RBF neural network with chaotic time series for deformation forecasting of underground powerhouse surrounding rocks
Wang Ji-wei · Engineering Journal of Wuhan University · 2009
Chaotic time series theory is introduced into solving the prediction question of nonlinear chaotic time series.The fractal dimension D and largest Lyapunov exponent λ1 that mean chaotic characterization are calculated;The neural network can overcome the shortcoming that the conventional models must be the combination of linearity and nonlinearity of imput data,and has the strong ability of self adaptive learning and remembering.Through combining chaos characteristic,one dimension time series is reconstructed into multidimensional phase spaces to optimize RBF neural network.The model RBF neural network with chaotic time series is combined.Results show the errors between prediction and measured values are all less than 6%,and post forecast deformation is descend with time which is fit with engineering fact.