Application of Aritificial Neural Network method in construction control of continual bridge

This paper takes the Nenjiang river continuous girder bridge as an example to discuss a method of carrying out predictive analysis on deflection in different construction stages by means of Aritificial neural network technology. Based on the foundation of finite element model, training samples for RBF network are obtained. BP and RBF neural network are used to forecast the deflection of bridge girder separately. The theoretical data of former several construction stages was used for network training, so that the construction deflections of subsequent construction stages will be forecasted. Comparing between predicted values and measured value, the reliability of Neural Network forecast method is verified. It concluded that, forecasting by ANN has the advantages of high accuracy, comprehensiveness, and reliability, even BP neural network has a slight advantage than RBF neural network, which is beneficial to improve the quality construction monitor control.