Authors: Peter Blaser, Andrew Larson, James Parker, Ali Akhavan, Niraj Mehta
CPFD Software, 1255 Enclave Parkway, Houston, Texas, 77077, USA
Source: This paper was published in FBC-24.
Abstract: Simulations of fluidized bed systems are inherently complex, involving multiphase hydrodynamic, thermal, and chemical reaction mechanisms. Additionally, fluidized beds are utilized for a broad range of end-use processes spanning traditional and novel industries and application areas. As such, the role of the particulate phase also varies, with the particles serving diverse purposes including mixing, heat transfer, catalysis, reactant, or product, to name a few. Due to these complexities, simulation has had a smaller impact on the commercialization, scale-up, and troubleshooting of fluidized bed conversion processes than that realized in other chemical engineering applications.
GPU computing, originally used for computer graphics, is at the center of a parallel computing revolution that has occurred over the last decade. Until the 2010s, parallel computing was almost exclusively undertaken using clusters of computers, and later multiple cores on those computers or clusters thereof, collectively called CPU parallelization. Modern GPUs have thousands of compute cores (compared with tens in a CPU), and are rapidly overtaking CPU clusters for HPC applications and artificial intelligence / machine learning (AI/ML) tasks.
This paper explores how advances in GPU and multi-GPU computing have impacted 3D, transient simulations of fluidized bed conversion processes, resulting in simulations running up to 400x faster using GPUs in a single workstation compared with CPU-only performance. While speed enables faster solutions, and marginally faster time-to-market for new technologies, the real benefit is that previously intractable problems are now possible to solve with meaningful resolution (spatial, temporal, physical model complexity, chemical reaction mechanism detail).
Sample case studies using the commercial Barracuda Virtual Reactor® software are shown for applications including the gasification of municipal solid waste streams, chemical looping combustion, and fluidized catalytic cracking. Implications on the breadth of R&D activities, and the acceleration of technology development, commercialization, scale-up, and IP protection are discussed.