We have performed high fidelity direct numerical simulations (DNS) of published experimental (Gepperth, ICLASS, 2012) planar pre-filming airblast atomization configuration using our in-house Navier-Stokes solver ARCHER. The liquid/gas interface has been tracked using two different methods: coupled level set moment of fluid (CLSMOF) and coupled level set volume of fluid method (CLSVOF) methods. The CLSMOF method is a hybrid method between using conventional moment of fluid (MOF) (Dyadechko & Shashkov, JCP, 2008) and level set method. This method combines the advantage of sharp interface representation from level set as well as accurate capturing of the under-resolved thin liquid ligaments and filaments in the domain.

It has been shown that MOF method (Asuri Mukundan et al, ICCFD10, 2018; Asuri Mukundan et al, ICLASS, 2018) performs better than CLSVOF method (Ménard et al, IJMF, 2007) in terms of accuracy. But due to high computational cost of MOF method, it is imperative to develop a method that has same numerical accuracy and less computational cost. To this end, we have implemented a new CLSMOF method that uses MOF interface reconstruction method only for those under-resolved thin liquid structures while using a conventional level set method otherwise.

MOF method uses liquid volume fraction and phase centroids to reconstruct interface in each computational cell. A directionally split advection algorithm is used for transporting the liquid volume fraction (Weymouth & Yue, JCP, 2010) and the phase centroids are also advected in a directionally split manner using Eulerian Implicit-Lagrangian Explicit (EI-LE) method in our solver ARCHER. A consistent mass and momentum flux computation (Vaudor et al, C&F, 2017) is implemented in our ARCHER. In this work, the CLSVOF method of Ménard et al (IJMF, 2007) is used.

The DNS were performed for moderate Reynolds and Weber number using kerosene as fuel under aircraft altitude relight conditions (Warncke et al, IJMF, 2017). The ability of our solver to predict the droplet diameter distribution and droplet velocity distribution is demonstrated for both interface reconstruction methods by matching the results with the experimental data (Gepperth et al, ICLASS, 2012). The predominant sheet breakup mode is relatively more accurately captured with CLSMOF than with CLSVOF method. Additionally, the ligament lengths and velocity of the accumulated liquid at the trailing edge of the pre-filmer plate have been analyzed statistically by applying post-processing techniques on our DNS data that is consistent with that applied to experimental data.