ABSTRACT
The cable-stayed bridge is widely used for long-span bridges because of the superior performance of its long-span capability and relatively lower cost. However, due to the effects of environment corrosion and traffic growth, the reliability and safety of existing cable-stayed bridges decreases continuously. The traditional reliability analysis approaches are inappropriate for long-span cable-stayed bridges with high structural nonlinearities and complex component systems. A hybrid approach was presented for reliability assessment of long-span cable-stayed bridges based on artificial neural networks. The structural nonlinearity was considered in the kernel radial basis functions and the system behavior was captured by the networks. The proposed hybrid approach was verified in two numerical studies. Finally, the hybrid approach was utilized for the case study of a long-span prestressed concrete cable-stayed bridge. Parametric studies of the bridge were conducted to find the sensitive parameters impacting the bridge safety. The numerical results indicate that the sensitive factors are the cross-sectional area and the elasticity modulus of the stay cables.
