G. De Castro, R. Motta, A. Paixão, E. Moura, E. Fortunato, L. Bernucci
The increase in axle loads in heavy-haul railways demands sub-ballast materials capable of maintaining long-term performance under severe cyclic loading, particularly in tropical environments. This study addresses the long-term deformation behaviour of tropical lateritic soils used as sub-ballast layers in heavy-haul railway tracks subjected to increased axle loads. The investigation focuses on untreated lateritic soil and lateritic soil treated with 3% cement under rehabilitation conditions using a Rail-mounted Formation Rehabilitation Machine (RFRM). The methodological framework combines laboratory testing and advanced numerical modelling. Mechanical characterization and cyclic triaxial tests were conducted to evaluate resilient behaviour and permanent deformation, while a 3D finite element model incorporating the High-Cycle Accumulation (HCA) approach was calibrated and applied to the untreated lateritic soil to assess long-term deformation trends under 32.5 t and 40 t axle loads. The results indicate good agreement between numerical predictions and laboratory measurements for the untreated soil, validating the adopted modelling strategy within the converged stress states. The untreated soil compacted with a vibratory plate (RFRM) showed excessive permanent deformation under 40 t/axle loading, demonstrating the limitations of this compaction method for heavy-haul conditions. In contrast, the 3% cement-treated soil significantly reduced deformation and ensured stable long-term behaviour in the laboratory tests under both load levels. Overall, the study confirms that low cement content treatment markedly improves the resilience of tropical lateritic sub-ballast soils and that the HCA-based numerical framework is a useful comparative tool for evaluating the nonlinear long-term performance of untreated lateritic sub-ballast. The findings provide practical guidance for the design and rehabilitation of heavy-haul railway infrastructure in tropical regions.