(381f) Validation of a High-Performance Framework for Simulating the Hopper Discharge Flow of Milled Biomass

Variability of lignocellulosic biomass characteristics has been highlighted as one of the primary factors impacting feedstock bulk handling and transport, flowability, and ultimately integrated biorefinery (IBR) operational efficiency. Biomass characteristics including particle size, shape, and moisture content, which may vary considerably across common feedstocks, significantly affect flow behavior. When not accounted for, such variability can lead to unreliable IBR operation (e.g. flow arching, ratholing) and reduced production capacity, as many of the existing methods and tools developed for handling bulk solid flow were designed for relatively uniform particles with low compressibility. Numerical methods have emerged in recent years as an important tool for investigating such problems given their ability to provide detailed flow information for a wide range of problems. The discrete element method (DEM) is a widely used approach for simulating granular materials but suffers from high computational costs which often prevent its use in industry-scale simulations. In this study, we develop and apply an open-sourced GPU-accelerated DEM solver capable of performing high-fidelity large-scale simulations. The solver is used to investigate milled corn stover discharge from a wedge hopper. Glued and bonded sphere models, which represent particles as clusters of spheres that are either rigidly joined, or may bend, twist and deform, respectively, are incorporated to simulate variable particle shapes and sizes, while a liquid bridge model accounts for particle cohesion due to moisture content. Simulation parameters are obtained from a series of laboratory-scale experimental flow tests. The DEM solver is extensively validated by performing a series of axial compression tests followed by simulations of a pilot-scale wedge hopper pre-filled with corn stover. The discharge rate and arching distance are measured and compared against experimental results and are found to be in good agreement after careful parameter calibration. The validated model is then used to investigate the impact of important particle characteristics (e.g. particle size, shape, and moisture content) on bulk flow behavior. Finally, to assess the solver’s readiness to simulate industry-scale flow, the GPU speedup is compared against traditional CPU-based parallelism and it is observed that the performance enhancement when using GPU acceleration is significant, allowing for commercial-scale simulations of feedstock handling.