Exploring rock mass and machine interaction is not a straightforward task due to complexity and large variability of ground, geology and rock mass conditions even in short intervals. Researchers, however, rarely employ non-linear models to examine the determinants and make little effort to identify a superior prediction model. The main objective of this paper is to compare two TBM performance estimation models with especial task of determining Rate of Penetration (ROP), using Artificial Neural Network (ANN) and multiple linear regression. These approaches were applied to a database compiled from 121 tunnel sections along the Milyang hydro-tunnel for correlating ROP with basic RMR input parameters. Nonetheless, the poor relationship between ground water condition and ROP resulted in exclusion of this parameter from the further analysis and Uniaxial Compressive Strength (UCS), RQD, joint spacing and joint condition were used. The accuracy of developed model was discussed by some statistical measures including correlation coefficient (R), the Root Mean Squared Error (RMSE), the accuracy factor (Af), and the bias.

Results of this study showed that ANN model has lower RMSE values for in-sample and out-of-sample forecasting. This indicates that the non-linear ANN model generates a better ft and predict of the panel data set than the regression model, and ANN is capable of catching sophisticated non-linear integrating effects in TBM performance problem.

This content is only available via PDF.
You can access this article if you purchase or spend a download.