Adaptive mesh refinement for a particle level simulation of supercritical water fluidized bed reactors

High-precision computational fluid dynamics-discrete element method (CFD-DEM) simulation serves as a foundation for the detailed design of supercritical water fluidized bed reactors (SCWFBRs). However, as the scale of these reactors increases, the computational domain and reacting particles expand exponentially. Traditional adaptive mesh refinement is constrained in dealing with dense phase zones in SCWFBRs, as the nonlinear characteristics of the dense reacting particle flow render local-field error estimation ineffective. To address this limitation, an efficient and accurate CFD-DEM framework is established, integrating adaptive mesh refinement and heterogeneous computing. The mesh adaptation strategy is based on whole-field error estimation and is defined by a set of cell-marking criteria. The effect of the criteria on the accuracy and efficiency of the computational process is evaluated. This method reduces the demand for computational resources, thereby enabling the simulation of large-scale reactors at the particle level. The optimized adaptation strategy maintains an error of 1.67% while saving 51.59% of mesh cells required and 28.4% computation time. The primary factors influencing computational accuracy are identified as fluid velocity gradient and the particle reaction rate. The scalability of the adaptation strategy is also validated. For a fivefold radial scaled-up reactor, the cell count is reduced by 46.86% and the computation time by 22.2%, while the maximum error is limited to 2.79%. This work provides a solution that can consider both the computational scale and accuracy, thereby enabling the detailed design of large-scale SCWFBRs.