Predicting uniaxial compressive strength using Support Vector Machine algorithm

Abstract

Compressive strength is the most important parameter in rock since all loads will be transferred and rest on the rock which is based on the load bearing capacity of rock in compression. However, obtaining the compressive strength or mostly measured, the uniaxial compressive strength (UCS) from the laboratory test requires certain standard and also cost constrain. This paper presents the application of Support Vector Machine (SVM) algorithm to predict the UCS. An algorithm has been tested on a series of rock data using dry density and velocity parameters. The relationship between the dry density, sonic velocity, and UCS was analyzed using RapidMiner Studio software. From the result, it was found that SVM is capable of predicting the missing values with a prediction trend accuracy of 75%. The results obtained and observation made in this study suggests that SVM could be a reliable tool to predict the UCS of a given rock. More robust prediction can be established with bigger sample number. It is worth mentioning, that the program module that has been set up could be used repeatedly for other correlation problems.