液压泵和滚动轴承多种样本量的改进多任务故障诊断

PDF(2025 KB)

振动与冲击 ›› 2024, Vol. 43 ›› Issue (4) : 270-278.

论文

郑直1，曾魁魁1，何玉灵2，李克1，王志军1

作者信息 +

Improved multi-task fault diagnosis of hydraulic pump and rolling bearing with multiple sample sizes

ZHENG Zhi1, ZENG Kuikui1, HE Yuling2, LI Ke1, WANG Zhijun1

Author information +

文章历史 +

摘要

基于充足样本的多个设备元件导致多任务学习网络规模庞大，轻微和严重的跨元件零样本问题难度大。在多种样本量（充足样本和零样本）下，针对基于充足故障样本的多元件诊断网络规模过于庞大问题，引入MicroNet方法对多任务学习网络进行轻量化处理，然后利用热重启余弦退火算法优化上述网络，提出一种多任务轻量化学习网络模型，改善多任务学习网络的准确率和效率。针对更高难度的跨元件零样本问题，引入元学习方法进一步改进上述MT-MN-CA，进而提出一种改进多任务轻量化学习网络模型，解决轻微和严重的跨元件零样本问题。通过实测液压泵和滚动轴承故障验证所提两个网络模型的有效性和优越性，实验结果表明所提网络具有很高的实时性和准确率。

Abstract

The multiple equipment components based on sufficient samples result in a large scale of multi-task learning network, and the slight and serious cross-component zero-sample problem has not been studied and is difficult. Under multiple sample sizes (sufficient samples and zero-sample), aiming at the problem that the scale of multi-component diagnosis network based on sufficient fault samples is too large, the MicroNet method is introduced to lighten the MTL network, and then the network is optimized by cosine annealing with warm restart algorithm, the lightweight MTL network model is proposed, thus the accuracy and efficiency of MTL network are improved. Aiming at the more difficult cross-component zero-sample problem, the meta-learning method is introduced to further improve the above MT-MN-CA, and then the improved lightweight MTL network model is proposed to solve the slight and serious cross-component zero-sample problem. The effectiveness and superiority of the proposed two network models are verified by the measured faults of hydraulic pump and rolling bearing, the experimental results show that the proposed network has high real-time performance and accuracy.

导出引用

郑直1，曾魁魁1，何玉灵2，李克1，王志军1. 液压泵和滚动轴承多种样本量的改进多任务故障诊断[J]. 振动与冲击, 2024, 43(4): 270-278

ZHENG Zhi1, ZENG Kuikui1, HE Yuling2, LI Ke1, WANG Zhijun1. Improved multi-task fault diagnosis of hydraulic pump and rolling bearing with multiple sample sizes[J]. Journal of Vibration and Shock, 2024, 43(4): 270-278

参考文献

[1] HE Y, TANG H S, REN Y, et al. A deep multi-signal fusion adversarial model based transfer learning and residual network for axial piston pump fault diagnosis[J]. Measurement, 2022, 192: 110889. [2] 姜万录, 赵亚鹏, 张淑清, 等. 精细复合多尺度波动散布熵在液压泵故障诊断中的应用[J]. 振动与冲击, 2022, 41(08): 7-16. Jang Wanlu, Zhao Yapeng, Zhang Shuqing, et al. Application of refined composite multiscale fluctuation dispersion entropy in hydraulic pumps fault diagnosis[J]. Journal of Shock and Vibration, 2022, 41(08): 7-16. [3] KONG Y, CHU F L, QIN Z Y, et al. Sparse learning based classification framework for planetary bearing health diagnostics[J]. Mechanism and Machine Theory, 2022, 173: 104852. [4] 刘恒畅, 姚德臣, 杨建伟, 等. 基于多分支深度可分离卷积神经网络的滚动轴承故障诊断研究[J]. 振动与冲击, 2021, 40(10): 95-102. Liu Hengchang, Yao Dechen, Yang Jianwei, et al. Fault diagnosis of rolling bearings based on a multi branch depth separable convolutional neural network[J]. Journal of Shock and Vibration, 2021, 40(10): 95-102. [5] XIE Z L, CHEN J L, FENG Y, et al. End to end multi-task learning with attention for multi-objective fault diagnosis under small sample [J]. Journal of Manufacturing Systems, 2022, 62: 301-316. [6] 赵晓平, 吴家新, 钱承山, 等. 基于多任务深度学习的齿轮箱多故障诊断方法[J]. 振动与冲击, 2019, 38(23): 271-278. ZHAO Xiaoping, WU Jiaxin, QIAN Chengshan, et al. Multi-fault diagnosis for gearboxes based on multi-task deep learning[J]. Journal of Shock and Vibration, 2019, 38(23): 271-278. [7] LIU Z L, WANG H, LIU J J, et al. Multi-task learning based on lightweight 1DCNN for fault diagnosis of wheelset bearings[J]. IEEE Transactions on Instrumentation and Measurement, 2021, 70: 1-11. [8] GUO S, ZHANG B, YANG T, et al. Multi-task convolutional neural network with information fusion for bearing fault diagnosis and localization[J]. IEEE Transactions on Industrial Electronics, 2020, 67(1): 8005-8015. [9] YU W B, LV P. An end-to-end intelligent fault diagnosis application for rolling bearing based on MobileNet[J]. IEEE Access, 2021, 9: 41925-41933. [10] YAO D C, LIU H C, YANG J W, et al, A lightweight neural network with strong robustness for bearing fault diagnosis[J]. Measurement, 2020, 159: 107756. [11] YU Y, ZHAO J, TANG T, et al. Wasserstein distance-based asymmetric adversarial domain adaptation in intelligent bearing fault diagnosis[J]. Measurement Science and Technology, 2021, 32(11): 115019. [12] YAO D C, LI G Y, LIU H C, et al. An intelligent method of roller bearing fault diagnosis and fault characteristic frequency visualization based on improved MobileNetV3[J]. Measurement Science and Technology, 2021, 32: 124009. [13] CHEN Z Y, WU J, DENG C, et al. Deep Attention Relation Network: A Zero-Shot Learning Method for Bearing Fault Diagnosis Under Unknown Domains[J]. IEEE Transactions on Reliability, 2022: 1-11. [14] LV H X, CHEN J L, PAN T Y, et al. Hybrid attribute conditional adversarial denoising autoencoder for zero-shot classification of mechanical intelligent fault diagnosis[J]. Applied Soft Computing Journal, 2020, 95: 106577. [15] GAO Y P, GAO L, LI X Y, et al. A zero-shot learning method for fault diagnosis under unknown working loads[J]. Journal of Intelligent Manufacturing, 2020, 12(31): 899-909. [16] XU J, ZHOU L, ZHAO W H, et al. Zero-shot learning for compound fault diagnosis of bearings[J]. Expert Systems with Applications, 2022, 190: 116197. [17] LI Y S, CHEN Y P, DAI X Y, et al. MicroNet: Improving image recognition with extremely low flops[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021: 468-477. [18] LOSHCHILOV I, HUTTER F. Decoupled weight decay regularization[C]//Proceedings of the International Conference on Learning Representations, 2019. [19] FINN C, ABBEEL P, LEVINE S. Model-agnostic meta-learning for fast adaptation of deep networks[C]//Proceedings of the International Conference on Machine Learning, 2017: 1126-1135.