Author(s)

J. Naprstek & P. Prochazka

Abstract

The paper deals with behavior of scalar functional serving for optimization of system with stochastically perturbed parameters. The functional should be approximated, especially in the stationary point area by a quadratic hy- per surf ace, at least. This fact leads to a considerably non-gaussian nature of the random part of the functional, which comes up to the Pearson's %% process. The distance between the true optimum and the most probable state is increasing with increasing number of input parameters burdened by imperfections. As an example, a special form of the cost functional is studied. It represents the variance of stresses of a cylindrical thick-walled layered structure that are related to the eigenstrains in individual layers. 1 Introduction The qual

Keywords