Page 173 - 《振动工程学报》2026年第3期
P. 173
第 3 期 孟祥恒,等: 自编码器在机械设备未知故障检测中的应用 773
德国帕德伯恩大学的公开轴承数据集以及 CWRU chine monitoring data[J]. IEEE Transactions on Instru⁃
的轴承故障数据集分别进行了试验验证,通过与单 mentation and Measurement, 2021, 70: 3512512.
解码器模型及其他常见的异常检测方法对比,进一 [10] QIAN J C, SONG Z H, YAO Y, et al. A review on
autoencoder based representation learning for fault de⁃
步体现出多解码器在未知故障检测方面的优势。同
tection and diagnosis in industrial processes[J]. Chemo⁃
时构造了多个不同的数据集,在将不同故障作为未
metrics and Intelligent Laboratory Systems, 2022,
知故障的情况下,验证了本文所提方法的有效性与 231: 104711.
实用性。 [11] WU J Y, ZHAO Z B, SUN C, et al. Fault-attention
generative probabilistic adversarial autoencoder for ma⁃
参考文献: chine anomaly detection[J]. IEEE Transactions on In⁃
dustrial Informatics, 2020, 16(12): 7479-7488.
[12] CHOW J K, SU Z, WU J, et al. Anomaly detection of
[1] TUERXUN W, XU C, GUO H Y, et al. Fault diagno⁃
defects on concrete structures with the convolutional au⁃
sis of wind turbines based on a support vector machine
toencoder[J]. Advanced Engineering Informatics, 2020,
optimized by the sparrow search algorithm[J]. IEEE
45: 101105.
Access, 2021, 9: 69307-69315.
[13] ZHANG C, HU D, YANG T. Anomaly detection and
[2] LIU R N, YANG B Y, ZIO E, et al. Artificial intelli⁃
diagnosis for wind turbines using long short-term memo⁃
gence for fault diagnosis of rotating machinery: a review
ry-based stacked denoising autoencoders and XGBoost
[J]. Mechanical Systems and Signal Processing, 2018,
[J]. Reliability Engineering & System Safety, 2022,
108: 33-47.
222: 108445.
[3] LECUN Y, BENGIO Y, HINTON G. Deep learning
[14] YIN C Y, ZHANG S, WANG J, et al. Anomaly de⁃
[J]. Nature, 2015, 521(7553): 436-444.
tection based on convolutional recurrent autoencoder for
[4] JIA F, LEI Y G, LIN J, et al. Deep neural networks:
IoT time series[J]. IEEE Transactions on Systems,
a promising tool for fault characteristic mining and intel⁃
Man, and Cybernetics: Systems, 2022, 52(1): 112-122.
ligent diagnosis of rotating machinery with massive data
[15] PANZA M A, POTA M, ESPOSITO M. Anomaly
[J]. Mechanical Systems and Signal Processing, 2016,
detection methods for industrial applications: a compara⁃
72-73: 303-315.
tive study[J]. Electronics, 2023, 12(18): 3971.
[5] HAN T, ZHANG L W, YIN Z J, et al. Rolling bear⁃
[16] PUKELSHEIM F. The three sigma rule[J]. The
ing fault diagnosis with combined convolutional neural
American Statistician, 1994, 48(2): 88-91.
networks and support vector machine[J]. Measure⁃
[17] HOU Y D, WANG J J, CHEN Z Q, et al. Diagnosis⁃
ment, 2021, 177: 109022.
former: an efficient rolling bearing fault diagnosis meth⁃
[6] KHORRAM A, KHALOOEI M, REZGHI M. End-
od based on improved Transformer[J]. Engineering Ap⁃
to-end CNN+LSTM deep learning approach for bear⁃
plications of Artificial Intelligence, 2023, 124: 106507.
ing fault diagnosis[J]. Applied Intelligence, 2021, 51
[18] ZHANG W, PENG G L, LI C H, et al. A new deep
(2): 736-751.
learning model for fault diagnosis with good anti-noise
[7] PANG J X, PU X K, LI C G. A hybrid algorithm in⁃
and domain adaptation ability on raw vibration signals
corporating vector quantization and one-class support
[J]. Sensors, 2017, 17(2): 425.
vector machine for industrial anomaly detection[J].
[19] VAN DER MAATEN L, HINTON G. Visualizing da⁃
IEEE Transactions on Industrial Informatics, 2022, 18
ta using t-SNE[J]. Journal of Machine Learning Re⁃
(12): 8786-8796.
search, 2008, 9(11): 2579⁃2605.
[8] LIU F T, TING K M, ZHOU Z H. Isolation-based
anomaly detection[J]. ACM Transactions on Knowl⁃ 第一作者: 孟祥恒(2000—),男,硕士研究生。
edge Discovery from Data, 2012, 6(1): 1-39. E-mail:mengxh@bupt.edu.cn
[9] LI C G, GUO L, GAO H L, et al. Similarity-mea⁃ 通信作者: 许方敏(1982―),男,博士,副教授。
sured isolation forest: anomaly detection method for ma⁃ E-mail:xufm@bupt.edu.cn

