基于神经网络算法的大跨径斜拉桥随机地震效应主梁动力可靠度研究

Random Seismic Effect of Long-span Cable-stayed Bridge on Dynamic Reliability of Main Beam Based on Neural Network Algorithm Study

为解决大跨径斜拉桥在随机地震作用下动力可靠度计算费时费力的问题,基于首次超越准则,构建了考虑地震随机性和结构本身参数随机性的双重随机功能函数,提出了使用神经网络算法获取结构内力响应均方根并使用线性规划优化方法求解最小可靠度的方法,并以某大跨径斜拉桥为算例进行了算法验证。计算结果表明:相比于传统的响应面法,基于神经网络算法的动力可靠度求解方法在满足结构计算精度的同时,具有更快的拟合速度和更高的计算效率。

In order to solve the problem of time-consuming and labor-intensive calculation of the dynamic reliability of long-span cable-stayed bridges under random earthquakes,based on the first transcendence criterion,a double random function function considering the randomness of earthquakes and the randomness of parameters of the structure itself was constructed,and a neural network was proposed.The algorithm obtains the root mean square of the internal force response of the structure and uses the linear programming optimization method to solve the minimum reliability method,and verifies the algorithm by taking a large-span diagonal cable as an example.The calculation results show that,compared with the traditional response surface method,the dynamic reliability solution method based on neural network algorithm not only meets the structural calculation accuracy,but also has a faster fitting speed and higher calculation efficiency.