[论文解读] Extremely fast simulations of heat transfer in fluidized beds

T. Lichtenegger, Stefan Priker|arXiv (Cornell University)|Jun 2, 2017

Granular flow and fluidized beds参考文献 8被引用 4

一句话总结

本文提出循环CFD（rCFD）方法，通过利用递归图识别的重复流动模式，实现流化床中传热过程的极快速模拟。通过外推这些模式，该方法在保持预测颗粒在冷却空气中流化过程中温度演变的高精度的同时，相比传统CFD-DEM模拟实现了约100倍的加速。

ABSTRACT

Besides their huge technological importance, fluidized beds have attracted a large amount of research because they are perfect playgrounds to investigate highly dynamic particulate flows. Their over-all behavior is determined by short-lasting particle collisions and the interaction between solid and gas phase. Modern simulation techniques that combine computational fluid dynamics (CFD) and discrete element methods (DEM) are capable of describing their evolution and provide detailed information on what is happening on the particle scale. However, these approaches are limited by small time steps and large numerical costs, which inhibits the investigation of slower long-term processes like heat transfer or chemical conversion. In a recent study (Lichtenegger and Pirker, 2016), we have introduced recurrence CFD (rCFD) as a way to decouple fast from slow degrees of freedom in systems with recurring patterns: A conventional simulation is carried out to capture such coherent structures. Their re-appearance is characterized with recurrence plots that allow us to extrapolate their evolution far beyond the simulated time. On top of these predicted flow fields, any passive or weakly coupled process can then be investigated at fractions of the original computational costs. Here, we present the application of rCFD to heat transfer in a lab-scale fluidized bed. Initially hot particles are fluidized with cool air and their temperature evolution is recorded. In comparison to conventional CFD-DEM, we observe speed-up factors of about two orders of magnitude at very good accuracy with regard to recent measurements.

研究动机与目标

• 为克服传统CFD-DEM模拟在模拟流化床中传热等慢过程时的计算瓶颈。
• 实现长期传热模拟，而这些模拟在标准CFD-DEM中因时间步长过小和数值成本过高而不可行。
• 开发一种方法，通过递归分析将快速粒子动力学与慢热过程解耦。
• 将rCFD方法与近期实验室规模流化床中颗粒温度演变的实验测量结果进行验证。
• 证明一旦识别出相干流动结构，被动过程（如传热）即可在计算成本大幅降低的情况下进行模拟。

提出的方法

• 进行传统CFD-DEM模拟以捕捉瞬态动力学并识别流化床中的重复流动模式。
• 构建递归图以检测并表征随时间重复出现的相干粒子结构。
• 利用递归结构外推流场，使其远超原始模拟持续时间，从而有效预测长期流场行为。
• 在外部推的流场基础上，采用被动标量方法模拟传热，显著降低计算成本。
• 该方法依赖于主要流动结构周期性重复的假设，从而可在不重新模拟完整动力学的前提下实现预测性外推。
• 通过将预测的颗粒温度演变与实验室规模流化床的实验数据对比，对方法进行验证。

实验结果

研究问题

• RQ1 基于递归的流场外推是否能够实现流化床中长期传热过程的准确且高效模拟？
• RQ2 rCFD在多大程度上可降低计算成本，同时保持预测颗粒温度演变的精度？
• RQ3 rCFD方法在多大程度上能再现实验室规模流化床中传热的实验测量结果？
• RQ4 重复流动模式在实现快速动力学与慢热过程解耦中起到何种作用？
• RQ5 rCFD框架是否可应用于流化床中传热以外的其他弱耦合过程？

主要发现

• 与传统CFD-DEM模拟相比，rCFD方法在流化床传热模拟中实现了约两个数量级（即约100倍）的加速。
• 对初始高温颗粒在冷却空气中流化时的温度演变预测，与近期实验测量结果高度一致。
• 该方法成功捕捉了颗粒的长期热弛豫行为，而无需对整个过程进行全尺度重新模拟。
• 递归图的使用实现了对相干流动结构的准确识别与外推，为高效热过程模拟奠定了基础。
• 该方法在表征底层流动动力学后，对被动过程（如传热）的模拟表现出高保真度。
• 结果证实，rCFD是模拟复杂颗粒系统中慢速、多尺度过程的可行且高效替代方案。