Swaminathan N., Bray K.N.C. (eds.) Turbulent Premixed Flames

Подождите немного. Документ загружается.

5.1 Utilization of Hot Burnt Gas for Better Control of Combustion 373

Figure 5.8. Local efﬁciency response to the equivalence ratio of the tertiary injection in the

1D virtual combustor.

only a little larger but almost equivalent amount. The lean-mixture injection provides

a higher working gas temperature that is, higher energy, whereas it achieves an

equivalent performance regarding the NO emission compared with the air injection.

The preceding simple simulations indicate the possible advantages of staged

lean-mixture injection. The virtual 1D combustor achieves low-NO

combustion with

the operating range from 900 K (air injection case) to 1600 K (lean-mixture injection

case) in the burnt gas temperature. The combustion is stabilized by the primary stage

rather than rich combustion, and the succeeding combustion by staged injection of a

very lean mixture increases the total energy of the system without signiﬁcant increase

of NO

or CO. Figure 5.8 shows the variation of the efﬁciency at the three axial

locations in the virtual 1D combustor with the change of the equivalence ratio of the

tertiary injection mixture. The efﬁciency is deﬁned by the combustion completeness

in terms of the CO mole fraction as, 1 − x

∞

, where x

is the CO mole fraction

at the axial measurement location and x

∞

is the CO

mole fraction when all the fuel

completely reacts to products. When the tertiary injection mixture is very lean, the

CO formation is very small and then the efﬁciency is high. As the tertiary injection

374 Future Directions

Figure 5.9. (Left) Photograph and (right) schematic illustration of staged tubular combustor

with premixing tubes for primary mixture and mixture injection tubes for secondary mixtures.

mixture becomes rich, the efﬁciency ﬁrst drops down. The CO production at the

early stage of reaction, which is observed at around 5-cm distance in Fig. 5.7(a),

becomes larger, while the consumption by the succeeding reactions is slow, and then

a signiﬁcant amount of CO remains at the measurement location. If the richness of

the tertiary injection mixture is increased then the CO consuming reactions become

faster to increase the efﬁciency at the measurement location. The efﬁciency in terms

of CO reaction completeness has a strong dependency on the equivalence ratio of the

tertiary injection mixture, which results from the balance between the ﬁnite reaction

rates of CO and the combustor size. The local efﬁciency drop is also observed in

the experiments, as presented in Fig. 5.4(a), and may be attributed to the same

mechanism. This example indicates the possibility of optimal design and control of

staged injection of lean mixtures. There will be some parameters to be optimised

that control the performance of the combustor. The impact of these parameters on

the performance should be estimated at an early stage of the system design.

5.1.2 Application of the Concept to Gas Turbine Combustors

This new low-emissions combustion concept was successfully applied to a unique

lean–lean axially staged combustor installed in commercial liquid-fueled, regenera-

tive 300-kW gas turbines [18, 19]. A direct photograph and a schematic illustration of

the combustor are presented in Fig. 5.9, showing the arrangement of the premixing

tubes for the primary mixture and the mixture injection tubes for the secondary

mixture on the combustor liner dome. Premixed mixtures are prepared by atomiz-

ing liquid fuels at the inlets of the premixing and mixture injection tubes. Premixed

ﬂames of lean mixtures are stabilized at the exits of the premixing tubes. The mixture

injection tubes extend about half the full length into the primary combustion zone.

The jets of the secondary mixture, lean to ultra-lean in fuel composition, are di-

rected towards the two volumes of the primary mixture burned gas by the V-shaped

ﬂow deﬂectors attached to the exits of the mixture injection tubes. Figure 5.10 is

5.1 Utilization of Hot Burnt Gas for Better Control of Combustion 375

Figure 5.10. Photo showing phenomena in combustion chamber with secondary mixtures

being injected into volumes of hot burned gas produced by ﬂame combustion of primary

mixtures.

a photograph showing the phenomena in the combustion chamber, installed in the

combustor test rig, operating with kerosene at simulated engine full power condi-

tions of 920 K inlet air temperature and 0.4-MPa pressure [20]. The primary burners

on the end wall were not visible because of the radiation from the volumes of the

primary combustion products (dotted circle). The mixture injection tubes are seen

on the top and bottom and the details of the exit of these injection tubes are clearly

seen. Four dilution air holes on the combustor liner and three thermocouple rakes

for the measurement of the exit gas temperature proﬁles are also seen. The arrows

starting from the exit of the mixture injection tubes show the directions of the sec-

ondary mixture jet. No ﬂames are seen to be stabilised at the exit of the secondary

mixture injection tubes, as intended while the combustor was designed.

The NO

and CO emissions levels of the gas turbine are well below 15 and 35 ppm

(15% O

), respectively, for kerosene over a range of loads from 50% to 100%, as

shown in Fig. 5.11. Other liquid fuels such as gasoline and solvents disposed from

factories are also used maintaining low levels of NO

and CO emissions. For more

than 10 years of operation, neither auto-ignition nor ﬂashback in the tubes has been

experienced. Combustion is smooth and free from oscillatory combustion over the

full range of power.

5.1.3 Numerical Simulation towards Design Optimization

Thanks to the development of computer power, parallel computing techniques, and

numerical methods, numerical simulation is becoming available as a tool for engi-

neering design in some ﬁelds. It is expected to assist the optimal design and control

376 Future Directions

0 50 100 150

Power (kW)

200 250 300

GT Engine : RGT3R

CO Target 50ppm

Target 20ppm

, CO (@15%O2 ppm)

Regerative

Out put : 300kW

Combustor Mod : MOD15

Fuel : Kerosene

Figure 5.11. Measured NO

and CO emissions when power is increased by increasing sec-

ondary fuel ﬂow while primary fuel ﬂow is maintained in lean–lean two-stage combustion.

of staged lean-mixture injection. The major issues in this combustion conﬁguration

are rapid mixing at each staged injection and the combustion reactions of the mix-

ture composed of hot burnt gas and injected fuel-lean mixture. The requirement of

a numerical simulation is to estimate the previously mentioned mixing process and

chemical reactions that take place in real-size combustors at an acceptable level and

throughput.

The mixing process contains two parts. One is the atomization and vaporisation

process of a liquid fuel jet mixing into air to make the lean mixture for injection. The

other is the mixing of the lean mixture and the hot burnt gas. The two may occur

sequentially and simultaneously. The former process is not yet clearly understood

or well modeled. Some detailed and large-scale simulations have been conducted to

understand the mechanism and to model the phenomena into CFD (computational

ﬂuid dynamics) friendly formulations [21–23]. Figure 5.12, for example, shows a

snapshot of an atomising liquid jet reproduced by two-phase ﬂow simulation using

about 6 billion cells, which includes the ligament production process from the jet

and the droplet production process from the ligament. The latter mixing process is

strongly affected and controlled by turbulence caused by the interaction between

burnt gas ﬂow and the injection jet. As with other mixing problems in the absence

of combustion, the improvement of the turbulent ﬂow model is persistently the key

for the estimation, whichever approach, either RANS (Reynolds-averaged Navier–

Stokes) or LES (large-eddy simulation), one uses.

Concerning the combustion reaction simulation, one of the critical problems is

the use of practical fuels that contain heavy hydrocarbons; for example, the carbon

number of kerosene is 10–15. It is not easy to understand the rigorous chemical

5.1 Utilization of Hot Burnt Gas for Better Control of Combustion 377

Figure 5.12. Detailed simulation of atomization process of liquid jet [23]. The suface colour

shows the axial velocity. (See colour plate.)

kinetics and to include the huge number of species and elementary reactions in a

numerical simulation for the optimal design and the further control process of staged-

injection combustion, although these elementary reactions can be important. Given

the proper chemical kinetics, the combustion modelling will be the most important

issue, as is now the case in many combustion simulations. The combustion we have

to deal with may be categorised into a different type from that described by the

concept of ﬂamelets [24]. The reactants are not pure unburnt gas components at

low temperature but the mixture of the hot products and radicals produced by the

previous stage and the injected lean mixture. The combustion occurring in the staged-

injection location may be so-called ‘mild combustion’ or ‘ﬂameless combustion’

[6, 7, 25, 26], including, auto-ignition processes, to which the conventional concept of

a ﬂamelet may not be applied rigorously. The combustion may take place in rather

thick layers, and propagative ﬂame may not exist. Direct numerical simulation (DNS)

with detailed chemistry and transport properties, which is becoming possible for real-

size combustion phenomena [27–29], is expected to be the most promising approach

to estimate combustion conﬁgurations such as staged lean-mixture injection. But

even today’s Peta FLOPS computer power is not enough to simulate the previously

mentioned phenomena in practical scale combustors with detailed chemistry of a

practical fuel. The situation will not drastically change even in the next Exa FLOPS

era. Hence, for the near term, the modelling of the combustion of the hot burnt

gas and injected lean mixture is an important and indispensable research subject,

as well as the development of improved numerical techniques such as discretization

methods. The chemistry reduction and tabulation approaches [30, 31] may be of help

378 Future Directions

in the construction of models. DNS applied to rather fundamental conﬁgurations

will support the construction and validation of these models.

5.2 Future Directions and Applications of Lean Premixed Combustion

By J. F. Driscoll

Combustion technology in the future is expected to progress more and more in the

direction toward lean premixed turbulent combustion for several reasons. Premixing

creates short, compact ﬂames that allow the size of an engine to be reduced. Smaller

engines are needed for small business jets, for mobile gas turbine power generators,

and for small automobile engines. Premixing also provides inherently low levels of

nitric oxide, CO, and hydrocarbon pollutants. Government regulations of pollutant

levels are expected to tighten for automobiles, aircraft, and ground-based power

generators. Another motivating factor is the attempt to reduce the ‘carbon footprint’

and thus the grams of CO

emitted per kiloJoule of energy delivered. There is

evidence [32] that signiﬁcant reductions in CO

emissions are possible if a lean

premixed recuperative gas turbine power plant is operated using a hydrogen-CO

synfuel, as subsequently discussed. Commercial aircraft engines also are moving

to premixed technology. A lean premixed prevapourised (LPP) combustor in the

General Electric GEnx engine will be operated in the next few years in the Boeing

787 aircraft [32]. In the automotive engine area, the HCCI engine represents another

new advance in lean premixed combustion technology.

However, this trend towards increased levels of premixing also leads to several

new research challenges. Partially premixed combustion often occurs in LPP and

HCCI devices, and it presents modelling challenges because any instantaneous re-

action interface can contain regions, of premixed ﬂames, which propagate as waves,

and non-premixed regions, which do not propagate, so the t wo regions are governed

by different physics. Premixed ﬂames are inherently more difﬁcult to stabilize in a

high-speed ﬂow than non-premixed ﬂames, so the trend towards premixed devices

makes it especially important to develop models that can realistically and reliably

predict ﬂame lift-off, blowout limits, and ﬂashback, and such a model is not available

today. This issue is discussed in Subsection 5.2.2. Flame-stability complications arise

because engines usually employ complex bluff-body geometries, swirl, and recircu-

lation zones to assist in ﬂame stabilisation. Another future driving force is the need

for a model that can reliably predict the onset of combustion instabilities that are

coupled with the acoustic pressure waves, which often arise when premixing is em-

ployed. Subsection 5.2.3 lists some emerging experimental technologies; the future

direction of measurements is to provide high-speed imaging of unsteady combustion

processes with improved spatial and temporal resolution. Subsection 5.2.4 contains

a discussion of some fundamental issues that still require further work to improve

our ability to model and understand premixed turbulent combustion.

5.2.1 LPP Combustors

Several types of gas turbine combustors have been developed that use premixing to

reduce both the NO

emissions as well as the size of the reaction zone. Three exam-

ples are called the dry low-NO

(DLN) injector [32–35], the low-swirl combustor,

5.2 Future Directions and Applications of Lean Premixed Combustion 379

(a) (c)

(d)

(b)

Figure 5.13. Lean premixed technology within two commercially available gas turbine com-

bustors: (a,b) DLN device operated on gaseous fuel (natural gas) for ground-based power

generation (schematic of Maughan et al. [33]). (c,d) GE-TAPS operated on liquid Jet-A fuel

for aviation applications (schematics of Mongia [37] and Dhanuka et al. [38]). PRZ, CRZ,

and LRZ respectively mean primary, corner, and lip recirculation zones.

which was developed by R.K. Cheng [36], and the TAPS (twin annular premixed

swirler) mixer that is described by Mongia [37] (see Section 4.2 also). DLN and

low-swirl devices operate on gaseous fuel (including H

/CO syngas mixtures) for

ground-based power generation, and TAPS is certiﬁed for use in commercial aircraft

that are run on liquid Jet-A fuel.

The DLN fuel injector [32–35] is shown in Fig. 5.13(a). It contains two annular

ducts that are shown in Fig. 5.13(b). The inner duct contains a small ﬂow rate of

pure fuel that is gaseous and swirling. This pilot fuel burns downstream as a non-

premixed pilot ﬂame. Most of the fuel is injected into the outer duct as a fuel jet

into a cross-ﬂow of air. This main fuel and air premix a few millimeters upstream

of the premixed main ﬂame. The primary zone equivalence ratio is designed to vary

from 0.5 to 0.8, and the attempt is to keep the maximum values of the instantaneous

gas temperatures below 1800 K to reduce thermal NO

formation. Another goal is

to keep the post-ﬂame temperatures between 1500 and 1800 K to oxidise the CO

without producing additional NO

. Using natural gas fuel, very low NO

levels of

5 ppm were reported [33].

The DLN device also has been used to burn a synfuel mixture of H

and CO

in a gas turbine power generator [32, 34]. DLN was reported [34] to provide a

signiﬁcant reduction in the cost of capturing CO

. Using a regenerative cycle, natural

380 Future Directions

Figure 5.14. Fluorescence images of ﬂames in lean premixed devices: (a) Instantaneous OH

PLIF in gaseous fueled LP combustor of Stopper et al. [39]; (b) time-averaged formaldehyde

ﬂuorescence of main and pilot ﬂames in a TAPS LPP combustor fueled with liquid Jet-A, by

Dhanuka et al. [40].

gas was converted to synfuel and the exhaust gases were recirculated. The DLN

injector provides the ﬂexibility to optimize the pressures, temperatures, and CO

concentrations within the regenerative cycle so that CO

could be more efﬁciently

removed by both absorbing solvents and a CO

membrane. The use of synfuel

creates two challenges: Flashback occurs for large H

/CO ratios, and for small H

/CO

ratios the challenge is to consume all of the CO. Synfuel can contain different

concentrations of H

and CO so its ﬂame speed can vary; it important to have

models that can predict the effects of varying ﬂame speed on engine performance.

The low-swirl combustor of R.K. Cheng and co-workers [36] has similarities with

the DLN device in that it operates on natural gas or gasesous synfuels and it achieves

low NO

by employing unique methods to stabilise ultra-lean combustion. Diverging

walls and swirl force the axial velocity to decrease as the reactants approach a

relatively ﬂat premixed ﬂame; this avoids ﬂashback because the ﬂame must propagate

back against progressively larger reactant velocities. It was selected for modelling

studies because of its simple geometry and because a fairly complete data set of ﬂow

properties was measured [36]. Another fairly complete database was measured for a

lean premixed design by Stopper et al. [39] and one OH image is seen in Fig. 5.14(a).

For aircraft applications, TAPS is a type of LPP device that is operated on liquid

Jet-A fuel [37]. TAPS provides a 50% reduction in NO

emissions compared with

the previous non-premixed ‘rich dome’ technology, and it is currently employed in

the GEnx engine. Figure 5.13(d) shows that the TAPS injector creates an inner pilot

ﬂame that surrounds the cone of Jet-A fuel drops that are created by an atomiser.

Most of the fuel enters in the main fuel stream. This main fuel jet is atomised when

5.2 Future Directions and Applications of Lean Premixed Combustion 381

0.4 0.45 0.5 0.55

Fuel Equivalence Ratio

CO in exhaust (ppm)

Increase in UHC

oxidation time

(a) Vorticity contours

(b) Corbon Monoxide

Flame surface

Figure 5.15. LES results for Gasesous-Fueled LP gas turbine combustors, similar to the

DLN design of the previous ﬁgure. (a) Instantaneous vorticity and ﬂame surface from the

G equation solutions of Yang and co-workers [42]; (b) exhaust CO computed using the G

equation coupled with a ﬂamelet library, by Eggenspieler and Menon [41].

it interacts with a high-speed cross-ﬂow of heated air. The cross-ﬂow velocity is large

enough to efﬁciently atomise and prevapourise the fuel, and it creates a sufﬁciently

large lift-off distance between the main ﬂame and the fuel injection location. This

distance provides the premixing that is necessary to achieve the proper fuel-lean

conditions before reactions occur. As with the DLN device, the goal is to keep

temperatures in the premixed main ﬂame and post-ﬂame gases between 1500 and

1800 K to minimize NO

while maximising the oxidation of CO. Residence times of

post-ﬂame gases in the three recirculation zones shown in Fig. 5.13(d) are important.

A complete database of velocity and ﬂame properties in a TAPS LPP combustor was

measured by use of particle image velocimetry (PIV) and PLIF diagnostics [38, 40].

The dotted lines in Fig. 5.13(d) represent the main s hear layer that provides the

necessary premixing; the main ﬂame is anchored near the low-velocity edge of the

shear layer. Unfortunately the shear layer also contains a periodic stream of strong

vortices that tend to create oscillations in the ﬂame anchoring position and may lead

to combustion instabilities at off-design conditions.

The modeling of LPP devices requires proper simulation of several highly un-

steady features, including the acoustic pressure ﬁeld, the precesion of the recircu-

lation zone, the shedding of vortices in the shear layer, and the unsteady locations

of ﬂame anchoring. The primary recirculation zone (PRZ) has a toroidal shape and

has a ﬂuid dynamic centreline that precesses about the geometric centreline. LPP

premixed ﬂames can ﬂash back, but only to the locations where the main fuel jet is

injected into to the main air ﬂow, so this distance is purposely minimised. The main

air velocity is designed to be larger than any possible turbulent burning velocity in

order to prevent ﬂashback. One research issue is that these devices are difﬁcult to

model because partially premixed combustion often occurs. More research is needed

to optimise the uniformity of the temperature ﬁeld, to optimise the ﬂame holding,

and to avoid a combustion instability known as engine growl [37].

Figure 5.15 shows two examples of LES results from Yang, Menon, and co-

workers [41, 42], who simulated a lean premixed gas turbine combustor that is

operated on gaseous fuels. The G equation (see Section 2.2) was used to track the

ﬂame surface at the resolved scale in order to ensure that this surface has the proper

wave physics; the surface must propagate normal to itself and form cusps that are

382 Future Directions

due to Huygen’s Principle. The chemistry was included in a premixed ﬂamelet library

that depends on a variable (G) that indicates the distance from any point to the ﬂame

surface. Partial premixing is included by deﬁning a conserved scalar and by adding

a library for nonpremixed ﬂamelets. The idea is that a ﬂamelet solution provides

the number of kilograms per second of CO that is produced per unit ﬂamelet area.

This number is then multiplied by the predicted ﬂame surface area per unit volume,

which is deduced from the mean gradient of the G ﬁeld. This yields the computed

volumetric CO reaction rate. The post-ﬂame reaction rates and residence times also

are modeled to account for NO

that is created and CO that is oxidised in the post-

ﬂame gases. This approach inherently contains assumptions about several values of

means and variances that result in the prediction of how much of the fuel burns in

the premixed mode and how much burns in the non-premixed mode.

The DLN and TAPS devices previously described contain ‘partially premixed

combustion’ (PPC). Normally the term “premixed ﬂame” implies that fuel and air

are fully mixed with no stratiﬁcation in the reactants. Therefore the term PPC is

used to refer to reactants that are not fully mixed. That is, PPC consists of stratiﬁed

premixed or stratiﬁed non-premixed ﬂames that often are close enough to each

other to interact. One example of PPC occurs when the fuel equivalence ratio varies

along a reaction layer and there are regions where the reactant mixture lies within

the ﬂammability limits, so a stratiﬁed premixed ﬂame exists. In other regions the

reactants lie outside of the ﬂammability limits so a stratiﬁed non-premixed ﬂame

exists. Where local extinction occurs along the reaction layer, two edge ﬂames may

propagate towards each other to reconnect the layer. The similar case of edge (triple)

ﬂames [43] near the base of a lifted non-premixed jet ﬂame represents another

example of PPC. In addition, PPC has been used to describe a rich premixed Bunsen

ﬂame that interacts with a downstream non-premixed ﬂame; the latter exists at the

boundary between the rich combustion products and the surrounding air.

A number of new technologies are needed to address the research challenges

of modelling partially premixed combustion. Experimental methods are needed

to quantify how much of the fuel burns in the premixed mode and in the non-

premixed mode. To do so, measurements of the ﬂame index would be helpful. The

ﬂame index is the product of the local gradients of the fuel and the oxidiser mass

fractions. A positive index identiﬁes a premixed ﬂamelet whereas a negative index

identiﬁes a non-premixed ﬂamelet. The DNS of Takeno and co-workers [44] provided

computed values of the ﬂame index. To assess such simulations, fuel gradients can be

measured by use of line Raman [45] or PLIF of fuels (or fuel markers) that ﬂuoresce.

Measurements of the oxidiser gradients are still a challenge. Reaction-rate intensity

can be identiﬁed by employing simultaneous OH–formaldehyde PLIF diagnostics

[46], and this may provide another way to differentiate between premixed and non-

premixed combustion. Other experimental challenges are the application of PIV

and PLIF to the high-pressure, dense spray environment of LPP devices and the

measurement of scalar gradients. At high pressures the background radiation from

soot precursors presents a formidable interference for both PIV and PLIF signals. In

dense sprays, PIV may incorrectly track the drop velocity instead of the gas velocity.

Models of the chemistry in LP devices [41, 42] considered two regions: the

primary reaction layers (ﬂamelets) within the ﬂame brush and the post-ﬂame gases,

where distributed secondary reactions continue to form NO

and to oxidise CO.