基于优化A-BiLSTM的滚动轴承故障诊断

doi:10.13229/j.cnki.jdxbgxb.20221339

摘要/Abstract

摘要：

为提高超参数设置的效率及其与模型的适配性，改善人工设置模型参数的高成本和低效率问题，提出一种基于蜜獾算法（Honey badger algorithm， HBA）优化注意力双向长短时记忆网络（HBA-A-BiLSTM）的滚动轴承故障诊断方法。首先，通过HBA对A-BiLSTM模型进行最优超参数组合搜寻，然后基于最优超参数下的A-BiLSTM模型进行故障诊断性能测试。最后，基于不同工况的数据集进行模型泛化能力测试。采用CWRU数据集对所提方法的故障诊断效果进行验证，利用诊断精度以及混淆矩阵进行评价。实验结果表明，与其他群智能优化算法相比，蜜獾算法搜索全局性能好，收敛速度快，优化后的最终模型的故障诊断准确率达到了99.5%，具有良好的效果，且在不同工况下能够实现稳定、准确的故障诊断性能，泛化能力强。

关键词: 故障诊断, 蜜獾算法, 参数优化, 双向长短时记忆网络, 注意力机制

Abstract:

In order to improve the efficiency of hyperparameter setting and its adaptability to the model， and break the high cost and low efficiency of manual parameter setting， a fault diagnosis method for rolling bearings based on the honey badger algorithm （HBA） optimizing bi-directional long short-term memory （BiLSTM） with attention mechanism （HBA-A-BiLSTM） is proposed. Firstly， search the optimal hyperparameter combination of the A-BiLSTM model through HBA. Secondly， the fault diagnosis performance is tested based on the A-BiLSTM model under the optimal hyperparameters. Finally， the generalization ability of the model is tested based on the datasets under different working conditions. The CWRU dataset is used to verify the fault diagnosis effect of the proposed method， which is used the diagnostic accuracy and the confusion matrix to evaluate. It is shown that， compared with other swarm intelligence optimization algorithms， the HBA has better global searching performance and faster convergence speed. The fault diagnosis accuracy of the optimized model has reached 99.5%， which has a good effect， also under different working conditions， it can achieve stable and accurate fault diagnosis performance， and has strong generalization ability.

Key words: fault diagnosis, honey badger algorithm, parameters optimization, bidirectional long short-term memory network, attention mechanism

中图分类号:

TH133.3

余萍,赵康,曹洁. 基于优化A-BiLSTM的滚动轴承故障诊断[J]. 吉林大学学报(工学版), 2024, 54(8): 2156-2166.

Ping YU,Kang ZHAO,Jie CAO. Rolling bearing fault diagnosis based on optimized A-BiLSTM[J]. Journal of Jilin University(Engineering and Technology Edition), 2024, 54(8): 2156-2166.

图/表 21

图1

图2

图3

图4

图5

图6

图7

表1

图 8

表2

图9

图10

图11

图12

表3

图13

表4

图14

图15

表5

图16

参考文献 18

1	程军圣,于德介,杨宇.基于EMD和SVM的滚动轴承故障诊断方法[J]. 航空动力学报,2006,21(3):575-580.
	Cheng Jun-sheng, Yu De-jie, Yang Yu. Fault diagnosis of roller bearings based on EMD and SVM[J]. Journal of Aerospace Power, 2006, 21(3): 575-580.
2	李红贤, 韩延, 吴敬涛, 等. 基于ICA包络增强MEMD的滚动轴承故障诊断[J]. 航空动力学报, 2021, 36(2): 405-412.
	Li Hong-xian, Han Yan, Wu Jing-tao, et al. Rolling bearing fault diagnosis based on MEMD with ICA envelop enhancement[J]. Journal of Aerospace Power, 2021, 36(2): 405-412.
3	武昆, 徐元博, 杨娜. 时变滤波经验模态分解与对称差分解析能量算子在轴承故障诊断中的应用[J]. 噪声与振动控制, 2020, 40(5): 101-107.
	Wu Kun, Xu Yuan-bo, Yang Na. Application of time-varying filtering empirical mode decomposition and symmetrical difference analytic energy operator in fault diagnosis of bearings.[J]. Noise and Vibration Control, 2020, 40(5):101-107.
4	刘泽锐, 邢济收, 王红军, 等. 基于VMD与快速谱峭度的滚动轴承故障诊断[J]. 电子测量与仪器学报, 2021, 35(2): 73-79.
	Liu Ze-rui, Xing Ji-shou, Wang Hong-jun, et al. Fault diagnosis of rolling bearings based on VMD and fast spectral kurtosis[J]. Journal of Electronic Measurement and Instrumentation, 2021, 35(2): 73-79.
5	Li Z, Wang Y, Ma J. Fault diagnosis of motor bearings based on a convolutional long short-term memory network of bayesian optimization[J]. IEEE Access, 2021, 9: 97546-97556.
6	邓飞跃, 吕浩洋, 顾晓辉, 等. 基于轻量化神经网络Shuffle⁃SENet的高速动车组轴箱轴承故障诊断方法[J]. 吉林大学学报: 工学版, 2022, 52(2): 474-482.
	Deng Fei-yue, Hao-yang Lyu, Gu Xiao-hui, et al. Fault diagnosis of high⁃speed train axle bearing based on a lightweight neural network Shuffle⁃SENet[J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(2): 474-482.
7	Tuerxun W, Chang X, Guo H Y, et al. Fault diagnosis of wind turbines based on a support vector machine optimized by the sparrow search algorithm[J]. IEEE Access, 2021, 9: 69307-69315.
8	董绍江, 朱朋, 裴雪武, 等. 基于子领域自适应的变工况下滚动轴承故障诊断[J]. 吉林大学学报: 工学版, 2022, 52(2): 288-295.
	Dong Shao-jiang, Zhu Peng, Pei Xue-wu, et al. Fault diagnosis of rolling bearing under variable operating conditions based on subdomain adaptation[J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(2): 288-295.
9	Yin A, Yan Y, Zhang Z, et al. Fault diagnosis of wind turbine gearbox based on the optimized LSTM neural network with cosine loss[J]. Sensors, 2020, 20(8): No.20082339.
10	Chen X, Zhang B, Gao D. Bearing fault diagnosis base on multi-scale CNN and LSTM model[J]. Journal of Intelligent Manufacturing, 2021, 32: 971-987.
11	杨振波, 贾民平. GA-1DLCNN 方法及其在轴承故障诊断中的应用[J]. 东南大学学报: 英文版, 2019, 2019 (1): 36-42.
	Yang Zhen-bo, Jia Min-ping. GA-1DLCNN method and its application in bearing fault diagnosis[J]. Journal of Southeast University(English Edition), 2019, 2019(1): 36-42.
12	蔡赛男, 宋卫星, 班利明, 等. 基于鲸鱼算法优化 LSSVM 的滚动轴承故障诊断[J]. 控制与决策, 2022, 37(1): 230-236.
	Cai Sai-nan, Song Wei-xing, Ban Li-ming, et al. Fault diagnosis method of rolling bearing based on LSSVM optimized by whale optimization algorithm[J]. Control and Decision, 2022, 37(1): 230-236.
13	Zheng J, Gu M, Pan H, et al. A fault classification method for rolling bearing based on multisynchrosque-ezing transform and WOA-SMM[J]. IEEE Access, 2020, 8: 215355-215364.
14	Li J, Chen W, Han K, et al. Fault diagnosis of rolling bearing based on GA-VMD and improved WOALSSVM[J]. IEEE Access, 2020, 8: 166753-166767.
15	Hashim F A, Houssein E H, Hussain K, et al. Honey badger algorithm: new metaheuristic algorithm for solving optimization problems[J]. Mathematics and Computers in Simulation, 2022, 192: 84-110.
16	Chadha G S, Panambilly A, Schwung A, et al. Bidirectional deep recurrent neural networks for process fault classification[J]. ISA Transactions, 2020, 106: 330-342.
17	Zhang Z Y, Yin A J, Tan J. Improved DBN method with attention mechanism for the fault diagnosis of gearboxes under varying working condition[J]. Journal of Vibration and Shock, 2021, 40(14): 47-52.
18	梁海涛, 王立纲, 王亮, 等. 基于ARCN模型的轴承故障诊断[J]. 航空动力学报, 2021, 36(9): 1793-1803.
	Liang Hai-tao, Wang Li-gang, Wang Liang, et al. Bearing fault diagnosis based on ARCN model[J]. Journal of Aerospace Power, 2021, 36(9): 1793-1803.

相关文章 15

[1]	郭昕刚,程超,沈紫琪. 基于卷积网络注意力机制的人脸表情识别[J]. 吉林大学学报(工学版), 2024, 54(8): 2319-2328.
[2]	王长建,刘久明,张锦洲,李斌. 基于高速摄影技术的行星减速箱故障激光序列脉冲诊断方法[J]. 吉林大学学报(工学版), 2024, 54(7): 1869-1875.
[3]	孙铭会,薛浩,金玉波,曲卫东,秦贵和. 联合时空注意力的视频显著性预测[J]. 吉林大学学报(工学版), 2024, 54(6): 1767-1776.
[4]	高云龙,任明,吴川,高文. 基于注意力机制改进的无锚框舰船检测模型[J]. 吉林大学学报(工学版), 2024, 54(5): 1407-1416.
[5]	李房云,夏容,张怡欣. 考虑电池荷电状态的混合动力汽车复合电源协同控制[J]. 吉林大学学报(工学版), 2024, 54(4): 1114-1119.
[6]	李晓旭,安文娟,武继杰,李真,张珂,马占宇. 通道注意力双线性度量网络[J]. 吉林大学学报(工学版), 2024, 54(2): 524-532.
[7]	霍光,林大为,刘元宁,朱晓冬,袁梦,盖迪. 基于多尺度特征和注意力机制的轻量级虹膜分割模型[J]. 吉林大学学报(工学版), 2023, 53(9): 2591-2600.
[8]	郭晓新,李佳慧,张宝亮. 基于高分辨率网络的视杯和视盘的联合分割[J]. 吉林大学学报(工学版), 2023, 53(8): 2350-2357.
[9]	唐菲菲,周海莲,唐天俊,朱洪洲,温永. 融合动静态变量的滑坡多步位移预测方法[J]. 吉林大学学报(工学版), 2023, 53(6): 1833-1841.
[10]	田彦涛,黄兴,卢辉遒,王凯歌,许富强. 基于注意力与深度交互的周车多模态行为轨迹预测[J]. 吉林大学学报(工学版), 2023, 53(5): 1474-1480.
[11]	于贵申,陈鑫,武子涛,陈轶雄,张冠宸. AA6061⁃T6铝薄板无针搅拌摩擦点焊接头结构及性能分析[J]. 吉林大学学报(工学版), 2023, 53(5): 1338-1344.
[12]	吕卫,韩镓泽,褚晶辉,井佩光. 基于多模态自注意力网络的视频记忆度预测[J]. 吉林大学学报(工学版), 2023, 53(4): 1211-1219.
[13]	田彦涛,许富强,王凯歌,郝子绪. 考虑周车信息的自车期望轨迹预测[J]. 吉林大学学报(工学版), 2023, 53(3): 674-681.
[14]	欧阳丹彤,孙睿,田新亮,高博涵. 基于集合阻塞的不确定系统中传感器选择方法[J]. 吉林大学学报(工学版), 2023, 53(2): 547-554.
[15]	江晟,王鹏朗,邓志吉,别一鸣. 基于深度学习的交通事故救援图像融合算法[J]. 吉林大学学报(工学版), 2023, 53(12): 3472-3480.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

故障标签	故障类别	样本长度
0	正常（Normal）	1 000
1	内圈（IR0.007）	1 000
2	滚动体（B0.007）	1 000
3	外圈（OR0.007）	1 000
4	内圈（IR0.014）	1 000
5	滚动体（B0.014）	1 000
6	外圈（OR0.014）	1 000
7	内圈（IR0.021）	1 000
8	滚动体（B0.021）	1 000
9	外圈（OR0.021）	1 000

超参数	最小值	最大值
学习率/lr	0.001	0.01
训练批量大小	1	200
迭代次数	10	200
隐层1神经元个数	20	300
隐层2神经元个数	20	300
全连接层神经元个数	30	300

模型	最优参数组合	准确率
HBA-A-BiLSTM	｛0.008，74，58，21，23，96｝	99.50%
WOA-A-BiLSTM	｛0.006，43，65，27，17，62｝	95.45%
SSA-A-BiLSTM	｛0.004，56，87，19，27，103｝	94.55%

模型	最优参数组合	准确率
HBA-A-BiLSTM	｛0.008，74，58，21，23，96｝	99.50%
HBA-BiLSTM	｛0.006，76，116，22，12，55｝	98.29%
HBA-LSTM	｛0.006，97，77，30，20，60｝	96.90%

模型	1 HP/%	2 HP/%	3 HP/%
LSTM	76.15	79.15	58.50
BiLSTM	90.20	88.00	67.80
A-LSTM	88.30	84.55	68.20
A-BiLSTM	97.15	94.95	80.10