Large Eddy Simulation of Hydrogen-EnrichedMethane-Air Premixed Flames in a

Confined Swirl BurnerDavid Cicoria, and C.K. Chan

Abstract—Large eddy simulation (LES) is employed to in-vestigate the effects of hydrogen addition on lean CH4-airturbulent confined swirling premixed flames for mixtures upto 50% of hydrogen in volume. The subfilter combustion termrepresenting the interaction between turbulence and chemistryis modelled using the PaSR model, along with complex chem-istry using a skeletal mechanism based on GRI-MECH3.0. Atconstant swirl number and equivalence ratio, results show asimilar trend as in experiments concerning the flame shapeand stabilization mechanism. Increasing hydrogen in the fuelmixture leads to an increase of turbulent flame speed, reactivity,flame temperature and NO emissions. Streamline profiles ofmean velocity highlight the change in flame shape at higherhydrogen content due to a big jump in velocity inside the flame,and the structural modification of the outer recirculation zone.Independently of the equivalence ratio, the mean flame shapeis shorter and turbulent flame thickness decreases as hydrogenis increased. Also, heat release becomes less sensitive to thechange of equivalence ratio when hydrogen is more present inthe fuel mixture.

Index Terms—Turbulent premixed combustion, large eddysimulation, hydrogen-enrichment of methane flames, swirl,finite rate chemistry, flame stability

I. INTRODUCTION

HYDROGEN-enrichment consists of a solution to en-hance stability and reduce pollutant emissions of lean

premixed flames. Hydrogen enrichment has been found toextend the lean flammability limit and to reduce CO, NOx

and UHC emissions in spark ignition engines [1], [2] and ingas turbines [3], [4]. Several studies have focused on funda-mental aspects of swirling H2-CH flames. It has been pointedout that hydrogen-enriched flames exhibit greater laminarflame speeds, increased resistance to strain and extendedlean flammability limits [5], [6], [7]. With the increase ofhydrogen in the fuel mixture, more NOx emissions have beenobserved in the upstream region of the flame, whereas furtherdownstream, similar levels have been reported for differentmixtures at the same adiabatic flame temperature [6]. Tofurther characterize the effects of swirl on flow dynamicswith hydrogen-enrichment, Kim et al. [8], [9] carried outtwo studies of confined and non-confined premixed swirling

Manuscript received February 26, 2017; revised March 24, 2017. Thiswork was partially supported by a studentship of The Hong Kong Polytech-nic University and grants from the Research Committee of The Hong KongPolytechnic University (Grant No G-UC26 and G-YBCL).

D. Cicoria is with the Department of Applied Mathematics, TheHong Kong Polytechnic University, Kowloon, Hong Kong, e-mails:david.cicoria@connect.polyu.hk, david.cicoria@gmail.com

C.K. Chan is with the Department of Applied Mathematics, TheHong Kong Polytechnic University, Kowloon, Hong Kong, e-mail:ck.chan@polyu.edu.hk

flames. They found out that higher combustibility of hydro-gen enhances flame stability with hydrogen addition, bothat low and high swirl strengths to the mixture. Hydrogenaddition led to a shift of the reaction zone upstream of theflame, causing an increase in NO formation in this area dueto higher peak temperatures of flame. However, their studiesonly considered hydrogen in the fuel mixture for up to 9% involume. In this paper, different characteristics of the flamesuch as flame stability and pollutant emissions are studiedby means of LES. Fuel mixtures of methane-hydrogen upto 50% of hydrogen in volume are considered to furtherexamine the effects of hydrogen in a confined premixed swirlburner.

II. LES COMBUSTION MODELLING AND NUMERICALPROCEDURE

In this paper, compressible Navier-Stokes equations withenergy and species transport conservation equations are usedwith a low Mach number approach. The open source codeOpenFOAM (Open Field Operation and Manipulation) [10],based on finite volume is used in this study. For the con-vective and diffusive terms, 2nd order accuracy is achievedin space using central difference scheme with linear inter-polation, along with a normalized variable diagram (NVD)scheme. PISO algorithm is chosen to solve the unsteadyNavier-Stokes equations using 3 pressure and momentumcorrectors (1 outer and 3 inner iterations) with residualscriterion of 10−9. Time integration is carried out using animplicit 2nd order quadratic backward approximation. Thepressure equation is solved using a generalized geometric-algebraic multi-grid (GAMG) solver along with a Gauss-Seidel smoother. Concerning the velocity field and otherscalar fields, the code uses a preconditioned bi-conjugategradient (PBiCG) solver for skew-symmetric matrices withthe diagonal incomplete LU (DILU) matrix used as pre-conditioner. The maximum CFL (Courant-Friedrichs-Lewy)number allowed in the simulations is 0.4. LES equationsof reacting flows are obtained by both Favre and implicitfiltering operations and are written as

∂ρ∂t + ∂

∂xi(ρui) = 0

∂∂t (ρui) +

∂∂xj

(ρuiuj) = − ∂p∂xi

+ ∂∂xj

[τij − ρ(uiuj − uiuj)]

∂∂t (ρYk) +

∂∂xi

(ρuiYk) = − ∂∂xi

[Vk,iYk + ρ(uiYk − uiYk)

]+ωk ; k = 1, ., N ,

∂∂t (ρh

s) + ∂∂xi

(ρuihs) = ∂∂xi

[λ ∂T∂xi

− ρ(uihs − uihs)]

−N∑k=1

hokωk ; k = 1, ., N ,

(1)

where usual symbols are used to describe the compressibleNavier-Stokes equations. Thermophysical properties of the

Proceedings of the World Congress on Engineering 2017 Vol II WCE 2017, July 5-7, 2017, London, U.K.

ISBN: 978-988-14048-3-1 ISSN: 2078-0958 (Print); ISSN: 2078-0966 (Online)

WCE 2017

species k are calculated using JANAF tables. The unresolvedReynolds stress tensor (uiuj − uiuj) is modelled usingthe dynamic Smagorinsky model. The unsolved speciesflux (uiYk − uiYk) and the enthalpy flux (uihs − uihs)are described using a simple gradient assumption, giving(uiYk− uiYk) = −νsgsSct

∂Yk∂xi

and (uihs− uihs) = −νsgsPrt∂hs

∂xi.

The filtered mass and heat diffusion fluxes are alsospecified using a simple gradient hypothesis leading toVk,iYk = −ρDk

∂Yk∂xi

and λ ∂T∂xi = λ ∂T∂xi . Concerning thediffusion coefficients Dk of the different species k, they areconsidered as constant for each of the species, obtained on a1D laminar freely propagating flame using CANTERA [11].They are defined considering the median values of Schmidtnumbers Sck [12] such as Dk = ν/Sck. Finally, the filteredreaction rate of species k, ωk is modelled using the finiterate chemistry Partially Stirred Reactor (PaSR) approachderived by Karlsson [13] along with the skeletal mechanismof Karalus [14] to handle CH4-H2 mixtures. For finite ratechemistry, integration of the source terms ωk is realizedby ISAT-CK7 [15], with a tolerance error of 1 ×10−4 incombination with the “SkeletalGRI” [14] mechanism. Thenumerical procedure used in this work has been previouslyvalidated by the authors [16] and Fooladgar et al. [17].

III. COMPUTATIONAL SETUP

The swirl burner used in this work is based on the reduced-scale model of Nogenmyr et al. [18]. Four tangential inletsand a radial swirler allow the burner to be operated fromlow to high swirl numbers. The main benefit of studying areduced-scale combustor resides in the potential reproductionof this computational domain for CFD studies. Indeed, inimplicit filtering of LES as used in this work, the filter isof the same size as the mesh. By using reduced models,it allows to mesh the computational domain with cell sizeof around that of laminar flame thickness and allowing theuse of simpler combustion models. Fig. 1 shows a schematicrepresentation of the burner.

By definition, the swirl number Sw is specified as the ratioof the axial flux of angular momentum to the product of theaxial flux of axial momentum with a characteristic radius R.For a tangential jet-induced swirl flow, it can be assessed as

Sw =

∫ R0UxUθr

2dr

R∫ R0U2xrdr

≈ πD

16WH

D −W(mx/mr + 1)2

, (2)

where D is the diameter of the axial inlet, W and H arerespectively the width and height of the tangential jets andmx/mr is the ratio of the axial to the tangential mass flowrates. An overall Reynolds number Re is then defined as

Re =mx + mr

µ(πD/4), (3)

where µ is the dynamic viscosity of the unburned mixture.Flames with a global equivalence ratio φ are investigated fordifferent mixtures composed of methane-hydrogen-air withthe volume fraction β of hydrogen in the CH4-H2 blenddefined by β = XH2/(XH2 +XCH4).

In the following section, the computational setup is firstpresented for the LES simulations with the specification ofthe different boundary conditions. Comparison of the resultsobtained in the simulations with experiments is only visual.

Fig. 1: A schematic drawing of the burner. All units are inmillimeter.

Concerning the prediction of species NO, the results are inthe same range as what has been measured by Fooladgar [19]and they are only studied numerically. A range of flow andflame configurations have been simulated by Cicoria [20],and configurations of simulations in this paper are given inTable I.

TABLE I: Parameters for the different configurations ofstudy.

Sw φ β (%) Re0.7 0.85 0 60000.7 0.85 10 60000.7 0.85 20 60000.7 0.85 30 60000.7 0.85 40 60000.7 0.85 50 6000

0.7 0.95 0 60000.7 0.95 20 60000.7 0.95 40 6000

Fig. 2 shows an overview of the computational domainwith the location of the flame in a premixed case, repre-sented by the instantaneous fuel mass fraction at 0.01. Thenumerical domain is composed of the swirler, combustionchamber and a large rectangular region downstream of thecombustor exit to consider the presence of air atmosphereand to diminish the impact of outflow boundary conditionsin the combustion chamber. Dimensions of the computationalconfiguration are given in Fig. 1.

The computational mesh consists of 2.5 million hexahedralcells, which are body fitted inside the domain. Inside theburner and combustor, the resolution is around ∆x = 0.5mm per cell, leading to 34 cells over the diameter of theaxial inlet and to maximum ratio of filter size to thermalflame thickness of 2 for the different reacting flows. For allwall boundary conditions, a boundary layer constituted of

Proceedings of the World Congress on Engineering 2017 Vol II WCE 2017, July 5-7, 2017, London, U.K.

ISBN: 978-988-14048-3-1 ISSN: 2078-0958 (Print); ISSN: 2078-0966 (Online)

WCE 2017

Proceedings of the World Congress on Engineering 2017 Vol II WCE 2017, July 5-7, 2017, London, U.K.

ISBN: 978-988-14048-3-1 ISSN: 2078-0958 (Print); ISSN: 2078-0966 (Online)

WCE 2017

Proceedings of the World Congress on Engineering 2017 Vol II WCE 2017, July 5-7, 2017, London, U.K.

ISBN: 978-988-14048-3-1 ISSN: 2078-0958 (Print); ISSN: 2078-0966 (Online)

WCE 2017

Proceedings of the World Congress on Engineering 2017 Vol II WCE 2017, July 5-7, 2017, London, U.K.

ISBN: 978-988-14048-3-1 ISSN: 2078-0958 (Print); ISSN: 2078-0966 (Online)

WCE 2017

Proceedings of the World Congress on Engineering 2017 Vol II WCE 2017, July 5-7, 2017, London, U.K.

ISBN: 978-988-14048-3-1 ISSN: 2078-0958 (Print); ISSN: 2078-0966 (Online)

WCE 2017