Common methods, such as manual inspection, image recognition, and fixed-point monitoring, struggle to comprehensively and accurately evaluate the status of the Overhead Contact System (OCS) compensation device and mid-point anchor. Therefore, a state anomaly diagnosis method based on data feature recognition is proposed. Firstly, two types of typical data features are summarized during abnormal conditions of both the compensation device and the mid-point anchor. Specifically, three feature quantities are defined to describe the Overhead Contact Line (OCL) height mutation in the mid-point anchor area, three to indicate the difference in OCL height on both sides of the anchor section, and three to describe the pantograph-catenary contact force mutation in the mid-point anchor area. Secondly, a sample library is constructed based on a large amount of detected data. Outlier diagnosis methods are employed to check the reliability of the samples, and Principal Component Analysis (PCA) is applied to reduce strongly correlated feature quantities, thereby forming a training sample library. Finally, a classification model is trained using Random Forest (RF), and the classification model is evaluated using overall accuracy, false positive rate, and true negative rate. The feature quantities, classification model, and field trial verification of the proposed method are conducted. The results indicate that each feature quantity exhibits good separability, verifying the reliability and universality of the summarized data features. RF achieves a classification accuracy of over 95% in both types of sample training. The diagnosis results obtained using the proposed method are highly accurate in field verification, demonstrating strong feasibility in diagnosing abnormal conditions of both the compensation device and the mid-point anchor.
LIUGuangdong, WUMingli, WEIMin, et al. Design and Application of Online Monitoring System for Catenary Compensation Device [J]. Electric Drive for Locomotives, 2021 (3): 140-145. in Chinese
WENChenglin, FeiyaLÜ, BAOZhejing, et al. A Review of Data Driven-Based Incipient Fault Diagnosis [J]. Acta Automatica Sinica, 2016, 42 (9): 1285-1299. in Chinese
[7]
王超.基于多源数据的异常检测关键技术研究[D].成都:电子科技大学,2019.
[8]
WANGChao. Research on Key Technologies of Anomaly Detection Based on Multi-Source Data [D]. Chengdu: University of Electronic Science and Technology of China, 2019. in Chinese
ZHANGLei. Research on Fault Diagnosis of Traction Power Supply System Based on PSO-LSSVM [J]. Electric Drive for Locomotives, 2019 (3): 51-55, 59. in Chinese
GUOBingting, CHENShuixuan, HONGZhaobin, et al. A Shallow Neural Network for Bearing Fault Diagnosis Based on Bayesian Hyperparameter Optimization [J]. Journal of Xiamen University of Technology, 2022, 30 (1): 10-15. in Chinese
LIUXiaoxiong, WUYan, SHIJingping, et al. Method of Fault Diagnosis Using Bayesian Optimization Algorithm [J]. Application Research of Computers, 2009, 26 (1): 129-131. in Chinese
YANGJianwei, CAIGuoqiang, YAODechen, et al. Fault Diagnosis Method for the Rolling Bearing of Railway Vehicle Based on Wavelet Packet Transform and BP Neural Network [J]. China Railway Science, 2010, 31 (6): 68-73. in Chinese
[17]
吉鹏霄,张桂林.电气化铁路接触网[M].3版.北京:化学工业出版社,2015:99-100.
[18]
JIPengxiao, ZHANGGuilin. Electrified Railway Overhead Contact System [M]. 3rd ed. Beijing: Chemical Industry Press Co., Ltd., 2015: 99-100. in Chinese
NIULiubin, HonglinZÜ, XUXiaodi, et al. Estimation Method and Application of Depth of Rail Corrugation Based on Vertical Vibration Acceleration of Axle Box [J]. China Railway Science, 2023, 44 (1): 25-38. in Chinese
[21]
YANH S, NIUY J, CHENL T, et al. Location Algorithm of Transfer Stations Based on Density Peak and Outlier Detection [J]. Applied Intelligence, 2022, 52 (12): 13520-13532.
[22]
EMERSONW R. Causation and Pearson's Correlation Coefficient [J]. Journal of Visual Impairment & Blindness, 2015, 109 (3): 242-244.
[23]
PARKH C, LEEG. Comparison of Incremental Linear Dimension Reduction Methods for Streaming Data [J]. Pattern Recognition Letters, 2020, 135 (8): 15-21.
[24]
杨学兵,张俊.决策树算法及其核心技术[J].计算机技术与发展,2007,17(1):43-45.
[25]
YANGXuebing, ZHANGJun. Decision Tree and Its Key Techniques [J]. Computer Technology and Development, 2007, 17 (1): 43-45. in Chinese
WANGHaiying, ZHOUWei, ZHANGWenxuan, et al. Intelligent Recognition Method for Mast Position of Overhead Contact Line [J]. China Railway Science, 2019, 40 (1): 111-116. in Chinese
[28]
LEOBreiman. Random Forests [J]. Machine Learning, 2001, 45 (1): 5-32.
[29]
SVETNIKV, LIAWA, TONGC, et al. Random Forest: a Classification and Regression Tool for Compound Classification and QSAR Modeling [J]. Journal of Chemical Information and Computer Sciences, 2003, 43 (6): 1947-1958.
[30]
AARTHYS T, MAZHER IQBALJ L. Time Series Real Time Naive Bayes Electrocardiogram Signal Classification for Efficient Disease Prediction Using Fuzzy Rules [J]. Journal of Ambient Intelligence and Humanized Computing, 2021, 12 (5): 5257-5267.
[31]
杜坤.基于弓网接触力及接触线高度制定线索张力的评价依据[D].成都:西南交通大学,2021.
[32]
DUKun. Evaluation Basis of the Catenary Tension Based on the Contact Force and the Height of Contact Line [D]. Chengdu: Southwest Jiaotong University, 2021. in Chinese
[33]
JUWARAL, SAHA-CHAUDHURIP. A Hybrid Covariate Micro Aggregation Approach for Privacy-Preserving Logistic Regression [J]. Journal of Survey Statistics and Methodology, 2022, 10 (3): 568-595.
SONGJindong, ZHUJingbao, LIShanyou. Continuous Prediction Method of Earthquake Early Warning Magnitude for High-Speed Railway Based on Support Vector Machine [J]. China Railway Science, 2021, 42 (3): 156-165. in Chinese
[36]
CAIY D, HUANGJ Z, YINJ F. A New Method to Build the Adaptive K-Nearest Neighbors Similarity Graph Matrix for Spectral Clustering [J]. Neurocomputing, 2022 (493): 191-203.
[37]
ALIAKBARI SANIS, KHORRAMA, JAFFARIA, et al. Development of Processing Map for InX-750 Superalloy Using Hyperbolic Sinus Equation and ANN Model [J]. Rare Metals, 2021, 40 (12): 3598-3607.