Railway Operation Interference Identification and Data Reconstruction Based on Observation Data from Geomagnetic Stations—Taking the Lushi Geomagnetic Station as an Example
XIE Jiaxing
HOU Bowen
ZHANG Hanbo
CHENG Na
QIN Pu
LIANG Xiangdong
Abstract：As Haoji Railway is only 130 m away from Lushi Geomagnetic Station,the geomagnetic observation data of Lushi Station has been seriously disturbed during construction and operation,resulting in unavailable data.The purpose of this paper is to conduct in-depth analysis on the typical events that caused the change of the second sampled geomagnetic data of Lushi Station during the operation of Haoji Railway.Based on the analysis of the characteristics of the time domain and frequency domain of the geomagnetic data of Lushi Station,the FFT observation anomaly recognition method based on the triple mean square error principle is proposed to distinguish the observation anomaly caused by the railway operation from the normal magnetic disturbance for identification.The actual reconstruction effect of spatial weighting method,BP neural network algorithm and XGBoost strong machine learning algorithm in data processing is verified by data reconstruction for the disturbed observation data and missing data of Lushi Station.The results show that when the XGBoost strong machine learning algorithm is applied to the data reconstruction of observation anomalies and missing records.,through data simulation,the reconstructed data has a high degree of coincidence with the original data in a small time scale,achieving good results.
Keywords：Geomagnetic observationData reconstructionMachine learningSpatial weightingBP neural networkGBDT algorithm
Publication Date：2025-12-30
Online Publishing Date：2025-12-15(First online date of this platform, not the publication date of the document)
Pages：11( 17-27 )