With the proliferation of massive state monitoring data, the health status monitoring of critical components in complex electromagnetic energy equipment demands increasingly stringent requirements for real-time and reliable performance. Edge intelligence technology, dedicated to decentralizing intelligent algorithms and computational resources to the device end, presents a promising approach to addressing the challenges of fault diagnosis in such systems. This paper introduces a novel lightweight fault diagnosis method leveraging edge intelligence, which processes data near its source, mitigating issues related to limited resources in industrial embedded systems and the delays caused by massive data transmission.
At the heart of our proposed method is the application of compressive sensing during data acquisition. This technique effectively compresses dense, multivariate monitoring data into a sparse form, significantly reducing data volume while preserving critical fault-related information. The subsequent integration of deep extreme learning machines (DELM) and kernel functions allows for a profound exploration of the relationships between the compressed data and potential faults. The model is then refined through lightweight techniques, making it suitable for deployment on edge computing devices, thus minimizing the demands on data transmission, computation, and storage resources.
The effectiveness of our method is empirically validated through rigorous testing on an EdgeBoard AI computing platform, a device designed to emulate real-world industrial conditions. The results are compelling: the method achieves a diagnostic accuracy exceeding 99%, with a diagnosis time well within the millisecond range, underscoring its potential for real-time industrial applications. A detailed analysis of the impact of compression ratio on diagnostic efficiency reveals that an 80% compression rate offers the optimal balance between accuracy and speed. Moreover, the incorporation of particle swarm optimization (PSO) into the model further enhances its performance by dynamically optimizing key parameters, leading to superior generalization and classification accuracy.
The robustness of our method is further demonstrated through its application on a fault data simulation platform, where it is compared against state-of-the-art fault diagnosis techniques. The results indicate that our method not only requires less input data dimensionality but also significantly outperforms other methods in terms of diagnostic time and accuracy. The use of temporal-spatial features, extracted through a combination of convolutional neural networks (CNN) for spatial learning and simple recurrent units (SRU) for temporal dynamics, contributes to the model's high performance. The SRU, in particular, stands out for its ability to mitigate the gradient disappearance problem and reduce computational costs compared to traditional recurrent neural networks like LSTM and GRU. The model's resilience to input data noise, as evidenced by the inclusion of Gaussian noise in the tests, further attests to its robustness in practical scenarios.
The study concludes that the edge intelligence-based fault diagnosis method is not only a significant advancement over existing methods in terms of real-time processing and diagnostic precision but also substantially alleviates the computational and transmission burdens associated with fault diagnosis. The successful validation on both the EdgeBoard AI platform and the fault data simulation platform confirms the method's practical applicability and its potential to revolutionize the field of intelligent fault diagnosis for electromagnetic energy equipment.
[1] 马伟明. 关于电工学科前沿技术发展的若干思考[J]. 电工技术学报, 2021, 36(22): 4627-4636.
Ma Weiming.Thoughts on the development of frontier technology in electrical engineering[J]. Transactions of China Electrotechnical Society, 2021, 36(22): 4627-4636.
[2] Tidriri K, Chatti N, Verron S, et al.Bridging data-driven and model-based approaches for process fault diagnosis and health monitoring: a review of researches and future challenges[J]. Annual Reviews in Control, 2016, 42: 63-81.
[3] 任翔宇, 秦勇, 王彪, 等. 基于博弈映射学习的多传感源信息融合三相电机智能故障诊断方法[J]. 电工技术学报, 2023, 38(17): 4633-4645.
Ren Xiangyu, Qin Yong, Wang Biao, et al.An intelligent multi-sensor information fusion fault diagnosis method of three-phase motors based on game mapping learning[J]. Transactions of China Electrotechnical Society, 2023, 38(17): 4633-4645.
[4] 李世晓, 杜锦华, 龙云. 基于一维卷积神经网络的机电作动器故障诊断[J]. 电工技术学报, 2022, 37(增刊1): 62-73.
Li Shixiao, Du Jinhua, Long Yun.Fault diagnosis of electromechanical actuators based on one-dimensional convolutional neural network[J]. Transactions of China Electrotechnical Society, 2022, 37(S1): 62-73.
[5] Donoho D L.Compressed sensing[J]. IEEE Transactions on Information Theory, 2006, 52(4): 1289-1306.
[6] Chen Junxin, Sun Shuang, Zhang Libo, et al.Compressed sensing framework for heart sound acquisition in Internet of medical things[J]. IEEE Transactions on Industrial Informatics, 2022, 18(3): 2000-2009.
[7] Tang Jiexiong, Deng Chenwei, Huang Guangbin.Extreme learning machine for multilayer perceptron[J]. IEEE Transactions on Neural Networks and Learning Systems, 2016, 27(4): 809-821.
[8] Deng Shuiguang, Zhao Hailiang, Fang Weijia, et al.Edge intelligence: the confluence of edge computing and artificial intelligence[J]. IEEE Internet of Things Journal, 2020, 7(8): 7457-7469.
[9] Houssein E H, Gad A G, Hussain K, et al.Major advances in particle swarm optimization: theory, analysis, and application[J]. Swarm and Evolutionary Computation, 2021, 63: 100868.
[10] 黄凯, 丁恒, 郭永芳, 等. 基于数据预处理和长短期记忆神经网络的锂离子电池寿命预测[J]. 电工技术学报, 2022, 37(15): 3753-3766.
Huang Kai, Ding Heng, Guo Yongfang, et al.Prediction of remaining useful life of lithium-ion battery based on adaptive data preprocessing and long short-term memory network[J]. Transactions of China Electrotechnical Society, 2022, 37(15): 3753-3766.
[11] 刘会兰, 许文杰, 赵书涛, 等. 面向高压断路器故障分类的电流-振动信号类聚几何敏感特征优选方法[J]. 电工技术学报, 2023, 38(1): 26-36.
Liu Huilan, Xu Wenjie, Zhao Shutao, et al.Optimization method of clustering geometric sensitive features of current vibration signals for fault classification of high voltage circuit breakers[J]. Transactions of China Electrotechnical Society, 2023, 38(1): 26-36.
[12] Cao Siyu, Zou Feng, Chen Debao, et al.A new hybrid prediction model with entropy-like kernel function for dynamic multi-objective optimization[J]. Applied Intelligence, 2023, 53(9): 10500-10519.
[13] Ghojogh B, Ghodsi A, Karray F, et al. Reproducing kernel Hilbert space, mercer’s theorem, eigenfunctions, nyström method,use of kernels in machine learning: tutorial and survey[EB/OL].2021: 2106.08443.https://arxiv.org/abs/2106.08443v1
[14] 文云峰, 赵荣臻, 肖友强, 等. 基于多层极限学习机的电力系统频率安全评估方法[J]. 电力系统自动化, 2019, 43(1): 133-140.
Wen Yunfeng, Zhao Rongzhen, Xiao Youqiang, et al.Frequency safety assessment of power system based on multi-layer extreme learning machine[J]. Automation of Electric Power Systems, 2019, 43(1): 133-140.
[15] 宋向金, 孙文举, 刘国海, 等. 深度子领域自适应网络电机滚动轴承跨工况故障诊断[J]. 电工技术学报, 2024, 39(1): 182-193.
Song Xiangjin, Sun Wenju, Liu Guohai, et al.Across working conditions fault diagnosis for motor rolling bearing based on deep subdomain adaption network[J]. Transactions of China Electrotechnical Society, 2024, 39(1): 182-193.
[16] Jia Ning, Cheng Yao, Liu Yunyang, et al.Intelligent fault diagnosis of rotating machines based on wavelet time-frequency diagram and optimized stacked denoising auto-encoder[J]. IEEE Sensors Journal, 2022, 22(17): 17139-17150.
[17] 杨洁, 万安平, 王景霖, 等. 基于多传感器融合卷积神经网络的航空发动机轴承故障诊断[J]. 中国电机工程学报, 2022, 42(13): 4933-4942.
Yang Jie, Wan Anping, Wang Jinglin, et al.Aeroengine bearing fault diagnosis based on convolutional neural network for multi-sensor information fusion[J]. Proceedings of the CSEE, 2022, 42(13): 4933-4942.
[18] Hu Zhongxu, Wang Yan, Ge Mingfeng, et al.Data-driven fault diagnosis method based on compressed sensing and improved multiscale network[J]. IEEE Transactions on Industrial Electronics, 2020, 67(4): 3216-3225.
