Reliability optimization design method based on multi-level surrogate model

In this work, a genetic-algorithm-based Kriging model with multi-point addition sequence optimization strategy is addressed to make up for the shortcomings of Kriging model with single point criterion. This approach combines the multi-point addition strategy with genetic algorithm to enable the Kriging model to efficiently capture the globally optimal solution. Based on this, a multi-level surrogate method is presented by employing a local surrogate model to modify the Kriging global surrogate model, and then applied to design optimization to improve the accuracy and efficiency of global optimization. Meanwhile, a reliability design optimization method based on multi-level surrogate model is studied by dealing with the reliability constraints with an adaptive reliability penalty function. Numerical examples show that the proposed method can find the optimal solution of the object problem with the least calculation cost under the condition of satisfying the reliability constraint.