Sensitivity analysis methods for reliability problems

In reliability problems an important aspect is the study of the influence of parameter changes on the target reliability which is often more an operational value. Since in complex problems such quantities have to be computed often, efficient methods for doing this are desirable. Here some methods for sensitivity analysis are outlined. For a FORM/SORM analysis it is possible to get sensitivity measures from the Lagrange mul- tiplier at the beta point. This allows simple estimates for the sensitivities with respect to parameters without additional computations. In the general case the partial derivatives and sensitivities of the failure probability with respect to parameters are given by surface integrals over the limit state surface. Such integrals can be transformed into domain inte- grals over the safe domain using the divergence theorem (Gauss-Ostrogradsky theorem). By modifying the integrands in a suitable way, it is possible to modify the integration domains such that the integrals can be estimated in a more efficient way.     «

In reliability problems an important aspect is the study of the influence of parameter changes on the target reliability which is often more an operational value. Since in complex problems such quantities have to be computed often, efficient methods for doing this are desirable. Here some methods for sensitivity analysis are outlined. For a FORM/SORM analysis it is possible to get sensitivity measures from the Lagrange mul- tiplier at the beta point. This allows simple estimates for the sensiti...     »