Simulation of atomization : from DNS to industrial applications MUSAF III - 29/09/2016 D. Zuzio, J.-L. Estivalèzes, O. Rouzaud, P. Gajan, P. Villedieu PhD/postdoc : G. Blanchard, I. Marter, A. Orazzo, S. Thuillet CEDRE development team 1
Part 1 INTRODUCTION Part 2 DNS APPROACH Part 3 INDUSTRIAL APPROACH 2 MUSAF III ONERA Toulouse, 29/09/2016
Part 1 INTRODUCTION 3 MUSAF III ONERA Toulouse, 29/09/2016
Introduction Atomization in aeronautical injectors Different atomization processes Pressure injectors Atomizing films Sheared liquid sheets Cross-flow jets Fuel Air Fuel Air Fuel 4 MUSAF III ONERA Toulouse, 29/09/2016
Introduction Atomization : a multi-scale problem Atomization is a complex phenomenon spanning several space and time scales ~1 cm ~10 cm ~0.1 mm ~0.01 mm ~10 mm High fidelity DNS simulations : solve all the scales Multi-scale, multi-solver approach: solve the largest scales of primary atomization, model the sub-grid formation of droplets 5 MUSAF III ONERA Toulouse, 29/09/2016
Part 2 DNS APPROACH 6 MUSAF III ONERA Toulouse, 29/09/2016
DNS approach Building blocks for a DNS code for assisted atomization I. Incompressible flows and interface tracking II. Mass conservation III. Multi-scale IV. Momentum conservation/consistency V. Complex geometries 7 MUSAF III ONERA Toulouse, 29/09/2016
DNS approach Solutions I, II I. Incompressible flows and interface tracking Projection method on Cartesian MAC Interface tracking algorithm (LS, VOF ) Mass / momentum / pressure decomposition Fixed density NS equations II. Mass conservation CLSVOF* LS function (φ): precisely computes the interface characteristics VOF function (C): assures mass conservation PLIC : sub-grid 2 th order resolution of the front Geometrically accurate flux integration *Sussman JCP 2000, Menard IJMF 2006 8 MUSAF III ONERA Toulouse, 29/09/2016
DNS approach Solutions III III. Multi-scale Massively parallel, AMR Dispersed phase coupling + oct-tree AMR CLSVOF Lagrangian particle Resolved droplet formation CLSVOF Refined mesh Droplet individuation Tag propagating algorithm Size and shape criteria Lagrangian droplet tracking Advanced 2W coupling Advection to de-refined zones drag force model Zuzio, Estivalèzes, C&F 2016, in review 9 MUSAF III ONERA Toulouse, 29/09/2016
DNS approach Atomization results Air/water sheet, a = 300 µm density ratio 100 U g,eq = 40 m/s U g,eq = 80 m/s Zuzio, Estivalèzes, C&F 2016, under review 10 MUSAF III ONERA Toulouse, 29/09/2016
DNS approach Atomization results 11 MUSAF III ONERA Toulouse, 29/09/2016
DNS approach Atomization results Preliminary study of droplets distribution U g,eq = 60 m/s Mesh size Δx = 40 μm Δx = 40 μm Good reproduction of the distribution peak Resolved distribution 14 MUSAF III ONERA Toulouse, 29/09/2016
DNS approach Atomization results Breakup Length Oscillation Frequency Good qualitative results but: Higher frequency Lower breakup length Limitations of the simulation: Density ratio? Simple geometry? 13 MUSAF III ONERA Toulouse, 29/09/2016
DNS approach Solutions IV IV. Consistent/conservative momentum discretization Local co-localized mass/momentum solve Consistent and geometrical accurate flux computation (PLIC reconstruction) Space/time integral of mass/momentum fluxes 14 MUSAF III ONERA Toulouse, 29/09/2016
DNS approach Solutions III Pure high density (10 3 ) shearing flow High density droplet advection ρl/ρg = 10 6 ρl/ρg = 10 9 10 passes, ρl/ρg = 10 9 I. Lagrange, A. Orazzo, ONERA 2016 16 MUSAF III ONERA Toulouse, 29/09/2016
DNS approach Solutions V V. Immersed interface method Discrete forcing approach : sharp interface representation Example: Poisson equation 1. Definition of new Level Set function φsol to set the position of the solid boundary 2. Evaluation of the length fraction Li,j=f(φsol) of the face covered by the solid 16 MUSAF III ONERA Toulouse, 29/09/2016
DNS approach Work in progress D. Zuzio, ONERA 2016 I. Lagrange, A. Orazzo, ONERA 2016 18 MUSAF III ONERA Toulouse, 29/09/2016
Part 3 INDUSTRIAL APPROACH 19 MUSAF III ONERA Toulouse, 29/09/2016
Large scale approach Combustion chambers simulation and atomization Combustion chamber LES No atomization modelling Reactive two phase LES simulation of a TLC burner, Luc-Henry Dorey PhD Thesis, ONERA, 2011 19 MUSAF III ONERA Toulouse, 29/09/2016
Large scale approach Multi-solver strategy Resolved primary atomization : multi-fluid solver ( ) One equation for each phase/specie Implicit interface representation Solve large scales only Spray : dispersed phase solver ( ) Lagrangian approach Resolution of primary atomization largest scales Subgrid droplet generation Spray evolution Experimental data, DNS results (Liquid structure sizes, velocity/vorticity field ) 20 MUSAF III ONERA Toulouse, 29/09/2016
Large scale approach Atomization model Production of droplet in under-resolved (smeared) zones Keeps intact resolved structures 2-way coupling in mass, momentum and energy Re-impact model Dispersed phase solver Separate phases solver Under-resolved zones activation of model, injection of droplets PhD G. Blanchard, ONERA-DGA, 2014 21 MUSAF III ONERA Toulouse, 29/09/2016
Industrial applications Sheared liquid sheet/reimpact Planar liquid sheet ONERA experience (SHAPE) Thèse G. Blanchard, ONERA-DGA, 2014 23 MUSAF III ONERA Toulouse, 29/09/2016
Industrial applications Sheared liquid sheet/reimpact Breakup Length Oscillation Frequency Thèse G. Blanchard, ONERA, 2014 24 MUSAF III ONERA Toulouse, 29/09/2016
Industrial applications CEDRE - Cross flow jet Flow instabilities Cross flow jet behaviour in presence of flow perturbations CEDRE large scale simulation Inlet Numerical domain Liquid jet Outlet A. Desclaux, V. Bodoc, P. Gajan, ONERA, 2015 Réf : Anderson et al. / ASME / 2001 ; Song et al. / ASME / 2013 27 MUSAF III ONERA Toulouse, 29/09/2016
Industrial applications CEDRE - Cross flow jet D. Zuzio, S. Thuillet, O. Rouzaud, P. Gajan, ONERA, 2015 28 MUSAF III ONERA Toulouse, 29/09/2016
Industrial applications CEDRE - Cross flow jet Jet trajectory analysis S. Thuillet PhD, ONERA, 2015 29 MUSAF III ONERA Toulouse, 29/09/2016
Conclusions et perspectives Some conclusions DNS has great potential for understanding and quantifying atomization process LSS computation of atomization is a convenient way to introduce unsteady injection in two-phase LES simulation of combustors (multi-fluid / dispersed phase) Some perspectives The DNS simulations could allow precise measurements and local correlations for large scale models The atomization models have to be filled with relevant physical parameters, coming hopefully from the unsteady local flow The spray distribution has still to be validated in steady and unsteady regimes 30 MUSAF III ONERA Toulouse, 29/09/2016
Thank you for your attention 31 MUSAF III ONERA Toulouse, 29/09/2016