Quantitative Inversion of Mean Grain Size from Conventional Well Logs Using Complete Ensemble Empirical Mode Decomposition （CEEMDAN）and Principal Component Analysis（PCA）

Mean grain size (Mz) is a key indicator of depositional processes and reservoir quality, yet its continuous characterization is commonly limited by the availability of core or cuttings data. This study presents a new method for predicting Mz from conventional well logs by integrating Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) and Principal Component Analysis (PCA). A density–neutron separation parameter is first constructed to capture lithological and grain-framework variations. Multi-scale components sensitive to grain-size changes are then extracted from this parameter using CEEMDAN, and the dominant controlling features are identified through PCA. The resulting model enables continuous Mz prediction along the wellbore. Comparisons with measured data demonstrate that the proposed approach reliably captures vertical grain-size variations, providing a practical and robust solution for quantitative grain-size characterization and supporting detailed reservoir analysis and geological modeling.