IOP Conference Series: Earth and Environmental Science

Related content
- Inner flow characteristics of self-priming
Research on design multi-points performance irrigation pumps
D H Xu, H Li and L Li
curves of pump - Static stress and modal analysis on the
impeller of screw centrifugal pump
S Q Yuan, T Li, J P Yuan et al.
To cite this article: L Li et al 2012 IOP Conf. Ser.: Earth Environ. Sci. 15 032018
- Experimental studies on the optimization
design of a low specific speed centrifugal
pump
J F Zhang, Y Yuan, S Q Yuan et al.
View the article online for updates and enhancements.

This content was downloaded from IP address 27.59.163.167 on 11/04/2019 at 07:18

26th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 15 (2012) 032018 doi:10.1088/1755-1315/15/3/032018

Research on design multi-points performance curves of pump

L Li, H Li and D H Xu
Research Center of Fluid Machinery Engineering and Technology of Jiangsu
University Zhenjiang, Jiangsu, 212013, China

E-mail: [email protected]

Abstract. The centrifugal pump’s performance curves are the most important chart to reveal
relevance of different performance parameters. They can show pump’s function
comprehensively and graphically. However these curves can’t be predicted and designed
precisely, duo to the complicated flow inside the impeller and the incomplete way to design
pump. The complete shape of performance curves can be gained only after test. With the
development of industry, many applications need the pump operating well at different flow
conditions. It means the pump’s performance curves should pass some specific points. This is a
problem to the designer who still uses the traditional way to design pumps. In this paper, the
Design of experiments was applied to arrange a plan of experiments. Because the theory
equations of performance curves contain many geometry factors of impeller, changing these
factors have different influence on the shape of curves, the relationship between geometry
factors and the performance under different operation points been attained after using variance
analysis to deal with experiment data. The relevant regression models and graphs were drawn
to help understand these relationships. Depending on the predicted values of geometry factors
pump’s impeller was redesigned, and the pump’s performance been simulated for saving time
and cost. Test shows that the shape of performance curves satisfy design objective, this
example can be taken as a reference of pump’s specific designs.

1. Introduction
Generally the centrifugal pumps performance curves which set the flow Q as abscissa and head H,
efficiency η , shaft power P, NPSH as ordinate are the most important chart data of pumps and it
shows the relationship of different performance parameters. However the traditional way of pump
design are rely on one flow point and the flow in the impeller is complicated, the shape of
performance curves at off-design operation points can’t be predicted or designed precisely. The details
of these curves only can be gained after test. Based on the traditional way to design pumps which
performance curves asked to pass multi specific points is difficult. Though as known form the theory
equations of performance curves, many geometric factors of impeller have particular influence on the
shape of curves. It is possible to control these curves shape by picking properly geometry factors and
selecting their values optimized. Yuan Shouqi studied the hump of performance curves of low ns
centrifugal pump when they are operating at the low flow areas, supplied advises which mainly talking
about change values of some geometric factors to eliminate this hump phenomenon [1]. Fu Qiang gets
a specific curve by combing different impeller in a multistage pump [2]. Some scholars studied the
relevance between the wrap angle of the blade and impeller diameter and the performance curves [3-4].
The similar research also include performance of pumps under the off-design operation fields [5-6].
But these previous study didn’t refer to the design multi-points of performance curves, they only have

Published under licence by IOP Publishing Ltd 1

26th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 15 (2012) 032018 doi:10.1088/1755-1315/15/3/032018

given roughly describe about the whole shape of curves influenced by the geometric factors. Recently
CFD is already applied in the design of pumps extensively [7-9]. Many software can help designer to
save time and cost of experiments, like ANSYA or FLUENT, etc. So ANSYS is chosen as the text
environment of this experiment.
2. Theoretical analysis
The theory equations of pumps flow-head curves can be gained under some conditions. The vu1 is
supposed as 0, and take accordance of limited impeller’s influence, H t − Qt which is a straight line can
be gained as Eq.(1).
u2  Q 
Ht =  σu 2 − t cot β 2  (1)
g  F2 
The hydraulic loss, volume loss in the impeller been taken away from the Eq. (1). Then the theory
equation is shown in Eq. (2).
u2  Q 
H =  σu 2 − cot β 2  − h f − h j − q (2)
g F2 

The coefficient F2 described as:

F2 = 2πR2b2ψ 2 (3)
o
The H t − Qt is a rise line when the impeller outlet blade angle β 2 < 90 . H t − Qt is a declining line
o o
when β 2 = 90 , H t − Qt is a horizontal line when β 2 = 90 ,based on the Eq.(1). The coefficient ψ 2 is to
describe the size between adjacent blades, so it’s effective to change ψ 2 by selecting different blade
wrap angle. The H t − Qt turned more horizontal by increasing b2 .So it is possible to create the expect
curves shape by selecting the properly factors and giving them optimized values.

3. Design of experiments
DOE is a useful way to study the right plan of experiments and analysis the text data. It can get
knowledge and information from a test through input the different values of factors and observe the
change of output responses.
Table 1 Level of factors
-1 0 1
A/° 150 165 180
B/° 30 33 36
C/mm 4 5.5 7
Table 2 Detail of factorial experiment
order A B C order A B C
1 1 -1 -1 6 1 1 -1
2 1 -1 1 7 -1 1 1
3 0 0 0 8 0 0 0
4 1 1 1 9 -1 -1 1
5 -1 1 -1 10 -1 -1 -1

2
26th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 15 (2012) 032018 doi:10.1088/1755-1315/15/3/032018

The importance of different input factors can be confirmed, and the way that how these factors to
influence output responses also can be gained. DOE is already employed in many fields [10-13]. It is
efficient to obtain favorably test results via arranging a good plan of experiments which can help
saving cost and time. In this paper Factorial design were applied to plan a experiments. To control the
performance curves to pass specific points, the relationship between chosen geometric factors of
impeller and the performance at the specific operation points should be obtained, or the regression
equations should be fitted. So three impeller geometric factors been selected in this case, they are
impeller outlet blade angle β 2 (B), blade wrap angle θ (A), blade outlet width b2 (C).The responses are
value of efficiencies when the flow is at 0.5Q, 0.8Q, Q, 1.1Q and 1.3Q. The model of experiment was
set as 23+2.The design factors of pump which was applied to test are as follows: Q is 24.75m3/h, H is
51.06m and shaft speed P is 2900r/min. The values range of these three input factors are shown in
Tab.1. The detail of experiment is listed in table2.

4. Numerical simulation
The calculating domains were created based on the pump geometries and the whole flow domains
consisted of three fluid domains: inlet section, impeller, volute with outlet section, shown in Fig.1. The
inlet section and outlet were respectively extended to three times and five times their diameters so that
the flow can be fully developed at the two sections. After modeled in Pro/e, the computational
domains are imported to CFX for further processing. Then the surface roughness was set as 25 micron
for all the flow domains. The mesh generation process was carried out by the ICEM CFD, and the
whole flow fields were meshed unstructured grids. The Grids number of impeller is 0.57million and
the whole grids number is 1.3million. The pitch change for the interface between rotor and stator was
set as Frozen Rotor. The turbulence is set as low with the intensity of 1%. The flow through these ten
pumps was simulated with the commercial CFX 12.1, which used the finite volume method to solve
the Reynolds averaged Navier-Stokes equations for 3D incompressible steady flow. The flow model
was applied standard k − ε model. The convergence precision was set as 10-4.

4.1. Simulation results

The simulation results of flow-efficiency curves of ten impellers are shown in Fig.1. The efficiency of
impeller 1,5,10 is highest at the operation point 0.5Q, but their efficiency is reduced below 50% at the
point 1.3Q, shown in fig.1.

Fig. 1 Flow-efficiency curves of impellers

3
26th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 15 (2012) 032018 doi:10.1088/1755-1315/15/3/032018

The curves shape of impeller 1,5,10 is high at the low flow operation point, and drop down quickly
at the large flow operation point. In contrary, the curves shape of impeller 2, 4 low at the 0.5Q, but rise
highly at the 1.3Q, the two impeller have the highest efficiency at the point 1.3Q,but their efficiency
start very low at the point 0.5Q.shown in Fig.1.All the ten flow-efficiency curves can be classify as
two groups, the first group which contain impeller 2, 4 is starting low but rising quickly, the second
group which contain impeller 1, 5, 10 is standing high at the beginning but dropping deeply at the
large flow operation fields.
In order to analyse the influence of impeller geometries on the pump performance, the inner flow
patterns were analysed for analysing the impeller performance. The velocity vector graphics and
pressure distribution graphics are shown in figure 2, 3 respectively.

impeller1 impeller5 impeller10 impeller2 impeller4

Fig. 2 Velocity vector graphics of impellers at 0.5Q

impeller1 impeller5 impeller10 impeller2 impeller4

Fig. 3 Pressure distribution of impellers at 0.5Q
Because the flow decreased much and was deviated the plan route in the impeller, which was
causing mass of turbulence and backflow in the flow passage shown in Fig.1.The efficiency of
impellers became very low at the 0.5Q, compared with design point Q. The area of backflow in
impeller2, 4 almost occupied all the passage, is larger than the backflow area of impeller 1, 5, 10, that
caused a great deal of hydraulic losses. The pressure of second group impellers at the outlet is higher
than first group, which means the second group can supply more energy to the fluid so they have a
better performance at point 0.5Q than first group, shown in Fig.2. Hence decreasing the b2 can make
the curves tally at the low flow point for the second group impeller having a narrow outlet width,
shown in Tab.2. The blade wrap angle could increase due to impeller 1, 2 which own a largest blade
wrap angle have the best performance near the design point Q to improve the whole fields efficiency.
The performance of impellers at point 1.3Q is shown in Fig.4, 5.

impeller1 impeller5 impeller10 impeller2 impeller4

Fig. 4 Velocity vector graphics of impellers at 1.3Q

4
26th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 15 (2012) 032018 doi:10.1088/1755-1315/15/3/032018

impeller1 impeller5 impeller10 impeller2 impeller4

Fig. 5 Pressure distribution of impellers at 1.3Q
When the flow is at 1.3Q point, impeller 2, 4 which own wide b2 and large θ have better
performance than the second group impellers shown in Fig.4, 5. The passage of second group blocked
by the redundant flow, it caused pressure down at the volute and outlet. So an ideal shape of flow-
efficiency curve couldn’t be gained unless balance the different geometric factors well and select the
values of factors optimal.

5. variance analysis
In order to express the function of geometric factors clearly, the Praeto diagram, main effect and
interaction effect chart of responses been drawn, shown in Fig.6, 7 ,8. The interaction effect of factors
defined as AB, AC, BC.
3.18 3.182 3.182
C C A
A A C
B B AC
AC AC BC
BC AB B
AB BC AB
0 10 20 30 40 50 0 1 2 3 4 5 6 7 8 9 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
0.5Q 0.8Q 1.3Q

Fig. 6 Praetor diagram of responses

To simplify regression model needs removing some input factors which are not significant factors.
It is obvious to know the importance of each factor from Fig.6. The t-test absolute values of effects
been set as abscissa, the importance of input factors been ranked as ordinate. So the first one factor
who have the mostly influence on response 0.5Q is C, namely b2. The interaction effects have more
influence on response with the flow increasing, shown in Fig.6.
0.5Q 1.3Q
A B A B
52 54
48 52
44 50
48
40 46
150 165 180 30 33 36 150 165 180 30 33 36
C C
52 54
48 52
44 50
48
40 46
4.0 5.5 7.0 4.0 5.5 7.0

Fig. 7 Main effect chart of 0.5Q and 1.3Q responses

The effect of b2 and β 2 at 0.5Q and 1.3Q is in contrary, shown in Fig.7. It is essential to selecting the
properly values of b2 and β 2 to get a balance shape of performance curve.

5
26th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 15 (2012) 032018 doi:10.1088/1755-1315/15/3/032018

0.5Q 1.3Q
30 33 36 4.0 5.5 7.0
30 33 36 4.0 5.5 7.0
50 55
A 45 50
A
40
45
55
50
B 45 B 50
40 45

C C

Fig. 8 Interaction effect chart of 0.5Q, 1.3Q responses

After collecting information of input factors and response, 0.5Q、 0.8Q、1.3Q been set as the
control points on the curves. Because there is no one impeller which efficiency exceeded 50% at both
0.5Q and 1.3Q point. All of the flow-efficiency curves shapes are asymmetric between 0.5Q to 1.3Q.
So in that case, the improvement plan is to design an impeller which has a flow-efficiency curve
passing or up 50% value of efficiency at both point 0.5Q and 1.3Q, also keeps the design point
efficiency not dropping down. The regression model of response is as follows:
y = β 0 + β1 x1 + β 2 x2 + β12 x1 x2 + ... + ε

the β x ， β x stand for the main effect, the β x x stands for the interaction effect, the ε is error. The
1 1 2 2 12 1 2
accuracy of regression model can be evaluated by some coefficient. As known from the Squares of the
factorization formulation of regression analysis, relevance parameters can be defined as follows:
SSTotal = SS Model + SS Error

the square of R been defined as R-Square, R-sq.

SS Model SS
R 2 = R − sq = = 1 − Error
SSTotal SSTotal

the regression model is more accuracy when the R-sq close 1. The square of Radj been defined as:
2 SS Erroe (n − p )
R − sq(adj ) = Radj = 1−
SSTotal (n − 1)

the values of R-sq and R-sq (adj) more closer means the regression model better. The n is number of
observed value. The p is total number of regression model. The value of parameter R-sq (adj) was
gained after removing some unimportant factors. Tab.3 is the comparison of parameters belonging to
0.5Q, 0.8Q and 1.3Q responses, respectively.

Table 3 Comparison of coefficient

R-sq R-sq(adj)
0.5Q 99.72% 99.58%
0.8Q 97.53% 95.55%
1.3Q 78.96% 6213%
After all test data were treated by variance analysis, the regression model of 0.5Q, 0.8Q, 1.3Q can
been obtained as follows:
Y0.5Q = 69.304 + 0.0498 A − 0.235 B − 4.341C

6
26th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 15 (2012) 032018 doi:10.1088/1755-1315/15/3/032018

Y0.8Q = 50.411 + 0.077 A − 0.09 B − 1.096C

Y1.3Q = 49.897 − 0.152 A − 3.207C + 0.044 AB − 0.08 B

6. result of experiments
Based on the requirements of this case, the efficiency of 0.5Q and 1.3Q are required as the same, the
values of them need to around 50%. The efficiency of 0.8Q point shouldn’t low than the average value
of ten impellers. So the expectedvalues of Y were drafted as Tab.4.
Table 4 Optimizing response values
responses Lower limit Expect value Higher limit
0.5Q 48 50 52
0.8Q 54 55.5 60
1.3Q 48 50 52
Depending on simultaneous solving regression models, the values of A, B, C were gained as 171,
34.8, 4.5 respectively, namely blade wrap angle is 171°, impeller outlet angle is 34.8°, blade outlet
width is 4.5mm. Through the same test environment the optimized impeller was modelled and
numerical simulated. The performance of new impeller is shown in Fig.9.

Fig. 9 Comparison of flow-efficiency curves

The flow-efficiency curves of optimized impeller is gently and symmetrical shown in Fig.9. The
values of 0.5Q, 1.3Q, 0.8Q are 51.37%, 52.38%, 56.19%. The efficiency absolute difference between
0.5Q and 1.3Q is 1.01, satisfying the design objective. The optimized impeller has the best efficiency
at the point Q because the efficiency of 0.8Q is controlled well. The favourable shape of flow-
efficiency curve of optimized impeller shows this experiment is successful.

7. Conclusions
This experiment indicates that multi-points performance curves of centrifugal pumps can be designed
by combining design of experiment and numerical simulation. The relationship between impeller
geometric factors and performance at different operation point can be gained applying variance
analysis to deal with test data. The main effect and interact effect of impeller geometric factors is
changing gradually with the flow increasing. Balancing efficiency at low and large flow depends on

7
26th IAHR Symposium on Hydraulic Machinery and Systems IOP Publishing
IOP Conf. Series: Earth and Environmental Science 15 (2012) 032018 doi:10.1088/1755-1315/15/3/032018

selecting the optimized values of different geometry factors, which can based on the result of solving
regression models.
The importance sequence of factors at 0.5Q is C, A, B. When the flow increased to 1.3Q the
sequence became A, C, AC. The velocity vector chart shown that decreasing C can narrow the passage
to reduce the hydraulic lose caused by mass turbulence and backflow existing. Selecting the proper
values of A, B can control the area ratio of impeller outlet and inlet to make the flow smooth and
decrease the friction loses. The values of optimized impeller efficiency at point 0.5Q, 1.3Q, 0.8Q are
51.37%, 52.38%, 56.19%, respectively. The flow-efficiency curves of optimized impeller is gently and
symmetrical which reach the design objective.

References
[1] Yu S Q and Zhang Y Z 1998 J. China agricultural machinery 29(2) 171-173
[2] Fu Q and Zhu R S 2010 J.Drainage and irrigation machinery engineering 28(3) p 185-189
[3] Yany H and Liu C 2007 J. Chinese journal of mechanical engineering 43(10) 166-169
[4] Tan M G and Liu H L 2009 J. Drainage and irrigation machinery 27( 5) 314-318
[5] Wang H and Tsukamoto Hiroshi 2003 J. Fluids engineering 125( 9) 767-778
[6] Wuibaut G 2002 J. Fluids engineering 124(9) 791-797
[7] Liu J R, Zhang L S and Xiang H J 2010 J. Drainage and irrigation machinery engineering 28(5)
394-397
[8] Zhang D S and Shi W D 2008 J.China rural water and hydropower 32(7) 95-97
[9] Pan Z Y and Xie R 2009 J. Drainage and irrigation machinery 27( 5) 319-322
[10] Ren K, Wang J J and Wu D H 2010 J. China mechanical engineering 21(3) 368-372
[11] Hou B X,Chen J C 2010 J.Science and technology information 29(2) 156-158
[12] Chen Y 2010 J. Applied statistics and management 29(5) 820-828