DEVELOPMENT EMPIRICAL MODELS FOR ESTIMATING TBM PENETRATION RATE IN ROCK MASS

Abstract

Estimating the Rate of Penetration (ROP) of Tunnel Boring Machines (TBMs) is crucial in rock excavation and boreability, as it serves as a key input for scheduling and executing tunneling projects. Despite advancements in tunneling technologies, achieving a reliable ROP estimate remains challenging. In present one of the difficult tasks is to quantifying the properties of rock discontinuities to be use as input for the TBM performance models. Another issue is that utilizing the combination of intact rock, rock mass properties and TBM specification into the TBM performance prediction models. This thesis aims to develop several empirical models using intact rock, rock mass, disc cutter and TBM specifications for predicting TBM penetration rates in hard rock conditions. In order to achieve this goal, extensive field and laboratory tests were conducted across six TBM projects, including the Queens Water Tunnel (USA), Manapouri hydropower tunnel (New Zealand), Miryang dam tunnel (South Korea), and three Iranian water tunnels (Karaj, Zagros and Ghomrood), resulting in a database of more than 550 cases. Each case consists of the intact rock, rock mass properties and TBM field parameters. The collected data underwent preprocessing and analysis. The raw data were thoroughly examined with the literature to ensure the research output was valuable and acceptable. Common TBM performance models used by researchers and industry were critically reviewed. In order to develop the models, first of all, a rock mass fracture index (FI) was introduced using a weighting method. Then, numerous linear and nonlinear multivariable regression models were developed using the SPSS program. At least a hundred models were generated for the aim; then the best twenty of them were chosen for each case by means of the linear multiple regression (LMR) and Non-linear multiple regression (NLMR) to be evaluated and introduced. After examination of LMR and NLMR models, with the top five models evaluated using performance indices and a total ranking approach introduced in the thesis, the rest of them are provided in the appendices. Six statistical measures—coefficient of determination, root mean square error, mean absolute deviation, mean absolute percentage error, relative root mean square error, and variance accounted for—were used to rank the models. The study confirmed that estimating TBM penetration rate is a complex, non-linear issue due to the heterogeneous nature of rock masses. The most accurate model was a NLMR with fewer input variables, achieving the highest-ranking score of 42 (Eq.5.17; Model 1-NLMR). This model, incorporating intact rock strength, a weighted fracture index, and operational machine parameters, was ultimately identified as a reliable tool for TBM tunneling. Finally, it is concluded that the introduced rock fracture index (FI) together with weighted fracture index (WFI) could be used for quantifying the discontinues of rock mass into the TBM performance prediction model to represent the rock mass parameters. Further, it is found that the developed model could be used for similar rock type and geological formation in practice; More, it is suggested that the alpha angle, that is the angle between plane of weakness including joint, faults and shear zone and TBM driving direction, could be added as an input into the model to obtain the more accurate result for project scheduling in practice.