在役桥梁全过程FNM风险概率估计法

基于时变可靠度理论,考虑三级工作模式,提出在役桥梁全过程FNM (Finite element-Neural network-Monte-Carlo simulation) 风险概率估计法。在风险识别的基础上,通过蒙特罗抽样获得桥梁使用周期内某一时间的n组风险因素随机样本。n为蒙特卡罗法模拟所需样本数的经验下限。设神经网络训练所需样本数的经验上限为N,通过有限元计算获得此N组样本。剩余n-N组样本由神经网络模型获得。由此即可得到蒙特卡罗法模拟所需的n组样本。实桥算例以一座运营了8年、主跨80m的吊杆砼拱桥为对象,预测了该结构的承载力和风险概率。算例分析表明,该方法科学合理、简便有效,能为在役桥梁的加固决策提供参考依据。

FNM Method for Estimating Risk Probability of Existing Bridge in Full Range

A new method for risk probability assessment is brought forward in this paper, which is called FNM (Finite element-Neural network-Monte-Carlo simulation) risk probability method for existing bridge. Based on time-dependent reliability theory, taken the “safe-middle-failure” three-grade work mode into account, n groups samples in service period are obtained by risk identification, where n is experiential minimum limit that is needed by Monte-Carlo simulation (MCS). Let the maximum limit of neural network’s training sample number equal to N, this groups samples can be worked out by finite element method. The rest n-N samples can be achieved by neural network. FNM method can work out bearing capability, risk probability and remaining life of existing bridge in full-range. A case study is investigated using FNM method to estimate risk probability of existing bridge. The bridge is a concrete arch bridge with 80-meter-span, which operates for 8 years. The actual example shows that FNM method is reasonable, simple and effective. And reference basis for maintenance decision of the bridge was provided.