基于BP神经网络和粒子群算法的钢管混凝土拱桥可靠度分析

Reliability analysis of CFST arch bridges based on BP neural network and PSO algorithm

文章针对大跨径钢管混凝土拱桥结构可靠度求解困难的问题,将BP神经网络与粒子群算法引入拱桥可靠度分析领域,首先利用BP神经网络对结构极限状态函数进行拟合,将高度非线性的极限状态方程显式化,然后采用粒子群算法全局搜索验算点并求解可靠指标。计算分析结果表明,BP神经网络和粒子群算法弥补了传统可靠度分析方法的不足,提高了计算精度,为大跨径桥梁结构可靠度研究提供了新的思路和手段。

In order to solve the difficulty of reliability calculation for the long-span concrete filled steel tubular（CFST） arch bridges, the BP neural network and particle swarm optimization（PSO） algorithm were applied to the reliability analysis of arch bridges. Firstly, BP neural network was used to fit the limit state function, making the highly nonlinear limit state equation explicit. Then the PSO method was used to globally search the design points and calculate the reliability index. The analysis results showed that the BP neural network and PSO algorithm made up for the deficiency of the traditional re- liability analysis methods and improved the calculation accuracy, thus providing a new thought and means for the study of the reliability of long-span bridges.