engleski

Validation and verification of modified DRM with respect to its application in the reliability based design approach

Two novel methods for fast reliability and sensitivity calculations are presented. They are based on the raw statistical moment representation of the performance function and fitting the performance function’s Probability Density Function (PDF) to calculated moments. This alternative approach to structural reliability calculation is validated regarding its computational efficiency and verified regarding accuracy. Raw statistical moments of performance function are calculated using the univariate Dimension Reduction Method (DRM). For numerical integration, instead of normalized moment-based quadrature rule, integration points and their corresponding weights are obtained using Gauss-Hermite integration. A few numerical examples are presented and obtained results are compared with the Monte Carlo Simulation (MCS) and the First Order Reliability Method (FORM). The method, due to small number of evaluations needed, can be very useful for the multi-criteria design processes using reliability/risk based design attributes and constraints. Since most problems include correlated variates, the fast sensitivity analysis method (SENSIT) capable of contrasting the importance of particular variates are applied and their sensitivity regarding relevant correlation coefficients is determined. Both methods are parts of the new software FASTREL to be integrated into Design Support System OCTOPUS.

Modified Dimension Reduction Method; Reliability analysis; Sensitivity analysis; Reliability Based Design

nije evidentirano