Edge-side Fault Diagnosis of Wind Turbine Gearboxes by Low-pass Screening Neural Architecture Search

doi:10.3901/JME.2025.07.361

Abstract

Abstract: Edge-side fault diagnosis requires lightweight deep models. They are, typically, empirically handcrafted by experts, which is time-consuming and labor-intensive. The configurable resource capacity for the edge hardware is not considered in manual lightweight models; therefore, they may not meet deployment requirements. Here, a method based on a low-pass screening neural architecture search is proposed. Fault-diagnosis models are automatically designed for edge hardware considering the hardware configurable resource capacity. First, an empirically inspired search space is designed to reduce the search difficulty in lightweight models. Meanwhile, a low-pass screening reward function is modeled to guide an agent iteratively screening lightweight diagnostic models meeting the hardware configurable resource capacity condition during the search process. Finally, Pareto-optimal domination is used to obtain a competitive Pareto-optimal solution set, providing an optimal model for the edge hardware to achieve fault diagnosis of wind-turbine gearboxes. The feasibility and effectiveness of the method were verified on a gearbox’s test from a drivetrain diagnostics simulator and on measured wind-farm case. The results indicate that the searched models are superior to the competing models in terms of accuracy, FLOPs, and parameters. Particularly, in the application case, LSNAS-Netb achieved 3.06% and 3.65% higher accuracies compared with the deep model GoogLeNet-v1 and edge-side-friendly MobileNet-v2, respectively, with 15.56×and 6.19×fewer parameters and 5.47×and 1.18×fewer FLOPs, respectively.

Key words: edge-side fault diagnosis, neural architecture search, low-pass screening, lightweight models, wind-turbine gearboxes

CLC Number:

TH17

WU Yanling, TANG Baoping, DENG Lei, FU Hao. Edge-side Fault Diagnosis of Wind Turbine Gearboxes by Low-pass Screening Neural Architecture Search[J]. Journal of Mechanical Engineering, 2025, 61(7): 361-372.

References

[1] 金晓航，孙毅，单继宏，等. 风力发电机组故障诊断与预测技术研究综述[J]. 仪器仪表学报，2017，38(5)：1041-1053. JIN Xiaohang，SUN Yi，SHAN Jihong，et al. Fault diagnosis and prognosis for wind turbines：An overview[J]. Chinese Journal of Scientific Instrument，2017，38(5)：1041-1053.
[2] WANG Tianyang，HAN Qinkai，CHU Fulei，et al. Vibration based condition monitoring and fault diagnosis of wind turbine planetary gearbox:：A review[J]. Mechanical Systems and Signal Processing，2019，126：662-685.
[3] LIU Zhiliang，WANG Huan，LIU Junjie，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.
[4] LIAO Yabin，RAGAI I，HUANG Ziyun，et al. Manufacturing process monitoring using time-frequency representation and transfer learning of deep neural networks[J]. Journal of Manufacturing Processes，2021，68(8)：231-248.
[5] DENG Jin，JIANG Wenjuan，ZHANG Ye，et al. HS-KDNet：A lightweight network based on hierarchical-split block and knowledge distillation for fault diagnosis with extremely imbalanced data[J]. IEEE Transactions on Instrumentation and Measurement，2021，70：1–9.
[6] 孟子尧，谷雪，梁艳春，等. 深度神经架构搜索综述[J].计算机研究与发展，2021，58(1)：22-33. MENG Ziyao，GU Xue，LIANG Chunyan，et al. Deep neural architecture search：A survey[J]. Journal of Computer Research and Development，2021，58(1): 22-33.
[7] WANG Ruixin，JIANG Hongkai，LI Xingqiu，et.al. A reinforcement neural architecture search method for rolling bearing fault diagnosis[J]. Measurement，2020，154：107417.
[8] WU Yanling，TANG Baoping，DENG Lei，et al. Distillation-enhanced fast neural architecture search method for edge-side fault diagnosis of wind turbine gearboxes[J]. Expert Systems with Applications，2022，208：118049.
[9] SUTTON R，BARTO A. Reinforcement Learning：An Introduction[M]. Second Edition. Cambridge，MA，MIT Press，2018.
[10] MA Ailong，WAN Yuting，ZHONG Yanfei，et al. SceneNet: Remote sensing scene classification deep learning network using multi-objective neural evolution architecture search[J]. ISPRS Journal of Photogrammetry and Remote Sensing，2021，172：171-188.
[11] QIU Zhengzhong，BI Wei，XU Dong，et al. Efficient self-learning evolutionary neural architecture search[J]. Applied Soft Computing，2023，146：110671.
[12] 王辉，徐佳文，严如强. 基于多尺度注意力深度强化学习网络的行星齿轮箱智能诊断方法[J]. 机械工程学报，2022，58(11)：133-142. WANG Hui，XU Jiawen，YAN Ruqiang. Multi-Scale attention based deep reinforcement learning for intelligent fault diagnosis of planetary gearbox[J]. Journal of Mechanical Engineering，2022，58(11)：133-142.
[13] ZHONG Zhao，YANG Zichen，DENG Boyang，et al. BlockQNN：Efficient block-wise neural network architecture generation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence，2021，7(43)：2314- 2328.
[14] YANG Shangshang，TIAN Ye，XIANG Xiaoshu，et al. Accelerating evolutionary neural architecture search via multi-fidelity evaluation[J]. IEEE Transactions on Cognitive and Developmental Systems，2021，4(14)： 1778-1792.
[15] PETCHROMPO S，COIT D. W，BRINTRUP A，et.al. A review of Pareto pruning methods for multi-objective optimization[J]. Computers & Industrial Engineering，2022，167：108022.
[16] 彭富强，于德介，武春燕. 基于自适应时变滤波阶比跟踪的齿轮箱故障诊断[J]. 机械工程学报，2012，48(7)：77-85. PENG Fuqiang，YU Dejie，WU Chunyan. Self-adaptively time-varying filter based order tracking method and its application in gearbox fault diagnosis[J]. Journal of Mechanical Engineering，2012，48(7)：77-85.
[17] SALAMEH J P，CAUET S，ETIEN E，et al. Gearbox condition monitoring in wind turbines：A review[J]. Mechanical Systems and Signal Processing，2018，111：251-264.
[18] ZHAO Minghang，KANG M，TANG Baoping，et al. Multiple wavelet coefficients fusion in deep residual networks for fault diagnosis[J]. IEEE Transactions on Industrial Electronics，2019，66(6)：4696-4706.
[19] BAKER B, GUPTA O, NAIK N, et al. Designing neural network architectures using reinforcement learning[J]. ArXiv，2017，1-18.
[20] HE Kaiming，ZHANG Xiangyu，REN Shaoqing，et al. Delving deep into rectifiers: surpassing human-level performance on imagenet classification[C]//IEEE International Conference on Computer Vision，2015：1026-1034.
[21] CHAKRABORTY S，SHAMRAT F. M. J. M，AFRIN S，et al. Industrial fault detection using transfer learning models[C]//International Conference on Smart Electronics and Communication，2021，Trichy，India.
[22] LI Xiaohui，WAN Shaoke，LIU Shijie，et al. Bearing fault diagnosis method based on attention mechanism and multilayer fusion network[J]. ISA Transactions，2022，128：550-564.
[23] LI Xueyi，KONG Xiangwei，ZHANG Jiqiang，et al. A Study on fault diagnosis of bearing pitting under different speed condition based on an improved inception capsule network[J]. Measurement，2021，181(2)：109656.
[24] WEN Long，LI Xinyu，GAO Liang. A transfer convolutional neural network for fault diagnosis based on ResNet-50[J]. Neural Computing & Applications，2020，32(10)：6111-6124.
[25] WU Jie，TANG Tang，CHEN Ming，et al. A study on adaptation lightweight architecture based deep learning models for bearing fault diagnosis under varying working conditions[J]. Expert Systems with Applications，2020，160(1)：113710.
[26] SANDLER M，HOWARD A，ZHU Menglong，et al. MobileNetV2：Inverted residuals and linear bottlenecks[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition，2018，Salt Lake City，UT，USA. 2018：4510-4520.
[27] WANG Yanxin，YAN Jing，LIU Tingliang. Comprehensive intelligent diagnosis for mechanical and insulation faults of power equipment in the power internet of things context[C]//2020 IEEE International Conference on High Voltage Engineering and Application ，2020，Beijing, China. 2020：1-4.
[28] PEZZOTTI N,LELIEVELDT B,LAURENS V,et al. Approximated and user steerable tSNE for progressive visual analytics[J]. IEEE Transactions on Visualization and Computer Graphics，2018，23(7)：1739-1752.