Structural reliability analysis based on improved parallel sampling algorithm

Aiming at the problem of computational inefficiency of traditional reliability analysis methods when dealing with large-scale structural systems, this paper proposes a reliability analysis method that improves the parallel sample algorithm. The method adopts the Kriging model to fit the performance function and introduces the mean square error to evaluate the uncertainty of the sample points; In addition, the positional relationship between the sample points is taken into consideration to avoid overfitting the model; To improve the efficiency of the model fitting, the sample points are divided into multiple clusters using the Mini batch k-means clustering algorithm in parallel computation. Finally, it is verified by multiple random variable cases that the method can significantly improve the computational efficiency while guaranteeing the relative error compared with other reliability analysis methods.