Prediction of Wood Pellet Bulk Density Using Geometrical and Mass Flow Features: A Preliminary Study

Bulk density (BD) is a key quality parameter of wood pellets, influencing handling, storage and volumetric energy density. The standard procedure (ISO 17828) is accurate but time-consuming and subject to operator variability. This preliminary study introduces the mass flow rate (MFR) as a new variable, not included in the standard, and presents a prototype built for its controlled measurement. Thirty commercial wood-pellet samples were characterized for MFR, pellet length distribution metrics and BD. After multivariate analysis, the relationships between MFR, dimensional descriptors and BD were explored through correlation analysis. As a first exploratory test, Support Vector Regression, Extreme Gradient Boosting and a weighted ensemble were evaluated by nested repeated cross-validation, with Gaussian noise-based augmentation applied to training data only. The mean out-of-fold MAE was 17.5 kg/m3 for the ensemble, 17.8 kg/m3 for SVR and 18.3 kg/m3 for XGB, with overlapping confidence intervals and considerable variability across data partitions. Nevertheless, the errors are contained and comparable to those attributable to operator variability and lower than the features standard deviation (24.3 kg/m3). The results suggest that MFR, combined with dimensional descriptors, can support BD estimation, although larger datasets are needed to establish a robust relationship.