Elastic modulus prediction of limestone using multiple variable regression and artificial neural network

It is important to determinate the elastic modulus of intact rock in many engineering projects relating to rock such as tunnels, slopes and foundations. Determination of this parameter need of high quality core samples and sophisticated test equipment because these make performing this test difficult and costly. There for in current years, researches have tried to produce relationships for prediction of this parameter using physical and index characteristics of rocks. So here an attempt has been made to predicate the elastic modulus of limestone by uniaxial compress strength, porosity and longitudinal wave velocity. So we used multiple variable regression method and perceptron artificial neural network with 3-4-1 architecture. Data base contains 123 samples that 70% and 30% of them has been used for train and test respectively in ANN. The models have been compared using coefficients of R2, RMSE and VAF. The coefficient of determination (R2) using multiple variable regression was 0.738. it was 0.805 and 0.832 for ANN train data set and ANN test data set respectively. This shows the ANN method have more accuracy than regression.