Prediction Method for Rock Drillability Level in Complex Formations Based on Local Weighted Regression Linear

Interbedded dense gas sand mudstones in Shaximiao formation, central Sichuan, exhibit abnormal drill bit wear and pose challenges in drill bit selection. Conventional methods for evaluating drilling feasibility suffer from limited parameters and poor reliability. To achieve accurate prediction of drilling feasibility levels, this study proposes a rock drilling feasibility level prediction method based on logging parameters such as acoustic travel time (AC), density (DEN), resistivity (RT), and natural gamma (GR), employing Gaussian weighted function combined with local weighted linear regression (LWLR). The research findings indicate that the model achieves the highest accuracy when the scaling factor k in the weight function is 0.03, with a goodness of fit of 0.98 and a root mean square error of 0.385. Comparative analysis with conventional empirical formulae and multivariate nonlinear fitting demonstrates that the proposed model outperforms them, with a goodness of fit higher than 0.962 for the former and 0.944 for the latter. Those indicate the effectiveness of the proposed model in predicting rock drilling feasibility levels and providing a theoretical basis for drill bit selection in the Shaximiao formation in central Sichuan.