Accuracy evaluation of railway track irregularity inspection using bogie-mounted IMUs of different grades

Railway track irregularity inspection is critical for operational safety and ride comfort. As inspection time windows become increasingly constrained, onboard inspection systems mounted on in-service trains have attracted growing attention because they can improve inspection efficiency. In such systems, inertial measurement units (IMUs), when integrated with laser sensors, provide an effective solution for dynamic track geometry measurement, especially for high-grade IMUs. However, the high cost of high-grade IMUs limits large-scale deployment. In this study, we evaluated the measurement accuracy of onboard track irregularity inspection systems based on 2D laser sensors and IMUs of different grades, together with different processing schemes. The results show that, under the adopted semi-physical evaluation framework based on signal grafting and after constant bias calibration, the system combining a tactical-grade IMU with 2D laser sensors has the potential to achieve measurement accuracies for gauge, cross level, and lateral/vertical irregularities comparable to those obtained using a navigation-grade IMU, satisfying the requirements of the relevant Chinese national standards. For micro-electro-mechanical system (MEMS) IMUs integrated with the same laser sensors, the accuracy of cross level measurement is inherently constrained by attitude estimation accuracy. However, for the other track geometry parameters, acceptable measurement accuracy can still be maintained after removing constant biases and omitting attitude-based gravity compensation, which mitigates attitude-error amplification but should be regarded as a trade-off that may reduce waveform fidelity. These results clarify the achievable measurement capability of IMUs of different grades within the adopted semi-physical evaluation framework and provide practical guidance for the design of cost-effective onboard track irregularity inspection systems.