Support vector regression for structural reliability analysis

Support vector regression (SVR), which is a novel regression method and a potential alternative to artificial neural networks (ANN), due to its excellent learning capacity and generalization capability with a small amount of samples, has not been widely applied to structural reliability analysis currently. Thereby, two approaches based on SVR are proposed for structural reliability analysis, i.e., SVR-based MCS and SVR-based FORM. In the proposed methods SVR is employed to approximate the implicit performance function. Examples with simple performance functions are used to demonstrate the application of SVR in the structural reliability analysis. Some comparisons among the classical structural reliability analysis methods and the proposed approaches show that highly accurate approximation both in the performance functions and failure probabilities can be obtained by the SVR with a small amount of information.