Research On Prediction Model Based On Chaotic Time Series For Shield Segment Structure Stress

Shield construction at some important area like starting district, the air shaft and contacting gallery might cause some changes at segment stress, leading to deformation of segment. Even more, some cracks might appear at bad situation, and causing disaster like phenomenon of water burst and sand boiled in the tunnel. With the development of monitoring technology, construction safety has been guaranteed to a limited extent. But in the case of the risk response mechanism is not perfect, using forecast model to get more risk response time can make it easier to ensure the construction safety and reduce the loss caused by the accident.Based on engineering data collected in Line No.4 of Wuhan Metro, the general rules of segment structure stress timing change is pointed out in this thesis. Segment structure stress has been proved to have chaos property from qualitative and quantitative angles.Combining chaotic theory and BP neural network, a chaotic time series forecasting model is constructed to predict segment structure stress.2000 groups of segment structure stress data is randomly assigned to calculate the important parameters of segment structure stress chaotic time series based on MATLAB.The embedded dimension m=5 and delay time ?=3 obtained by C-C method are used to reconstruct phase space. The result of maximum Lyapunov index of the chaos time sequence calculated by Wolf method is ?=0.6487>0, which proves that the time series have chaos property. The reconstructed phase space provides input and output data, input layer number for the BP neural network. After the training, using MSE, MPE, maximum and minimum absolute error, maximum and minimum percentage error to measure the accuracy of result. The results MSE=0.0313<0.05, MPE=0.042%<0.1% show that the model of chaos time sequence prediction can assure a good accuracy of the prediction of duct piece structure stress.