| description abstract | Abstract. Ambient drying is a low-cost and scalable process where moisture evaporates under room temperature and humidity conditions, making it attractive for biomass material manufacturing due to its low energy consumption and ability to preserve structural integrity. However, ambient drying is often characterized by variable drying kinetics and prolonged drying time, which complicates process monitoring and prediction. In this study, we model the temporal evolution of biomass weight during ambient drying using a gravimetric, data-driven approach. A real-time gravimetric sensing setup is custom-built to collect dynamic weight changes. In particular, the effects of cellulose content and fan speed are investigated, as they play critical roles in determining the initial mass and the drying rate. Subsequently, building upon an asymptotic drying formulation, a hierarchical nonlinear mixed-effects (NLME) model with a curvature parameter is proposed to capture both population-level drying trends and sample-specific heterogeneity. Quantitatively, the proposed NLME model reduces the prediction mean squared error by approximately 20% compared with conventional nonlinear regression models. Furthermore, by integrating a Bayesian updating mechanism, the prediction error is reduced by over 60% as real-time measurements are incorporated during the ongoing drying process. Beyond predictive accuracy, the proposed framework provides interpretable insights into how material composition and airflow conditions influence initial weight, drying rate, and moisture retention behavior. The modeling approach is adaptable to new biomass formulations and drying conditions, offering a generalizable and data-efficient solution for monitoring, prediction, and future control of ambient biomass drying processes. | |
