Analysis of Neural Network Models in Prediction of Ground Surface Settlement Around Deep Foundation Pit

摘要

During the foundation pit excavation, the prediction of ground surface settlement around deep foundation pit is directly related to the safety of the foundation pit excavation, surrounding buildings and pipelines, but the ground surface settlement of foundation pit has the characteristics of nonlinear and fuzzy. So it is necessary to monitor and predict the excavation settlement according to the excavation conditions, the surrounding environment, security level and other buildings around. Neural network can simulate any unknown system of complex polygene conveniently and high precision. GRNN and two improved BP neural network prediction models are established to predict settlement in this paper. The ground surface settlement around a deep foundation pit is predicted with all main influential factors being taken into account properly. The three neural network prediction models - GRNN, PSO-BP and GA-BP prediction model are analyzed in principle and network architecture design. And they are used to predict ground surface settlement for an engineering example in Beijing. The prediction results show that neural network have high feasibility and reliability in predicting ground surface settlement around deep foundation pit, and neural network will have better application prospect in the field of geotechnical in-situ testing & monitoring.