基于注意力机制与深度学习算法的机床主轴系统故障辨识

doi:10.12382/bgxb.2021.0202

摘要/Abstract

摘要： 针对具有复杂非线性特点的数控机床主轴系统整体动态退化故障较难辨识及故障研究难度大的问题，从数据分析入手，提出一种基于注意力机制与深度学习算法的智能化故障辨识方法，研究机床主轴系统的整体故障辨识问题。该方法设计了注意力机制的研究框架，将研究问题分为全局纵向大分类区间和局部横向细粒度区间两个维度：采用训练并调优后推理平均绝对误差达到0.028 7的门控循环单元模型，辨识出大分类区间的全局性退化故障；采用鲁棒性强且辨识准确率达99.7%的残差网络模型，在sym8小波基自适应软阈值降噪的基础上对局部细粒度区间故障进行准确细节辨识。结果表明：该方法可量化地辨识出主轴系统的整体故障；所提注意力机制可使大分类区间无法准确辨识的故障在细粒度区间得到有效区分，类内数据增长梯度由6.6%增加到43.8%；通过对机床主轴系统实际使用中在空载状态下遇到的不对中和局部共振等典型故障，以及在负载加工状态下故障的辨识研究，验证了所提方法的有效性与准确性。

关键词: 机床主轴系统, 故障辨识, 注意力机制, 门控循环单元模型, 残差网络模型

Abstract: The overall dynamic degradation fault of the numerically-controlled machine tool spindle system with complex nonlinear characteristics is difficultly identified and investigated.An intelligent fault identification method based on attention mechanism and depth learning algorithm is proposed to stud the overall fault identification of spindle system by starting with data analysis.The proposed method is used to design the research framework of attention mechanism，and divide the research problems into global vertical large classification interval dimension and local horizontal fine-grained interval dimension.The gated recurrent unit model with reasoning average absolute error of 0.028 7 after training and tuning is used to identify the global degradation faults in large classification interval.The residual network model with strong robustness and identification accuracy of 99.7% is used to accurately identify the local fine-grained interval faults， based on sym8 wavelet basis adaptive soft threshold noise reduction.The results show that the proposed method is used to quantitatively identify the overall fault of spindle system. The proposed attention mechanism is used to effectively distinguish the faults that cannot be accurately identified in the large classification interval in the fine-grained interval，and the data growth gradient in the category increases from 6.6% to 43.8%. The effectiveness and accuracy of the proposed method are verified by studying the typical faults，such as misalignment and local resonance encountered in the actual use of the machine tool spindle system under no-load，and the fault identification under loading.

Key words: machinetoolspindlesystem, faultidentification, attentionmechanisms, gatedrecurrentunitmodel, ResNetmodel

中图分类号:

TG519.1

王伟平, 王琦, 于洋. 基于注意力机制与深度学习算法的机床主轴系统故障辨识[J]. 兵工学报, 2022, 43(4): 861-875.

WANG Weiping, WANG Qi, YU Yang. Fault Identification of Machine Tool Spindle System Based on Attention Mechanism and Deep Learning Algorithm[J]. Acta Armamentarii, 2022, 43(4): 861-875.

参考文献

［1］黄志坚.机械设备振动故障监测与诊断[M].北京：化学工业出版社，2017.
HUANG Z J.Vibration fault monitoring and diagnosis of mechanical equipment[M].Beijing:Chemical Industry Press 2017. (in Chinese)
[2］ BENTLY D E，HATCH C T.Fundamentals of rotating machinery diagnostics[M]. New York，NY，US:ASME Press，2003.
[3］周林，赵杰，冯广飞.装备故障预测与健康管理技术[M].北京：国防工业出版社，2015.
ZHOU L，ZHAO J，FENG G F.Failure prediction and health management technology of equipment[M].Beijing: National Defense Industry Press，2015.(in Chinese)
[4］杨周，朱云鹏，张义民，等.基于Copula函数的数控车床可靠性综合分配方法[J].兵工学报，2016，37(1):131-140.
YANG Z，ZHU Y P，ZHANG Y M，et al.A comprehensive reliability allocation method of numerical-controlled lathes based on Copula function[J].Acta Armamentarii，2016，37(1):131-140. (in Chinese)
[5］王福元，赵建社，唐霖.多系统集成电解加工控制策略与故障诊断方法研究[J].兵工学报，2015，36(6):1074-1081.
WANG F Y，ZHAO J S，TANG L. Research on control strategy and fault diagnosis method of multi-system for electrochemical machining[J].Acta Armamentarii，2015，36(6):1074-1081.(in Chinese)
[6］邓超，陶志奎，吴军，等.基于性能退化的数控机床剩余寿命预测[J].机械工程学报，2018，54(17):181-189.
DENG C，TAO Z K，WU J，et al. Residual life prediction of NC machine tool based on performance degradation[J].Journal of Mechanical Engineering，2018，54(17):181-189.(in Chinese)
[7］林滨，陈东，韩植林，等.铝合金薄壁圆柱壳切削稳定性研究[J].兵工学报，2012，33(7):847-851.
LIN B，CHEN D，HAN Z L，et al. Research on the cutting stability of thin cylindrical shell of aluminum alloy [J].Acta Armamentarii，2012，33(7):847-851. (in Chinese)
[8］常丽萍，常亚南，杨雷雷.内圈倾斜角对高速角接触球轴承动态特性的影响[J].兵工学报，2021，42(6):1138-1147.
CHANG L P，CHANG Y N，YANG L L. The influence of inner ring tilt angle on dynamic characteristics of high-speed angular contact ball bearings[J].Acta Armamentarii，2021，42(6):1138-1147. (in Chinese)
[9］ PRABHU RAJA V，THIRUMALAIMUTHUKUMARAN M，THULA P R，et al.A strategy for minimization of thermal error in headstock assembly of CNC lathe[J].Machining Science and Technology，2019，23(1):57-58.
[10］邱继伟，罗海胜.结构可靠性分析的自适应共轭非线性近似方法[J].兵工学报，2021，42(2):146-152.
QIU J W，LUO H S.Self-adaptive conjugate nonlinear approximation method for structural reliability analysis[J].Acta Armamentarii，2021，42(2):146-152. (in Chinese)
[11］ SHENG Z Q，DAI S L，LIU Y C，et al.Characteristic analysis and optimization of CNC machine tool spindle based on finite element method[J].Advanced Materials Research，2012，507:217-221.
[12］ ZHAO H T，YANG J G，SHEN J H. Simulation of thermal behavior of a CNC machine tool spindle[J].International Journal of Machine Tools and Manufacture，2007，47(6):1003-1010.
[13］高峰，程明科，李艳，等.永磁同步型磨削电主轴偏心振动分析及实验[J].仪器仪表学报，2019，40(2):38-50.
GAO F，CHENG M K，LI Y，et al. Eccentric vibration analysis and test of permanent magnet synchronous grinding motorized spindle[J].Chinese Journal of Scientific Instrument，2019，40(2): 38-50. (in Chinese)
[14］ TIAN Y，MA J，LU C，et al.Rolling bearing fault diagnosis under variable conditions using LMD-SVD and extreme learning machine[J].Mechanism and Machine Theory，2015，90: 175-186.
[15］ LU C，WANG Z Y，ZHOU B.Intelligent fault diagnosis of rolling bearing using hierarchical convolutional network based health state classification[J].Advanced Engineering Informatics，2017，32:139-151.
[16］ DING Y，MA L，MA J，et al. Intelligent fault diagnosis for rotating machinery using deep Q-network based health state classification: a deep reinforcement learning approach[J].Advanced Engineering Informatics，2019，42:100977.
[17］ LU C，WANG Z Y，QIN W L，et al. Fault diagnosis of rotary machinery components using a stacked denoising autoencoder-based health state identification[J].Signal Processing，2017，130:377-388.
[18］ TRAN H，NOORI M，ALTABEY W A，et al.Fault diagnosis of rotating machinery using wavelet-based feature extraction and support vector machine classifier[J].High Speed Machining，2017，3(1): 23-41.
[19］ SHARMA R B，PAREY A.Modelling of acoustic emission generated in rolling element bearing[J].Applied Acoustics，2017，144:96-112.
[20］朱敏，许爱强，许晴，等.基于改进多层核超限学习机的模拟电路故障诊断[J].兵工学报，2021，42(2):356-369.
ZHU M，XU A Q，XU Q，et al.Fault diagnosis of analog circuits based on improved multilayer kernel extreme learning machine[J].Acta Armamentarii，2021，42(2):356-369. (in Chinese)
[21］ YANG W A，XIAO M H，ZHOU W，et al.Trace ratio criterion-based kernel discriminant analysis for fault diagnosis of rolling element bearings using binary immune genetic algorithm[J].Shock and Vibration，2016:8631639.

[1]	于博文，吕明. 改进的YOLOv3算法及其在军事目标检测中的应用[J]. 兵工学报, 2022, 43(2): 345-354.
[2]	赵书彬，徐诚，蒋弘毅，步春辰. 模块化枪族智能化辅助设计策略与方法[J]. 兵工学报, 2021, 42(4): 723-733.
[3]	穆思奇，林进健，汪海泉，魏雄志. 基于改进YOLOv4的X射线图像违禁品检测算法[J]. 兵工学报, 2021, 42(12): 2675-2683.
[4]	刘宗凯，刘满红，郭正阳，李占江. 全附体潜航器螺旋桨桨叶损伤故障辨识机理[J]. 兵工学报, 2021, 42(12): 2710-2721.