基于灰色模型与LSTM网络的旋转机械轴承寿命预测

doi:10.12305/j.issn.1001-506X.2021.08.39

摘要/Abstract

摘要：

大型机械设备中旋转机械占到总量的80%, 为及时掌握其工作状态, 开展如何旋转机械轴承的寿命预测精度的仿真研究。首先, 通过可靠性数值(confidential value, CV)量化评估工作状态; 然后, 利用数据变换和累加积分的方法优化数据平滑性与背景值来改进灰色模型; 并与长短时记忆网络结合为新预测模型来预测系统工作状态; 最后, 将平均绝对百分比误差等3种性能指标与单一模型对比, 将预测失效时刻与全卷积层神经网络算法和无迹粒子滤波算法对比。结果表明, 组合模型预测退化趋势3种指标的平均值优于3种单一模型; 组合模型预测的失效时刻相比于另外两种改进算法更准确。

关键词: 旋转机械, 轴承, 寿命预测, 预测精度, 灰色模型, 长短时记忆网络

Abstract:

Rotating machinery account for 80 percent of the total large mechanical equipment. In order to grasp the working condition in time, the simulation research about how to improve the accuracy of life prediction for the bearings in rotating machineries is carried out. Firstly, the confidential value (CV) is used to quantify the evaluation of the working status. Then, the method of data transformation and cumulative integration is used to optimize the smoothness of the data and the background value to improve the grey model. And the long-short term memory network (LSTM) is combined with a new prediction model to predict the working state of bearings. Finally, the average absolute percentage error and other two performance indicators are compared with the single models, and the predicted failure time is compared with fully convolutional layer neural network algorithm and unscented particle filter algorithm. The results show that the average value of the three indicators for predicting the degradation trend of the combined model is better than that of the three single models. The failure time predicted by the combined model is more accurate than the two improved algorithms.

Key words: rotating machinery, bearing, life prediction, prediction accuracy, grey model, long-short term memory (LSTM) network

中图分类号:

TH17

舒涛, 张一弛, 丁日显. 基于灰色模型与LSTM网络的旋转机械轴承寿命预测[J]. 系统工程与电子技术, 2021, 43(8): 2355-2361.

Tao SHU, Yichi ZHANG, Rixian DING. Life prediction of bearings in rotating machinery based on grey model and LSTM network[J]. Systems Engineering and Electronics, 2021, 43(8): 2355-2361.

图/表 10

图1

图2

图3

图4

表1

表2

图5

图6

表3

表4

参考文献 18

1	任利娟. 滚动轴承性能退化评估与剩余寿命预测[D]. 济南: 山东大学, 2019.
	REN L J. Rolling bearing performance degradation assessment and remaining life prediction[D]. Jinan: Shandong university, 2019.
2	郭锐, 赵之谦, 贾鑫龙, 等. 基于ANFIS的外啮合齿轮泵寿命预测研究[J]. 仪器仪表学报, 2020, 41 (1): 223- 232.
	GUO R , ZHAO Z Q , JIA X L , et al. External gear pump life prediction based on ANFIS study[J]. Journal of Instruments and Meters, 2020, 41 (1): 223- 232.
3	黎慧, 张国文. 基于灰色模型的滚动轴承剩余寿命预测[J]. 机械设计与研究, 2018, 34 (1): 113- 116, 120.
	LI H , ZHANG G W . Based on grey model to predict the residual life of the rolling bearing[J]. Journal of Mechanical Design and Research, 2018, 6 (1): 113- 116, 120.
4	ALI B J , CHEBEL M B , SAIDI L , et al. Accurate bearing remaining useful life prediction based on Weibull distribution and artificial neural network[J]. Mechanical Systems and Signal Processing, 2015, 56 (8): 150- 172.
5	石慧, 曾建潮. 考虑突变状态检测的齿轮实时剩余寿命预测[J]. 振动与冲击, 2017, 36 (21): 173- 184.
	SHI H , ZENG J C . Considering gear real-time residual life prediction of mutation state detection[J]. Journal of Vibration and Shock, 2017, 36 (21): 173- 184.
6	MOHSEN M , MOHAMMAD J S . Bearing remaining useful life prediction under starved lubricating condition using time domain acoustic emission signal processing[J]. Expert Systems with Applications, 2021, 30 (18): 168- 170.
7	KHAZAEE M , BANAKAR A , GHOBADIAN B , et al. Remaining useful life (RUL) prediction of internal combustion engine timing belt based on vibration signals and artificial neural network[J]. Neural Computing and Applications, 2020, 20 (4): 16- 20. doi: 10.1007/s00521-020-05520-3
8	张继冬, 邹益胜, 邓佳林, 等. 基于全卷积层神经网络的轴承剩余寿命预测[J]. 中国机械工程, 2019, 30 (18): 2231- 2235.
	ZHANG J D , ZOU Y S , DENG J L , et al. Layer based on the convolution of the neural network bearing the residual life prediction[J]. China Mechanical Engineering, 2019, 30 (18): 2231- 2235.
9	文娟, 高宏力. 一种基于UPF的轴承剩余寿命预测方法[J]. 振动与冲击, 2018, 37 (24): 208- 213, 243.
	WEN J , GAO H L . A bearing residual life prediction method based on UPF[J]. Journal of Vibration and Shock, 2018, 37 (24): 208- 213, 243.
10	刘颉. 基于振动信号分析的旋转机械故障诊断方法研究[D]. 武汉: 华中科技大学, 2018.
	LIU J. Based on the analysis of vibration signal of rotating machinery fault diagnosis method research[D]. Wuhan: Huazhong University of Science and Technology, 2018.
11	申中杰, 陈雪峰, 何正嘉, 等. 基于相对特征和多变量支持向量机的滚动轴承剩余寿命预测[J]. 机械工程学报, 2016, 49 (2): 183- 189.
	SHEN Z J , CHEN X F , HE Z J , et al. Characteristics and multivariate support vector machine (SVM) based on relative residual life prediction of the rolling bearing[J]. Journal of Mechanical Engineering, 2016, 49 (2): 183- 189.
12	ALI B , MORELLO C . Accurate bearing remaining useful life prediction based on Weibull distribution and artificial neural network[J]. Mechanical Systems & Signal Processing, 2016, 56 (3): 150- 172.
13	张亢, 程军圣, 杨宇. 基于局部均值分解与形态学分形维数的滚动轴承故障诊断方法[J]. 振动与冲击, 2013, 32 (9): 90- 94.
	ZHANG H , CHENG J S , YANG Y . Credits fractal dimension based on local mean decomposition and form of the rolling bearing fault diagnosis methods[J]. Journal of Vibration and Shock, 2013, 32 (9): 90- 94.
14	周建民, 黎慧, 张龙, 等. 基于EMD和逻辑回归的轴承性能退化评估[J]. 机械设计与研究, 2016, 32 (5): 72- 75, 79.
	ZHOU J M , LI H , ZHANG L , et al. Bearing performance degradation assessment based on EMD and logistic regression[J]. Journal of Mechanical Design and Research, 2016, 32 (5): 72- 75, 79.
15	钟鑫, 刘文彬, 杨剑锋. 基于逻辑回归的滚动轴承性能退化评估[J]. 科技信息, 2010, 18 (16): 504- 505.
	ZHONG X , LIU W B , YANG J F . Rolling bearing perfor-mance degradation assessment based on logistic regression[J]. Journal of Information Science and Technology, 2010, 18 (16): 504- 505.
16	Bearing data set[EB/OL]. NASA Ames Prognostics data repository[2020-11-04]. http://ti.arc.nasa.gov/tech/dash/poe/prognostic-data-repository.
17	DING N , LI H L , YIN Z W , et al. Journal bearing seizure degradation assessment and remaining useful life prediction based on long short-term memory neural network[J]. Measurement, 2020, 16 (4): 126- 129.
18	SAMUEL V , CHARLES D , JALAL F , et al. The degrees of freedom of partly smooth regularizers[J]. Annals of the Institute of Statistical Mathematics, 2017, 69 (4): 26- 27. doi: 10.1007/s10463-016-0563-z

隐层神经元数量/个	训练轮次/轮	训练数据个数/个	x_MAPE
1	700	420	0.025 3
2	700	420	0.024 4
3	700	420	0.021 8
4	700	420	0.019 9
5	700	420	0.023 1

组别	IGM-LSTM模型			GM(1, 1)模型			IGM(1, 1)模型			LSTM网络模型
组别	x_RMSE	x_MAPE/%	P	x_RMSE	x_MAPE/%	P	x_RMSE	x_MAPE/%	P	x_RMSE	x_MAPE/%	P
701~720	0.020	4.100	0.959	0.068	8.933	0.910	0.026	5.031	0.950	0.023	3.682	0.963
721~740	0.017	4.035	0.960	0.048	7.042	0.929	0.024	5.222	0.948	0.025	3.901	0.961
741~760	0.019	3.781	0.962	0.044	7.563	0.924	0.030	4.184	0.958	0.022	3.632	0.964
761~780	0.006	2.992	0.970	0.059	6.054	0.940	0.022	3.964	0.960	0.018	3.053	0.970
781~800	0.013	2.451	0.976	0.102	9.220	0.910	0.054	5.895	0.941	0.029	4.184	0.958
801~820	0.015	2.867	0.971	0.086	6.763	0.932	0.032	3.313	0.967	0.015	1.999	0.980
821~840	0.017	2.720	0.973	0.112	10.012	0.900	0.014	2.750	0.973	0.028	4.024	0.960
841~860	0.033	3.011	0.970	0.133	11.232	0.888	0.026	3.013	0.970	0.032	3.452	0.966
861~880	0.024	2.053	0.980	0.097	9.441	0.910	0.034	3.562	0.964	0.028	3.563	0.964
881~900	0.016	2.007	0.980	0.085	9.061	0.909	0.027	4.767	0.952	0.031	3.791	0.962
901~920	0.020	1.799	0.982	0.069	8.550	0.915	0.018	2.948	0.971	0.027	3.460	0.965
921~940	0.025	1.391	0.986	0.072	8.354	0.917	0.023	3.754	0.963	0.020	2.772	0.972
941~960	0.014	1.074	0.989	0.090	9.132	0.910	0.015	2.083	0.979	0.032	2.981	0.970
961~984	0.012	1.152	0.989	0.093	7.420	0.926	0.023	2.232	0.978	0.022	2.472	0.975
AVG	0.018	2.531	0.975	0.083	8.481	0.915	0.026	3.760	0.962	0.023	3.354	0.964

阶段	全卷积层神经网络		UPF		IGM-LSTM
阶段	x_MAPE/%	P	x_MAPE/%	P	x_MAPE/%	P
961~968	2.214	0.978	1.883	0.981	1.496	0.985
969~976	6.598	0.934	1.080	0.990	1.205	0.988
977~984	4.732	0.953	3.412	0.966	0.637	0.994
AVG	4.435	0.955	2.078	0.979	1.152	0.989

真实结果	预测结果
真实结果	IGM-LSTM	全卷积层神经网络	UPF
9 645	9 621	9 486	9 586

[1]	吉瑞萍, 张程祎, 梁彦, 王跃东. 基于LSTM的弹道导弹主动段轨迹预报[J]. 系统工程与电子技术, 2022, 44(6): 1968-1976.
[2]	侯召国, 王华伟, 周良, 付强. 基于改进深度残差网络的旋转机械故障诊断[J]. 系统工程与电子技术, 2022, 44(6): 2051-2059.
[3]	孙世岩, 张钢, 梁伟阁, 佘博, 田福庆. 基于时间序列数据扩增和BLSTM的滚动轴承剩余寿命预测方法[J]. 系统工程与电子技术, 2022, 44(3): 1060-1068.
[4]	陈云翔, 饶益, 蔡忠义, 王泽洲. 基于改进相似性的装备部件剩余寿命预测及经济性储备策略[J]. 系统工程与电子技术, 2021, 43(9): 2688-2696.
[5]	宋兆理, 贾祥, 郭波, 程志君. 基于贝叶斯融合与仿真的系统剩余寿命预测[J]. 系统工程与电子技术, 2021, 43(6): 1706-1713.
[6]	赵洪利, 陈天铭, 郑涅. 基于综合指标多阶段相似的发动机寿命预测[J]. 系统工程与电子技术, 2021, 43(5): 1430-1436.
[7]	赵俊龙, 李伟, 王泓霖, 黄腾, 甘奕夫, 王也. 基于长短时记忆网络的雷达波形设计[J]. 系统工程与电子技术, 2021, 43(2): 376-382.
[8]	王泽洲, 陈云翔, 蔡忠义, 项华春, 王莉莉. 基于比例关系加速退化建模的设备剩余寿命在线预测[J]. 系统工程与电子技术, 2021, 43(2): 584-592.
[9]	张一凡, 张双辉, 刘永祥, 荆锋. 基于注意力机制的堆叠LSTM网络雷达HRRP序列目标识别方法[J]. 系统工程与电子技术, 2021, 43(10): 2775-2781.
[10]	刘凯, 张斌, 黄青华. 基于TCNN-BiLSTM网络的调制识别算法[J]. 系统工程与电子技术, 2020, 42(8): 1841-1849.
[11]	李京峰, 陈云翔, 项华春, 蔡忠义. 基于LSTM-DBN的航空发动机剩余寿命预测[J]. 系统工程与电子技术, 2020, 42(7): 1637-1644.
[12]	蔡忠义, 王泽洲, 张晓丰, 李岩. 隐含非线性退化设备的剩余寿命在线预测方法[J]. 系统工程与电子技术, 2020, 42(6): 1410-1416.
[13]	张奡, 王治华, 赵振平, 秦续国, 吴琼, 刘成瑞. 两阶段非线性Wiener过程退化建模与可靠性分析[J]. 系统工程与电子技术, 2020, 42(4): 954-959.
[14]	陈闯, 陆宁云, 姜斌, 邢尹. 单部件加速退化系统的视情维修策略优化[J]. 系统工程与电子技术, 2020, 42(3): 613-619.
[15]	黄魁, 苏春. 基于灰色神经网络组合模型的故障预测[J]. 系统工程与电子技术, 2020, 42(1): 238-244.