(470c) Direct Numerical Simulation of Bioreactors Using GPUs

Turbulent fluid flow in bioreactors occurs over a range of length and time scales. In the upper limit, these systems contain macro eddies with length scales comparable to the vessel diameter and time scales comparable to the tank blend time. In the lower limit, these systems contain micro eddies with length and time scales characterized by the Taylor and Kolmogorov scales. In chemical reactors and combustion systems, process outcomes are governed by fluid dynamics occurring across the entire spectrum of eddy scales: the macro scales describe how systems blend, while the microscales describe how species will react. From a modeling perspective, direct numerical simulations are required to obtain the complete insights required to properly predict reactor performance.

In this work, we show how GPUs (as opposed to CPUs) can make DNS of industrial mixing systems practical and timely. We begin by introducing the concepts governing GPU-performance, as applied lattice-Boltzmann simulations. We then present various criteria for monitoring the convergence from large-eddy simulation (LES) to direct numerical simulation, within the context of the Taylor and Kolmogorov scales. We then apply these criteria to study blending, energy dissipation, and reaction rates in a benchtop bioreactor in the LES and DNS regimes. The effects of simulation resolution on runtime and results are discussed.