Functional quantization of probabilistic life-cycle performance models

In life-cycle civil engineering, the uncertainty in the deteriorating performance (e.g., reliability) of structures is often described by means of time-dependent models that have random parameters. Most of the numerical models of individual structures (e.g., bridges) are already complex, and when a life-cycle performance analysis of a network of systems is carried out, the complexity and the computational time become very high. Thus, for simulation-based probabilistic analyses, the number of deterministic runs that can actually be performed is limited. Then, a method that can provide the most accurate results given a moderate sample size is of paramount importance. In this paper, a technique called Functional Quantization by Infinite-Dimensional Centroidal Voronoi Tessellation is used for the optimal selection of the life-cycle profiles samples, which are non-Gaussian and non-stationary random functions. Functional Quantization provides a set of realizations of the stochastic life-cycle process and their associated relative weights, which can be used to obtain the optimal probabilistic representation of the process at hand.