Reliability analysis of underground cavern using Gaussian process classification
Guoshao Su, Yilong Xiao, Zineng Xia · 2011
Aiming to the difficulties resulted by implicit performance function and the limitations of traditional methods for reliability analysis of underground cavern, a new method based on Gaussian process classification (GPC), which is a newly developed machine learning method, is proposed for structural reliability analysis of underground cavern. The implicit performance function of underground cavern is reconstructed by GPC model based on small amount of training datasets which are built by finite element numerical simulation. Thus, the implicit performance function is approximated by GPC model with explicit formulation. Then, Monte Carlo Simulation method is applied to get the failure probability and reliability index of underground cavern using the approximated performance function by GPC. The study results show that the method can get accurate and computationally efficient results for reliability analysis of underground cavern.