Active learning polynomial chaos expansion for reliability analysis by maximizing expected indicator function prediction error

Abstract

Assessing the failure probability of complex aeronautical structure is a difficult task in presence of uncertainties. In this paper, active learning polynomial chaos expansion (PCE) is developed for reliability analysis. The proposed method firstly assigns a Gaussian Process (GP) prior to the model response, and the covariance function of this GP is defined by the inner product of PCE basis function. Then, we show that a PCE model can be derived by the posterior mean of the GP, and the posterior variance is obtained to measure the local prediction error as Kriging model. Also, the expectation of the prediction variance is derived to measure the overall accuracy of the obtained PCE model. Then, a learning function, named expected indicator function prediction error (EIFPE), is proposed to update the design of experiment of PCE model for reliability analysis.This learning function is developed under the framework of the variance-bias decomposition. It selects new points sequentially by maximizing the EIFPE that considers both the variance and bias information, and it provides a dynamic balance between global exploration and local exploitation. Finally, several test functions and engineering applications are investigated, and the results are compared with the widely used Kriging model combined with U and expected feasibility function learning function. Results show that the proposed method is efficient and accurate for complex engineering applications.

AMRBF-SS: Subset Simulation with Active Learning and Multiple Kernels Radial Basis Function for Small Failure Probability Prediction

“…Other surrogate models, such as the polynomial chaos expansion (PCE), have also received attention (Sudret and Mai 2014). Recent developments in sparse Bayesian PCE have facilitated reliability analyses (Marelli and Sudret 2018;Cheng and Lu 2020;He et al 2020). However, the PCE faces the same challenge for highdimensional problems.…”

Support vector machine in structural reliability analysis: A review

“…Most of the real structures involved complex geometry and nonlinear material behaviours that required a computationally demanding finite element (FE) analysis or other numerical techniques for response evaluation. Different metamodeling approaches e.g., response surface method (RSM) [5,6], radial basis functions networks (RBFN) [7], polynomial chaos expansion (PCE) [8,9], multivariate adaptive regression splines (MARS) [10], Kriging method [11,12], artificial neural networks (ANN) [13,14], etc., were developed to address the computational challenge of large complex SRA problems. However, such metamodels were developed following the empirical risk minimization principle.…”

Active Learning-based Domain Adaptive Localized Polynomial Chaos Expansion

“…Active learning for reliability analysis with PCE was used e.g. in [16,17,18]. For general UQ studies, some recent studies have focused on general sequential sampling for PCE based on spacefilling criteria or alphabetical optimality [19,20].…”