Model Reduction Opportunities in Detailed Simulations of Combustion Dynamics

Abstract

Rocket and gas turbine combustion dynamics involves a confluence of diverse physics and interaction across a number of system components. Any comprehensive, self-consistent numerical model is burdened by a very large computational mesh, stiff unsteady processes which limit the permissible time step, and the need to perform tedious, repeated calculations for a broad parametric range. Predictive CFD models rely on very large scale simulations and advanced hardware. Reduced Basis Methods (RBM) have grown in usage during the past decade, as promising new techniques in making large simulations more accessible. These methods create models with far fewer unknown quantities than the original system, by generating "proper" fundamental solutions and their Galerkin projections, while guaranteeing accuracy and computational efficiency. RBMs seek to reproduce full CFD solutions, rather than solutions to a simplified or linearized set of equations. We present here some recent work in this area, focusing on approaches to model large scale combustor systems. The maturation of methods leading to LES-based turbulent combustion modeling is discussed, and model reduction goals and strategies are explored from the perspective of applicability in real life problems in both gas turbine, as well as rocket engines.

Investigations and Improvement of Robustness of Reduced-Order Models of Reacting Flow

“…Reduced-order models have proven to be efficient in reducing complex partial differential equations (PDE) to low dimensional ordinary differential equations (ODE) in non-reacting flow problems such as flow control [9][10][11] and aeroelasticity [12,13]. Recent studies have extended ROMs to reacting flow problems [14,15] and preliminary explorations of the POD/Galerkin technique have been carried out [16,17] using a simple 1D solver to establish a basic approach for ROM construction and its characteristics in reacting flows. More recently, the current authors initiated the first attempt to apply ROM techniques on relevant combustor flow simulations [18,19], which identified steep temperature gradients as one of the underlying challenges of reacting flow ROM development.…”

Exploration of POD-Galerkin Techniques for Developing Reduced Order Models of Nonlinear Euler Equations

“…ROM techniques have wide application in non-reacting flow problems including flow control [4][5][6] and unsteady aeroelasticity [7,8]. Recently application has been extended to combustion problems [9,10]. A preliminary exploration of the POD/Galerkin technique using a model reaction-advection scalar equation in developing valid and capable reduced-order model was performed by Huang et al [11], which mainly assesses the capability of the ROM for predicting responses at target frequencies.…”