Time-accurate Modeling of Multicomponent Fluid Mechanical Systems using GPUs

The behavior of multicomponent fluid mechanical systems, including liquid/solid, liquid/liquid, and liquid/gas mixtures, is governed by flow physics that operate over a range of length and time scales. At the smallest scales, process outcomes are governed by Kolmogorov-sized eddies and particle relaxation times. At the largest scales, system performance is governed by tank residence times, blend times, and the dimensions of the system. These smallest and largest scales are coupled by a continuous cascade of fluid motion and eddy break-up.

The fidelity of a computational model is correlated to the range of length/time scales it can capture simultaneously. Although models for predicting the mean flow field have existed for decades, their inability to capture transient eddy structures limits their applicability to multicomponent process modeling. In this presentation, we show how modern CFD algorithms operating on graphics processing units (GPUs) can enable real-time multicomponent flow modeling in industrial systems. Following an overview of the underlying theory and implementation, we validate this approach by comparing real-time model predictions to measured data obtained from particle suspension systems and gasified bioreactors.