Efficient reliability analysis of laminated composites using advanced Kriging surrogate model

To demonstrate a computational efficient and accurate approach for reliability analysis of laminated composites, an advanced Kriging model is applied to approximate the mechanical model of the structure. To construct the surrogate model, the structural response is simulated through finite element method based on classic theory of laminates. Tsai-Wu criterion is adopted to define limit state function in reliability analysis. A high quality surrogate is achieved using a probabilistic classification function together with a metric for refinement of the model. This method uses just a limited number of finite element analyses to construct the surrogate model which verifies the computational efficiency of the method. In addition, high accuracy is achieved because the enrichment technique of experiment points in the vicinity of the limit state function allows using Monte Carlo simulation to calculate the probability of failure. Numerical examples of neural network based surrogates are compared with advanced Kriging surrogate. The comparison results verify the computational cost efficiency and high accuracy of proposed approach for reliability analysis of laminated composites.