CORRELATION AND CLASTERING OF SUMATRA SHALE CLAY FACIES USING X-RAY DIFFRACTION SPECTRUM AND MACHINE LEARNING

• Main Author: Cindrawati, Maria
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/68642
• Institution: Institut Teknologi Bandung

Shale gas is an unconventional hydrocarbon energy that is formed and trapped in the surface of the shale formation matrix. The purpose of this study was to correlate and cluster clay facies in shale rocks in the X and Y fields. This study included analyzing of clay minerals by X-ray diffraction, calculating the relationship between porosity and mineral clay percentage, and quantifying the correlation between samples by machine learning. The percentage of clay minerals in field X is dominated by Illite and Kaolinite minerals, while field Y is dominated by Illite and Chlorite minerals. The correlation shows that the relationship between field samples X and Y has a strong correlation, the correlation value is 0.556 for samples 8 and 14 of the Brownshale Formation. The relationship between porosity and mineral content shows that Illite percentage is inversely proportional to porosity, while Kaolinite is directly proportional to porosity. The classification of mineral percentages with SVM shows that clay at long distances has a strong correlation of more than 50%, while between formations there is also an overlapping correlation of around 50%. The analysis of the number of clusters using K-Means with elbow and silhouette score methods shows that the number of clusters of two is still less representative for differentiating formation classes. This is because K-Means is an unsupervised machine learning algorithm that can only determine the number of clusters based on the user's wishes. Class analysis using PCA shows an intersection between the two classes, which is about 50% of the total sample data from different fields and formations. Analysis of data processing on correlation values and machine learning shows that samples from different fields and formations have a correlation of around 50%. This shows that the clay mineral content between the two formations still has a similarity of around 50%.