressing system disturbance. Therefore, it is important to explore regression models to predict tool wear of continuous variables. The objective of this work is to propose a regression-based framework for predicting tool wear in CNC systems. We use support vector regression (SVR) as the ML algorithm, to provide insights into and quantify tool wear prediction in the machining process.
As shown in Fig.1, in the proposed framework, we collect, tabulate, and label audio and vibration data of the CNC process under varying tool wear and explore the performance of the proposed framework using regression. First, audio and vibration signals are converted to frequency domain through Fast Fourier Transform (FFT) to reduce the complexity of the audio and vibration time waveforms collected. Then, Principal Components Analysis (PCA) is used to eliminate features of small variance, increase the information content, reduce overfitting and computational effort. The selected features of audio and vibration signals are modeled and analyzed with an SVR model. The framework of Fig.1 was applied to a Mazak Variaxis 630-5X II T Mill/Turn Machine 40 hp 5-axis CNC machine. The experimental facility was instrumented to acquire process data for machine power, spindle power, spindle vibration, and to record audio/video and is MTConnect capable. Moreover, the machine was controlled to perform milling at various spindle speeds, widths of cut, depths of cut, and feed rates. Application of the framework of Fig. 1 for tool wear monitoring will be presented. In terms of the accuracy in tool wear prediction, it will be shown that the audio and vibration signals are sufficiently rich in information about tool wear so that a very accurate model (0.013 and 0.007 in RMSE, respectively) could be developed with those signals. Computation of the relative error between real tool wear measurement and predicted tool wear showed 1.54% error in audio signal-based regression and 3.1% error in vibration signal-based regression. Fusion of these signals further improved the accuracy of the regression model. In conclusion, the acoustic signals that often lead engineers and operators identify issues and anomalies in the operation of precision machining equipment were used successfully to learn ML models to predict tool wear in a very accurate manner. Extrapolation of these models in time to be useful in machine prognostics is currently underway and will be discussed in this presentation.
Acknowledgment: This material is based upon work supported by the U.S. Department of Energyâs Office of Energy Efficiency and Renewable Energy (EERE) under the Advanced Manufacturing Office Award Number DE-EE0007613.
Disclaimer: This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.â
 H. Shao, H. L. Wang, and X. M. Zhao, âA cutting power model for tool wear monitoring in milling,â Int. J. Mach. Tools Manuf., vol. 44, no. 14, 2004, doi: 10.1016/j.ijmachtools.2004.05.003.
 Z. Li, R. Liu, and D. Wu, âData-driven smart manufacturing: Tool wear monitoring with audio signals and machine learning,â J. Manuf. Process., vol. 48, 2019, doi: 10.1016/j.jmapro.2019.10.020.
 A. Jegorowa, J. GÃ³rski, J. Kurek, and M. Kruk, âUse of nearest neighbors (K-NN) algorithm in tool condition identification in the case of drilling in melamine faced particleboard,â Maderas Cienc. y Tecnol., vol. 22, no. 2, 2020, doi: 10.4067/S0718-221X2020005000205.
 D. Shi and N. N. Gindy, âTool wear predictive model based on least squares support vector machines,â Mech. Syst. Signal Process., vol. 21, no. 4, 2007, doi: 10.1016/j.ymssp.2006.07.016.
 F. Jia, Y. Lei, J. Lin, X. Zhou, and N. Lu, âDeep neural networks: A promising tool for fault characteristic mining and intelligent diagnosis of rotating machinery with massive data,â Mech. Syst. Signal Process., vol. 72â73, 2016, doi: 10.1016/j.ymssp.2015.10.025.