Identification of structured features for positioning in the guideway maintenance of high-speed maglev system

The geometrical deviations of the track are crucial for safe operation and are typically manageable only at the ends of the beams. This necessitates accurately correlating the mileage information from the on-board track inspection data to the beam joints. Under high-speed (up to 500 km/h) and complex operating conditions, achieving localization without excessive reliance on the operation and control system has become a significant challenge. To address this issue, this article decomposes the localization problem of mileage data into track- magnet gap signal with structured features identification and mileage positioning problems. Among them, the signal identification task is primarily implemented by a two-stage unsupervised algorithm consisting of isolation forest and K-means algorithms, while the mileage positioning task enhances computational efficiency and accuracy through feature matching. Through testing 118 trips of track inspection data on the Shanghai maglev demonstration operation line over the past 20 years, the results show that the proposed method exhibits high accuracy and fast computational efficiency.