A DBN-PSOSA hybrid response surface method for bridge reliability

Abstract

When solving bridge reliability problems, the traditional response surface method has a highly nonlinear implicit function, which results in a low fitting accuracy and difficulty in meeting the requirements. Therefore, the dynamic Bayesian network (DBN), which is suitable for solving a multi-state unit or system uncertainty problems, is selected to construct the response surface of the implicit function. In this study, the DBN model is combined with the particle swarm optimisation based on simulated annealing (PSOSA) algorithm to improve the optimisation efficiency of model parameters and allow the constructed implicit function to simulate the structure limit state function. The DBN-PSOSA hybrid response surface analysis method is proposed for bridge failure probability calculations. A numerical example is given to demonstrate the effectiveness of the proposed method, and the reliability of an actual bridge project is analysed. The results indicate that this method has a higher calculation accuracy and efficiency compared to the traditional response surface method, and is easy to combine with the existing general finite element analysis software to achieve the rapid analysis of bridge structure reliability.