Abstract: Surrounding rock pressure is a one of significant feedback information after tunnel excavation. Affected by excavation and other uncertain factors, the monitoring data of rock pressure is an unsmooth time series, which includes trend term and random part. The trend part of the data can be fitted with BP(back propagation) neural network and the random part is processed by a normal ARMA(auto regressive moving average) model. Combined with a progression arithmetic, time series analysis of rock pressure in a tunnel in Zhejiang Province is carried out. The results show that the method has a high prediction accuracy, is in a maximum relative error of 3.73%, and can be used in practical engineering.

Key words: artificial neural network, ARMA, time series, surrounding rock pressure, progression arithmetic