LAUSR

With emphasis on long-span bridges, the dynamic responses of bridges without considering random traffic flows were found to be different from actual situations. The introduction of a random traffic flow model provides a new approach for the random analysis of bridge structure responses under vehicle loads. In this paper, the finite element and intelligent ant colony optimization-back propagation neural network (ACO-BPNN) models were used to study the dynamic responses of long-span bridges. The computational model was also validated by an experimental test. To confirm the validity of the proposed ACO-BPNN model after parameter selection, it was compared with the traditional back propagation neural network (BPNN) model and the genetic algorithm-back propagation neural network (GA-BPNN) model. BPNN, GA-BPNN, and ACO-BPNN adopt the same network topology structure to predict the dynamic responses of the long-span bridge. When the ACO-BPNN model conducted the iteration to the 130th generation, a training error of 0.009 was found to be smaller than the set critical error. In this manner, the computational accuracy was increased, and the optimized time was reduced. In addition, only 0.4 hours were spent in using the proposed ACO-BPNN model to predict the dynamic response of the long-span bridge. In the case of the same computer performance, it took 4.5 h to use the finite element model to predict the dynamic response of the long-span bridge. The advantage of the proposed ACO-BPNN model in predicting the performance of large-scale complex structures such as long-span bridges was clearly found.