Forecasting coal and rock dynamic disaster based on adaptive neuro-fuzzy inference system

Abstract

Forecasting model of coal rock electromagnetic radiation was built with combining time series analyze and adaptive neuro-fuzzy inference system (ANFIS). In the first, coal rock electromagnetic radiation phase space was reconstructed through Takens theory, and time delay and embedding dimension are determined by mutual information method and false nearest neighbor method respectively. Then, the forecasting model of coal rock electromagnetic radiation was constructed via ANFIS in the reconstruction phase space, and the parameters of ANFIS are tuned by hybrid learning algorithm. Finally, the simulation results and comparison analysis are presented, the training and checking root mean squared error are 0.0248 and 0.0286 respectively, which indicates that the ANFIS has better learning ability and generalization performance, thus, the model is creditable and feasible. © 2010 Springer-Verlag Berlin Heidelberg.