Identification of critical samples of stochastic processes towards feasible structural reliability applications

• New identification framework for stochastic processes and non-linear dynamic systems.
• Fast and robust pre-screen of Monte Carlo samples.
• Benefits of fully probabilistic concept are maintained.
• Multi-disciplinary application possibilities.

This paper contributes to the structural reliability problem by presenting a novel approach that enables for identification of stochastic oscillatory processes as a critical input for given mechanical models. The proposed method is based on a graphical representation of such processes utilizing state of the art image processing and pattern recognition techniques, leading to a set of finite rules that consistently identifies those realizations of stochastic processes that would lead to a critical response of a given mechanical model. To examine the validity of the suggested method, large sets of realizations of artificial non-stationary processes were generated from known models, several criteria for critical response were formulated and the results were statistically evaluated. The promising results suggest important applications that would dramatically decrease computational costs e.g. in the field of probabilistic seismic design. Further examination may lead to a formulation of a new class of importance sampling techniques.