Thermally induced rail defects—including track buckling in extreme heat and pull-aparts in extreme cold—remain a critical safety concern for the U.S. rail network. These failures arise from thermally induced longitudinal forces in continuous welded rail (CWR), governed by rail neutral temperature (RNT), ambient temperature variation, and track restraint. Reliable estimation of internal rail stress, from which RNT can be inferred, is therefore essential for safe CWR management. However, existing methods for assessing RNT are intrusive, localized, labor-intensive, or dependent on uncertain installation records, limiting their scalability for network-level monitoring [2,3].
This project proposes a physics-informed digital twin framework for nondestructive estimation of internal rail stress using non-contact vibration measurements. The approach leverages the acoustoelastic sensitivity of low-frequency flexural waves in rails, where axial stress induces measurable shifts in phase velocity and wavenumber [1]. The field system consists of three integrated components: (1) synchronized dual actuators to generate controlled excitation, (2) Scanning Laser Doppler Vibrometry (SLDV) to capture spatially resolved wavefields along the rail head and web, and (3) a physics-informed digital twin that models the coupled effects of axial stress, track-support stiffness, and damping.
A key technical challenge is the reliable measurement of wave propagation characteristics, particularly phase velocity, in the low-frequency regime (<10 kHz). Conventional single-input methods (e.g., hammer impact or single-shaker excitation) produce mixed forward- and backward-propagating waves that are strongly influenced by boundary reflections, making direct phase-velocity estimation difficult. While prior studies have demonstrated stress-sensitive wavelength measurements using harmonic excitation and laser vibrometry, they remain limited by wave interference effects.
To overcome this limitation, this project introduces a synchronized dual-actuation strategy to generate steady-state traveling waves over a scanned rail segment. This approach suppresses reflection-induced interference and enables direct, spatially resolved estimation of phase velocity. The measured wavefields will be integrated with a reduced-order, physics-informed digital twin to infer internal rail stress without requiring baseline RNT information.
This work establishes a scalable, nondestructive framework for rail stress monitoring and advances wave-based sensing methodologies for distributed structural state estimation in transportation infrastructure.
[1]Malladi, Vijaya VN Sriram, Mohammad I. Albakri, Manu Krishnan, Serkan Gugercin, andPablo A. Tarazaga. “Estimating experimental dispersion curves from steady-state frequencyresponse measurements.” Mechanical Systems and Signal Processing 164 (2022): 108218.
[2]Sun, L. J., Li, Z. W., Zhu, W. F., He, Y. L., Fan, G. P., Fang, W. P., & Shao, W. (2021). A methodfor long-term on-line monitoring of temperature stress of continuously welded rail. Advances inMechanical Engineering, 13(8), 16878140211041432.
[3]Phillips, R., Bartoli, I., Coccia, S., Lanza di Scalea, F., Salamone, S., Nucera, C., … & Carr, G.(2011, June). Nonlinear guided waves in continuously welded rails for buckling prediction. In AIPConference Proceedings (Vol. 1335, No. 1, pp. 314-321). American Institute of Physics.