中文摘要：

以全桥安全性评价为总体目标，在以往桥梁安全性评价方法的基础上，引入人工神经网络理论，并结合层次分析法，提出了基于径向基网络的铜管混凝土拱桥安全性评价方法。从影响钢管混凝土拱桥安全性的承载能力、承重构件损伤以及外观损伤等3个主要方面进行考虑，分别建立RBF神经网络安全性评价模型，采用现场实测数据评价结果作为神经网络训练和检验样本，对神经网络进行学习训练，获取专家的经验知识和直觉思维，建立高度非线性的输入与输出的映射关系。通过仿真得到桥梁承载能力、承重构件损伤、外观损伤以及成桥状态下最终的安全性信息。以武汉市晴川桥为例进行工程实例分析，分析结果表明，该方法较好地反映了钢管混凝土拱桥结构的安全性状况。

英文摘要：

In order to gain the result of the safety assessment of CFST arch bridge, based on the traditional safety evaluation methods, artificial neural network （ANN） and analytic hierarchy process （AHP）, a safety evaluation method based on radial basis function （RBF） neural network was presented. With consideration of several key factors which affect the whole safety state such as load bearing capability, damage of bearing component and damaged state of appearance, RBF neural network models were built and trained by checking samples gained from the data measured on the spot. Then the experience and instinct thought of experts were gained and the highly nonlinear mapping relationship between the input and output factors was established. Finally, the whole safety evaluation of load bearing capability, damage of bearing component, damaged state of appearance and completed state of arch bridge were gotten by the method of simulation. The result of safety evaluation of Qingchuan Bridge in Wuhan city shows that the approach is highly potential and practical to evaluate the safety of long-span concrete filled steel tube arch bridge.