Track Irregularity Estimation Using Multi-Source On-Board Vibration Signals

Online high-precision measurement of track irregularities is an essential requirement for railway condition monitoring. Existing inversion and estimation methods suffer from several limitations, including the high cost of offline inspections, the cross-scale frequency mismatch between high-frequency on-board accelerations and low-frequency track irregularities, and the lack of physical constraints in purely data-driven models, restricting their application in real-time measurement scenarios. This paper proposes a track irregularity inversion method integrating kinematic physical priors, termed RCLNet. Specifically, trapezoidal numerical double integration is utilized to construct multi-scale acceleration-velocity-displacement features, mitigating the cross-scale mapping challenge from a mechanistic perspective. Furthermore, the model integrates a Temporal Convolutional Network, a Convolutional Block Attention Module, and a Bidirectional Long Short-Term Memory network to simultaneously achieve local feature extraction, key information enhancement, and long-range spatial dependency modeling. Ultimately, high-precision inversion of vertical track irregularities is achieved in both the time and spatial domains. Building upon this, a dual-dimensional systematic validation of the proposed RCLNet model is conducted by combining simulation data generated from an Abaqus-based vehicle-track coupled dynamics model with field-measured on-board vibration signals from high-speed railways. The results indicate that the coefficient of determination R² of the proposed inversion method reaches 0.94, an improvement of 16 percentage points over the baseline model with a single acceleration input. Compared with other classical models, the proposed method achieves superior inversion accuracy and real-time performance. The inference time for a single sample is merely 20 ms, demonstrating excellent robustness even under complex field conditions.

Authors

Institutions

Publication Details

Journal
Measurement Science and Technology
Published
2026-11-20
DOI
https://doi.org/10.1088/1361-6501/aec4fc
Primary Topic
Railway Engineering and Dynamics
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Track Irregularity Estimation Using Multi-Source On-Board Vibration Signals

Daojin Yao, Wentao Dong, Xiaoming Wang, Linya Liu et al.
3 links
Measurement Science and Technology
Railway Engineering and Dynamics
article

Track Irregularity Estimation Using Multi-Source On-Board Vibration Signals

Daojin Yao, Wentao Dong, Xiaoming Wang, Linya Liu, Yifeng Huang, Shengnan Wu
article en

Abstract

Online high-precision measurement of track irregularities is an essential requirement for railway condition monitoring. Existing inversion and estimation methods suffer from several limitations, including the high cost of offline inspections, the cross-scale frequency mismatch between high-frequency on-board accelerations and low-frequency track irregularities, and the lack of physical constraints in purely data-driven models, restricting their application in real-time measurement scenarios. This paper proposes a track irregularity inversion method integrating kinematic physical priors, termed RCLNet. Specifically, trapezoidal numerical double integration is utilized to construct multi-scale acceleration-velocity-displacement features, mitigating the cross-scale mapping challenge from a mechanistic perspective. Furthermore, the model integrates a Temporal Convolutional Network, a Convolutional Block Attention Module, and a Bidirectional Long Short-Term Memory network to simultaneously achieve local feature extraction, key information enhancement, and long-range spatial dependency modeling. Ultimately, high-precision inversion of vertical track irregularities is achieved in both the time and spatial domains. Building upon this, a dual-dimensional systematic validation of the proposed RCLNet model is conducted by combining simulation data generated from an Abaqus-based vehicle-track coupled dynamics model with field-measured on-board vibration signals from high-speed railways. The results indicate that the coefficient of determination R² of the proposed inversion method reaches 0.94, an improvement of 16 percentage points over the baseline model with a single acceleration input. Compared with other classical models, the proposed method achieves superior inversion accuracy and real-time performance. The inference time for a single sample is merely 20 ms, demonstrating excellent robustness even under complex field conditions.

Measurement Science and Technology
East China Jiaotong University (CN)
National Key Research and Development Program of China
Industry, innovation and infrastructure
Railway Engineering and Dynamics
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon