Piecewise-linear system reliability estimates

A procedure for approximating nonlinear component limit state functions in system reliability analyses is presented. The procedure idealizes each component limit state surface in a piecewise-linear manner with a collection of hyper-planes. Guidelines for optimally locating the hyper-planes are provided and an innovative algorithm for combining the half-spaces defined by the hyper-planes to represent arbitrarily shaped failure domains is described. Numerical examples are used to demonstrate the accuracy of the proposed procedure. The results of the examples indicate that the system reliability estimate provided by the proposed procedure can be significantly more accurate than that obtained from a conventional first-order system reliability analysis.