Estimation of Uniaxial Compressive Strength Using General Regression Neural Network

Uniaxial compressive strength (UCS) represents the ultimate load-bearing capacity of intact rock material when subjected to unconfined compression along a single principal stress direction, with the specimen's lateral deformation unrestricted and free to expand. UCS serves as a fundamental index property for rock mass classification systems (e.g., RMR, Q-system) and empirical design of underground excavations. The direct measurement of UCS is time-consuming and costly. The General Regression Neural Network (GRNN) is employed here to predict UCS from data collected on ۳۶۳ rock samples from various locations worldwide. Given that the designed network was not exposed to the test data (۳۰% of the available dataset) during the training phase, the results indicate high accuracy of the developed GRNN model, with correlation coefficients (Rs) of approximately ۰.۹۸ for the training phase and ۰.۹۳ for the testing phase.

کلیدواژه ها:

Uniaxial compressive strength (UCS) ، Radial Basis Function Neural Network ، General Regression Neural Network (GRNN) ، Machine Learning