On cyclic accumulation models for degradation of railway foundations

Abstract: The degradation of railway foundations due to the repeated loading of traffic induces maintenance and safety obligations for the track operator. One major contributor to track degradation is the accelerated settlement rates in soft soils below the track that lead to alignment issues, especially at stiffness transitions. Furthermore, due to the low stiffness of most soft soils, significant ground vibrations are emitted and lead to low critical train velocities that avoid track resonance. In addition, the dynamic properties of the soil, particularly weakly bonded soft natural clays, are subject to significant alternations over the lifetime of the railway structure due to cyclic traffic loading. The soil/foundation is thus a major source of degradation issues that, as opposed to track and subgrade-related causes, are largely controlled by local site conditions. This thesis aims at identifying a proper sensitivity analysis method in geotechnical Finite Element Analysis (FEA) for optimal use of advanced constitutive soil models. For this purpose, first the viscoplastic Creep-SCLAY1S model is evaluated for a boundary value problem. The objective was ultimately addressed by implementing two Global Sensitivity Analysis (GSA) methods for quantifying the uncertainties of Creep-SCLAY1S. The common GSA method of Sobol was benchmarked against Experimental design in a lab-scale numerical model of Constant Rate of Strain (CRS). The Sobol method has proven to be computationally expensive for sensitivity analysis of advanced constitutive models using FEA. The spatial sensitivity measures of Sobol and Experimental design indicate that they are not altogether distinct. Thus, Experimental design represents a more feasible approach by using less resources, such as computational time and required storage. Furthermore, temporal Sensitivity Analysis (SA) has demonstrated the importance of the entire time domain spectrum, particularly for factor fixing purposes. The second part of the study proposes a revised strain accumulation model that has been validated using new data on Swedish natural clay for cyclic loads with low amplitude. The model presented herein offers a strong basis for the accurate prediction of strain accumulation in soft clays beneath embankments subjected to a significant number of loading cycles. In general, the knowledge gained in this research contributes to a better understanding of comprehensive numerical models in Geotechnics and can be a helpful prior to inverse modelling, data assimilation, and Random Finite Element Method (RFEM).