A NEW ADAPTIVE MULTISCALE METHOD FOR DIRECT NUMERICAL SIMULATION OF SHEARED LIQUID SHEET
DOI: 10.1615/ICHMT.2015.IntSympAdvComputHeatTransf.1860 pages 17951796
JeanLuc Estivalezes ONERA, the French Aerospace Lab, Toulouse, France
Davide Zuzio ONERA, the French Aerospace Lab, Toulouse, France
RésuméThe detailed numerical simulation of aeronautical combustion chambers has become an important topic in the last years, answering to the need to enhance the engine efficiency and to reduce the polluting agents. The airblast injectors are among the most common devices found in aeronautical engines, exploiting the assisted atomization to pulverize the fuel. A way to perform multiscale simulations of such systems is to use a computational mesh with a variable space step where the finest mesh is used for a detailed description of the liquid/gas interface. However, the cloud of droplets resulting from the atomization can spread across a large area of the computational domain. As the small size of the droplets would require to refine dramatically the large domain occupied by the cloud, this strategy is not well adapted to the simulation of the cloud. An alternative numerical approach is to use a dispersed phase model for the smallest inclusions. First, sufficiently small droplets keep a spherical shape due to surface tension. Second, the derefinement of the mesh in the cloud produces large mesh cells for which the dispersed phase model hypothesis are satisfied.
The present work uses an adaptive mesh refinement (AMR) method to obtain a global mesh composed of block of Cartesian structured meshes of various sizes. A coupled Levelset/VolumeofFluid (CLSVOF) method is used to track the multiphase information. In the present case, the finest blocks are dynamically generated at the liquid/gas interface where the discretization of the conservation equations have to take into account strong gradients of various physical quantities (velocity, viscosity, density, interface curvature...). This approach is well adapted to simulate all the scales of the atomization process which generally occurs in a reduced zone. The smallest liquid inclusions are treated as Lagrangian particles and therefore do not trigger the mesh refinement anymore. Such an Eulerian/Lagrangian coupled approach has been previously studied by some authors among others Hermann[2010], Eckett[2013]. These Eulerian/Lagrangien couplings allow large and small inclusions (with respect to the mesh size) to be tracked. However, for what we call medium inclusion (with a characteritics length between 1 and 3 grid step), it seems that both models are not appropriate. The discrete point Lagrangian dispersed phase hypothesis is not verified while the Eulerian modeling is not accurate at this scale. The present work proposes to use a new modeling with an Eulerian/Lagrangian method (ELM) to treat the medium inclusions inaccurately treated with Eulerian (EM) and Lagrangian (LM) methods. It is based on a projection of liquid density and momentum of the Lagrangian droplet on the underlying Eulerian grid allowing a conservative treatment of density and momentum of the twophase flow. The whole methodology is applied to the simulation of an airblast planar injectors. As an example, we show in figure 1 late stage of head collision of two droplets at We=100, the green particles are the Lagrangian one, in figure 2 is shown the atomization of a plane liquid sheet sheared by two coflowing air flow where, here again the green particles correspond to the Lagrangian droplets coming from primary atomization.
