Time-space adaptive numerical methods for the simulation of combustion fronts

Description

This paper presents a new computational strategy for the simulation of combustion fronts based on adaptive time operator splitting and spatial multiresolution. High-order and dedicated one-step solvers compose the splitting scheme for the reaction, diffusion, and convection subproblems, to independently cope with their inherent numerical difficulties and to properly solve the corresponding temporal scales. Adaptive and thus highly compressed spatial representations for localized fronts originating from multiresolution analysis result in important reductions of memory usage, and hence numerical simulations with sufficiently fine spatial resolution can be performed with standard computational resources. The computational efficiency is further enhanced by splitting time steps established beyond standard stability constraints associated to mesh size or stiff source time scales. The splitting time steps are chosen according to a dynamic splitting technique relying on solid mathematical foundations, which ensures error control of the time integration and successfully discriminates time-varying multi-scale physics. For a given semi-discretized problem, the solution scheme provides dynamic accuracy estimates that reflect the quality of numerical results in terms of numerical errors of integration and compressed spatial representations, for general multi-dimensional problems modeled by stiff PDEs. The strategy is efficiently applied to simulate the propagation of laminar premixed flames interacting with vortex structures, as well as various configurations of self-ignition processes of diffusion flames in similar vortical hydrodynamics fields. A detailed study of the error control is provided and show the potential of the approach. It yields large gains in CPU time, while consistently describing a broad spectrum of space and time scales as well as different physical scenarios.

Abstract

Accepted for Publication. This research was supported by a fundamental project grant from ANR (French National Research Agency - ANR Blancs): Séchelles (PI S. Descombes - 2009-2013) and by a DIGITEO RTRA project: MUSE (PI M. Massot - 2010-2014). The support of the France-Stanford Center for Interdisciplinary Studies through a collaborative project grant entitled "Multi-scale mathematical modeling and numerical methods for multiphase and reactive flows" (PIs: P. Moin and M. Massot) has been very helpful. We also acknowledge the computational resources of the Mesocentre of Ecole Centrale Paris where some of the simulations were performed.

Abstract

International audience

Additional details