Mohammadpour, Mobarakeh1, Arashpour, Mehrdad1, Roshan, Hamid2 and Masoumi, Hossein1
1Department of Civil Engineering, Monash University, Melbourne, Australia 2School of Mineral and Energy Resources Engineering, UNSW, Sydney, Australia
Geophysical logs have been routinely performed in coal mines for many years. Compressional velocity is one of the important characteristics which can be measured using sonic log. Despite importance of P-wave velocity and its application in geophysical and geomechanical studies; some boreholes in coal mines do not have P-wave velocity or sonic log in their log suits; as a result, empirical correlations were commonly used to estimate P-wave velocity. However, these models are mostly local correlations which were derived for specific areas.
In this study Machine Learning based Gaussian Processes Regression was used to predict the P-wave velocity. Gamma, two density logs with different resolutions (Long Spaced Density and Short Spaced Density) and depth were applied as the input parameters. These three logs are the most common ones which are extracted from the coal mines. The model was generated using the data obtained from six boreholes in one of the Australian coal mines in Queensland. The data were divided into two groups including 35382 points for the training of the model and 11794 points for the testing. Root mean square deviation (RSME) and coefficient of determination (R2) were calculated to evaluate the accuracy of the proposed model.
Mobarakeh Mohammadpour is a PhD student at the at the Department of Civil Engineering, Monash University.