轴承作为最常用的重要支撑元件，在机械设备中使用极其普遍，对其进行故障诊断一直是提高经济效益和保障运行安全的研究热点。在运行过程中，滚动轴承受转速、载荷和润滑状态的影响，发生故障在所难免。振动信号中包含丰富的机械运行状态信息，通过对其分析能够判别滚动轴承是否故障。然而，在滚动轴承的复合故障状态下，其振动信号成分复杂，故障特征难以准确提取，对后续故障诊断带来了极大挑战。为此，本文研究滚动轴承振动信号的故障特征提及诊断方法，在单一和复合故障模式下，以小波分析和神经网络为基础，研究不同故障模式下的滚动轴承精确故障诊断及其优化技术，主要研究内容包括：

（1）振动信号特征提取研究。针对单一故障，提出采用小波消噪-小波包分解对振动信号提取能量特征，并对二者的小波(包)基函数及分解层数进行优选。针对复合故障，提出采用小波包联合AR(Auto-Regressive)谱提取振动信号能量熵，并优选小波包基函数及分解层数。对上述两种不同故障模式下的特征提取方法进行实验验证，结果证明两种方法均能准确提取相应故障模式特征。

（2）故障诊断模型研究。采用壳体单路振动加速度信号，开展故障诊断模型研究。针对滚动轴承单一故障，提出一种自适应无速度项粒子群算法(Adaptive No Velocity Term Particle Swarm Optimization, ANVTPSO)优化BP神经网络(Back Propagation Neural Network, BPNN)阈值和权值参数的ANVTPSO-BPNN故障诊断模型。针对滚动轴承复合故障，提出一种自组织映射网络(Self-Organizing Maps, SOM)与BPNN组合的SOM-BPNN网络模型，并采用ANVTPSO优化SOM-BPNN的阈值和权值参数，构建ANVTPSO-SOM-BPNN故障诊断模型。对ANVTPSO-BPNN、ANVTPSO-SOM-BPNN模型进行实验验证，结果表明在滚动轴承单一故障及复合故障模式下两种模型诊断准确率相较BPNN均有明显提升。

（3）多振动信息融合诊断研究。滚动轴承-转子复合故障相较单一故障振动信号较为复杂，采用单路信号容易造成故障特征提取困难，无法进行故障准确诊断。提出采用双路加速度信号以及单路加速度联合位移信号两种方式进行不同模式复合故障诊断，对各单路信号采用ANVT-PSO-SOM-BPNN方法进行故障诊断，进而通过Dempster-Shafer (D-S)证据理论对诊断结果进行决策层融合，实现对滚动轴承-转子复合故障的可靠诊断。针对多振动信息融合的滚动轴承-转子复合故障诊断方法进行了实验验证，结果表明在加速度联合位移信号方案下，融合后的故障诊断准确率达到98.28%，相较双路加速度信号方案下提高了6.86%。

As a key supporting component, the rolling bearing is widely used in rotating machinery and the fault diagnos of rolling bearing has always been a research hotspot to improve the economic benefit and ensure the operation safety. In the working process, the failure of rolling bearing is inevitable due to the change of speed, load and lubrication state. Vibration signal reflects rich mechanical operation information, through which fault location and type identification of rolling bearing can be realized. However, in complex fault states, the vibration signal components of rolling bearing are complex, and the fault features are difficult to extract accurately, which brings great challenges for fault diagnosis. Therefore, this paper studies the fault feature extraction and diagnosis method of rolling bearing based on vibration signal. Under single and composite fault modes, based on wavelet analysis and neural network, this paper studies the precision fault diagnosis method and optimization scheme of rolling bearing under different fault modes. The main research contents include:

Study of vibration signal feature extraction. Under the single fault and composite fault modes of rolling bearing, two feature extraction methods are proposed: for the single fault of rolling bearing, wavelet analysis is proposed to denoise the vibration signal, and then the energy features are extracted through wavelet packet decomposition, and the basis functions and decomposition levels of the two are optimized. For the composite fault of rolling bearing, wavelet packet combined with AR (Auto Regressive) spectrum is proposed to extract the energy entropy feature of vibration signal, and the wavelet packet basis function and decomposition level are optimized. The experimental results show that the two methods can accurately extract the features of the corresponding fault modes.

Fault diagnosis model study. The fault diagnosis method of rolling bearing is studied by using the vibration acceleration signal of single shell. For single fault of rolling bearing, an ANVTPSO-BPNN fault diagnosis model based on adaptive no velocity term particle swarm optimization (ANVTPSO) is proposed to optimize the threshold and weight parameters of BP neural network (BPNN). In the composite fault mode of rolling bearing, a SOM-BPNN network model combining self-organizing maps (SOM) and BPNN is proposed. ANVTPSO is used to optimize the threshold and weight parameters of SOM-BPNN, and ANVTPSO-SOM-BPNN fault diagnosis model is constructed. The performance of ANVTPSO-BPNN and ANVTPSO-SOM-BPNN models is verified by experiments. The results show that under the single and composite fault mode of rolling bearing, the diagnostic accuracy of both models was significantly improved compared with BPNN.

Multi-vibration information fusion diagnosis study. Compared with a single fault, the situation of rolling bearing rotor composite fault is more complex. Using a single sensor signal is easy to cause the difficulty of fault feature extraction and accurate fault diagnosis. Two ways of dual acceleration signal and combined acceleration displacement signal are proposed for different mode composite fault diagnosis. ANVTPSO-SOM-BPNN method is used for fault diagnosis of each single signal, and then the decision-making layer fusion of diagnosis results is carried out through Dempster-Shafer (D-S) evidence theory, so as to realize the reliable diagnosis of rolling bearing rotor composite fault. The effect of the rolling bearing rotor composite fault diagnosis method based on multi vibration information fusion is verified by experiments. The results show that the diagnosis accuracy after fusion is 98.28% in the acceleration combined displacement signal mode, which is 6.86% higher than that in the dual acceleration signal mode.

[1]张迪. 滚动轴承的多点复合故障诊断研究[D]. 大连：大连理工大学，2019.

[2]Gałęzia A, Gryllias K. Application of the Combined Teager-Kaiser Envelope for bearing fault diagnosis[J]. Measurement, 2021, 182: 109710.

[3]Hoang D T, Kang H J. A survey on deep learning based bearing fault diagnosis[J]. Neurocomputing, 2019, 335: 327-335.

[4]Yu K, Lin T R, Tan J W, et al. An adaptive sensitive frequency band selection method for empirical wavelet transform and its application in bearing fault diagnosis[J]. Measurement, 2019, 134: 375-384.

[5]Shi J J, Liang M, Guan Y P. Bearing fault diagnosis under variable rotational speed via the joint application of windowed fractal dimension transform and generalized demodulation: A method free from prefiltering and resampling[J]. Mechanical Systems&Signal Processing, 2016, 68(6): 15-33.

[6]池永为. 滚动轴承故障的振动特性分析与智能诊断方法研究[D]. 杭州：浙江大学，2018.

[7]刘东东，程卫东，温伟刚，等. 基于包络解调滤波的滚动轴承复合故障诊断[J]. 中南大学学报(自然科学版)，2018, 49(4): 881-887.

[8]崔玲丽，吴春光，邬娜. 基于EMD与ICA的滚动轴承复合故障诊断[J]. 北京工业大学学报，2014, 40(10): 1459-1464.

[9]Silva J L H, Cardoso A J M. Bearing failures diagnosis in three-phase induction motors by extended Park's vector approach[C]. Industrial Electronics Society, 32nd Annual Conference of IEEE, 2005: 2585-2590.

[10]Padinharu D K K, Bandaru V K, Nilajkar V S. Diagnosis of bearing thermal anomalies in an electrical machine[P]. US: US20120109569, 2012.05.03.

[11]李凤林. 基于改进快速峭度图的高速列车滚动轴承复合故障诊断[D]. 成都：西南交通大学，2019.

[12]Chen K, Li X B, Wang F, et al. Bearing fault diagnosis using Wavelet analysis[C]. Quality, Reliability, Risk, Maintenance, and Safety Engineering (ICQR2MSE), 2012 International Conference on. 2012: 699-702.

[13]于江林，戴光，王美波，等. 滚动轴承非接触式声学故障诊断方法研究及应用[J]. 声学技术，2009, 28(1): 34-38.

[14]Vashisht R K, Peng Q. Crack detection in the rotor ball bearing system using switching control strategy and Short Time Fourier Transform[J]. Journal of Sound and Vibration, 2018, 432: 502-529.

[15]杨伟康. 基于时频分析的直升机传动系统齿轮箱故障诊断软件开发[D]. 长沙：湖南大学，2014.

[16]杨贤昭. 基于经验模态分解的故障诊断方法研究[D]. 武汉：武汉科技大学，2012.

[17]吴雅朋，王吉芳，徐小力，等. 基于小波分析的FastICA联合降噪方法在滚动轴承故障诊断中的应用研究[J]. 中国机械工程，2017, 28(18): 2183-2188+2197.

[18]马朝永，盛志鹏，胥永刚，等. 基于自适应频率切片小波变换的滚动轴承故障诊断[J]. 农业工程学报，2019, 35(10): 34-41.

[19]李志农，刘跃凡，胡志峰，等. 经验小波变换-同步提取及其在滚动轴承故障诊断中的应用[J]. 振动工程学报，2021, 34(6): 1284-1292.

[20]Liu T T, Sang Y. Fault diagnosis of rolling bearing based on empirical wavelet transform and fuzzy function[J]. IOP Conference Series: Materials Science and Engineering, 2020, 768(5): 052065.

[21]He S L, Chen J L, Zhou Z T, et al. Multifractal entropy based adaptive multiwavelet construction and its application for mechanical compound-fault diagnosis[J]. Mechanical Systems and Signal Processing, 2016, 76-77: 742-758.

[22]Wang H C. A frequency slice wavelet transform based on wavelet de-noising using neighboring coefficients method and its application in feature extraction of rolling bearing early weak fault[J]. Journal of Vibroengineering, 2020, 22(2): 383-392.

[23]Pan R, Xu T, Liu Y. The Fault Diagnosis Method for Roller Bearing with Harmonic Wavelet Packet and DT-SVM[J]. Applied Mechanics and Materials, 2015, 3764(724): 238-241.

[24]任学平，王朝阁，张玉皓，等. 基于双树复小波包自适应Teager能量谱的滚动轴承早期故障诊断[J].振动与冲击，2017, 36(10): 84-92.

[25]马本栋，胡书举，孟岩峰，等. 基于多小波包-能量算子解调的滚动轴承复合故障诊断[J]. 噪声与振动控制，2018, 38(4): 160-164.

[26]彭畅，柏林，刘小峰. 基于鲁棒性小波包峭度图的滚动轴承故障诊断[J]. 振动.测试与诊断，2016, 36(1): 11-16+194.

[27]Venkatasubramanian V , Chan K. A neural network methodology for process fault diagnosis[J]. Aiche Journal, 1989, 35(12): 1993-2002.

[28]吕楠，姚平喜. 基于BP神经网络的滚动轴承故障诊断[J]. 煤矿机械，2020, 41(8): 172-173.

[29]Liu J F, Yu X. Rolling Element Bearing Fault Diagnosis for Complex Equipment based on MFMD and BP Neural Network[J]. Journal of Physics Conference Series, 2021, 1948(1): 012113.

[30]黄文超，王林军，刘晋玮，等. 基于多特征融合与GA-BP模型的滚动轴承故障识别[J]. 机床与液压，2021, 49(06): 170-173+89.

[31]鞠晨，张超，樊红卫，等. 基于小波包分解和PSO-BPNN的滚动轴承故障诊断[J]. 工矿自动化，2020, 46(8): 70-74.

[32]王丽，王晓凯. 一种非线性改变惯性权重的粒子群算法[J]. 计算机工程与应用，2007, 43(4): 47-48+92.

[33]朱雅敏，薛鹏翔. 基于学习因子自适应改变的粒子群算法研究[J]. 陕西科技大学学报(自然科学版)，2015, 33(4): 172-177.

[34]李国华，付振芳，曾璇，等. EEMD联合SOM的电机滚动轴承故障诊断[J]. 噪声与振动控制，2020, 40(4): 87-91.

[35]Yao X Q, Tang G, Hu X. Research on Fault Diagnosis of Rolling Bearing Based on SOM Neural Network[J]. International Core Journal of Engineering, 2019, 5(9): 159-166.

[36]王海林，张春光，唐超尘，等. 基于小波分析和RBF神经网络的轴承故障诊断研究（英文）[J]. 机床与液压，2020, 48(24): 182-187.

[37]Ding S, Chang X H, Wu Q H. Fault Diagnosis of Induction Motors Based on RBF Neural Network[J]. Applied Mechanics and Materials, 2013, 462-463: 85-88.

[38]Abbasion S, Rafsanjani A, Farshidianfar A, et al. Rolling element bearings multi-fault classification based on the wavelet denoising and support vector machine[J]. Mechanical Systems & Signal Processing, 2007, 21(7): 2933-2945.

[39]Fan H W, Xue C Y, Zhang X H, et al. Vibration Images-Driven Fault Diagnosis Based on CNN and Transfer Learning of Rolling Bearing under Strong Noise[J]. Shock and Vibration, 2021, 2021: 1-16.

[40]胡晓依，荆云建，宋志坤，等. 基于CNN-SVM的深度卷积神经网络轴承故障识别研究[J]. 振动与冲击，2019, 38(18): 173-178.

[41]石江波，杨兆建，郭伟杰，等. 电流与振动信号融合的转子系统故障诊断研究[J]. 机械设计与制造，2018, (8): 19-21+25.

[42]刘希亮，陈桂明，李方溪，等. 基于多传感器信息融合的贝叶斯网络故障诊断方法研究及应用[J]. 机械科学与技术，2013, 32(1): 91-95.

[43]杨永旭，陈旭辉. 模糊集理论在多传感器信息融合中的应用[J]. 计算机应用与软件，2011, 28(11): 122-124.

[44]Lei Y G, Lin J, He Z J, et al. A method based on multi-sensor data fusion for fault detection of planetary gearboxes[J]. Sensors, 2012, 12(2): 2005-17.

[45]袁小宏，赵仲生，屈梁生. 粗糙集理论在机械故障诊断中的应用研究[J]. 西安交通大学学报，2001, 35(9): 954-958.

[46]Shafer G. A mathematical theory of evidence[M]. Princeton: Princeton university press, 1976.

[47]王远霏，裴春兴，孙海荣. 基于加权改进D-S证据融合理论的地铁车辆转向架轴承故障诊断方法[J]. 北京交通大学学报，2018, 42(6): 75-82+90.

[48]Khazaee M, Ahmadi H, Omid M, et al. Classifier fusion of vibration and acoustic signals for fault diagnosis and classification of planetary gears based on Dempster-Shafer evidence theory[J]. Proceedings of the Institution of Mechanical Engineers, Part E: Journal of Process Mechanical Engineering, 2014, 228(1): 21-32.

[49]孙伟超，李文海，李文峰. 融合粗糙集与D-S证据理论的航空装备故障诊断[J]. 北京航空航天大学学报，2015, 41(10): 1902-1909.

[50]Sun R, Huang H Z, Miao Q. Improved information fusion approach based on DS evidence theory[J]. Journal of mechanical science and technology, 2008, 22(12): 2417-2425.

[51]谭青，向阳辉. 加权证据理论信息融合方法在故障诊断中的应用[J]. 振动与冲击，2008, 27(4): 112-116.

[52]Li S B, Liu G K, Tang X H, et al. An ensemble deep convolutional neural betwork model with improved D-S evidence fusion for Bearing fault diagnosis[J]. Sensors, 2017, 17(8): 1-19.

[53]王亚涛. 变转速工况下的滚动轴承复合故障诊断方法研究[D]. 北京：北京化工大学，2020.

[54]邓飞跃. 滚动轴承故障特征提取与诊断方法研究[D]. 北京：华北电力大学，2016.

[55]刘飞. 基于解卷积方法的风电机组滚动轴承复合故障诊断研究[D]. 呼和浩特：内蒙古工业大学，2019.

[56]付士鹏. 基于LabVIEW的旋转机械振动监测与故障诊断的研究[D]. 保定：华北电力大学，2015.

[57]张俊. 基于小波分形和神经网络的滚动轴承故障诊断[D]. 北京：北京交通大学，2009.

[58]姜小荧. 基于小波分析的滚动轴承故障诊断方法的研究及应用[D]. 大连：大连理工大学，2005.

[59]臧玉萍，张德江，王维正. 小波分层阈值降噪法及其在发动机振动信号分析中的应用[J]. 振动与冲击，2009, 28(8): 57-60+198-199.

[60]Liu X L, Liu H G, Yang J H, et al. Improving the bearing fault diagnosis efficiency by the adaptive stochastic resonance in a new nonlinear system[J]. Mechanical Systems & Signal Processing, 2017, 96: 58-76.

[61]王普，李天垚，高学金，等. 分层自适应小波阈值轴承故障信号降噪方法[J]. 振动工程学报，2019, 32(03): 548-556.

[62]张鑫媛. 基于数据驱动的采煤机摇臂传动系统故障诊断方法研究[D]. 西安：西安科技大学，2019.

[63]李东钰，田慕琴，宋建成，等. 基于最优小波基选取的掘进机振动信号去噪方法[J]. 工矿自动化，2016, 42(10): 35-39.

[64]王婉秋，杨松. 结构故障诊断中的小波分解层数选取方法研究[J]. 强度与环境，2009, 36(4): 34-41.

[65]刘朦月. 基于振动信号的电机轴承故障诊断方法研究[D]. 哈尔滨：哈尔滨工业大学，2016.

[66]田慕玲，王晓玲. 电机故障诊断中的小波分析方法及小波基选取[J]. 煤矿机械，2007, 28(5): 176-178.

[67]郑小霞，钱轶群，王帅. 基于优选小波包与马氏距离的滚动轴承性能退化GRU预测[J]. 振动与冲击，2020, 39(17): 39-46+63.

[68]张玲玲，赵懿冠，肖云魁，等. 基于小波包-AR谱的变速器轴承故障特征提取[J]. 振动.测试与诊断，2011, 31(4): 492-495+537.

[69]Kankar P K, Sharma S C, Harsha S P. Fault diagnosis of rolling element bearing using cyclic autocorrelation and wavelet transform[J]. Neurocomputing, 2013, 110(8): 9-17.

[70]杨禄. 基于优化BP神经网络的遥感影像荒漠化分类与变化监测研究[D]. 桂林：桂林理工大学，2019.

[71]刘天舒. BP神经网络的改进研究及应用[D]. 哈尔滨：东北农业大学，2011.

[72]胡旺，李志蜀. 一种更简化而高效的粒子群优化算法[J]. 软件学报，2007, 18(4): 861-868.

[73]魏星，舒乃秋，崔鹏程，等. 基于改进PSO-BP神经网络和D-S证据理论的大型变压器故障综合诊断[J]. 电力系统自动化，2006, 30(7): 46-50.

[74]温正，孙华克. MATLAB智能算法[M]. 北京：清华大学出版社，2017.

[75]仝秋娟，赵岂，李萌. 基于自适应动态改变的粒子群优化算法[J]. 微电子学与计算机，2019, 36(2): 6-10+15.

[76]董健，钦文雯，李莹娟，等. 基于改进反向传播神经网络代理模型的快速多目标天线设计[J]. 电子与信息学报，2018, 40(11): 2712-2719.

[77]周晔锡. 基于SOM-BP串联神经网络的中速磁浮列车悬浮控制器故障诊断方法研究[D]. 北京：北京交通大学，2019.

[78]陈海周. 旋转机械传动件微弱故障融合诊断方法研究[D]. 重庆：重庆大学，2017.

[79]李善，谭继文，俞昆. 神经网络和改进D-S证据理论相结合的滚动轴承复合故障诊断研究[J]. 机床与液压，2018, 46(1): 153-157+184.

[80]Li Y B, Chen J , Ye F, et al. The Improvement of DS Evidence Theory and Its Application in IR/MMW Target Recognition[J]. Journal of Sensors, 2016, (6): 1-15.

[81]Wang P, Shang C X, Han Z Z. Target identification based on evidence theory of improved basic probability assignment [M]. Electronics, Communications and Networks IV. 2015: 1171-1175.

[82]Fan H W, Yan Y, Zhang X H, et al. Composite fault diagnosis of rolling bearing based on optimized wavelet packet AR spectrum energy entropy combined with Adaptive No Velocity Term PSO-SOM-BPNN[J]. Journal of Sensors, 2021, 2021: 1-15.