Investigation on the Nonlinear Time Series Predication of Monitoring Data in Geotechnical Engineering

According to the characteristic of monitoring data in geotechnical engineering, three nonlinear time series prediction models: whole-region methods (linear function, exponential function, Gauss function and Fourier function), local-region methods (linear function, power function and exponential function) and chaos neural network based on the local-region method are built up by introducing the theory of nonlinear time series. These methods are applied to predict the displacement of outer monitoring point TP/BM27 in the 17-17 section in high slope of Three Gorges permanent ship lock. The result indicates that the deviation between the prediction displacement in three models and monitoring data is small and the law of the prediction displacement in whole-region methods is incompletely consistent with that of observation displacement and chaos neural network based on the local-region method is better than the whole-region methods.