Journal of Chemical Engineering of Japan, Vol. 44, No. 11, pp.

840–844, 2011 Research Paper

Comparison of Power Number for Paddle-Type Impellers by Three Methods

Ming-hui XIE1,2, Guo-zhong ZHOU 2, Jian-ye XIA1,
Chen ZOU 2, Pei-qing YU 2 and Si-liang ZHANG1
1
State Key Laboratory of Bioreactor Engineering,
College of Biotechnology,
East China University of Science and Technology,
Shanghai 200237, China
2
Wenzhou Great Wall Mixer Design Institute,
Wenzhou 325019, China

Keywords: Power Number, Empirical Correlations, Computational Fluid Dynamics, Paddle Impeller

Power numbers in a baffled stirring tank with paddle-type impellers were measured over a range of Reynolds
numbers, from laminar- to turbulent-flow regions. The impellers studied included a two-blade flat-paddle im-
peller, a 45° two-blade pitched-paddle impeller, a 45° four-blade pitched-blade turbine, and a two-stage pitched-
blade turbine. The impeller-diameter-to-tank-diameter ratio was 0.5 to 0.6. The measured power numbers were
compared with those derived from two other methods: prediction by using correlations and computational fluid
dynamics (CFD). Power number correlations showed that the values estimated from the empirical correlations
proposed by Nagata agreed closely with those measured in the laminar-flow region but deviated significantly in
the transitional- and turbulent-flow regions. The power numbers derived from the correlation proposed by
Kamei and Hiraoka agreed well with the experimental results in the laminar- and turbulent-flow regions. The
power number in a stirring tank was also simulated by the CFD method. The numerical results showed satisfac-
tory agreement with the experimental data over a wide range of Reynolds numbers.

(CFD) techniques are being used as substitutes for ex-

Introduction
periments to calculate power consumption and identify
Mixing and stirring tanks are widely used in many mixing characteristics (Shekhar and Jayanti, 2002; Driss
industries such as chemical, biochemical, food, and et al., 2006; Suzukawa et al., 2006). One advantage of
water-treatment industries. Because of the lack of accu- CFD-based prediction methods is that they are quick and
rate methods for power consumption calculation, many easy to perform with models of various scales. Shekhar
industrial mixing operations are inefficient and cause and Jayanti (2002) studied the power number of an
wastage of energy. Power number is the basic parameter eight-blade paddle impeller in an unbaffled vessel by
used to design an agitator. This parameter also influ- using CFD methods and found that the simulated data
ences the initial equipment investment and running were lower in value than those predicted by the Nagata
costs. However, very few research papers have discussed correlation.
impeller power numbers in recent years. The purpose of the present study is to compare the
Paddle and turbine impellers are commonly used in power numbers (NP) obtained from experiments, empiri-
industrial mixing operations. Methods for calculating the cal correlations, and CFD methods for different im-
power consumption of these impellers have been exten- pellers. In this study, a two-blade flat-paddle impeller (2-
sively studied and reported in the literature. To estimate FP), a two-blade pitched (45°)-paddle impeller (2-PP), a
power consumption, the correlation method developed four-blade pitched (45°)-blade turbine (4-PBT), and a
by Nagata (1975) for two-blade paddle impellers has tra- two-stage 4-PBT have been used.
ditionally been used. Kamei et al. (1996) and Hiraoka et
al. (1997) developed a new correlation method to esti-
1. Method
mate the power consumption of paddle, anchor, and heli-
cal ribbon impellers. Computational fluid dynamics 1.1 Experiments
As shown in Figure 1, the stirring tank used is a
Received on June 7, 2011; accepted on July 7, 2011 flat-bottom cylindrical tank with an inner diameter D of
Correspondence concerning this article should be addressed to 0.58 m; the liquid height H is equal to D for the one-
G. Z. Zhou (E-mail address: [email protected]); J. Y. Xia stage impellers, and H  1.2D for the two-stage impeller.
(E-mail address: [email protected]). Four baffles of width BW  D / 10 were installed with an
Presented at the 3rd Asian Conference on Mixing, at Jeju Island,
October, 2010
off-wall clearance of 0.008 m. The impeller-diameter-to-

840 Copyright © 2011 The Society of Chemical Engineers, Japan

 Table 1 Detailed parameters for the impellers Table 2 Nagata power number correlation for paddle im-
pellers
Number Name d/D b/d C/D h/D
Unbaffled condition
1 2-FP 0.6 0.109 0.35
2 2-PP 0.6 0.109 0.35 NP0  A / Re  B[(1000  1.2Re0.66) / (1000  3.2Re0.66)]p
3 4-PBT 0.6 0.092 0.35 (H / D)(0.35  b / D)sin1.2 θ
Double-stage
4 0.5 0.083 0.30 0.6 A  14  (b / D)[670(d / D  0.6)2  185]
4-PBT
B  10[1.3  4(b /D 0.5)  1.14(d /D)]
2

p  1.1  4(b / D)  2.5(d / D  0.5)2  7(b / D)4

Re  Nd 2ρ / μ
Fully baffled condition (Nagata, 1975)

NPmax is calculated through NP0 calculation formula under

unbaffled condition, where Re is substituted by Reθ .
Reθ  104(1  sin θ)25(d / D  0.4)2 / (b / D)
 {(b / D) / [0.11(b / D)  0.0048]}
Partly baffled condition (Wang, 2002)
(NPmax  NP) / (NPmax  NP0)  [1  2.9(BW / D)1.2nb]2

Fig. 1 Schematic diagram of the experimental apparatus

Table 3 Kamei and Hiraoka power number correlation for
paddle impellers
tank-diameter ratio d / D was between 0.5 and 0.6, and
the impeller off-bottom clearance C was 0.35D for the Unbaffled condition
one-stage impellers and 0.3D for the two-stage impeller.
The detailed parameters for the impellers are shown in NP0  {[1.2π 4β 2] / [8d 3 / (D2H)]}f
Table 1. The experimental procedure has been described f  CL / ReG  Ct{[(Ctr / ReG)  ReG]1  ( f∞ / Ct)1 / m}m
in detail elsewhere (Chen et al., 2010). The impeller ReG  {[πη ln(D / d )] / (4d / β D)}Re
Reynolds number Re was calculated using the relation
Re  Nd 2ρ / μ, where N is the rotational speed; ρ , the CL  0.215η np(d / H )[1  (d / D)2]
liquid density; and μ , the liquid viscosity. The power  1.83(b sin θ / H)(np / 2 sin θ )1 / 3
consumption P was determined from measurements of Ct  [(1.96X 1.19)7.8  (0.25)7.8]1 / 7.8
the shaft torque M and rotational speed N, using the for- m  [(0.71X 0.373)7.8  (0.333)7.8]1 / 7.8
mula
Ctr  23.8(d / D)3.24(b sin θ / D)1.18X 0.74
P  2π NM (1) f∞  0.0151(d / D) Ct0.308
and the NP value of the impeller was calculated from the X  γ np0.7b sin1.6 θ / H
following relation:
β  2 ln(D / d) / [(D / d)  (d / D)]
NP  P / ρ N 3d 5 (2) γ  [η ln(D / d ) / ( β D / d 5)]1 / 3
1.2 Correlation method η  0.711{0.157  [np ln(D / d )]0.611} / {np0.52[1  (d / D)2]}
Empirical correlations are commonly used to pre- Fully baffled condition
dict the NP values of impellers because they are simple
and time-saving. The Nagata correlation and the Kamei Flat paddle: NPmax  10(np0.7b / d)1.3 (np0.7b / d )  0.54
and Hiraoka correlation (see Tables 2 and 3) are used for
paddle impellers with various geometries. In this study,  8.3(np0.7b / d ) 0.54  (np0.7b / d )  1.6
we assumed that Re  200 in the laminar-flow region  10(np b / d )
0.7 0.6
1.6  (np0.7b / d)
and calculated NP under unbaffled conditions in this re- Pitched paddle: NPmax = 8.3(2θ / π )0.9(np0.7b sin1.6 θ / d)
gion. For Re  200, we calculated NP under fully baffled
Partly baffled condition
or partly baffled conditions.
To obtain the NP correlations for the two-stage im- NP  [(1  x3)1 / 3]NPmax
peller, we referred to the Design Handbook for Practical x  4.5(Bw / D)nb0.8 / {(2θ / π )0.72NPmax0.2}  NP0 / NPmax
Subject of Chemical Engineering (Part II) (Wang, 2002).

841
Fig. 2 Comparison of power numbers of 2-FP obtained by Fig. 3 Comparison of power numbers of 2-PP obtained by
the three methods the three methods

In this investigation, NP for the two-stage 4-PBT was mately 10% larger than the measured values. There is
twice that of the one-stage impeller. good agreement between the Kamei and Hiraoka corre-
1.3 Numerical method lation values and the experimental values in the transi-
1.3.1 Computational grids Grids for the tank domain tional-flow region, where Re is in the range 300 to 2000.
and the impeller domain were generated using the com- The Nagata correlation predicts higher values in the
mercial software GAMBIT 2.3 (ANSYS, Inc.). The en- transitional-flow region, while CFD predicts lower val-
tire tank was modeled. The grid density was about ues. In the turbulent-flow region, the simulated values
550,000 tetrahedral cells for the one-stage impellers and and correlation values are about 10–15% lower than the
about 670,000 tetrahedral cells for the two-stage im- experimental values.
peller. In order to achieve high resolution, the grids were For 2-FP, the Kamei and Hiraoka correlation gives
refined near the impeller, rotating interface, and tank good results over a wide range of Re values, from the
wall. laminar-flow region to the turbulent-flow region. The
1.3.2 Calculation method All the calculations were Nagata correlation and CFD provide accurate predic-
carried out with the commercially available software tions in the laminar- and turbulent-flow regions.
FLUENT6.3 (ANSYS, Inc.), which is a general-purpose 2.2 Power number of 2-PP
computer program, using a finite volume method. The NP values of 2-PP from CFD, the Nagata corre-
Multiple reference frame (MRF) was employed for the lation, the Kamei and Hiraoka correlation, and experi-
flow between the rotating impellers and the stationary ments are compared in Figure 3. In the laminar-flow re-
baffles. The pressure-velocity coupling was derived gion, the Nagata correlation and the Kamei and Hiraoka
using the SIMPLE algorithm. An accurate second-order correlation give NP values that are about 23% and 15%
upwind scheme was used for the discretization method. larger than the experimental value, respectively. The
The laminar model was chosen for laminar flow (Re  CFD results agree well with the experiment values. In
200), and the standard k-epsilon turbulence model was the transitional-flow region and the low-Re turbulent-
chosen for transitional and turbulent flows. flow region, all the three methods overpredict NP, espe-
cially the Nagata correlation. However, in the high-Re
turbulent-flow region, the data from the three methods
2. Results and Discussion
correlate well with the data from the experiments, pre-
2.1 Power number of 2-FP dicting an NP that is about 5–10% lower than the value
The NP values of 2-FP obtained by CFD, the Nagata obtained from the experimental data.
correlation, and the Kamei and Hiraoka correlation are 2.3 Power number of 4-PBT
compared with the experimental results in Figure 2. It The Nagata correlation is usually used for two-
can be seen that every method predicts the correct trend blade impellers, whereas in the present study, we used a
of the NP curve. NP initially decreases as Re increases four-blade pitched blade turbine. As Nagata suggested,
and then becomes reasonably constant in the turbulent- we changed the four-blade PBT to a two-blade pitched
flow region. However, the values obtained by the three paddle that has an equivalent blade area.
methods show different deviations in different flow re- As shown in Figure 4, the Kamei and Hiraoka cor-
gions. In the laminar-flow region, the NP values derived relation value shows the largest deviation from the val-
from the three methods agree well with the experimental ues determined by other methods in the laminar-flow re-
values. The values determined by the Nagata correlation gion. The deviation is about 10–20% from the experi-
and the Kamei and Hiraoka correlation are approxi- mental values. The results obtained for the other three

842 JOURNAL OF CHEMICAL ENGINEERING OF JAPAN

 correlation is nearly 50–80% higher than the experimen-
tal values. Therefore, this method should not be used
under the present experimental conditions. In the turbu-
lent-flow region, the Kamei and Hiraoka correlation val-
ues and the CFD predictions correspond well with the
experimental values.

Conclusions
NP values of paddle and turbine impellers in a baf-
fled stirred tank were measured over a wide Re range,
from laminar- to turbulent-flow regions. Different corre-
lation results and CFD results were compared with the
experimental data.
Fig. 4 Comparison of power numbers of 4-PBT obtained by The NP values derived from the Nagata correlation
the three methods for the 2-FP, 2-PP, and 4-PBT impellers agreed well with
the experimental values in the laminar- and turbulent-
flow regions, but there was a large deviation in the
transitional-flow region. For the two-stage 4-PBT in a
baffled agitated tank, the Nagata correlation holds
good only in the laminar-flow region.
The NP values determined from the Kamei and
Hiraoka correlation were consistent with the experimen-
tal values in the laminar-, transitional-, and turbulent-
flow regions for all the impellers researched. Therefore,
the Kamei and Hiraoka correlation can be used for accu-
rately predicting NP for paddle or turbine impellers
under baffled conditions.
The NP values obtained from the CFD method
agreed well with the measured values over a wide Re
Fig. 5 Comparison of power numbers of two-stage 4-PBT range. Thus, CFD can be used as a powerful tool not
obtained by the three methods only for analyzing impeller flow fields but also for power
consumption research.

methods correlate well with each other. However, in the Nomenclature

transitional-flow region, the Nagata correlation gives BW  baffle width [m]
higher values than the other methods. The results are b  height of the impeller blade [m]
about 10–30% higher than the experimental values. In C  clearance between the bottom and the impeller [m]
the turbulent-flow region, the simulated values and cor- D  tank diameter [m]
d  impeller diameter [m]
relation values are about 10–15% lower than the experi- f  friction factor [—]
mental values. Therefore, under the present experimental H  liquid height [m]
conditions, the Nagata correlation with the suggested h  distance between blades [m]
blade area method is useful in the laminar-flow region, M  shaft toque [N · m]
but there are large deviations in the transitional- and tur- N  impeller rotational speed [s1]
NP  power number [—]
bulent-flow regions. NP0  power number under unbaffled conditions [—]
2.4 Power number of two-stage 4-PBT NPmax  power number under fully baffled conditions [—]
The NP values of 4-PBT from CFD, the Nagata cor- nb  number of baffle plates [—]
relation, the Kamei and Hiraoka correlation, and experi- np  number of impeller blades [—]
P  power consumption [W]
ments are compared in Figure 5. As with the previous
Re  Reynolds number [—]
results, the three methods give a very good prediction of ReG  modified Reynolds number [—]
NP for the two-stage 4-PBT in the laminar-flow region.
The Kamei and Hiraoka correlation values are about θ  impeller blade angle [°]
15% higher than the experimental values. For the Nagata μ  liquid viscosity [Pa · s]
correlation values and the simulation data, the deviation ρ  liquid density [kg · m3]
is about 5% from the experimental values. When Re  Literature Cited
200, the Nagata correlation can be used for the two- Chen, Z. X., M. H. Xie and G. Z. Zhou; “Power Curves of 24 Types of
stage 4-PBT, but at higher Re values, NP from the Nagata Mixers,” Chem. Eng., 38, 38–43 (2010)

VOL. 44 NO. 11 2011 843

Driss, Z., S. Karray and H. Kchaou; “Computer Simulations of Lami- Ltd., Tokyo, Japan (1975)
nar Mixing within a Pitched-Blade Paddle Vessel,” the Interna- Shekhar, S. M. and S. Jayanti; “CFD Study of Power and Mixing Time
tional Conference on Advances in Mechanical Engineering and for Paddle Mixing in Unbaffled Vessels,” Chem. Eng. Res. Des.,
Mechanics, pp. 1–6, Hammamet, Tunisia (2006) A80, 482–498 (2002)
Hiraoka, S., N. Kamei and Y. Kato; “Power Correlation for Pitched Suzukawa, K., S. Mochizuki and H. Osaka; “Effect of the Attack
Blade Paddle Impeller in Agitated Vessels with and without Baf- Angle on the Roll and Trailing Vortex Structures in an Agitated
fles,” Kagaku Kogaku Ronbunshu, 23, 969–975 (1997) Vessel with a Paddle Impeller,” Chem. Eng. Sci., 61, 2791–2798
Kamei, N., S. Hiraoka and Y. Kato; “Effects of Impeller and Baffle Di- (2006)
mensions on Power Consumption under Turbulent Flow in an Agi- Wang, F. H.; Design Handbook for Practical Subject of Chemical
tated Vessel with Paddle Impeller,” Kagaku Kogaku Ronbunshu, Engineering (Part II), pp. 3–25, Academy Press, Beijing, China
22, 249–256 (1996) (2002)
Nagata, S.; Mixing-Principles and Applications, pp. 24–58, Kodansha

844 JOURNAL OF CHEMICAL ENGINEERING OF JAPAN