The time series prediction of tunnel surrounding rock deformation based on FA-NAR dynamic neural network

The deformations of tunnels surrounding rock have the characteristics of dynamicity, sensitivity to time and space, nonlinearity, high complexity, etc. In order to improve the prediction accuracy of the surrounding rock deformation, the firefly algorithm(FA) was used to determine the delay order and number of hidden layer elements. And the prediction was carried out by nonlinear auto-regressive(NAR) dynamic neural network. A prediction model of tunnel surrounding rock deformation based on FA-NAR dynamic neural network was proposed. The deformation monitoring data of Beishan exploration tunnel was used to predict. And the prediction results of FA-NAR dynamic neural network algorithm were compared and analyzed with the BP neural network algorithm. It shows that the predicted values of FA-NAR dynamic neural network are basically consistent with the measured values. Its mean absolute error and mean relative error are around 1/5 and 1/4 of BP neural network respectively, which proves that the FA-NAR dynamic neural network algorithm model has higher prediction accuracy than the BP neural network algorithm model. The FA-NAR dynamic neural network algorithm model can solve the problem of surrounding rock deformation prediction well. It not only reduces the blindness of human input network parameters, but also improves the learning ability and prediction accuracy of the network. © 2019, Science Press. All right reserved.