Time Series Forecast of Foundation Pit Deformation Based on PSO-LSSVM
Jin Cao · Control Engineering of China · 2015
The deform data of foundation pits acquired by field measurement contains internal mechanics evolutionary information. Aiming at the complexity of influential factors of foundation pit deformation, the highly nonlinear monitoring data and extra learning problem of the artificial neural network, a time series forecast method of foundation pit deformation based on PSO-LSSVM model is proposed. Wherein, the parameter of least squares support vector machine is optimized by particle swarm and the phase space reconstruction theory is used to preprocess data. Through analyzing the measured data, the deformation forecast model is established and applied to the dynamic design and information construction, which is of great significance to ensure the safety of foundation pit. This method has been successfully applied in the deep horizontal displacement forecast of one foundation pit in Kunming. Through constantly using the measured data of early working stages, the model is established, and then the later working stage deformation is forecasted and satisfying results are obtained.