Lefebvre A.H., Ballal D.R. Gas Turbine Combustion: Alternative Fuels and Emissions

Подождите немного. Документ загружается.

90 Gas Turbine Combustion: Alternative Fuels and Emissions, Third Edition

Sovran and Klomp tested more than 100 diffuser geometries, all having an

inlet radius ratio within the range of 0.55–0.70. The experimental program was

conducted with inlet Mach numbers less than 0.30, Reynolds numbers from

4.8 × 10

to 8.5 × 10

, and a single inlet velocity prole. Their measurements of

diffuser effectiveness show broad agreement with those obtained by Reneau

et al. [5] for two-dimensional diffusers. Analysis of the experimental data over

a range of area ratios from 1.4 to 3.0 shows that the optimal value of nondimen-

sional length, corresponding to the

line, may be approximated as

Lh AR

51=−

(

)

(3.23)

Sovran and Klomp concluded that area ratio and nondimensional wall

length are the main determinants of optimum geometry for straight-walled

annular diffusers. Thus, the optimum geometry for a xed wall length

occurs at an area ratio that is virtually independent of the combination of

wall angles and radius ratios employed. This clearly simplies the treatment

of annular-diffuser geometries.

Performance charts for annular diffusers of the type obtained by Sovran and

Klomp [1] and Takehira et al. [14] are in widespread use, but their application

is limited to ow situations where the inlet boundary layer has been allowed

to develop naturally in approach sections. The data presented in these charts

tend to be pessimistic in relation to the performance actually achieved on the

engine [9]. Lakshminarayana and Reynolds [16] have shown that when an

annular diffuser is sited downstream of a compressor rotor, there is a redis-

tribution of radial turbulence energy that delays separation and improves the

radial velocity prole. These improvements allow larger area ratios to be used

than those of the optimum geometries given by standard performance charts.

3.6 Effect of Inlet Flow Conditions

From the very earliest studies on diffusers, dating back almost a century, it

was realized that inlet conditions have an important inuence on the subse-

quent development of ow within the diffuser. Most of these early studies

were conned to conventional diffuser geometries, but more recent work

has shown that the performance of gas turbine diffusers is also strongly

affected by inlet ow conditions. For example, Stevens et al. [17,18] and Klein

[19] have demonstrated the inuence of the wakes produced by compressor

outlet blade vanes on diffuser performance, whereas Lohmann et al. [20] and

Carrotte et al. [21] have examined the effects of inlet swirl on ow conditions

within and downstream of various prediffuser congurations. In addition,

many combustors contain struts, either within the prediffuser or between its

Diffusers 91

exit plane and the liner dome, which are there mainly for structural reasons,

but also to accommodate supply lines to the bearing compartment. The pres-

ence of these struts in the compressor efux creates more wakes in the ow.

Another important parameter inuencing ow stability and diffuser

performance is the radial distribution of velocity at the compressor exit. On

many aircraft engines, the compressor outlet velocity is both peaked and

asymmetric and, therefore, subject to appreciable variation with changes in

aircraft speed and altitude. Radial velocity proles can also vary between

development and production engines and even among engines from the same

production line. Usually, the velocity prole at the diffuser inlet peaks toward

the outer diffuser wall. In any case, the propensity for separation is enhanced

at the low-velocity wall and suppressed at the high-velocity wall [9]. Faired dif-

fusers are especially susceptible to the adverse effects on ow stability arising

from variations in inlet radial velocity prole and this is the main reason why

they are no longer favored for application to modern engine combustors.

Although the radial velocity prole is of paramount importance, many

other inlet ow parameters can have a signicant effect on both diffuser

and overall combustor performance. They include Reynolds number, Mach

number, turbulence, and swirl.

3.6.1 reynolds Number

The inuence of Reynolds number is most pronounced when the inlet

boundary layer is not fully developed. Under these conditions, an increase

in Reynolds number improves performance by reducing the boundary-

layer thickness and increasing the turbulence level. At Reynolds numbers

larger than 3 × 10

at the diffuser inlet, the performance of conical diffus-

ers is insensitive to variations in Reynolds number [22]. For annular diffus-

ers, Reynolds number has little or no effect on performance for Re > 5 × 10

(based on the hydraulic mean diameter at inlet). Only a limited amount of

data is available for dump diffusers, but Hestermann et al. [23] found no

effect of Reynolds number on performance for values of Re in the range from

9.2 × 10

to 1.6 × 10

for both small and large dump gaps.

Typical values of Reynolds number in gas turbines are in the order of sev-

eral millions and turbulence levels are relatively high. Therefore, combustor

diffusers are unlikely to be Reynolds-number sensitive. However, as pointed

out by Adkins [24], care must be taken during component testing when air

densities may be close to normal ambient and Reynolds numbers are much

lower than on the engine. Measurements of performance carried out at such

conditions could give pessimistic results.

3.6.2 Mach Number

The ow characteristics and performance of diffusers are insensitive to Mach

number when it is below around 0.3. Between this value and 0.6, performance

92 Gas Turbine Combustion: Alternative Fuels and Emissions, Third Edition

can actually improve slightly because much of the pressure recovery occurs

nearer the diffuser inlet, thereby providing more opportunity for settling of

the ow in the downstream portion of the diffuser. Above a Mach number of

0.6, the pressure gradient near the inlet of a straight-walled diffuser becomes

excessive and the performance begins to deteriorate. Large-scale separation

can occur when the Mach number approaches 0.7, and the performance falls

off dramatically. As combustor diffusers always operate at Mach numbers

below 0.4, these compressibility effects have little practical signicance.

3.6.3 Turbulence

Moore and Kline [25] were among the rst to examine the inuence of turbu-

lence on diffuser performance. Their tests on two-dimensional congurations

showed that turbulence has little effect on the line of rst stall; however, the

divergence angle at which fully developed stall occurs is signicantly increased,

as shown in Figure 3.8. Later work by Hoffmann et al. [26–28] on two-dimen-

sional diffusers conrmed these early ndings. Their results show that increas-

ing the turbulence intensity up to 3.5% can improve performance markedly.

For a diffuser Ν/W

ratio of 15 and a divergence angle of 20°, an increase in

turbulence level raised the pressure recovery coefcient from 0.58 to 0.71.

L/W

High inlet turbulence

Low inlet turbulence

4816 25

No separation

Large transitory

stall

Two-dimensional

steady stall

Hysteresis

Jet flow

100

200

300

Divergence angle, 2θ

Figure 3.8

Regions of two-dimensional plane-wall diffuser ow. (From Moore, C.Α. and Kline, S.J., “Some

Effects of Vanes and of Turbulence in Two-Dimensional Wide-Angle Subsonic Diffusers,”

NACA TN 4080, 1958. With permission from NASA.)

Diffusers 93

Similar performance improvements from raising turbulence levels are

also observed in annular combustors. For example, Stevens and Williams

[15] investigated the performance of an annular diffuser with con-

stant inner diameter and found that increasing the turbulence intensity

improved ow stability at the diffuser exit. Recent work by Hestermann

and Wittig [23] has shown that combustor dump diffusers also exhibit this

same tendency.

In general, by delaying ow separation in the prediffuser and attening

the velocity prole at exit, an increase in turbulence level enhances pred-

iffuser performance and also improves ow conditions into the inner and

outer annuli.

Compressor-generated turbulence seems to be especially benecial to dif-

fuser performance. It produces improvements that are even larger than those

achieved by grid- or spoiler-generated turbulence of the same intensity. This

is attributed to the special type of turbulence created by the compressor,

which has a large component in the radial direction [16].

3.6.4 Swirl

Some degree of swirl is normally present in the efux from compressors. It

is generally regarded as undesirable due to its deleterious effect on tempera-

ture pattern factor at the turbine inlet, although it has been suggested that in

wide-angled diffusers, swirl can sometimes prove advantageous by inhibit-

ing ow separation. Comparatively little evidence is available on the effects

of swirl on diffuser performance, but Lohmann et al. [20] have shown that

inlet swirl angles can increase as the ow is diffused due to the decrease in

axial velocity and radius that is experienced by the ow passing to the inner

annulus. Quantitative evidence in support of this nding has been provided

by Carrotte et al. [21], who carried out an experimental investigation on sev-

eral modern dump diffuser congurations with inlet conditions being gener-

ated by an axial-ow compressor. For all the geometries tested, their results

show how the presence of a small amount of inlet swirl (≈3°), typical of that

in a gas turbine engine, can result in large swirl angles (≈15°) being generated

further downstream, especially in the ow within the inner annulus.

3.7 Design Considerations

3.7.1 Faired Diffusers

Many gas turbine combustors now in service employ a “faired” type of dif-

fuser, in which the main objective is to achieve a gradual reduction in ow

velocity without inducing ow separation (stall) and with minimum loss in

94 Gas Turbine Combustion: Alternative Fuels and Emissions, Third Edition

pressure. A typical faired diffuser for an annular combustor is shown in

Figure 3.9. Note in this gure that a section is located at the combustor inlet

where the ow velocity is maintained sensibly constant at the compressor

outlet value. This section provides a settling length whose primary function

is to dissipate the large radial and circumferential ow distortions that exist

in the compressor efux due to the wakes of the compressor outlet guide

vanes and to the secondary vortices of the last row of rotor blades. To allevi-

ate the adverse effects of these ow distortions on diffuser and combustor

performance, Lefebvre [29] recommends that the settling length should be

twice the chord width of the exit blade or vane.

The need for a settling length has not found universal acceptance. Based on

the results of his own research, and that of other workers, Klein [9] has argued

that a settling length is not required because, although the mixing losses asso-

ciated with the decay of ow distortions create high losses in the prediffuser,

they also generate a fairly uniform velocity prole at the diffuser outlet, which

results in lower losses and more stable ow conditions in the regions down-

stream. According to Klein, inlet ow distortions decay more rapidly when their

level is high. Thus, a settling length offers no advantage because the resulting

weaker ow distortions entering the prediffuser will decay at a slower rate.

These observations of Klein are of considerable practical interest and

importance, but they are relevant only to ow in a single channel. If the ow

is divided, either by vanes or the combustor snout, instabilities could arise

due to unequal diffusion down the separate channels. Further information

on the role of the settling length may be found in Klein [9].

A faired diffuser incorporates at least three separate diffusion regions, as

illustrated in Figure 3.9. The rst region is located immediately downstream

of the settling length (if any), and its purpose is to achieve some reduction in

velocity, typically around 35%, before the air reaches the “snout,” which is

formed by two cowlings that extend upstream from the liner dome. At the

snout, most of the air is subjected to a change in direction as it enters the dif-

fusing passages, which are created in the space between the cowlings and the

combustor casing. It is important that this change in ow direction be carried

Regions of

diffusing flow

Settling length

Snout

Figure 3.9

Faired diffuser.

Diffusers 95

out at constant velocity to reduce the possibility of breakaway at this point.

Some rounding of the snout lip can be helpful, especially if the compressor-

outlet velocity prole is subject to variation with changes in engine operating

conditions. However, care must be taken to ensure that it creates no sudden

local increase in divergence angle; otherwise, ow separation may be initi-

ated. Downstream of the ow split produced at the snout is a second diffusion

region. From this region, air ows around a bend at constant velocity into the

third diffusion region, which is formed by the annular space between the

liner and the casing. Beyond this region, further diffusion takes place as more

air ows into the liner, thereby reducing the ow velocity in the annulus.

Many faired diffusers incorporate a fourth diffusion region that affects

only the center portion of the air entering the snout. In early combustor

designs, the snout air comprised around 10% of the total combustor airow.

Today, due to the large and increasing air requirements of airblast atomizers

and other fuel preparation devices for low pollutant emissions, this gure

is appreciably higher. It is customary to t a short annular diffuser at the

entry to the snout, as illustrated in Figure 3.10. This allows the snout air to

be further decelerated before being “dumped” into the dome region. The

primary purpose of this additional diffusion zone is to provide a higher and

more uniform pressure in the air supply to the atomizers, air swirlers, and

dome-cooling slots.

Clearly, a basic requirement of faired diffusers is that all diffusion regions

must perform without stall. Lines of rst stall for conventional conical,

two-dimensional, and annular congurations are shown in Figure 3.3.

Pressure losses in each diffuser section may be calculated using perfor-

mance charts of the type presented in References [1–9,30,31] and illustrated

in Figures 3.5 and 3.7.

The faired diffuser comprises a number of subdiffusers whose pas-

sage heights are quite small. In consequence, their dimensions, and hence

Diffusing passage

Plenum chamber

Figure 3.10

Illustration of pressure-balancing slots.

96 Gas Turbine Combustion: Alternative Fuels and Emissions, Third Edition

also their effective area ratios, are especially subject to manufacturing

variations, differential thermal expansion between the liner and casing,

and the thermal distortion that occurs with changes in engine operating

conditions. Another drawback is that the distribution of air between the

inner and outer annuli, which governs the ow distribution throughout

the liner, is very sensitive to variations in inlet velocity prole. The effect

is minor when the inlet velocity is low, because the fraction of total pres-

sure that is tied up in kinetic energy is so small that the static pressure

drop across the combustor effectively controls its airow distribution.

In consequence, the airow pattern tends to remain constant, regardless

of variations in inlet velocity prole. However, when the inlet velocity

is high, a signicant proportion of the total pressure is in the form of

dynamic pressure, and the static pressure drop across the combustor is

correspondingly lower. Under these conditions, it is the distribution of

dynamic pressure in the inlet airstream that controls the airow distri-

bution throughout the combustor. For a at or symmetrical velocity pro-

le, this would present no special problems because the extra velocity

could readily be accommodated by an increase in diffuser pressure loss.

Unfortunately, an increase in compressor outlet velocity is almost invari-

ably accompanied by a deterioration in velocity prole, to the extent that

often makes it virtually impossible to achieve a balanced airow distribu-

tion inside the liner.

To balance the aerodynamic ow pattern within the liner, it is necessary

to attain symmetry in terms of the mass ows of the opposing jets, the pen-

etration of these jets, and their momentum. If the air entering the combus-

tion chamber has a at velocity prole, it is possible to achieve matching, or

symmetry, on all three counts. However, if the velocity prole is distorted,

i.e., peaked either to the outside or the inside of the diffuser centerline,

then one can achieve symmetry only with two of the three parameters.

The designer can choose which two to balance by suitable alterations to

passage areas and by modications to liner hole sizes, but the fact remains

that a distorted velocity prole at the combustor inlet always results in an

unbalanced liner airow pattern.

To some extent, the problems described above can be alleviated by arrang-

ing for the snout diffuser to accept an airow that exceeds its normal require-

ments. The excess air then ows into the inner and outer passages through

short slots, as illustrated in Figure 3.10. With this conguration, the snout acts

as a plenum chamber, supplementing the airow in both liner passages—but

especially in whichever passage is decient in air due to a shift in inlet veloc-

ity prole. Another drawback to faired diffusers is the risk of air leakage

between the burner feed arm and the hole in the combustor snout through

which it is inserted.

In summary, the faired diffuser has the great advantage of low pressure

loss, some one-third less than that of the dump diffuser. This important

Diffusers 97

attribute is, however, more than offset by the following drawbacks that vir-

tually prohibit its application to modern aircraft engines:

1. Excessive length

2. Performance and ow stability are very sensitive to changes in inlet

velocity prole

3. Performance is very susceptible to thermal distortions and manufac-

turing tolerances

3.7.2 Dump Diffusers

The annular dump diffuser concept is illustrated in Figure 3.11. The unit

consists of a short conventional diffuser in which the air velocity is reduced

to about 60% of its initial value. In complete disregard for conventional dif-

fuser design principles, the air is then “dumped” and left to divide and ow

around the liner dome before entering the two annuli that surround the liner.

The standing vortices created by projection of the prediffuser walls into the

dump region help to maintain a uniform and stable division of ow around

the liner. Due to the sudden expansion at the prediffuser outlet, the dump

conguration has an inherently higher pressure loss than the faired diffuser,

typically by around 50%. However, this penalty is more than compensated

by substantial savings in length and weight. These attributes are especially

advantageous for aircraft applications. The dump diffuser has the further

important advantage of producing a stable ow pattern, which is fairly insen-

sitive to manufacturing tolerances, differential thermal expansions between

the liner and combustor casing, and variations in inlet velocity prole.

In an early study on dump diffusers, Fishenden and Stevens [32] examined

the inuence on performance and stability of diffuser geometry, distance

Prediffuser

Outer annulus

Inner annulus

Figure 3.11

Dump diffuser.

98 Gas Turbine Combustion: Alternative Fuels and Emissions, Third Edition

between prediffuser outlet and liner dome (dump gap), and the division of

ow between the inner and outer annuli. They found that nearly all the static

pressure rise occurs in the prediffuser, and most of the loss in total pressure

occurs in the dump and settling regions. It was also found that the presence

of the liner has a benecial effect on the performance and stability of the ow

in the prediffuser. In particular, it tends to suppress ow separation, which

allows divergence angles and area ratios to be made larger than those of

conventional annular diffusers of the same relative length. Hestermann et al.

[23] also found that ow separation in the prediffuser could be suppressed

by reducing the dump gap. The larger the divergence angle, the smaller the

dump gap must be in order to prevent ow separation. For a divergence angle

of 22°, they found that ow separation could only be suppressed by reducing

to below unity. However, as Carrotte et al. [21] have shown, if the dump

gap is made too small, the overall dump diffuser pressure loss will increase

due to excessive local ow acceleration and an increase in ow turning and

curvature around the liner dome. These considerations suggest that for any

given diffuser conguration there will be a certain value of dump gap ratio

), depending on the amount of prediffusion and the liner depth ratio,

at which the overall diffuser pressure loss is a minimum. This is conrmed

in the experimental data of Honami and Marioka [33]. These data also show

that the optimum value of d

decreases with an increase in the prediffuser

divergence angle.

3.7.2.1 Inﬂuence of Liner Depth Ratio

For aircraft applications, the constant desire to reduce engine length and

weight has increased the demand for shorter length diffusers. At the same

time, the requirements of lower lean blowout limits and better relight capa-

bility at high altitudes, along with the continuing pressure to reduce pollut-

ant emissions, have created a trend toward liners of increasing depth. For

example, the early studies of Fishenden and Stevens [32] were conducted on

a diffuser that had a liner depth ratio (dened as the ratio of the liner depth

to the passage height at the prediffuser inlet, D

) of 3–5, whereas in more

recent work by Carrotte et al. [21], the liner depth ratio was 5.5. An increase

in this ratio results in larger amounts of ow turning and curvature around

the liner dome, which, as discussed above, impairs diffuser performance by

raising the pressure loss.

The measurements of Srinivasan et al. [34] on a 60°-sector dump dif-

fuser model that incorporated a cowl to pass 20% dome ow indicated a

60% increase in pressure loss when the liner depth ratio was increased from

3.1 to 4.1. Stevens and Wray [18] also observed a much higher loss than had

been measured in previous work by Fishenden and Stevens [32], which they

attributed to an increase in liner depth ratio from 3.5 to 5.5.

In an attempt to clarify the inuence of liner depth on diffuser perfor-

mance, Klein [9] examined measured values of loss coefcient, λ, from several

Diffusers 99

sources, all relating to dump diffusers of similar design and the same dump

gap ratio of unity. The results of this study indicate a general increase in λ of

around 60% as the liner depth ratio is raised from 3.5 to 5.5. It is clear, there-

fore, that liner depth ratio should be regarded as an important geometric

parameter affecting diffuser performance.

3.7.3 Splitter Vanes

The notion of using splitter vanes to create a multiple passage diffuser of

large total divergence angle is by no means new, and was demonstrated

successfully by Cochran and Kline [35] as early as 1958. In their design,

the vanes were arranged symmetrically, as shown in Figure 3.12, and their

length and divergence angle were chosen so that each vane passage operated

near the line of rst stall. By this means, near-optimum pressure recoveries

and unstalled ow were obtained for total divergence angles of up to 42°.

The vanes were also found to improve the outlet velocity prole, making it

more uniform. Subsequent tests by Miller [36] on three-vane diffusers dem-

onstrated satisfactory performance up to a total divergence angle of 50°.

The combustor prediffuser of the NASA/GE “Energy Efcient Engine”

employs a single splitter vane to achieve the desired large area ratio of 1.8

and to direct the ow toward the two dome regions as well as to the inner

and outer annuli. The ow split between the two diffusing passages is such

that the inner duct passes 52% of the total air mass ow and the outer duct

48%. The presence of the splitter vane allows a length reduction of 50% rela-

tive to a single annular diffuser of the same area ratio. The overall pressure

loss coefcient for the entire combustor inlet is 0.30, which corresponds to

35% of the total combustor pressure loss [9,37].

The General Electric LM6000 dry low-emissions combustor, shown in

Figure 9.30, requires almost twice the volume of the conventional combustor

it replaces in order to meet its emissions goals. The use of three splitter vanes

Figure 3.12

Diffuser incorporating splitter vanes.