Machine Learning–Based Real‐Time Lithology Identification From Drilling Data in the Ca Tam Offshore Field

ABSTRACT Accurate real‐time identification of lithological layers during drilling is essential for optimizing borehole stability, reducing operational risks, and improving reservoir characterization. This study develops a data‐driven approach for immediate lithology classification using standard drilling parameters combined with supervised machine learning (ML) techniques. The method was applied to time‐series datasets from wells in the Ca Tam oil field, offshore Vietnam—an area characterized by complex and heterogeneous stratigraphy. Four drilling parameters were selected as model inputs: weight on bit (WOB), torque, standpipe pressure (SPP), and rate of penetration (ROP). After preprocessing, including noise and outlier removal using the modified Z ‐score method, four ML algorithms were trained and evaluated to classify lithology into three categories: sand, claystone, and shale. The tested models included Random Forest (RF), Extreme Gradient Boosting (XGBoost), support vector machine (SVM), and artificial neural network (ANN). Among these algorithms, the RF model delivered the highest and most consistent performance, achieving an accuracy of 0.873 on the test set, slightly outperforming XGBoost (0.87). The small gap between RF's training and testing accuracy indicates strong generalization ability with minimal overfitting. Importantly, when independently validated on the blind well CT‐108, the RF model maintained high reliability with an accuracy of 84.23%, confirming its robustness under real drilling conditions. The results demonstrate that the RF model is highly suitable for deployment in real‐time lithology prediction systems. Such a predictive framework enables drilling operators to proactively adjust key parameters (e.g., WOB and rotary speed [RPM]), improving operational efficiency, enhancing borehole stability, mitigating risks such as stuck pipe incidents, and significantly reducing drilling time and costs in the Cuu Long Basin.