Artificial intelligent techniques for prediction of rock strength and deformation properties-A review

In rock design projects, a number of mechanical properties are frequently employed, particularly unconfined compressive strength (UCS) and deformation (E). The researchers attempt to conduct an indirect investigation since direct measurement of UCS and E is time-consuming, expensive, and requires more expertise and methodologies. Recent and past studies investigate the UCS and E from rock index tests mainly P-wave velocity (Vp), slake durability index, Density, Shore hardness, Schmidt hammer Rebound number (Rn), unit weight, porosity (e) point load strength (Is(50)), and block punch strength index test as its economical and easy to use. The evaluation of these properties is the essential input into modern design methods that routinely adopt some form of numerical modeling, such as machine learning (ML), Artificial Neural Networking (ANN), finite element modeling (FEM), and finite difference methods. Besides, several researchers evaluate the correlation between the input parameters using statistical analysis tools before using them for intelligent techniques. The current study compared the results of laboratory tests, statistical analysis, and intelligent techniques for UCS and E estimation including ANN and adaptive neuro-fuzzy inference system (ANFIS), Genetic Programming (GP), Genetic Expression Programming (GEP), and hybrid models. Following the execution of the relevant models, numerous performance indicators, such as root mean squared error, coefficient of determination (R2), variance account for, and overall ranking, are reviewed to choose the best model and compare the acquired results. Based on the current review, it is concluded that the same rock types from different countries show different mechanical properties due to weathering, size, texture, mineral composition, and temperature. For instance, in the UCS of strong rock (granite) in Spain, ranges from 24 MPa to 278 MPa, whereas in Malaysian rocks, it shows 39 MPa to 212 MPa. On the other side, the coefficient of determination (R2) correlation for the UCS also varies from country to country; while using different modern techniques, the R2 values improved. Finally, recommendations on material properties and modern techniques have been suggested.

Automated summary

In rock design projects, mechanical properties such as unconfined compressive strength (UCS) and deformation (E) are frequently used. Due to the challenges of direct measurement, researchers often rely on indirect investigations using rock index tests. These properties play an essential role in modern design methods involving numerical modeling techniques. The current study compares laboratory tests, statistical analysis, and intelligent techniques for estimating UCS and E, and highlights the importance of considering variations in rock types and applying modern techniques to improve accuracy.