Numerical part
Although gas-phase turbulent combustion modeling have been widely studied with a wide range of suggested closures, relatively few studies have been dedicated to turbulent spray combustion and acoustic because of modeling difficulties. When undergoing turbulent fluctuations or acoustic waves, droplets tend to form clusters leading to local high level of droplet density. The present work points out evaporating droplet influence on mixture fraction and proposes how to introduce these effects in turbulent combustion models.
Month 12 (D 4.2): Development of a DNS Database
This work considers a geometry involving one non-homogeneous direction: the spatially decaying turbulence (SDT). It simulates a grid-turbulence with a high kinetic energy at the inlet that decays along the streamwise direction. Thus the droplets undergo a natural polydispersion process. Even if the injected spray is monodisperse, polydispersion occurs because of joint effect of droplet evaporation and their mixing by turbulence structures.
To evaluate the spray impact on the mixing, the injected turbulence properties remain the same in all the configurations. Test cases are thus performed with the same injected turbulence and the observed differences occur because of two joint effect: the turbulence modification induced by droplet momentum and the local evaporation rate resulting from the droplet dispersion.
Considering data storage, memory and time consumption, we assure a compromise with a configuration that computes a million of droplets tracked individually in a Lagrangian frame that evolves onto a 129x65x65 Eulerian mesh. Then, to respect DNS restriction, the size of the domain is (respectively) 16 l_t x 8 l_t x 8 l_t for an injected spectral turbulence characterized by an acoustic Reynolds number Re = 5000 and a 50% turbulent rate. COntact for the database: Julien.Reveillon@coria.fr
Month 24 (D 4.6 & 4.7): Analysis of subgrid models for spray evaporation
In the context of either Reynolds averaged Navier-Stokes calculations (RANS) or Large Eddy Simulation, non-premixed (or partially premixed) turbulent combustion usually adopts the mixture fraction concept.This conserved scalar indicates the mixing between reactives: $Z = 0$ in the oxidizer stream and $Z=1$ in the gaseous fuel stream. To provide information on reactant mixing through the computational domain, one strategy is to presume the form of the probability density function (PDF). The beta-PDF is the most commonly used shape. It requires two input parameters: the large scale level of the mixture fraction, and its subgrid fluctuations Z_v. All the models that have been found in the litterature to close the mixture fraction evolution and specific models developped in the framework of the European project have been tested thanks to the database developped during year 1. Details may be found in our 24th month or final report.
Month 36 - extended Month 42 (D4.15):
Combustion instabilities are observed in numerous industrial systems and more particularly in aeronautical engines. They create many undesirable effects as, for example, an increase of wall heat fluxes, flames extinction and flashback, or strong vibrations of the mechanical structure, which can lead to its destruction. In spite of many research tasks based on this topic, these instabilities are difficult, if not impossible, to predict.
To begin with, it is necessary to identify the phenomena responsible for the presence of combustion instabilities and their consequences on various processes such as injection, atomization, spray evaporation, reactants mixing, chemical reactions, interactions between flames and walls, etc. For that reason, multiple theoretical, experimental and numerical works are dedicated to the understanding and the analysis of the fundamental physical mechanisms of the couplings between these various processes and acoustic phenomena in the chamber.
To our knowledge, there is no numerical simulation dedicated to the analysis of the interactions between spray combustion and acoustic instabilities. The objective of this task is two-fold: first to demonstrate the capability of a DNS solver to capture the complex spray/flame/acoustic interactions and then, to focus on the impacts of velocity modulations on reaction rate through an analysis of the transfer function. A classical Bunsen configuration has been selected and experimental comparisons were made possible thanks to the data of the EM2C laboratory, Ecole Centrale Paris obtained in the framework of this project. The final report shows a complete validation of the simulations with the experimental data of EM2C laboratory. The validation has been made for both gaseous and two-phase flows and all the reduced frequencies up to a value of 30 through the transfer function amplitude and phase.