A hybrid modelling approach for prediction of UCS of rock materials

Abstract

It is necessary to make an accurate assessment of uniaxial compressive strength (UCS) for rock mass classification and rock engineering design. However, there are many shortcomings in the conventional tests for UCS of rocks. The aim of this study is to present a hybrid model by integrating the genetic algorithm (GA) into the least-squares support vector machine (LSSVM) to predict the UCS of rock materials. The GA technique was utilized to improve the forecasting accuracy of the proposed LSSVM. To develop the proposed hybrid GA-LSSVM model, four main factors including the block punch index, point load strength, Schmidt rebound hardness and ultrasonic P-wave velocity were considered as input variables, while the UCS of rock materials was the output. A comparison was conducted among the proposed GA-LSSVM, the adaptive neurofuzzy inference system, the fuzzy inference system, the artificial neural network and the statistical method in accordance with three statistical indexes. The results of the comparisons show that the developed GA-LSSVM model has great potential to accurately estimate the UCS of rock materials.