Sammendrag
In thermal systems for transportation and power generation turbulence and combustion are ubiquitous and very often inextricably coupled processes, down to the smallest scales of continuum mechanics in non-linear fashion. Conventional approaches to Computational Fluid Dynamics (CFD), traditionally based on coarse modelling assumptions and low-order numerical methods, are not able to provide accurate predictive capabilities outside model-specific narrow bands of applicability nor deliver any physical insight about the behavior of the rate-controlling processes that govern turbulent combustion at small time and length scales. Today, highly-resolved, large-scale Direct Numerical Simulation (DNS) can be paired with increasingly affordable (and accessible!) high-performance computational resources and this combination has the potential to fill knowledge and methodology gaps, still present in mature combustion technologies, to achieve needed improvements and breakthroughs in power generation and propulsion. It will be shown that DNS is able to provide an unprecedented level of insight and detail to isolate and understand the mechanistic causality between turbulence, mixing and chemical reactions. Furthermore, these first principles DNS are becoming increasingly capable to extend beyond basic setups of purely academic interest and investigate configurations of industrial relevance: examples are provided of three-dimensional DNS with detailed chemical kinetics that are used to investigate 1) the stabilization of turbulent non-premixed flames in the wake of a transverse fuel jet in oxidant cross-flow (JICF), 2) the upstream propagation (flashback) of turbulent premixed flames in fully developed channel flow and 3) the stabilization mechanism of auto-igniting flames in a staged (reheat) combustion system. Ultimately, exa-scale computing, to be deployed throughout Europe, the USA and Asia in the early 2020s, will theoretically enable DNS of turbulent combustion in parameter regimes that are still inaccessible today. However, making the transition to exa-scale poses a number of algorithmic, software and technological challenges. As Moore’s Law and Dennard scaling come to an end exa-scale computing will be achieved only through massive concurrency. Addressing issues of data movement, power consumption, memory capacity, interconnection bandwidth, programmability, node failure and scaling is critical to ensure that future combustion simulations can take advantage of the computer architectures of the future. © 2019, American Institute of Aeronautics and Astronautics Inc, AIAA. All rights reserved.