Solving the bottleneck of industrial-scale biomass feeding: An experimental-informed modeling study of pellet flow

Herbaceous energy crops like miscanthus offer significant potential for bioenergy but suffer from poor flowability during material handling and feeding, leading to costly process upsets such as jamming and clogging. This study investigates the use of a densification (i.e., pelletization) method as an additional preprocessing step after milling to overcome these challenges, comparing the flow performance of miscanthus ( Miscanthus × giganteus ) pellets against baseline milled miscanthus. A pilot-scale ring die pellet mill was used to produce pellets from 6-mm milled miscanthus, which were then characterized for physical properties including diameter, length, density, and stiffness. Using a physical experiment-informed modeling approach, the study developed discrete element method simulations to analyze the criticality of pellet properties (aspect ratio, stiffness, friction) and wedge hopper geometries (wall inclination angle, opening width) on discharge flow rate. Simulations revealed that initial packing of pellets can occasionally induce jamming but is only likely with long pellets. Results demonstrated that pellet length and hopper opening are the most influential parameters, and pellets, especially short, stiff pellets, significantly outperformed milled miscanthus, achieving continuous, stable flow rates exceeding the industrial target of 2000 metric tons/day (or 83 metric tons/hour). More importantly, hopper flow rate of pellets can be accurately controlled with outlet opening for design needs. A deep neural network model was subsequently trained on the simulation data to create a predictive design chart. The findings confirm that pelletization effectively eliminates downtime risks associated with herbaceous biomass, validating its benefit as a promising process for scalable biorefinery operations. • Herbaceous biomass feeding problem is addressed using a computational framework. • A calibrated DEM model accurately captures miscanthus pellet flow behavior. • Deep neural networks were trained on extensive DEM simulation datasets. • The DNN surrogate model accurately predicts hopper feeding flow rate of pellets. • Pelletization is essential to eliminate downtime risks despite extra processing cost.