Life-cycle oriented strain reconstruction of CRTSIII slab tracks using quasi-distributed fiber Bragg grating sensing and deep learning toward sustainable high-speed rail

The widespread implementation of ballastless slab track systems has positioned the CRTS III structure as a key component in long-life, low-maintenance, and sustainable high-speed railway infrastructure. While the system exhibits strong mechanical resilience and reduced maintenance demand, long-term service conditions—particularly repeated dynamic loading—may lead to cumulative deterioration, underscoring the need for continuous monitoring to support life-cycle performance evaluation. Quasi-distributed fiber Bragg grating (FBG) sensing enables real-time internal strain assessment, yet sensor degradation or interfacial debonding can result in missing measurements, affecting the integrity of long-term structural health monitoring (SHM) records. This study investigates deep learning-based reconstruction of incomplete strain data obtained from embedded FBG sensors in CRTS III slab track structures. Full-scale cyclic loading tests provide reference strain sequences obtained from FBG sensors installed within the self-compacting concrete layer as well as the underlying baseplate. A set of deep learning models comprising CNN-based architectures, LSTM temporal networks, and GRU recurrent structures are trained on complete sequences and tested under artificially constructed conditions with partial data loss using standard regression metrics. Results demonstrate that the developed approach markedly improves the completeness and robustness of FBG-based monitoring records. The findings support life-cycle oriented SHM, enabling more effective condition-based maintenance, extended reuse of slab track components, and lower material usage achieved via data-informed, low-emission maintenance strategies for high-speed rail systems.