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

Подождите немного. Документ загружается.

4.2 Application of Lean Flames in Aero Gas Turbines 323

pilot fuel is often used as a coolant, adding complexity to an already complicated

assembly, and of course, adding to heat pickup by the pilot fuel.

2. Main injector fuel purging and reignition has an impact on emergency acceler-

ation time, which is an airworthiness issue. To avoid compressor surge, staging

must take place smoothly. An initial fuel spike must be avoided. There is less

than a 1-s margin between the acceleration time achieved by high-bypass-ratio

engines and the airworthiness requirement. The implications are that the fuel

manifolds cannot be purged, for to reﬁll them very quickly without creating a

fuel surge when the liquid reached the nozzles would be very difﬁcult.

3. During approach, frequent, large changes in thrust will give rise to cyclic staging

at low altitude with the aircraft ﬂaps in a high-drag conﬁguration. Staging and any

purging process must be completely reliable. The situation is more hazardous

in heavy rain and is exacerbated by the fact that all engines undergo thrust

changes simultaneously. Extinction could occur simultaneously in all engines,

a situation from which recovery would be impossible. It is probably necessary

to stage injectors in several groups to avoid excessively depleting pilot ﬂow to

furnish the main injectors with adequate fuel to light. This complicates the fuel

control system, increasing the likelihood of component failure. It is important to

remember that thrust changes must be smooth and predictable. The pilot must

have virtually perfect control of engine behaviour.

4. The TET proﬁle may change substantially during staging, resulting in deteriora-

tion of hot end components.

5. Fuel manifolds are exposed to high ambient air temperature and radiation from

the HP engine casing, which can reach a temperature of 600

◦

C. Employing fuel

control valves remotely from the engine core creates a fuel heating problem in

the main manifold. The use of a single manifold in which fuel ﬂows at all ﬂight

conditions requires the fuel staging valves to be located in a hostile environment.

Concentric manifolds could be used so that the pilot fuel cools the main, but

complexity and cost make that option unattractive.

These are not insoluble problems, but they increase the scale of the necessary tech-

nology development programmes. Understanding these complex, critical issues helps

to explain why progress is seemingly slow.

4.2.7 Performance Compromise after Concept Demonstration

Research programmes necessarily concentrate on those aspects of combustor per-

formance in which the stakeholders are interested – currently emissions reduction –

and use test rigs that do not perfectly replicate the engine operating conditions, ge-

ometry, and compressor delivery ﬂow velocity proﬁle, wakes, and turbulence. Often

the engine operating pressure and mass ﬂow are higher than those of available rig

test facilities. The core compressor power input for a large aeroengine is more than

100 MW. After the successful achievement of emissions targets in, say, a single-sector

high-pressure combustion test rig, the combustor will be redesigned in annular form,

with representative diffuser geometry, and at the scale required by the engine project.

Optimisation of the combustor to achieve altitude relight, to eliminate combustion

instability, to improve cooling, to adjust outlet temperature distribution to suit the

324 Lean Flames in Practice

9% cooling

29%

6% - cooling

100%

70% - high power

22% - low power

15%

turbine

19%

Figure 4.55. Typical ﬂow modulation required in a variable-geometry combustor. (Courtesy

of Rolls–Royce plc.)

turbine, to reduce cost and weight, to incorporate thermal management into the fuel

injector, and ﬁnally to operate the combustor in the engine environment, give rise to

loss of emissions performance. It is r easonable to suppose that the ﬁnal NO

will be

50% higher than that predicted from sector rig tests, and sometimes it could double.

4.2.8 Lean-Burn Options

VARIABLE GEOMETRY. Theoretically, variation of the airﬂow distribution in the com-

bustor would solve the problem of operability, and this was demonstrated experi-

mentally by R-R in the Japanese government funded hypr programme during the

1990s. [208]. Altitude relight was achieved at 15-Km altitude, adequate idle efﬁciency

was achieved, and NO

at cruise measured in a single-burner rig was well below the

target EI of 5. Geometry was not adjusted during test, but the combustor was re-

conﬁgured and then tested at different conditions. When taken to the next stage –

a three-burner sector – severe combustion instability occurred. Design studies per-

formed in that programme failed to ﬁnd a practical method of varying geometry in

an engine. Figure 4.55 shows the range of air ﬂows required. The main points to note

are these:

1. The effective area change in the fuel injector is so great that dilution ﬂow must

also be changed to avoid an unacceptable increase in combustor pressure loss.

2. The combustor is annular. Modulation of the inner dilution jets by mechanical

means appears impractical.

3. Any mechanical mechanism for ﬂow modulation must withstand vibration and

operate reliably at temperatures up to 650

◦

4. Working clearances between moving parts allow leakage, impairing accuracy of

ﬂow control.

4.2 Application of Lean Flames in Aero Gas Turbines 325

Radial air swirlers

Combustion air

Fuel Manifold

Fuel spray

Combustion

air

Fuel–air mixture

Impingement cooled

dome

Ceramic

thermal

coating

Fuel–air

mixture

Figure 4.56. Macrolaminate injection system invented and developed by Parker Hanniﬁn

Corporation. (Courtesy of Parker Hanniﬁn Corp.)

5. Fluidic devices such as vortex ampliﬁers waste half the available wall pressure

loss to achieve ﬂow stability, and to achieve substantial ﬂow variation with the

available wall pressure loss of 3%, they have to be far too big to ﬁt within the

available space.

MULTIPLE INJECTORS. GE, UTC, and NASA with Parker Hanniﬁn [209] investigated

the idea of using up to 1000 small injectors combined with fuel staging to achieve

rapid, uniform mixing of fuel and air, with short residence time. This approach is

used successfully for burning gas in the GE LM6000 engine.

The obstacles to be overcome for aeroengine application are

The heat transfer to fuel, leading to injector coking; removal of the injectors for

cleaning is impractical (probably true for most ultra-low-NO

systems);

The depth of the combustor renders diffuser design and accommodation within

the bypass duct difﬁcult and adds weight.

Parker Hanniﬁn Corporation developed Macrolaminate injectors, in which a large

array of injector elements is chemically etched into sheets, subsequently laminated

by diffusion bonding or brazing (see Fig. 4.56). Both air and fuel passages can thus

326 Lean Flames in Practice

Figure 4.57. Radially fuel-staged double-annular combustor. (Courtesy of Rolls-Royce

Deutschland)

be formed. Heat transfer from air to fuel must be limited by careful design. No

application is known to the author, but the concept offers ﬂexibility, low cost, and

uniform distribution of fuel and air. Performance of this injection system was demon-

strated in high-pressure research combustion rigs, and the emissions performance

shows promise [210]. Very low NO

was demonstrated at pressures up to 28 atm

and with an inlet temperature of 870 K. There was no evidence of ﬂashback into the

swirl modules. High efﬁciency was also achieved at cruise conditions, so the concept

appears well suited to fuel the main zone of a staged combustor.

An aspect of performance not discussed in the available literature, to the knowl-

edge of the author, is that of altitude relight performance. The author believes that

this would prevent application to the pilot zone of an aero gas turbine. The achieve-

ment of ignition at above 8000-m altitude requires a substantial recirculation volume

in the primary reaction zone. The length of this zone in a combustor with a single ring

of injectors is typically half the combustor depth, being determined by the length

of the recirculation generated by the fuel injector and surrounding air swirlers. The

recirculation length is proportional to the diameter of the injector. If, for example,

the scale of the fuel injection device is reduced to one quarter, then the length of the

recirculation-zone will also be reduced by 75% and with it the recirculation-zone

volume. Altitude relight could not be achieved.

The concept lends itself to an axially staged arrangement with a separate, con-

ventionally fuelled pilot zone similar in layout to that shown in Fig. 4.57 or to an

arrangement in which the Macrolaminate nozzles are grouped around separate pilot

injectors in a single annulus.

WATER INJECTION. Direct water injection into the combustor is a practical and

powerful way to reduce NO

at take-off. Substantial NO

reduction can be achieved

4.2 Application of Lean Flames in Aero Gas Turbines 327

with a water-to-fuel ratio of around 1:1. If intimately mixed, a reduction in ﬂame

temperature of around 300 K can be achieved, and the volume of steam produced

also reduces the oxygen and nitrogen concentrations in the combustion gas by about

10%, reducing reaction rate by 20%. NO

is approximately halved by a reduction in

ﬂame temperature of 100 K, so to a ﬁrst approximation a 10-fold reduction in NO

can be expected. Reductions of this order have been demonstrated in engine tests.

A modern large aircraft with take-off thrust of 1000 kN will consume around

600 kg fuel during the ﬁrst minute from the start of take-off. It would thus require this

weight of water and would require a water tank and pumps, adding weight that would

be carried throughout the ﬂight. This increases fuel consumption, further adding to

take-off weight and emissions.

Freezing of the water has to be avoided in extreme weather conditions, and

concern has been expressed about the possibility of the wrong liquid being put into

the wrong tank during fuelling. This may be very unlikely and would be hindered by

baulking features on delivery nozzles. The author is not qualiﬁed to judge the risk,

but the consequences of a mistake would clearly be very serious.

The idea has come up for discussion periodically but always the disadvantages

have been considered too great. Increasing concern about NO

could shift the

balance.

LPP FUEL-STAGED COMBUSTION. LPP combustion has achieved very low NO

high-pressure test rigs. Fuel staging could take the form of a central pilot zone within

each premixing duct outlet or a separate ring of conventional fuel injectors in a

double-annular conﬁguration. An example of this type is shown in Fig. 4.57.Of

these, the former appears preferable. The pilot ﬂame can be used to stabilise the

LPP ﬂame and to aid smooth light across during staging. The practical difﬁculties

with LPP combustion are the risks of auto-ignition and ﬂashback at high power,

poor turn-down ratio, high CO at cruise, combustion instability, and the problem of

dispersing the fuel uniformly in a very short time.

AUTO-IGNITION. There are very large differences in auto-ignition delay times for avi-

ation fuels published over the last 40 years, arising from differences in the design of

the test apparatus, the difﬁculty of the experiments, and the stochastic nature of the

process. The various types of apparatus employed – shock tubes, piston compression

devices, static bombs into which fuel is injected into preheated, pressurised air, and

steady-ﬂow combustion tunnels provide data over different temperature and pres-

sure ranges. The most difﬁcult conditions for obtaining data appear to be between

850 and 1100 K with pressures above 20 bars – just the range of conditions at which

data are needed for LPP system design.

The steady-ﬂow research rigs are closest in geometry to gas turbine premixing

ducts. Effort is made to achieve uniform fuel distribution and air velocity across

the duct, and of course there is no constraint on upstream settling length or on the

complexity and size of the fuel injection apparatus.

If the auto-ignition delay time is affected by the variation in rig design, how

much more is it likely to be affected by differences in premix duct design in gas

turbines? They are constrained to ﬁt within a length of less than 100 mm, are fed

with a non-uniform velocity proﬁle from a compressor immediately upstream, they

may have to employ swirl to distribute the fuel, with wakes from swirl vanes, and

328 Lean Flames in Practice

the fuel distribution may vary greatly with engine power because of the variation of

momentum ratio between air and fuel.

To quantify this problem we compare results published by leaders in this ﬁeld.

Some excellent work done by Spadeccini and TeVelde [211] of UTC was published

by NASA in 1980. The value of this paper is enhanced by the publication of the

measured – not merely correlated – data. Although tests were conducted over the

range 700–1000 K, almost no auto-ignition events could be obtained at above 900 K,

and there are no data at above 20 bars with a temperature higher than 800 K.

A wide range of delay time pressure dependence is reported. Spadaccini and

TeVeld e [211] report a P

−2

dependence. Others, like Hayashi of JAXA report [212]

−1

, and results collected from various sources by Coats [213] of the UK Royal Air-

craft Establishment (RAE) also show P

−1

dependence. These observations illustrate

the difﬁculty of formulating design rules. As most of the high-temperature data were

obtained at 20 bars or less, and with large uncertainty on the pressure exponent,

one cannot estimate auto-ignition delay time at 50 bars reliably. There are ratios

greater than 2 in ignition delay time measured in different test facilities at the same

conditions. Coats [213] included data obtained up to 60 bars, but the data reported

in that survey from the mid-1980s indicate signiﬁcantly longer delay times than the

more recent publications. What safety margin should we apply in application of the

experimental data to a real design?

For the moment let us not compound the problem by extrapolation. There are

a lot of data at around 800 K and 20 bars corresponding to altitude cruise for a large

engine and max take-off for an engine of below 65-kN thrust. Application of the

Spadeccini and TeVelde correlation [211] gives a range of delay time from 0.75 to

1.25 ms, Hayashi et al. [212] reports 1.1 ms, and data collected by Coates [213]give

a value of around 5 ms.

Bearing in mind that some compromise of performance may be expected in a

premix duct designed for an engine, a safety factor of 50% relative to the short-

est time measured in a laboratory experiment appears reasonable. This implies a

duct residence time of 0.5 ms at 20 bars and 800 K. Assuming a ﬂow velocity of

100 m/s in the premixing duct, the available mixing length is 50 mm. The author’s

experience with experimental, lean premixed combustors suggests that this estimate

is conservative, but conservatism is necessary for the sake of safety. Auto-ignition

delay experiments are conducted for short periods, and the experiment ends as soon

as ignition occurs. A gas turbine operates for thousands of hours, heat from the

ﬂame can soak back into the premix duct, raising the boundary-layer temperature,

compressor surge can reverse the ﬂow in the duct temporarily – if ﬂashback occurs

then detection and shutdown are necessary or the ﬂame must blow out of the duct.

Compressor surge can be caused by bird ingestion during take-off. It is not a rare

event in aircraft safety terms, in which a probability of 10

−8

is considered remote.

It is clear from this preliminary assessment that LPP combustion may be imprac-

tical at higher pressures and temperatures than those previously assumed. For a large

engine during take-off and climb, DI would be necessary, but to achieve reasonably

low NO

, even this must be a fuel-staged system. It does not appear practical to

incorporate three kinds of injection system in one engine – a diffusion burner for op-

eration at and below ﬂight idle, an LPP duct injector for cruise, and a LBDT burner

using the same air as the LPP burner for take-off and climb. The author believes

4.2 Application of Lean Flames in Aero Gas Turbines 329

that application of LPP should be limited to engines of below 20 OPR and thus to

regional aircraft of 50 seats or fewer.

The engine manufacturer has to consider the implications of thrust growth,

mentioned in the introduction of this section. It would be very costly to change

the combustion system because a step in thrust growth rendered an existing LPP

combustor unsafe to use. For example, the Rolls-Royce BR700 series of engines

use the same combustor and range in thrust from 66 to 93 kN and in pressure ratio

from 24 to 32, the higher pressure ratio adding about 75 K to the combustor entry

temperature.

SUPERSONIC CIVIL TRANSPORTS – A SPECIAL CASE. Although no such aircraft are

planned at present, they have been discussed repeatedly, discussions that have stim-

ulated substantial research programmes, such as the NASA High Speed Research

initiative of the 1990s and the Japanese Government programmes HYPR (1989–

1999) and ESPR (2000–2004), sponsored by NEDO.

Supersonic transports must be treated as a special case for several reasons:

They ﬂy at high altitude, within the lower part of the ozone layer, which is

sensitive to NO

concentration.

They consume around 50% more fuel per passenger kilometre than a subsonic

aircraft.

During cruise, ram precompression raises the combustor inlet temperature to

over 850 K, considerably higher than in a subsonic aircraft. To limit cruise tem-

perature, the SLS pressure ratio is low around 10:1 and because the atmospheric

pressure at 15 km is only 0.115 bar, the combustor pressure at cruise will be

around 10 bars. It is therefore probable that engines of this type, in which high

pressure and high temperature do not occur simultaneously, could utilise LPP

combustion.

PRACTICAL LIMITATIONS GOVERNING THE DESIGN OF A PREMIXING DUCT. For the pur-

pose of illustration through which some understanding of the design problem may be

gained, let us consider a simple, annular premixing duct supplied with fuel through

radial jets. A swirling fuel preﬁlmer like that used in many airblast injectors, sand-

wiched between airstreams, could also be used, but radial dispersion across the

airstream is difﬁcult to achieve in a short distance. Cross-stream injection achieves

efﬁcient atomisation and mixing when optimised. Some excellent research was per-

formed by Rachner et al. [214] on this subject.

The difﬁculty in the application of cross-stream injection is that jet penetration

decreases with engine power. Whereas the air velocity is almost independent of

power and thus air momentum varies with pressure, fuel momentum varies approx-

imately with the square of the engine pressure. Jets optimised for cruise will over

penetrate at take-off. However, the following issues have to be taken into account:

1. Mixing in the duct: The high velocity in the duct allows little pressure loss for the

generation of turbulence. The dynamic head in the duct proposed for illustration

is 3% of the total pressure. The available combustor pressure loss is about 4.5%.

Any device used to produce turbulence must not create wakes in which fuel

could auto-ignite.

330 Lean Flames in Practice

2. Premixing duct ﬂow: An engine suitable for a 50-seat regional jet such as the GE

CF34-3 delivers 40.9-kN SLS thrust, at 21 OPR, with bypass ratio of 6.2. SLS

SFC is 0.34 kg/kg/h, so fuel ﬂow at full power is about 0.4 kg/s, 90% of which

would pass through the premixing ducts, 10% through pilot nozzles. Combustor

air mass ﬂow would be about 17 kg/s, and total premixing duct ﬂow would be

9.5 kg/s, giving an equivalence ratio of 0.65.

3. Duct ﬂow passage: The design constraints are as follows:

a. Experience shows that fuel jets less than 0.5-mm-diameter block easily.

b. For a reasonable fuel pressure drop over the power range, the total number

of jets must be limited, and they have to be shared among the premix ducts.

The larger the number of ducts, the more widely spaced the jets are. There

comes a point when circumferential uniformity of the fuel distribution is

compromised.

c. There must be enough space in the centre of the premix duct to accommodate

a pilot injector, allowing for double skinning of the fuel passages for thermal

insulation.

d. The ﬂow passage must contract in area by at least 10% between fuel injection

plane and outlet to prevent ﬂashback and to suppress wakes or separations

in which auto-ignition could occur.

An example calculation is as follows.

Dynamic head 3% P

comb

Premixing duct effective area 89 cm

Number of premix ducts 12

Fuel jet diameter 0.5 mm

Fuel jet C

0.75

Fuel P 8 bars

Fuel ﬂow 0.36 kg/s

Fuel ﬂow per jet 0.0048 kg/s

Number of jets 72

Number of jets per duct 6

O.D. of duct 40.0 mm

I.D. of duct 27.4 mm

Radial thickness of insulation

between pilot and main passages 5 mm

Max O.D. of pilot 17.4 mm

Min O.D. of pilot to pass 10% ﬂow 13 mm

Duct height 7.2 mm

Jet pitch 14.3 mm

Jet penetration reaches almost to the full duct height, based on the correlation

in [214], and is thus satisfactory.

4. Discussion. A jet pitch of twice the duct height is larger than one would like, but

to reduce the diameter of the duct, so that height increases and pitch reduces,

leaves insufﬁcient space for a pilot and results in inadequate jet penetration. At

high-altitude cruise, the jet penetration is reduced. In a business jet application,

in which the cruise altitude might be 13 km, fuel pressure drop could fall to

4.2 Application of Lean Flames in Aero Gas Turbines 331

less than 1 bar and the f uel jets would then penetrate to only about 60% of the

duct height. Regional jets usually cruise at below 9000 m. Because business jets

ﬂy very few hours per year, they do not contribute signiﬁcantly to the global

emissions problem. Deterioration in cruise NO

should not be a concern. The

radial FAR proﬁle would be biased to the inside at cruise, which might help

combustion efﬁciency.

The preceding simple calculation indicates that a practical premix duct geometry

can be formulated for an engine of modest pressure ratio. LPP combustion applied

to engines of around 20 OPR should reduce LTO and cruise NO

by up to 70%

compared with today’s values. The gain is compromised by the fact that the premix

duct cannot be used at idle and during descent, and by the need for a diffusion pilot

over the entire ﬂight envelope.

COMBUSTION INSTABILITY. Because the mixture entering the combustor is uniform

and lean, small perturbations in AFR affect the rate of heat release t hroughout

the burning zone in the same sense. In a DI combustor, some zones are richer

and some are leaner than stoichiometric, so the response to a change of AFR in

these zones is of opposite sense. Pressure perturbations generated in the combustor

modulate the airﬂow in the premixing duct, but not the fuel ﬂow; hence there is a

strong coupling between heat release and pressure. This may end up with serious

consequences if the local conditions allow the amplitude of the pressure wave to grow

in time. This includes ﬂashback in the premixing duct, damage to the combustor, and

damage to sensitive external components such as fuel and oil pipes, which is due

to vibration, leading to fracture. Every lean-burn system with which the author

has been involved during the past 20 years has shown a propensity to such cyclic

instability. An important aspect of the problem is that there are several possible

modes of instability. Some occur only in combustion rigs, some in engines, and

these instabilities often depend on the inlet and outlet acoustic boundary conditions.

Engine combustors are choked at outlet, and rig combustors operate with an outlet

Mach number of around 0.3 and exhaust into a large volume. A rig combustor is

usually fed with air from a large plenum. An engine combustor is closely coupled to

the compressor. Thus instabilities may appear and seriously hinder rig development,

which would not occur in an engine, and instabilities may appear for the ﬁrst time

when the combustor is engine tested. The risk must be assessed and included in the

engine demonstration and product development programme. CFD tools can help

the product development programme by answering many ‘what if ...’questions. A

good example of this is in the previous chapter on thermoacoustic instability. One

should, however, understand that the quality and reliability of a CFD solution very

much depends on the ability of various submodels to capture the relevant physical

processes correctly and accurately.

4.2.9 Conclusions

The introduction of lean DI combustion systems can be expected to reduce NO

about 40%. For this technology to substantially permeate the entire world ﬂeet will

take at least 25 years. The only way in which NO

can be further reduced is by the

332 Lean Flames in Practice

application of lean, premixed, prevapourised combustion. This may be feasible in

engines of up to 20:1 pressure ratio, including supersonic civil aircraft, should these

ever again be built. It would be wise to gain service experience in such engines before

attempting to apply the technology at higher pressure ratios.

All lean-burn combustion systems require staged fuelling and thus add com-

plication, weight, and cost and tend to reduce reliability. Reliability of engines for

large aircraft is not only a safety concern but a concern for dealing with very large

numbers of stranded passengers at airports and the rescheduling of their ﬂights. One

may think that these are problems to be dealt with by airlines, but inevitably such

issues have large economic implications for engine manufacturers.

Although future increases in pressure ratio and temperatures will exacerbate

the emissions problem, it is likely that compressor-disk material will limit the OPR

to around 55 for several years. NO

cannot be controlled if the TET exceeds about

1850 K. Because the optimum TET is linked to pressure ratio and bypass ratio,

emissions control may limit engine cycles within the next 10 years.

REQUIRED EXPERIMENTAL RESEARCH EMPHASIS. It should be clear from the forego-

ing discussion that emissions reduction in aircraft gas turbines requires the consid-

eration of all aspects of combustor performance and fuel injector design. Issues like

altitude relight and thermal management are perceived by the research community

to be the concern of engine designers and manufacturers, and thus to belong to work

around technology readiness level 5.

The author believes that these issues need

to be evaluated at least theoretically at technology readiness level 3 or 4, so that

research may be directed to address the areas of greatest risk or uncertainty.

With lean-burn systems, the injector is of paramount importance, as it controls

the entire combustion process. A single research project may focus on a very small

area of the subject, but the researcher should understand how his or her work

relates to the total problem. Collaboration among universities, research institutions,

and industry is essential if the aim is to make improvements in environmental impact.

The improvement will come from applications in engines, and it is in the transition

from the simpliﬁed research rig to the engineering of a reliable product that all may

be lost. Simpliﬁcation and compromise are necessary in research, but there is a point

at which simpliﬁcation renders the experiment worthless.

For example, the investigation of the condition of fuel leaving a premix duct

by the application of phase-Doppler anemometry on a cold, atmospheric rig, using

water instead of fuel, would be a complete waste of time. Droplet mass and fuel–

air momentum ratio are nowhere near engine values, droplet trajectory bears no

relation to a practical case, there is no evaporation, and most of the atomised liquid

runs down the walls of the duct and is inefﬁciently reatomised off the outlet inner

and outer rims. If there is swirl in the duct, the liquid will be centrifuged onto the

outer wall. Yet to conduct such an experiment with kerosene at conditions for which

auto-ignition is possible is unsafe. Good experiments that are within the capability of

university research departments and that offer adequate access for instrumentation

can be designed through conversation between engine combustion specialists and

TRL, NASA deﬁned technology Readiness Levels from 1 to 9 [215], and their deﬁnitions are now

widely accepted. The concept is very useful for controlling ﬁnancial risk.