Predicting the impact of short-duration densification on timber set-recovery using generalized additive model and transformed ordinary least squares regression

Abstract This study examined the effects of varying compression ratios (40%, 50%, and 60%) applied during the short-duration densification on the set-recovery and density of laminas of Paraserianthes falcataria , a low-density timber species. The densification process was comprised of pressing, venting, and cooling phases. Kruskal-Wallis and Dunn tests evaluated the statistical differences in set-recovery and density between different compression ratios. Modelling approaches, such as Generalized Additive Model (GAM) and Transformed Ordinary Least Squares (TOLS) regression, were developed to predict set-recovery outcomes across a broader compression ratio range (10%–80%). The non-parametric tests revealed that increased compression ratios significantly influence set-recovery and density, with a minimal increase in set-recovery and a consistent density over time. As for the dual-modelling approaches, GAM outperformed TOLS in all the model metrics. Nevertheless, both models obtained low metrics values, which were indicative of high reliability and accuracy. The overall pattern from both models suggests that set-recovery increased as the compression ratios increased. These insights are crucial for enhancing sustainable timber production, indicating potential advancements in manufacturing high-performance, eco-friendly wood materials.