This is the authors' abstract. We don't add key points for this paper.
Compressional sonic logs is one of the important logs for subsurface characterization, reservoir evaluation, and wellbore stability analysis. However, acquiring these logs is often challenging due to logistical constraints. This study explores the application of machine learning (ML) techniques to predict compressional sonic logs using conventional well logs from five wells. The methodology involves data preprocessing, feature selection, and training various regression models, including Random Forest, CatBoost, XGBoost, K-Nearest Neighbors (KNN), Support Vector Machines (SVM), and Deep Neural Networks (DNN). Model performance is optimized through hyperparameter tuning and evaluated using correlation coefficients and root mean square error (RMSE) metrics. Results indicate that ensemble models (Random Forest, CatBoost, XGBoost) achieve the highest accuracy, with correlation coefficients ranging from 89 to 89.6% and RMSE between 5.85 and 6.03. Additionally, feature engineering and data cleaning significantly improve model performance, while input scaling is essential for SVM, KNN, and DNN models. Incorporating blind well testing further enhances reliability. This study presents a robust ML-based workflow for predicting compressional sonic logs, offering a cost-effective solution for reservoir management and geomechanical analysis.
Answers are grounded in this paper and cite their sources.
No takes yet. Share an insight, caveat, or question.
Saleh et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: