Application of response surface method based on new strategy in structural reliability analysis

Abstract

In traditional reliability analysis methods, Monte Carlo method and finite element methods require a large amount of computational cost, and response surface method is an effective method to solve this problem. This method uses polynomial fitting instead of the exact limit state function (LSF) used in the Monte Carlo method, which greatly reduces the computational complexity. In this paper, a new response surface method is proposed for the following aspects, (1) A method of sample point location selection based on the sensitivity and importance of each random variable is proposed; (2) A method is proposed to calculate the experimental point according to the Euclidean distance between the experimental point and the limit state function and the relative position between the two design points; (3) A new search strategy for exploration area is proposed; (4) A new iterative control strategy is adopted; (5) The fractional order polynomial is used to replace the integer order polynomial, and a quadratic response process is added. This makes the response surface fit the LSF better, The validity of the method is proved by four examples.