Vectorial generative adversarial surrogate modeling reliability evaluation framework for engineering structural systems

Abstract

To effectively evaluate the multi-objective reliability of engineering structural systems, the vectorial generative adversarial surrogate modeling (VGASM) reliability evaluation concept is proposed by integrating the surrogate model, matrix theory, generate adversarial strategy, Copula thought, and linkage sampling technology. In this concept, the surrogate model is employed as a basis function; the matrix theory is applied to establish the vectors and cell arrays of variables and hyperparameters; the generate adversarial strategy is adopted to determine the hyperparameters; the Copula thought and linkage sampling technology are utilized to simultaneously implement multi-objective correlated reliability analysis. Under this concept, the vectorial generative adversarial Kriging (VGAK) reliability evaluation method is developed by combining the Kriging model with VGASM reliability evaluation concept. In addition, a mathematical example and two engineering cases (i.e., aeroengine exhaust gas temperature and turbine blisk multi-failures) are employed to validate the proposed method. The reliability devel of aeroengine exhaust gas temperature and turbine blisk are 0.9989 and 0.9984, when the allowable values of EGT, deformation, and strain are 950 °C, 1.9253 × 10−3 m, and 5.2898 × 10−3 m. Besides, the presented method holds excellent modeling and simulation properties by comparing multiple methods. The study can provide theoretical references for the multi-objective reliability evaluation of engineering structural systems.