基于贝叶斯优化的SWDAE-LSTM滚动轴承早期故障预测方法研究

PDF(2232 KB)

振动与冲击 ›› 2021, Vol. 40 ›› Issue (18) : 286-297.

论文

石怀涛1，尚亚俊1，白晓天1，郭磊1，马辉2

作者信息 +

Early fault prediction method combining SWDAE and LSTM for rolling bearings based on Bayesian optimization

SHI Huaitao1, SHANG Yajun1, BAI Xiaotian1, GUO Lei1, MA Hui2

Author information +

文章历史 +

摘要

针对滚动轴承的早期故障特征较弱，在强噪声背景下难以有效提取以致生命周期很难准确预测的问题，提出了一种基于贝叶斯优化（BO）的滑动窗堆叠去噪自编码器（SWDAE）和长短期记忆（LSTM）网络的早期故障预测模型。使用滑动窗算法保留具有非线性特征和时序特征的历史正常数据，输入到模型中进行训练，使模型学习滚动轴承的正常运行状态趋势。将滚动轴承运行的数据输入到训练好的SWDAE-LSTM模型中进行实时在线监控，利用模型的预测值与真实值的残差检测滚动轴承早期故障。针对模型超参数组合选择困难的问题，使用贝叶斯优化算法对模型的超参数进行调优。最后，使用美国辛辛那提大学智能维护中心(IMSCenter)的轴承全生命周期数据以及机械故障综合模拟实验装置获取的数据进行仿真实验验证。结果表明，使用贝叶斯优化算法进行智能调参的模型和基于时域指标的方法对比，可以更早的有效检测出滚动轴承的早期故障并具有很强的鲁棒性。与其余深度学习方法比较，其模型的诊断准确率高于其他方法，进一步证明了其有效性和可靠性。

Abstract

The early fault characteristics of rolling bearings are usually rather weak and it is difficult to effectively extract under the background of strong noise so that the life cycle is difficult to accurately predict.Aiming at this,an initial fault prediction model based on Bayesian optimization(BO) consisting of sliding window stacked denoising auto encoder (SWDAE) and long short-term memory(LSTM) network was proposed.The sliding window algorithm was used to retain historical normal data with non-linear characteristics and time series characteristics, so,when the processed data were input into the model for training, the model was able to learn the normal running state trend of the rolling bearing.The data of the rolling bearing operation were then input into the trained SWDAE-LSTM model for real-time online monitoring, and the residual of the predicted value and the true value of the model was used to detect the early failure of the rolling bearing.Aiming at the difficulty in selecting the hyperparameter combination of the model, a Bayesian optimization algorithm was used to tune the hyperparameter of the model.Finally, the bearing full life cycle data from the University of Cincinnati Intelligent Maintenance Center (IMSCenter) and the data from a mechanical failure integrated simulation experiment device were used for simulation verification.The results show that the model of intelligent parameter adjustment using Bayesian optimization algorithm and the method based on time domain index can detect the early fault of the rolling bearing effectively and have strong robustness.Compared with other deep learning methods, the diagnostic accuracy of the model is higher than that of other methods, which further proves its validity and reliability.

导出引用

石怀涛，尚亚俊，白晓天，郭磊，马辉. 基于贝叶斯优化的SWDAE-LSTM滚动轴承早期故障预测方法研究[J]. 振动与冲击, 2021, 40(18): 286-297

SHI Huaita, SHANG Yajun, BAI Xiaotian, GUO Lei, MA Hui. Early fault prediction method combining SWDAE and LSTM for rolling bearings based on Bayesian optimization[J]. Journal of Vibration and Shock, 2021, 40(18): 286-297

参考文献

［1］SANZ J, PERERA R, HUERTA C.Fault diagnosis of rotating machinery based on auto-associative neural networks and wavelet transforms［J］.Journal of Sound and Vibration, 2007, 302(4/5)：981-999.
［2］SHI H T, BAI X T.Model-based uneven loading condition monitoring of full ceramic ball bearings in starved lubrication［J］.Mechanical Systems and Signal Processing, 2020, 139：106583.
［3］YU K, LIN T R, MA H, et al.A combined polynomial chirplet transform and synchroextracting technique for analyzing nonstationary signals of rotating machinery［J］.IEEE Transactions on Instrumentation and Measurement, 2019,69(4):1505-1518.
［4］SONG L Y, WANG H Q, CHEN P.Step-by-step fuzzy diagnosis method for equipmentbased on symptom extraction and trivalent logic fuzzy diagnosis theory［J］ IEEE Transactions on Fuzzy Systems, 2018,26:3467-3478.
［5］CHEN P, SONG L Y, WANG H Q.Vibration-based intelligent fault diagnosis for rollerbearings in low-speed rotating machinery［J］.IEEE Transactions on Instrumentation and Measurement ，2018,67(8):1887-1899.
［6］HAO Y S, SONG L Y, CUI L L, et al.A three-dimensional geometric features-based SCA algorithm for compound faults diagnosis［J］.Measurement,2019, 134:480-491.
［7］CUI L L, LI B B, MA J F,et al.Quantitative trend fault diagnosis of a rolling bearing，basedon Sparsogram and Lempel-Ziv［J］.Measurement，2018，128:410-418.
［8］YANG X Q, DING K, HE G L, et al.Double-dictionary signal decomposition method basedon split augmented Lagrangian shrinkage algorithm and its application ingearbox hybrid faults diagnosis［J］.Journal of Sound and Vibration，2018, 432:484-501.
［9］CUI L L, ZHANG Y, ZHANG F B,et al.Vibration response mechanism offaulty outer race rolling element bearings for quantitative analysis［J］.Journal of Sound and Vibration， 2016,364:67-76.
［10］CHEN Z Q, LI C, SANCHEZ R V.Gearbox fault identification and classification with convolutional neural networks［J］.Shock and Vibration, 2015(2):1-10.
［11］袁云龙.基于峭度-小波包分析的滚动轴承故障诊断［J］.新技术新工艺, 2008(5)：43-46.
YUAN Yunlong.Fault diagnosis of rolling bearing based on kurtosis-wavelet packet analysis［J］.New Technology & New Process, 2008(5)：43-46.
［12］李恒，张氢，秦仙蓉，等.基于短时傅里叶变换和卷积神经网络的轴承故障诊断方法［J］.振动与冲击， 2018, 37(19)：124-131.
LI Heng, ZHANG Qing, QIAN Xianrong, et al.Fault diagnosis method for rolling bearings based on short-time Fourier transform and convolution neural network［J］.Journal of Vibration and Shock, 2018, 37(19)：124-131.
［13］ZHANG X Y, LIANG Y T, ZHOU J Z, et al.A novel bearing fault diagnosis model integrated permutation entropy, ensemble empirical mode decomposition and optimized SVM［J］.Measurement, 2015, 69：164-179.
［14］QU Z D, HAGHANI P, WEINSTEIN E, et al.Syllable-based acoustic modeling with CTC-SMBR-LSTM［C］∥IEEE Automatic Speech Recognition and Understanding Workshop.Okinawa：IEEE, 2017.
［15］SHI H T, GUO L, TAN S, et al.Rolling bearing initial fault detection using long short-term memory recurrent network［J］.IEEE Access, 2019, 7：171559-171569.
［16］杜小磊, 陈志刚, 许旭, 等.基于小波卷积自编码器和 LSTM 网络的轴承故障诊断研究［J］.机电工程，2019, 36(7)：663-668.
DU Xiaolei, CHEN Zhigang, XU Xu, et al.Research on bearing fault diagnosis based on wavelet convolution autoencoder and LSTM network［J］.Journal of Mechanical & Electrical Engineering, 2019, 36(7)：663-668.
［17］赵建鹏，周俊.基于长短时记忆网络的旋转机械状态预测研究［J］.噪声与振动控制，2017, 37(4)：155-159.
ZHAO Jianpeng, ZHOU Jun.State prognosis of rotary machines based on long/short term memory recurrent neural network ［J］.Noise and Vibration Control, 2017, 37(4)：155-159.
［18］SHI H T, BAI X T, ZHANG K, et al.Influence of uneven loading condition on the sound radiation of starved lubricated full ceramic ball bearings［J］.Journal of Sound and Vibration, 2019, 461：114910.
［19］YOUNG S R,ROSE D C,KARNOWSKI T P,et al.Optimizing deep learning hyper-parameters through an evolutionary algorithm［C］∥The International Conference for High Performance Computing,Networking,Storage and Analysis.Austin：Association for Computing Machinery, 2015.
［20］HUANG Q J, MAO J L, LIU Y.An improved grid search algorithm of SVR parameters optimization［C］∥IEEE 14th International Conference on Communication Technology.Chengdu：IEEE, 2012.
［21］QU J, CHEN H Y, LIU W Z, et al.Application of support vector machine based on improved grid search in quantitative analysis of gas［J］.Chinese Journal of Sensors & Actuators, 2015, 28(5)：774-778.
［22］BERGSTRA J, BENGIO Y.Random search for hyper-parameter optimization［J］.Journal of Machine Learning Research, 2012, 13(2)：281-305.
［23］YAN F L, LIN Z B, WANG X Y, et al.Evaluation and prediction of bond strength of GFRP-bar reinforced concrete using artificial neural network optimized with genetic algorithm［J］.Composite Structures, 2017, 161：441-452.
［24］LI Y, LU G, ZHOU L H, et al.Quantum inspired high dimensional hyperparameter optimization of machine learning model［C］∥IEEE 2017 International Smart Cities Conference (ISC2).Wuxi：IEEE, 2017.
［25］李甜甜.基于改进粒子群算法的超参数优化问题的研究［D］.西安：西安电子科技大学, 2019.
［26］CHOI D, CHO H, RHEE W.On the difficulty of DNN hyperparameter optimization using learning curve prediction［C］∥TENCON 2018-2018 IEEE Region 10 Conference.Jeju：IEEE, 2018.
［27］崔佳旭, 杨博.贝叶斯优化方法和应用综述［J］.软件学报, 2018, 29(10)：3068-3090.
CUI Jiaxu, YANG Bo.Survey on Bayesian optimization methodology and applications［J］.Journal of Software, 2018, 29(10)：3068-3090.