Study On Dynamic Prediction Of Monitoring Information For Deep Foundation Pit Based On ICBP Network Neural And Genetic Algorithm

Monitoring, design and construction are the three basic factors of guaranteeing foundation pit's safety. Dynamic design and information construction method become more and more universal. The key of this technology is the prediction of the following which using the prophase construction information. It is often unable to get accurate numerical solutions by tradition prediction method for its own limitation and the uncertain factors in the course of constructing. Network neural is fit for dealing with nonlinear problems especially for its good ability for nonlinear mapping.ICBP (Improved Circular Backpropagation) neural network which was advanced based on the BP neural network is applied to forecast analysis. In order to improve the structure of ICBP neural network, genetic algorithm is applied. Two forecast models are set up based on the ICBP neural network and genetic algorithm. One is based on the landscape orientation extend theory which set up unlined relation among parameters using known dates, the other is based on the time serial theory which considering discounted least squares(DLS) theory, i.e. system state later phase is forecast by former information. At the same time, the mufti-step predictor is applied to achieve dynamic prediction.Freezing and role-pole method is used in Runyang Bridge Nancha South deep foundation pit firstly in China. The load is complex during the pit excavation. Analysis and prediction of monitoring information is the important guarantee for engineering safety. Two models are both applied in the pit's forecasting analysis. The displacement of supporting structure, the pore water pressure and the supporting force are all forecasted. And the results show that the method is viable. According to the dynamic prediction of the monitoring information, the Naomao deep foundation pit is stable all the time after being reinforced.