Static system reliability analysis of cable-stayed bridge based on improved BP neural network

In order to enhance the efficiency of static system reliability calculation for cable-stayed bridges, a system reliability analysis model was developed based on an improved back propagation (BP) neural network. By introducing the genetic algorithm (GA) into the BP neural network, the limit state functions of key cable-stayed bridge components could be efficiently reconstructed, the design points could be captured rapidly, and the algorithm of GA-BP-GA-Monte Carlo (GBGMC) for component reliability index calculation was established. The rectified β-unzipping method was used to select candidate failure components, and the structure was modified by assuming in turn failure in the potential failure elements. The primary failure modes of the cable-stayed bridge were identified, upon which the fault tree was subsequently constructed. Based on the equivalent linear functions of the failure modes and correlation coefficients, the differential equivalent recursion algorithm was employed to calculate the reliability of the structural system. The effectiveness and accuracy of GBGMC were verified through a reliability study of three numerical examples. The proposed system reliability analysis method was used to evaluate the structural failure history of a cable-stayed bridge with a main span of 448 m. The component reliability indexes in each failure stage were calculated, and the structural system failure tree of the cable-stayed bridge was created. The structural system reliability index was efficiently calculated, and important components controlling the system safety were identified. Research results show that the computational error of GBGMC is within 0.3%, which is better than other traditional methods.The deflection reliability index at the main span center of the cable-stayed bridge is the lowest for the normal utilization limit state, which is 2.7. In terms of the ultimate limit state, the reliability indexes of the cables in the center of the main span, the main girder at the tower, and the lower section of the tower's cable anchorage area all pose a high risk of failure. Their respective reliability indexes are 3.1, 3.6, and 3.9, respectively. Nineteen main failure modes for the cable-stayed bridge at ultimate limit state are obtained, and the system reliability index is 3.8. Research results provide a theoretical basis of system safety control for the design optimization and maintenance decision of cable-stayed bridges.