In this paper we study rare events associated to the solutions of an elliptic partial differential equation with a spatially varying random coefficient. The random coefficient follows the lognormal distribution, which is determined by a Gaussian process. This model is employed to study the failure problem of elastic materials in random media in which the failure is characterized by the criterion that the strain field exceeds a high threshold. We propose an efficient importance sampling scheme to compute the small failure probability in the high threshold limit. The change of measure in our scheme is parametrized by two density functions. The efficiency of the importance sampling scheme is validated by numerical examples.
Introduction.The study and computation of rare events in stochastic systems have received intensive attention in recent years. Rare events, though they do not occur often, represent the most severe consequence of uncertainty and random effects. The study of these rare events hence gives crucial understanding and has important applications. However, due to the small probabilities of occurrence of such events, quantification casts a serious challenge for conventional probabilistic methods. For example, a direct Monte Carlo strategy to estimate the vanishing small probability will require a huge number of sample points to give estimates with small relative error; in other words, the huge relative variance of these estimators makes them incapable of accurate prediction.In this work, we aim at developing an efficient important sampling strategy to study the rare events associated with a materials failure problem. The method we develop in this work applies to the general linear elasticity model. For simplicity, we will restrict our discussions here to a scalar model in two dimensions, which can be viewed as a model for out-of-plane deformation of an elastic membrane under external forcing. Similar equations also arise from other contexts, such as groundwater hydraulics and electrostatic response of a planar media. Let D ⊂ R 2 be an open domain with smooth boundary, which is the equilibrium configuration of the membrane. We consider an out-of-plane displacement field u given by the following boundary