LABORATORY INVESTIGATION OF LOW SWIRL INJECTOR FOR LEAN PRE-MIXED GAS TURBINES


SEMINAR REPORT Submitted by
KIRAN .G.
S7M2
ROLL NO-30
Department Of Mechanical Engineering
College of Engineering, Thiruvananthapuram – 16
OCTOBER 2010

[attachment=7052]


ABSTRACT

Laboratory experiments have been conducted to investigate the fuel effects on the

turbulent premixed flames produced by a gas turbine low-swirl injector (LSI). The

lean-blow off limits and flame emissions for seven diluted and undiluted hydrocarbon

and hydrogen fuels show that the LSI is capable of supporting stable flames that

emit <5 ppm NOx (@ 15% O2). Analysis of the velocity statistics shows that the

nonreacting and reacting flowfields of the LSI exhibit similarity features. The turbulent

flame speeds, ST, for the hydrocarbon fuels are consistent with those of methane/air

flames and correlate linearly with turbulence intensity. The similarity feature and linear

ST correlation provide further support of an analytical model that explains why the LSI

flame position does not change with flow velocity. The results also show that the LSI

does not need to undergo significant alteration to operate with the hydrocarbon fuels but

needs further studies for adaptation to burn diluted H2 fuels.























INDEX OF CONTENTS

Title Page no

Title Page 1
Acknowledgements 3
Abstract 4
Index of Contents 5
Nomenclature 6
List of tables, figures 7
Introduction 8
Background 9
Experimental system and diagnostics 11
Results 14
4.1 Flame stability and lean blow-off 14
4.2 Flowfield analysis 17
4.2 (I) centerline profiles for non-reacting flows 17
4.2 (II) ).Radial profiles of the non-reacting flows 19
4.2 (III) Centerline profiles for reacting flows 20
4.2 (IV) Radial profiles of reacting flames 22
4.3 Turbulent flame speed 24

5.Conclusions 26

6.Scope 27

7.References 28












NOMENCLATURE

Symbols

ar – normalized radial divergence rate

ax - normalized axial divergence rate

Ls- Length of swirler section

Ø – Equivalence ratio

q’ – 2D Turbulent kinetic energy

R - ratio of the radii of the centerchannel and injecto
r
Rc - Radius of centerchannel

S –Swirl number

SL – Laminar flame speed

ST – turbulent flame speed

T0,P0,U0-Stagnation temperature , pressure ,velocity respectively

Tad – Stoichiometric adiabatic flame temperature

u’ – Linear flame velocity

v’ – Radial flame velocity

x0 - virtual origin of the divergent flow

α - Vane angle


Abbreviations

DLN : Dry Low NOx

LBO : lean blow-out

LSB : Low-swirl Burner

LSI : Low-swirl injector

PIV : Particle Image Velocimetry
LIST OF TABLES AND FIGURES

FIGURES PAGE NO

Fig 1:Schematics and photographs of the low-swirl injector 10
Fig 2.Experimental setup 11
Fig. 3. (a) LSI lean blow-off limits for natural gas at STP and elevated T0 and P0 15


Fig. 3.(b) LSI lean blow-off limits for fuels of Table 1 at STP. 15

Fig. 4(a). NOx emissions from LSI for the hydrocarbon fuels of Table 1. 16

Fig. 4(b)CO emissions from LSI for the hydrocarbon fuels of Table 1. 16

Fig. 5. Comparison of NOx data 17

Fig. 6. Centerline profiles of the non-reacting flows. 18
Fig. 7. Radial profiles of the non-reacting flows at x = 15 mm. 19

Fig. 8. Centerline profiles of eight flames with 9.2 <U0 < 9.5 m/s 21
Fig . 9. Radial profiles of eight flames with 9.2 < U0 < 9.5 m/s at x = 15 mm. 23

Fig. 10. Correlation of flame speeds measured from LSI and LSB 25

TABLES


Table 1 : List of properties for the fuels used in this study 13
Table 2 : List of experimental parameters for the fuels studied 25












1.INTRODUCTION


Power generation turbines operating on natural gas are subjected to stringent emission

rules and many urban areas have NOx requirements of <5 ppm ( corrected to 15% O2).

Recent research has led to development of effective control technologies based on lean

premixed combustion , such as catalytic combustors , trapped vortex combustors , and metal

fiber combustors. Low-swirl injector (LSI) provides another option that avoids altering

engine layout or operating cycle. As more mid-size turbines are deployed in locations

with readily available alternate fuels such as landfills , paper mills , and oil platforms ,

meeting emissions goals while using different fuels presents great challenges. This is due

to differences in combustion properties and their interactions with turbulence that affect

flame stability, emissions, and turndown performance . Our goal is to investigate the fuel

effects on turbulent premixed flames in the LSI to develop an engineering method to

adapt it to operate on alternate fuels. The approach is to investigate lean blow-out (LBO),

emissions and the flowfield characteristics to gain the fundamental insights for

optimizing the LSI for fuel-flexibility.















2.BACKGROUND


Lean premixed combustion is a proven dry low-NOx (DLN) method for natural gas-

powered turbines. Most DLN engines emit NOx < 25 ppm and CO < 50 ppm (both @ 15%

O2). But attaining ultra-low emissions of <5 ppm NOx requires that turbines operate at

conditions close to the lean blowoff limit (LBO) where combustors are susceptible to

combustion oscillations. Results at turbine conditions (500 < T0 < 730 K, 6 < P0 < 15 atm,

12 < U0 < 48 m/s) show that the LSI produces stable flames with NOx and CO below 2

ppm (@15% O2) at the leanest conditions. Further work has led to an LSI that has been

evaluated in a single cylindrical combustor and in a multi-injector annular combustor at

simulated engine conditions. The study showed that the LSI has good performance

characteristics, and is stable over a wide range of conditions where NOx < 5 ppm and CO

well below the acceptable limit of 400 ppm. The flame does not have a propensity to

become unstable towards blowoff or show undesirable injector-to injector interaction.

The heart of the LSI is a swirler evolved from atmospheric low-swirl burners. The

swirler section is 2.8 cm long (Ls), and has an outer radius of 3.17 cm and sixteen

curved vanes (vane angle α = 42° at the exit) attached to the outer surface of a Rc = 2 cm

centerchannel. The open centerchannel allows a portion of reactants to remain unswirled

and this nonswirling flow inhibits flow recirculation and promotes formation a divergent

flowfield, a key feature of the flame stabilization mechanism . To control the mass ratio,

m=mc/ms, between the flows through the centerchannel, mc, and the swirled annulus, ms, a

perforated screen is fitted at the entrance of the centerchannel.

The swirl number definition is:

S=2/3 tan⁡〖α (1-R^3)/(1-R^2+m^2 〖(1⁄(R^2-1))〗^2 R^2 )〗



Here the ratio of the radii of the centerchannel and injector, R, is 0.63 and the screen

blockage controls m and hence S. Fitted with a 58% blockage screen, this LSI has

S = 0.Otherwise, all dimensions e.g. exit tube length of li = 9.5 cm and a 45° tapered edge,

remain unchanged.








Fig 1:Schematics and photographs of the low-swirl injector














3.EXPERIMENTAL SYSTEM AND DIAGNOSTICS




Fig 2.Experimental setup

For the lean blowout (LBO) and Particle Image Velocimetry (PIV) investigations, the LSI

was mounted on a cylindrical settling chamber. Air (up to 1800 LPM) enters at the side of

a 25.4 cm diameter chamber and flows into the LSI via a centrally placed 30 cm long

straight tube. Air flow is adjusted by a valve and monitored by a turbine meter, and fuel

(Table 1) is injected in the air supply to ensure a homogeneous mixture for the injector.

Both the fuel and the PIV seeder flows are controlled by mass flow controllers and set

according to a predetermined value of Ø with a PC .The fuels listed in Table 1 consist

of hydrocarbons,N2 and CO2-diluted CH4 to simulate landfill and biomass fuels, H2-

enriched CH4 to simulate refinery gas and CO2-diluted H2. Variations in the combustion

properties are shown in Table 1 by the stoichiometric adiabatic flame temperatures, Tad,

and laminar flame speeds, SL. The Wobbe Index is used commercially as an indicator of

fuel . Emission measurements were performed with a Horiba PG-250 analyzer, calibrated

using 7.9 ppm NO in N2 and 31.8 ppm CO in N2 . The instrument has an accuracy of

±0.5 ppm for NOx. To measure emissions, flames were enclosed in a 16 cm diameter, 20

cm high quartz tube and sampled with a probe placed a few centimeters above the center

of the tube. The collected exhaust gas was cooled and water was removed with a

dessicant before it flowed into the analyzer . To facilitate PIV data collection, the non-

reacting flows and the flames were not enclosed. PIV system and data analysis has a

New Wave Solo PIV laser with double 120 mJ pulses at 532 nm and a Kodak/Red Lake

ES 4.0 digital camera with 2048 by 2048 pixel resolution. The optics were configured

to capture a field of view of 13 cm by 13 cm. A cyclone particle seeder seeds the air

flow with 0.3 µm Al2O3 particles. Data analysis was performed on the 224 image pairs

recorded for each experiment using software developed by Wernet. Using 64 • 64 pixels

cross-correlation interrogation regions with 50% overlap, this rendered a spatial resolution

of approximately 2 mm.














Table 1: List of properties for the fuels used in this study



Fuel composition
Tad at Ø = 1 K SL at Ø = 1 m/s Wobbe index
kcal/Nm
CH4 2230 0.39 11542
C2H4 2373 0.74 14344
C3H8 2253 9.45 17814
H2 2318 2.50 9712
0.5 CH4/0.5 CO2 2013 0.20 4182
0.6 CH4/0.4 N2 2133 0.31 6026
0.6 CH4/0.4 H2 2258 0.57 10130
0.5 H2/0.5 CO2 1693 0.56 1432





















4.RESULTS

Flame stability and lean blowoff


Flame stability and LBO were determined at volumetric flow rates 300 < Q < 1880 LPM,

corresponding to bulk flow velocities of 3 < U0 < 9 m/s. Figure 2a shows LBO data for

methane. The open flame data at STP are shown as the baseline. The data at higher inlet

temperatures and pressures (1–14 atm , 620–770 K) were obtained from enclosed

configurations simulating a gas turbine combustor and they show the lowest LBO occur

at heated atmospheric tests in a quartz rig . These data also show that the LSI can

operate up to U0 = 85 m/s, and that LBO remains relatively insensitive to U0. This is a

desirable feature for turbines for it indicates that the LBO will not edge closer to the

operating point of the combustor when the load increases. In Fig. 2b LBO values are

essentially the same for CH4, C3H8, 0.5 CH4/0.5 CO2, and 0.6 CH4/0.4 N2. The dilution of

CH4 by inerts has no observable effect on LBO. LBO is slightly lower for C2H4 and

0.6 CH4/0.4 H2, which have higher flame speeds than the other fuels. The LBO values

for H2 are very low and do not show a significant effect due to dilution. However, the

stability ranges for H2 fuels are limited because the flames tend to reattach to the burner

rim at Ø> 0.30. NOx and CO emissions from flames at Q = 1500 LPM (U0 = 7 m/s) are

shown in Fig. 3. Only data for the hydrocarbon fuels are plotted as emissions from H2

fuels were below detectable limits. For the hydrocarbon fuels, NOx has an exponential

dependence on Ø, and at a given Ø, emissions show a dependence on Wobbe Index,

consistent with the higher heat content of these fuels. However, the significant implication

of these data is that regardless of fuel content the LSI supports stable flames emitting <5

ppm NOx and the conditions are well above the LBO point. Flame temperature is an

important parameter in NOx formation in the LSI. The plot of NOx vs. Tad in Fig. 4

shows that NOx correlates well with Tad and is consistent with data at high T0, P0 and

U0. LSI flow has little or no recirculation, which explain why the NOx production

depends primarily on flame temperature.



Fig. 3. (a) LSI lean blow-off limits for natural gas at STP and elevated T0 and P0


Fig. 3.(b) LSI lean blow-off limits for fuels of Table 1 at STP.




Fig. 4(a). NOx emissions from LSI for the hydrocarbon fuels of Table 1.







Fig. 4(b)CO emissions from LSI for the hydrocarbon fuels of Table 1.







Fig. 5. Comparison of NOx data




4.2. Flowfield analysis


Table 2 shows the PIV experimental conditions consisting of three non-reacting flows

and sixteen flames. For hydrocarbon flames, their stoichiometries were set at the

conditions where NOx ≈ 5 ppm to compare them at the conditions that meet the

emission goals. For the diluted hydrogen fuels, flames at Ø = 0.25 and 0.30 were

studied.

4.2.(i).The centerline profiles for non-reacting flows

The centerline profiles for three non-reacting flows are compared in Fig. 5. Two

parameters were introduced to characterize the nearfield region. The first is the virtual

origin, x0, of the divergent flow, obtained by extrapolating the linear velocity decay

region downstream of the exit (Fig. 5a), and second is the slope of the linear extrapolation

that quantifies the normalized axial divergence rate, ax = dU/dx/U0. Values of x0 and ax

for the three flows are given in Table 2 and they are very close. Profiles of the

normalized 2D turbulent kinetic energy, q’ = ((u’² + v’²)½)/2 of Fig. 5b shows that within

the linear velocity decay region, turbulence along the centerline remains constant. These

characteristics can be attributed to the effect of annulus swirling flow. In the farfield, slight

increases in q’/U0 at x > 60 mm are consistent with the formation of a very weak

recirculating zone.






Fig. 6. Centerline profiles of the non-reacting flows.







4.2.(ii).Radial profiles of the non-reacting flows

Radial profiles of the non-reacting flows at x = 15 mm all exhibit similarity behavior. In

Fig. 6a, the U/U0 profiles have a flat central region corresponding to the centerchannel

non-swirling flow flanked by two velocity peaks, corresponding to the swirling flow. In

Fig. 6b, linear distribution of the V/U0 profiles within the center region (_15 < r < 15 mm)

show that the normalized radial divergence rates ar = dV/dx/U0 are about half that of ax.

The q’/U0 profiles (Fig. 6c) have relatively flat distributions in the center regions

surrounded by intense turbulence peaks. These velocity statistics show that the LSI

produces a uniform central region with low shear stresses for flame stabilization.



Fig. 7. Radial profiles of the non-reacting flows at x = 15 mm.

4.2.(iii).Centerline profiles for reacting flows


Centerline profiles for reacting flows of eight flames with 9.3 < U0 < 9.6 m/s are shown

for clarity. Despite the large difference in the farfield, all U/U0 of Fig. 7a have linear

velocity decays near the LSI exit. The positions where profiles deviate from linear decay

trends correspond to the leading edges of the turbulent flame zones. From these centerline

profiles, ax and x’ for the nearfield linear decay regions can be deduced. Results listed in

Table 2 show that the flames increase both ax and x0 to demonstrate an influence of the

flame on mean characteristics of the upstream reactant flow. For hydrocarbon flames, the

majority of the ax values are around -0.014mm¯¹ compared to ax = -0.085 mm¯¹ for

the non-reacting flows. For the diluted H2 flames, the increases in ax are smaller, averaging

-0.011 mmˉ¹ and their U/U0 profiles have different shapes than the hydrocarbon flames.

This seems to be associated with the lower heat release compared to the hydrocarbon

flames. Though the hydrocarbon flame profiles are consistent in the nearfield, their farfield

features show dependence on heat release. Significant flow accelerations are found only in

the C2H4 and 0.5 CH4/0.5 H2 flames, while other hydrocarbon flame profiles have

relatively flat distribution .The corresponding q’/U0 profiles of Fig. 7b show that the

fluctuation levels at the LSI exit are slightly higher than in the non-reacting flows.But

the anisotropic ratio u’:v’ remains unchanged. The q0/U0 levels remain relatively flat

through the flame brushes and the increases in the farfield at x > 80 mm corresponds to

flames that produce weak recirculation.





















Fig. 8. Centerline profiles of eight flames with 9.2 <U0 < 9.5 m/s

4.2.(iv).Radial profiles of reacting flames


Figure 8 shows radial profiles at x = 15 mm for flames of Fig. 7. These positions are below

the flame brushes so that the results can be compared with those of Fig. 6. Although the

U/U0 profiles in Figs. 8a and 6a have similar features, there are quantitative differences.

Within the central flat regions, U/U0 levels decrease to 0.5 for the two diluted H2 flames,

and 0.3 for hydrocarbon flames. These changes correspond to increases in ax and x’. The

center regions are also slightly wider than in the non-reacting flows. Another difference is

peak velocity in the surrounding swirl annulus increasing from U/U0 = 1.2 in non-reacting

flows to 1.5 in the flames. The V/U0 profiles of Fig. 8b all collapse onto a consistent

distribution, giving further evidence for flow similarity in the divergent flow regions

upstream of the flames. The slopes of the center region are also larger, but the 2:1 ratio

between ax and ar is preserved. Another observable effect of the flame is that the

minimum and maximum V/U0 values corresponding to the U/U0 peaks also increase to

show higher radial outflow. In Fig. 8c,the q’/U0 levels in the center region are more

scattered due to the influence of flames but the overall shape remain the same as in

Fig. 6c. Our flowfield analysis indicates that the overall effect of the flame is that of an

aerodynamic blockage against the flow out of the LSI. The net effects are a systematic

shift of the divergence flow into the LSI, increases in the divergence rates, and increases

in U and V in the swirl regions. These effects are weaker for flames with low heat

releases. Despite these systematic changes, the similarity features of the center region are

preserved.





Fig . 9. Radial profiles of eight flames with 9.2 < U0 < 9.5 m/s at x = 15 mm.

4.3. Turbulent flame speed

The turbulent flame speed, ST is the basic turbulent flame property that explains the LSI

stabilization mechanism because the freely propagating flame settles at the point within

the center divergent flow region where the mean flow velocity is equal and opposite to

ST. The fact that the LSI supports stable flames from 3 < U0 < 85 m/s indicates that the

ST deduced from the LSI has practical engineering significance, and provides necessary

insight for further development. From previous studies using LSBs with air-jets, it has

been shown that ST/SL correlates linearly with u’/SL. More recent data from the CH4/air

LSI flames at 7 < U0 < 22 m/s and from two 5.08 cm ID LSBs of R = 0.8 and 0.6 give

further support to this correlation. The ST deduced from the current data are listed

in Table 2. ST is defined by the velocity at the point where the centerline U0 profile

deviates from its initial linear decay. The effects of fuel composition on ST are shown

by their values listed in Table 2. Despite the low heat release rates, the ST of the diluted

H2 flames are higher than the ST of the hydrocarbon flames. In Table 2, only the u’/ST

and ST/SL for the hydrocarbon flames are listed because reliable SL data for very lean

diluted H2 mixtures are not available. From Fig. 9 it can be seen that the ST of the

hydrocarbon flames are consistent and they are well within the experimental scatter. The

inclusion of the twelve hydrocarbon flames did not affect the correlation of

ST/SL = 1 + 2.16 u’/SL. Although the ST for diluted H2 cannot be compared directly with

hydrocarbon flame data, the fact that their ST are higher strongly suggests that their

turbulent flame speeds will not be consistent with those in Fig. 9.





Fig. 10. Correlation of flame speeds measured from LSI and LSB.


Table 2 List of experimental parameters for the fuels studied
Fuel Ø U0(m/s) ax(mmˉ¹) x0(mm) ST(m/s) u’/SL ST/SL
none 0 6.76
7.47
9.21 -0.0086
-0.0085
-0.0082 -21.41
-23.45
-24.62
CH4 0.73 6.23
9.27 -0.0141
-0.0134 -38.93
-38.81 1.40
1.97 2.43
2.99 6.03
8.49
C2H4 0.62 6.32
9.40 -0.0140
-0.0130 -33.57
-45.88 1.62
2.17 2.30
3.00 6.23
8.35
C3H8 0.69 6.23
9.30 -0.0131
-0.0134 -40.92
-42.84 0.92
1.20 1.80
2.24 3.67
4.80
0.5 CH4/0.5 CO2 0.83 6.27
9.50 -0.0131
-0.0154 -42.10
-38.70 1.00
1.46 3.18
4.51 7.11
10.43
0.6 CH4/0.4 N2 0.76 6.24
9.40 -0.0142
-0.0142 -38.94
-42.75 1.16
1.56 2.45
3.69 6.44
8.67
0.6 CH4/0.4 H2 0.62 6.58
9.13 -0.0108
-0.0120 -55.95
-45.08 2.43
2.24 2.14
2.91 6.50
10.18
0.5 H2/0.5 CO2 0.25

0.3 6.48
9.55
6.56
9.38 -0.0121
-0.0102
-0.0110
-0.0094 -32.89
-34.08
-27.27
-33.70 1.42
2.91
2.54
4.00


5. Conclusions

Laboratory experiments have been performed to investigate the fuel effects on a low-

swirl injector developed for natural gas turbines. The experimental fuels comprise a

typical range (characterized by the Wobbe indices of 1430–17800 kcal/Nm3) for on-site

power generation. The LBO experiments show that the LSI with S = 0.57 supports stable

flames for all seven fuels. The stability range for 0.5 H2/0.5 CO2 flames is limited to

Ø < 0.3 where NOx emissions are below detectible limits. NOx emissions from the

hydrocarbon flames show an exponential dependence on Ø and correlate with Tad and

are consistent with previous measurements at 500 < T0 < 700 K and 6 < P0 < 15 atm.

Despite the variations in fuel properties, the LSI is capable of supporting stable

hydrocarbon flames that emit NOx < 5 ppm and CO well below acceptable limits.

Analyses of the non-reacting and reacting flowfields indicate that the overall effect of the

flame is that of an aerodynamic blockage against the flow supplied through the LSI. The

net result is a systematic shift of the divergence flow into the LSI, increases in the

divergence rates and increases in the mean axial and radial velocities in the swirl

annulus region. These effects are weaker for the flames with lower heat releases.

However, the virtual origin of the flow divergence, x0, and its nondimensional stretch rates

ax show that the similarity features of the nearfield region are preserved. The turbulent

flame speeds, ST, of the hydrocarbon fuels are consistent with those of methane/air

flames. The similarity features and linear ST correlation provide further support of an

analytical model that explains why the lifted LSI flame does not shift with U0. This

study shows that the LSI does not need to undergo significant alterations to operate with

the hydrocarbon fuels, but need further studies for adaptation to burn diluted H2 fuels.


6.Scope


1.Lowering swirl number from 0.54 to 0.43 generates more lifted flames and postpones
flame attachment to φ= 0.4 when flames are not enclosed
a LSI with S = 0.51 offers best performance for laboratory studies
b LSI for H2 is not significantly different than LSI for hydrocarbons


2.Corner recirculation zone formed at the combustor entrance promotes H2 flame
attachment
a Eliminating the sharp corner with a diffuser cone is a solution to mitigate H2
flame attachment


3. Fuel combustion chamber, and nozzle for mixing liquid fuel and air in the fuel
combustion chamber uses lean direct injection combustion for advanced gas turbine
engines, including aircraft engines. Advanced gas turbines benefit from lean direct wall
injection combustion. The lean direct wall injection technique of the present invention
provides fast, uniform, well-stirred mixing of fuel and air.






7.REFERENCES



sciencedirect.com

asme.com

M.R. Johnson, D. Littlejohn, W.A. Nazeer, K.O.Smith, R.K. Cheng, Proc. Combust.

Inst. 30 (2005)


W.A. Nazeer, K.O. Smith, P. Sheppard, R.K.Cheng, D. Littlejohn, in: ASME Turbo

Expo2006: Power for Land, Sea and Air, ASME, Barcelona, Spain, 2006.



.D. Bradley, Proc. Combust. Inst. 24 (1992)



6 P. Strakey, T. Sidwell, J. Ontko, Proc. Combust.Inst. 31 (2) (2007) 3173–3180.