结合域对抗自适应的刀具磨损预测方法

董靖川; 谭志兰; 王太勇; 武晓鑫

doi:10.13433/j.cnki.1003-8728.20200614

结合域对抗自适应的刀具磨损预测方法

doi: 10.13433/j.cnki.1003-8728.20200614

天津大学机械工程学院，天津　300350

基金项目: 国家自然科学基金面上项目（51975402）

详细信息

作者简介:
董靖川（1983−），高级工程师，博士，研究方向为数控技术、测控技术，new_lightning@aliyun.com

中图分类号: TP183
计量
- 文章访问数: 197
- HTML全文浏览量: 221
- PDF下载量: 72
- 被引次数: 0
出版历程
- 收稿日期: 2021-04-24
- 网络出版日期: 2023-03-27
- 刊出日期: 2023-02-25

Prediction Method of Tool Wear Combined withDomain Adversarial Adaptation

School of Mechanical Engineering, Tianjin University, Tianjin 300350, China

摘要

摘要: 数控加工中存在刀具几何误差及安装误差、刀具及工件材料性能的随机波动等因素，导致刀具之间的磨损过程与监测信号上存在较大差异的问题，使得刀具磨损值难以精确预测。为此，本文提出了一种结合域对抗自适应的多尺度分布式卷积长短时记忆网络模型（Multiscale time-distributed convolutional long short-term memory，MTDCLSTM）。将加工过程中采集到的多传感器信号作为模型输入，通过域分类器与预测器之间的对抗学习，提取出可有效表征刀具磨损且与域无关的多尺度时空特征，经预测器的非线性映射，实现对刀具磨损值的精确预测。实验结果表明，结合域对抗自适应的MTDCLSTM模型预测性能明显优于分布式卷积神经网络、长短时记忆网络、卷积神经网络与支持向量机模型。与基于迁移成分分析的支持向量回归模型相比，本文模型的均方根误差与平均绝对误差分别降低了59.8%和62.5%，决定系数提高了66.1%，可有效缩小刀具个体之间的差异，提高磨损值预测精度。
- 刀具磨损 /
- 域对抗自适应 /
- 多尺度时空特征 /
- 分布式卷积神经网络 /
- 长短时记忆网络
Abstract: In the CNC machining, due to the factors such as the geometric and the installation error of the cutting tool, the random variations in the material properties of the workpiece and the cutter, the wearing process and the monitoring signals between the individual tools have large differences, which makes the tool wear values to difficultly predict. To address this problem, a multiscale time-distributed convolutional long short-term memory model (MTDCLSTM) combined with domain adversarial adaptation is proposed. With the multi-sensor signals obtained in the machining as the model input, the model extract multiscale spatio-temporal features that can effectively characterize the tool wear and are independent of the domain through the adversarial learning between the domain classifier and the predictor. The accurate prediction value of the tool wear can be obtained by using the non-linear mapping of the predictor. Experimental results show that the prediction performance of the MTDCLSTM model combined with domain adversarial adaptation is significantly better than the time-distributed convolutional neural networks, long short-term memory neural networks model, convolutional neural networks and support vector machine models. Comparing with the support vector regression model based on transfer component analysis, the root mean square error and average absolute error of the present model were reduced by 59.8% and 62.5%, respectively, and the coefficient of determination was increased by 66.1%, which means the present model can effectively reduce the difference between the individual tools and improve the accuracy of the wear prediction.
- tool wear /
- domain adversarial adaptation /
- multiscale spatio-temporal features /
- time-distributed convolutional neural networks /
- long short-term memory neural networks

HTML全文

图 1 结合域对抗自适应的MTDCLSTM模型结构

下载: 全尺寸图片幻灯片

图 2 分布式卷积神经网络结构

下载: 全尺寸图片幻灯片

图 3 刀具磨损实验平台

下载: 全尺寸图片幻灯片

图 4 3把刀具的振动监测信号Q-Q图

下载: 全尺寸图片幻灯片

图 5 3把刀具的磨损曲线

下载: 全尺寸图片幻灯片

图 6 不同尺度下模型的预测结果

下载: 全尺寸图片幻灯片

图 7 本文模型在6个实验任务上的预测结果

下载: 全尺寸图片幻灯片

表 1 实验加工参数

加工参数	数值
主轴转速/（r·min⁻¹）	10400
进给速度/（mm·min⁻¹）	1555
径向切宽/mm	0.125
轴向切深/mm	0.2

下载: 导出CSV

表 2 域对抗自适应对模型性能的影响

模型尺度	单尺度	单尺度*	双尺度	双尺度*	三尺度	三尺度*	四尺度	四尺度*
RMSE	16.26	13.44	17.96	11.98	13.35	9.16	12.24	11.77
MAE	13.04	10.36	14.26	9.04	10.39	6.97	9.63	8.97
R²	0.76	0.83	0.71	0.88	0.84	0.93	0.88	0.89
注：*为该模型使用了域对抗自适应。

下载: 导出CSV

表 3 本文所提出的模型与其他模型对比结果

模型	RMSE	MAE	R²
TDCNN	11.81	8.74	0.88
LSTM	16.11	11.41	0.78
CNN	38.34	31.50	−0.41
SVR	23.30	16.76	0.53
TCA-SVR	22.80	18.60	0.56
本文模型	9.16	6.97	0.93

下载: 导出CSV

参考文献(17)

[1]	KONG D D, CHEN Y J, LI N. Gaussian process regression for tool wear prediction[J]. Mechanical Systems and Signal Processing, 2018, 104: 556-574 doi: 10.1016/j.ymssp.2017.11.021
[2]	程灿, 李建勇, 徐文胜, 等. 基于支持向量机与粒子滤波的刀具磨损状态识别[J]. 振动与冲击, 2018, 37(17): 48-55 + 71 CHENG C, LI J Y, XU W S, et al. Tools' wear state recognition based on support vector machine and particle filtering[J]. Journal of Vibration and Shock, 2018, 37(17): 48-55 + 71 (in Chinese)
[3]	肖鹏飞, 张超勇, 罗敏, 等. 基于自适应动态无偏最小二乘支持向量机的刀具磨损预测建模[J]. 中国机械工程, 2018, 29(7): 842-849 XIAO P F, ZHANG C Y, LUO M, et al. Modeling method for tool wear prediction based on ADNLSSVM[J]. China Mechanical Engineering, 2018, 29(7): 842-849 (in Chinese)
[4]	XU X W, WANG J W, MING W W, et al. In-process tap tool wear monitoring and prediction using a novel model based on deep learning[J]. The International Journal of Advanced Manufacturing Technology, 2021, 112(1-2): 453-466 doi: 10.1007/s00170-020-06354-y
[5]	CAI W L, ZHANG W J, HU X F, et al. A hybrid information model based on long short-term memory network for tool condition monitoring[J]. Journal of Intelligent Manufacturing, 2020, 31(6): 1497-1510 doi: 10.1007/s10845-019-01526-4
[6]	QIAO H H, WANG T Y, WANG P. A tool wear monitoring and prediction system based on multiscale deep learning models and fog computing[J]. The International Journal of Advanced Manufacturing Technology, 2020, 108(7-8): 2367-2384 doi: 10.1007/s00170-020-05548-8
[7]	董靖川, 徐明达, 王太勇, 等. 分布式卷积神经网络在刀具磨损量预测中的应用[J]. 机械科学与技术, 2020, 39(3): 329-335 DONG J C, XU M D, WANG T Y, et al. Application of distributed convolutional neural network in wear prediction of tool[J]. Mechanical Science and Technology for Aerospace Engineering, 2020, 39(3): 329-335 (in Chinese)
[8]	陈仁祥, 吴志元, 胡小林, 等. 深度特征联合匹配的不同刀具间磨损状态识别[J]. 仪器仪表学报, 2020, 41(12): 138-145 CHEN R X, WU Z Y, HU X L, et al. Wear state recognition for different tools based on the joint matching of depth characteristics[J]. Chinese Journal of Scientific Instrument, 2020, 41(12): 138-145 (in Chinese)
[9]	PAN S J, TSANG I W, KWOK J T, et al. Domain adaptation via transfer component analysis[J]. IEEE Transactions on Neural Networks, 2011, 22(2): 199-210 doi: 10.1109/TNN.2010.2091281
[10]	李广. 结合迁移学习的机床刀具磨损预测研究[D]. 成都: 电子科技大学, 2020 LI G. Research on tool wear prediction of machine tool based on transfer learning[D]. Chengdu: University of Electronic Science and Technology of China, 2020 (in Chinese)
[11]	MAO W T, HE J L, ZUO M J. Predicting remaining useful life of rolling bearings based on deep feature representation and transfer learning[J]. IEEE Transactions on Instrumentation and Measurement, 2020, 69(4): 1594-1608 doi: 10.1109/TIM.2019.2917735
[12]	GANIN Y, USTINOVA E, AJAKAN H, et al. Domain-adversarial training of neural networks[J]. The Journal of Machine Learning Research, 2016, 17(1): 2096-2030
[13]	蔡伟立, 胡小锋, 刘梦湘. 基于迁移学习的刀具剩余寿命预测方法[J]. 计算机集成制造系统, 2021, 27(6): 1541-1549 CAI W L, HU X F, LIU M X. Prediction method of tool remaining useful life based on transfer learning[J]. Computer Integrated Manufacturing Systems, 2021, 27(6): 1541-1549 (in Chinese)
[14]	DE OLIVEIRA DA COSTA P R, AKÇAY A, ZHANG Y Q, et al. Remaining useful lifetime prediction via deep domain adaptation[J]. Reliability Engineering & System Safety, 2020, 195: 106682
[15]	RAGAB M, CHEN Z H, WU M, et al. Adversarial transfer learning for machine remaining useful life prediction[C]//2020 IEEE International Conference on Prognostics and Health Management (ICPHM). Detroit: IEEE, 2020: 1-7
[16]	The Prognostics and Health Management Society. 2010 PHM society conference data challenge[EB/OL]. (2010-05-18). https://www.phmsociety.org/competition/phm/10
[17]	万鹏, 李迎光, 刘长青, 等. 基于域对抗门控网络的变工况刀具磨损精确预测方法[J]. 航空学报, 2021, 42(10): 318-331 WAN P, LI Y G, LIU C Q, et al. Method for accurate prediction of tool wear under varying cutting conditions based on domain adversarial gating neural network[J]. Acta Aeronautica et Astronautica Sinica, 2021, 42(10): 318-331 (in Chinese)