Chapter 5

Pump losses

5.1 Loss types

5.2 Mechanical losses
5.3 Hydraulic losses
5.4 Loss distribution as function of specific speed
5.5 Summary
 5. Pump losses

5. Pump losses Euler head

Recirculation losses
H Leakage
As described in chapter 4, Euler’s pump equation provides a simple, loss- Flow friction
Incidence
free description of the impeller performance. In reality, because of a number Pump curve

of mechanical and hydraulic losses in impeller and pump casing, the pump H
performance is lower than predicted by the Euler pump equation. The losses
cause smaller head than the theoretical and higher power consumption, see
figures 5.1 and 5.2. The result is a reduction in efficiency. In this chapter we
describe the different types of losses and introduce some simple models for
calculating the magnitude of the losses. The models can also be used for
analysis of the test results, see appendix B.
Q
Figure 5.1: Reduction of theoretical Euler
head due to losses.
5.1 Loss types Q
Distinction is made between two primary types of losses: mechanical losses
and hydraulic losses which can be divided into a number of subgroups. Table Shaft power P2
P Mechanical losses
5.1 shows how the different types of loss affect flow (Q), head (H) and power Disk friction
consumption (P2). Hydraulic losses
Hydraulic power Phyd
P
Smaller Higher power
Loss Lower head (H)
flow (Q) consumption (P2)
Mechanical Bearing X
losses Shaft seal X
Hydraulic Flow friction X
losses Mixing X
Recirculation X Q
Incidence X
Disk friction X
Leakage X
Q
Chart 5. 1: Losses in pumps and their influence on the pump curves. Figure 5.2: Increase in power consumption
due to losses.

Pump performance curves can be predicted by means of theoretical or em-

pirical calculation models for each single type of loss. Accordance with the
actual performance curves depends on the models’ degree of detail and to
what extent they describe the actual pump type.

78 78
 Figure 5.3 shows the components in the pump which cause mechanical and
hydraulic losses. It involves bearings, shaft seal, front and rear cavity seal, in-
let, impeller and volute casing or return channel. Throughout the rest of the
chapter this figure is used for illustrating where each type of loss occurs.

Volute

Diffuser

Inner impeller surface

Outer impeller surface

Front cavity seal

Inlet

Bearings and shaft seal

Figure 5.3: Loss causing

components.

79 79
 5. Pump losses

5.2 Mechanical losses

The pump coupling or drive consists of bearings, shaft seals, gear, depending
on pump type. These components all cause mechanical friction loss. The
following deals with losses in the bearings and shaft seals.

5.2.1 Bearing loss and shaft seal loss

Bearing and shaft seal losses - also called parasitic losses - are caused by
friction. They are often modelled as a constant which is added to the power
consumption. The size of the losses can, however, vary with pressure and
rotational speed.

The following model estimates the increased power demand due to losses
in bearings and shaft seal:

Ploss, mechanical = Ploss, bearing + Ploss, shaft seal = constant (5.1)

where
V2
H loss, friktion
Ploss, mechanical = ζ ⋅ H dyn, inpower
= Increased = ζ ⋅demand because of mechanical (5.2)
loss [W]
2g
Ploss, bearing = Power lost in bearings [W]
Ploss, shaftH = Power LV 2lost in shaft seal [W]
loss, pipe = f
seal (5.3)
Dh 2 g

Dh = 4 A (5.4)
5.3 Hydraulic O losses
Hydraulic VD harise on the fluid path through the pump. The losses occur
Re losses
= (5.5)
because of frictionν or because the fluid must change direction and velocity
on its path 64 the pump. This is due to cross-section changes and the
flaminarthrough
= (5.6)
passage throughRethe rotating impeller. The following sections describe the
individual hydraulic losses depending on how they arise.
Q (10/3600) m3 s
Mean velocity: V = = = 3.45 m s
A π
0.032 2 m2
4
VD h 3.45m s ⋅ 0.032m
Reynolds number: Re = = = 110500
ν 1 ⋅ 10 −6 m 2 s
80 0.15mm 80
Relative roughness: k/D = = 0.0047
 5.3.1 Flow friction
Flow friction occurs where the fluid is in contact with the rotating impel-
ler surfaces and the interior surfaces in the pump casing. The flow friction
causes a pressure loss which reduces the head. The magnitude of the friction
loss depends on the roughness of the surface and the fluid velocity relative
to the surface.

Model
Flow friction occurs in all the hydraulic components which the fluid flows
through. The flow friction is typically calculated individually like a pipe fric-
tion loss, this means as a pressure loss coefficient multiplied with the dy-
Ploss, mechanical = Ploss, bearing + Ploss, shaft seal = constant (5.1)
namic head into the component:
V2
H loss, friktion = ζ ⋅ H dyn, in = ζ ⋅ (5.2)
2g
2
where H loss, pipe = f LV (5.3)
Dh 2 g
ζ = Dimensionless loss coefficient [-]
Hloss,friktion
4A
Hdyn, in D= =Dynamic head into the component [m] (5.4)
V
h
O
= Flow velocity into the component [m/s]
VD h
Re = (5.5)
ν
64
flaminar
The friction loss= thus grows quadratically with the flow velocity, see figure
(5.6) 5.4.
V
Re
Figure 5.4: Friction loss as function of
Loss coefficients can be found e.g. in (MacDonald, 1997). Single components the flow velocity.
Q (10/3600) m3 s
such asMean
inlet and outerVsleeve
velocity: = =which are not directly
= 3.45affected
m s by the impeller
A π
0.032 2 m
can typically be modelled with a 4constant
2
loss coefficient. Impeller, volute
housing and return channel will on the contrary typically have a variable loss
coefficient. When the flow 3.45m
VD hin the s ⋅ 0.032m
Reynolds number: Refriction
= = impeller is calculated, the relative
= 110500
ν
velocity must be used in equation (5.2). 1 ⋅ 10 −6
m 2
s
0.15mm
Relative roughness: k/Dh = = 0.0047
32mm

81 81
LV 2 2m ⋅ ( 3.45 m s) 2
 5. Pump losses

Ploss, mechanical = Ploss, bearing + Ploss, shaft seal = constant (5.1)

5.3.2 Mixing loss at cross-section expansion V2

H loss, friktion = ζ ⋅ H dyn, in = ζ ⋅ (5.2)
Velocity energy is transformed to static 2g pressure energy at cross-section ex-
pansions in the pump, LVsee
2
the energy equation in formula (2.10). The conver-
H loss, pipe = f (5.3)
sion is associated with Dh 2agmixing loss.

Dh = is 4A (5.4)
A2
The reason O that velocity differences occur when the cross-section ex- A1
pands, see figureVD h 5.7. The figure shows a diffuser with a sudden expansion
Re = (5.5) V1 V2
beacuse all water ν particles no longer move at the same speed, friction occurs
64
between the molecules in the fluid which results in a diskharge head loss.
flaminar = (5.6)
Even though theRe velocity profile after the cross-section expansion gradually
is evened out, see figure 5.7, a part of the velocity energy is turned into heat
A2
Q
energy instead of static pressure energy. m3 s
(10/3600)
Mean velocity: V = = = 3.45 m s A1 A2
A π
0.032 2 m2
Mixing loss occurs at different places 4 in the pump: At the outlet of the im- A1

peller where the fluid flows intoVD the volute

3.45mcasing or return channel as well
s ⋅ 0.032m
Reynolds
as in the diffuser.number: Re = h
= = 110500
ν 1 ⋅ 10 m 2 s
−6

When designing the hydraulic components,

0.15mm it is important to create small Figure 5.7: Mixing loss at cross-section
Relative roughness: k/Dh = = 0.0047 expansion shown for a sudden expansion.
and smooth cross-section expansions as possible.
32mm

Model 2 2m ⋅ ( 3.45 m s) 2
Pipe loss: H loss, pipe = f LV = 0.031 = 1.2 m (5.7)
D h 2 g is a function
The loss at a cross-section expansion 0.032mof⋅ 2the dynamic
⋅ 9.81 m s head into
2

the component.

V 12
H loss, expansion = ζ ⋅ H dyn,1 = ζ ⋅ (5.8)
2g
2
where  A 
ζ = 1 − 1  (5.9)
 A2into
V1 = Fluid velocity  the component [m/s]
2
 A  V2
The pressure
H loss, contraction = 1 − 0ζ depends
loss coefficient ⋅ 0 on the area relation between the com-
(5.10)
ponent’s inlet and outlet  asAwell
2  as2 g evenly the area expansion happens.
how

V 22
H loss, contraction = ζ ⋅ H dyn,2 = ζ ⋅ (5.11)
2g
86 2 86
ws2 w 1 − w 1, kanal
 LV 2 = 0.031 2m ⋅ ( 3.45 m s) 2
Pipe loss: H loss, pipe = f = 1.2 m (5.7)
D h 2g 0.032m ⋅ 2 ⋅ 9.81 m s
2

V 12
H
For a sudden = ζ ⋅ H
expansion, as dyn,1 = ζ ⋅
shown in figure (5.8)
loss, expansion
2 g 5.7, the following expression is used:
2
 A 
ζ = 1 − 1  (5.9)
 A 2 
2
where  A  V2
H loss, contraction = 1 − 0  ⋅ 2 0 (5.10)
A1= Cross-section areaat inlet A 2  [m ]2 g
A2= Cross-section area at outlet [m2]
V 22
H loss, contraction = ζ ⋅ H = ζ ⋅
The model gives a good estimate of2 gthe head loss at large expansion
dyn,2 (5.11)
ratios
(A1/A2 close to zero). In this case the loss coefficient 2
is ζ = 1 in equation (5.9)
which means that= almostw 2 w − w
ϕ s the = ϕ entire dynamic head into the component is
1 1, kanal
H loss, incidence (5.12)
lost in a sharp-edged diffuser. 2 ⋅g 2⋅g

For small incidence = k 1 ratios

H loss,expansion ⋅ (Q − Qas well)2 +as
k 2 for other diffuser geometries
design
(5.13) with
smooth area expansions, the loss coefficient ζ is found by table lookup
(MacDonalds)
Ploss, disk or= by
kρ measurements.
U 32 D2 ( D2 + 5e )
m
 2ν ⋅ 10 6 
k = 7.3 ⋅ 10 −4
  (5.14)
 U 2 D2 
(n3D52 )A
5.3.3 Mixing loss
(Ploss, disk )A =at( Pcross-section
)
loss, disk B
contraction (5.15) Contraction
(n3D52 )B
Head loss at cross-section contraction occurs as a consequence of eddies being A1

created Q inimpeller
the flow A0 A2
= Q when it comes close to the geometry edges, see
+ Q leakage figure 5.8.
(5.16)
It is said that the flow ’separates’. The 2 reason for this is that the flow because V1 V0 V2
2 ( D2 − Dgap )
2

of the local gap = H stat,

H stat, pressure impeller − ω
gradients fl longer adheres in parallel to the(5.17)
no surface but
8g
instead will follow curved streamlines. This means that the effective cross-
section Harea which the V 2flow L V 2 + 1.0 isVreduced.
+ fexperiences
2
It is said that a(5.18)
contraction
stat, gap = 0.5 Figure 5.8: Loss at cross-section contraction.
2g s 2g 2g
is made. The contraction with the area A0 is marked on figure 5.8. The contraction
accelerates the flow and it must therefore subsequently decelerate again to
(5.19)
fill the cross-section.
2gH stat, gapA mixing loss occurs in this process. Head loss as a
V=
consequence off Lcross-section contraction occurs typically at inlet to a pipe
s + 1.5
and at the impeller eye. The magnitude of the loss can be considerably reduced
Q leakagethe
by rounding = inlet
VA gapedges and thereby suppress separation. If the inlet is
adequately rounded off, the loss is insignificant. Losses related to cross-section
contraction is typically of minor importance.

87 87
 Relative roughness: k/Dh = = 0.0047
32mm
Q (10/3600) m3 s
Mean
5. Pump velocity: V = = = ⋅ 3.45
( 3.45mmss) 2
Pipe losses
2 2m
loss: H loss, pipe = Af LV π =0.032
0.0312 m2 = 1.2 m (5.7)
D h 2 g4 0.032m ⋅ 2 ⋅ 9.81 m s
2

VD h 3.45m s ⋅ 0.032m
Model Reynolds number: Re = ν =2 = 110500
V1 1 ⋅ 10 −6 m 2 s
Based on H loss, expansion = ζit⋅ is
experience, H dyn,1 = ζ ⋅ that the acceleration of the fluid
assumed (5.8)from V
2g 1
to V0 is Relative
loss-free, whereas the 0.15mm
subsequent mixing loss depends on the area
roughness:
2 k/Dh = = 0.0047
ratio now  A1  to the contraction 32mm
ζ =compared
1 − A0 as well as the dynamic head in the
A2 
(5.9)
contraction:
2LV 2 2m ⋅ ( 3.45 m s) 2
Pipe loss: H loss, pipe =Af0  V =2 0.031 = 1.2 m (5.7)
H loss, contraction = 1 − D h 2⋅ g 0 2
0.032m ⋅ 2 ⋅ 9.81 m s (5.10)
 A2  2g

where VV22 2
H
H=loss,
loss, contraction = ζ
= ζ⋅H ⋅ H dyn,2 = ζ ⋅
= ζ ⋅ 2 g1 [m/s] (5.11)
(5.8)
V0 Fluid velocity
expansion indyn,1
contraction 2g Pressure loss coefficient ζ
A0/A2 = Area ratio 2[-] 2 2
 A1  w s w 1 − w 1, kanal
Hζ loss,
= incidence
1 − = ϕ 2 ⋅ g = ϕ (5.12)
(5.9)
 A2  2⋅g A1 A2 A1 A2
The disadvantage of this model is that it assumes knowledge of A0 which is
2
not directly measureable.  The AQ following
V2 2 alternative formulation (5.13)
is therefore
H loss, = k ⋅ (
contraction =1 1 −
loss, incidence
Q − 0
 ⋅ ) 0+ k 2
design (5.10) 1,0
AR = A2 /A1 AR = A1 /A2
often used:  A 2  2 g
0,8

Ploss, disk = kρ U D2 ( D2 + 5eV

3
2 ) 2 0,6
H loss, contraction = ζ ⋅ H dyn,26 =m ζ ⋅ 2 (5.11)
− 4  2ν ⋅ 10 
2g 0,4
k = 7.3 ⋅ 10   (5.14)
 U22D2  w − w
0,2
where 2
w 1 1, kanal 0
= ϕ outs of=the
H loss, incidencehead
Hdyn,2 = Dynamic ϕ component [m] (5.12)
2 ⋅ g (n3D52 )A 2 ⋅ g 0 0,2 0,4 0,6 0,8 1,0
(Ploss,
V2 = Fluid disk )A = out
velocity ( Ploss,of )B component
diskthe [m/s] (5.15) Area ratio
(n3D52 )B
H loss, incidence = k 1 ⋅ (Q − Q design ) + k 2 2
(5.13) Hloss,contraction = ζ . Hdyn,2
Q impeller = Q + Q leakage (5.16)
Figure 5.9 compares loss coefficients at sudden cross-section expansions Hloss,expansion = ζ . Hdyn,1
Ploss, disk == H
H kρ U 32 D2 (−D2ω+2 5(eD)22 −D2gap ) (5.17)
and –contractions
stat, gap as functionmof
stat, impeller fl the area ratio A /A between the
8g 1 2
inlet and
outlet. As  2 ν ⋅ 10 6
 Figure 5.9: Head loss coefficents at sudden
k =shown,
7.3 ⋅ 10 the − 4
 2loss coefficient,
 and thereby also the head loss, is in
(5.14)
U2 D2LV 2than in expansions.2 cross-section contractions and expansions.
generalHsmaller at V 
contractions V This applies in particular
stat, gap = 0.5 +f + 1.0 (5.18)
2g s 2g 2g
at large area ratios. 3 5
(n D2 )A
(Ploss, disk )A = ( Ploss, disk )B (5.15)
(n3D52(5.19)
)B
The head loss 2gH coefficient
stat, gap for geometries with smooth area changes can be
V=
found by Q impeller =L Q+ 1.5
table flookup. + QAs mentioned earlier, the pressure loss in (5.16)
leakage a cross-sec-
s
tion contraction can be reduced2 to ( Dalmost zero by rounding off the edges.
2 − Dgap )
2 2
H
Qstat, gap = H VA impeller − ω fl
stat,gap (5.17)
leakage
8g
2 2 2
Hstat, gap = 0.5 V + f L V + 1.0 V (5.18)
2g s 2g 2g

88 (5.19) 88
2gH stat, gap
 5.3.4 Recirculation loss
Recirculation zones in the hydraulic components typically occur at part
load when the flow is below the design flow. Figure 5.10 shows an example
of recirculation in the impeller. The recirculation zones reduce the effec-
tive cross-section area which the flow experiences. High velocity gradients
occurs in the flow between the main flow which has high velocity and
the eddies which have a velocity close to zero. The result is a considerable Recirculation zones

mixing loss.

Recirculation zones can occur in inlet, impeller, return channel or volute

casing. The extent of the zones depends on geometry and operating point.
When designing hydraulic components, it is important to minimise the size Figure 5.10: Example of recirculation in
of the recirculation zones in the primary operating points. impeller.

Model
There are no simple models to describe if recirculation zones occur and if so to
which extent. Only by means of advanced laser based velocity measurements
or time consuming computer simulations, it is possible to map the recirculation
zones in details. Recirculation is therefore generally only identified indirectly
through a performance measurement which shows lower head and/or higher
power consumption at partial load than predicted.

When designing pumps, the starting point is usually the nominal operating
point. Normally reciculation does not occur here and the pump performance
can therefore be predicted fairly precisely. In cases where the flow is below
the nominal operating point, one often has to use rule of thumb to predict the
pump curves.

89 89
 Dh 2 g

Dh = 4 A (5.4)
O
5. Pump lossesVD h
Re = (5.5)
ν
64
flaminar = (5.6)
Re

Q (10/3600) m3 s
loss V = A = π
Mean velocity:
5.3.5 Incidence = 3.45 m s
0.032 2 m2
Incidence loss occurs when there is4 a difference between the flow angle and
blade angle at the impeller or guide
VD h vane leading
3.45m edges. This is typically the
s ⋅ 0.032m
case at Reynolds or when Re
part loadnumber: =
prerotation=exists. = 110500
ν 1 ⋅ 10 m 2 s
−6

A recirculation 0.15mm
Relativezone occurs on
roughness: k/Done
h =
side of the=blade
0.0047 when there is difference
32mm
between the flow angle and the blade angle, see figure 5.11. The recirculation
zone causes a flow contraction after the blade leading edge.2 The flow must
2
LVcontraction 2m ⋅ ( 3.45 m s) Figure 5.11: Incidence loss at inlet to impeller
once again
Pipedecelerate after
loss: H loss, pipe = f the = 0.031 to fill the entire blade = channel
1.2 m (5.7) or guide vanes.
D h 2g 0.032m ⋅ 2 ⋅ 9.81 m s
2
and mixing loss occurs.

At off-design flow, incidence losses V also

2 occur at the volute tongue. The de-
H loss, expansion = ζ ⋅ H dyn,1 = ζ ⋅ 1 (5.8)
signer must therefore make sure that 2 gflow angles and blade angles match
each other so the incidence 2 loss is minimised. Rounding blade edges and vo-
 A1 
ζ = 1
lute casing tongue − can reduce the incidence loss.
A2 
(5.9)

2
Model  A  V2
H loss, contraction = 1 − 0  ⋅ 0 (5.10)
The magnitude of the incidence A 2  loss 2 gdepends on the difference between rela-
tive velocities before and after the blade leading edge and is calculated using
the following model=(Pfleiderer V 22
H loss, contraction ζ ⋅ H dyn,2 =og
ζ ⋅Petermann, 1990, p 224): (5.11)
2g

W1
2
w2 w 1 − w 1, kanal

,ka
H loss, incidence = ϕ s = ϕ

na
(5.12)

l
2 ⋅g 2⋅g
β´1
β1
where H loss, incidence = k 1 ⋅ (Q − Q design ) + k 2
2
(5.13)
W
1

ϕ = Emperical value which is set to 0.5-0.7 depending on the size of the recir-
Figure 5.12: Nomenclature for incidence loss
culation
Ploss, diskzone
= kρafter
U 32 Dthe blade leading edge. model.
2 ( D 2 + 5e )
ws= difference between relative m velocities before and after the blade edge
− 4  2ν ⋅ 10 
6
using
k = vector
7.3 ⋅ 10calculation,
 see figure 5.12. (5.14)
 U 2 D2 
(n3D52 )A
(Ploss, disk )A = ( Ploss, disk )B (5.15)
(n3D52 )B

90 Q impeller = Q + Q leakage (5.16) 90

 2
 A  V2
H loss, contraction = 1 − 0  ⋅ 0 (5.10)
 A2  2g

V 22
H loss, contraction = ζ ⋅ H dyn,2 = ζ ⋅ (5.11)
2g
Incidence loss is alternatively modelled as a parabola 2 with minimum at the
w 2 w 1 − w 1, kanal
best efficiency point.
H loss, incidence = ϕThe incidence
s
=ϕ loss increases quadratically with
(5.12)the dif-
ference between the design 2 ⋅ g flow and 2the
⋅ g actual flow, see figure 5.13. Hloss, incidence

H loss, incidence = k 1 ⋅ (Q − Q design )2 + k 2 (5.13)

where k2
Qdesign disk = kρ
P=loss,Design U 32[m
flow ( D 2 + 5e )
D23/s]
m
Qdesign Q
k1 = Constant [s22/m 5 6 
ν ⋅ 10
]
k = 7.3 ⋅ 10 − 4   (5.14) Figure 5.13: Incidence loss as function of
k2 = Constant [m]  U 2 D2  the flow.

(n3D52 )A
(Ploss, disk )A = ( Ploss, disk )B (5.15)
(n3D52 )B

Q = Q + Q leakage (5.16)
5.3.6 Diskimpeller
friction
2 − Dgap )
2 2
Disk friction is the 2 ( Dconsumption
H stat, gap = Hincreased power
stat, impeller − ω fl
which occurs on(5.17)
the shroud
8g
and hub of the impeller because it rotates in a fluid-filled pump casing. The
fluid in the cavity between 2 impeller
2 and 2pump casing starts to rotate and
Hstat, gap = 0.5 V + f L V + 1.0 V (5.18)
2 g s
creates a primary vortex, see section 1.2.5.2 g 2 g The rotation velocity equals the
impeller’s at the surface of the impeller, while it is zero at the surface of
the pump casing. The average velocity (5.19) of the primary vortex is therefore as-
2gH stat, gap
sumed to V =be equal to one half of the rotational velocity.
f Ls + 1.5 e
Secondary
vortex
The centrifugal
Q leakage =force
VA gapcreates a secondary vortex movement because of the
difference in rotation velocity between the fluid at the surfaces of the impel-
ler and the fluid at the pump casing, see figure 5.14. The secondary vortex in-
creases the disk friction because it transfers energy from the impeller surface
to the surface of the pump casing.

The size of the disk friction depends primarily on the speed, the impeller di-
ameter as well as the dimensions of the pump housing in particular the dis-
Figure 5.14: Disk friction on impeller.
tance between impeller and pump casing. The impeller and pump housing
surface roughness has, furthermore, a decisive importance for the size of the
disk friction. The disk friction is also increased if there are rises or dents on
the outer surface of the impeller e.g. balancing blocks or balancing holes.

91 91
 0.15mm
Relative roughness: k/Dh = 2 = 0.0047
V 232mm
H loss, contraction = ζ ⋅ H dyn,2 = ζ ⋅ (5.11)
5. Pump losses 2g
2 2m ⋅ ( 3.45 m s )
2
LV 2w =−0.031
w s=2 f D 2 g
Pipe loss: H loss, pipe 1 w 1, kanal = 1.2 m (5.7)
H loss, incidence = ϕ = ϕh 2
0.032m ⋅ 2 ⋅ 9.81 m s (5.12)
2 ⋅g 2⋅g
Model
Pfleiderer
H loss,and Petermann (1990, p. 2 322) use the following model to deter-
incidence = k 1 ⋅ ( Q − Q design ) + V 1k2 2 (5.13)
mine the increased
H loss, expansion = ζ
power ⋅ H consumption
dyn,1 = ζ ⋅ caused by disk friction: (5.8)
2g
Ploss, disk = kρ U 2 2 D 2 ( D 2 + 5e )
3
 A 
ζ = 1 − 1  m
(5.9)
A  2ν ⋅ 10 6 
k = 7.3 ⋅ 102− 4   (5.14)
  U2 D2  2
A0 V0 2
H
where loss, contraction  = 1 −  ⋅ (5.10)
 A 2 (n3D52 )2Ag
D2 = Impeller = ( Ploss, disk[m]
(Ploss, disk )diameter )B (5.15)
A
(n3D52 )B
e = Axial distance to wall at the Vperiphery 2 of the impeller [m], see figure
H contraction = ζ ⋅ H dyn,2 = ζ ⋅ 2 (5.11)
5.14 Qloss, impeller = Q + Q leakage 2 g (5.16)
U2 = Peripheral velocity [m/s] 2 ( D2 −D2 ) 2
H stat, gap = H stat, impeller
w 2 −2 ω fl νw=10 1−w
2 gap (5.17)
ν = Kinematic
H loss, incidenceviscosity
= ϕ s [m= /s], ϕ 8 g [m /s] for water at 20°C.(5.12)
-6 1, kanal
2

k = Emperical value 2 ⋅ g 2⋅g

2 2
Hstat, gap = equals
m = Exponent 0.5 V 1/6 L V
+ ffor smooth V2
+2 1.0surfaces and between 1/7(5.18)
to 1/9
H loss, incidence = k21 g⋅ (Q − Qs design2g ) + k 2 2 g (5.13)
for rough surfaces

If changes Ploss,are2gH
disk =
made
3
kρ U 2 Dthe
stat, gapto
(5.19)
2 + 5e ) of the impeller, calculated disk friction
2 ( Ddesign
V= m
Ploss,disk,A can be fscaled L +− 41.5
 2toν ⋅estimate
10 6
the disk friction Ploss,disk,B at another impel-
k = 7.3 ⋅ 10 s   (5.14)
ler diameter or speed:  U 2 D2 
Q leakage = VA gap
(n3D52 )A
(Ploss, disk )A = ( Ploss, disk )B (5.15)
(n3D52 )B
The scaling
Q equation
= Q + can
Q only be used for relative small design changes.
(5.16)
impeller leakage

( D22 −D2gap )
H stat, gap = H stat, impeller − ω 2fl (5.17)
8g
2 2 2
Hstat, gap = 0.5 V + f L V + 1.0 V (5.18)
2g s 2g 2g
5.3.7 Leakage
Leakage loss occurs because of smaller (5.19) circulation through gaps between
2gH stat, gap
V = and fixed parts of the pump. Leakage loss results in a loss in ef-
the rotating
ficiency because f Lsthe
+ 1.5flow in the impeller is increased compared to the flow
throughQthe entire pump:
leakage = VA gap

92 92
  2ν ⋅ 10 6 
k = 7.3 ⋅ 10 − 4   (5.14)
U DQ  (10/3600) m3 s
Mean velocity: V =2 2 = = 3.45 m s
A 3 5π 0.032 2 m2
(n D24)A
(Ploss, disk )A = ( Ploss, disk )B (5.15)
(n3D52 )B
VD h 3.45m s ⋅ 0.032m
Reynolds number: Re = = = 110500
Q impeller = Q + Q leakage ν 1 ⋅ 10 −6 m 2 s (5.16) Figure 5.15: Types of leakage

( D22 −D2gap )
H stat, gap =roughness:
where Relative H stat, impeller −k/D ωh2fl = 0.15mm = 0.0047 (5.17)
8g
32mm
Qimpeller = Flow through impeller [m3/s], Q = Flow through pump [m3/s] , Qleakage
= Leakage Hstat,flow [m3/s] V 2 + f L V 2 2+ 1.0 V 2
gap = 0.5 2m ⋅ ( 3.45 m s) 2 (5.18)
Pipe loss: H loss,2 g = f s 2LVg = 0.031 2g = 1.2 m (5.7) Qleakage,1
pipe
D h 2g 0.032m ⋅ 2 ⋅ 9.81 m s
2
Leakage occurs many different places in the pump and depends on the pump
(5.19)
type. Figure 5.15 shows
2gH stat, gap where leakage typically occurs. The pressure differ-
V= 2 Leakage between impeller eye and
ences in the pump L which drives the Vleakage flow as shown in figure 5.16.
f s +=1.5
H loss, expansion ζ ⋅ H dyn,1 = ζ ⋅ 1
(5.8) pump casing.
2g
= VA 2gap the impeller and the casing at impeller eye and
Q leakagebetween
The leakage  A 
throughζaxial = 1 relief
− 1 are typically of the same size. The leakage flow Qleakage,2
A between
(5.9)
 2 
guidevane and shaft in multi-stage 2
pumps are less important because both
pressure difference =and1 −  A  V
gap area⋅ are smaller.
2
Q
H loss, contraction 
0

0
(5.10) leakage,1

 A 2  2 g
To minimise the leakage flow, it is important to make the gaps as small as
V2 Leakage above blades in an open
possible. H loss,
When the=pressure
contraction ζ ⋅ H dyn,2 = ζ⋅ 2
difference across the gap is large, (5.11)
it is in par- impeller
2g
ticular important that the gaps are small.
2
w2 w 1 − w 1, kanal
H loss, incidence = ϕ s = ϕ (5.12)
Qleakage,2
Model 2 ⋅g 2⋅g
The leakage can be calculated by combining two different expressions for
in =head
k 1 ⋅ (Q − Q design +leakage,1
2
H loss, incidence
the difference across the) Qgap: k 2 The head difference generated
(5.13) by
Qleakage,1 Qleakage,3
the impeller, equation (5.17) and the head loss for the flow through the gap
equation (5.18).
Ploss, Both expressions
disk = kρ U 2 D2 ( D 2 + 5e )
3
are necessary to calculate the leak flow.
Leakage between guidevanes and shaft in a
m
multi-stage pump
− 4  2ν ⋅ 10 
6
k = 7 . 3 ⋅ 10
In the following an example 
 U D of  the leakage between impeller eye(5.14)
and pump
 2 2 
housing is shown. First the difference in head across the gap generated by
5 Qleakage,2
the impeller is calculated. The (n3D 2 )A difference across the gap depends on
head
(Ploss, disk )A = ( Ploss, disk )B (5.15)
the static head above the impeller (n D2 )B and of the flow behaviour in the cavity
3 5
Qleakage,4
between
Qleakage,1impeller and pump casing: Qleakage,1
Q impeller = Q + Q leakage Qleakage,1
(5.16) Qleakage,3

( D22 −D2gap )
H stat, gap = H stat, impeller − ω 2fl (5.17) Leakage as a result of balancing holes
8g
2 2 2
Hstat, gap = 0.5 V + f L V + 1.0 V (5.18)
2g s 2g 2g

93 (5.19) 93
2gH stat, gap
 2 2g 2m ⋅ ( 3.45 m s) 2
Pipe loss: H loss, pipe = f LV = 0.031 = 1.2 m (5.7)
D h 2g 0.032m
2 ⋅ 2 ⋅ 9.81 m s
2
w 2 w 1 − w 1, kanal
5. Pump losses
H loss, incidence = ϕ s =ϕ
2 ⋅g 2⋅g
(5.12)

V2
H
Hloss, expansion==kζ⋅ ⋅( Q
H dyn,1
− Q = ζ )⋅2 + 1k (5.8)
(5.13)
loss, incidence 1 design 2 g2
where 2
ωfl Pζ ==1 A1  3 velocity of the fluid in the cavity between impeller
Rotational
−
 = kρ U 2 D2 ( D2 + 5e ) (5.9)
 and
loss, disk
A2pump
 casing m [rad/s]
 2ν ⋅ 10  2 6
Dgap k = = 7. 3 Inner
⋅ 10 − 4diameter
  A 0 of the V0gap
2 [m] (5.14)
H
Hstat, impellerloss,
=contraction
Impeller = U
1 −
 static D
2 2 head ⋅ rise [m] (5.10)
 A2  2g
(n3D52 )A
(
The headloss,P )
difference
disk A = ( P across
loss, )
disk B the gap (5.15)
5 Vcan also be calculated as the head loss of
2
Low pressure High pressure

H loss, contraction = ζ ⋅ H dyn,2 (= n3Dζ ⋅2 )B 2 (5.11)

the flow through the gap, see figure2 g 5.17. The head loss is the sum of the fol- Figure 5.16: The leakage is drived by the
pressure difference across the impeller.
lowing Q three
impellertypes= Q of+Q losses:
leakage Loss due to sudden 2 contraction when(5.16)
the fluid
w 2 w 12 − w21, kanal
runs into the gap,=friction
H loss, ϕ s loss
= ϕ between
( D −D fluid ) and wall, and mixing loss due
(5.12)
gap = H stat, impeller
H incidence 2 ⋅ g − ω 2fl 2 2 ⋅gap g (5.17)
to suddenstat,expansion of the outlet of8the
g gap.

H loss, incidence = kV1 2⋅ (Q − QL design

2 )2 + k 2 (5.13)
Hstat, gap = 0.5 + f V + 1.0 2 V (5.18)
2g s 2g 2g
where Ploss, disk = kρ U 2 D2 ( D2 + 5e )
3

f = Friction coefficient m (5.19)

2gH stat,− 4gap 2[-]
ν ⋅ 10 6 
V
k = = 7 . 3 ⋅ 10   (5.14)
L = Gap length [m]
f Ls + 1.5  U 2 D2 
s = Gap width [m]
V = Fluid leakage = VA
Q velocity in gap
gap [m/s] (n3D52 )A
(Ploss, )
disk A = ( P )
loss, disk B (5.15)
Agap = Cross-section area of gap (n3D 5
[m2 )2B]
Q impeller = Q + Q leakage (5.16)
The friction coefficient can be set to 0.025 or alternatively be found more Figure 5.17: Pressure difference across the
precisely in gapa Moody ( D22 −5.6.
D2gap )
H stat, = H stat,chart, see 2figure
impeller − ω fl (5.17) gap through the friction loss consideration.
8g

By isolating the velocity 2 V in the

2 equation
2 (5.18) and inserting H from
Hstat, gap = 0.5 V + f L V + 1.0 V stat,gap
(5.18)
2 g s 2 g 2 g
equation (5.17), the leakage can be calculated: L

(5.19)
2gH stat, gap
V=
f Ls + 1.5
s

Q leakage = VA gap

Dspalte
D2

94 94
 5.4 Loss distribution as function of specific speed
The ratio between the described mechanical and hydraulic losses depends
on the specific speed nq, which describes the shape of the impeller, see sec-
tion 4.6. Figure 5.18 shows how the losses are distributed at the design point
(Ludwig et al., 2002).

Flow friction and mixing loss are significant for all specific speeds and are the
dominant loss type for higher specific speeds (semi-axial and axial impellers).
For pumps with low nq (radial impellers) leakage and disk friction on the hub
and shroud of the impeller will in general result in considerable losses.

At off-design operation, incidence and recirculation losses will occur.

η [%] 100 Mechanical loss

Leakage loss
95
Disk friction
90
Flow friction and mixing losses
85
Hydraulic efficiency
80

75

70

65

60 Figure 5.18: Loss distribution in a centrifugal

55 pump as function of specific speed nq
10 15 20 30 40 50 60 70 80 90 (Ludwig et al., 2002).
nq [min -1]

5.5 Summary
In this chapter we have described the individual mechanical and hydraulic
loss types which can occur in a pump and how these
� � � � � ���losses affect flow, head

and power consumption. For each loss type we �have

�� made a simple physical
� � �� ���� � � ��� �
description as well as shown in which hydraulic
� � ��components
� the loss typi-
cally occurs. Furthermore, we have introduced some simple models which
can be used for estimating the magnitude of the losses. At the end of the
chapter we have shown how the losses are distributed depending on the
specific speeds.

95 95