Evaluation of bearing capacity of medium-small span old bridges based on GANN

Based on the complementation of advantages and disadvantages of genetic algorithms (GA) and neural network (NN), a new hybrid genetic algorithms and neural network model (GANN) is presented in this paper. In this model, the relationship model is established by error back propagation network (BP), the connection weights and thresholds of BP are optimized by GA, and then the precision of model is increased by BP. It not only avoids the deficiency of BP and GA, but also gives full play to the global searching capacity of the GA and local searching capacity of BP network. Aiming at decreasing the high-expenditure and heavy-workload of loading test of bridges, the new hybrid GANN bearing capacity evaluation model of medium-small span old bridge was established. In this model, the bearing capacity was evaluated by 8 easily measured damage indexes, and the calibration coefficient of the bridge bearing capacity (η) was used as the evaluation output index directly. After an evaluation analysis of bearing capacity, the result indicates that the evaluation model is scientific, accurate, available and convenient.