ABSTRACT
Given that the dam monitoring model based on the conventional Radial Basis Function (RBF) neural network cannot screen the remarkable factors and is liable to fall into the local optimum, this paper establishes a fusion monitoring model that combines the Mean Impact Value (MIV), improved fruit Fly Optimization Algorithm (FOA), and the RBF neural network. First, the MIV is introduced to screen three kinds of forecast factors, namely, water pressure, temperature, and aging. Then, the improved FOA algorithm is adopted for searching the optimal spread value of the RBF network. By using these two methods, the MIV-improved RBF neural network model is set up. To verify the validity of the model, taking the displacement monitoring data of the gravity dam into account, the multiple linear regression model, the conventional RBF model, and the MIV-improved RBF model are also built. The calculation results show that the MIV-improved RBF neural network model has characteristics such as great generalization ability, stable prediction, and high precision. Furthermore, this model can be applied to the dam deformation for monitoring and warning.
