Multi-Offset Recurrent Neural Network Model for Displacement Prediction of High Wall Rock Mass

It´s an important research project to forecast deformation of high wall of underground house during designing and constructing. The neural network is optimized and the multi-offset recurrent neural network is built to predict deformation. The maximum predictable number of days is calculated by calculating the maximum Lyapunov exponent λ₁, and the structure of neural network is optimized through chaotic characteristics. The example shows that the errors between prediction values and measuring ones are all no more than 10%, so the precision is high and results are credible on real time.