Data-driven optimization of reliability using buffered failure probability

Design and operation of complex engineering systems rely on reliability optimization. Such optimization requires us to account for uncertainties expressed in terms complicated, high-dimensional probability distributions, which only samples or data might be available. However, using often degrades the computational efficiency, particularly as conventional failure is estimated indicator function whose gradient not defined at zero. To address this issue, by leveraging buffered probability, paper develops method (BORM) efficient, data-driven reliability. The proposed formulations, algorithms, strategies greatly improve efficiency thereby needs nonlinear problems. In addition, an analytical formula developed estimate sensitivity, a subject fraught with difficulty when probability. thoroughly investigated context many different leading novel measure tail-heaviness called tail index. accuracy methodology are demonstrated three numerical examples. Although they show slight deviations from target because sampling errors other inaccuracies, results underline unique advantages potentials analysis.