Structural Reliability Analysis of Complex Systems: Applications to Offshore and Composite Structures

This thesis aims at developing new methodologies for the reliability analysis of structural systems with applications to offshore and aeronautical fields. In general, sructures of practical interest are complex redundant systems, in which more than one element is required to fail in order to have catastrophic failure. Moreover, ramdomness inherently exists in both material properties and external loads. As a result, complex structural systems are typically characterised by a huge number of possible failure sequences, of which only some are most likely to occour. Therefore, for an efficient risk analysis, only the dominant failure modes need to be considered, so as to minimise the number of failure paths as well as the computational costs associated to their enumeration and evaluation. However, although several techniques have been developed for the identification of the critical failure sequences, these methods are still either time-demanding or prone to miss potential failure modes. These challenges motivated the first part of the thesis, in which the merits of a risk assessment framework recently developed for truss and frame structures are here investigated in view of its extensive application to the offshore field. To this end, the case study of a jacket-type platform under an extreme sea state is considered. First, the dominant failure modes of the structure are rapidly identified by a multi-point parallel search employing a genetic algorithm. Then, a multi-scale system reliability analysis is performed, in which the statistical dependence among both structural elements and failure modes is fully considered through simple matrix operations. Finally, the accuracy and the efficiency of the proposed approach are successfully validated against crude Monte Carlo simulation. In the second part of the thesis, system reliability theory is applied to the uncertainty quantification of the longitudinal tensile strength of UniDirectional (UD) composites, a structural component very common in aircraft structures. Predictive models for size effects in this class of materials are paramount for scaling small-coupon experimental results to the design of large composite structures. In this respect, a Monte Carlo progressive failure analysis is proposed to calculate the strength distributions of hierarchical fibre bundles, which are formed by grouping a predefined number of smaller-order bundles into a larger-order one. The present approach is firstly validated against a recent analytical model to be later applied to more complex load-sharing configurations. The resulting distributions are finally used to analyse the damage accumulation process and the formation of clusters of broken fibres during progressive failure.