Intelligent Lithology Prediction While Drilling Using Engineering Parametes

Traditional lithology identification mainly relies on logging-while-drilling (LWD) data. However, the high operation costs and low transmission efficiency of LWD limit its field application To verify the feasibility of lithology identification using only drilling engineering parameters, true triaxial drilling experiments were conducted in the laboratory. Rock samples of different lithologies were drilled under varying weight-on-bit (WOB) and rotational speed conditions, during which engineering parameters such as WOB, rotary speed, and torque were collected and analyzed to reveal their response patterns to lithological differences. On this basis, an intelligent lithology prediction model while drilling was constructed using XGBoost, with engineering parameters as feature variables and lithology as classification labels, and its accuracy was evaluated. The experimental results show that under different WOB conditions, torque decreases with increasing rotary speed, while rotary speed and rate of penetration (ROP) exhibit a positive correlation. Under the same WOB, the ROP of low-hardness rocks is significantly higher than that of high-hardness rocks, and the increase in ROP with rotary speed is more pronounced for low-hardness rocks. The XGBoost lithology identification model, built on engineering parameters, achieves a recognition accuracy of over 90%. Studies have shown that the intelligent lithology identification method while drilling based on engineering parameters can accurately predict changes in lithology, offering a new approach for lithology logging in drilling operations.