Novel efficient method for structural reliability analysis using hybrid nonlinear conjugate map-based support vector regression

The estimation of the failure probability for complex systems is a crucial issue for sustainability. Reliability analysis methods are needed to be developed to provide accurate estimations of the safety levels for the complex systems and structures of today. In this paper, a novel hybrid framework for the reliability analysis of engineering systems and structures is extended to reduce the computational burden. The proposed hybrid framework is named as SVR–CFORM and consists of coupling two parts: the first is an enhanced first-order reliability method (FORM) using nonlinear conjugate map (CFORM); the second is an artificial intelligence technique called support vector regression (SVR). The conjugate FORM (CFORM) is adaptively formulated to improve the robustness of the original iterative FORM algorithm, whereas the SVR technique is used to enhance the efficiency of the reliability analysis by reducing the computational burden. The performance of the proposed SVR–CFORM formulation is compared in terms of efficiency and robustness with several FORM formulas (i.e. HL–RF, directional stability transformation method, conjugate HL–RF and finite step length) through different numerical/structural reliability examples. Results indicate that the proposed SVR–CFORM formulation is more accurate and efficient than other reliability methods. Based on the comparative analysis results, the SVR technique can highly reduce the computational costs and accurately model the response of complex performance functions, while the iterative CFORM formulation found to provide stable and robust reliability index results compared to the others reliability methods.