چكيده لاتين
Despite the widespread use of tunnel boring machines (TBMs) in tunneling projects, accurately estimating machine performance, particularly under complex geological conditions, remains challenging. Prediction of TBM performance is a key factor for planning, cost estimation/control, and selecting proper machine specifications. Over the past decades, numerous models have been proposed to predict TBM performance. However, due to the involvement of various factors including intact rock and rock mass properties as well as machine operational and design characteristics, no comprehensive model has been developed that considers all influential factors.
Given that different rock mass classification systems are commonly used in many rock engineering projects due to their simplicity, global acceptance, and the availability of effective parameters, they are suitable methods for estimating TBM performance. Among the most commonly used rock mass classification systems, the RMR (Rock Mass Rating) shows a better correlation with TBM penetration rates. This is because of using the uniaxial compressive strength (UCS) as an input parameter in the RMR classification system. The objective of this study is to modify the RMR classification system to develop a new model for predicting TBM performance in hard rock using statistical analyses and artificial intelligence algorithms. For this purpose, data from 10 tunneling projects with different geological conditions and machine types have been obtained from pertinent research groups and compiled in a database. This database includes geological and geotechnical characteristics of the rock mass and operation data and the actual performance of tunnel sections. Various statistical analyses were first conducted to evaluate the relationship between rock mass engineering geological properties and TBM performance. The possibility of using the RMR input parameters to develop performance estimation relationships using ML-based regression analysis was then examined, and new relationships for estimating the Field Penetration Index (FPI) based on the RMR input parameters were proposed. Given that different rock types have varying textures, structures, and mineral compositions, and respond differently to machine shear forces, the effect of rock type was also incorporated into the machine performance prediction relationships. These relationships can be particularly useful in the design and planning stages of a tunneling project.
Furthermore, since the RMR classification system was developed for design applications involving the estimation of rock mass properties, stability conditions, and tunnel support, its input parameter rating is also based on this purpose. This might be the reason for the low correlation between RMR values and TBM performance. Therefore, it seems possible to achieve an optimized RMR by adjusting the RMR input parameters rating and the internal weighting of each parameter, tailored to the aim of this study, which is to predict TBM performance in hard rock. The approach used to develop this optimized RMR involves using nonlinear regression algorithms, decision tree and Random Forest, Ultimately, the RMRTBM system is proposed to predict TBM performance in hard rock.
