A combination model of multiple regression and rock engineering systems (MR-RES) to identify main parameters effect value on tunnel face advance

To accurately predict the advance of a tunnel excavated by the drilling and blasting method, various parameters related to the rock and the operational conditions of the project should be taken into account. In this paper, comprehensive model was developed a to investigate the effects of different parameters on the advancement of such a tunnel. To achieve this goal, we conducted a systematic study at the tailrace tunnel of the Azad Dam in Iran. Rock properties, including the rock mass rating (RMR) and tunneling quality index (Q), as well as operational conditions such as blasting specific charge (q) and tunnel face area (A), were measured to establish comprehensive datasets for prediction. A total of 86 tunneling data points were collected and considered in this study. A novel model was developed, combining multiple regression (MR) and rock engineering systems (RES), to estimate tunnel face advance. The RES coding method was improved by incorporating a multiple regression model. The proposed coding method creatively assesses the influencing parameters, providing the advantage of accommodating uncertainties in the RES analysis. It achieves this by modeling the relationship between the explanatory (independent) variables and response (dependent) variables, thereby quantifying the interaction matrix. To evaluate the accuracy of the proposed models for both MR and RES datasets, we used the coefficient of determination (R2), a significant statistical criterion. A comparison of the values predicted by the models demonstrated that RES offers a more suitable performance than MLR for predicting tunnel advance. Sensitivity analysis of the MR-RES models reveals that the effective parameters on tunnel advance, in descending order of influence, are RMR (35.62%), Q (28.6%), q (20.35%), and A (15.42%). This hybrid method can be developed in other field of engineering without human judgment and consider statistical background of the data.