XRD AND MACHINE LEARNING FOR CLAY MINERAL CHARACTERIZATION IN SHALE GAS

• Main Author: Pratiwi, Ratih
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/68621
• Institution: Institut Teknologi Bandung

Shale gas is unconventional gas obtained from shale rock reservoirs that can be used as an alternative energy source for the declining world energy supply. Therefore, studies related to shale gas have become something important to do. One of the scopes of shale gas studies is the mineral content of shale rocks. In this study, the X-Ray Diffraction (XRD) technique is used to identify the mineral content of shale rocks. Machine learning is used to characterize and correlate the clay mineral content of XRD results as the main constituent of shale rocks between fields and between formations. The Support Vector Machine (SVM), Self-Organizing Map (SOM), and K-Means algorithms are used to classify and cluster clay facies from field-X and field-Y shale lithologies located in the Central Sumatra basin. The XRD results show that the percentage of clay in the field-X sample is dominated by illite and kaolinite, while the field-Y sample is dominated by illite and chlorite. The relationship between mineral content and porosity shows that the percentage of illite is inversely proportional to the porosity, while the percentage of kaolinite is directly proportional to the porosity. Data analysis shows that there is a fairly strong correlation value between the field-X and Y samples, which is worth 0.566. The SVM classification shows that about 50% of the feature pair dataset can separate the brown shale and upper red bed classes. Clustering with SOM and K-Means has been able to divide data points into two clusters although it is still not accurate because SOM and K-Means are unsupervised machine learning algorithms. In the future, it will need to be tested to perform clustering analysis using a larger number of clusters.