Development and Use of LES for Diesel Engine CFD

Abstract

Large eddy simulation is a promising advancement for IC engine modeling. This requires appropriate models for a many physical processes. This paper classifies and briefly describes the major types of models for turbulence, combustion, and sprays. Representative references are listed with an emphasis on engine applications. Recommendations for modeling approaches to use in engine LES are also provided. Introduction LES Turbulence Models It is generally agreed that the next generation of The turbulence model in LES simulations is for turbulence modeling in computational fluid dynam- sub-grid stresses in the momentum equation. It is ics (CFD) for many applications will be some form the primary model in LES since it should allow for of large eddy simulation (LES) [1,2]. For the ap- flow structures in the solution. Table 1 classifies propriate applications, LES can offer significant the major approaches to LES turbulence models advantages over traditional Reynolds Averaged and briefly states advantages and disadvantages. Navier Stokes (RANS) modeling approaches. In This table does not list more esoteric but includes internal combustion (IC) reciprocating engines LES only modeling approaches that are likely to find can be used to study cycle-to-cycle variability, use in practical applications. provide more design sensitivity for investigating both geometrical and operational changes, and LES Combustion Models produce more detailed and accurate results. There There is a rich tradition of combustion modeling are also characteristics of IC engines such as inhe- [10] much of which has been adapted for engine rent unsteadiness and a moderately sized domain simulations. In many cases, users have simply that are well suited to LES. However, there can be taken RANS combustion models and used them many difficulties in using LES for the multi-phase, with LES turbulence models. This is a type of hybrreacting, turbulent flows in complex internal com- id approach and has proven useful. However, as bustion engine geometries. In addition, LES in IC LES use in engines matures, more attention is engines is new and there are potential uncertain- being given to either adjusting or even reformulatties and ambiguities since a generally accepted ing basic combustion models for use in LES. Table „best practice‟ is still developing. This report lists 2 classifies and briefly describes the major apcurrent LES models that could have application to proaches to combustion modeling that have either IC engines and briefly evaluates their suitability been used or could be used for engine CFD. and potential predictive capability for engine CFD. T4 Detached Eddy T3 Dynamic Smagorinsky

Keywords

• Disadvantages

Transport equations 0

Model type TVuisrcbousleitnyt T1 None Numerical viscosity only T2 Smagorinsky Yes

T5 k-equation LES T3 Dynamic Smagorinsky

T6 Dynamic kequation LES T7 Dynamic Structure

Refs.

General Approaches and

Specific Models C1: Direct Integration “CHEMKIN” or other

stiff ODE integrator P, D P, D

P P, D P, D

Uses detailed kinetic mechanisms; no special modeling required.

Better computational efficiency for detailed chemistry. Uses flamelet concepts to model subgrid mixing (C4d).

Simple; uses both kinetic and turbulent time scales.

Improves on Magnusson by integrating towards current equilibrium state.

Flamelet approaches. Sound mathematical descriptions. Sound modeling of turbulence effects on flame front.

Similar to C4a for premixed flames but reduces grid resolution requirements.

Similar to G-equation approach (C4b) but uses the flame area for a more physical description. Can incorporate detailed chemistry through flamelet library. Uses prescribed PDF to model subgrid mixing effects.

Tries to improve on mixture fraction models (C5d) by using values from the reaction zone. Provides direct closure without models for reaction terms. Ignores subgrid turbulence effects. Better suited for homogeneous combustion.

Computationally expensive.

Not really a CFD method

since model is not applied to each grid cell.

Requires using same time scales for all reactions within individual grid cells.

Still requires same time scales.

Transport equations require modeling of scalar flux, source terms, and sink terms.

No detailed chemistry. Better for high Reynolds number flows. Requires high grid resolution to resolve flame front Not suited for detailed chemistry. Requires model for turbulent flame speed. (similar to C4b)

C5: PDF Transport

LES Fuel Spray Models

Until recently there has been little published work on spray models for LES simulations. Most simulations of applications have used existing RANS type spray models with simple modifications for use with LES turbulence and combustion models. Most practical applications for IC engines use the Lagrangian spray parcel methodology originally developed for RANS approaches (see Reitz [ 21 ]) in which the CFD grid is not resolved around individual spray particles. In this context the spray modeling issue is how to represent the sub-grid interaction of the Lagrangian spray particles with the continuous gas phase. This interaction includes momentum transfer (e.g. drag), kinetic energy transfer, heat and mass transfer during evaporation, and models for atomization, breakup, and collisions. This is a very extensive list of complex physical processes that require modeling. This is probably why most spray models are extensions of RANS approaches and, only recently, has work has been done on developing new spray model formulations specifically for LES applications.

Model type RANS correlations

S2 LES modifications

S3 Volume of Fluid

(VOF) S4 Continuous phase, non-particle models (Full Eulerian)

References

Recommendations [8] K. Sone , S. Menon, “ Effect of Subgrid Modeling on The more advanced differential LES turbulence the In-Cylinder Unsteady Mixing Process in a Direct models (T5-T7) should be used . These do not Injection Engine,” J. Eng. Gas Turb. Power , Vol.

require extremely fine grids and work well on the 125 , pp. 435 - 443 ( 2003 ).

courser grids commonly found in engine applica- [9] EE .quPaotmiornanNionng-aVnisdcoCs. Jit .yRLuEtSlanMdo,d“Ael, D ” yAnIAamAiJcoOunrnea-l, tions. The most common LES models use simple Vol. 40 , No. 4, pg. 689 - 701 ( 2002 ).

viscosity formulations (T2, T3), do not take advan- [10] Heinz Pitsch , “ Large-Eddy Simulation of Turbulent tage of LES concepts, and perform poorly in en- Combustion,” Annu. Rev. Fluid Mech . 38 : 453 - 82 gine applications unless unrealistic grid resolutions ( 2006 ).

are used . [11] R. Jhavar and C.J. Rutland , “ Using Large Eddy Use transport based combustion models (C4) Simulations to Study Mixing Effects in Early Injecsince these models benefit from the large scale tion Diesel Engine Combustion,” SAE 2006-01- flow structures that occur in LES simulations . Use 0871 , ( 2006 ).

LES specific modification for major terms within the [12] tCio .nHienrgaaDrtIaDniedsNe.l EPnegteinrse, A“SpipmlyuilnagtinagNtheewCMoomdbeulfso-r models such as mixing time scales, scalar dissipa- the Conditional Scalar Dissipation Rate,” SAE Paper tion rate, turbulent flame speeds, scalar flux , etc. 2001-01-1001 ( 2001 ). The development of LES spray models is signif- [13] Takeshi Miyamoto , Tadashi Tsurushima, Naoki icantly behind work on other LES models . Thus, it Shimazaki, Akira Harada, Satoru Sasaki, Koichi is hard to make a strong recommendation for use Hayashi, Yasuo Asaumi, and Yuzo Aoyagi, “Modelin IC engine applications . In practice, it appears ing Ignition and Combustion in Direct Injection that using RANS based spray models with some Compression Ignition Engines Employing Very Early minor modifications to work with LES turbulence Injection Timing”, JSME International Journal Series models may be adequate for the time being . [14] BS,.CV.oKl.o4n5g ,,NZo.. H4a ,nppa. n8d72R -. 8D8 . 0R ( e2i0tz0 ,2“)T. he development and application of a diesel engine and com-

[1] R. S , .Rogallo, P. Moin, “ Numerical simulation of tion” , SAE Paper 950278 ( 1995 ). turbulent flows . Ann. Rev. Fluid Mech . 16 , 99 - 137 [15] R. Steiner , C. Bauer , C. Krüger , F. Otto , U. Maas, ( 1984 ). “3D-Simulation of DI-Diesel Combustion Applying a

[2] Meneveau , C. & Katz , J. “ Scale-invariance and Progress Variable Approach Accounting for Comturbulence models for large-eddy simulation , Ann. plex Chemistry, ” SAE Paper 2004 - 01 - 0106 ( 2004 ) Rev . Fluid Mech. 32 , 1 ( 2000 ) [16] L. Liang , R. D. Reitz , C. O. Iyer and J. Yi , “Modeling

[3] F. F. Grinstein , C. Fureby, “ Recent Progress on Knock in Spark-Ignition Engines Using a G-equation MILES for High Reynolds Number Flows,” Transac- Combustion Model Incorporating Detailed Chemical tions of the ASME , Vol. 124 ,pp 849 - 861 ( 2002 ). Kinetics,” SAE Paper 2007 - 01 - 0165 ( 2007 ).

[4] Philipp Adomeit , Oliver Lang, Stefan Pischinger, [17] O. Vermorel , S. Richard , O. Colin , C. Angelberger Richard Aymanns, Markus Graf and Georg Stapf, and A. Benkenida , D. Veynante , “ Multi-Cycle LES “ Analysis of Cyclic Fluctuations of Charge Motion Simulations of Flow and Combustion in a PFI SI 4- and Mixture Formation in a DISI Engine in Stratified Valve Production Engine , ” SAE 2007 - 01 -0151 Operation,” SAE 2007- 01 - 1412 ( 2007 ). ( 2007 )

[5] Haworth D , Jansen K. , “Large-eddy simulation on [18] B. Hu , C.J. Rutland , T. Shethaji, “ Combustion Modunstructured deforming meshes: towards reciprocat- eling of Conventional Diesel-type and HCCI-type ing IC engines,” Comput. Fluids 29 ( 5 ): 493 - 524 Diesel Combustion with Large Eddy Simulation,” ( 2000 ). SAE Paper 2008 - 01 - 0958 ( 2008 ).

[6] Christian Hasse , Volker Sohm, Bodo Dur, “De- [19] S.H. Kim , K.Y. Huh , R. A. Fraser , “ Numerical Pretached eddy simulation of cyclic large scale fluctua- diction of the Autoignition Delay in a Diesel-Like Entions in a simplified engine setup , ” Int. J. of Heat vironment by the Conditional Moment Closure Modand Fluid Flow 30 , pp. 32 - 43 ( 2009 ). el, ” SAE 2000 - 01 - 0200 ( 2000 )

[7] F. Brusiani and G. M. Bianchi , “LES Simulation of [20] E. H. Kung and D. C. Haworth , „ Transported ProbaICE Non-Reactive Flows in Fixed Grids,” SAE 2008- bility Density Function (tPDF) Modeling for Direct01-0959 ( 2008 ). Injection Internal Combustion Engines,” SAE 2008 - 01 - 0969 ( 2008 )

[21] R.D. Reitz , “ Controlling D.I. Diesel Engine Emissions Using Multiple Injections and EGR,” Comb. Sci. & Tech ., Vol 138 , pp. 257 - 278 ( 1998 ).

[22] T. Hori , J. Senda , T. Kuge and H. Gen Fujimoto “Large Eddy Simulation of Non-Evaporative and Evaporative Diesel Spray in Constant Volume Vessel by Use of KIVALES,” SAE 2006- 01 - 3334 ( 2006 ).

[23] N. Bharadwa , C. J. Rutland , S. Chang , “ LES Modeling of Spray Induced Turbulence Effects,” SAE 2009 - 01 -0847, to appear Int. J. of Engine Research ( 2009 ).

[24] D. Gosmen , H. Weller, “ Large eddy simulation of primary diesel spray atomization , ” SAE 2004 - 01 - 0100 ( 2004 ).

[25] F. Laurent , M. Massot , P. Villedieu, “ Eulerian multifluid modeling for the numerical simulation of coalescence in polydisperse dense liquid sprays , ” J. Comp. Physics , Vol. 194 , 2 , pp 505 - 543 ( 2004 )