公路工程

针对大跨度斜拉桥地震易损性研究中计算量大、计算成本高的问题，提出了一种基于改进BP神经网络响应面的地震易损性计算模型。首先推导了氯离子侵蚀机理和钢筋材料性能退化模型。其次提出了基于自适应遗传算法优化BP神经网络的结构易损性计算方法，采用BP神经网络拟合了两种锈蚀模式结构在地震作用下的地震响应特性，并通过蒙特卡洛法(Monte Carlo Sampling, MCS)计算了不同损伤水平的结构地震易损性指标。最后以某大跨度斜拉桥为工程背景，计算了考虑钢筋锈蚀程度的桥墩结构时变地震易损性曲线。结果表明：基于改进BP神经网络建立的结构地震响应代理模型拟合精度较高，泛化能力较强；当考虑轻微损伤和中等损伤时，钢筋锈蚀程度对地震损伤超越概率的影响较小，当考虑严重损伤和完全破坏时，随着服役年限的增加，钢筋锈蚀程度对结构抗震性能的影响越来越大；当地面峰值加速度为1.2g时，考虑严重损伤和完全破坏下，服役时间100 a的结构损伤概率相较服役时间50 a的分别高了12.85%和18.16%;考虑的损伤指标越高时，钢筋锈蚀程度对结构抗震性能的影响越大。

Aiming at the problems of large amount of calculation and high cost in the seismic vulnerability study of long-span cable-stayed bridges, a seismic vulnerability calculation model based on Improved BP neural network response surface is proposed. Firstly, the mechanism of chloride ion erosion and the degradation model of steel material properties are derived. Secondly, a structural vulnerability calculation method based on BP neural network optimized by adaptive genetic algorithm is proposed. The seismic response characteristics of two kinds of corroded structures under earthquake action are fitted by BP neural network, and the seismic vulnerability indexes of structures with different damage levels are calculated by Monte Carlo sampling(MCS). Finally, taking a long-span cable-stayed bridge as the engineering background, the time-varying seismic vulnerability curve of pier structure considering the degree of reinforcement corrosion is calculated. The results show that the surrogate model of structural seismic response based on Improved BP neural network has high fitting accuracy and generalization ability; when considering slight damage and moderate damage, the influence of reinforcement corrosion degree on seismic damage exceedance probability is small. When considering severe damage and complete damage, the influence of reinforcement corrosion degree on seismic performance of structure is more and more significant with the increase of service life; when the ground peak acceleration is 1.2g, the damage probability of the structure with service time of 100 a is 12.85% and 18.16% higher than that with service time of 50 a, respectively, considering the severe damage and complete failure; the higher the damage index, the greater the impact of steel corrosion on the seismic performance of the structure.