PREDICTION OF SCALABLE LOSS IN FRANCIS RUNNERS OF

DIFFERENT SPECIFIC SPEED

T.KITAHORA, J.KUROKAWA, M.MATUMOTO

Yok-ohama NatiOnal UniVersiOl
1 56 Tok-iwadai, Hodogapa-k-u, Yok-oharna 240, Japan
R. SUZUKI
Fuji Ele ctric Co. ,Ltd.
l-1 Tanabeshinden, Kawasak-i-ku, awasak'i 210, Japan

1. Introduction

Hydraulic losses are composed of scalable loss and non-scalable loss. For the
determination of scale-up formulae in hydraulic turbines, it is important to predict the
scalable loss, that is the friction losses in a runner and t"Ixed chamels, as precisely as
possible. However, it is still diffioult to determine them both theoretioally and
experimentally, partly because the bbundary layer is too thin to prepare sufficient
mesh size for CFD, and partlv. because the scalable loss oannot be separated
experimentally i'rom the measured total hydraulic loss.
Scale-up f･ormulae have long been discussed in IEC-WG18[2] and IAHR-WG3,
and also in JSME[3]･ In the JSME Standard S-008 "Perf ormance Conversion Method
for Hydraulic Turbines and Pumps" , the new method of theoretically derived scale-up
formulae are presented[5]. In order to further improve the acouraoy of scale-up, it is
necessary to develop a method of more precisely predicting soalable loss.
The present study presents a predietion of scalable loss in a runner chaunel of
arbitrary three-dimensional shape. The prediction is composed of a 3 -D Euler
calculation for the main flow in a runner chamel and the 3 -D boundary la_Ver
oalculation, The latter is based on the momentum integral equati･on by use of an
orthogonal curvilinear coordinates.
The present method is applied for five Francis runners of different specific
speed, and the calculated results of scalable losses are compared with the results by
conventional two-dimensional loss analysis.

2. Boundary layer equation

A three-dimensional boundary layer along a Francis turbine runner is analyzed taking

323
E. C brera et al. (edsJ. Hydralelic Machinery and Cavitation, 32 332.
C 1 996 Kluwer Academic Publishers. Printed in the Netherlands.
the surface curvature into account. Z-axis of the
ZA . .･-1¥

y
absolute Cartesian coordinate ( , Y,Z) is set on the . ' _ , ,.
.･

r
rotating shaft and the rotating local relative .... ... .

curvilinear orthogonal coordinates (x,y,z) is set on .

z ' , .-
x.. Soli¥ d wall
the runner wall as shown in Fig.1. Each . '.=-. ' ' ' '
'.

intersection of calculated grid lines on the walls is ' :':.

selected as the origin of this looal coordinates. For
simplicity_ of theory, the direotions of the grid x / Y
lines on the wall are neglected, and the main tlow J igure I. CoQrdinates system
direction at the outside of boundary layer is taken
as the x direction in each origin. Moreover, the axis perpendicular to this dir ction on
the w all is defined as z-axis and the nonnal direction to the wall is taken as y- axis.
The Navier-Stokes equations and the continuity equation applied to this
coordinate system are given as follows.

6u+uau+v6u+w6u - c ,uw+ K ,w2 =- 1 ap+v 6a;u +f...+f,.. (1)

p ax

a +uw +va +w
w aza /cw uw+/ca 2 w1 up+v
a+f,.,+f...
gw _
az
2)

au
ax + av 6w /c u 'c*w=0

To stabilize the calculation, the method of calculating unsteady flow is adopted. In

this method, the real length is used for the length of each axis, theret ore the scale
t'actors are 'fixed to unity, because the grid interval
distance and the boundary layer thickness are much
y

s'maller than the radius of ourvature of the axis. The ' U

velocity v in y-direction is determined by the t ollowing
equation derived from Eq.(3) neglecting the last two

v=-
As shown in Fig.2, I /n th-power law model is adopted z,
for the velocity profile u in the x-direction. The Figure 2. Velooity distribution
Marger's, shown as the followings, model is adopted in boundary layer
[

for secondary flow in z-direction.

u[l

= - tart
pREDlcnoN OF SCALABLE LOss DT FRANCIS RUNNERS OF DF ;RENT SpECIFIC SPEED 325

Her , 6 is the boundary. IaV. er thickness, and is the angle between the direction
of the main flow U at the outside of boundary layer and the direction of the wall
stream line.
Boundary. Iayer integral equations in x and z directions are obtained by
substituting the above velocity. ' distributions into Eqs. (1) and (2), and by integrating
them from the wall to the edge of boundary layer under the assumption that the main
flow pressure prevails into the wall surface. The pressure terms can be deleted by
oombining the equation of the main flow with that of boundary layer. The following
equations are finally obtained for the boundary layer.

Un+1 at + axCfU C_.+U Ce 6zC4+1cJC

ax az
-lc.C5 2Q ax aY ax aX (6)

C
= p
*- aX az a az 3
ac a a
t 3 axCr azC5+1c.C4+,c.Cl

2Q az aYaY6xaz
2 aX (7)
C
1
= p
' 6X 6x
Here, the coefficients C l C 5 in the above equations are expressed as follows.

l fs(U2-u2)dy = 2U2
C =
n +2
2f
f f

C = (U-u) dy =
8 2U n3tanP
C3= ow dy = 6n3+1lv2+6n+1
s U2 n 8tan
C4= ou , dy_= 3n3+1ln2+12n+4
f:
-

5= w2 dy
0n5+137n4+225n3+170n2+60n+8
and /c * and /c , in Eqs. (6) and (7) are the ourvature of the z-axis and x-axis
respectively. is rotating angular velocity of runner, and th last tenns in Eqs. (6)
and (7) show the decrease of the Corioli's force in the boundary layer.
The following experimental formula by Ludwieg-Tillman is adopted for wall
shear stress T * in the main flow direction.

) Ua ).678( "2 .2e8

l ,= p U2 XO. 123 X 10 ' (9)
ll

'n
e =
n+1 )(n+2)

Equation (3) of continuity is also integrated, yielding the following equation.

 ax C2- U - C3- Ic.C2 +/c.C8+0.0306U(He- 3) o. 658 = O (1 l)
6x 6z
Here, He is the entrainment of the main flow and the t'ollowing equation by Standen
is assumed.

H,=
n

At'ter all, the variables to bc determined b _! use of the numerical method are 6 , n
and tan in Eqs. (6), (7) and (1 1 ).

3. Finite difference method

The mam flov i velocrty U at each @ @

intersection of curvilinear grid lines, shown d z
in Fig.3, on the wall is given by the main
flow calculation. In the present study, the d
main tlo i' velocity. was determined bV. R
solving the Euler's equation by use of a
finite difference method.
As t or the boundary. Iayer calculation, Fig:3 Local upstream FDM
Eqs.(6),(7). and (ll) are converted into the
t inite dit t erence equations under the local coordinates sV_ stem. The upstream
difference scheme is used to stabilize the calculation. The upstream reference point
used i or FDM is set on the intersection between the lines connecting t our grid points
@ and the main flow line passing the origin. Each variable at the upstream
reference point is determined b_y interpolation from the values of two grid points )
and @ . And also, the ret'erence point in the direction perpendicular to the main flow
is similariy detennined for the secondary_ flow. The implicit method of Crank-
Nicolson is used for unsteadV flow calculation.
For the inlet boundary. condition, very small values of 6 ,tl=5 and tan =0 are
given. The same values are also given to the whole area as the initial condition of the
unsteady tlow calculation. It is generally dit ficult to determine the flow characteristics
near the comer. In the present study the equations on each wall surface is solved
separately under the assumption of tan =0 at the comer boundary_ .
The solutions at each time step are detennined by applV_ ing an iterative method
to these equations, and the final steady solutions can be detennined when the
unsteady solution does not change any more.

4. Application to Francis Turbine Runner and Mam Flow Charactenstrcs

The preceding c'alculation is applied to five runners of model Francis tul aines. The
design parameters of these models are shown in Table I . The turbine E is well known
PREDlcnoN OF SCAu BLE LOSS IN FRANCIS RUNNERS OF DIFFERENT SPECFIC S PEED 327

pressure suri ace crown surf ace Band surf'ace

Suction suri'ace

r-z plane r- n. plane e -z plane t (' : T ¥ 1,-'-[ !

r-z plane

r.
:¥. ¥. '¥:
./
" I'/'t
¥ " " tttil

tli
.¥.. ." ' ' ""'¥, 2
l.
l'
.
, . .¥
' l":¥/ ¥
.,ll , , t )

¥'
tl t'.
'. *

:,t
.,,
･ "

¥tl ¥.,' , I '

tt.
,tt
l(
¥l
.

'l4;i L//' 1,_

I ' :
l
! ll l
,

l
¥ { l'
;

¥¥¥ ¥ . ¥ l)f.j/:i'i/'
:: E! :r

"'¥ ¥ L ""*/ / '1. . '

''/' ._('
. .//i.'///
/ " t
/

' t/i/(a) Velocity

. ¥ .vector¥.¥ ¥flow ' r
of main ):) "("'
._'
/..i_-/"
- *'
' .l
"- ._'
:- ;-
/-'/_);- '/r' ' _ "'/'
/: (r/';-1(1 ..t ;-'_'('_-'/ ;:_'

'"-'-''- '' -'*"_l

-0.6 _o.6
'l
/

:0'8 -'O'8
/]

¥
/

-0.6
-1 oJ /1 - I 'O
'
-0.6 -0.8

- .O
_ ¥: - I 'e
/
-1'8

¥ , _
' '-1'4 , : 0'6
ll

-1'8- ; : ' " r

¥¥' . ' 16
l
¥¥'
-
_ _. 1
7"
¥¥¥___' '_/'
//1/'// _O 8
_1'O
_ 12 l r'/ /
'// _1'2
r' - -¥'_
¥.
¥"¥¥ .f. '
1'2t
- I .2
/

¥> ¥ ¥
' ¥ l'
¥ ¥ ¥:¥L
-1'1 ¥¥ i
¥ti
'
¥1

l }'

it i!
j

. t'
i !'
-A.4 -2.8
¥:

-1'2 l ¥T
"'
¥t"_ / l -a.6
-2.4
- I '6
-1'4 .-2.8
_2i4):Jj/
o s_d; 11 : 5 / rr-1'6
¥¥ J
/

-1'8

-t6 ¥a.2/4///4 ¥/
-2.2 -2.0 1'8 -a'O -
///: _ 2.0
f__/ __l
- 2.2
l-'
(c) Equi pressure line when pressure rize due to ceirit ugal force is neglected
Figure 4. Main flow characteristics (Runner E)
328 T. KrrAHORA ET AL.

Table l. Parameters of model Francis turbines

P : Potentral Flow E Euler

as CTAMM Runner.
The boundary. Iayer characteristics are analyzed only at the maximum et'i iciencv_' point.
In order to show the main fiow characteristics, the results of the nmner E, as an
example, are illustrated in Figa.4. These results are obtained by numerical calculation
of a three-dimensional Euler's equation under the assumption of wall slip condition.
The calculated velocity vector of the main flow is shown in Fig.4(a) and the contour
of equi-pressure lines are shown in Fig.4(b). Here, U 2 is the peripheral velocitv_ of
the runner at reference diameter. Four surfaces of a runner channel are developed and
are pro.i,ected from the inside of the channel. The main flow velocity U at the edge of
boundary layer is used as the input data of the following boundary calculation.
The calculated L esults of the main flovv in the runner ex'it section is compared
with the measured data in Fig.5 t or the case of runner E, The horizontal axis is the
distance from the outer wall of draft tube. The numerical results are averaged over the
vd' ne-to-vane channel. The calculated results agree well with the experimental results.

o.s

'o
s
o
c: 0_4
oc
S

o
o
h
.t o.2
oo
o
>
¥
c:t) o.o
.5
tc:

o
oo
o '0.2
co
co
oh
,:.

'0.4

o.o 0.6 0.8

0.2 t .o

Arc lengLh

(C p :Pressure>C 3 :axiai velocity.C s :Tangential velocity.C ' meridlonal velocrty)

_ rn'
Figvre 5. Outlet Flow characteristice in runner E and comparison with measurments
PREDICT1ON OF SCALABLE LOSS IN FRANCIS RUNNERS OF DIFFERENT SPEC田C SPEED 329

Pressuresur血CC Crown sur鉛ce Suction Surぬce Band surfacc

猷α2
 アー2Plane  アーθPlane   γ一zplalle θ・zplalle

、＼

 ／

     、
   α
  2！
4硬   

0．2
＼
 
  

し︾’
 ’

．
＼
ぐ

︾α−
o．4
纏

 
      0，49

 ㍉
  、￥
 X、
  

＼
0．6

一 δの2×10。2
（a）Boundary layer thicknes，

多
         号・で1．7
    ∼

2。1
 肖

 1・31
3 イ
f．7

  1．7
獣、
2．5

，1。5    ！
  ノ！1 筆．71
 、竃
1 ／
1．3

1．3

      ＼   z
ノノヘー＾．
       ＼

、！ヲ“
1．5

電．3

           1．7
 ／＼1・3 ＼ 。、

＼
   13一、、一
、／

！．一ノ／ ￥
（b）Shape fゑctor， 丑   1。3
ぜ￥−

〉
。り﹃

盟、

乃g麗κ6：． Results食om3．D boundary layer calculation （Rumer E）
330 T. KrrAHORA ET AL.

5. Results' and Discussions of Boundary Layer Calculation

To discuss the boundary_ Iayer characteristics, the contours of the equal boundary layer
thickness 6 , of the equal shape factors H, and the wall stream lines are illustrated in
Fig.6 for the case of ruuner E. The boundary layer calculation were performed
without an_v trouble, although reverse flow occurs near the crown wall of the pressure
side in the main flow.
On the vane suction surface and on the band surface of the boundary layer grows
from the leading edge, but on the vane pressure surface it grows t rom the entrance of
'the crown surface, since the main flow enters into the pressure surface through the
crown surface as shown in Fig.6(a).
The Corioli's foroe works strongly in the main flow where the velocity is large,
and balanoes with both the pressure gradient and the inertia force. Because the
Corioli's force works weakly near the wall where flow is slow, the wall stream line
tums to the opposite direction of Corioli's force, or tums to the direction in whioh the
real pressure decreases if the pressure rise due to centnfugal force is removed as
shown in Fig.4(c).
On the crown surface shown in Fig.6(b), the shape factor becomes very lar ge
near the comer between the inlet and the pressure surface where flow velocity is
slow, and ' the boundary layer fluid is transferred along the wall stream line whioh
deviated from the pressure side to the suction side between vanes. This flow behavior
is mainly caused by the pressure gradient from the pressure side to the suotion side
rather than the Corioli ' s foroe.
In the suction surface of the vane, the boundary layer gradually grows toward the
downstream near the leading edge, but the stream lines tums to the band side with an
increase in the boundary layer thickness. This is because the pressure decreases from
the crown side to the band side due to the meridian ourvature of chaunel.
Using the present method, the hydraulio energ_y loss due to friction can be
predicted by use of energy thickness of the boundary layer. The t riction loss on the
wall can be estimated as a energy defect at the channel exit. The energy det ect is
calculated from the energy thickness e * at the exit surface for the case of three
dimensional boundary layer.

e･=I8 ･f o V' (U2- V2 ) dy 13)

U

Here, V' is the velocity component perpendicular to V, which is velocity in the

boundary layer(Fig.2), on the exit section.
On the other hand, in the performance conversion, the ratio 6 B , of friction
head to effective head is very convenient for standardization, as adopted in IEC Code
f

or JSME Code. The ratio 6 E , is obtained by the following equation.

E,= (Ue6') dl lgQH (14)

PREDICTION OF SCALABLE LOS S D( FRANCIS RUNNERS OF DF RENT SPECIFIC SPEED 331

n 2
,:

,-,
'

w 1.5
co 3-D 1'-D
( Band Wal l
1

L ,crown wal l
O SUot i on
o.
5

surface
_ OJ I-
:T D:
Ps:essure
surfaoe
o

20 30 40 50 60 70 80n *Q90
Figure 7. Friction loss in runner
6

'-1
oe

LJ
5

UJ
CO Draft tube
4

Oas i ng
3

Stay vanes
2

Guide vanes
1

Runner
o

20 30 40 50 60 70 80 90
n SQ
Figure 8. Friction loss of turbine

Here, I is the distance along the contour of the exit wall. The eurvilinear integration
conceming I is peri onned along the exit wall edge of all channels of the runner.
The present calculation was applied to all turbine rumers A-E and the calculated
friotion loss ratio 6 E , are plotted against specific speed n , Q Fig. 7 . The notations
O [] are the results from the present calculation, and the notation e (1 1/L are
those of the conventional two dimensional analysis, in which the flow is assumed to
be along the grid line, and the energy equation and Rotta's equation are used.
The total friction loss in a ruuner becomes larger with an increase in the specific
speed , but the scatter of the data is large because of the difference in the design
philosophy and manufacturing tolerance of each turbine. The solid line indicates total
friction loss obtained by two dimensional analysis.
Figur e.7 shows that the total friction loss from two dimensional analV_ sis becomes
larger than that t'rom the present analysis. Due to the secondary flow, the boundarv.
layer thickness must be thinner in the thee-dimensional analysis than in the tw o
dimensional analysis, because the fluid in the boundary layer is removed bV_ the
secondary. fiow in the 3Danal sis.
 Figure.7 also reveals that the t riction loss on the vane occupies large part of the
total t riction loss of a runner. On the other hand, on the crown surface, the main flow
velocity is slow and the wall stream line leans. Theret'ore the boundary_ Iayer thickness
in the exit section is thin and the i riction loss becomes thus very small.
The i riction loss ratio 6 E of the casing, stay vanes, guide vanes, and drat t tube
of turbines t rom A to D ai'e shown in Fig.8. All the results except t or loss of the
runner are obtained by_' two dimensional calculation. Figure.8 reveals that the friction
loss on guide vanes becomes smaller with an increase in specific speed. The notation
O shows the sum of fiiction losses on all sulfaces obtained by_ using two
dimensional analV.'sis. The dotted line is the averaged total scalable loss and is seen to
be almost independent of the specit '1c speed, and 6 E takes the value around 0.037.

6. Conclusions

The calcul.･rtion method of thiee dimensional bOundany layer developed b),' use of
upstream FDM with local curvilinear coordinate system, is applied to t 'rve types of
Francis turt,ines. The i ollowing conclusions are obtained.
( l) The present method can be solved without trouble, even if complicated tTow
including reverse flo¥v is induced such as the flow in the pressul e suri ace.
(2) The pr sent method is able to estimate the intluence of the secondal .f tlo v, which
is pell)endicular to the main flow. The t'riction loss obtained by this method becomes
smaller than that obtained by t¥vo dimensional method.
(3) The avel aged total tnction lot;ts I atio on all t;urtaces obtamed by using t¥vo
dimensional analysis is about 0.037 in Francis turbine, and it is almost independent of
the specit lc speed.

Acknolvledgements

Calculation was perfonned as a part of the (activity of JSME WCT "Peltolll:1ance

Conve'rsioll Method fof Hydratilic Turbines and Pumps" . The main flow data for 3D
Euler, 3D potential ilow and 2D boundary. IaY. er data vere of t ered bV the committee
memberS. Df. Nagat'il_ji and Mt'. Suzuki in Toshiba Co.Ltd.. Mr. Miur('i ill Hitachi
Co.Ltd. and Mr. MiV. aga¥va in MitSLtbishi Heavl,y InduStiy Co.Ltd. . The atrthorS ¥vould
like to express sincere gTatitude to the WCT members.

Ref erences
l Alakawa C.. Same.jim'a, lvl.. et al.. A 3D Ettl*1 soluuon of Fl mcls !unnel usm5' ps*udo
compressibility, 3D-Comput'ation of Incompressib]e Intemal Flows' (NirNTFl¥･i39), (1989:),65-69
2. IEC drafi code 4(CO)47-1987.Detemlination of the prototype peribml'ance from modei tests of
hvdraulic machines with consideration of scale efi cts
3. JSlvlE S008(1989), Perfbmlance conversion method for hv. draulic turbines and pttmps
4. Suzuki, R_> An 'assessment of the loss distribution in Fr',ulcis Tatrbines, Proc. of 18th IAHR S¥.'mpo.,
(1996)
_ . Ida T., et al Recent development of studles on scals eflbct in Jap'an, Proc. of 18th IAHR Sympo__