Modeling rock mass deformation modulus using adaptive techniques

The deformation modulus (Em) of rock mass is an important parameter used in designing underground excavations. It models the mechanical response of rock mass due to excavation and can be determined directly using large scale in-situ tests which are often time consuming and expensive. To overcome this issue, several empirical equations are usually employed. However, these existing equations are suitable for certain types of rock masses posing limitations. Therefore, this paper intends to investigate alternatives for estimating the Em using adaptive techniques namely, the Adaptive Neuro-fuzzy Inference systems (ANFIS) and Multivariate Adaptive Regression Spline (MARS). Available data on the Em was employed to establish the models. The input parameters used to develop the models included the uniaxial compression strength, rock quality designation, discontinuity characteristics and the rock mass rating index. The performances of proposed models were evaluated using various performance indices namely the variance account for (VAF), root-mean square error (RMSE), and the coefficient of determination (R²). The results indicated good accuracy. Overall, the MARS model showed lower performance compared with the ANFIS model but the MARS model was able to produce easy-to-interpret.