Enhancing of clay volume prediction using machine learning Techniques and well logging data

Abstract

The values related to rock properties, such as volume shale (VSH), are considered fundamental indicators in the characterization of geological formations during petroleum operations. These values are usually measured using logging tools during drilling or through precise laboratory techniques. However, these methods are not always available and are often costly in terms of time and resources. This necessitates the use of alternative solutions based on artificial intelligence that provide accurate results in less time and at lower cost.This study aims to apply predictive models using machine learning algorithms to estimate shale volume in the reservoir based on a limited number of input variables. The focus was placed on using only two inputs: TNPH and RHOZ, and training different models such as XGBoost, Random Forest, and SVR.The performance of the models was evaluated using three main metrics: Mean Absolute Error (MAE), Mean Squared Error (MSE), and the Coefficient of Determination (R²). The results showed that the Random Forest model with a tree depth of 3 achieved the best performance with R² = = 0.7535, while the Linear Regression model produced stable and acceptable results. On the other hand, the XGBoost model did not perform as expected, indicating the importance of matching the model to the nature of the data.This study confirms the effectiveness of artificial intelligence techniques in supporting the prediction of subsurface formation properties and opens the door for their future use in improving geological characterization accuracy and reducing reliance on traditional, high-cost methods.