基于深度迁移学习的电梯故障诊断研究

近年来，随着社会的不断进步以及城市现代化的飞速发展，高层建筑不断增加，电梯已成为了城市中不可缺少的垂直交通工具。电梯的安全维护工作也随之变得至关重要。目前，电梯的维护管理模式主要是人工定期维保，但由于电梯数量非常巨大，传统的人工维保不仅会耗费大量的运维成本，而且容易错过电梯故障维修的最佳时间节点，无法及时排除安全隐患。针对这些问题，本文提出了一种基于ALSTM-SEFCN神经网络的深度迁移学习电梯故障诊断方法，旨在实现电梯信号的实时监测及故障诊断，提升电梯运行的安全可靠性。

本文提出的电梯故障诊断方法不需要复杂的数据预处理操作和过多的专家知识，通过非侵入式传感器采集电梯的多维电流信号作为数据集，借助ALSTM-SEFCN神经网络提取信号样本的时序故障特征，以有监督学习的方式实现了电梯的故障检测与识别。本文以采集的真实电梯数据作为实验数据集，将本文提出的方法与其他网络模型进行了比较，证实了该电梯故障诊断方法的有效性。为了观察各层网络的特征提取能力，理解模型的分类过程，采用t-SNE方法对模型的各层输出进行了可视化分析。

针对传统机器学习方法样本需求量大，而实际的电梯故障诊断问题中异常训练样本缺乏导致模型性能降低的问题，本文将深度迁移学习方法引入到模型中，建立了基于深度迁移学习的电梯故障诊断方法。本文收集了不同规格和模式的电梯数据作为源数据集和目标数据集，针对不同的迁移学习场景，比较了不同迁移策略下模型的故障诊断性能，为实际场景下的迁移学习策略选择提供了参考。

为了进一步验证本文提出的基于深度迁移学习的故障诊断方法的实际价值，利用该方法设计了电梯故障双层诊断模型，结合滑动窗口的数据实时切分方法，搭建了电梯实时状态监测故障诊断模型系统并进行了模型部署上线测试，该系统实现了电梯实时状态信息显示、故障诊断以及故障信息统计等功能。

In recent years, with the continuous progress of society and the rapid development of urban modernization, urban population density is increasing day by day, which leads to the rapid increase of high-rise buildings and elevators. Nowadays, elevators have become an indispensable vertical transportation in cities. Therefore, the safety maintenance of elevators has become very important. At present, manual maintenance is the mainly maintenance and management mode of elevators. However, due to the huge number of elevators, traditional manual maintenance will not only consume a lot of maintenance costs, but also easily miss the best time node for elevator fault maintenance, resulting in failure to eliminate potential safety hazards in time. In response to these problems, this paper proposes a deep transfer learning elevator fault diagnosis method based on ALSTM-SEFCN neural network, which aims to realize real-time monitoring and fault diagnosis of elevator signals and improve the safety and reliability of elevator operation.

The elevator fault diagnosis method proposed in this paper does not require complex data preprocessing operations and excessive expert knowledge. The multi-dimensional current signal of the elevator is collected as a data set through non-invasive sensors, and the time series fault characteristics of the signal samples are extracted with the help of ALSTM-SEFCN neural network, to realize the fault detection and identification of elevators in a supervised learning manner. In this paper, we collected real elevator data as the experimental data set. The method proposed in this paper is compared with other network models, and the results confirm the effectiveness of the elevator fault diagnosis method. In order to observe the feature extraction ability of each layer of the network and understand the classification process of the model, the t-SNE method is used to visually analyze the output of each layer of the model.

In view of the problem that traditional machine learning methods require a large number of samples, and the lack of abnormal training samples in the actual elevator fault diagnosis problem can lead to model performance drops, this paper introduces the deep transfer learning method into the model, and establishes an elevator fault diagnosis method based on deep transfer learning. This paper collects elevator data of different specifications and modes as the source data sets and target data sets. For different transfer learning scenarios, the fault diagnosis performance of the models under different transfer strategies is compared, which provides a reference for the selection of transfer learning strategies in practical scenarios.

In order to further verify the practical value of the fault diagnosis method based on deep transfer learning proposed in this paper, using this method combine with the real-time data segmentation method of sliding window to design a double-layer real-time status monitoring and fault diagnosis system for elevators. The system has carried out the model deployment and online test, and realized the functions of elevator real-time status information display, fault diagnosis and fault information statistics.

[1] 2019-2025 年中国电梯行业市场及竞争发展趋势研究报告[R].深圳市盛世华研企业管理有限公司,2019.
[2] 国家质检总局.质检总局关于 2016 年全国特种设备安全状况情况的通报[J].中国特种设备安全,2017,33(4).
[3] ISERMANN R. Process fault detection based on modeling and estimation methods—A survey [J]. Automatica, 1984, 20(4): 387-404.
[4] ISERMANN R. Parameter adaptive control systems:-a review on methods and application [J]. IFAC Proceedings Volume s, 1985, 18(5): 373-8.
[5] BEARD R V. Failure accomodation in linear systems through self￾reorganization [D]; Massachusetts Institute of Technology, 1971.
[6] 姜新通,刘钊铭,陈言,等.基于参数估计算法的异步电机定子电流故障诊断[J].电气自动化,2019,5.
[7] 魏婧雯.储能锂电池系统状态估计与热故障诊断研究[D].合肥:中国科学技术大学,2019.
[8] CHOW E Y, LOU X-C, VERGHESE G C, et al. Redundancy relations and robust failure detection [M]. Detection of Abrupt Changes in Signals and Dynamical Systems. Springer. 1985: 275 -93.
[9] 郭玉英,朱正为,靳玉红.基于等价空间算法的飞控系统故障诊断[J].西南科技大学学报,2007,22(2):48-51.
[10] 刘平华.飞机发动机故障诊断专家系统研究[D].青岛:青岛科技大学,2017.
[11] 赵磊磊.基于专家系统的船舶电力系统故障诊断方法研究[D].镇江:江苏科技大学,2018.
[12] ZHI-LING Y, BIN W, XING-HUI D, et al. Expert system of fault diagnosis for gear box in wind turbine [J]. Syst ems Engineering Procedia, 2012, 4: 189 -95.
[13] MA D, LIANG Y, ZHAO X, et al. Multi-BP expert system for fault diagnosis of powersystem [J]. Engineering Applications of Artificial Intelligence, 2013, 26(3): 937-44.
[14] LEE W-S, GROSH D L, TILLMAN F A, et al. Fault tree analysis, methods, and applications ߝ a review [J]. IEEE Transactions on Reliability, 1985, 34(3): 194-203.
[15] 谈 娟.基于解析模型的飞控系统故障诊断技术研究[D].南 京:南 京 航 空 航 天 大学,2020.58参考文献
[16] 王俊芳.基于模糊神经网络的信息融合在配电网故障诊断中的应用[D];重庆:重庆大学,2013.
[17] 李心一,谢志江,罗久飞.加窗插值快速傅里叶变换在滚动轴承故障诊断中的应用[J].中国机械工程,2018,29(10):1166-72.
[18] 李志农,朱明,褚福磊,等.基于经验小波变换的机械故障诊断方法研究[J].仪器仪表学报,2014,35(11):2423-32.
[19] 张承彪,罗运柏,文习山.主成分分析在变压器故障诊断中的应用研究[J].高电压技术,2005,(08):9-11.
[20] ZHAO Y, ZHOU M, XU X, et al. Fault diagnosis based on space mapping and deformable convolution networks [J]. IEEE Access, 2020, 8: 212599 -607.
[21] GAO Y, WANG L, LIU Y, et al. Design of elevator fault diagnosis cloud system based on smart phones; Proceedings of the 2019 Chinese Control and Decision Conference (CCDC), F, 2019 [C]. IEEE.
[22] LI W, HUANG R, LI J, et al. A perspective survey on deep transfer learning for fault diagnosis in industrial scenarios: Theories, applications and challenges [J]. Mechanical Systems and Signal Processing, 2022, 167: 108487.
[23] HABIC V, SEMENOV A, PASILIAO E L. Multitask deep learning for native language identification [J]. Knowledge -Based Systems, 2020, 209: 106440.
[24] ZHOU D, MA S, HAO J, et al. An electricity load forecasting model for Integrated Energy System based on BiGAN and transfer lear ning [J]. Energy Reports, 2020, 6: 3446 -61.
[25] GOPALAKRISHNAN K, KHAITAN S K, CHOUDHARY A, et al. Deep convolutional neural networks with transfer learning for computer vision -based data-driven pavement distress detection [J]. Construction and Building Materials, 2017, 157: 322-30.
[26] FATEH A, FATEH M, ABOLGHASEMI V. Multilingual handwritten numeral recognition using a robust deep network joint with transfer learning [J]. Information Sciences, 2021, 581: 479 -94.
[27] 段礼祥,谢骏遥,王凯,等.基于不同工况下辅助数据集的齿轮箱故障诊断[J].振动与冲击,2017,36(10):104-8.
[28] 陈仁祥,陈思杨,胡小林,等.源域多样本集成GFK的不同工况下滚动轴承寿命状态识别[J].振动工程学报,2020.
[29] 沈飞,陈超,严如强.奇异值分解与迁移学习在电机故障诊断中的应用[J].振动工程学报,2017,30(1):118-26.
[30] SHAO S, MCALEER S, YAN R, et al. Highly accurate machine fault diagnosis using deep transfer learning [J]. IEEE Transactions on Industrial Informatics, 2018, 15(4): 2446-55.59参考文献
[31] LIU J, ZHANG Q, LI X, et al. Transfer learning -based strategies for fault diagnosis in building energy systems [J]. Energy and Buildings, 2021, 250: 111256.
[32] HE Z, SHAO H, ZHANG X, et al. Improved deep transfer auto -encoder for fault diagnosis of gearbox under variable working conditions with small training samples [J]. IEEE Access, 2019, 7: 115368-77.
[33] HOPFIELD J. Neural networks and physical systems with emergent collective computational abilities [J]. Proceedings of the National Academy of Sciences, 1982, 79(8): 2554-8.
[34] JORDAN M I. Serial order: A parallel distributed processing approach [M]. Advances in Psychology. Elsevier. 1997: 471 -95.
[35] ELMAN J L. Finding structure in time [J]. Cognitive Science, 1990, 14(2): 179-211.
[36] BENGIO Y, SIMARD P, FRASCONI P. Learning long -term dependencies with gradient descent is difficult [J]. IEEE Transactions on Neural Networks, 1994, 5(2): 157-66.
[37] ABBASIMEHR H, SHABANI M, YOUSEFI M. An optimized model using LSTM network for demand forecasting [J]. Computers & Industrial Engineering, 2020, 143: 106435.
[38] SU C, HUANG H, SHI S, et al. Neural machine translation with Gumbel Tree -LSTM based encoder [J]. Journal of Visual Communication and Image Representation, 2020, 71: 102811.
[39] GAO D, ZHU Y, REN Z, et al. A novel weak fault diagnosis method for rolling bearings based on LSTM considering quasi-periodicity [J]. Knowledge -Based Systems, 2021, 231: 10 7413.
[40] BAHDANAU D, CHO K, BENGIO Y. Neural machine translation by jointly learning to align and translate [J]. arXiv preprint arXiv:14090473, 2014.
[41] FUKUSHIMA K. Neocognitron: A hierarchical neural network capable of visual pattern recognition [J]. Neural Networks, 1988, 1(2): 119-30.
[42] LECUN Y, BOTTOU L, BENGIO Y, et al. Gradient-based learning applied to document recognition [J]. Proceedings of the IEEE, 1998, 86(11): 2278 -324.
[43] SZEGEDY C, LIU W, JIA Y, et al. Going deeper with convolutions; Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, F, 2015 [C].
[44] HU J, SHEN L, SUN G. Squeeze -and-excitation networks; Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, F, 2018 [C].
[45] 姚晨.基于深度学习的风机故障诊断与预警[D];北京:华北电力大学(北京),2021.60参考文献
[46] KALMAN R E. A new approach to linear filtering and prediction problems [J]. 1960.
[47] KING G, ZENG L. Logistic regression in rare events data [J]. Political Analysis, 2001, 9(2): 137-63.
[48] VAN DER MAATEN L. Accelerating t-SNE using tree-based algorithms [J]. The Journal of Machine Learning Research, 2014, 15(1): 3221-45.
[49] PEZZOTTI N, LELIEVELDT B P, VAN DER MAATEN L, et al. Approximated and user steerable tSNE for progressive visual analytics [J]. IEEE Transactions on Visualization and Computer Graphics, 2016, 23(7): 1739-52.
[50] WEISS K, KHOSHGOFTAAR T M, WANG D. A survey of transfer learning [J]. Journal of Big Data, 2016, 3(1): 1-40.
[51] SUKHIJA S, KRISHNAN N C, SINGH G. Supervised heterogeneous domain adaptation via random forests; Proceedings of the IJCAI, F, 2016 [C].