Simulation of the Delft Jet-in-Hot-Coflow burner using transported ...

COMBURA’12Combustion ReseaRCh and appliCation

octo

ber

&34Kasteel VaesharteltMaastrichtThe Netherlands

i

COMBURA 2012 was organized on behalf of STW-Platform Clean and Efficient Combustion Technology Foundation STW NVV (Nederlandse Vlamvereniging) Dutch Section of the Combustion Institute and was organized by Theo van der Meer, University of Twente Howard Levinsky, University of Groningen and KEMA Nederland BV Pepijn Pronk, Tata Steel Bart Somers, Eindhoven University of Technology Leo Korstanje, Technology Foundation STW

Platform Clean and Efficient Combustion Prof.dr.ir. Th.H. van der Meer (chairman) Universiteit of Twente Prof.dr.ir. R.S.G. Baert TNO Automotive Dr.ir. M.F.G. Cremers KEMA Nederland BV Ir. J.J. van Dijk Agentschap NL Prof.dr. L.P.H. de Goey Eindhoven University of Technology Ir. M. van Hal DAF Trucks Ir. M. Houkema ECN Dr.ir. W. de Jong Delft University of Technology Prof.dr. H.B. Levinsky University of Groningen and

KEMA Nederland BV Dr.ir. L. Post Shell Global Solutions Int. BV Dr.ir. P. Pronk Tata Steel Dr.ir. C.J.A. Pulles KIWA Gas Technology Prof.dr. D.J.E.M. Roekaerts Delft University of Technology Dr. L.J. Korstanje (secretary) Technology Foundation STW Secretariat Astrid van der Stroom Technology Foundation STW P.O. Box 3021 3502 GA Utrecht The Netherlands Tel: +31 (0)30 600 1 297 Fax:+31 (0)30 601 44 08 E-mail: [email protected]

Photo front cover: Room for ID's, Nieuwegein

ii

Abstracts of CCC projects, 10 October 2012 3

1. flexFLOX - Flameless combustion conditions and efficiency improvement of 4 single- and multi-burner-FLO X furnaces in relation to changes in fuel and oxidizer composition

Luis Arteaga-Mendez, Delft University of Technology Effect of fuel and oxidizer composition on jet-in-coflow flames Eun-Seong Cho, Delft University of Technology Local temperature characteristics in a flameless oxidation furnace

Gerasimos Sarras, Delft University of Technology Simulation of the Delft Jet-in-Hot-Coflow burner using transported PDF method and 3D-FGM tabulated chemistry

2. HiTAC Boiler - Heavy fuel-oil combustion in a HiTAC boiler 10 Hugo Rodrigues, Delft University of Technology

Spray flame experiments in a jet- in-vitiated-coflow burner in flameless conditions Sanglong Zhu, University of Twente Heavy fuel-oil combustion in a HiTAC boiler

3. MILDNOx - NO Formation and fuel flexibility in dilute combustion 13 Ebrahim Abtahizadeh, Eindhoven University of Technology

Fuel flexibility and NO Formation in dilute combustion Farimah Pouyandeh, Eindhoven University of Technology Investigation of fuel flexibility on the auto ignition behavior of MILD combustion in Jet-in-hot-coflow burners

4. MoST - Multi-scale modification of swirling combustion for optimized gas turbines 17 Thiago Cardoso de Souza, Eindhoven University of Technology

Resonant turbulence in premixed combustion 5. XCiDE - Crossing the Combustion modes in Diesel Engines 19 Ulaş Egüz, Eindhoven University of Technology

Baseline n-heptane case (Spray H) simulations with the FGM method Niels Leermakers, Eindhoven University of Technology Combustion phasing controllability with dual fuel injection timings

6. ALTAS - Advanced low NOx flexible fuel gas turbine combustion, aero 23

and stationary Andrea Donini, Eindhoven University of Technology Numerical simulations of premixed turbulent combustion using the Flamelet Generated Manifold approach with heat loss inclusion

7. BiOxyFuel - Torrefied biomass combustion under oxy-fuel Conditions in coal fired 25

power plants Eyerusalem Gucho, University of Twente Co-combustion characteristics and kinetics of torrefied beech wood with bituminous coal 8. ULRICO - Ultra Rich Combustion of Hydrocarbons and Soot Formation 27 Michael Stöllinger, Delft University of Technology

ULRICO: Ultra rich combustion of hydrocarbons and soor formation – numerical modeling using PDF methods Mark Woolderink, University of Twente ULRICO: Ultra Rich Combustion of hydrocarbons and soot formation – design of a flameless oxidation burner

iii

Abstracts of presentations, 4 October 2012 31

1. Jan Anker, (NUMECA) Combustion modeling of gaseous flames 32

2. Peter-Christian Bakker, (Eindhoven University of Technology) 34

Application of Partially Premixed Combustion using low octane fuels in a heavy-duty diesel engine

3. Maurice Corvers (Eindhoven University of Technology) 36

Optical diagnostics on Partially Premixed Combustion in a heavy-duty diesel engine 4. Sander Gersen, (DNV KEMA) 38 Autoignition properties of gaseous fuels at conditions relevant to engines and turbines

5. Francisco Hernandez Perez, (Eindhoven University of Technology) 41 Large-Eddy Simulation of a laboratory-scale flame in the mild combustion regime 6. Jim Kok, (University of Twente) 42

Research on gas turbine combustors at University of Twente 7. Dr. Werner Krebs, (Siemens) 44

Combustion technologies for future gas turbines: design methodology and validation 8. Lu, Jie, (Delft University of Technology) 45

CARS measurements of temperature in a flameless oxidation furnace 9. Mohammad Mir Najafizadeh, (Delft University of Technology and IUST Teheran) 47

Chemical kinetic analysis in turbulent lifted flames in a hot coflow 10. Kirti Bhushan Mishra, (BAM Federal Institute for Materials Research and Testing) 49

Applications of peroxy-fuels in vehicle propulsion 11. Ferry Tap, (Dacolt) 51 FGM modelling of a gas turbine model combustor 12. Jeroen Vancoillie, (Ghent University) 53

Fundamental study of methanol flames by use of contained explosions and flat flame burner measurements

13. Prof. Luc Vervisch, (INSA/CORIA, Rouen) 55 Combustion, flames and burner design: challenges and computing tools 14. Moresh Wankhede, (Dacolt) 57

Strategies for combustor design 15. Hai Wu,(Tata Steel) 59 Enhance radiant heat transfer by helical inserts in an annealing furnace 16. Ron Zegers, (Eindhoven University of Technology) -- Flame lift-off as function of ignition delay for cyclic oxygenates

iv

Abstracts of posters 60 1. Cemil Bekdemir, Eindhoven University of Technology 61

Tabulated chemical kinetics for efficient and detailed simulations of Diesel engine combustion

2. Richard Haas-Wittmüß, OWI Oel-Waerme-Institut 63

Liquid fuel burner for the measurement of adiabatic laminar burning velocities 3. Carlo Locci, IFP Energies Nouvelles 64

Towards LES simulation of flameless combustion: diluted homogeneous reactors (DHR) tabulation applied to Sandia Flames D and F

4. Evren Ozcan, University of Technology 65 Numerical modelling of self-excited thermo-acoustic instabilities in the DESIRE

Combustor

5. Luck Peerlings, Eindhoven University of Technology 66 Flame saturation current as a measure of the flame thermo-acoustic behavior

6. Joost Sallevelt, University of Twente 68 Numerical and experimental study of ethanol combustion in an industrial gas turbine 7. F. Schlösser, OWI Oel-Waerme-Institut 69

Development of a compact combustion chamber for liquid fuels with high energy densities

8. Juan Carlos Roman Casado, University of Twente 70 Experiments with the LIMOUSINE combustor (NOTE: new version sent) 9. Santosh Kumar, University of Twente 72

Numerical simulation of a stable/unstable combustion in a bluff-body stabilized combustor

List of participants COMBURA 2012 75

Abstracts of CCC-projects

3

TNF11 “Eleventh International Workshop on Measurement and Computation of Turbulent Flames”, July 26–28, 2012, Darmstadt, Germany

Effect of fuel and oxidizer composition on jet-in-coflow flames L. D. Arteaga Mendez1, M. J. Tummers1 and D. J. E. M. Roekaerts1

1 Laboratory for Aero & Hydrodynamics, Faculty of Mechanical, Maritime and Materials Engineering, Delft University of Technology

[email protected] Keywords: Flameless combustion, fuel flexibility, jet in hot coflow

Flameless oxidation can contribute to reducing NOx emissions while delivering high fuel efficiency. In this combustion technology, the reactants are injected at high momentum so that they are mixed with combustion products before reacting. Important benefits may result if applied to non-conventional fuels such as biogas, coke oven gas and refinery gas. To support this development, the effect of hydrogen and carbon dioxide addition on the stabilization mechanism of natural gas diffusion flames using either air and or hot, lean combustion products as oxidizers was studied using high speed recordings of the flame luminescence and particle image velocimetry. The hot coflow has relatively low oxygen content mimicking the characteristic oxidizer conditions in flameless oxidation. Figure 1 presents the visual appearance of four exemplary flames that were considered in this study. Addition of hydrogen to DNG results in flames that stabilize closer to the fuel pipe nozzle while increasing the visible flame length. Addition of carbon dioxide to DNG has the opposite effect.

Fig. 1 Visual appearance of four studied flames: a) 10/90 CO2/DNG in cold coflow, b) 10/90 CO2/DNG in hot coflow, c) 10/90 H2/DNG in cold coflow and d) 10/20 H2/DNG in hot coflow. For the hot coflow the O2 concentration is 9.92% (mass basis) and the temperature is 1460 K. The window height is approximately 70 cm.

Fig. 2 Sequence of flame luminescence images showing the evolution of an autoignition kernel for a DNG flame in hot coflow. The time between images is 4 ms.

Flames oxidized in cold air have a sharp but irregular boundary between the reactants and the flame zone. The position of the sharp boundary is fluctuating in time. The fuel composition affects the mean liftoff height and the RMS value of the fluctuations in liftoff height. These fluctuations increase when the liftoff height increases due to the presence of larger turbulent structures at greater distances from the fuel pipe nozzle.

Flames oxidized in hot coflow with low oxygen concentration exhibit a different stabilization mechanism. In this type of flames stabilization is achieved by autoignition kernels that grow while being convected downstream. The sequence of images of the flame luminescence in Fig. 2 shows the evolution of an autoignition kernel with time intervals of 4 ms for a DNG flame in hot coflow. Figure 3 shows the probability of flame luminescence as a function of height above the fuel pipe nozzle. The liftoff height was defined as the height where the probability of flame luminescence is 50%. Hydrogen addition strongly influences the stabilization position of the flame. The liftoff height is reduced by a factor of five for

4

NF11 “Eleventh International Workshop on Measurement and Computation of Turbulent Flames”, July 26–28, 2012, Darmstadt, Germany

hydrogen addition of 10% as compared to the pure DNG flame at identical Reynolds

0 50 100 150 2000

0.5

1

P in

−

25%H

210%H

25%H

2 DNG

50 100 150 200 2500

0.5

1

P in

−

z in mm

10%CO2

20%CO2

30%CO2

DNG

a)

b)

Fig. 3 Probability of flame luminescence as a function of the axial position. a) Results for H2/DNG fuel mixtures oxidized in hot coflow, b) Results for CO2/DNG fuel mixtures oxidized in hot coflow.

number. The effect of carbon dioxide addition on the liftoff height is not as substantial as that of hydrogen addition. Increasing the carbon dioxide content in the fuel increases the liftoff height. However, for a carbon dioxide addition of 30% the increase in liftoff height is only 10% (as compared to that for pure DNG at identical Reynolds number). Figure 4 shows the mean flow field for a 25/75 H2/DNG flame oxidized in cold coflow (left) and hot coflow (right). In the lower region the coflow has a negative radial velocity component due to the entrainment of jet fluid. Near the liftoff height (horizontal magenta line in Fig. 4) the radial velocity changes sign due to sudden heat release and thermal expansion.

−20 −10 0 10 20

0

10

20

30

40

50

60

70

80

r in mm

Mean Velocity in m/s

0

2

4

6

8

10

12

14

16

18

20

−20 −10 0 10 20

0

10

20

30

40

50

60

70

80

r in mm

z in

mm

Mean Velocity in m/s

0

5

10

15

20

25

30

35

b)b)a)a)

Fig. 4 Mean flow field for a 25/75 H2/DNG flame oxidized in a) cold coflow and b) hot coflow. The arrows indicate the velocity vectors and the solid lines denote streamlines. The solid magenta line marks the mean lift off height.

Figure 5a shows the images of the seed particles when these are illuminated in the PIV laser light sheet. The boundary between the relatively cold reactants and the hot combustion products is marked by a green line. Figure 5b shows the corresponding instantaneous flow field with the above mentioned boundary superimposed. The stabilization point is located in the lean side of the shear layer between the jet and coflow fluid. Large turbulent structures interact with the reaction zone deforming it and changing its position. The fuel composition has a substantial impact on the flame stability for flames in cold coflow. Hydrogen addition to the fuel will substantially increase the stability while carbon dioxide addition has the opposite effect. In flames oxidized in hot coflow, only hydrogen addition has a strong influence on the growth velocity of autoignition kernels and the stream wise extend of the region where the autoignition events occur while carbon dioxide addition to the fuel will slightly shift the reaction zone downstream.

Fig. 5 a) Raw PIV image with green line marking the boundary between the reactants and the hot combustion products and b) the instantaneous flow field of a 25/75 H2/DNG flame in cold coflow. ACKNOWLEDGEMENTS This CCC project (flexFLOX) is supported by STW, NVV, Tata Steel, Shell, TNO, Numeca Int., and WS Gmbh.

5

Combura’12 symposium, Kasteel Vaeshartelt, Maastricht, The Netherlands, October 3-4, 2012

LOCAL TEMPERATURE CHARACTERISTICS IN FLAMELESS OXIDATION FURNACE

Eun-Seong Cho, Arjen Jansen, Wiebren de Jong and Dirk J.E.M. Roekaerts Process & Energy, 3mE, Delft University of Technology, 2628 CA Delft, The Netherlands ([email protected])

Key words: Flameless oxidation, local temperature measurement, CFD simulation

Flameless Oxidation (FLOXTM [1]) is a promising combustion technology to accomplish high efficiency and low emissions. High momentum injection of the fuel and air entrain the flue gas through internal recirculation, thus diluting the oxygen concentration in the local combustion zone. This leads to a more distributed heat release rate, avoiding high peak temperature, causing reduction of the thermal formation of NOx. Combined with a high preheat temperature of the combustion air, this combustion technique achieves a high efficiency. The characteristics of multi-burner flameless oxidation furnace have been investigated [2] in a semi-industrial 300kWth regenerative FLOXTM system at TU Delft. Burner configurations and operating modes (parallel and staggered) have been compared for furnace efficiency, emissions (NO, CO) and temperature uniformity ratio. Additionally, the excess air ratio (λ) and cycle time have been varied. Also, a two burner pair 200kWth experiment has been conducted to evaluate the characteristics of partial combustion load of real industrial operating condition [3]. Figure 1 shows the schematic diagram of the two burner pair case.

Figure: 1 Schematic diagram of regenerative multi-burner flameless oxidation furnace.

A sstudy of the local combustion characteristics is needed to understand flameless combustion in detail. For this study a 3-D CFD simulation was performed in flameless conditions and local flame temperature was measured using a long S-type thermocouple. The measured results are used for the validation of numerical models. Figure 2 shows a schematic diagram of the local measurement campaign. The temperature was measured (a) in the near burner region and (b) in the direction of the fuel jet axis.

Figure: 2 Schematic diagram of local temperature

measurements campaign.

The simulations were performed [4] using FLUENT 6.3. The realizable k-ε turbulent model was used. For the combustion of natural gas the Smooke mechanism (16 species, 46 reactions) was applied. The turbulence-chemistry interaction is taken into account by using the Eddy Dissipation Concept model. Radiative heat transfer is calculated with discrete ordinate method using a WSSG model. A heat transfer coefficient was used to represent wall heat loss.

Figure: 3 Temperature contour of 3-D numerical

simulation result (K).

6


Figure 3 shows temperature contour in two orthogonal planes through the burner. It shows that large differences in mean temperature are absent. A zone with higher temperature appears at the wall opposite to the burner.

Figure 4 show the furnace inside temperature contour in the near burner region in parallel (a) firing and (b) regenerating conditions by S-type long thermocouple measurement. The X-axis represents the distance from the side wall (x=-750 mm) to the furnace center (x=0 mm). The Z-axis shows the distance from the burner. At each position temperature was measured for around 5 minutes before moving to the next position. Temperature fluctuates are accompanying the regeneration cycle, with the firing period showing high temperature and the regenerating period showing low temperature in the near burner region.

Z di

stan

ce [m

m]

0

100

200

300

400

500

600

700

700 800900100011001200

(a) Parallel Firing

X distance [mm]

-750 -600 -450 -300 -150 0

Z di

stan

ce [m

m]

0

100

200

300

400

500

600

700

700 800900100011001200

(b) Parallel Regenerating

X distance [mm]

-750 -600 -450 -300 -150 0

Figure: 4 Temperature contour in near burner region.

Figure 4 (a) firing condition shows the high temperature region in front of the fuel nozzle position around z=250 mm from burner surface. In that region fuel and oxidizer are mixed and ignition happens. In the regenerating condition (Figure 4 (b)) the fuel nozzle region shows low temperature because fuel temperature supply with ambient temperature.

Figure 5 shows the temperature profile along the burner central axis, from burner nozzle to the opposit

wall. Temperature increases until 750 mm from the burner and then decreases. The highest temperature is over 1200oC, being 200oC higher than the average furnace temperature.

Parallel

Axial distance from burner fuel nozzle [mm]

0 300 600 900 1200 1500

Tem

pera

ture

[o C]

700

800

900

1000

1100

1200

1300

Max temp. (firing)Min temp. (regenrating)

Figure: 5 Temperature profile of fuel jet axis with distance from the burner.

The temperature profile from the CFD simulation is not in good agreement with the measured temperature profile. The numerical simulation results show more delayed combustion phenomena than experimental results. Model improvement is needed, in particular in the area of modeling of ignition in turbulent conditions, in order to better estimate the temperature field during flameless operation.

ACKNOWLEDGEMENT

The flexFLOX project (10428) is financially supported by the Technology Foundation STW and the Dutch Flame Foundation NVV.

REFERENCE

[1] J.A. Wünning, J.G. Wünning. Flameless oxidation to reduce thermal NO-formation. Prog. Energy Combust. Sci. 23 (1997) 81-94.

[2] E.-S. Cho, B. Danon, W. de Jong, D.J.E.M. Roekaerts. Behavior of a 300kWth regenerative multi-burner flameless oxidation furnace. Applied Energy 88 (2011) 4952-4959.

[3] E.-S. Cho, D. Shin, J. Lu, W. de Jong, D.J.E.M. Roekaerts. Configuration effects of natural gas fired multi-pair regenerative burners in a flameless oxidation furnace on efficiency and emissions. Submitted to Applied Energy (2012).

[4] A. Jansen. Numerical study on a multi-burner flameless oxidation furnace in relation to change in fuel composition. Master Thesis, TU Delft (2012).

7

Simulation of the Delft Jet-in-Hot-Coflow burnerusing transported PDF method and 3D-FGM

tabulated chemistry

G. Sarras1, M.K.Stöllinger2, D.J.E.M. Roekaerts1,2

1Department Process and Energy, Delft University of Technology,Leeghwaterstraat 44, 2628 CA Delft, [email protected] of Multi-Scale Physics, Delft University of TechnologyLeeghwaterstraat 39, 2628 CB Delft, The Netherlands

1. MotivationFlameless combustion is a clean combustion technique where the oxidizer and/or fuel are diluted withhot combustion products before entering the main reaction zone. This technique relies on exhaust gasheat recovery and high recirculation ratios. In order to mimic the important characteristics of flamelesscombustion without the complications of a real furnace, a simplified laboratory scale configuration hasbeen realized in the Delft Jet-in-Hot-Coflow (DJHC) burner. The DJHC burner creates a turbulentdiffusion flame of a gaseous fuel in a coflowing oxidiser stream of high temperature with low oxygenconcentration. In this conditions the reaction rates are lower due to oxygen dilution as compared toconventional diffusion flames. The Reynolds numbers in all experimentally studied DJHC flames are notvery high (Rejet < 104). To handle adequately the influence of turbulent fluctuations on mean reactionrates is a modeling challenge.

Transported probability density function (PDF) method allows to include the effects of turbulence -chemistry interaction in the Reynolds Averaged Navier Stokes (RANS) framework. The distinct advantageof this method is that the mean reaction rate is treated exactly for complex finite rate chemistry. However,the high computational costs related to the integration of the reaction rates can result in prohibitivecomputational time. A promising method of simplified chemistry is based on the so called FlameletGenerated Manifold (FGM) [1].

2. Modeling approachA hybrid finite-volume transported PDF method is used [2]. It combines an underlying Reynolds stressmodel with the Monte Carlo simulation of velocity and scalar statistics. The joint velocity-compositionPDF or the joint composition PDF is represented by a number of particles and stochastic differentialequations are solved for particles properties. Statistics of averaged quantities are estimated simply byan ensemble averaging. The local thermochemical state of the mixture is assumed to evolve on a FGMtable.

To construct the FGM that is suitable to describe the chemistry in the DJHC two main issues haveto be considered. First, the spatial non-uniformity of the coflow composition and temperature requirestwo mixture fractions to describe the mixing in the DJHC. A suitable FGM has to be at least threedimensional (two mixture fractions and a progress variable). At any point of the flow field, the 3-streammixing is represented by two mixture fractions Z1 and Z2. Z1 describes the mixing between the fuel andthe coflow stream, while Z2 the entrainment of the surrounding air and the radial variation of oxygenconcentration at the coflow inlet. The radial variation of the enthalpy at the coflow inlet is implicitlyaccounted through the second mixture fraction Z2. Second, to adequately describe the auto-ignitioneffects which have been observed experimentally in the DJHC, unsteady non-premixed flamelet solutionsare used to construct the FGM. The FGM is based on one dimensional, unsteady, non-premixed flameletsbetween the fuel (Z1 = 1) and a certain coflow composition. The unsteady flamelet solutions are obtainedwith the CHEM1D code [3].

A suitable progress variable is defined as Y = YH2MH2

+ YH2O

MH2O+ YCO2

MCO2, where Mi denotes the molecular

weight of the considered species. The time coordinate is then transformed into the reaction progress

18

variable e.g. T (Z1, t) → T (Z1, Y ) and all relevant thermo-chemical variables are stored in a 2D FGMtable. Several 2D FGM tables are created corresponding to different values of Z2. A total of eight 2DFGM tables are used to construct a 3D FGM. A single strain rate a = 100s−1 was used in all the flameletcalculations corresponding to an axial average of the mean scalar dissipation rate at the stoichiometricmixture fraction. Using the 3D FGM all thermo-chemical variables are given by e.g. Yi = Yi(Z1, Z2, Y ).To improve the memory efficiency, the 3D FGM is tabulated using the Delft FLAME code [4].

Every Monte-Carlo particle carries two mixture fractions Z1, Z2 and a progress variable Y . Thismeans that every particle now evoles in composition space according to the general equation:

dφ∗α,mix = θ∗α,mixdt+ Sα(~φ)dt, (1)

where ~φ ≡ (Z1, Z2, Y ) and α = 1, 2, 3. Also, θ∗α,mix represents the IEM, EMST or modified CD micro-mixing model and S3(~φ) the chemical source term of the progress variable. The mixture fractions arenot affected by chemical reactions (S1,2(~φ) = 0) and their values only change according to the chosenmicro-mixing model.

Figure 1: Radial profiles of the mean temperature at several axial locations for IEM, EMST and CDmicromixing models. The dashed line denotes simulation results with joint velocity-compositionPDF(JVCPDF), the solid line denotes simulation results with joint composition PDF (JCPDF) and thecircles denote experimental results [5].

References[1] J. A. van Oijen, F. A. Lammers, and L. P. H. de Goey. Modeling of complex premixed burner systems

by using flamelet-generated manifolds. Combustion and Flame, 127(3):2124 – 2134, 2001.

[2] B. Naud, C. Jiménez, and D. Roekaerts. A consistent hybrid PDF method: implementation detailsand application to the simulation of a bluff-body stabilised flame. Progress in Computational FluidDynamics, 6:147–157, 2006.

[3] CHEM1D. A one-dimensional laminar flame code. Eindhoven University of Technology.www.combustion.tue.nl/chem1d.

[4] T.W.J. Peeters. Numerical modeling of turbulent natural-gas diffusion flames. PhD thesis, TechnischeUniversiteit Delft, 1995.

[5] E. Oldenhof, M.J. Tummers, E.H. van Veen, and D.J.E.M. Roekaerts. Role of entrainment in thestabilisation of jet-in-hot-coflow flames. Combustion and Flame, 158(8):1553 – 1563, 2011.

29

Spray flame experiments in a jet-in-vitiated-coflow burner in

flameless conditions

H. Rodrigues, M.J. Tummers and D.J.E.M. Roekaerts

Laboratory for Aero & Hydrodynamics, Faculty Mechanical, Maritime and MaterialsEngineering, Delft University of Technology

September 8, 2012

The goal of this work is to determine the fea-tures of ethanol turbulent spray flames in ExcessEnthalpy Combustion (EEC) combustion and ob-tain a dataset that can be used for further develop-ment and validation of advanced combustion mod-els. In combination with regenerative burners, EECallows for the desired combination of low nitrogenoxide (NOx) and improved efficiency. This combus-tion process produces low amounts of nitric oxidefor light oil products, however, combustion of otherliquid fuel oils revealed significant differences andthe effects of the droplet size and dispersion are notyet understood.

Here we report first experimental results ondroplet statistics in a jet-in-vitiated-coflow burneroperating in EEC conditions. Under hot-vitiatedconditions a low-luminescent highly transparentflame with no sharp flame-front is observed. Theflame forms a clearly lifted flame at some distancefrom the nozzle with a faint-blue part in the lowerregion and high luminosity at the edges. The lift-off height does not vary with increasing injectionpressure.

Droplet velocity and diameter measurementsperformed with Phase Doppler Anemometry(PDA) are presented in figure 1. Due to the ini-tial high velocity ratio between droplets and gas, arapid momentum transfer from the droplets to thegas occurs, leading to deceleration of the droplets.The measured mean droplet diameter and the spraydispersion show good symmetry with respect tothe axis of the burner. By visual inspection itis observed that some droplets reach the regionoutside the spray cone. The obtained dataset on

Z=35mm

Z=20mm

Z=12mm

Z=8mm

−30 −20 −10 0 10 20 30

20

40

60

D10

[um

]

Radius [mm]

−40 −30 −20 −10 0 10 20 30 400

5

10

15

20

Uz [m

/s]

Radius [mm]

z=8mmz=12mmz=20mmz=35mm

Figure 1: Spray flame in hot-vitiated coflow (top) andcorrespondent mean droplet distribution (bottom)

spray statistics forms a first essential ingredient of adatabase for model validation. However, the resultsalso allow an analysis of the feasibility of detailedcharacterisation of inflow boundary conditions ofthis type of spray flame.

Acknowledgement

This work is supported by the Technology Founda-tion STW, Stork Thermeq and Shell.

1

10

Heavy fuel-oil combustion in a HiTAC boiler

Shanglong Zhu1, Artur Pozarlik1, Theo van der Meer1*,

Hugo Rodrigues2, Mark Tummers2, Dirk Roekaerts2, Bart Venneker3

1. University of Twente; 2. Delft University of Technology; 3. Stork Thermeq B.V.

[email protected]

Key words: High temperature air combustion, spray combustion, boiler

The High Temperature Air Combustion (HiTAC) process lends itself ideally for the

combustion of all sorts of “difficult” fuels, ranging from low-calorific gases such as

waste-gases, to heavy fuel oil. Especially for heavy fuel oil, expectations are that in

combination with HiTAC these can be utilized for steam generation with very low harmful

emissions such as NOX, CO and particulates. And the key features of this high-efficiency

combustion process can be utilized to lead to simpler, cheaper and more reliable designs of

boilers, with very low emissions of harmful species.

This project concerns the extension of the application of HiTAC to heavy-oil combustion

processes in a boiler. In this poster an overview of the project progress is given. To generate

the knowledge needed to develop and design such a boiler, experimental and computational

investigations are being made of turbulent ethanol spray flame under HiTAC conditions,

which is also called “the Delft spray flame”. However, since little is known about spray

combustion under HiTAC condition to date, validation of the computational models involving

turbulence, atomization, evaporation, combustion, radiative heat transfer, etc., under

conventional conditions is required, and a conventional methanol spray flame in a chamber at

the National Institute of Standards and Technology, which is also called “the NIST flame”,

was simulated and compared to the measured data for validation.

The NIST flame experiment was carried out in a combustion chamber. Swirling combustion

air passes through the outer annulus passage at ambient pressure and temperature. A

pressure-jet nozzle forms a hollow-cone methanol spray at ambient temperature. In the

present study, the previous simulations of the NIST flame were studied and the features of this

flame, including the boundary conditions of the inlet air and the spray, were analyzed to relate

the experiment and simulations. We performed a numerical simulation in ANSYS Fluent with

the steady flamelet model in order to include detailed chemistry and the influence of the

evaporation on mixture fraction variance was investigated. Predictions of the mean velocity

components of air flow and droplets, droplet number density, and Sauter Mean Diameter

(SMD) at various elevations were compared with the experimental data and they showed

good agreements.

In the Delft flame experiment, the liquid ethanol spray created by a pressure atomizer is

surrounded by a co-flow produced by a pre-combustor in an open system. Various

temperature and O2 concentration of the co-flow are generated to investigate their influences

on spray combustion. The measured data are used for defining boundary conditions and

further validation of the simulations.

Preliminary simulation results of the Delft flame showed minor influence of the ambient air

11

mailto:[email protected]

on the flame zone, which indicates that the combustion characteristics depend only on the

co-flow. With the steady laminar flamelet model, the predicted cold inner zones in the flame

and flame profiles are validated, and the preliminary simulation results with different

temperature (300K, 600K, 900K, 1200K and 1500K) and O2 concentration (21%, 18%, 15%,

12%, 9% and 6%) of the co-flow shows that the averaged peak temperature increases with the

increasing temperature of the co-flow and decreases with the decreasing O2 concentration of

the co-flow. Under HiTAC conditions, the temperature is more uniformly distributed and the

flame is more stable. Further validation will be made with the results of recent (PDA) and

planned (PDA+LDA, CARS) measurements and it will be used to obtain a better

understanding of the influences of the temperature and O2 concentration on light fuel-oil

spray combustion.

A field test with heavy fuel-oil will be carried out in a 9MW boiler, and the water-steam

system is optimized. An oil gun with industrial atomizer, such as steam-blast atomizer is used,

and surrounded by separated air flow, primary air and secondary air. 3D simulations of the

boiler are made with two-step global reaction mechanism and empirical droplet size

distribution. Similar to the simulation results of the Delft flame, preliminary results with the

existing burner showed that the increasing of the temperature of the combustion air leads to

higher peak temperature, while reducing O2 concentration of the combustion air results in

more uniform temperature distribution, but instead of thermal NOx, fuel NOx is dominant in

heavy fuel-oil combustion, and soot formation increases with low O2 concentration. As a

result, further investigation of the HiTAC boiler will focus on the mixing of flue gas with the

combustion air and/or heavy fuel-oil, and the research on the Delft flame and the 3D

simulation of the boiler will be used for better understand of heavy fuel-oil under HiTAC

condition and the corresponding HiTAC burner design.

Acknowledgement

The authors would like to thank STW/NVV/Stork Thermeq for sponsoring this project.

12

��

�

��

��

�� !��"��

�� #�$��%�!��&

��'�

��!��"��#��$��#��%��!��&'�'(�)*��*+ �("��#��!��$��#��%,��!��-.�./,,��!��

��"�/�&'�'(�) -�-. ��/,��!��

��(�� )��

� ��

��

��

��

��

� ��

��

��

��

� � ��

��

��

��

��

��

��

��

��

��

��

� ��

�

��*�� &��!�� % � ��

0��%��1'

� ��

��

��

��

��

��

��

��

� ��

��

��

�� !��

�� !� �� !��

��

�� " ��

�� !��

�� #� ��

�!� ��

� ��

�� $%��

��

�

��*��

��&��%%��!��%��%

��'

��

� ��

� ��

� ��

� ��

��

� ��

��

��

��

��

� �� !��

13

��

�

��

� ��

�� !��

��*�� '��

�& �� %%�� !�� %�� % ��

��'

�� &��

��

��

�� '��

� �� '��

��

��

�

��*�� +�� %%��!��%��%

��'

��

�� #��

�� '��'�

�� (��)!�

� ��

� ��

� �� *!��

�� $��

��

��

� ��

��

� ��

�

�

��*��,��

�� %%�� !�� %�� % ��

��'

�

�-��.%�/0�1��2�

�� +�� +�,�-��

��

�

14

��

��

��

��

��

��

��

��!�"��"#� ��$%��&��$�'��

��"��!��&��%��&��(��"��)*�*+�,-��-. �+!��&��

�

�� !�� !��"

��

�� #��$" %��

��

�� " &��

��

��

� ��

�� "��

��

�� "��

��

� �� '� ��(�� "

)�� *��!� +� �#�*!+$ �� &�,�� " ��

�� ")��

�� -�� " .��

��

��

�� !/��")��

��

�� #��0 ��

�� !/$ �� -��

#+!1�+/!2��+3!4��567!/$"8��

�� ,��

�� ")�� !/ � ��

��

�� "-��

��

��

��")��(��

��

�� ")��9��

� ��

�� / �� *!+ ��

#4"17 �� $")�� 0�� #��0$��

�� : 66�� "

8��" ��

��!/")��

�� !/��+!1��+!1��

�� !/� ��

�� "

8�� "�� !/��

��"�: 66��

8��/"�� !/��

��"�: 66��

8��"/ ��

��!/�� 67"��

�� !/ ��

��")��

��8��

��+!�� "

15

��

8��3"�� )��!/��

�� "�: 66�� "

) ��

�� !/ ��

��)" �� ;

�):��#)#<��$=)#<�6$$

8��" 3 ��

�� !/ ��

��"!��!/��

�� ")��

�� !/ ��

��!/��>��

�� "

8��"1��8��"5��

��

�� "��

��"-� � ��

��

� ��

�� ,��"

��"#� ��

�

� ��

��,��

��*!+��"��0�� "

!�� "&��

�� !/ ��

��" -� �� !/ ��

��

�� " )��

�� !/��

��

�*!+ ��"-��

��

�� "

8��1"+��

�� 67�!/��: 66�� "

8��5"+��

�� 67�!/��: �� "

16

Combura’12 symposium, Maastricht, The Netherlands, October 03-04, 2012

RESONANT TURBULENCE IN PREMIXED COMBUSTION

T. Cardoso de Souza1, R.J.M. Bastiaans

2, B.J. Geurts

1,2, L.P.H. de Goey

2

1Mechanical Engineering, Eindhoven University of Technology, P.O. Box 513 5600 MB Eindhoven, The Netherlands 2Laboratory of Thermal Engineering, University of Twente, P.O.Box 217, 7500 AE Enschede, The Netherlands

Key words: lean premixed combustion, resonant turbulence

Several authors have recently documented [1,2,3] that turbulent flows when subjected to a periodical perturbation might present a maximum response to the acting forcing depending on certain conditions. The flow response can be characterized by the bulk of turbulent kinetic energy, or the flow dissipation rate. Experimental results have shown [3,4] that the flow might exhibit distinct responses depending if the frequency of the forcing is higher or lower than the frequency of the large turn over eddies. For instance, when the frequency, ω, of the forcing is sufficiently small the flow will simply follow the pattern of the introduced modulation, whereas when the frequency of this perturbation is high then the flow response drops according with a 1/ω decay [1,3,4]. Between both limiting cases, a maxima response of the flow is expect to occur. This phenomenon is called resonant turbulence. The fact that the flow response might be controlled by perturbing the flow with the appropriate length scales brings the idea to use the concept of resonant turbulence to improve the effectiveness of the combustion process. The main goal of the STW project MoST is to study the application of this concept under the context of turbulent premixed combustion. The reason for such strategy relies essentially in the possibility to increase flame surface by promoting more flame front wrinkling due to the

introduction of the appropriate turbulent length scales. We started to investigate the effect of the flow perturbations by first performing DNS considering the case of a turbulent mixing layer subject to a periodical forcing. Next, the imposed perturbations were carried out under the context of a turbulent Bunsen flame. A 3D view of the computational domain considered for both cases is shown in Fig.1, together with the imposed flow profile. The dashed line denotes the slot region. In Fig.2 is shown an example of 2D contours for the perturbations imposed at the inflow plane. This vortex pattern was obtained for the case with length scales K0 = 2π/D where D is the diameter of the slot. In this case six integer modes in the x-direction are imposed at the inflow plane. Regions in blue correspond to negative vorticity, whereas in red with positive vorticity. So far, we have only considered the case of steady-perturbations. First results have shown, Fig.3, that the dynamic of the flow differs strikingly when the cold flow turbulent mixing layer is subject to these perturbations.

Figure 3 2D vorticity contours of the flow for different perturbation length scales.

To characterize the flow response to these perturbations we compute dissipation rate as shown in Fig.(4) .

Figure 1 Numerical grid considered for the DNS of the turbulent mixing layer and for the Bunsen flame.

Figure 2 An example of a flow forcing imposed at the inflow plane. The vortices introduced have a specific length scales.

17


Figure 4 Results for the spatially averaged dissipation rate for the range of wave numbers considered in the simulations. Results obtained in a XY plane at z=3D(△ symbols), and at z = 1.5D ( circle symbols).

Results in Fig.4 show that the flow presents a maximum dissipation rate when the size of the coherent structures added at the inflow plane is similar to the size of the inflow slot. On next, DNS of a turbulent Bunsen flame subject to a similar forcing was also considered prescribing the same inflow and boundary conditions as for the cold flow case. First preliminary results are show respectively in Figs. 5-6.

Figure 5 2D vorticity contours of the Bunsen flame case. In left is shown the vorticity field with no forcing, and in right with the forcing imposed at the inflow. The forcing length scale in this case corresponds to have the same diameter of the slot width.

Results in Fig.5 shows that the dynamics of the flow is significantly different compared to the cold flow case, moreover in Fig. 6 a comparison with a small range of modes indicates that the strong response of the dissipation rate for the combustion case is not evident as in the cold case.

Figure 6 Results for the spatially averaged dissipation rate for the combustion case obtained in a XY-plane at z = 1.5D.

Thus, additional quantities should also be used to characterize the response of the system in case of combustion, for instance the flame surface density or the turbulent burning velocity. These quantities can provide a direct guidance of the potential effect of the introduction of the appropriate length scales to optimize turbulent combustion. Time dependent forcing is currently being considered. REFERENCE

[1] A. K. Kuczaj, B. J. Geurts, D. Lohse, Response maxima in time-modulated turbulence: Direct numerical simulations, Europhysics Letters 73 (6) (2006) 851–857 [2] D. Lohse, Periodically kicked turbulence, Physical Review E 64 (4) (2000) 4946–4949. [3] O. Cadot, J. H. Titon, D. Bonn, Experimental observation of resonances in modulated turbulence, Journal of Fluid Mechanics 485 (2003) 161–170. [4] H. E. Cekli, C. Tipton, W. van de Water, Resonant enhancement of turbulent energy dissipation, Physical Review Letters 105 (2010) 044503.

18

Baseline n-heptane case (Spray H) simulations with the FGM method

U. Egüz*, S. Ayyapureddi, C. Bekdemir, L. M. T. Somers, L. P. H. de Goey Eindhoven University of Technology, Department of Mechanical Engineering

www.combustion.tue.nl

The operation of diesel engines combines turbulent multi-phase flow including a very high injection pressure with the non-linear combustion event. Highly accurate numerical models are paramount to comprehend the mixing and combustion processes inside the cylinder. To reduce modeling assumptions detailed chemistry information can be utilized to predict the engine characteristics like the auto-ignition delay time and emissions or to capture fuel flexibility effects in a reliable way. However, direct implementation of detailed chemical kinetics is far from being practical due to the enormous number of species and elementary reactions in chemical mechanisms. Reduction techniques are commonly applied to overcome this problem.

Spray H (baseline n-heptane) cases of Engine Combustion Network (ECN) are investigated with the Flamelet Generated Manifold (FGM) method in a constant volume combustion chamber. In the FGM method, all thermo-chemical properties are stored as a function of controlling variables, here the mixture fraction (Z) and the progress variable ( ). Two approaches are used to construct the FGM tables, igniting counterflow diffusion flamelets (ICDF) and homogeneous reactors (HR). There is however a close connection between the two ‘generators’ which can best be illustrated by what has come to known as the ‘flamelet equations’,

2

2i i

iY Yt Z

(1)

and a similar equation for the temperature. The scalar dissipation 2

2 ZDx

connects the local flow field

(straining) to the gradient in composition space. An ICDF can be described appropriately by this equation whereas it reduces to an HR in the limit of zero scalar dissipation.

The 3D-RaNS simulations are executed with the commercial CFD code StarCD. After optimizing the settings by performing a sensitivity study for the non-reacting case, the FGM method is applied as the combustion model for the reacting cases. O2 sweep is performed for the trend study of the reacting cases.

Figure 1: Ignition delay as a function of O2 concentration for experiments (blue), HR (red) and ICDF (black) approaches (FGM table resolution is 101(uniform) x 101(quadratic) for Z and , respectively).

To capture ignition delay times in a reliable way, a sensitivity study is performed to determine the FGM table resolution. It is observed in Figure 1 that using quadratic discretization in progress variable space, i.e. more points in

19

early stages of chemistry, is essential to predict accurate ignition timings. Once quadratic spacing is applied, the ignition delay time results show good correlation with those of experiments.

(a) (b) Figure 2: The lift-off-length as a function of (a) O2 concentration (b) ignition delay for experiments (blue), FGM-HR (red),

FGM-ICDF (black) methods.

According to the definition agreed upon by the ECN, the lift-off length (LOL) is defined as the point where any cell in the domain reaches YOH = 0.00025. Figure 2 displays that the trend of the LOL is also predicted well with both systems. However, the quantitative difference between the two FGM methods is higher than that observed for the ignition delay. In order to comprehend this difference better figure 2(b) is plotted. There exists a linear correlation between ignition delay and LOL for the experiments. This linear relation is clearly observed in HR based simulations as well although there is a slight under prediction for LOL and ignition delay results. In contrast, the ICDF based FGM simulations over-predict the lift-off length and expresses a slightly different sensitivity to ignition delay. Here, the main trend breaker is the [O2]:12% case. To investigate the origin of these differences between the two FGM strategies, the OH distribution in the tables is investigated. In figure 3, OH mass fraction contours as a function of temperature and Z are presented. Although the two strategies have similar ignition delay timings, species concentration predictions can be quite different, leading to diverse flame structures and LOL results.

(a) (b) Figure 3: OH mass fraction contours in the FGM tables as a function of temperature and Z for (a) HR (b) ICDF systems.

In the future work a temperature sweep will be added to the ambient O2 concentration sensitivity study. Meanwhile, the study will be extended to Spray A where n-dodecane is used as the fuel instead of n-heptane.

20

Combustion Phasing Controllability with Dual Fuel Injection Timings

C.A.J. Leermakers1,*, L.M.T. Somers1, B.H. Johansson1,2

1 Department of Mechanical Engineering, Eindhoven University of Technology, The Netherlands


2 Department of Energy Sciences, Lund University, Sweden www.energy.lth.se

Introduction

Reactivity Controlled Compression Ignition through in-cylinder blending of gasoline and diesel to a desired reactivity has previously been shown to give low emission levels, combined with an effi-ciency advantage. To determine the possible via-bility of the concept for on-road application, a de-termination of the control space of injection pa-rameters with respect to combustion phasing is presented.

The experiments have been performed on a heavy duty test engine, equipped with an intake port fuel injection (PFI) system for gasoline and a common-rail direct injection (DI) system for diesel.

Measurement matrix and procedure For all measurements the following conditions

are kept constant: 1200 rpm (σ=0.44 rpm), Fuel flow of 1.23 g/s (σ=0.022 g/s), 2 bar absolute intake pressure (σ<0.005 bar), 1.13 bar abs. exhaust pressure (σ=0.014 bar), 62wt% heavily cooled EGR (σ=0.9wt%), 306 K intake temperature (σ=2.4 K), Port injected gasoline of 80% of injected mass.

For these constant load, speed and ambient conditions, which result in a lambda value of 1.60 (σ=0.03), four diesel injection strategies are inves-tigated:

First, 20 wt% of injected mass is injected in a single diesel injection. For injecting such small amounts, a diesel injection of 1000 bar is used. This is the minimal pressure to have stable opera-tion of the injector, using a 500 microsecond (=3.6CAD) actuation duration. The start of injector actuation (SOA) is swept from -40 to -90 degrees aTDC, with 10 degree increments.

In the second and third strategies the 20 wt% diesel is equally divided over two injections. To enable stable operation of the injector, the injection pressure has to be lowered to 500 bar, to obtain a sufficiently long actuation duration. In the second strategy the late injection is fixed at -10 degrees aTDC, with an early injection variation from -40 to -90, with 10 degree increments. In the third strategy the early injection is fixed at -70 degrees aTDC, with a late injection variation from -25 to -5, with 5 degree increments.

The fourth strategy is derived from the third

one, with the early injection fixed at -70 degrees aTDC, with a late injection variation from -25 to -5, with 5 degree decrements and the early-late mass balance is shifted to 70:30.

Results: Combustion phasing control The combustion phasing response of all four in-

jection strategies is combined in Figure 1. As dis-cussed above, the single and early injections have an inverse effect on combustion phasing. A first order fit of the measured points gives a quantifica-tion of this negative slope, defined as

50ID

CAsSOA

.

Furthermore the linear association of the re-sponse is not very strong and in a double injection strategy, the response of CA50 on a variation of the first injection is (very) weak.

Figure 1 – Timing of CA50, for injection timing variations

in 4 different injection strategies. Marker and vertical errors depict the mean and standard deviation, respec-tively, of 50 measured cycles per operating point. The given slope and coefficient of determination (R2) are

based on a linear regression fit.

The response to a variation in the late injection has a positive IDs , which is nearly exactly linear and has a larger value compared to the early injec-tion variation strategies. Furthermore, above it was shown that the more fuel is admitted in the second injection, the larger IDs is. Therefore the third strategy was found to be most favorable with re-spect to combustion phasing response.

* Corresponding author: [email protected] COMBURA 2011, the ReeHorst, Ede

21

Results: Efficiency For the double injections, the injection pressure

had to be lowered to 500 bar, and together with the very short injections, this appears to be dramatic for the completeness of combustion (combustion efficiency). As can be expected, this has a signifi-cant effect on the gross indicated fuel efficiency.

All double injection strategies have an indicated efficiency of about 10 percent lower than the single injection strategy. Other tests have shown that the dual-fuel concept, even with double injections, is possible of producing very high efficiencies, and thus low fuel consumption. However, because of the low completeness of combustion, this is not achieved in the present investigation.

Results: NOx and smoke emissions It is general practice to plot smoke emissions

versus nitrogen oxides emissions to see how dif-ferent strategies behave with respect to the com-mon NOx-soot trade off. From Figure 2 it can be seen that the present injection strategies largely escape from this trade-off, with both smoke and NOx emissions being near zero.

Figure 2 – Smoke emissions vs. nitrogen oxides for

injection timing variations in 4 different injection strate-gies. Euro VI emission levels depicted by purple box.

Therefore, the chosen injection strategy does not have a big impact. For smoke, also the com-bustion phasing has a minor effect, whereas for nitrogen oxides the emission levels increase with advancing combustion, but remain reasonably low.

Results: Maximum pressure rise rate Largely premixed combustion can lead to unac-

ceptably high pressure rise rates. From Figure 3, it shows that the maximum pressure rise rate is largely independent from the chosen injection strategy, but mainly depends on the resulting com-bustion phasing.

For all strategies the pressure rise rates are ef-ficiently suppressed by the high dilution rates used. Therefore, it is desired to have an injection strate-gy that offers a wide range of control. With such an injection strategy, combustion phasing can be

shifted such that the maximum pressure rise rate always stays below acceptable levels.

Figure 3 – Maximum pressure rise rate vs. CA50 for

injection timing variations in 4 different injection strate-gies.

Conclusions

A variation in the timing of the first or single diesel injection has an opposite effect on com-bustion phasing. The response is reasonably linear, but the sensitivity of the first injection is weak.

The sensitivity of the late injections is positive and larger in absolute value compared to the early injections variation strategy. Furthermore the sensitivity correlates with the amount inject-ed in the second injection. As such the third strategy is most favorable.

All three double injection strategies give very poor combustion efficiency. For these double injections, injection pressure had to be lowered to 500 bar, and together with the very short in-jection this results in a low combustion efficien-cy.

Because of the high dilution level and largely premixed mixture, all present injection strate-gies break with the common soot-NOx trade-off, with both smoke and NOx emissions being near or below upcoming legislated levels.

For all strategies, the pressure rise rates are efficiently suppressed by the high dilution rates used, and mainly depends on combustion phas-ing. Therefore it is desired to have an injection strategy that offers a wide range of control.

Acknowledgments This project was funded by STW project 10417. The authors kindly acknowledge Bas van den Berge for performing the measurements.

Reference C.A.J. Leermakers, L.M.T. Somers and B.H. Johansson, “Combustion Phasing Controllability with Dual Fuel Injec-tion Timings”, SAE Technical Paper 2012-01-1575, (2012).

22


Numerical Simulations of Premixed Turbulent Combustion Using the Flamelet Generated Manifold Approach With Heat Loss Inclusion

A. Donini, R.J.M. Bastiaans, J.A. van Oijen and L.P.H. de Goey

Combustion Technology, Eindhoven University of Technology, Postal Address P.O. Box 513 5600 MB Eindhoven

Key words: Gas turbine combustor, fuel flexibility, fuel composition, NOx reduction, turbulent combustion, Computational Fluid Dynamics (CFD), Large-Eddy simulation (LES), Reynolds Averaged Navier-Stokes (RANS), chemical reduction techniques, Flamelet-Generated Manifolds (FGM), combined combustion modes, combustor design.

Gas turbine combustion is the most important energy conversion method in the world today. Using gas turbines, large scale, low emission energy production is possible. Nitrogen oxide emissions are one of the most important technology drivers for combustion systems today. For land based engines, low NOx emissions can be achieved by very lean premixed combustion (dry low NOx). For most aero applications, gas turbines are the only option to achieve the required thrust. Conventional combustion systems are based on diffusion mode combustion, but to be able to meet emission requirements, the next generation of aero gas turbines will be based on lean premixed combustion technology.

In the field of stationary land based engines, Siemens Power Generation (SPG) is very active and would like to increase their turbulent combustion knowledge to predict engine performance. This is mainly in connection with techniques to include high levels of hydrogen in the fuel.

Currently at Siemens Power Generation, SPG, Reynolds Averaged Navier-Stokes (RANS) based CFD is used with simple combustion models. One of the aims of the current project is to enhance the CFD capabilities to support the design of future combustors. It is in the interest of SPG to gain the knowledge to be able to use CFD as a tool to optimize designs as well as to obtain fundamental knowledge to arrive at possible solutions for performance and emission issues. Here different types of RANS and LES (Large-eddy simulation) techniques become an important item. The goal is to develop models that are capable of representing these effects at high temperatures and pressures. For SPG, methods are developed for commercial CFD codes, e.g. CFX (RANS and LES versions).

In the current project detailed knowledge for modeling of combustion with alternative fuels will be developed. This is done by means of detailed descriptions in the framework of Computational Fluid Dynamics (CFD). The ultimate goal is to predict the

combustion process of gas turbines, including complex physical real fuel phenomena (temperature-traverse, NOx, preferential diffusion, thermo-diffusive effects, soot, ignition, extinction etc.). This requires a method in which different combustion modes can be captured, different fuels can be covered and extinction, ignition, heat loss and slow chemistry effects can be included. To that end the promising technique of flamelet-generated manifolds (FGM) will be extended in this project. The technique is developed in its basic form at TU/e and has been continuously tested and extended to more general situations over the last years.

In the present paper a computational analysis of a confined premixed turbulent methane/air and hydrogen/air jet flames is presented. In this scope, chemistry is reduced by the use of the Flamelet Generated Manifold (FGM) method, and the fluid flow is modeled in a RANS and LES context. A generic lab scale burner for high-velocity preheated jets (as schematically shown in Figure 1) is used for validation. It consists of a rectangular confinement, and an off-centre positioning of the jet nozzle enable flame stabilization by recirculation of hot combustion products.

Figure 1: The burner and the combustion

chamber.

23


Flame structures were visualized by OH* chemiluminescence imaging and planar laser-induced fluorescence of the OH radical. Laser Raman scattering was used to determine concentrations of the major species and the temperature. Velocity fields were measured with particle image velocimetry (Figure 2).

Figure 2: Averaged velocity profile of the central

section. FGM is a chemistry reduction method that

combines the advantages of chemistry reduction and flamelet models. The approach is based on the idea that the most important aspects of the internal structure of the flame fronts should be taken into account. In the FGM technique the progress of the flame is generally described by a few control variables, for which a transport equation is solved during run-time. The flamelet system is computed in a pre-processing stage, and a manifold with all the information about combustion is stored in a tabulated form. During run-time only equations for the control variables are solved, using the database to retrieve all necessary information to update the solution. In the present implementation the reaction evolution is described by the reaction progress variable, the heat loss is described by the enthalpy and the turbulence effect on the reaction is represented by the progress variable variance. The turbulence-chemistry interaction is considered through the use of a presumed pdf approach, a simple but non-trivial model that is often used for chemistry modeling. This research attempts to apply the FGM chemistry reduction method coupled with RANS and LES models, in order to predict the evolution and description of a turbulent jet flame in high Reynolds number flow conditions, including the important

effect of heat loss to the walls. Comparison of various mean fields (velocities, temperatures and compositions) with RANS and averaged LES results are shown. The use of FGM as combustion model shows that combustion features in gas turbine conditions can be reproduced with a reasonable computational effort. Additionally, LES-FGM shows accurate description of the combustion process in complex combustion systems, including pollutants behavior.

The current project is heavily supported by Siemens Power Generation, SPG, and Rolls-Royce Deutschland, RRD. The vision of these companies is that a joint project performed in collaboration with the Combustion Technology group of Eindhoven University of Technology, can create added value to their individual competitiveness in CFD based gas turbine design. It has to be mentioned that this is a unique development dictated by the challenging changes with which gas turbine developers are confronted in the near future. The benefit in a competitive situation of a joint project becomes significant with the existence of common fundamental problems, which can be tackled within a unified framework. The methods developed by the Combustion Technology group of TU/e will provide such an approach.

ACKNOWLEDGEMENT The authors would like to thank STW, Siemens Power Generation (SPG), and Rolls-Royce Deutschland (RRD) for sponsoring this project.

24


Co-combustion characteristics and kinetics of torrefied beech wood with bituminous coal

Eyerusalem M.Gucho1, Eddy A. Bramer, Gerrit Brem

University of Twente, Department of Thermal Engineering, Drienerlolaan 5, 7522 NB Enschede, The Netherland

Because of fossil fuels depletion and global warming, many researchers has turned their attention on the study of renewable energy fuels like biomass. Biomass is a clean fuel with low sulphur and nitrogen content and CO2 neutral (as the carbon dioxide emitted to the environment during the combustion process for electricity generation is captured during photosynthesis). Nevertheless, biomass offers many advantages regarding its environment benefits, there are some drawbacks of the biomass fuels that hurdles its expansion in the energy utilization market. These drawbacks of biomass fuel are: less energy density (high cost during transportation), fibrous (expensive grinding cost), hydrophilic (difficult to store it outside). Thermal pre-treatment of the biomass is one way to reduce these disadvantages of the biomass fuel.

Torrefaction is a thermal pre-treatment of biomass at a temperature between 220-300 oC under inert atmosphere. Torrefaction improves some properties of the biomass like the energy density, grindability property and hydrophobicity. This thermal pre-treatment makes biomass to be easily utilized in different application, co-firing with coal, gasification, for household heating and so on. The main application area for torrefied biomass is in co-combustion with pulverized coal power plants, since the properties of the biomass enhanced during torrefaction are mainly related to this specific application. Therefore, study on the influence of torrefaction to the combustion characteristics and kinetics are important during co-/combustion with coal.

The scope of the present work was to investigate the influence of torrefaction on the combustion characteristics and the kinetics of torrefied beech wood and bituminous coal in non-isothermal thermogravimetric method (TGA) under air atmosphere. Simultaneously, for co-combustion study, among the torrefied beech samples, two samples (lightly and severely torrefied beech wood) were selected so that to get insight on the extreme torrefaction conditions . These torrefied beech wood were then blended with coal in 25, 50 and 75 wt% to study the co-combustion behaviour and its kinetics.

The experimental result showed that light torrefied fuels has shown increase in combustion reactivity while severe torrefied fuel show a steadily decrement. The torrefied biomass combustion takes place in general two stage, between 210-400 oC the volatile are released and between 390- 540 oC the rest char gets combusted. On contrary, coal combustion takes place in one stage between 300-640 oC and the coal/torrefied wood blend shown up to three stages. Reaction kinetics were then calculated by the Coats–Redfern method to study the responsible mechanism for oxidation of the samples. The kinetics were analysed using single order reaction for both volatile and char combustion stages. Range of activation energy in volatile combustion phase were 120-80 KJ/mol-1 and char combustion phase was between 17-92 KJ/mol-1 as the torrefaction gets from light to severe conditions. In the case of coal/torrefied biomass combustion, no significant synergetic effect between coal/ light torrefied beech wood blend were realised, however for coal/severe torrefied biomass blend clear interaction was detected for blending ration up to 25 wt% of torrefied beech wood.. *Corresponding author. Address: University of Twente, Drienerlolaan 5, 7522 NB Enschede , The Netherlands. Tel: +31 53 489 3564; ; Fax: +31 53 489 3663 E-mail address [email protected]

25

mailto:[email protected]


26

ULRICO: ULTRA RICH COMBUSTION OF HYDROCARBONS AND SOOT FORMATION

NUMERICAL MODELING USING PDF METHODS

M.Stöllinger1, D.J.E.M. Roekaerts1, M.H.F.Woolderink2 and J.B.W.Kok2

1Department of Process and Energy, Delft University of Technology, Mekelweg, 2, 2628 CD Delft, The Netherlands

[email protected]

2Thermal Engineering, University Twente, P.O. Box 217, 7500 AE Enschede, The Netherlands

Key words: Ultra rich combustion, natural gas, syngas, soot, flame sampling, numerical modeling

Introduction Ultra rich combustion (partial oxidation) of natural gas is a process applied to produce synthesis gas (or syngas). This gas is composed primarily of hydrogen and carbon monoxide. Syngas represents the intermediary step from hydrocarbons to bulk chemicals and synthetic fuels. The reactor design and the operating conditions have to ensure a high conversion of natural gas to hydrogen and carbon monoxide. In addition to this, the hydrogen to carbon monoxide ratio in the syngas is relevant for the downstream application of the syngas produced. The syngas soot content is also of concern, in view of the syngas fouling the reactor system and to minimize the downstream effort of soot removal. The operating conditions for the large scale application of the partial oxidation process are characterized by turbulent flow and a high fuel to oxidizer ratio (“rich” combustion), far beyond the stoichiometric ratio. The goal of this research project is twofold: development of computational models that can be used as a design tool for the partial oxidation reactor and the development and use of a reliable measurement system to quantify relevant soot properties such as the size distribution. This poster reports on model development. Model description The non-premixed process is studied by means of a transported probability density function (PDF) method. The gaseous chemistry is described by a non-adiabatic Flamelet Generated Manifold (FGM) method. The FGM is based on steady and unsteady non-premixed laminar flamelets which are obtained for different values of strain rates and enthalpy loss using a detailed chemical mechanism. The FGM is parametrized by the mixture fraction, a reaction progress variable and the enthalpy loss. The nucleation, surface growth and oxidation of soot particles depend non-linearly on the gas phase temperature and composition. Moreover, these processes occur on larger time scales than the gaseous chemical reactions. This implies that the soot number density and the soot mass concentration cannot be directly related to the mixture fraction. Instead, additional transport equations for the soot variables have to be solved. To treat the effects of the turbulent fluctuations of the gas phase temperature and composition on the soot process in closed form, an equation for the joint PDF of mixture fraction, enthalpy, soot number density and soot mass concentration is solved.

27

The radiative heat transfer is significant in sooting processes. To account for the radiative heat transfer, the Reynolds averaged radiative transfer equation (RTE) is solved by means of a discrete transfer method. The adopted PDF approach allows treating the emission turbulence-radiation interaction in closed form. Results The PDF method has been used in simulations of two laboratory-scale non-premixed methane-air flames respectively at one and three bar pressure. Figure 1 shows a comparison of the mean soot volume fraction results obtained with an FGM and a strain rate based flamelet chemistry model in the high-pressure flame. In the flamelet model, the effect of enthalpy loss on the chemical composition has been neglected. The concentration of the soot precursor species acetylene is very sensitive to heat loss and is significantly over predicted in the flamelet model. This explains the over prediction of the soot volume fraction in figure 1.

The sensitivity of the soot volume fraction predictions on the radiative heat transfer is strong in particular in the three bar flame where the soot concentration is about 10 times larger than in the atmospheric flame. Moreover, the correlation between the mixture fraction and soot volume fraction fluctuations is presented. This correlation has to be modeled in presumed shape PDF methods and the commonly adopted models will be analyzed. Conclusion The transported PDF study of soot formation in non-premixed combustion provides valuable insight into the tight coupling between fluctuating gas phase temperature and composition and the soot formation.

The analysis of the PDF results is used to verify and improve modeling assumptions, used in simpler turbulence-chemistry models. The developed numerical models are sufficiently general to serve as a valuable design tool for a great variety of partial oxidation reactors. ACKNOWLEDGMENT This project is supported by Technology Foundation STW and Shell.

Figure 1: Comparison of mean soot volume fraction results obtained with FGM and flamelet reduced chemistry in the p=3bar flame.

28

ULRICO: ULTRA RICH COMBUSTION OF HYDROCARBONS AND SOOT FORMATION

DESIGN OF A FLAMELESS OXIDATION BURNER

M.H.F.Woolderink1, J.B.W.Kok1, M.B. Holtkamp1, M.Stöllinger2 and D.J.E.M. Roekaerts2

1Laboratory of Thermal Engineering, University of Twente, P.O. Box 217, 7500 AE Enschede, The Netherlands

2Department of Process and Energy, Delft University of Technology,

Mekelweg, 2, 2628 CD Delft, The Netherlands [email protected]

Key words: Ultra rich combustion, natural gas, syngas, soot, numerical modeling, flameless oxidation

Introduction Ultra rich combustion (partial oxidation) of natural gas is a process applied to produce synthesis gas (or syngas). This gas is composed primarily of hydrogen and carbon monoxide. Syngas represents the intermediary step from hydrocarbons to bulk chemicals and synthetic fuels. The reactor design and the operating conditions have to ensure a high conversion of natural gas to hydrogen and carbon monoxide. In addition to this, the hydrogen to carbon monoxide ratio in the syngas is relevant for the downstream application of the syngas produced. The syngas soot content is also of concern, in view of the syngas fouling the reactor system and to minimize the downstream effort of soot removal. The operating conditions for the large scale application of the partial oxidation process are characterized by turbulent flow and a high fuel to oxidizer ratio (“rich” combustion), far beyond the stoichiometric ratio. The goal of this research project is twofold: development of computational models that can be used as a design tool for the partial oxidation reactor and the development and use of a reliable measurement system to quantify relevant soot properties such as the size distribution. This poster reports on the design of a flameless oxidation burner. Burner Design The majority of soot particles are formed in the flame zone. It is therefore very important to minimize the soot leaving the flame zone in the burner design. In flameless oxidation combustors there is no apparent flame zone visible. The combustion is characterized by smooth temperature and species concentrations gradients. This is achieved by recirculating product gases which dilute and heat up the reactants. This gives no problems in lean combustion, because the product gases are inert. In rich combustion however, the product gas, synthesis gas, is highly reactive. It is therefore a challenge to dilute the fresh reactants with synthesis gas without combusting it. Taking above in account, a burner was designed with help of flow and combustion simulations in ANSYS CFX. In this burner the oxidizer and fuel enter the combustion chamber, mix with product gases and react in a specific way such that the synthesis gas does not react with the oxidizer to water and carbon dioxide. Results ANSYS CFX with the BVM combustion model was used for simulating the combustion in the flameless oxidation burner described in the previous section. In table 1 the mass fractions for a specific set of operating conditions on the outlet and averaged in the domain of the premix and flameless oxidation

29

combustors are shown. It can be seen that the hydrogen and carbon monoxide mass fractions on the outlet of the flameless oxidation burner are significantly higher than those of the premixed burner, while the average acetylene mass fraction in the domain is drastically lower. Acetylene is the most important soot precursor. Y_H2_outlet Y_CO_outlet Y_C2H2_average Premixed combustion 0.024 0.193 0.049 Flameless oxidation 0.028 0.230 0.011 Difference +17% +19% -77.5%

Table 1: Hydrogen and carbon monoxide mass fractions on the outlet and the average acetylene concentration in the domain of the flameless oxidation and premixed combustors. Conclusion A flameless oxidation burner for the partial oxidation of natural gas to synthesis gas was designed. Simulations with this burner show that in comparison with the premixed burner, the syngas output increases with almost 20% and the acetylene concentration decreases with almost 80%. Acetylene is the most important soot precursor, therefore a decrease in soot formation is also expected. ACKNOWLEDGMENT This project is supported by Technology Foundation STW and Shell.

30

Abstracts of presentations

31

Evaluation of RANS-based Combustion Modeling of Gaseous Flames

J. E. Anker, N. Beishuizen, K. Claramunt, Ch. Hirsch*

NUMECA Int., Chaussée de la Hulpe/Terhulpsesteenweg 189, BE-1170 Brussels, Belgium

www.numeca.com *[email protected]

Introduction The Flamelet Generated Manifolds method (FGM), the hybrid BML-flamelet/turbulent flame speed closure (TFC) approach, and the eddy dissipation model (EDM) have been implemented into the unstructured simulation system FINETM/Open. Although LES simulations for combustion are gaining larger industrial acceptance, design considerations still require rapid turnovers and RANS based combustion models are an adequate tool, provided they are sufficiently validated. The interrelations and differences of the mentioned models, as well as their conceptual strengths and weaknesses are first discussed from a theoretical point of view. Then results from the simulation of various combustion test cases ranging from elementary flames to industrial configurations like furnaces and combustors are used to demonstrate the practical advantages and shortcomings of the models for gaseous combustion processes. Modeling Methods In the FGM and in hybrid BML-flamelet/TFC approaches a transport equation for the progress variable, the mixture fraction, and the mixture fraction variance are solved, which makes these models computationally efficient. The dependent thermochemical properties as well as the reaction source term are retrieved from a manifold spanned by the mixture fraction and one (or several) progress variables in the FGM approach. This manifold is generated by remapping a library of premixed or non-premixed flamelets, which have been generated by the use of a 1D chemistry code. Contrary to this, in the TFC model the reaction source term for the progress variable equation is modeled as a function of the laminar flame speed and the turbulent time scale. In the compressible EDM approach a transport equation is solved for each species, which is present in the modeled chemical system. The thermochemical properties and the source terms are computed in the CFD solver in dependence of the concentrations, temperature and the turbulent state. Results To assess the performance of the implemented combustion models, a variety of elementary test cases have been carried out. A comparison of the computational results for well-established fundamental test cases with experimental data confirms the validity of the models for gaseous flames. Among other cases, TNF’s Flame D, Sidney/Sandia’s Bluff-body stabilized flame, and TU Darmstadt’s stratified flames have been simulated. Fig. 1 shows the simulated temperature contour of Flame D. To test the performance of the implemented models on complex test cases, the reactive flow fields of the industrial furnace of Sandia’s Burner Engineering Research Laboratory (BERL) [3], of DLR Stuttgart’s premixed combustor [4], and of the generic gas turbine (GGT) combustor [5] of EKT/TU Darmstadt have been simulated. Figure 2 shows the simulated temperature field in the GGT combustor and in Fig. 3 the flame front in the DLR combustor is visualized. The computational results are compared to measurement data. The differences and similarities of the models are discussed and conclusions are drawn with regard to the performance and reliability of the various models in the different combustion regimes.

Fig. 1: Computed temperature contours in Flame D (EDM approach)

Fig. 2: Computed temperature contours in the GGT combustor of TU Darmstadt (FGM technique)

Fig. 3: Computed flame front (c = 0.8) inDLR Stuttgart’s premixed combustor (TFC model)

32

Acknowledgement The financial support received from the European Union in the Marie-Curie COMBINA IAPP project is gratefully acknowledged.

References [1] Van Oijen, J. A. (2002): Flamelet-Generated Manifolds: Development and Application to Premixed Laminar Flames, PhD Thesis, TU Eindhoven, 2002 [2] Van Oijen, J. A.; de Goey, L. P. H. (2000): Modelling of Premixed Laminar Flames using Flamelet-Generated Manifolds, Combustion Science and Technology, Vol. 161(1), pp. 113-137 [3] Kaufman, K. C.; Fiveland, W. A.; Peters, A.A.F; Weber, R. (1994): The BERL 300 kW Unstaged Natural Gas Flames with a Swirl-stabilized Burner, Babcock & Wilcox’s Research and Development Division, Alliance, OH, USA [4] Meier, W.; Weigand, P.; Duan, X. R.; Giezendanner-Thoben, R. (2007): Combust. Flame, Vol 150, pp. 2-26 [5] Janus, B.; Dreizler, A.; Janicka, J. (2004): ASME Paper GT2004-53340

33

Application of Partially Premixed Combustion using low octane fuels in a heavy duty engine

P.C. Bakker∗, C.A.J. Leermakers and B.H. Johansson

Department of Mechanical Engineering, Eindhoven University of Technology, The Netherlandswww.combustion.tue.nl

IntroductionStringent emission legislations are forcing automo-

tive industry to come up with innovative solutions to re-duce harmful emissions. Kalghatgi et al. proposed thatgasoline like fuels might reduce emissions, in particu-lar soot, by means of elongating the available time formixing[1]. They achieved low emissions and high effi-ciency simultaneously. Partially Premixed Combustion(PPC) has shown the potential of high efficiency, emis-sions of both nitrogen oxides (NOx) and soot below fu-ture emissions regulations and acceptable acoustic noise.Low-octane-number gasoline fuels were shown to be mostsuitable for this concept, with the reactivity determiningthe possible load range. At best this load ranges fromidle to full load, without major modifications to the en-gine setup. Although PPC has shown its potential, severalquestions have still not been solved.

Experimental ApparatusThe experiments are conducted on a dedicated test rig

which is based on a DAF XE 355c engine (see Table 1). Itis a six cylinder, 12.6 liter engine of which only cylinder1 will be used for combustion concept testing. Cylinders2 and 3 are the so-called pumping cylinders and those areused for supplying EGR. Cylinders 4, 5 and 6 operate un-der the stock ECU and together with a Schenck W450 dy-namometer they are used for controlling the engine speed.Figure 1 is a schematic overview of this setup.

Table 1: Engine specifications

Bore [mm] 130Stroke [mm] 158Compression ratio [-] 15.7

Figure 1: Schematic overview of the test engine

Fuel CharacteristicsLow octane fuels are known for their extended igni-

tion delay with respect to EN590. Two fossil fuel based,naphtha like fuels are selected and their composition isdetermined using gas chromatography - mass spectrome-try. These naphtha fuels (respectively abbreviated as NB1and NB2) could be produced in refinery with minor ad-justments.

On top of that, a promising biofuel (n-butanol or BuOH)is investigated in various blending ratios with EN590. Ta-ble 2 shows some specifications of the test fuels.

Table 2: Fuel specifications

Properties NB 1 NB 2 BuOHBoiling range [◦C] 98 - 140 142 - 200 117Aromatics <0.5% ~15% 0%LHV [MJ/kg] 43.48 43.31 33.02

Both the naphtha blends suffer from insufficient lu-bricity and therefore, a lubricity additive(Infineum) is addedto those fuels.

Measurement outlineThe authors decided to put emphasis on mid range

loads, i.e. 8 - 16 bar gross IMEP using 50% of EGR.Influence of combustion phasing, injection strategy andfuel pressure is of particular interest for characterizing thePPC potential of the fuels. Moreover, dilution effects arealso investigated. Finally, load sweeps will be conducted.

Results and DiscussionSensitivity of ignition delay

Pure butanol had lack of control and ignitability issuesat 8 bar IMEP. Therefore, the diesel content is increasedgradually to gain control over the start of combustion.Figure 2 depicts the observed effects in terms of the sensi-tivity of ignition delay, a term introduced by Leermakerset al.[2]

Fuel pressure effectsLowering the fuel pressure traditionally results in increasedsoot emissions. Sooting tendency can be evaluated bykeeping all parameters constant except for the fuel pres-sure (see Figure 3).

∗Corresponding author: [email protected] 2012, Kasteel Vaeshartelt, Maastricht

34

−15 −10 −5 00

2

4

6

8

10

12

SOI [°CA]

CA

50 [

°CA

]

Sensitivity of ignition delay at 8 bar IMEP

BuOH80BuOH70BuOH60BuOH50S

ID = 1.076

SID

= 1.16

SID

= 1.308

SID

= 1.425

Figure 2: Increasing the diesel content yields a more 1:1correlation between CA50 and SOI

500 600 700 800 900 1000 1100 1200 1300 1400 15000

0.2

0.4

0.6

0.8

1

pfuel

[bar]

ISP

M [

g/k

Wh

]

Fuel pressure sensitivity at 8 bar IMEP and CA50 = 4 °CA

EN590NB2NB1BuOH60

Figure 3: Differences in sooting tendency

Heat releaseFigure 4 reveals that the extended ignition delay of thebutanol blend results in short burn durations. NB1’s pre-mixed peak is slightly larger than with NB2 which couldbe caused by NB1’s lower boiling range (i.e. more in PPCregime).

−10 −5 0 5 10 15 20 25

0

200

400

600

800

1000

CA [°CA]

Hea

t rel

ease

[J/°

CA

]

Heat release at 8 bar IMEP and CA50 = 6 °CA

NB2NB1BuOH60EN590

Figure 4: Premixedness increases from EN590 to NB2,NB1 and finally BuOH60

Single injection load sweepThe capabilities of the fuels at 8 bar IMEP have beenproven, but additional measurements at increased loadsshould be performed to gain insight in the applicabilityof those fuels for PPC purposes. Figure 5 depicts the de-creasing premixedness when load increases for the 60%butanol blend.

−10 −5 0 5 10 15 20 25

0

200

400

600

800

1000

CA [°CA]

Hea

t rel

ease

[J/°

CA

]

Heat release during load sweep for BuOH60 at CA50 = 8 °CA

8 bar10 bar12 bar14 bar16 bar

increasing load

Figure 5: Heat release switches from premixed to classi-cal diesel combustion when load increases

ConclusionsBoth the naphtha blends have shown a similar reactivityto EN590. Nevertheless, the higher volatility might en-hance the mixing with air. Additional experiments shouldreveal whether naphthas are really beneficial over tradi-tional diesel, but reductions in soot have already been ob-tained. Butanol on the other hand is more promising dueto its enormous soot reducing potential and renewability.Reactivity might be an issue, but a smart inlet air heateror smart blending with diesel could cope with this.

AcknowledgementsThis project is supported by the Dutch Technology

Foundation STW project 10417, DAF Trucks N.V., Shell,Delphi and Avantium. The authors appreciate the supportof the technicians of the Eindhoven Combustion Technol-ogy group: Gerard van Hout, Bart van Pinxten, Hans vanGriensven and Theo de Groot.

References[1] SAE Technical Paper 2006-01-3385.

[2] SAE Technical Paper 2012-01-1575.

35

* Corresponding author: [email protected] Combura 2012, Kasteel Vaeshartelt, Maastricht

Development of optical diagnostics on Partially Premixed Combustion in a heavy-duty diesel engine

M.M.H. Corvers1,*, C.A.J. Leermakers1, N.J. Dam1, B.A. Albrecht2, L.P.H. de Goey1, B.H. Johansson1,3

1 Department of Mechanical Engineering, Eindhoven University of Technology, the Netherlands

www.combustion.tue.nl 2 DAF Trucks N.V., Eindhoven, the Netherlands

www.daf.nl 3 Department of Energy Sciences, Lund University, Sweden

www.ce.energy.lth.se

Introduction

In the battle against NOx and soot emissions, highly-diluted, partially premixed combustion has been proposed to get rid of the NOx-soot trade-off experienced in conventional diesel combustion. Fuel stratification has a large influ-ence on the combustion properties, it should be low enough to initiate combustion at multiple locations at once and suppress local tempera-tures. On the other hand, fuel should be stratified enough to not have it all ignite at once. Engine experiments have shown that simultaneous reduction of both NOx and soot emissions is pos-sible [1]. The origin of this effect can be studied further by using an optical engine. The CH radical is known to be formed in the early stages of hydrocarbon combustion, there-fore it is a good indicator of the time and loca-tion of the start of combustion and therewith the fuel stratification. High-speed laser-induced fluorescence can be used to visualize the crank-angle resolved spatial distribution of the CH radical. In this project, the visualization method will first be validated using a Bunsen type flame. If these validation tests have a positive result, the experiments will be extended to an optical engine. Laser-induced fluorescence of CH

Planar Laser-induced fluorescence (PLIF) is based on the theory that molecules and/or at-oms can be excited to a higher energy level by aiming a laser beam with a certain wavelength at them. The excited molecules and/or atoms will naturally fall back to their ground energy level by emitting the excess energy in the form of light. The wavelength of the emitted light is rep-resentative of the type of molecule and/or atom. CH radicals can be excited using a laser wave-length of 387 nm (B2Σ-X2Π (0,0) band). The emitted light from the CH radicals has a wave-length around 431 nm.

Experimental Setup For the excitation of the CH radicals, a high-

speed dye laser is pumped by a dual cavity high-speed Nd:YAG laser. This system is able to pro-duce laser pulses with a wavelength of 387 nm with an energy of 90 µJ/pulse at 6000 Hz. The beam of laser pulses is converted to a sheet of approximately 2 cm high by using lens optics. A Bunsen type flame is used as a source of CH rad-icals. The CH-PLIF emission at 431 nm is cap-tured using a high-speed camera equipped with a high-speed image intensifier (IRO). In front of the lens (f/2.8) a band-pass filter of 427±5 nm is placed to block flame luminosity and scattered laser light. The experimental setup used during the validation process is shown in Figure 1.

Figure 1: Experimental setup of validation process

The Bunsen burner is fueled by a propane tank and various air/fuel compositions can be inves-tigated. Results

The equipment used proved to be sufficient for visualizing CH-PLIF emissions in the Bunsen type flame. Figure 2 shows an example of an acquired image. For better visualization, post processing of the images is needed.

36

The left image in Figure 2 shows the original image as taken by the high-speed camera, after applying a color map. The next step in improving the image quality is to reduce noise. This is done using a 2D Wiener filter, which calculates the local signal density to maintain and is able to maintain the signal contours. In the last step is the image is smoothened by using a median fil-ter.

Figure 2: Improving the image quality

As a result, the CH-PLIF emission is clearly visi-ble as a cone in the final image. This cone is the inner flame front of a Bunsen burner flame, also shown in Figure 3 as the light blue part.

Figure 3: The Bunsen type flame with a pilot

The CH radical concentration in the outer flame front (the dark blue part in Figure 3) is not suffi-cient to produce a CH-PLIF emission signal that is visible to the used detection system. The CH-PLIF emission from Bunsen type flame is also used to create a CH-excitation spectrum by varying the laser wavelength from 385 nm to 393 nm. The obtained spectrum is compared to a simulated CH excitation spectrum made with LIFBASE software [2]. This comparison proved that the captured signal is indeed CH-PLIF emis-sion and also that the strongest signal is ac-quired when using an excitation wavelength of 387.204 nm. Conclusion

The conclusions that can be drawn after the validation process are that producing and cap-turing CH-PLIF is possible with the used equip-ment. The next step is to extend these results to experiments on an optical engine. This step is briefly discussed next.

Work in progress The optical engine that will be used is a one-

cylinder engine with a DAF MX heavy-duty diesel engine cylinder head, driven by an electric mo-tor. The fuel used is n-heptane. Optical access from the side is acquired by placing windows in an elongated cylinder wall as shown in Figure 4.

Figure 4: Schematic overview of the optical access in the heavy-duty diesel engine.

The laser beam as shown in figure 1 is rerouted to the optical engine and converted to a laser sheet of approximately 2 cm high before enter-ing the engine through one of the windows per-pendicular to the camera view. This is shown in Figure 5.

Figure 5: Experimental setup of engine experiments

Acknowledgements

This project is funded by the Dutch Technol-ogy Foundation STW (Project 10417). DAF Trucks N.V., Shell Global Solutions, Avantium Chemicals B.V. and Delphi are also acknowl-edged for their contributions to the project. The authors kindly appreciate the support of the technicians of the Eindhoven Combustion Tech-nology group: Bart van Pinxten, Hans van Griensven, Theo de Groot, Paul Bloemen and Gerard van Hout.

References [1] C.A.J. Leermakers, C.C.M. Luijten, L.M.T. Somers, G.T.

Kalghatgi, B.A. Albrecht. “Experimental Study of Fuel Composition Impact on PCCI Combustion in a Heavy-Duty Diesel Engine”. SAE Technical paper 2011-01-1351, 2011.

[2] J. Luque and D.R. Crosley. “Lifbase: Database and spectral simulation (version 2.0.64)”, 1999.

37

"Autoignition properties of gaseous fuels at conditions relevant to engines and turbines" S. Gersena and H.B. Levinskya,b

a DNV KEMA Energy & Sustainability ., P.O. Box 2029, 9704 CA Groningen, The Netherlands bLaboratory for High Temperature Energy Conversion Processes, University of Groningen,

Nijenborgh 4, 9747 AG Groningen, The Netherlands The drive towards sustainability, increasing fuel prices and the depletion of local fuel reserves have resulted in an increasing trend in the diversification of fuels for combustion equipment. Regarding operational aspects, changes in the autoignition delay time of the fuel can result in engine knock in internal combustion engines or pre-ignition in lean-burn premixed gas turbines. Both phenomena should be avoided since they can physically damage the equipment. Clearly, information about the autoignition properties of fuels is essential to guarantee safe and efficient operation of current and future combustion equipment. Moreover, understanding the autoignition behavior of fuels is an integral aspect of the development and benchmarking of chemical oxidation mechanisms of fuels. Central to the analysis of ignition behavior is data from well-defined experiments. Autoignition delay times of stoichiometric ( =1) and fuel-lean ( =0.5) CH4, nC4H10, iC4H10, and H2 as well as pure fuels, binary mixtures of CH4 with n-C4H10, i-C4H10, nC5H12, iC5H12, neo-C5H12, H2, CO, NO2 and tertiary mixtures with H2 and CO measured in a Rapid Compression Machine (RCM) at temperatures ranging from 660-1150K and pressures up to P 80 bar will be presented. Pure n-butane and iso-butane exhibit a negative temperature coefficient (NTC) region (figure 1) and at low temperatures two-stage ignition is observed for both fuels. Experiments show that the addition of small fractions butane and pentane (1-3%) to methane results in a substantial reduction in the ignition delay time. Interestingly, in contrast to the pure butanes, methane/butane mixtures show no NTC region in our range of measurement conditions. No significant difference is found between the ignition promoting effect of the isomers of butane. However, the mixtures of the pentane isomers with methane do show differences: the ignition promoting effects of iso-pentane and n-pentane are similar, while the addition of neo-pentane to methane results in a substantially smaller reduction in the ignition delay time. In general, good agreement is found between measurements and calculations for all pure n-butane, iso-butane/methane, n-butane/methane, n-pentane/methane, iso-pentane/methane and neo-pentane/methane mixtures. The effects of CO addition on the ignition of H2, up to 50% CO in the fuel, and CH4, up to 20% CO in the fuel, are observed to be negligible both experimentally and computationally for the conditions studied here. Replacement of methane by hydrogen lowers the autoignition delay time. The addition of Syngas (CO/H2) to methane results in ignition behavior that resembles an equivalent methane/hydrogen fuel mixture with the same hydrogen fraction. In contrast to results previous presented in the literature, we thus observe no inhibiting effect from CO for the conditions in our experiments. Changing the reaction rate of HOCHHOCHH 23232 by reducing the pre-

exponential factor within the uncertainty reported in the literature yields excellent agreement for all H2, H2/CO, CH4/CO and CH4/CO/H2 mixtures studied. We also discuss the observation that autoignition delay times measured in different experimental venues can substantially differ from each other due to facility-specific differences. As an example, for H2/O2 mixtures, non-ideal behavior (observed as an increase in pressure) in shock tube data at lower

38

temperatures/longer delay times tends to shorten the delay time, as compared to the ideal constant-volume assumption, while non-ideality in RCM measurements (a decrease in pressure/temperature in the adiabatic core) tends to lengthen the delay times. This latter effect has been recognized as essential for modeling RCM data for several years, while the consideration of non-ideality in modeling shock tube data is relatively recent. Thus, the respective non-idealities in the two experimental venues drive the determined ignition delay times systematically apart, and can easily reach factors of 5 or higher. parameters such as activation energy that are derived directly from experimental data should be used with caution. In addition the impact of small amounts of NO2 to methane/oxidizer mixtures will be presented. The results show (figure 2) that the addition of 100 and 270 ppm NO2 to methane reduces the ignition delay time substantially, up to more than a factor of two. Moreover the results show that the ignition promoting effect of NO2 increase with increasing temperature (900-1050K) at pressures ranging from 25-50 bar. Although the model predicts the observed trend in decreasing the ignition delay time with increasing NO2 fraction, the computations overestimate the effects of NO2 addition on ignition.

Figure 1: Measured (filled symbols) and calculated (open symbols) total autoignition delay times for stoichiometric n-butane and iso-butane fuel as function of temperature at a fixed pressure Pc = 30.

0.1

1

10

100

1000

600 650 700 750 800 850 900 950 1000

Tim

e, m

s

Temperature, K

n-butane

iso-butane

n-C4H10 (calc.)

iso-butane (calc.)

39

Figure 2: Symbols denote measured auto ignition delay times in methane/NO2/O2/N2/Ar mixtures. Lines denote calculated autoignition delay times at a fixed pressure Pc 40. Acknowledgement

This research has been financed by a grant from the Energy Delta Gas Research (EDGaR) program. EDGaR is co-financed by the Northern Netherlands Provinces, the European Fund for Regional Development, the Ministry of Economic Affairs, Agriculture and Innovation and the Province of Groningen. We also gratefully acknowledge the financial support from the N.V. Nederlandse Gasunie.

0.1

1

10

100

0.94 0.96 0.98 1 1.02 1.04 1.06 1.08 1.1 1.12

Tim

e, m

s

1000/Tc (K-1)

0 ppm NO2100 ppm NO2270 ppm NO20 ppm NO2 (Gersenet al.)100 ppm NO2 (Gersen et al.)270 ppm NO2 (Gersen et al.)

40

Lar ge-EddySimulation of a Laboratory-Scale Flame in the MILDCombustion Regime

F. E. Hernández-Pérez, J. A. van Oijen, and L. P. H. de Goey

Combustion Technology, Eindhoven University of Technology, P.O. Box 513, 5600 Eindhoven, The Netherlands

Moderate and intense-low dilution (MILD) combustion is characterized by preheating and dilutionof reactants, producing more homogeneous temperature fields with lower peak temperatures as com-pared to conventional combustion modes. Consequently, MILD combustion leads to decreased pollutantemissions and noise levels [1]. Although large-eddy simulation (LES) has emerged in recent yearsas a promising method for the prediction turbulent reacting flows, accurate modelling of the flame-turbulence interaction at the subfilter scale (SFS) level remains a major challenge. In this research,the modelling of MILD combustion using LES with presumed probability density functions (PDF) andFlamelet-Generated Manifold (FGM) [2] tabulated chemistry is investigated. The LES-FGM formu-lation is applied to the Delft jet-in-hot-coflow (DJHC) burner configuration [3]. In particular, a casecorresponding to a fuel-inlet Reynolds number of approximately 4500 (DJHC-I) is considered. In thepresent study, beta and top-hat PDFs are implemented and compared. Furthermore, the tabulation ofchemical kinetic terms is based on igniting counterflow diffusion flames. A parallel, block-based adap-tive mesh refinement (AMR), finite-volume scheme is used for the numerical solutions of the Favre-filtered Navier-Stokes equations along with the additional equations for the SFS modelling [4]. TheFavre-filtered transport equations are solved employing a second-order accurate scheme in both timeand space. Details on the adopted modelling strategies, the numerical scheme and comparisons of theLES predictions to experimental data will be given in the presentation.

References

[1] Antonio Cavaliere and Mara de Joannon. Mild combustion.Prog. Energy Combust. Sci., 30:329–366, 2004.

[2] J. A. van Oijen and L. P. H. de Goey. Modelling of premixed laminar flames using flamelet-generatedmanifolds.Combust. Sci. Tech., 161:113–137, 2000.

[3] E. Oldenhof, M. J. Tummers, E. H. van Veen, and D. J. E. M. Roekaerts. Ignition kernel formationand lift-off behaviour of jet-in-hot-coflow flames.Combust. Flame, 157:1167–1178, 2010.

[4] F. E. Hernández Pérez.Subfilter Scale Modelling for Large Eddy Simulation of Lean Hydrogen-Enriched Turbulent Premixed Combustion. PhD thesis, University of Toronto, 2011.

1

41

Gas Turbine Combustion Research at the University of Twente, 1989-2012.

Combura 2012 Summary of a presentation by Jim Kok.

In gas turbine engines the combustors are the most critical components to design. In contrast with the design of compressors and turbines their design does not revolve around optimizing efficiency, but design efforts are targeted at clean and stable combustion and good quality of the hot gas flow through the combustor into the turbine. This needs to be achieved in all circumstances: summer or winter, part load and nominal load. Here severe problems can be encountered due to combustion dynamics, inducing high amplitude pressure oscillations in the combustor. These oscillations may develop into a limit cycle of such high saturated amplitude that they can damage the engine, due to fatigue failure, in a matter of minutes. Hence the design of a gas turbine combustor needs to fulfill all the requirements on clean combustion, hot gas characteristics, wall temperatures, combustion dynamics and structural vibration at part load and nominal load. It is important to emphasize that combustion processes in a gas turbine engine take place in a confined and compact volume, at elevated pressure and air inlet temperature, with significant wall heat loss and high turbulence and mixing intensity. This requires specific attention with a view to the models used and experiments performed. At the University of Twente the following topics are explored both theoretically and experimentally on turbulent combustion processes in gas turbine engines:

Combustion Dynamics in Natural Gas and Syngas combustors. Liquid fuel combustion in sprays. Acoustic phenomena in Combustors. Vibration and Mechanical Failure of Combustor structures. Formation and Emission of NOx, CO and UHC. Soot formation and deposition.

Figure above: Rolls-Royce DLN combustor: Axial velocity (L); Mixture fraction (M);Temperature (R). Theoretical work: Numerical modeling of combustion, heat transfer and acoustics. In the last decade of the previous century much work was performed at the UT on the development of a reaction progress variable model that was able to predict the partially premixed turbulent combustion of natural gas and syngas, taking into account effects of heat loss. This type of model is very suitable for combustion processes that occur in the distributed regime, which is typical for a gas turbine. In order to predict the emission of NOx and CO, detailed chemical reaction schemes were employed. The UT performed much work to develop a numerical code that specifies the optimal weight coefficients for the composed mass fraction components of the reaction progress variables. The reaction progress variable is then based on one or several composed mass fraction variables. This way a large detailed chemical reaction scheme can be projected on a small number of reaction progress variables. For each variable transport equations can be derived and numerically solved for a flame. The species mass fractions can be determined in several ways from the progress variables. The University of Twente model calculates this correlation by means of assumption of partial equilibrium of fast chemical reactions. The slow reactions determine the evolution of the progress variable. Another way that can be used is the Flame Generated Manifold, or a method like the Turbulent

42

Flame speed Closure that correlates this with laminar flame data and the gradient of progress variable and turbulent flame speed. At the University of Twente this model was employed initially to predict the steady combustion field and the emission of NOx, CO and UHC. At the start of this century the focus of the work moved to the transient behavior of turbulent flames. The reaction progress variable model was used in transient simulations to assess the dynamic characteristics of the flame and its flow interaction with the combustor. For the stability prediction of combustion in the combustor the Flame Transfer Function is important to calculate, because it links the response of the rate of combustion to flow perturbations in the burner inlet passage. Apart from the numerical flow simulation also the calculation of acoustics is important. Much progress is made in solving both the phenomena at turbulent flow velocity scale and at the much larger acoustic velocity scales. Still however acoustic network models render important information on stability that cannot be obtained otherwise.

Numerical modeling research on combustion dynamics was performed on laboratory scale test rigs like the LIMOUSINE and DESIRE rigs, but also on industrial gas turbine combustors. The following examples can be mentioned: The Siemens V94.2 (Puertollano IGCC), Rolls-Royce, GE frame 9 DLN1 and P&W P200. This was performed in close cooperation with companies like Siemens, Rolls-Royce, Endesa, Electrabel, Sulzer, Ansaldo Thomassen and INNECS. The work on the P&W combustor was done in cooperation with NLR and the Royal Dutch Air Force. For INNECS a 2 MW/4 bar low NOx burner was developed based on the DESIRE design burner and suitable for medium to high calorific gas. The figure on the previous page shows numerical simulation results of the FLAMESEEK project using the reaction progress variable combustion model on a DLN NG fired Rolls-Royce combustor.

The work on combustion dynamics involves also modeling of two way Fluid Structure Interaction. Next to the work on combustion dynamics also models are developed for liquid fuel combustion and soot formation in ultra rich flames. Experimental research. For exploration of physical processes and for model validation, experimental work and facilities are very important. The University of Twente has a laboratory available for operation of combustors at 6 bar pressure and with the use of Natural Gas, Hydrogen and Carbon Monoxide. Three rigs are used specifically to study combustion dynamics: DESIRE test rig: 500 kW/5 bar Natural Gas fuel, air inlet 300 Celsius. HEGSA test rig: 100 kW/5 bar Syngas fuel, air inlet 250 Celsius. Limousine test rig: 80 kW/1 bar Methane fuel, air inlet 250 Celsius. Here the Flame Transfer Function and combustion dynamics are determined with the use of pressure transducers, a Photo Multiplier Tube and a home developed siren or MOOG valve. The production of syngas from natural gas and the formation of soot in ultra rich combustion conditions is explored with the ULRICO test rig: 300 kW syngas/6 bar Natural Gas fuel. Here the gas composition is measured with a Fourier Transform Infra Red analyser and the soot particle distribution is measured with a SMPS analyser. Funding. The last decade the majority of the work was funded by European budgets in 7 projects. The largest project was LIMOUSINE (2008-2012), with 5 PhD students at the UT and all 18 PhD students coordinated by the UT. The latest EC funded project that started this year is COPA, with 3 PhD students at the UT. Here is targeted either the simulation of a full engine and/or the modeling of multiphysics phenomena. Massive parallel methods are envisaged using 100,000 cores.

43

Combustion Technologies for future gas turbines : design methodology and validation

Werner Krebs (SIEMENS AG, Germany)

The future energy production is dominated by the increase share of renewables which provide fluctuating energy production driven by the availability of solar radiation and wind. In this context the energy consumption and the energy production are not synchronized and compensating devices need to be provided. Gas turbines are seen as a suitable technology solution for filling the gap between energy production and consumption due to their high flexibility and comparable low cost. However the development targets for gas turbines will be redirected towards higher operational flexibility and fuel flexibility whilst maintaining high efficiencies and low overall cost of the power plant. Regarding combustion technologies the focus will be set on increasing fuel flexibility meeting different kinds of fuel specification including a wider variety of natural gas blends as well as admixing of H2. In addition the stable combustion operation range needs to be extended towards lower firing temperatures to meet part load emissions of future power plants. The presentation will provide the technical boundary conditions and needs for the development of new combustion technologies for future gas turbines. An overview on different fuel specifications is presented. Details on the development cycle for new gas turbine combustion technologies is given including the aspects of the new Clean Energy Center for validation of new gas turbine combustion technologies. Another major contribution to the development are more accurate prediction tools which need to better represent details of fuel as well as operational aspects. As example recent developments for predicting CO emissions at partload as well as recent investigations on nonlinear thermoacoustics are given. Finally the technology solutions of SIEMENS gas turbines are outlined and potential design hypothesis are given.

44

CARS MEASUREMENTS OF TEMPERATURE IN A FLAMELESS OXIDATION FURNACE

J. Lu, E.-S. Cho, E.H. van Veen, W. de Jong, D.J.E.M. Roekaerts Department Process & Energy, Delft University of Technology, The Netherlands

[email protected]

This paper reports on an experimental study on flameless combustion in a Multi‐burner Excess

Enthalpy Furnace. The objective is to measure the local and instantaneous temperature

fluctuations in a furnace with two pairs of regenerative FLOX® burners using Coherent anti‐

Stokes Raman Spectroscopy (CARS) [1], each burner pair having a thermal input 100 kWth. In

the recent past, also 3–burner pair experiments (3x100 kWth) have been carried out using this

furnace. So the experimental series described in this paper also sheds light on the effect of a

lower power per unit volume, which is relevant for real industrial furnace operation conditions.

The application of flameless oxidation (FLOX® [2]), also known as HiTAC or MILD combustion, in

furnaces is an effective combustion technology to realize high efficiency for enhancing the heat

transfer combined with the advantages of low emissions of NOx and CO. It is based on a

combination of an increase of air temperature above the self‐ignition temperature of the

reactant mixture with a low oxygen concentration in the flame zone. Separate injection of fuel

and oxidizer with high momentum leads to entrainment of flue gas and in‐furnace recirculation

thus lowering the local oxygen concentration and providing spatially wider distributed heat

release and reduced peak temperature.

Fig.1 Overview of CARS experiment

The CARS experimental setup consists of an injection‐seeded Nd:YAG laser(532nm), a modeless

dye laser(607nm), various optics, transporting rails, a spectrometer/CCD combination and a

computer which is shown in Figure 1. The experiments were carried out for a fixed burner

configuration, but for two firing modes (parallel, staggered) and two cycle times (20s, 60s).

Measurements were made in the near‐burner zone along a line 250mm away from the furnace

wall and crossing the mixing jets. Separate measurement series of were obtained while the

burner was firing and while it was generating. The temperature at a number of points was

45

recorded at a frequency of 10Hz. As a result the probability density function of the local

instantaneous temperature was obtained for operation of the furnace in multiple conditions [3].

These results can be used for validation of numerical predictions using CFD.

Fig.2 Mean temperature comparisons Fig.3 Temperature trends during start‐up

In the CFD (RANS, EDC) simulation result in the right‐bottom and right‐top of Figure 2, it can be

observed that the mean temperature sequence of positions A, B, C, D is TD>TB>TC>TA. In the

experimental result shown in the left hand side of Figure 2, the same trend could be observed.

Figure 3 shows the comparison between a thermocouple result and a CARS result in both flame

mode (used during start up of the furnace) and flameless mode. Tfurnace stands for the

thermocouple temperature while Tfiring stands for the CARS result in firing mode and Tregen stands

for the CARS result in regeneration mode. Tflox stands for the CARS result in FLOX mode. It can be

seen that on average the CARS result is higher than the thermal couple result. This is mainly due

to the difference in location (spatial inhomogeneity of mean temperature) of both

measurements. To further explore this, a series of local thermal couple experiments was made

to get the temperature profile in the near burner zone, and to compare it with the CARS result.

In the experiment it was also observed that the fluctuations of temperature in the staggered

mode are higher than that in parallel mode, at least when the burner operates in flame mode.

The occurrence of fluctuations with high temperature can provide an explanation of the

previous observation that the NOx emission in staggered mode is higher than in parallel mode

despite the temperature of the staggered mode being lower.

In future work, more experiments will be carried out on fuel flexibility and using local probe

measurements. The experimental results will be compared with numerical model predictions

References

[1] E.H. van Veen, D. Roekaerts, Applied Optics 44 (2005) 6995‐7004

[2] J.A. Wünning, J.G. Wünning, Prog. Energy Combust. Sci. 23(1997) 81‐94.

[3] E.‐S. Cho, B. Danon, W. de Jong, D.J.E.M. Roekaerts, Applied Energy 88 (2011) 4952‐4959

46

Chemical Kinetic Analysis in Turbulent Lifted Flames in a Hot Coflow S. M. Mir Najafizadeh1,2, M. T. Sadeghi2, D.Roekaerts1

Iran University of Science & Technology, School of Chemical Engineering, Tehran, Iran

Delft University of Technology, Process and Energy, Mekelweg 2, 2628 CD Delft, Netherlands

Pilot, swirl and bluff-body stabilized flames are widely used for flame stabilization in practical combustors and furnaces, by mixing hot combustion products and cold reactants, leading to a continuous process of ignition. To investigate the mechanisms controlling the flames in an environment of hot gases, Jet-in-Hot-Coflow (JHC) flames are employed since they can emulate properly the conditions of non-premixed jet flames in hot coflow of the post combustion gases while the flow field is decoupled from the chemistry. The governing mechanism for flame stabilization in JHC flames has been investigated by numerous experimental and numerical studies [1-4]. According to these studies, for a wide range of oxygen concentration in the hot coflow, autoignition is the main mechanism for flame stabilization. Studies of the budgets of convection, diffusion and reaction in the species transport for turbulent lifted flames in hot coflow indicate that near the flame base the reaction term is balanced by the convective term with minimal contribution of axial diffusion. This behaviour implies the occurrence of autoignition. It differs from the case of flames stabilized by premixed flame propagation which are characterised by a diffusive-reactive balance, preceded by a convective-diffusive balance. Autoignition is a transient process initiated from a slowly reacting state and eventually leading to a fully burning state corresponding to the combustion at high temperature. The reactions controlling autoignition may be different from those in high temperature combustion, and the representation of finite rate effect is important. The difference could be used for the recognition of autoignition in turbulent lifted flames in the hot coflow. To investigate autoignition from this point of view through modeling of the turbulent reacting flow, an accurate representation of both the chemical mechanism and the turbulence-chemistry interaction is required. In the present study, the case of the hydrogen oxidation is chosen for understanding of autoignition process since the chemistry is more accurately known than the chemistry of hydrocarbon fuels and at the same time hydrogen oxidation mechanism constitutes a part of the chemistry of hydrocarbon fuels. The flame studied is known as the Cabra flame [4]. A variety of modelling methods have already been evaluated for representing the highly non-linear interaction between turbulence and chemical reaction in lifted flames in the hot coflow. Among them, the transported PDF approach has the advantage of representing reaction exactly without modeling assumptions. Using this method it is possible to accurately represent the trends in lift-off height with coflow temperature [2]. The present work combines the PDF approach with reaction analysis to investigate in detail the ignition process. Reaction rate analysis answers the question what is the relative contribution of different reaction channels. A post-processing code, written in FORTRAN, was developed in the present study to evaluate the reaction rates at different locations in the flame. This analysis allows determining the dominant chemistry at the flame base. The particle properties calculated by the Monte Carlo solver are post-processed in order to obtain the reaction rates. To obtain the mean reaction rates, mass weighted averaging is used in each cell in the computational domain. Then the noise in the averaged results is reduced using a least-squares smoothing filter. To examine the contribution of each reaction in the production and consumption of H radicals, reaction rates at two locations are compared: one at the flame base and one at a position where the high temperature flame is dominant. Fig. 1 shows the radial profile of mean reaction rates of several elementary reactions leading to the production or consumption of H radicals. The figure illustrates that the reactions responsible for the productions of the H radicals are mainly R2 and R3, given by

O + H2 = H + OH (R2) OH + H2 = H + H2O (R3)

47

while they are consumed by reactions R1, R9 and R11, given by H + O2 = O + OH (R1)

H + O2 + M = HO2 + M (R9) HO2 + H = 2OH (R11)

Reactions R1 and R9 are well known in gas kinetic processes. They compete for H radical in the induction stage during ignition process. R1 is the chain branching reaction which directly increases the amount of radicals in the radical pool, along with reactions R2 and R3, while reaction R9 (which acts initially as a chain termination reaction) produces the less reactive HO2 radical reducing the reactivity of the system. It was observed that near the nozzle exit plane (not shown in a figure), reaction R9 is dominant while the contribution of reactions R1 and R11 is negligible. Reaction R9 acts in this region as a termination reaction and HO2 is acting as a sink species for the H radical. This behaviour leads to a delay in ignition in this region. Near the flame base, while reaction R9 is still dominant relative to R1, reaction R11 becomes increasingly important and affects the consumption of H radical. HO2 radical does not behave as a pure sink species in this region and it is decomposed into two OH by R11. Those OH radicals then take part in reaction and eventually boost chain branching process and induce thermal runaway. This behavior of elementary reactions has been reported in the literature and is attributed to autoignition. Downstream of the flame base, (near x/D=18 at Fig. 1), reactions contributing in chain branching process, R1, R2 and R3, are dominant while the contribution of R9 and R11 are relatively small. This dominant chemistry implies the existence of a high temperature flame.

Fig1.Contribution of elementary reactions in the production and consumption of H radicals

It is concluded that reaction rate analysis confirms the occurrence of autoignition at the flame base of a turbulent lifted flames in the hot coflow, with the chain branching reaction, R11 playing an essential role. References 1. E. Oldenhof, M. J. Tummers, E. H. Van Veen and D. J. E. M. Roekaerts, Ignition Kernel formation and lift-off behaviour of jet-in-hot-coflow flames, Combustion and Flame, 157 (2010), pp. 1167-1178. 2. R. L. Gordon, A. R. Masri, S. B. Pope and G. M. Goldin, A Numerical Study of Auto-ignition in Turbulent Lifted Flames Issuing into a Vitiated Coflow, 11 (2007), pp. 351-376. 3. C. N. Markides, E. Mastorakos, An Expermental Study of Hydrogen Auto-ignition in a turbulent Coflow of Heated Air, Proceedings of the Combustion Institute, 30 (2005), 883-891. 4. R. Cabra, T. Myhrvold, J. Y. Chen, R. W. Dibble, A. N. Karpetis, and R. S. Barlow, Simultaneous Laser Raman-Rayleigh-Lif Measurements and Numerical Modelling Results of a Lifted Turbulent H2/N2 Jet Flames in a Vitiated Coflow, 29 (2002) pp. 1881-1888.

48

Applications of peroxy-fuels in vehicle propulsion

Kirti Bhushan Mishra∗, Klaus-Dieter Wehrstedt

Division 2.2 “Reactive Substances and Systems”BAM Federal Institute for Materials Research and Testing

Unter den Eichen 87, 12205 Berlin, Germany

1. Introduction

The combustion of petroleum derived conventional fuels (diesel, gasoline etc.) takes place relatively at a slowerrate. In many applications they are forced to combust faster in order to match the specific power/output requirement.To accelerate the combustion process fuel (and oxidiser) supply has to be correspondingly increased by some means.These imply the necessity of high pressure fuel feed pump, air compressor (super or turbo charger) and a large volumeof chamber to combust these mixtures [1]. As an outcome the overall compactness of the vehicle declines. Further-more, as an additional consequence in form of emissions e.g. COx and NOx occur as a result of fast combustion offuel and air [2].We investigate combustion of a class of energetic materials commonly known as organic peroxides (or peroxy-fuels)as a main fuel in vehicle propulsion [2, 3, 4]. Peroxy-fuels are typically known as reaction enhancer’s, polymerizingagents and cross linkers. Sometimes, they have been also utilised in small amounts (1% to 5%) to conventional fuelssuccessfully. It was found that combustion process was improved and emission was drastically turned down. There-fore, they are termed as combustion improvers as well. Also it was shown that with a peroxy-fuel an engine can berun anaerobically [5].

2. Present Work

We have recently investigated four different peroxy-fuels as listed in Table 1. The liquid fuels are burned in formof pool flames (non-premixed flames). All experimental conditions were kept similar for all fuels. The characteristicsof interest are mass burning rates, flame lengths, flame temperatures and thermal radiation. These parameters weremeasured with calibrated instruments. The reproducibility of the measured data is also ensured. The emphasis wasgiven to learn the mass, momentum and energy transfer for the above fuels. The first result on mass burning rate(Fig. 1) against the pool diameter clearly show the overwhelming large amount (3 to 114 times) mass transfer withinsame time in case of peroxy-fuels compared to kerosene. Due to higher burning rates the visible flames are 5 times

Figure 1: Mass burning rates (left) and flame lengths (right) of kerosene and peroxy-fuel pool fires [3]

larger than in case of kerosene (Fig. 1). A relationship between pressure drop 4p and mass flow rate m′′

f for anincompressible fluid states that

4p ∝(m

′′

f

)2

and q ∝ m′′

f . (1)

Correlating eqs. (1) to both fuels reveal that a significant amount of pressure drop (consequently pump power andnoise) can be saved for a fixed heat release rate due to combustion q (kW/m2) [2].

∗Corresponding author: [email protected], Mastricht, The Netherlands

49

3. Application in Internal Combustion (IC) Engines

Based on the characteristics we have developed a concept of an IC engine operating purely with peroxy-fuels. Thevarious components and operation methods of such engines are discussed in detail in [2]. Only a summary is providedhere. The power P developed by a typical IC engine and the compression ratio r are defined as

P ∝ p VS and r =VS + VC

VC(2)

where p is the mean effective pressure developed in the engine cylinder and VS is the stroke and VC is clearancevolume of the cylinder. It can be futher shown that for gasoline and diesel engines p and VS is dependent on the amountof intake charge i.e. mixture of fuel and air in the following ways

P ∝ p ∝(m

′′

f

)2

and P ∝ VS ∝ m′′

f . (3)

According to eqs. (2) and (3) a peroxy-fuel based engine will consume 5 to 114 times less amount of fuel for a con-stant pressure cycle (diesel) and therefore VS will also be reduced in the similar fashion. In case of a constant volumecycle (gasoline) p can be reduced by 9 to 5000 times depending on the selection of a particular peroxy-fuel (Table 1).Hence, the overall and volumetric efficiencies of the engine will be increased by a factor of 3 to 114 [2].

Table 1: Properties of investigated peroxy-fuels [2, 3] and of general hydrocarbon fuels [4]Type of fuel → INP TBPEH TBPB DTBP Kerosene Gasoline DieselParameter ↓Formula C18H34O4 C12H24O3 C11H14O3 C8H18O2 C12H26 C4 to C12 C8 to C25

Molar mass (g/mol) 314.5 216.32 194.2 146.2 170 100–105 ∼200Active oxygen (%) 5.09 7.40 8.24 10.94 - - -Enthalpy of 30100 34455 30113 36600 46300 44400 45400combustion (kJ/kg)TSADT (K) 293 308 338 358 - - -Air to fuel ratio (-) 10.7 10.52 9.23 10.86 15.6 14.7 14.7m

′′

f (kg/m2s) (d = 6 cm) 4 0.53 0.83 0.18 0.012 0.055 0.035INP: Di-(3,5,5-trimethylhexanoyl) peroxide; TBPEH: tert-Butyl peroxy-2-ethylhexanoate; TBPB: tert-Butyl peroxy-benzoate; DTBP: Di-tert-butyl peroxide.

Due to the low air to fuel ratio (Table 1) and the reduced volume in case of peroxy-fuel combustion the overallemission of NOx can also be stepped down significantly. Moreover, the oxygenated quality of the fuel helps to estab-lish oxy-fuel like combustion [2].The safety in dealing with peroxy-fuels are of prime concern. In Table 1 the self-accelerating decomposition temper-atures TSADT of peroxy-fuels are given. For safety reasons during transportation and processing the temperature of aperoxy-fuel should not go beyond the same [2, 3].

4. Conclusion

A concept of an internal combustion engine operating with peroxy-fuels is developed and it has been shown thatthe engine requires considerable less amount of fuel for a given power. This consequently enhances the compactness.It can further be shown that with peroxy-fuels the emissions can also be controlled upto a great extent.We encourage partners from industries/institutes to come forward, collaborate and exploit the exisisting knowledge.

References

[1] C. F. Taylor, The internal combustion engines in theory and practice, MIT Press, ISBN-0-262-20051-1.

[2] K. B. Mishra, K.-D. Wehrstedt: Komponente fur Verbrennungsmotoren zum Betrieb mit Peroxy-Kraftstoffen,German patent DE 10 2011 051 228.4 (filed).

[3] K. B. Mishra: Experimental investigation and CFD simulation of organic peroxide pool fires (TBPB andTBPEH), BAM- Dissertation Series 63 Berlin 2010, ISBN 978-3-9813550-6-2.

[4] www.afdc.energy.gov/pdfs/fueltable.pdf.

[5] H.O. Pritchard, Anaerobic operation of an internal combustion engine, US Patent No.4800847.2

50

FGM modelling of a gas turbine model combustor

F.A. Tap*, M.J. Wankhede, W.J.S. Ramaekers and P. Schapotschnikow

Dacolt International B.V.

Grote Looiersstraat 28a

6211 JJ Maastricht,

The Netherlands

*Corresponding author: [email protected]

A gas turbine model combustor is simulated with unsteady RANS CFD using the Flamelet-Generated Manifold (FGM) combustion model and Reynolds Stress Model (RSM) for turbulence. The purpose of this study is two-fold. First, to develop an unsteady RANS based CFD simulation strategy to capture the combustor flow field. And second, to numerically characterize the complex reacting flow field and compare it against experimental data. The FGM [1] reduction method generates a low-dimensional manifold based on one-dimensional premixed flame structures. The use of this approach results in a large reduction in the number and stiffness of required transport equations, while it still includes detailed reaction kinetics and species diffusion in reaction layers. Dacolt has implemented the generation of FGM tables, based on the Chem1D flame solver [2], in Tabkin®. Tabkin is Dacolt’s dedicated Software-as-a-Service package for generating CFD look-up tables. The commercial CFD solver ANSYS® FLUENT® is used. The FGM model is implemented through Dacolt’s progress variable / mixture fraction framework of UDFs developed for the Dacolt PSR+PDF combustion model [3]. Special care has been taken in NOx modelling following the approach of Vreman et. al. [4]. The FGM tables generated with Tabkin for the present study make use of the GRI 3.0 reaction mechanism for natural gas combustion [5]. Simulation results of a gas turbine model combustor are then compared to detailed experimental results [6]. This experimental combustor set-up ensures a well defined set of boundary conditions for CFD simulation and analysis. The 3D computational domain is shown in Figure 1 below. The burner consists of an air plenum common to two concentric radial air swirlers and an annular ring of small fuel channels between the air swirlers through which methane is supplied. The two air streams and the fuel stream all discharge and mix through the nozzle into the square combustion chamber where the main combustion

51

process takes place in partially-premixed mode. The burnt gases exit at high velocity through a circular exhaust tube further downstream the chamber.

Figure 1: Computational domain of the model combustor

The computational mesh used for the CFD analysis is built with ICON FOAMpro. A hex-dominant mesh with polyhedral/tetrahedral transitions between refinement zones is generated with a total cell count of ~7.5 million cells. Overall the mesh size and topology used is designed to provide reasonable accuracy of flame structure/shape and near-injector flow-field in particular. The global equivalence ratio of the flame investigated in this study is 0.65 and the swirl number is approximately 0.9. Overall, this study provides a first performance analysis of the URANS/FGM approach in terms of flow-field, temperature, species and NO predictions, for an adiabatic swirling turbulent flame representative of a gas turbine combustor. Acknowledgements Ansaldo Energia and Ansaldo Thomassen are gratefully acknowledged for their support of this work. DLR is also kindly acknowledged for making the experimental data and geometry available. References

1. J.A. van Oijen and L.P.H. de Goey, Combust. Sci. Technol. 161 (2000), pp. 113–137. 2. Tap, F. and Schapotschnikow, P., SAE Technical Paper 2012-01-0152. 3. Vreman AW, Albrecht BA, van Oijen JA, de Goey LPH, Bastiaans RJM, Combust. Flame 153

(2008) 394-416. 4. http://w3.wtb.tue.nl/en/research/research_groups/combustion_technology/research/flamecodes/c

hem1d/ 5. http://www.me.berkeley.edu/gri_mech/version30/text30.html 6. Weigand, P., Meier, W., Duan, X. R., Stricker, W., Aigner, M., Combust. Flame 144 (2005) 205-

224.

52

Fundamental study of methanol flames by use of contained explosions and flat flame burner measurements J. Vancoillie*, S. Verhelst [email protected] Department of Flow, Heat and Combustion Mechanics, Ghent University Sint-Pietersnieuwstraat 41, 9000 Gent, Belgium The use of methanol in spark-ignition engines forms a promising approach to decarbonizing transport and securing domestic energy supply. The physico-chemical properties of this fuel enable engines with increased performance and efficiency compared to their fossil fuel counterparts. An engine cycle code valid for methanol-fuelled engines could help to unlock their full potential. However, the development of such a code is currently hampered by the lack of a suitable correlation for the laminar flame speed of methanol-air-diluent mixtures and its dependence on flame stretch.

The laminar burning velocity UL is a key parameter characterizing the combustion behavior of a combustible mixture. This property is dependent on the pressure, temperature and mixture composition (fuel type, equivalence ratio and amount of diluents). Whereas the laminar burning velocity at standard conditions provides invaluable information on the combustion properties and the underlying oxidation chemistry of the given fuel, it is crucial to quantify the effects of pressure and unburned mixture temperature for engine simulation purposes.

Since many of the existing experimental data for this property are compromised by the effects of flame stretch and instabilities, this study was aimed at obtaining new, accurate data for the laminar burning velocity of methanol-air-diluent mixtures using two distinct experimental methods, complemented by chemical kinetics simulations performed with the methanol oxidation mechanism of Li et al. [1].

A first experimental method consisted of non-stretched flames that were stabilized on a perforated plate burner at atmospheric pressure (see Figure 1). The Heat Flux method was used to determine burning velocities under conditions when the net heat loss from the flame to the burner is zero. Equivalence ratios and initial temperatures of the unburned mixture ranged from 0.7-1.5 and 298-358 K respectively. Additionally the effects of diluting methanol-air flames with water vapor (up to 20 mole%) and nitrogen (up to 10 mole%) were examined.

A second series of experimental results was obtained using the spherically expanding flame technique. By recording spherically expanding flames in a optically accessible fan stirred bomb (see Figure 2) through Schlieren photography, the laminar burning velocity of methanol-air mixtures and its dependence on flame stretch (expressed by the Markstein length 𝐿𝑏) could be measured for equivalence ratio between 0.8 and 1.4, unburned mixture temperatures between 303 and 383 K and pressures between 1 and 10 bar.

53

Figure 1: Perforated plat burner used for measuring UL by applying the Heat flux method

Figure 2: Combustion vessel used for measuring UL by applying the spherically expanding flame technique

In lean conditions the correspondence of the Heat Flux method results with recent literature data and chemical kinetics calculations was very good, whereas the expanding flame technique resulted in UL values that were 5-10 cm/s lower. For rich conditions both methods support the higher burning velocity as predicted by several chemical kinetics mechanisms, as opposed to the lower values reported in previous experimental works.

The effects of unburned mixture temperature and pressure on the laminar burning velocity of methanol were

analyzed using the correlation 𝑈𝐿 = 𝑈𝐿0 ∙ �𝑇𝑢𝑇𝑢0�𝛼∙ � 𝑝

𝑝0�𝛽

. Whereas most existing expressions for the temperature

exponent α assume a linear decrease with increasing equivalence ratio, both the modeling and experimental results produce a minimum in α for slightly rich mixtures. The experimentally derived values for β agreed well with the modeling results and showed only a minor dependence on equivalence ratio.

The effect of dilution on UL was well predicted by the chemical kinetics calculations for both N2 and H2O dilution. The relative impact of thermal and chemical effects of dilution was estimated computationally. For both N2 and H2O the chemical effect was shown to be negligible for diluents ratios considered here (< 20 mole%). Based on the modeling results, an explicit correlation was proposed that describes the effect of dilution on SL in terms of diluent molar content, diluent specific molar heat capacity, equivalence ratio and unburned mixture temperature. Very good agreement was obtained between the correlation and the modeling data.

Experimental results for the Markstein length Lb were subject to large uncertainty bounds. Nevertheless the reported trends for Lb in terms of equivalence ratio, temperature and pressure were well reflected in our measurements. The Markstein length decreased as a function of equivalence ratio and pressure. It was almost independent of temperature.

In conclusion, the current experimental database for UL confirms that the methanol oxidation mechanism of Li et al. [1] adequately simulates the laminar burning velocity of methanol-air-diluent mixtures as a function of equivalence ratio, diluents ratio, pressure and temperature. It can therefore be used with confidence to develop a laminar burning velocity correlation for use in engine cycle code valid for methanol-fuelled engines.

References

1. Li, J., Z.W. Zhao, A. Kazakov, et al., "A comprehensive kinetic mechanism for CO, CH2O, and CH3OH combustion", International Journal of Chemical Kinetics, 39(3), p. 109-136, 2007.

54

Combustion, flame and burner design:Challenges and computational tools

L. Vervisch, P. Domingo, G. Lodier, V. Moureau

National Institute of Applied Sciences (INSA) Rouen& CNRS-CORIA

Key issues and modeling challenges in numerical simulation of turbulent flamesmostly result from the strong multi-scale character of gaseous turbulent reactiveflows. Combustion chemistry involves a very large number of chemical species (about50 for simple hydrocarbon fuels) reacting over hundred (or even more) elementaryreaction steps, whose characteristic time scales spread over a very large spectrum,with the smallest of the order of a few microsecond. In terms of length scales, underatmospheric pressure, the thermal flame thickness is of the order of 100 micrometersfor usual fuels, while intermediate radicals evolve over lengths of the order of a fewmicrometers. In most real combustion systems, the turbulent fluctuations at largeand small scales feature length and time scales which fully overlap those of theflames, leading a strong increase of the overall burning rate thanks to turbulentmixing, but also to serious modeling challenges when the lack of mesh resolutiondoes not allow for describing accurately all these scales and their strongly non-linearinteractions.

A large variety of turbulent combustion modeling tools have been developed totackle the simulation of turbulent flames and help in the design of combustion sys-tems. Looking at the technical literature, the most frequently used, and sometimessuccessful closures, are based on the hypothesis that the burning rate is essentiallymixing controlled, with a cut-off time-scale provided by a simplified chemical kinet-ics. However, these approaches are limited by construction and provide misleadingresults when they are applied outside of the parameters range for which they havebeen tuned. On the other hand, how to precisely account for a more detailed descrip-tion of the flame physics in the design loop of real burners in complex geometries isstill an open question in many practical cases, and high pressure combustion makesthe problem even more complex.

Along these lines, some of the recent developments in LES of turbulent burnerswill be discussed (Figures 1 and 2), focusing on new approaches related to the down-sizing of detailed chemistry using optimization tools, and, to sub-grid scale modeling

1

55

that preserves the spectral properties of sub-filter (unresolved) length scales. A ten-tative schedule for the systematic application of these advanced methods to burnerdesign will also be given.

Figure 1: Iso-contour c = 0.8 of the progress variable in a 2.6 billon cells DNS of aswirling burner [1].

(a) 0.94 tTDC (b) 0.99 tTDC (c) 1.07 tTDC

Figure 2: Evolution of turbulence during ignition in a rapid compression machine,iso-contour of Q-criterion (Q = 15 · 106 s−2) colored by temperature. Flow goes fromleft to right [2].

References

[1] V. Moureau, P. Domingo, and L. Vervisch. From large-eddy simulation to directnumerical simulation of a lean premixed swirl flame: Filtered laminar flame-pdfmodeling. Combust. Flame, 158(7):1340–1357, 2011.

[2] G. Lodier, C. Merlin, P. Domingo, L. Vervisch, and F. Ravet. Self-ignitionscenarios after rapid compression of a turbulent mixture weakly-stratified intemperature. Combust. Flame, in press.

2

56

Strategies for Combustor Design

Dr. Moresh J. Wankhede

Combustion CFD Engineer Dacolt International B.V.

Maastricht 6211 JJ, The Netherlands

Since the 1950’s, gas turbine combustor technology has developed gradually and

continuously rather than any dramatic changes, which is why most combustor designs

resemble each other. In 2001, the Advisory Council for Aeronautical Research in Europe

(ACARE) laid down stringent fuel consumption and pollutant emissions targets for the year

2020. Year 2020 is not far off in terms of component development cycle times in the gas

turbine industry and new targets for 2050 are already under evaluation. Also, it is clear

that these current and upcoming stringent targets could only be realized by major step

change in gas turbine technologies and developing rapid and efficient component design

methodologies. Thus, the strategy employed during combustor design and development

would have a direct impact on the achievability of ACARE targets.

Combustor design is a complex procedure due to simultaneous involvement of many

conflicting performance requirements that are strongly coupled to each other. During the

design and development phase of gas turbine combustors, the use of computational fluid

dynamics (CFD) simulations of transient combustor aero-thermo-dynamics to provide an

insight into the complex reacting flow-field is expensive in terms of computational time. A

large number of such high-fidelity reactive CFD analyses of the objective and constraint

functions are normally required in combustor design and optimisation process. Hence,

traditional design strategies utilizing only high-fidelity CFD analyses are often ruled out,

given the complexity in obtaining accurate flow predictions and limits on available

computational resources and time. Further, to make traditional design approaches using

expensive CFD simulation efficient and practical in the context of combustor design, it

requires a methodology where the search algorithm is not coupled directly to expensive

CFD simulations.

57

In this study, computationally efficient strategies for combustor design are developed and

demonstrated; where pre-defined set of combustor CFD simulations within the target

design space can be represented by an intermediate approximate model on which a global

search is performed. The intermediate model is referred to as surrogate or response

surface model (RSM). Kriging RSM based high-fidelity and various co-Kriging RSM based

multi-fidelity strategies are developed and applied for the design of a two-dimensional test

combustor problem. The design and optimisation problem is set-up for two geometric

variables and a single-objective function. Initially, a Kriging based design strategy using

only high-fidelity simulations is applied for combustor design. Then, various co-Kriging

based multi-fidelity strategies consisting of two levels of fidelity; a fast but approximate

low-fidelity and an expensive but accurate high-fidelity combustor solution, are developed

and applied to perform combustor design optimisation. It is observed that Kriging strategy

outperforms the co-Kriging strategy within the current problem set-up and fixed

computational budget. However, both Kriging and co-Kriging based strategies find an

optimal combustor design very early on in the process.

Keywords: computational fluid dynamics, response surface model, Kriging, co-Kriging,

design optimisation, gas turbine and combustor

Further reading: [1] Wankhede, M. J., 2012, Multi-fidelity strategies for lean burn combustor design, University of Southampton, Faculty of Engineering and the Environment, PhD Thesis [2] Wankhede, M. J., Bressloff, N. W., Keane, A. J., 2011, Combustor design optimisation using co-Kriging of steady and unsteady turbulent combustion, Journal of Engineering for Gas Turbines and Power, GTP-11-1113, 133(12), (DOI: 10.1115/1.4004155) [3] Wankhede, M. J., Bressloff, N. W., Keane, A. J., Caracciolo, L., Zedda, M., 2010, “An analysis of unstable flow dynamics and flashback mechanism inside a swirl-stabilised lean burn combustor,” Proc. ASME Turbo Expo 2010: Power for Land, Sea and Air, GT2010-22253, Glasgow, UK

58


Enhance radiant heat transfer by helical inserts in an annealing furnace

H. Wu1, A. Radford2, A. Thomas2 1Tata Steel Research, Development & Technology, P.O.Box 10.000, 1970 CA IJmuiden, The Netherlands

2Tata Steel Strip Products UK, Port Talbot Works, Port Talbot, South Wales, SA13 2NG, United Kingdom

Key words: Furnaces, energy efficiency

Tata Steel is Europe’s second largest steel producer with its manufacturing operations primarily in the UK and The Netherlands. In cold rolling mills, cold rolled strip needs to be annealed to 600-850°C via either continuous annealing or batch annealing to restore its mechanical properties before being further processed. Radiant tube furnace is the type of furnace that is typically used in the continuous annealing process. In the furnace, heating energy is provided by combustion reaction taking place inside so called radiant tube units while the space between the tubes and the furnace enclosure is filled with protective gas to prevent strip from oxidation. Depending on furnace capacity, one furnace can contain a large number of such tubes, varying from a few tens to a few hundreds.

Natural gas is supplied to each radiant tube and fired through a burner device. Gas consumption is usually one of the major operational costs of the process. Thus, reducing the gas usage reduces not only production cost but also CO2 emission from the combustion.

There are a few options to reduce the gas consumption of the furnace by for example reducing heat losses through wall and leakage, adopting modern combustion technology, increasing heat recovery efficiency, etc. These measures require, however, substantial investment and weeklong production stop. Another option to place helical inserts inside the radiant tube appears to be a simple application and requires barely any modification to the existing installation. The inserts in helical shape can be made of steel or ceramic material which shall have high temperature resistance and radiative emissivity. It can potentially enhance heat transfer from the hot gas produced by combustion to the radiant tube and further heat radiation from the tube to the strip.

To quantify the effect, one set of inserts was installed and tested in the radiant tube test furnace located at the IJmuiden research centre. The test furnace contains a full scale radiant tube, a recuperative burner with a heating capacity of 150kW and associated equipment. The test results showed that

the inserts increased the metal temperature of the affected tube length by 27°C, as seen in Figure 1. The extent of the temperature increase was primarily influenced by burner load and prevailing surroundings temperature. Meanwhile, a few sets of inserts were trialled in the radiant tube furnace of the Continuous Annealing Process Line at the Port Talbot site. The radiant tubes that had the inserts installed were monitored and compared with the adjacent tubes that contained no insert. The comparisons showed the similar effect on the tube temperature from the inserts. The laboratory tests demonstrated also that the inserts introduced a small increase in system flow resistance. Further no negative impact was observed on combustion CO and NOx emissions.

Based on the test results, furnace simulations were carried out to define the best application scenario for the Continuous Annealing Process Line at Port Talbot. The outcomes indicated that the inserts could achieve 2-5% gas reduction depending on the tube location inside the furnace where the inserts were placed. Advices were given to apply the inserts to the area with a high installed heating capacity, operating constantly at a high burner load and seeing a high strip temperature. Based on this, the plant has decided to install more inserts to the strip exit end of the annealing furnace.

Radiant tube temperature profile

400

600

800

1000

0 2000 4000 6000 8000

Tube length [mm]

Met

al te

mpe

ratu

re [°

C]

no insertwith inserts

100% burner load

50% burner load

insert affected length

Figure 1: Insert effect on radiant tube temperature

59

Abstracts of posters

60

Tabulated Chemical Kinetics for Efficient and DetailedSimulations of Diesel Engine Combustion

C. Bekdemir1∗, L.M.T. Somers1, L.P.H. de Goey1, J. Tillou2, C. Angelberger21Department of Mechanical Engineering, www.combustion.tue.nl

Eindhoven University of Technology, The Netherlands2IFP Energies Nouvelles, Rueil Malmaison, France

A study will be presented on the application of a tabulated chemistry technique to model fuel spray igni-tion. The objective is to predict correct spray combustion physics. The FGM approach is applied to takeinto account detailed chemistry. This chemistry tabulation technique has been applied to RANS simula-tions of sprays and engine combustion before and proved to be effective in predicting the correct macro-scopic behavior [Bekdemir et al. PROCI 33 (2011) 2887-2894, Bekdemir et al. SAE 2010-01-0358].However, to gain more insight in the effect of for instance the way the chemistry table is generated,different engine operating points, used reaction mechanism on the ignition process and the flame struc-ture (premixed, partially-premixed, non-premixed), one needs spatially better resolved simulations. LEScan give such detail and becomes increasingly feasible for these applications. In this study, tabulatedchemical kinetics are applied to Large-Eddy Simulations (LES) of diesel spray combustion.

Numerical codeThe code that has been used in this study is AVBP (which is owned by CERFACS and IFP EnergiesNouvelles). AVBP is a parallel CFD code for reactive unsteady flow simulations on hybrid grids. It canhandle two-phase flows with an Eulerian formulation and has an injection model specially developedfor full cone spray simulations. The injection model relates injector parameters to two-phase boundaryinflow conditions that apply typically 10 times the nozzle orifice diameter downstream of the nozzle exit[Martinez et al. Fuel 89 (2010) 219-228].

FGM: tables and implementationFlamelet Generated Manifolds are pre-processed chemistry tables that contain information from de-tailed chemistry (DNS) simulations of 0D or 1D model problems. From experiments it can be con-cluded that non-premixed combustion is a major process in direct injection combustion. Therefore,in this analysis, 1D counterflow diffusion flames have been solved with our in-house code CHEM1D[www.combustion.tue.nl] to generate the FGMs. Unsteady diffusion flames have been used additionallyto steady diffusion flames at a broad range of strain rates (0.1 < a < 500). The igniting flame has beensolved for a single strain rate value of 500 [1/s]. All needed quantities (may be chemical source terms,species mass fractions, thermodynamic properties etc.) are stored as function of mixture fraction Z andprogress variable Y (here defined as the sum of CO2, CO and HO2).

Table 1: Spray H conditions: fuel = n-heptane, dnozzle = 100 µm, pinj = 150 MPa, ρ = 14.8 kg/m3, Tfuel = 373 K[www.sandia.gov/ecn].

case O2 [vol%] and Tamb [K]1-4, 8 0, 10, 12, 15, 21 vol% at 1000 K5-9 800, 850, 900, 1000, 1100 K at 21 vol%

All laminar CD-flame simulations have been performed with the Andrae and TRF reaction mecha-nisms. The Andrae mechanism [C&F 155 (2008) 696-712] contains 633 reactions among 137 species.The TRF mechanism is based on the n-heptane mechanism of Peters et al. [C&F 128 (2002) 38-59].The size is reduced leading to a mechanism with 248 reactions among 48 species.

During the LES simulations transport equations for Z and Y are solved, which are used to look-updesired quantities from the FGM table.

ResultsIn this section, the simulation results of Sandia’s spray H cases are compared to experimental data usinga Dirac δ assumption for the subgrid distribution of Z and Y. All simulations have been performed on

∗Corresponding author: [email protected]

COMBURA 2012

1

61

an unstructured grid with a mesh size of 80 µm at the nozzle exit that gradually increases to 800 µmdownstream. The case conditions are summarized in Fig. 1.

Ignition delay in experiments is defined at rapid pressure rise marking the beginning of second-stagehigh temperature chemistry. This point is not easily identified from simulations. Therefore, the ignitiondelay from simulations is defined at the steep increase of maximum temperature in the domain, whichis a sharp mark in temporal space. Lift-off length in experiments is defined at excited-state OH (OH*)which is found in high-heat-release regions. Since we do not have simulated OH* data readily available,a comparison has been done on basis of the temperature rise in the domain, which is also a measurefor heat-release.

10 12 15 210

0.2

0.4

0.6

0.8

1

1.2

1.4

ambient oxygen content [mole% O2]

igni

tion

dela

y [m

s]

measurementLES simulation: TRF (48 spec)LES simulation: Andrae (137 spec)

Figure 1: LES ignition delay time as a function of am-bient oxygen content with 2 reaction mechanisms.

800 850 900 1000 11000

0.5

1

1.5

2

2.5

ambient temperature [K]

igni

tion

dela

y [m

s]


Figure 2: LES ignition delay time as a function of am-bient temperature with 2 reaction mechanisms.

10 12 15 210

10

20

30

40

50

ambient oxygen content [mole% O2]

flam

e lif

t−of

f len

gth

[mm

]


Figure 3: LES flame lift-off length as a function of am-bient oxygen content with 2 reaction mechanisms.

800 850 900 10000

10

20

30

40

50

ambient temperature [K]

flam

e lif

t−of

f len

gth

[mm

]


Figure 4: LES flame lift-off length as a function of am-bient temperature with 2 reaction mechanisms.

Main characteristics, spray penetration depth, ignition delay time (Fig. 1 and 2), and flame lift-offlength (Fig. 3 and 4), are predicted very well, for a variation in ambient oxygen content and temperatureusing the δ approximation for subgrid chemistry. The studied reaction mechanisms perform similarlywell, except for the flame lift-off in the 800 K case, where the Andrae mechanism outperforms the TRFmechanism. Both mechanisms over-predict the measured ignition delay for this case.

DiscussionThe choice for subgrid chemistry closure for the used mesh seems to be important for the high reactivecases, especially for the stabilization. Therefore, the Dirac δ function, β-PDF, top-hat filter, and hybridtop-hat - δ filters have been compared a-priori. The β and top-hat shapes are substantially different,leading to large differences in chemical source term predictions. Filtering the chemical source term,which is narrow in Z-space, with a top-hat filter reduces its value up to 50%. And with the hybridmethod, the δ contribution at Z = 0 lowers the source so much that the spray does not ignite. LES withβ-PDF have been run on four cases. Although the ignition is retarded, it stabilizes the flames that didnot with the δ filter.

Overall, this study shows that LES of diesel spray combustion is feasible and predictive thanks totabulated chemical kinetics of igniting laminar diffusion flames (FGM), and offers the opportunity toproceed towards LES of full diesel engine combustion.

2

62

Liquid fuel burner for the measurement of adiabatic laminar burning velocities

R. Haas-Wittmüß, S. Cencin, R.T.E. Hermanns

OWI Oel-Waerme-Institut GmbH - Affiliated Institute of RWTH Aachen

Abstract

The heat flux burner is a research burner used to measure the adiabatic laminar burning veloci-ty SL. The laminar burning velocity is an important parameter of a fuel-oxidiser-mixture. Compu-tational fluid dynamics which include combustion processes rely on the accuracy of the laminar burning velocity. The TU Eindhoven (NL) redesigned the burner [1, 2, 3] developed by Botha and Spalding [4] to measure the laminar burning velocities of different fuels. The heat flux burner stabilises an adi-abatic, laminar, and flat flame on a burner plate. This flat flame approximates the ideal one-dimensional flame, necessary for the measurement of the laminar burning velocity. The heat flux burner mainly consists of a plenum chamber and the burner plate. The plenum chamber conditions the entering gaseous fuel-oxidiser mixture to a temperature above dew-point. The burner plate is heated by a fluid flow to 65 °C above the plenum chamber. The holes in the burner plate are 0.5 mm in diameter with a 0.7 mm diagonal pitch. Some holes are closed, as thermocouples are inserted to measure the temperature distribution across the sur-face. In a state, when the heat release of the preheated burner plate to the unburnt gas com-pensates the heat loss of the flame, a homogeneous temperature profile appears. This indicates the adiabatic state of the flame. By variation of the gas velocity different temperature profiles across the burner plate can be measured. The results are interpolated to zero heat loss, deliver-ing the adiabatic laminar burning velocity SL. Several research groups (i.e. TU Eindhoven (NL), University Lund (S), University Erlangen (D), TU Freiberg (D), and University Nancy (F)) examined different fuels in varying conditions (e.g. [5, 6]). At the OWI the focus of research is the conversion of liquid fuels, therefore a heat flux burner is adapted to operate with these. The vaporiser unit at the OWI is designed to overcome the current limitation to a fuel boiling temperature of 200 °C. With this setup liquid fuels up to middle distillates are to be investigated.

Literature

[1] Boschaart, K.J. and de Goey, L.P.H.: Detailed analysis of the heat-flux method for measuring burning velocities, Combust. Flame 132 (1-2), 2003, p. 170 - 180

[2] Boschaart, K.J. and de Goey, L.P.H.: The laminar burning velocitiy of flames propagating in mixtures of hydrocarbons and air measured with the heat flux method, Combust. Flame 136 (3), 2004, p. 261 - 269

[3] Van Maaren, A.: One-step chemical reaction parameters for premixed laminar flames. Ph.D. thesis, Technische Universiteit Eindhoven, 1994

[4] Botha, J.P. and Spalding, D.B.: The laminar flame speeds of propane/air mixtures with heat extraction from the flame, Proc. R. Soc. Lon. ser-A 255(1160), 1954, p. 71 – 95

[5] Naucler, J.D., Christensen, M., Nilsson, E.J.K, Konnov, A.A.: Laminar burning velocities of C2H5OH + O2 + CO2 flames, European Combustion Meeting 2011 Cardiff Van Lipzig, J.P.J., Nilsson, E.J.K., de Goey, L.P.H., Konnov, A.A.: Laminar burning velocities of n-heptane, iso-octane, ethanol and their binary and tertiary mixtures, Fuel 90, 2011, p. 2773 - 2781

63

Abstract COMBURA 2012 Towards LES simulation of flameless combustion: diluted homogeneous reactors (DHR) tabulation applied to Sandia Flames D and F

C. Locci*, O. Colin, J-B Michel

IFP Energies nouvelles 1 à 4 avenue du Bois Préau

92 800 Rueil-Malmaison, France Flameless combustion [Wunning 1996] represents nowadays an attractive technology to increase efficiency and reduce thermal NOx in furnaces. FC conditions are achieved by recirculating burnt gases and increasing the oxidant temperature. This results in a more diffused reactive zone and a more homogeneous radiative flux. CFD techniques are also required in order to develop this new technology and represent an useful tool to decrease R&D costs. In recent years, CFD research mainly focused on RANS techniques as they offer a good compromise between prediction accuracy and a relatively low CPU time compared to LES. On the other hand, RANS is known to poorly reproduce turbulent mixing which is a key parameter in flameless combustion. A more reliable tool for turbulence is therefore represented by LES turbulence solvers, even though more expensive from a computational point of view. Within this context, this work presents a new combustion recently developed for turbomachines [KIAI project 2011] and adapted here to flameless combustion. The model call Diluted Homogeneous Reactor (DHR) ,follows a tabulated approach and it is based on the tabulation of diluted homogeneous reactors. Tabulation is built by diluting fresh gases with hot burnt gases at different enthalpies. The subsequent auto-ignition is tabulated in a look-up table as a function of the initial parameters of the homogeneous reactors: mixture fraction mean and variance, progress variable, enthalpy. Tabulation allows to maintain a low CPU cost while allowing the use of complex chemistry, which is essential to accurately predict auto-ignition at low temperatures. Therefore this model allows to accurate predict auto-ignition and enthalpy losses, which are two key factors in flameless combustion. The model is validated on the two SANDIA flames D and F. A very good agreement is found for the Flame D mixture fraction field and a good accordance for temperature. Flame F also presents results in accordance with experimental data but a lack of precision is found for temperature in high strain rate zones. This deficiency is explained by the absence of strain description in the model. For NOx modelling the model of Ihme and Pitsch [Ihme 2008] is retained. It is observed that for the present adiabatic calculations, NOx are overpredicted compared to experimental data, but are in accordance with previous predictions shown in other works [Ihme 2008,Zoller 2011]. Introduction of radiations should lead to a better agreement.

• Corresponding author: [email protected]

64

Lean premixed combustion is introduced as an emission reducing technology for modern gas turbine engines. However, lean premixed combustion is more susceptible to thermo-acoustic instabilities, which occur due to a coupling between the unsteady flame heat release and the acoustic field inside the combustion chamber. A self-excited high amplitude limit cycle oscilla-tion can be triggered and its hazardous amplitude of vibration can cause mechanical failure in a fraction of time.

A poster is presented about a part of a master’s thesis work on thermo-acoustic instabilities. This master’s thesis is performed in the framework of the LIMOUSINE project, in which thermo-acoustic instabilities are studied in a multidisciplinary approach. A part of the experimental research is being done with the DESIRE combustor. One goal of the presented research is to ob-tain a numerical tool that can predict the thermo-acoustic behavior of the DESIRE combustor.

A numerical model is therefore established to predict the thermo-acoustic instabilities. Results of transient simulations show good agreement with experimental data. The hazardous frequen-cies are predicted with an accuracy of 3.3%. Amplitudes are less well predicted. This study has furthermore identified the relevance of acoustic boundary conditions and the influence of the Courant-Friedrichs-Lewy number.

Key words: Combustion, thermo-acoustics, numerical modelling, limit cycle

Numerical modelling of self-excited thermo-acoustic instabilities in the DESIRE combustor

E. Özcan[a][b], M. Kapucu[a], J.B.W. Kok[a]

[a] Laboratory of Thermal Engineering, University of Twente, P.O.Box 217, 7500 AE Enschede, The Netherlands[b] Twente Centre for Studies in Technology and Sustainable Development (CSTM), University of Twente, P.O.Box 217, 7500 AE Enschede, The Netherlands

65

Flame saturation current as a measure of the flame thermo-acoustic response.

L.B.W. Peerlings, M. Manohar, V.N. Kornilov, L.P.H. de Goey Combustion Technology, Eindhoven University of Technology, P.O. Box 513, 5600, the Netherlands.


Introduction

Manufacturers are in need for a cheap and robust detector of thermo-acoustic instabilities for active control purposes. This study explores the idea to use the ion concentration in flames as an indicator for thermo-acoustic instabilities. The thermo acoustic response of flat burner stabilized flames is characterized by flame transfer functions. The emitted OH* chemi-luminescence and the total ion concentration are used as a measure for the heat release and the results by both indicators are compared for various equivalence ratios (𝜙) and gas speeds (��). The thermo acoustic response of the flames is characterized using flame transfer functions in the frequency domain. The flame transfer function is defined as the ratio of the relative heat release perturbation to the relative velocity perturbations

𝑇𝐹(𝑓) ≔𝑞′(𝑓)/𝑞�𝑣′(𝑓)/��

The heat release is measured using OH* chemi-luminescence. The thermo acoustic response measured using the saturation current is

characterized in the same manner:

𝑇𝐹𝑖𝑠(𝑓) ≔𝑖𝑠 ′(𝑓)/𝚤𝑠�𝑣′(𝑓)/��

The saturation current is a measure of the total flame ion concentration

Experimental setup

The setup used to determine the thermo-acoustic response of flat premixed flame at atmospheric pressure is given in Figure 1. The equivalence ratio and gas speed of the incoming mixture are regulated using calibrated mass flow controllers. An air cooled electrode is placed above the flame parallel to the burner deck. The electrode consists of small hollow tubes through which the air coolant flows. The electrode is placed 10mm above the burner deck. The electric field is established by grounding the burner deck and attaching the air cooled electrode to a high voltage power supply. The applied electric field will create a body force upon the charged particles in the direction of the field lines. The charged particles will drift towards both electrodes and at the electrodes their charge will be neutralized. This results in a current flow between

Figure 1 Schematic representation of the experimental setup used to measure the thermo-acoustic response of flat flames

66

http://www.combustion.tue.nl/

the electrodes. If the field strength is large enough, all charged particles will be collected at the electrodes and the so called saturation current will flow. The saturation current is a measure of the total ion concentration in the flame. The thermo-acoustic response is measured by applying small flow oscillations (𝑣′/ 𝑣� ≈ 5%) to the gas flow using a loudspeaker connected to the gas supply hose. The resulting gas velocity fluctuations are measured using a constant temperature hot wire probe. The emitted OH* chemi-luminescence of the whole flame is recorded using a photo-multiplier tube equipped with an OH* filter (309nm narrowband). The oscillations in current are not easily measured because the air cooled electrode is at high voltage. Therefore the device measuring the current cannot directly be attached to the DAQ. To solve this problem a current probe is developed. This device uses an opto-coupler to optically transmit the measured current signal and insulate the high voltage line from the measurement line. The gain of the transfer function is determined by computing the amplitude spectrum of the signals using FFT. The computed gain will have an arbitrary value, because the relation between the sensor response [V] and the measured quantity in physical units is not explicitly known. The gain should equal one in the quasi steady state (𝑓 → 0), therefore the gain is signified by smoothly extrapolating the measured gain function to 𝑓 = 0. The phase of the transfer function is determined using cross correlation.

Results

In Figure 2 the thermo acoustic response of the flame determined using OH* chemi-luminescence and the saturation current as a measure of the flame response is given. The results show a good correspondence between the phase measured using OH* chemi-luminescence and the saturation current. A good correspondence in gain for both indicators of the flame response is found for lean flames. For near stoichiometric and rich flames a difference is observed in the measured gain. Additional measurements have been performed which indicate that the origin of the difference lies in the OH* chemi-luminescence intensity emitted from the secondary combustion zone.

Conclusion

The results using the saturation current as an indicator of the flame response are comparable with OH* chemi-luminescence measurements especially for lean flames. Therefore the total ion concentration can be used as an indicator of thermo-acoustic instabilities as it can accurately measure the gain and phase of the flame response at the frequencies of practical importance.

Acknowledgement

This study is supported by STW and part of project 10430B ''Suppression of thermo-acoustic instabilities in central heating equipment by burner design optimization and active control''

0 100 200 300 400 500 6000

0.5

1

1.5

2

2.5

3

Gai

n [-]

Frequency [Hz]0 100 200 300 400 500 600

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

Phas

e shi

ft /π

[rad

]

Frequency [Hz]Figure 2 Flame transfer functions of flat flames. Lines 𝑻𝑭, Symbols 𝑻𝑭𝒊𝒔.

Blue: 𝝓 = 𝟎.𝟖𝟎, 𝑣� = 𝟐𝟐.𝟓cm/s, (𝑺𝑳 = 𝟐𝟓.𝟖cm/s). Red: 𝝓 = 𝟎.𝟖𝟓, 𝑣� = 𝟐𝟔.𝟏cm/s, (𝑺𝑳 = 𝟐𝟗.𝟔cm/s). Green: 𝝓 = 𝟏.𝟎𝟓, 𝑣� = 𝟑𝟓.𝟏cm/s, (𝑺𝑳 = 𝟑𝟕.𝟐cm/s). Yellow: 𝝓 = 𝟏.𝟐𝟎, 𝑣� = 𝟑𝟏.𝟓cm/s, (𝑺𝑳 = 𝟑𝟐.𝟕cm/s)

67

Numerical and experimental study of ethanol combustion in an industrial gas turbine

J.L.H.P. Sallevelta, A.K. Pozarlika, M. Beranb, L.-U. Axelssonb, G. Brema

a Department of Thermal Engineering, University of Twente, 7522 NB Enschede, The Netherlands b OPRA Turbines B.V., 7554 TS Hengelo, The Netherlands

A substantial part of the power and heat in the world is generated by gas turbines. The current development towards cleaner and more sustainable energy production leads to a growing interest in gas turbines fired by non-conventional fuels, which demands for combustors with high fuel flexibility. The application of alternative fuels in gas turbines has a promising future, although extensive research is needed to get insight into the combustion quality when used in conventional or newly designed combustion chambers. In this study, the application of ethanol as a biomass-derived fuel in OPRA's 2 MW class OP16 radial gas turbine has been studied using experimental and numerical methods.

A modified OP16 combustor has been mounted in a test rig to conduct ethanol burning experiments at atmospheric pressure. This reverse-flow tubular combustor is suitable for both conventional and non-conventional fuels and operates in a diffusion mode. Exhaust gas temperature and emissions (CO, CO2, O2) have been measured for different operating conditions of the combustor. Furthermore, temperature data from the liner has been acquired by applying several types of thermochromic paint. The temperature profile indicated by the paint layers is used as a boundary condition in the model to account for heat losses through the liner.

The combustion tests have been simulated in ANSYS FLUENT for three different operating conditions considered in the experiments. Taking into account periodicity, the calculations have been performed using a 1/8th slice of the combustor geometry. The simulations are based on the RANS approach in combination with the SST k-ω turbulence model. Fuel enters the computational domain in the form of droplets, which are tracked as discrete particles while they interact with the gas phase. A flamelet library has been constructed based on non-adiabatic, steady flamelets to describe the chemical reactions taking place in the flame.

Finally, the simulation results have been compared with experimental data. The current study therefore indicates the performance of the industrial combustor when fired on ethanol and shows the predictive capabilities of the chosen CFD models. Furthermore, the numerical work can serve as a basis in the development of a model to simulate pyrolysis oil combustion, which will be the next step in the project. This research is part of the BE2.O program sponsored by the Province of Overijssel.

Ethanol combustion in the modified OP16 combustor.

68

Development of a Compact Combustion Chamber for Liquid Fuels with High Energy Densities

F. Schlösser, S. Hackhofer, R.T.E. Hermanns OWI Oel-Waerme-Institut GmbH

Affiliated Institute of RWTH-Aachen

Abstract

In the market segment of combined heat and power of liquid fuels, systems based on gas turbine pro-cesses can be applicable as well to the stationary as to the mobile sector. With regard to power densi-ty and system weight gas turbine systems have advantages over internal combustion engines or fuel cells. The present study aims at a combustion stabilisation according to the principle of Lean-Prevaporised-Premixed-Combustion (LPP). Due to the high massflow the time available for the fuel to vaporise and mix with the combustion air is limited to a timeframe smaller than one millisecond, which is close to the ignition delay time of the mixture [1]. The stabilisation of the reaction zone is reached through a swirl-ing flow stabilisation, leading to minimal pressure drop. In the present study the radial swirler is opti-mised to a level, which stabilises the reaction without causing re-ignition and leading to low emissions. According to the definitions mentioned by Beer [2] the global swirl numbers SN are between SN =0,5 (subcritical) und SN =1 (supercritical). The applied reactor wall is made of quartz glass which gives the opportunity to visually assess the flame stability during the operation. Additionally, emissions of CO and NOx are being monitored, since both of them can be applied to estimate the performance of the combustion process. This gives in the end, according to the definitions of Walsh [3], atmospheric com-bustion intensities of 50 MW atm-1 m-3 and combustor loading of 10 kg s-1 atm1,8 m3, whereby the emissions of carbon monoxide and nitrogen oxide could stay relative low.

References

[1] Schmitz, I.; Magere vorgemischte Verbrennung von leichtem Heizöl in Mikrogasturbinen, Sha-ker Verlag, ISBN-978-3-8322-8292-9, 2009

[2] Beer, J.M.; Combustion Aerodynamics, Applied Science Publishers LTD, ISBN 0-855334-513-9, 1972.

[3] Walsh, P,: Gas Turbine Performance, Blackwell Science Ltd, ISBN 0-632-06434-x,

69

INTROD

Gas turbi

other kin

specific p

kinds of f

continuo

modern

combusti

prevent t

or CO, th

loop b

aerodyna

is a limit

cause c

elements

the trans

The feed

participat

other du

cycle in a

must be

aerodyna

related t

difficult t

the geom

combusti

combusto

This work

Juan Carl

DUCTION

ne engines pr

nds of power

power, they

fuel and, can

us effort to

gas turbine

ion technolog

the formation

he flames is p

between he

amics and pre

t cycle of pre

racks and d

s of the turbin

ition pieces.

Figure 1: The

dback loop

ting different

ring the limit

an unstable c

broken at a

amics of the

to the burne

to change, th

metry and b

ion chamber.

or may chan

k was already

os Roman C

LaFa

resent many a

plants. They

are able to

be very effici

o reduce po

engines use

gies. While t

of pollutant s

prone to estab

eat release

essure waves

essure oscilla

damage to

ne, such as th

thermoacousti

(Figure 1) sh

elements int

t cycle. To pr

ombustor the

any of the li

fluid and he

er, and there

e acoustic fie

boundary con

Changes of th

nge the comb

y studied in a

Casado

aboratory of Taculty of Eng

UniEnsche

Email: j.c.ro

advantages ov

have very hi

burn differe

ient. Due to t

llutant speci

lean premix

this technolo

species like N

blish a feedba

fluctuatio

[1]. The resu

ations that m

the structu

he liner walls

ic loop

hows how t

teract with ea

revent the lim

e feedback lo

inks. While t

eat release a

efore., they a

eld is related

nditions of t

he length of t

bustion regim

previous pap

Thermal Enggineering Teciversity of Twede, The Netomancasado

ver

igh

ent

the

es,

xed

ogy

Ox

ack

ns,

ults

may

ural

or

the

ach

mit

oop

the

are

are

to

the

the

me.

per

of

LI

EX

ge

un

LI

m

on

lo

co

le

fla

w

gineering (Thchnology (CTwente therlands [email protected]

f the autho

MOUSINE com

XPERIMENT

The experim

eneric burner

nder the Eur

MOUSINE. T

methane and a

The flame

n the wake o

ocated betwe

ombustion ch

ngth can be t

ame is optical

indows.

Figure 2

an

J.B.W. K

hW), TW)

l

rs on a ea

mbustor[2] .

TAL SET UP

mental resear

, developed a

opean Union

he combusto

ir up to 70 kW

is partially pr

of an equilat

een the ple

hamber. The

tuned by addi

ly accessible t

: LIMOUSINE co

nd double liner

Kok

arlier version

rch was carrie

at University o

n Commission

or can burn m

W

remixed and

teral triangula

enum duct

e combustor

ing one extra

thanks to 4 qu

ombustor with

r configurations

n of the

ed out in a

of Twente

n program

mixtures of

stabilized

ar wedge,

and the

chamber

liner. The

uartz glass

h single

s

70

This c

prevent t

include

infinite s

self exite

intensifie

to record

narrow si

RESULT

The com

operating

High am

were pre

condition

unstable

configura

instability

pressure

thermal p

Figure 3 P

and air f

The freq

Helmholt

Where L

average s

same He

instability

aerodyna

body.

combustor has

the losses of

pressure tra

ide hoses, ph

ed OH* emis

ed CCD camer

d pictures o

ide.

TS AND DISC

bustion test

g condition in

mplitude limit

esent in both

ns range in

was reduc

ation. The m

y were lowe

spectrum in

power and air

Pressure in the

factor 1.20. Bla

dash lin

quencies we

tz number in e

L is the lengt

speed of sou

e number,

y to acousti

amic perturba

s special air a

acoustic ener

nsducer mou

hoto multiplie

ssion and the

ra with a CH*

f the flame

CUSION

were recorde

both burner.

cycle press

h combustors

which the

ced in the

easured freq

er as well. T

the downstr

r factor λ = 1.2

combustion ch

ack solid line: S

ne: double liner

ere normaliz

equation 1.

th of the co

nd. The main

and therefo

c phenomen

ations in the w

nd fuel inlets

rgy. The senso

unted in sem

er tube for t

ermocouples.

filter was us

also from t

ed for differe

ure oscillatio

. The operati

combustor

double lin

quencies of t

The shows t

eam for 50 k

2.

hamber for 50 k

Single liner. Red

r

zed using t

(1)

ombustor and

n peaks has t

ore, relate t

na and not

wake of the bl

to

ors

mi‐

the

A

sed

the

ent

ons

ing

is

ner

the

the

kW

kW

d

the

)

d c

the

the

to

uff

FU

Th

th

bo

In

is

Th

of

Ac

FP

pr

B

1.

2.

UTURE WO

he work in p

he heat releas

oth combusti

teraction bet

also being inv

he authors wo

f this researc

ctions – Netw

P7‐PEOPLE‐20

roject number

IBLIOGRAP

Lieuwen, combustioturbine cofor Gas Tuthe Asme,

J.C. RomaJ.B.W.Kostudy of thcombustioJaneiro.

RK and ACK

rogress includ

se spectra an

on chamber

ween pressur

vestigated.

ould like to ac

ch by the E

works for Init

007‐1‐1‐ITN, P

r 214905

HY

T., et al., A mon instability iombustors. Jouurbines and Po 2001. 123(1)

an Casado, P.Rk., Experimen

he effect of acoon stability in

KNOLEDGEM

des the comp

nd the flame

length. Fluid

re waves and

cknowledge th

C in the Ma

tial Training,

Project LIMOU

mechanism of in lean premixurnal of Enginower-Transac): p. 182-189. R.Alemela anntal and numeoustic time de

n ICSV 2011: R

MENTS

parison of

shape for

Structure

liner walls

he funding

arie Curie

under call

USINE with

xed gas neering ctions of

d erical elays on Rio de

71

NUMERICAL SIMULATION OF STABLE/UNSTABLE COMBUSTION IN A BLUFF-BODY STABILIZED COMBUSTOR Santosh Kumar and J.B.W Kok Department of Thermal Engg, University of Twente, The Netherlands e-mail: [email protected]

In the present study the impact of two different combustion models on prediction of limit cycle frequencies in a model combustor are analysed. CFD analysis is carried out on a partially premixed bluff body stabilized methane-air combustor, which is self-excited. Cost-effective URANS simulations are performed with SAS-SST turbulence model and two combustion models - (1) BVM model (also known as Turbulent Flamespeed Closure model) and (2) CFI model developed at the University of Twente. The results show that the BVM model predicts unstable combustion whereas the CFI model is stable at the same operating condition. The pressure spectrum of the BVM simulation shows two distinct peaks, the base frequency corresponds to the longitudinal acoustic mode of the downstream duct of the combustor and the second frequency is assumed to be due to non-linear effects in the coupling between combustion and acoustics of the system. To validate results the numerical source terms are compared to OH* chemiluminescence data from experiments. It is seen that the compactness of the BVM source leads to instability. Also acoustic pressure and velocity fields are reconstructed to get impedances at the burner plane and outlet. For unstable combustion the values of impedances matched with experimental findings. The analysis further

suggests that the upstream and downstream sections resonate at different frequencies. The reflection coefficient calculations at the burner plane show the decoupled behaviour of the combustor.

Results

Transient CFD analysis is used to study the combustion dynamics pressure fluctuations in the combustor. The predicted time series are analysed to find the instability frequencies. In Figure 6 the pressure spectrum obtained from transient numerical simulations is shown. The data shown here is at the same operating point but with two different combustion models. The BVM model predicts a unstable combustion at this operating point of 40 kW and λ = 1.60 and the CFI model predicts fairly stable conditions

72

with maximum amplitudes of 55 [Pa]. The pressure spectrum indicates two peaks f1=347 Hz and f2=695 Hz. The first peak f1 is assumed to be related to the quarter wave mode of the downstream section of the model combustor. This has been verified by 1D acoustic modelling of the duct taking into account the change in local speed of sound with temperature. The second frequency f2 is twice the f1 and is produced due to nonlinear flame dynamics. The frequency of f1 is expressed as the acoustic impedance at the flame-holder plane is too high and the downstream duct is decoupled from the upstream duct. The calculated impedance value (ρc) at the burner plane in the slots is 277 [kg/m2s]. Therefore the acoustic fluctuations in this case are limited to the downstream duct only hence frequency of f1.

Conclusions

A simple model combustor is used to analyse the combustion instability mechanism. Detailed CFD analysis is carried out with the ANSYS-CFX package and with the help of industrial standard and university developed combustion models (BVM and CFI model). The high resolution SAS-SST model is used for resolving the flow field. The steady state analysis shows that the BVM

model has overestimation of the source term. It predicts a higher rate of heat release and the heat release is concentrated near the bluff-body. The source term predicted by the CFI model is diffused in the wake of the bluff-body and as a consequence combustion dynamics is suppressed. Mean measured OH* chemiluminescence data pictures in stable conditions match closely with the data from CFI

73

model. Under transient conditions the effects of the source terms from the combustion models is seen. Due to compactness of the numerical source the BVM model simulations reaches LCO with high amplitude pressure fluctuations. The CFI model however predicts a stable combustion process, with very low amplitude pressure fluctuations. This stable behaviour comes from the spatially distributed source term, which has very small acoustic time scales and cannot cause

feedback. Thus the source term predictions are a critical process. The BVM model can be improved in this matter by having a better turbulent-chemistry database and probably LES simulations for accurate estimations of the mixing. The reflection coefficients at the combustor exit and flame holder locations further suggest that the downstream section is decoupled during LCO.

74

List of participants Combura 2012 3 October - 4 October 2012

75

S.E. Abtahizadeh Eindhoven University of Technology P. R. Alemela Ansaldo Thomassen B. Amini Ghent University J E.Anker NUMECA Int. L.D. Arteaga Mendez Delft University of Technology P.C. Bakker Eindhoven University of Technology R.J.M. Bastiaans Eindhoven University of Technology N.A. Beishuizen Bosch C. Bekdemir Eindhoven University of Technology K. Boonstra GDFSUEZ Energy Europe T.W.F.M. Bouten University of Twente G. Brem University of Twente R.T.F.W. Callaars BRACE Automotive B.V. T.Cardoso de Souza Eindhoven University of Technology E.S. Cho Delft University of Technology O.C. Colin IFPEN M.M.H. Corvers Eindhoven University of Technology J.P. van der Dennen CelSian Glass & Solar A. Donini Eindhoven University of Technology A. Fancello Eindhoven University of Technology

S. Gersen DNV KEMA L.P.H. de Goey Eindhoven University of Technology E.M. Gucho University of Twente R. Haas-Wittmuess OWI Oel-Waerme-Institut GmbH T. Hendriks Hendriks Engineering R. Hermanns OWI Oel-Waerme-Institut GmbH E. Hernandez Perez Eindhoven University of Technology D.P.L.F. Jansen Ansaldo Thomassen B. Johansson Eindhoven University of Technology W. de Jong Delft University of Technology M. Kapucu University of Twente J.B.W. Kok University of Twente J.N.A. Koomen Stork Technical Services V.N. Kornilov Eindhoven University of Technology L.J. Korstanje Technology Foundation STW W.K. Krebs SIEMENS AG C.A.J. Leermakers Eindhoven University of Technology C.L. Locci French Institute of Petroleum / University of Cottbus J. Lu Delft University of Technology L. Ma Delft University of Technology

76

M. Manohar Eindhoven University of Technology T.H. van der Meer University of Twente S.M. Mir Najafizadeh Delft University of Technology K. Mishra BAM Berlin W.H.L. Moonen Tata Steel J.A. van Oijen Eindhoven University of Technology E. Özcan University of Twente L.B.W. Peerlings Eindhoven University of Technology A.J. Plomp ECN J.L. Raas Nuon/Vattenfall W.J.S. Ramaekers Dacolt - Combustion & CFD H.R. Rodrigues Delft University of Technology D.J.E.M. Roekaerts Delft University of Technology J.C. Roman Casado University of Twente E. Russo Eindhoven University of Technology J.L.H.P. Sallevelt University of Twente G. Sarras Delft University of Technology F. Schloesser OWI Oel-Waerme-Institut GmbH A.V. Sepman University of Groningen M. Shahi University of Twente

Y. Shoshin Eindhoven University of Technology L.M.T. Somers Eindhoven University of Technology N. Speelman Eindhoven University of Technology G.G.M. Stoffels University of Twente A. van der Stroom Technology Foundation STW V. Tarband University of Twente O.J. Teerling Bekaert Combustion Technology BV D. Treur Shell Global Solutions M.J. Tummers Delft University of Technology J.V. Vancoillie Ghent University B.C.H. Venneker Stork Technical Services A.A. Verbeek University of Twente J.F. Vervisch CORIA CNRS & INSA de Rouen M. Wankhede Dacolt International BV H. Wu Tata Steel J.G.W. Wuenning WS GmbH R.P.C. Zegers Eindhoven University of Technology S. Zhu University of Twente

77

Simulation of the Delft Jet-in-Hot-Coflow burner using transported ...

Documents

Transcript of Simulation of the Delft Jet-in-Hot-Coflow burner using transported ...