Structural health monitoring of railway track lateral resistance using time-segmented train and earthquake excitations

Monitoring the global lateral resonance (GLR) of ballasted railway track provides a direct, non-intrusive indicator of lateral resistance and overall track stability, but no field-validated measurements of low-frequency lateral modes have previously been reported. This study demonstrates a dual-excitation operational modal analysis (OMA) method that combines broadband earthquake motions with routine train passages to identify and track the fundamental GLR frequency. A low-noise rail-mounted accelerometer system recorded 39 earthquakes and 64 train pass-bys on a curved mainline; signals were detrended, band-pass filtered (1-20 Hz), and analyzed using AR-Burg power spectral density estimation. A custom Python grid search evaluated 50 time-segmentation variants grouped in five families (A-E) to identify the most stable train segment. The earthquake records, free from modal mass addition and transient ballast stiffening, yielded a robust GLR reference of 7.01 Hz with a 5.3 % maximum intra-group deviation. Train data produced a median of 7.60 Hz, an absolute difference of only 0.59 Hz (approximate to 8.4 %) and within a +/- 15 % SHM tolerance. Among five time-segmentation windows, only segment C (first 5.5 s of the final 6 s) and segment D (final 5 s) satisfied the <= 15 % stability criterion. The results confirm that properly segmented train vibrations, triggered at 20 mm/s2 with a 14.5 s tail extension, can serve as reliable surrogates for continuous GLR monitoring. The validated framework enables predictive maintenance by flagging GLR frequency shifts that signal evolving ballast stiffness or lateral resistance loss, supporting safer and more cost-effective railway operations.