You are on page 1of 10

Proceedings of the Institution of Mechanical Engineers, Part A: http://pia.sagepub.

com/Power and Energy Journal of

Aerodynamic loss prediction of axial flow turbine blade rows with coolant injection
D Bohn and T. S. Kim Proceedings of the Institution of Mechanical Engineers, Part A: Journal of Power and Energy 1999 213: 93 DOI: 10.1243/0957650991537464 The online version of this article can be found at: http://pia.sagepub.com/content/213/2/93

Published by:
http://www.sagepublications.com

On behalf of:

Institution of Mechanical Engineers

Additional services and information for Proceedings of the Institution of Mechanical Engineers, Part A: Journal of Power and Energy can be found at: Email Alerts: http://pia.sagepub.com/cgi/alerts Subscriptions: http://pia.sagepub.com/subscriptions Reprints: http://www.sagepub.com/journalsReprints.nav Permissions: http://www.sagepub.com/journalsPermissions.nav Citations: http://pia.sagepub.com/content/213/2/93.refs.html

>> Version of Record - Mar 1, 1999 What is This?

Downloaded from pia.sagepub.com at INDIAN INSTITUTE OF TECH on November 5, 2011

93

Aerodynamic loss prediction of axial ow turbine blade rows with coolant injection
D Bohn* and T S Kim** Institute of Steam and Gas Turbines, Aachen University of Technology, Germany

Abstract: This work presents a calculation routine for the prediction of the aerodynamic loss of axial ow turbines with coolant injection, which can be easily incorporated into throughow calculation programs. The concept of the mixing layer is introduced and the one-dimensional method is used to estimate the local mixing pressure loss. For a real annular nozzle guide vane, calculations are performed with the aid of the mixing model and viscous loss correlations in order to predict loss at the blade mid-height. Predictions for various single-row congurations show favourable agreement with experimental results. Losses are also calculated for the multi-row conguration with a consideration of ow recirculation among cooling holes. It is found that the coolant recirculation may cause a considerable increase in the pressure loss when the amount of net coolant supplied is relatively small. The effect of coolant temperature is also examined for both single-row and multi-row cases. Keywords: axial ow turbine, coolant injection, mixing layer, pressure loss, recirculation Subscripts NOTATION A area c coolant fraction relative to inlet gas ow H total enthalpy m mass owrate M Mach number p static pressure P total pressure R gas constant T total temperature V velocity h k Z x y | ow angle between the main uid and the injected coolant specic heat ratio loss mixing ratio pressure loss coefcient fraction of the mainstream ow in the mixing layer c f g i j m ml ps s ss t ti v 0 1 2 3 coolant front part mainstream gas injection component suction component mixing effect mixing layer pressure side isentropic suction side total trailing edge injection viscous effect without injection blade inlet blade exit fully mixed state

INTRODUCTION

The MS was recei6ed on 14 September 1998 and was accepted after re6ision for publication on 17 December 1998. *Corresponding author: Professor, Institute of Steam and Gas Turbines, Aachen Uni6ersity of Technology, Templergraben 55, 52056 Aachen, Germany. ** Present address: Turbo and Power Machinery Research Center, Seoul National Uni6ersity, Seoul, Korea.
A05498 IMechE 1999

There have been numerous attempts to estimate the performance of axial ow turbines using throughow methods, such as that of the present authors [1]. Most of these studies have been made possible with the aid of many accumulated theoretical and experimental databases. For modern turbines, however, coolant injection
Proc Instn Mech Engrs Vol 213 Part A

Downloaded from pia.sagepub.com at INDIAN INSTITUTE OF TECH on November 5, 2011

94

D BOHN AND T S KIM

makes the ow through blades much more complicated, and this makes it difcult to predict precise aerodynamic loss. Accordingly, the present authors believe that a calculation routine which takes the characteristics of injected coolant into account and which can be easily adopted in the throughow programs is required. Several studies have been made on the topic of developing reliable methods of predicting the aerodynamic loss caused by coolant injection. Most of them were simplied theoretical studies, which were based on the fact that the locally injected coolant owrate is relatively small compared with the main gas owrate. Among others, the main ideas of Hartsel [2] and Prust [3] have been utilized in much subsequent research. This study aims at constructing an estimation routine for the aerodynamic loss caused by the mixing of gas and injected coolant, which serves as part of throughow analysis programs for cooled turbines. Therefore, the aerodynamic calculation routine should not be very complicated in order not to lose the advantages of the throughow method, namely the simplicity, short computing time and diverse applications. In this work, the mixing layer concept which is similar

to those of Hartsel [2] and Ito et al. [4] is introduced. The local mixing pressure loss is estimated by the one-dimensional method. For an annular nozzle guide vane, a mean line calculation using the mixing model and viscous loss correlations is made to predict the pressure loss at mid-height. The prediction results are compared with experiments for various single-row congurations and a fully lm-cooled conguration.

2 2.1

ANALYSIS Coolant injection

Fig. 1

Schematic diagram of the loss calculation model for axial turbine blades with surface injection

The throughow program that has been developed by the present authors can handle both streamline curvature analysis and mean line analysis. It mainly consists of the pressure loss and the ow angle correlations of Kacker and Okapuu [5]. The structure of the program and its validity were explained in the previous paper [1]. The calculation routine of this study for the effect of coolant injection is intended to be easily incorporated into the existing program. Figure 1 shows the main idea of the mixing pressure loss calculation of this study. A layer over each blade surface is dened as the mixing layer, in which mixing between part of the mainstream and the injected coolant takes place. The rest of the ow is called the unaffected mainstream, which is free from any loss. The parameter | represents the ratio of the mass ow of the uid in the mixing layer approaching the blade row to the total mainstream ow at the blade inlet. An explanation of the calculation procedure will be given for the general multi-row injection (injection from the suction and pressure side and also from the trailing edge). The single-row injection can be treated as its simplied case. The unaffected mainstream uid experiences an isentropic ow up to the trailing edge. The ow inside each mixing layer is isentropic where no mixing takes place, i.e. from the inlet to the rst injection row, along all paths between two adjacent rows and from the nal row to the trailing edge. Mixing between the uid inside the mixing layer and the injected coolant occurs at each injection hole exit. At the trailing edge plane, the unaffected mainstream is mixed with three ows: two streams from both surface mixing layers and the ow injected from the trailing edge. Every mixing is assumed to occur at the static pressure of the main uid, which is dened as the surface static pressure for surface injection and the static pressure of the unaffected mainstream for the mixing at the blade exit plane. In this study the one-dimensional method of Shapiro [6] in which the streamwise velocity ratio is taken into account is used for all local mixing process calculations.
A05498 IMechE 1999

Proc Instn Mech Engrs Vol 213 Part A


Downloaded from pia.sagepub.com at INDIAN INSTITUTE OF TECH on November 5, 2011

AERODYNAMIC LOSS PREDICTION OF AXIAL FLOW TURBINE BLADE ROWS

95

This idea is supported by the fact that the mixing mass ow ratio is small enough at each mixing position. Assuming that the mass ow ratio x is small and the injected uid mixes with the mixing layer uid at the local static pressure, the following equation is obtained for the total pressure change of the mixing layer uid, where positive differences mean pressure loss and temperature drop respectively: DP k DT V = M2 +kM 2 1 c cos h x P T Vg 2

(1)

mixing layers and the coolant from trailing edge injection) at the static pressure of the mainstream is applied. It is further assumed that the ow angles of all these streams are equivalent to that of the unaffected mainstream. Finally, the total pressure loss of the mainstream is estimated from the following equation, where each mixing ratio xti, xml,ps and xml,ss is dened as a mass ow ratio of each secondary stream (trailing edge ejected stream, pressure side mixing layer stream and suction side mixing layer stream) to the unaffected mainstream: DP k DT = M2 T P 2 +kM 2

where h is the angle of coolant injection measured from the local main ow direction. After manipulating the equation for the mass owrate of ideal gas, the following equation for the coolant is obtained: mc = Mc

'

1+

kc 1 2 Mc 2

' 



kc pA RcTc c c

(2)

where Ac is the cross-sectional area of the injection hole. With the given coolant mass ow of each injection hole, the Mach number of the coolant at the injection hole exit can be determined by this equation. Then, the total pressure of the coolant at the injection hole exit can be obtained. The velocity ratio in equation (1) can be expressed as a function of the total temperatures and Mach numbers of both main gas ow and injected coolant as follows: Vc = Vg

+ 1

     n
Vml,ss xml,ss Vg

Vti Vml,ps xti + 1 xml,ps Vg Vg (4)

'

kc Rc Tc Mc kg Rg Tg Mg

 '!

1 +[(kg 1)/2]M 2 g 1 +[(kc 1)/2]M 2 c

"

(3)

The above equation is applied to every mixing position over each blade surface successively up to the trailing edge. The important independent variables in calculating the total pressure loss are the static pressure and the Mach number of the main ow at each injection hole. In this study, an experimentally observed surface static pressure distribution is used, as will be explained later. Once the static pressure is determined, the surface Mach number is easily obtained. Several studies have been made on the effect of trailing edge injection. A study by Prust [7] compared the analytical results of Hartsel [2] and Prust [3] with the experimental results for various trailing edge slot geometries. Hartsels [2] method assumed mixing of coolant ow while Prusts [3] method assumed no mixing loss. Both methods turned out to have advantages and disadvantages, and no generally acceptable simple prediction means were attained. The trailing edge injection model of this study is similar to Hartsels model in that full mixing between the coolant and the main gas ow is assumed. However, this study does not solve the trailing edge momentum and energy balance equations as in Hartsels model, which does not seem to have been fully validated through experiments. The trailing edge mixing model of this study is as follows. At the trailing edge, one-dimensional mixing of the unaffected mainstream and other streams (those from both the
A05498 IMechE 1999

The velocity of the coolant injected from the trailing edge can also be calculated from equations similar to equations (2) and (3) with the aid of the injection hole geometry and the exit static pressure. Once the exit total pressure of each mixing layer is obtained as a result of successive mixing calculations through the surface, the velocity is determined with the aid of the given static pressure of the mainstream. By applying equation (4), the total pressure loss of the mainstream gas through the blade due to coolant injection is found. The parameter | is an important factor in the prediction. In this work, several analyses have been performed before the main analysis and it is concluded that the mixing loss is not very sensitively affected by the | value as stated by Hartsel [2]. However, it is found that care should be taken not to assign too small or too large a value of |. Accordingly, | is given by following the criterion that mixing ratios at the local injection position and the blade row exit are equivalent, as this avoids too large a value of the mixing ratio at any mixing position. For the single-row surface injection case, the following equation is used to determine | based on the criterion: c |+c = | 1 | (5)

When coolants are injected through several rows on both the pressure and the suction sides, two equations similar to the above equation are applied to determine the | values of both surfaces. In the calculation for multi-row injection, for a given total injection owrate an iteration is needed to assign the same total pressures for all injected coolant components. It is assumed that the pressure losses of ows through the injection holes are equivalent. Meanwhile, the coolant ow through each row is determined. However, it should be pointed out that, when the pressure inside the blade (air supply pressure) is lower than the
Proc Instn Mech Engrs Vol 213 Part A

Downloaded from pia.sagepub.com at INDIAN INSTITUTE OF TECH on November 5, 2011

96

D BOHN AND T S KIM

surface static pressure of an injection position, coolant cannot be injected through the hole and suction through the hole may occur. This possibility was briey stated by Koellen [8] and Koellen and Koschel [9] but they did not conrm the existence of suction nor its effect on experiment and prediction. The existence of ow recirculation (injection through some holes and suction through the others) has been reported by Kiock et al. [10]. Even for zero net injection, local injection and suction may occur. In this study, the calculation of the multi-row injection considers a full ow recirculation among injection rows. If the surface static pressure is higher than the supply pressure, the uid inside the mixing layer is assumed to ow into the blade (suction). The air supply pressure is determined such that the summation of the positive (injection) and the negative (suction) owrates equals the net injection owrate (the measured supply air amount). When the air supply pressure is high enough, no suction occurs. For single-row injection and multi-row injection without suction, the coolant injection temperature is equivalent to the coolant source temperature. Determination of the realistic local temperature distribution and the coolant temperature rise before injection is very complicated as has been acknowledged and is beyond the scope of this study. When recirculation occurs for multi-row injection, it is further assumed that every injecting coolant has a constant temperature which is obtained by an energy balance between the coolant source, the suction components from the mixing layer and the injection components. The recirculation of the multi-row case can be summarized by the following relations: m c,net =% mc,i % mc,j
i j

the method cannot be easily utilized in throughow analyses in which a radial loss distribution due to the three-dimensional ow effect should be considered. As mentioned above, the current mixing loss calculation routine is intended to be incorporated into the existing throughow program, which calculates the basic viscous losses with the aid of proven correlations. Accordingly, to maintain consistency, the viscous loss of this study is estimated using the correlations of Kacker and Okapuu [5]. They assumed that the loss consists of the prole, trailing edge, secondary and clearance losses. Since the ow along the blade mid-height passage is examined in this work, only the prole and the trailing edge losses are considered. The viscous loss of this study appears as follows using their correlations: DPv = DPprofile + DPtrailing where DPprofile = f(geometry, flow angle, Mach number) DPtrailing = f(geometry, flow angle, trailing edge thickness) The trailing edge loss is due to the existence of the nite thickness of the trailing edge, which results in complicated phenomena attributable to additional loss including the wake. When the coolant is injected through the trailing edge slot, the trailing edge thickness loss should be corrected. When a portion of the trailing edge area is occupied by the injected coolant, the effective thickness attributable to the pressure loss may be reduced. Therefore, an effective thickness value which is smaller than the real trailing edge thickness value is used in the trailing edge loss calculation using the correlation. It is further assumed that the effective thickness reduction is proportional to the injection intensity. When the injection velocity ratio Vc/Vg is greater than 1.0, the minimum effective trailing edge thickness obtained by subtracting the injection slot area from the real trailing edge area is used to determine the trailing edge thickness loss. In the velocity ratio range between 0 and 1, the effective thickness varies from the real thickness to the minimum effective thickness. After determination of both the viscous and the mixing loss components, the exit total pressure is determined as follows: P3 = P1 DP, DP = DPv + DPm (8) (7)

(6)
j

m c,netHc =% mc,i Hc,inj % mc,j Hc,j


i

where i and j indicate the injection and the suction components respectively, Hc denotes the given coolant supply enthalpy and Hc,inj is the constant enthalpy of injection ow.

2.2

Overall loss

The mixing pressure loss is superimposed on the viscous loss. The viscous loss may be estimated by performing an extra prole loss calculation using boundary layer parameters, as in the work of Goldman and Gaugler [11]. Even though this method seems good enough, it has two main shortcomings. Firstly, introducing the integral method in the throughow analysis programs requires too much extra computing effort. In addition, it is only applicable to two-dimensional ow. In the real ow through turbine blades, a strong three-dimensional effect exists, e.g. secondary and clearance ows. Thus,
Proc Instn Mech Engrs Vol 213 Part A

TEST VANE

The calculation scheme is applied to an annular guide vane, for which experimental results have already been extensively presented [8, 9]. The test data seem very
A05498 IMechE 1999

Downloaded from pia.sagepub.com at INDIAN INSTITUTE OF TECH on November 5, 2011

AERODYNAMIC LOSS PREDICTION OF AXIAL FLOW TURBINE BLADE ROWS

97

Table 1

Geometric data of the annular guide vane


44 mm 33 175 mm 33.3 mm 48.1 mm 8.42 mm 0.818 mm 10.21 mm 2 49.1

Blade height Number of blades Mid-height data Radius Space Chord Maximum prole thickness Trailing edge thickness Opening (throat) Blade inlet angle* Stagger* * Measured from the axial direction.

suitable for the validation of prediction methods because experiments were performed for a multi-row conguration as well as for various single-row congurations. Since the experimental loss data are presented only at the mean diameter and the ow at the blade mid-height is found to be free from the secondary ow effect, a mean line analysis considering only the prole and the trailing edge losses is carried out. The nozzle guide vane has an untwisted spanwise prole. The main geometric data are shown in Table 1. Figure 2 shows an isolated blade prole with injection row congurations such as the diameter, the number of holes and the injection angle. The conguration in the gure is for the fully lm-cooled test. In addition to the multi-row conguration in Fig. 2, every single-row conguration with only one row of cooling holes was separately manufactured and tested. In the case of multi-row injection, all six rows are grouped into two parts: the front part (A, B, E and F) and the rear part (C and D). The coolant for each part is supplied independently. This aims at investigating the effect of the coolant distribution on the pressure loss. Therefore, in the calculation the relations (6) should be applied to both parts. In the experiment, the coolant temperature was equivalent to that of the inlet gas ow and no results for the effect of the difference between

Fig. 3

Example of the variation in the blade surface pressure ratio (M2s = 0.8)

the coolant and the gas temperatures were presented. As acknowledged, well-organized aerodynamic experimental data for real engine conditions (a considerable temperature difference between the gas and the coolant) are difcult to obtain. In the present study, however, calculations are also made to see the effect of the temperature difference. As mentioned in Section 2.1, the static pressure distribution is required to calculate the velocity of the injected coolant. A couple of examples of the distribution of the wall isentropic Mach number are presented in the experimental results. Figure 3 shows an example of the distribution of the pressure ratio (the static pressure divided by the inlet total pressure) as a function of the axial chord. Injection locations are also shown for the nominal condition of an exit isentropic Mach number of 0.8. It is assumed that the qualitative shapes of the variation in the pressure ratio value are similar for every exit Mach number condition in the calculation range.

Fig. 2
A05498 IMechE 1999

Prole of a fully lm-cooled guide vane [8]


Proc Instn Mech Engrs Vol 213 Part A

Downloaded from pia.sagepub.com at INDIAN INSTITUTE OF TECH on November 5, 2011

98

D BOHN AND T S KIM

RESULTS AND DISCUSSION

The following loss coefcient which represents the total pressure loss through the blade is compared for experimental and calculation results: y= P1 P3 P1 p3 (9)

Before the main analysis on the blade with injection is performed, the pressure loss coefcients are calculated for the solid blade in order to conrm the reliability of the prediction of viscous loss. As will be shown briey later in Table 2, the viscous loss prediction is found to be good. 4.1 Single-row injection

First, the calculated pressure loss of trailing edge injection (conguration D) is compared with experiment and the results are shown in Fig. 4. The gure presents the pressure loss coefcient contributed by mixing only since y0 denotes the loss coefcient for a blade without injection. The mixing loss achieves a peak value and then decreases, which is well predicted by the calculation. When enough coolant is injected, the loss coefcient becomes negative, which means that the total pressure of the main uid increases owing to the momentum transfer from the injected coolant whose velocity is higher than that of the main uid. For the temperature ratio Tc/Tg of 1.0, the calculation with the trailing edge thickness correction explained in Section 3.2 shows better agreement with experiment than the calculation without the thickness correction. Another calculation shows that a lower coolant temperature

increases the pressure loss for a given coolant fraction. A lower coolant temperature increases the coolant density and thus reduces the coolant velocity. This leads to a higher momentum loss, which is represented by the second term in equation (4). This aerodynamic penalty outweighs the counter effect (the loss decrease) due to the temperature drop, which is represented by the rst term in equation (4). Figures 5 to 9 show the mixing loss for other singlerow congurations. Except for E and F, the existence of a maximum loss coefcient is clearly shown. It can be seen that the coolant fraction predicted for the maximum loss coefcient coincides well with the experimental value for every conguration. In congurations E and F, the direction of coolant injection is opposite to that of the main ow, which causes a monotonic increase in the loss with increasing coolant fraction. It is also well predicted by the current analysis. The prediction results of Koellen [8] and Koellen and Koschel [9] are also presented for congurations A and B. Their prediction also included the mixing layer concept. They used streamwise momentum balance equations for both the surface mixing and the blade row exit mixing by following the theory of Goldman and Gaugler [11]. Comparison between their and the present calculations reveals that the present one-dimensional mixing model is not less accurate than the other method. For A, B and C, the coolant temperature effect has the same tendency as that of trailing edge injection. The loss increase due to the lower coolant temperature coincides well with the observation of Ito et al. [4], who simulated the temperature difference between gas and coolant by using an injection uid of higher density. However, if the injection angle is opposite to the main ow direction as in the cases of E and F, a lower coolant temperature (and thus lower Vc/Vg) leads to two counter-effects:

Fig. 4

Mixing pressure loss coefcient for trailing edge injection (conguration D, M2s =0.81: cal.1, calculation with trailing edge loss correction; cal.2, calculation without trailing edge loss correction)

Fig. 5

Mixing pressure loss coefcient for conguration A (M2s = 0.81)


A05498 IMechE 1999

Proc Instn Mech Engrs Vol 213 Part A


Downloaded from pia.sagepub.com at INDIAN INSTITUTE OF TECH on November 5, 2011

AERODYNAMIC LOSS PREDICTION OF AXIAL FLOW TURBINE BLADE ROWS

99

Fig. 6

Mixing pressure loss coefcient for conguration B (M2s =0.82)

Fig. 8

Mixing pressure loss coefcient for conguration E (M2s = 0.81)

Fig. 7

Mixing pressure loss coefcient for conguration C (M2s =0.81) Fig. 9 Mixing pressure loss coefcient for conguration F (M2s = 0.81)

(a) a momentum loss decrease at the local mixing position [equation (1) with a negative velocity ratio term] and (b) a momentum loss increase at the row exit plane [equation (4) with a positive velocity ratio term]. These counter-effects lead to either overall loss increase (E) or overall loss decrease (F). 4.2 Multi-row injection

Table 2 shows the loss coefcients predicted for zero net injection. Also presented are the loss coefcients of the solid blade for comparison. Except for the transonic case (M2s = 0.98), the viscous loss prediction for the solid blade is very good. In the experiment, even without
Pressure loss coefcient y for zero net injection (Tc/Tg = 1.0)
y Fully lm-cooled blade M2s 0.6 0.8 0.98 Experiment 0.0450 0.0460 0.0480 Calculation 0.0440 0.0457 0.0463 Solid blade Experiment 0.0310 0.0330 0.0440 Calculation 0.0337 0.0349 0.0363

Table 2

As mentioned before, the injection rows of the multi-row conguration are separated into two parts, for which the coolant mass ows are independently controlled. In the experiment, only the owrates through the two coolant sources were measured and the individual owrate through each row was not reported. In addition, the coolant temperature was equivalent to the main ow temperature for every experiment.
A05498 IMechE 1999

Proc Instn Mech Engrs Vol 213 Part A


Downloaded from pia.sagepub.com at INDIAN INSTITUTE OF TECH on November 5, 2011

100

D BOHN AND T S KIM

Fig. 10

Pressure loss coefcient for multi-row injection with two coolant supplies: the effect of coolant distribution between supplies (M2s =0.81, net coolant fraction of 3 per cent)

net injection, the lm-cooled conguration resulted in a much higher loss coefcient. This seems to be caused by the ow recirculation among the cooling rows. It should be noted that the predicted difference between the loss coefcients of the fully lm-cooled conguration and the solid blade is comparable with the difference observed in the experiment. This conrms that ow recirculation exists in the experiment and its effect on the pressure loss should be included in the prediction. Figure 10 shows the result of the fully lm-cooled conguration with 3 per cent of net coolant fraction and an isentropic exit Mach number of 0.81. Experiments were performed by varying the coolant distribution between the two air supply sources (the front part and the rear part). The abscissa is the fraction of coolant supplied through the front part; for example, the value of 1.0 means that positive net injection (coolant supply) is given only through the front part. Firstly, it should be recognized from the experimental results that the loss coefcient of zero net injection is nearly equivalent to the average loss coefcient of 3 per cent net coolant injection, which clearly shows the effect of recirculated ow on the pressure loss. Even though local discrepancies exist, the average loss coefcient of the prediction is close to that of the experiment. In addition, the general tendency of the experiment, i.e. the loss coefcient is larger when the coolant fraction through the front part is higher than that for the opposite condition, is reproduced by the predictions. From this specic result only, it is difcult to conclude that the existence of ow recirculation is the only reason for loss increase. There could be other reasons that are not considered in this study such as the effect of the presence of injection holes on the boundary layer transition, which may contribute to the loss increase. HowProc Instn Mech Engrs Vol 213 Part A

ever, the good agreement between the experiment and the prediction by introducing the ow recirculation for zero net injection must be encouraging. A calculation with various net injection amounts was carried out and the result is shown in Fig. 11. In this case, the cooling hole congurations are the same as those of the former multi-row injection case, but a single coolant supply is assumed. Furthermore the coolant source temperature is varied. The loss value of zero net injection is comparable with that of the case with two coolant supplies. When the temperature ratio is high enough, the loss achieves a peak value and then decreases with increasing coolant fraction. As the coolant temperature becomes lower, the loss increases and ow recirculation occurs up to a higher net coolant fraction. Figure 12 demonstrates the individual injection fraction relative to the blade inlet gas ow for three cases. In the case of zero net injection, the energy balance assumption explained at the end of Section 2.1 leads to the same injection temperature for all three cases, which equals the coolant supply temperature. Since the static pressure of row F is the highest, positive injection through row F takes place last. Even though the static pressure of the trailing edge (D) is the lowest, its injection rate is not always the highest. At high net injection fractions, row C shows a higher injection rate than row D. This is due to the larger cooling hole area (see Fig. 2). This calculation result demonstrates the importance of the cooling hole geometry design. In practical engine conditions, ow recirculation between cooling holes should be avoided to secure a proper cooling effect all through the blade wall. Unfortunately, however, once the coolant owrate becomes very small for any reason, recirculation may occur and it reduces the stage efciency too. Consequently, to minimize the recirculation effect, care should be taken in designing the cooling holes and an analysis including the ow recirculation should be carried out.

CONCLUSIONS

A calculation routine to estimate the pressure loss of lm-cooled axial turbine blade rows has been constructed by introducing the mixing layer and by using the one-dimensional mixing model. Analyses have been performed on a real annular guide vane. The pressure loss was predicted and compared with experimental data for various single-row congurations and for a fully lm-cooled conguration. The agreement between prediction and experiment was good in the cases of singlerow congurations. It was found that the prediction of current analysis is not worse than the prediction using a momentum balance equation. The effect of trailing edge injection on the trailing edge loss was also examined. In the analysis of a fully lm-cooled conguration,
A05498 IMechE 1999

Downloaded from pia.sagepub.com at INDIAN INSTITUTE OF TECH on November 5, 2011

AERODYNAMIC LOSS PREDICTION OF AXIAL FLOW TURBINE BLADE ROWS

101

It is found that, when recirculation is considered, the pressure loss increase is explained well. The effect of the coolant temperature was also calculated for both the single-row and the multi-row cases and the results seem reasonable. The concept and calculation routine of this study are expected to be applied to the throughow analysis of cooled turbines.

ACKNOWLEDGEMENT Thanks are due to the Turbo and Power Machinery Research Center of Seoul National University for its support of T.S. Kims research abroad.
Fig. 11 Predicted pressure loss coefcient for multi-row injection with a single coolant supply (M2s = 0.81)

REFERENCES
1 Bohn, D. and Kim, T. S. A comparative throughow analysis of axial ow turbines. Proc. Instn Mech. Engrs, Part A, Journal of Power and Energy, 1998, 212, 141145. 2 Hartsel, J. E. Prediction of effects of mass-transfer cooling on the blade-row efciency of turbine airfoils. AIAA paper 72-11, 1972. 3 Prust, H. W. An analytical study of the effect of coolant ow variables on the kinetic energy output of a cooled turbine blade row. AIAA paper 72-12, 1972. 4 Ito, S., Eckert, E. R. G. and Goldstein, R. J. Aerodynamic loss in a gas turbine stage with lm cooling. J. Engng Power, 1980, 102, 964 970. 5 Kacker, S. C. and Okapuu, U. A mean line prediction method for axial ow turbine efciency. J. Engng Power, 1982, 104, 111 119. 6 Shapiro, A. The Dynamics and Thermodynamics of Compressible Fluid Flow, Vol. I, 1953 (Ronald Press, New York). 7 Prust, H. W. Cold-air study of the effect on turbine stator blade aerodynamic performance of coolant ejection from various trailing edge slot geometries. II: comparison of experimental and analytical results. NASA TM X-3190, 1975. 8 Koellen, O. Experimentelle und theoretische Analyse der aerodynamischen Verluste in lmgekuehlten Turbinenleitschaufeln. Dissertation, Rheinisch Westfalische Technische Hochschule, Aachen, 1986. 9 Koellen, O. and Koschel, W. Effect of lm cooling on the aerodynamic performance of a turbine cascade. AGARDCP-390, paper 39, 1985. 10 Kiock, R., Hoheisel, H., Dietrichs, H. J. and Holmes, A. T. The boundary layer behavior of an advanced gas turbine rotor blade under the inuence of simulated lm cooling. AGARD-CP-390, paper 42, 1985. 11 Goldman, L. J. and Gaugler, R. E. Prediction method for two-dimensional aerodynamic losses of cooled vanes using integral boundary-layer parameters. NASA technical paper 1623, 1980.

Fig. 12

Individual coolant fraction for multi-row injection with a single coolant supply (M2s =0.81)

the effect of coolant recirculation between cooling rows due to the differences in cooling hole surface static pressures is included. Even without a net coolant supply, the fully lm-cooled conguration exhibited a higher pressure loss than the uncooled solid blade did.

A05498

IMechE 1999
Downloaded from pia.sagepub.com at INDIAN INSTITUTE OF TECH on November 5, 2011

Proc Instn Mech Engrs Vol 213 Part A

You might also like