212

Numerical Study of Partial Admission Stages in Steam Turbine∗

(Efficiency Improvement by Optimizing Admission Arc Position)

Naoto SAKAI∗∗ , Tetsuya HARADA∗∗∗ and Yoshinobu IMAI∗∗∗

This paper deals with an application of computational fluid dynamics (CFD) to partial
admission stages in a steam turbine. The calculation of partial admission stages requires
unsteady analysis and full circle modeling. Therefore, quasi-3-dimensional (Q-3D) analysis
of the mean radius is conducted to reduce computational load. First, an experiment using the
air turbine is carried out. The result is in good agreement with the result of CFD analysis
under the same conditions as the experiment, and the application of the Q-3D method to
partial admission stage analysis is validated. Using this method, 2-stage analysis of partial
admission is conducted. The influence of the circumferential position of the admitted arc on
stage efficiency is discussed. The efficiency difference is related to the windage loss caused
by pressure distribution in the circumferential direction. It is found that there is an optimum
circumferential position of the admitted arc from the point of view of turbine efficiency.

Key Words: Steam Turbine, Computational Fluid Dynamics, Efficiency Improvement, Par-
tial Admission, Quasi-3-Dimensional Analysis

unadmitted ones. Therefore it is generally known that the

1. Introduction efficiency of a partially admitted stage is lower than that
Effective utilization of fossil fuel is becoming impor- of a fully admitted one. To improve the efficiency of a
tant from the environmental point of view as the global partial admission stage, experimental studies of partial ad-
warming due to CO2 emission becomes a serious problem. mission turbines have been made(1) . Although studies us-
Following this trend, the steam turbine, which is one of ing CFD analysis, which is often applied to other turbo-
the key hardware components of power generation plants, machinery(2) , are rare because it requires heavy computa-
is required to be more efficient. As one of the aids to effi- tional load due to unsteadiness and full circle modeling(3) .
ciency improvement of a steam turbine, the result of CFD It is difficult to carry out unsteady analysis of fully
(computational fluid dynamycs) analysis in a partial ad- 3-dimensional partial admission stages in a short period
mission row in a steam turbine is reported in the present of time at practical costs, although computers have been
paper. making great progress in this decade. Therefore, we use
The blade height of high-pressure stages of a small quasi-3-dimensional CFD analysis of the mean radius.
steam turbine is extremely low when steam is admitted in First, we conduct an air turbine experiment of one stage
a full arc. To avoid losses due to low blade height, partial in order to validate the quasi-3-dimensional analysis of
admission, in which steam is admitted only in one arc of a partial admission turbine stage. The result is in good
the turbine annulus, is commonly adopted. Rotor blades agreement with the result of CFD analysis under the same
in a partial admission row cause additional losses because conditions as the experiment. We are confident that this
they periodically pass through admitted arcs as well as the method can be applied to a partial admission stage.
By the above-mentioned method, we conduct CFD
∗
Received 20th September, 2005 (No. 05-4163) analysis of 2-stage partial admission turbine to elucidate
∗∗
KAWASAKI HEAVY INDUSTRIES, LTD., Technical its flow field. We find that the relative circumferential
Institute, 1–1 Kawasaki-cho, Akashi, Hyogo 673–8666, position of admitted arcs influences the turbine stage ef-
Japan. E-mail: sakai [email protected] ficiency.
∗∗∗
KAWASAKI HEAVY INDUSTRIES, LTD., Gas Turbine
and Machinery Company, 3 Higashikawasaki-cho, Chuo- Nomenclature
ku, Kobe, Hyogo 650–8670, Japan. √
E-mail: harada [email protected]; imai [email protected] Ca : isentropical velocity Ca = 2h

Series B, Vol. 49, No. 2, 2006 JSME International Journal

 213

G : normalized mass flow rate

H : static enthalpy
Ht : total enthalpy
∗
h : heat drop h = Ht0 − H2or4
l : axial chord length
o : throat width
p : pressure
t : pitch
u : blade speed
Vθ : tangential velocity
z : axial coordinate
ε : admission ratio
Ht0 −Ht2or4
ηt−t : normalized adiabatic efficiency ηt−t =
Ht0 −Ht∗2or4
Ht0 −Ht2or4
ηt−s : normalized adiabatic efficiency ηt−s = ∗
Ht0 −H2or4 Fig. 1 Test apparatus of air turbine
θ : tangential coordinate
∗ indicates isentropic change
Suffix
0 : inlet of the 1st stage
1 : outlet of the 1st nozzle
2 : outlet of the 1st blade
3 : outlet of the 2nd nozzle
4 : outlet of the 2nd blade
2. Computational Methods (a) Meridional view
In the numerical investigation, we use the compress-
ible Navier-Stokes code, CFX-5 ver. 5.7, which is com-
mercially available. CFX-5 solves the Reynolds Averaged
Navier-Stokes equations by a time-marching control vol-
ume method. The k – ω SST model is used, and all walls
are assumed to be smooth. In this study, turbine cascades
are modeled quasi-3-dimensionally where the models do
not have distributions of values such as pressure in the ra-
dial direction. The hub walls and shroud walls are treated
as free slip walls. Meridional contours in turbine cascades
are considered by changing the height of cells in the nu-
merical analysis. To grasp the flow field of a partial admis-
sion turbine cascade, unsteady simulations are conducted. (b) Blade-to-blade view

3. Air Turbine Experiment Fig. 2 Schematic of partial admission stage

An experiment using an air turbine which has a par-

tial admission stage was carried out in order to confirm
whether the numerical simulation method described in the
previous section yields an appropriate solution of partial
admission stages.
3. 1 Experimental apparatus
The cross section of a test rig is shown in Fig. 1,
and the details of the partial admission stage are shown
in Fig. 2. Although this test rig consists of a single stage
(one nozzle row and one blade row), the subsequent stages
of the control stage in a small steam turbine are usually
partial admission stages. Therefore, a dummy nozzle that
has only one passage with an admission ratio the same
as that of the first stage is equipped with the rig. Figure 3 Fig. 3 Geometry of blades

JSME International Journal Series B, Vol. 49, No. 2, 2006

214

Table 1 Specifications of cascades of air turbine experiment

and Table 1 indicate, respectively, the geometry of the cas- (a) Whole grids
cades and their specifications.
3. 2 Experimental method
Compressed air from a blower was introduced into
the inlet of the test rig, and its outlet was open to the
atmosphere. The experiment was conducted keeping the
pressure ratio constant. By changing the speed of the tur-
bine, the velocity ratio of u/Ca (Ca is the isentropic ve-
locity calculated from the head drop across a stage) could
be controlled.
3. 3 Numerical analysis
Numerical analysis was conducted by the method de-
scribed in section 2, using the quasi-3-dimensional model
of an air turbine with air as the working fluid. Figure 4
shows the numerical grids used in this analysis. The grid
consists of three regions: the nozzle region, the blade re-
gion and the dummy nozzle region. The nozzle region and (b) Around cascades
the dummy nozzle region are on a stationary frame and the Fig. 4 Numerical grids for cascades
blade region is on a rotating frame. Every grid consists
of an O-type mesh and an H-type mesh. The total ele- Ht0 − Ht4
ηt−s = .
ment counts in the nozzle region, the blade region and the Ht0 − H4∗
dummy nozzle region are, respectively, 84 498 (approxi- There is good agreement between the analytical results
mately 60 × 30 × 2), 677 280 (approximately 60 × 22 × 2), and the experimental ones.
and 79 086. The sum of the elements in the three regions Through comparison with results of the air turbine
is 840 864. experiment, it is confirmed that the quasi-3-dimensional
3. 4 Results method can simulate the flow pattern of partial admission
Figure 5 (a) shows velocity distribution at one mo- stages in a steam turbine.
ment of CFD analysis. In this figure, we can see the
4. Analysis of 2-Stage, 4-Blade Rows
high-velocity region caused by the sucking effect of par-
tial admission at the end of the arc of admission and the A quasi-3-dimensional analysis of an actual steam
low-velocity region at its beginning(4) . Figure 5 (b) shows turbine is carried out by the above-mentioned method. In
the pressure distribution at the inlet of the dummy nozzle. this study, the influence of the relative circumferential po-
The same tendency of pressure distribution can be seen sition of an admission arc is discussed. Therefore, 2 stages
between the analytical results and the experimental ones (two nozzle rows and two blade rows) are modeled.
at the outlet of the dummy nozzle. Next, analyses with 4. 1 Computational methods
velocity ratios, u/Ca, of 0.25 and 0.45 are conducted by Computational methods are the same as those in the
keeping the pressure ratio of the turbine inlet and outlet air turbine experiment. At the inlet, the steam pressure is
constant and by changing the speed of the turbine. Fig- 5.3 MPa and the steam temperature is 755 K. Static pres-
ure 5 (c) and (d) respectively indicate the efficiencies and sure at the outlet is approximately 2.0 MPa (the outlet con-
the mass flow rate. The charts are plot of the results of both dition is specified by the mass flow rate). Since the work-
CFD analysis and experiment. Values are normalized by ing fluid is superheated steam, the steam properties are
those of experiment with u/Ca = 0.45. Turbine efficiency taken into consideration. Gas properties such, as specific
is defined as heat, are variables and are functions of local pressures and

Series B, Vol. 49, No. 2, 2006 JSME International Journal

 215

temperatures. These variables are calculated by referring

to an internal steam table of the code(5) . Although more
memory is required, the computational load of analysis
considering steam properties is as heavy as that of ideal
gas analysis.
4. 2 Cascades
The overview of the model is shown in Fig. 6. The
model is composed of nozzle rows in which passages ex-
ist in one part of the turbine annulus and blade rows in
which passage exist in full circles. The outlet boundary of
the grid is set far from the cascades in order to reduce the
(a) Velocity distribution by CFD analysis; u/Ca = 0.45. influence of the boundary. Figure 7 shows numerical grids
Velocities are normalized by Ca of the first stage and Table 2 indicates the geometric pa-
rameter of the cascades. Every grid consists of an O-type
mesh and an H-type mesh. The total element counts of the
first nozzle region, the first blade region, the second nozzle
region, the second blade region and the dummy nozzle re-
gion are, respectively, 123 426 (approximately 70×20×2),
408 520 (approximately 40 × 16 × 2), 186 786 (approxi-
mately 75 × 25 × 2), 626 566 (approximately 40 × 20 × 2),
and 147 387. The sum of the elements in the five regions
is 1 492 685.

(b) Pressure distribution at dummy nozzle outlet; u/Ca = 0.45

(c) Mass flow rate and u/Ca Fig. 6 Analytical region

(d) Efficiency and u/Ca

Fig. 5 Results of experiment and CFD calculation
Fig. 7 Numerical grids

JSME International Journal Series B, Vol. 49, No. 2, 2006

216

Table 2 Data of cascades

Fig. 9 Total to total efficiency of each model

Fig. 10 Tangential velocity distribution in the axial direction

around the first blade

around the first stage blade row at the position opposite

that of the arc of admission. The direction of steam near
the blade is the same as that of blade rotation, but the flow
directions away from blades differ from each other. In par-
Fig. 8 Partial admission stage model
ticular, the flow direction of model B is the opposite to
that of blade rotation. Figure 11 indicates instantaneous
4. 3 Position of admission arcs streamlines of the three models. In model A, streamlines
Three models with different circumferential locations between the first blade row and the second nozzle row in
of admission arcs are prepared in order to study the ef- an unadmitted arc form several eddies, and steam does not
fect of the circumferential location on turbine stage per- flow in any particular direction. On the other hand, the
formance. These three models are shown in Fig. 8. The streamlines of model B and model C show that there exist
center of the second stage admission arc of model A is circumferential flows there, which are indicated by arrows
located at the same circumferential position of that of the in the figures. These flows result in the differences in the
first stage. The admission arc of the second stage is shifted stage efficiencies.
in the direction of rotation in model B, and in model C, the These windage flows are caused by pressure distri-
admission arc of the second stage is shifted in the direction bution in circumferential directions. Figure 12 shows the
opposite to that of rotation. pressure distribution of model B. In this figure, pressure is
4. 4 Results normalized so that the inlet pressure of the first stage and
Figure 9 shows the turbine efficiencies of the three the outlet pressure of the second stage are, respectively,
models. Turbine efficiency is defined as 1 and 0. In Fig. 12, a low-pressure region, which causes
Ht0 − Ht4 circumferential flow, is seen. The reason behind the for-
ηt−t = .
Ht0 − Ht∗4 mation of the low-pressure region is that the circumfer-
The efficiencies are normalized by that of model A. The ential position difference of the admission arcs does not
efficiency of model A is the highest among the three. The supply enough steam at the end of the admission arch of
reason for this can be explained in terms of the windage the second nozzle. When flow is insufficient in a nozzle
flow between nozzles and blades. Figure 10 indicates row, there is no pressure difference between the nozzle in-
a tangential velocity distribution in the axial direction let and the nozzle outlet. Under this condition, pressure at

Series B, Vol. 49, No. 2, 2006 JSME International Journal

 217

Fig. 13 Pressure distribution at the outlet of the first blade row.

The pressure difference in model B is largest

In summary, differences in the circumferential posi-

tion of the admission arc cause pressure variations at the
blade row exit, which leads to windage flows. Therefore
the efficiency of model A, whose pressure variation is the
smallest, is the best.
5. Conclusions
The quasi-3-dimensional analysis is validated as a
method of simulating the flow of a partial admission stage
of a steam turbine, by comparing the results of an ex-
periment using a single stage air turbine with those of
CFD analysis. Using the method, analyses of two-stage-
Fig. 11 Streamlines of each model. Note that windage flows turbines are conducted, and the circumferential position
of model B and model C are opposite. Streamlines of the admission arc is found to change windage flow and
are disconnected at the boundaries between a stationary turbine efficiency.
frame and a rotating frame
References
(1) Kovats, Z., Interferometrically Measured Partial Ad-
mission Impulse Blade Forces, Aero-Thermodynamics
of Steam Turbines, (1981), pp.43–46.
(2) Dawes, W.N., Development of a 3D Navier Stokes
Solver for Application to All Types of Turboma-
chinery, ASME Paper No.88-GT-70, Gas Turbine and
Aeroengine Congress, Amsterdam, The Netherlands,
(1988).
Fig. 12 Pressure distribution of Model B (3) He, L., Computation of Unsteady Flow through Steam
Turbine Blade Rows at Partial Admission, Proc. Insti-
the nozzle inlet is reduced. tution of Mechanical Engineers, Vol.211, Part A, Jour-
Figure 13 indicates the pressure distributions of the nal of Power and Energy, (1997).
(4) Balje, O.E., Turbomachines, (1981), pp.155–157, A
stages. The low-pressure region for model B is also seen
Wiley-Interscience Publication.
in this figure, which causes windage flow. In model C, (5) Sakai, N., Yoshida, H., Harada, T. and Imai, Y., Ap-
the low-pressure region is formed on the other side of the plication of CFD to Performance Evaluation of Steam
admitted arc, which causes windage flow in the opposite Turbine, Proc. International Conference on Power En-
direction. gineering, Kobe, Japan, Vol.2 (2003), pp.211–216.

JSME International Journal Series B, Vol. 49, No. 2, 2006