A new parameter perturbation method suitable for reliability analysis of large dynamic systems

Paper in proceeding, 2013

To decrease the computational efforts needed for reliability analysis of large-scale finite element (FE) models, fast simulation is a necessity. To achieve this, model reduction approach offers a systematic way to capture the main input-output behaviour of the original dynamic model with a simpler one. However to be useful for reliability analysis, the approximant must provide a good accuracy for a range of stochastic parameters variation. This problem is addressed by introducing a reduced modal basis based on a reference Linear Time Invariant (LTI) FE model.Amethod is employed that uses an iterative algorithm to obtain perturbed solutions that are due to the effects of parameter variation. The main idea is to iteratively approximate the perturbed solution in the space spanned by reduced basis. This technique is used to develop an adaptive reliability method. In a numerical example, this approach is applied to a railway track problem.