Professional Documents
Culture Documents
Limitations On Gas Turbine Performance Imposed by Large Turbine Cooling Flows
Limitations On Gas Turbine Performance Imposed by Large Turbine Cooling Flows
2000-GT-635
+ ++
J H Horlock , D T Watson and T V Jones*
ABSTRACT NOTATION
Calculations of the performance of modern gas turbines usually
include allowance for cooling air flow rate; assumptions are made for the
amount of the cooling air bled from the compressor, as a fraction of the A area
mainstream flow, but this fractional figure is often set in relatively B Biot number [ht/k]
arbitrary fashion. cp specific heat at constant pressure
There are two essential effects of turbine blade cooling: C,C’ constants
[i] the reduction of the gas stagnation temperature at exit from the h heat transfer coefficient
combustion chamber [entry to the first nozzle row] to a lower stagnation K constant
temperature at entry to the first rotor and M Mach number
[ii] a pressure loss resulting from mixing the cooling air with the p pressure
mainstream. Qnet net heat transferred
Similar effects occur in the following cooled blade rows. R gas constant
The paper reviews established methods for determining the amount of r pressure ratio
cooling air required and semi-empirical relations, for film cooled blading t thickness [of thermal barrier coating]
with thermal barrier coatings, are derived. Similarly, the pressure losses St Stanton number
related to elements of cooling air leaving at various points round the blade T temperature
surface are integrated over the whole blade. This gives another semi- V velocity
empirical expression, this time for the complete mixing pressure loss in w mass flow
the blade row, as a function of the total cooling air used. w+ non-dimensional quantity defined in equation [4]
These two relationships are then used in comprehensive calculations of W+ non-dimensional quantity defined in equation [10]
the performance of a simple open-cycle gas turbine, for varying x r (γ − 1)/γ
combustion temperature and pressure ratio. These calculations suggest that y ratio of velocities
for maximum plant efficiency there may be a limiting combustion ε0 overall cooling effectiveness
temperature [below that which would be set by stoichiometric εF film cooling effectiveness
combustion]. For a given combustion temperature, the optimum pressure η cooling efficiency
ratio is reduced by the effect of cooling air. ηc compressor isentropic efficiency
ηt turbine isentropic efficiency
ηth thermal efficiency
ξ coolant fraction [of mainstream gas flow]
φ angle of discharge of coolant
ψ quantity defined in equation [14]
µ quantity defined in equation [11]
ρ density
so that
[b] [wc/wg] = λ[cpg/cpc ] [hg /cpg ρgVg] (Tgi – Tbl)/ η (Tbl – Tci)
= λ[cpg/cpc ]Stg (Tgi– Tbl)/η (Tbl – Tci) [3]
Tgi and Tci are usually determined from and/or specified for the cycle
Fig 1 [a] Simple model for convective cooling [b] Simple model for film calculation so that the cooling effectiveness, ε0, implicitly becomes a
cooling [c] Simple model for transpiration cooling requirement (subject to Tbl which again can be assumed for a ‘level of
technology’).
analysis [which allows for external film cooling as well as internal If η and C are amalgamated into a single constant K then:
convective cooling] is given below. It is based on the assumption that the
external Stanton number [Stg], which is generally a weak function of the ξ = Kε0/ (1 − ε0) [6]
Reynolds number, remains constant as engine design parameters [Tcot and
r] are changed. But a variation of this approach, based on a constant
With η = 0.7 and C = 0.035 say, K = 0.05 and equation [6] then correlates
external heat transfer coefficient [hg], is also described in detail there.
well with values used by El-Masri [1987] and [1988].
A second less conventional model is to assume that hg is constant and
3.1 Convective Cooling Only
that the specific mass flow is in a choking condition so that, taking the
A simple heat balance for a typical convectively cooled blade [as
inlet gas conditions as representative,
illustrated in Fig 1a, which shows the notation] is
Qnet = wccpc (Tco - Tci) = wgcpg (Tgi – Tgo) = hg Asg (Tgi – Tbl) [1] ρgVg = [2/(γg + 1)] (γg + 1 ) / 2(γ – 1)
g ] pg(γg /RgTg)1/2
Here it is assumed that the temperature of the coolant does not fully It then follows that
0.05
0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
COOLING EFFECTIVENESS
For a cooled nozzle guide vane row, with internal cooling efficiency
η = 0.7, Fig. 2 shows plots of ξ against cooling effectiveness ε0, based on
equations [4] and [7a] and the values of C and C’ given above. Equation
[7a] also requires engine parameters, Tg and pg = r (bar) to be specified and
it was assumed that they were related by an approximate equation given
later [eqn [21] for maximum efficiency conditions. The pressure ratio r Fig 3 The Shapiro/Hartsel model for simple mixing
gives the cooling temperature T2 and the assumption of a blade
temperature of 1073K gave the cooling effectiveness. The two expressions Hartsel developed Shapiro’s table of influence coefficients to allow for
thus give quite similar forms for the cooling air fractions, but they will a difference between the total temperature of the injected flow [Tc] and the
differ as conditions move away from the optimum conditions for mainstream [Tg]:
maximum efficiency. We discuss later how well they compare with
practical designs. ∆pt/pt = - ξγMg2/2 { 1 + (Tc/Tg) – 2y cos φ} [17]
For film cooling
[i] ξ = CW+ [12a] Here y is the ratio of the velocity of the injected coolant and that of the
with C ≈ 0.03 free stream [y = Vc/Vg], Mg is the Mach number of that free stream and
or [ii] ξ = C’Tg1/2 W+/pg [13a] φ is the angle at which the cooling air enters the mainstream.
with C’ ≈ 0.0246. Clearly the value of y is critical; an approximation suggested by
For transpiration cooling Hartsel is to take [pt]c = [pt]g , so that Vc/Vg ≈ (Tc/Tg)1/2 . A more
ξ = C ln [1/(1 - ε0)] [16a] sophisticated approach would not assume the total pressures of coolant
and mainstream to be the same, but would derive them by accounting for
with C = 0.03.
losses as follows:
Since film cooling is now used in most gas turbines, we have adopted in the mainstream, ∆ptCC in the combustion process, and ∆ptN in the nozzle
the form of equations [12a] and [13a] for most of our subsequent row itself;
calculations. However, it was found that the cooling flows calculated from for coolant air tapped at pressures other than the compressor discharge
these equations are less than those used in recent and current practice in pressure, losses ∆ptD [in the ducting and disks before coolant enters the
which film cooling is employed. This is for two main reasons. blade itself], and ∆ptBL [in the blading heat transfer process].
[i] Designers are conservative, and choose to increase the cooling flows Discharge at any local static pressure would then give different Mach
[a] to cope with entry temperature profiles [the maximum temperature numbers Mc and Mg and different velocities Vc and Vg
being well above the mean] and local hot spots on the blade and El-Masri, Shrivastava and Maccallum [1987], and Day and Oldfield
[b] locally where cooling can be achieved with relatively small penalty on [1999] have used cascade data with and without coolant discharge to
mixing loss [and hence on polytropic efficiency – see below]. Thus obtain experimental relationships between ξ and the extra loss arising
regions remote from these injection points are cooled with this low loss from injection of the cooling air [but note that injection near the trailing
air. edge causes little total pressure loss - it may even reduce the basic loss in
[ii] In practice some surfaces in a turbine blade row will be convective the wake]. The basic uncooled loss is included in the value assumed for
cooled with no film cooling. The use of equation [12a] or [13a] for the polytropic efficiency in the subsequent cycle calculations; there is no
whole blade row assembly will therefore lead to the total cooling flow double counting involved in adding an extra penalty associated with the
being underestimated. [Film cooling leads to more efficient cooling, which mixing loss arising from cooling air injection into the mainstream.
is reflected in W+ being much less than w+; for the NGVs of a modern gas In practice the cooling flow leaves the blade surface at various points
turbine W+ may take a value of about 2 but w+ may be about 4]. round the blade profile [say an elementary amount dξ at a particular
third blade rows, it was assumed that the pressure ratio of the HP turbine
remained constant [rHP = 3] for varying r and Tcot, this figure reflecting
THERMAL EFFICIENCY
WB+ the first stage rotor row [the relative stagnation temperature], and 2073
hence the cooling fraction ξB, was assumed to be the arithmetic mean of 2273
the entry combustion temperature Tcot and the exit absolute temperature TE 41
(for an uncooled turbine stage of 50% reaction this relationship may be
shown to be exact). An alternative would have been to take absolute
stagnation temperature Trit from the iterative calculation and then
40
determine the relative stagnation temperature from representative velocity 15 20 25 30 35 40 45 50 55
triangles, but sample calculations showed that taking TgB as [Tcot + TE]/2 PRESSURE RATIO
was a good approximation even for the cooled turbine stage.
With the cooling fractions for all three rows thus obtained, the
corresponding mixing losses were obtained from the approximate Fig 4 Thermal efficiency as a function of combustion temperature Tcot and
integrated equation [18b]. It was argued that the mixing in the second and pressure ratio r [for cooling mass flow ratio ξ = 0.045W+]
third rows together constituted a pressure loss at exit from the HP turbine.
Equation [20] shows that both the entry and exit losses can be converted
into an efficiency reduction for the HP turbine. Numerical experiments
showed that virtually identical plant thermal efficiencies were obtained by
500
Tg = 1473
T g = 1673 48%
400
Tg = 1873
Tg = 2073 46% 2200 K
300
Tg = 2273 2000 K
44% 1800 K
THERMAL EFFICIENCY
200
42%
1600 K
Fig 5 Specific work as a function of combustion temperature Tcot and pressure 36%
32%
It is seen that for each value of Tcot , the thermal efficiency peaks at a 30%
pressure ratio of about 35. The maximum value of efficiency increases 0 10 20 30 40 50 60
PRESSURE RATIO
with Tcot up to 1900 K - 2000 K but then decreases slightly. These results
are surprising, in marked contrast to uncooled cycle calculations, in which Fig. 7 Thermal Efficiency for four cooled rows [first stage cooled from
efficiency continues to increase with Tcot for all pressure ratios. Some compressor discharge, second stage cooled by lower pressure air]. Convective
corresponding values of specific work [SW, kJ per unit of exhaust gas flow] cooling, ξ = 0.045 w+, calculations with Provision code.
are shown in Fig. 5. SW increases with Tcot for all pressure ratios.
The effect of changing the assumptions for determining coolant flows 6.3 The Relationship between Tcot and Trit
is illustrated in Fig. 6 for r = 30, which also shows the uncooled efficiency
for that pressure ratio. Use of the film cooling equations [12a], with C = In practice, it is usually argued [e.g.Walsh and Fletcher [1998]] that the
0.045, and [13a], with C’ = 0.037 [but with pg taken as p2 for all three cycle efficiency is strongly dependent on the first rotor inlet temperature
Trit [a fact emphasised by the frequent quotation of Trit as the maximum
cooled blade rows] leads to similar values of thermal efficiency at low Fig
turbine temperature]. Walsh and Fletcher suggest an empirical correlation
Tcot; but the second assumption, implying hg constant, gives lower values
between Trit and pressure ratio ropt at maximum efficiency, which may be
of efficiency at higher Tcot. The assumption of “ultimate” transpiration approximated to
cooling [equation [16a] gives thermal efficiencies close to the uncooled ropt = 16.4 [Trit/1300] 5/2 [21]
values, and represents an objective for designers, but one that is strictly
limited. This dependency on Trit is simply illustrated by an “air standard” analysis of
a cooled cycle [i.e. one in which a perfect gas is assumed as the working fluid],
6.2 Calculations Using the Provision Code with a cooled first inlet guide vane row. The thermal efficiency [ηth] at a
The Provision code has the facility for using cooling air tapped along particular combustion temperature, Tcot, may be shown to be the same as the
the compressor. It was used to determine the thermal efficiency with low efficiency of an uncooled cycle with the same rotor inlet temperature Trit, as
pressure air used to cool the second stage – third and fourth blade rows. follows. The cooled cycle efficiency is given by
Otherwise the assumptions were as in the earlier calculations, except that
convective cooling was assumed, and equation [4] used with C = 0.045. ηth = [wc + wg]{ηtTrit [1 – (1/x)] – T1[x – 1]/ηc}/wg [Tcot – T2] [22]
Fig. 7 shows the results obtained. Now the therrnal efficiency is
generally higher because the exergy loss in throttling the cooling air has
where x = r [γ − 1]/γ , ηc and ηt are the isentropic efficiencies of the
been eliminated, and less is used because it is at a lower temperature.
compressor and turbine respectively, wc is the amount of cooling air used
Maximum efficiency now occurs at slightly higher Tcot because of these
per unit air flow through the compressor and wg is the air heated to Tcot
effects, but the main point remains – there would appear to be a limit on
[i.e. (wg + wc) = 1 is the flow through the turbine after mixing at exit from
50 the NGVs].
But application of the steady flow energy equation across the cooled
45
NGVs yields
THERMAL EFFICIENCY
30
0.045W * Substitution of equation [24] into equation [22] gives
0.037[Tg1/2/pg] W *
ηth = {Trit ηt[1 – (1/x)] – T1[x – 1]/ηc}/ [Trit – T2] [25]
25
1000 1200 1400 1600 1800 2000 2200 2400 2600 2800