中文摘要：

结合现行规范及钢管混凝土拱桥的受力特点，给出钢管混凝土拱桥安全一中介一失效三级工作模式的确定准则．据此计算结构的损伤概率与失效概率及对应的可靠度指标，建立可靠度指标与三级工作模式的直接关系．基于三级工作模式，提出评估在役桥梁结构可靠度的有限元-神经网络-蒙特卡罗（FNM）法．根据时变可靠度理论，将结构继续服役期离散化，随机抽样生成蒙特卡罗法所需的最少样本，由随机有限元法计算训练样本，用神经网络扩展训练样本，最后通过蒙特卡罗法计算失效概率和损伤概率，得到结构时变可靠度，判断其工作状态．运用该方法对一座建成10年，主跨83．6m的钢管混凝土拱桥进行分析，结果表明，该桥主拱处于安全状态，这一结论与实际情况相符．说明FNM法对于评价在役桥梁的可靠性是可行的．

英文摘要：

Combined with current specifications and stress characteristics of concrete filled steel tubular （CFST） arch bridges, the determination principle of safe-middle-failure threestage mode is given. Accordingly, damage probability and failure probability and the corresponding reliability indices are calculated; a direct relationship between reliability indices and three-stage working status is made. Based on the three-stage working mode, a combined FNM （finite element-neural network- Monte-Carlo simulation） method is put forward to estimate the reliability of existing bridges. According to time-dependent reliability theory, subsequent service time is divided into several stages; minimum samples required by the Monte-Carlo method are generated by random sampling; training samples are calculated by the finite element method, and the training samples are extended by the neural network; failure probability and damage probability are calculated by the Monte-Carlo method. Thus, time dependent reliability indices are obtained, and the working status is judged. A case study is investigated to estimate the reliability of an actual bridge by the FNM method. The bridge is a CFST arch bridge with an 83.6 m span and it has been in operation for 10 years. According to analysis results, in the tenth year, the example bridge is still in safe status. This conclusion is consistent with the facts, which proves the feasibility of the FNM method for estimating the reliability of existing bridges.