Cerfacs Enter the world of high performance ...

Web link: https://cerfacs.webex.com/cerfacs/j.php?MTID=m80cd1664853c4618fe56dda8ca79fffd

Abstract: 

The aeronautical sector must reduce its pollutant emissions and its fuel consumption to face the ever more stringent environmental standards. To achieve this goal, lean combustion technologies, allowing a good combustion efficiency while limiting nitrogen oxides (NOx) and carbon monoxide (CO) emissions, are being developed. Those new generation combustors are nevertheless more likely to destabilize and the fuel injection system plays a significant part in stabilizing the flame, and thus maintaining an efficient combustion process. Because of their ability to offer stable performances for a large range of operating conditions, Airblast injectors have attracted much attention from the engine manufacturers. In such injectors, the fuel is introduced in the combustor as a thin liquid film flowing along the injector wall towards the diffuser lip, where it is finally atomized into small droplets. This PhD aimed at numerically studying the physical phenomena involved in Airblast injectors, and at developing the numerical tools allowing their characterization. With that objective, a methodology allowing to perform some direct numerical simulations of academic configurations has first been proposed. It was then applied in two distinct configurations to the study the liquid film dynamics and its atomization process using a Volume Of Fluid method and the NGA incompressible solver. In parallel, a simpler modelling method relying on phenomenological models has been developed in the Lagrangian solver of the compressible code AVBP. This approach allows to predict the atomization dynamics given the flow topology. Our main work consisted in developing and integrating the AutomaticPAMELA (Primary Atomization Model for prEfilming airbLAst injectors) model which proposes a local formulation and an automated determination of the PAMELA primary atomization model inputs (Chaussonet, 2016). Once validated on a well characterized experimental configuration, the model was applied to an industrial Airblast injector configuration and allowed us to predict spray characteristics in line with experimentally measured. While very efficient, this approach is valid for a restricted operating range, thus limiting its application to flows where the physical phenomena involved are well known. To address this issue, a twophase flow numerical method based on a multifluid diffuse interface approach has also been developed in AVBP during this PhD. It relies on the combination of a 4-equations multifluid model to describe the flow behavior, and a Noble-Able-Stiffened-Gas (NASG) thermodynamic closure able to represent the thermodynamic evolution in both the gaseous and the liquid phase. To avoid numerical diffusion of strong gradients at liquid/gas interface while ensuring the numerical stability, a Godunov numerical scheme with a HLLC Riemann solver designed for “node-centered” formulations has been integrated in the AVBP code. It required an important algorithmic work specially to implement a MUSCL reconstruction methodology in a high-performance computing framework, and thus ensuring a global second order accuracy. After some 1D and 2D validations, the method has finally been used to perform 2D simulations of a gas sheared liquid film and of a prefilming Airblast atomization process, highlighting its ability to address physical phenomena representative of real Airblast injector configurations.

keywords: CFD, two-phases flow, Airblast injector, atomisation

Jury: