International Journal of Heat and Technology

Vol. 39, No. 1, February, 2021, pp. 107-120

Journal homepage: http://iieta.org/journals/ijht

Analysis and Optimization of the Performances of the Centrifugal Compressor Using the
CFD
Brihmat Mostefa1*, Refassi Kaddour1, Douroum Embarek2, Kouadri Amar3
1
Department of Mechanical Engineering, LMSS Laboratory, University of Djillali Liabes, Sidi Bel – Abbes 22000, Algeria
2
Laboratory of Materials and Reactive Systems, Sidi Bel Abbes 22000, Algeria
3
Gas Turbine Joint Research Team, Faculty of Science and Technology, University of Djelfa, Djelfa 17000, Algeria

Corresponding Author Email: mostefa.brihmat@univ-sba.dz

https://doi.org/10.18280/ijht.390111 ABSTRACT

Received: 22 February 2019 Centrifugal compressors have been used in many areas of the machinery. The centrifugal
Accepted: 10 November 2020 compressor design is very complex, and a unique design system needs to be developed. A
centrifugal compressor design system should be easy to use in interface and also flexible
Keywords: for inputs and outputs. The design tool also needs to be able to predicate the compressor
turbomachinery designs, impeller, performance in a fairly accurate level. In this study, we have developed a general analyses
thermodynamics, aerodynamics, centrifugal and optimization approach in the design and performance analysis of centrifugal
compressor, CFD turbomachines. This approach is based on different methods starting from a 1D approach
up to the 3D study of the internal flow. It presents itself as a robust procedure for predicting
and understanding the phenomena associated with the operation of turbomachines, but also
for predicting performance. Current design system includes initial parameter studies,
throughflow calculation, impeller design. The main improvements of the design system are
adding the interface to allow users easy to use, adding the input and output capabilities and
modifying few correlations. Current design system can predict the blade loading and
compressor performance better compared with original design system. To check the
aerodynamic appearance of the centrifugal compressor impeller blades, we must change
the impeller dimensions and focus on changing axial length, but when changing the blade
numbers, the model that improved efficiency and power at the same time introduced a
design with a 0.274% and 10.735% improvement in each respectively in comparison to the
initial impeller at the design point.

1. INTRODUCTION main elements, namely, the rotating part or impeller, and

stationary parts; diffuser and manifold. An important part up
The design of the impeller is very critical for a compressor stream of the impeller is the inducer duct.
stage. The proper blade loading distributions with minimums Air is sucked into the impeller eye and whirled round at high
boundary layer distortions can significantly improve the speed by the vanes on the impeller disc. The normal practice
compressor performance. The estimations of the blade is to design the compressor so that about half the pressure rise
loadings are improved in the current studies to closer to CFD occurs in the impeller and half in the diffuser [12].
results. Many design optimizations and CFD analysis have Large variations in pressure contours are not observed when
been used for improving the impeller aerodynamic designs [1- crossing the boundary between the two major components, but
9]. Impeller aerodynamic design not only affects the impeller noticeable gradients are present at the diffuser outlet because
efficiency but also affects the performance of downstream of the sudden increase in width at the volute inlet. Still heavier
diffuser and volute. It is important for an impeller to achieve a distortions in the pressure are present at the volute tongue
high-efficiency design with relative uniform exit flow. The because of the large incidence, creating a separation like flow
design of centrifugal impeller is subject to multidisciplinary on the suction side of the tongue [13].
and multi-objective optimizations among aerodynamics, Niazi et al. [14] employed a sector-domain (or passage)
structure, and rotor dynamics. between two successive blades) in the impeller for numerical
Turbines produce the power by expanding fluids through calculations for the flow in an impeller of single-stage
drop pressure or head, and temperature. Turbines are widely centrifugal compressor. He generated a grid of 400,000 nodes;
used in aerospace, power generations, automobile, and next he increased the number of nodes till 1.8 million nodes in
aerospace, etc. Successful turbomachine designs are very a mesh sensitivity analysis. Code validation with achieved via
critical. In the concept design stage of the turbomachines, it is comparison with experimental data. He was able to simulate
very important to have a good estimation of the compressor or surge and studied surge control by air-injection. He also
turbine performance from choosing preliminary parameters examined other parameters such as velocity and total pressure.
[10, 11]. Aghaei et al. [15] studied how a good compressor can be
Centrifugal compressors are identified as radial designed and modeled with CFD steady models and to explain
turbomachines. This type of compressor is composed of three reasons for discrepancies between experiment (1D design) and

107
3D CFD analysis. Design/methodology/approach – A model thermal load results in non-uniform tip clearance profile.
with only one impeller channel was used to compare 1D Kim et al. [22]. in his article A series of aero-thermo-
design data, which were obtained from centrifugal compressor mechanical analyses were carried out to predict the running tip
design code, written and developed by the authors. The clearance and the effects of impeller deformation on the
numerical results with respect to performance data showed performance using two different centrifugal compressors
quite good agreement with experimental data at and near the (blade type A and B). In operation, impeller deformation due
operating point of best efficiency. Originality/value – This displacement occurs at the leading edge tip of the impeller
paper offers a combined 1D and 3D numerical approach in blade but maximum stress takes place at the blade root of the
turbo machinery design, especially in radial compressible impeller.
turbo machines design. The design of the impeller is very critical for a compressor
Xu and Muller [16] presented a detailed flow simulation in stage. The proper blade loading distributions with minimums
the volute of a single-stage centrifugal compressor. Their boundary layer distortions can significantly improve the
computationnels domain had 782,000 nodes. The “Frozen- compressor performance. The estimations of the blade
rotor” method is used at interfaces between rotating and loadings are improved in the current studies to closer to CFD
stationary elements. They considered a full-domain flow study results. Many design optimizations and CFD analysis have
and plotted the pressure distribution of the volute. been used for improving the impeller aerodynamic designs
Yutaka et al. [17] carried out a numerical and experimental [23-31]. Impeller aerodynamic design not only affects the
study of the centrifugal-compressor noise affected by the flow impeller efficiency but also affects the performance of
in the tapered diffuser. Experimentally, they presented downstream diffuser and volute. It is important for an impeller
advanced experimental setup. Numerically, 3D steady model to achieve a high-efficiency design with relative uniform exit
was used to simulate flow. The numerical model utilized a full- flow. The design of centrifugal impeller is subject to
domain mesh (single-stage) and a sliding mesh method. multidisciplinary and multi-objective optimizations among
Ling et al. [18] simulated a small centrifugal compressor aerodynamics, structure, and rotor dynamics.
undertaken at the MPG and KJ66 gas turbine design. They The objective of the approach presented in this article is
studied improving the efficiency of the compressor by study aerodynamic performance. For this, we used an
increasing the impeller diameter from 66 mm to 71 mm. They approach based on the reduction of the sources internal to the
found that the performance of the new design is better than the impeller (it is the object of a selection of the aerodynamic
old one within a certain operating range. parameters according to the acoustic computations), and using
An efficient optimization approach to centrifugal impeller the software ANSYS CFX and the method k - ε. For
blades is developed and applied to the design of a 3D impeller performance at the nominal point, we applied the criteria
using; UDM, CFD, regression analysis method, and genetic suggested by the literature, the main differences between the
algorithm. Global optimization of the approximate function is original impeller and the optimized one being the value of the
obtained by genetic algorithms. Taking maximum isentropic slowing down. To show the validity of the correlations and
efficiency as objective function, this optimization approach models proposed in this study the modifications have the main
has been applied to the optimum design of a certain centrifugal modified parameters correspond to the impeller: the axial
compressor blades [19]. The results, compared with those of length, the number of blades.
the original one, show that isentropic efficiency of the To optimize the impeller of this compressor and reduce
optimized impeller has been improved which indicates the turbulence losses, we have changed some parameters of the
effectiveness of the proposed optimization approach such as original turbocharger impeller. But we finally found that when
CFD, structural dynamics, and thermal analysis combined you change the numbers of the blades, you get good results for
with the emergence of improved optimization algorithms the performance of the central compressor.
makes it possible to perform blade shape automatic
optimization process.
Tang et al. [20] studied numerically a 3D impeller and 2. METHODOLOGY
vaneless diffuser of a small centrifugal compressor. The
influence of impeller tip clearance on the flow field of the 2.1 Description of the centrifugal compressor MAN-Turbo
impeller was investigated. Then, a new partially shrouded
impeller was designed. They used a sector-domain and The turbocharger man -turbo type RG028/02-l2 (Figure 1)
110,000 grid points. Numerical results show that the consists of a single helical gear wheel and compressor stages.
secondary flow region becomes smaller at the exit of the A large gear wheel drives a gear shaft. On the end faces of the
impeller. Better performance is achieved in comparison with pinion shaft are screwed in the axial direction, the moving
the un-shrouded impeller. wheels, with a Hirth lock. The axial forces of the pinion shaft
Centrifugal compressors are used in many heat pumps and are transmitted to the large gear by thrust collars. The thrust
refrigeration systems. Radial vaneless diffuser is a principal bearing of the shaft of the large gear wheel absorbs the axial
component in these compressors. Therefore, Shaaban [21] in force resulting therefrom. The compressor ratio R1S / R2 =
his article aims at improving the centrifugal compressor 0.53.
performance by optimizing the design of the radial vaneless • Mobile impeller
diffuser. Two radial vaneless diffuser geometries were Each compressor stage consists of a cantilevered mobile
proposed, investigated and numerically optimized. The impeller, surrounded by a volute casing. The housings are
optimization aimed at minimizing the diffuser loss coefficient supported by vertical flanges arranged on the speed adapter. In
and maximizing the pressure coefficient. Simulations were the housings are housed the blade supports and the input insert.
performed by solving the Reynolds averaged Navier–Stokes The movable impelles of stages 1, 2 are in closed
equation under 2D axisymmetric condition. To the construction. Each of these impelles is separated from its
combination of centrifugal force, aerodynamic pressure and volute casing by axial and radial seals. impeller linings are

108
 designed as non-contact labyrinth liners. They seal between • Generation of Geometry and CAD
the inlet and outlet sides of each stage of the compressor. The Figure 3 shows a 3D view of the original man-turbo
operating conditions of the centrifugal compressor studied in compressor impeller obtained from the ANSYS Workbench
this specification are presented in Table 1. software "CFX-BladeGEN".
Table 1. Compressor operating conditions Man-Turbo

Amount Specified value

total pressure Pt1 = 16 bar
total temperature Tt 1= 68.4℃
flow of mass m˙ = 0.28 kg/s
rotation speed N = 21592 tr/min
operating fluid Gaz
polytropic efficiency η=79.39

Figure 3. 3D visualization of the original impeller man turbo

CFX-BladeGEN

3. NUMERICAL ANALYSIS

The numerical simulation has proven to be well suited and

efficient for accomplishing the task of dimensioning and
optimizing a centrifugal wheel. However, the development of
computer resources and digital techniques have allowed the
emergence of CFD codes capable of analyzing the three-
dimensional structure of internal flows. In this work, two
objectives will be followed using CFD software. That is to say:
their use represents, at the same time, the finest optimization
but also the final step of the design process and the validation
of the previous steps. Presented CFD codes are commercial
software: CFX-BladeGEN, CFX-TurboGrid. Automatic or
operator-controlled links allow dialogue between CFX-Pre,
Figure 1. Sectional view of the speed adapter compressor CFX-Solver and CFX-Post.

Figure 2 shows the original impeller is given from 3.1 Governing equations
experimental executions.
Man-turbo centrifugal compressor, impeller geometry and • Continuity equation:
experimental design data. The main dimensions of the original
wheel are summarized in Table 2.  
+ ( Vi ) = 0 (1)
t xi

• 3D Navies-stokes equation:

 ( Vi )  p  ij
t
+
x j
( i j)=−
VV +
xi x j
(2)

where, p is the static pressure, and is the viscous stress tensor

given by

 Vi V j 2 Vl 
 ij =  + +  (3)
Figure 2. Original impeller geometry  x j xi 3 xl 

Table 2. Main dimensions of the original impeller • Energy equation:

Geometric data of the original impeller Dimensions  (E)    t 
outlet diameter D2 = 275 mm + Vi (  E + p ) =  K −  h j J i + (Vi ij )  h (4)
blade height at outlet b2 = 10 mm t xi xi  xi j 
number of blades ZPR = 18
metal angle at exit βp2 = 60º • Turbulence model Equations:
input radius for foot R1h R1h = 43 mm
1-Equation of turbulence energy (k):
input radius for blade head R1s R1S =73mm

109
     t  k  The optimum must be calculated by assuming the
(  k ) + (  kui ) =   +   − E (5) meridional velocity ratio (Rvm = C1s /0.5 (C1h + C1s)). All
t xi x j   k  x j  calculations made this optimum is expressed:

2-Equation of rate of dissipation of turbulence kinetic 2/3

energy (E): R m
R1S optimum = R1h 2
+ 2  vm  (8)
 1 
      E  E E2
(  E ) + (  Eui ) =   + t   + C1E C3E − C2 E (6)
t xi x j   E  x j  k k
The value obtained depends strongly on the meridian speed
ratio and the radius at the hub. Placing of the blades in the foot
The model constants 𝐶1E, 𝐶2E, 𝐶𝜇, 𝜎𝑘, and 𝜎𝜖 have the and at the head as a first approximation, it is possible to
following default values: 𝐶1E= 1.44, 𝐶2E = 1.92, 𝐶𝜇 = 0.09, evaluate the relative angle of the inlet flow β1 once R1h and R1S
𝜎𝑘 = 1.0 and 𝜎E = 1.3. fixed. Assuming the one-dimensional flow, we can show that:

• Methods Predimensioning 1D analysis U1S R1S

1S = tan −1 = tan −1
Pre-dimensioning is defined as the step that sets the main
characteristic dimensions of the machine at the stations
C1S (
m / 1 R 21S − R 21h ) (9)

recalled in Figure 4.
For the flow at the head, the choice of blade pitch can then
Impeller: spokes at the entry for the foot and the blade head:
be initiated by using a criterion for the incidence I (defined as
R1h and R1S, exit radius R2 and output blade height: b2. Wedge
the difference between the relative angle of the incident flow
angles of the blades βp1s, βp1h. Axial length Lax, the value of
and the pitch of the blade). Number of blades of the impeller
the angle βp2, is optionally refined during this step. This step
resulting from an optimization between friction and sliding is
relies mainly on two approaches: (I) the use of empirical
given by the relation:
criteria and (II) the use of 1D-prediction tools.
ZPR = 25*cos (β2)/nsq (10)

• Axial length
The axial length Lax is a parameter that significantly
influences the aerodynamic performance, but also: (I) the size
of the stage, (II) the dynamics of the shaft, (III) the mechanical
stresses in the impeller. From the point of view of
aerodynamics.

Figure 4. Schematic view of a centrifugal compressor stage

• impeller
Radius at the head the pre-dimensioning of the impeller
begins with the establishment of the radius at the input casing,
under the constraint given for R1h.
Japikse [32] recommend minimizing the number of mach in
mind the theoretical justification is to minimize friction losses,
related at the speed level.
Note that this is also beneficial for acoustics. For a feed
without pre-rotation, the relative velocity of the fluid at the top
of the blade is written:
Figure 5. Scheme for conducting impeller analysis by CFD

W1S = U1S 2 + C1S 2 (7) The axial length is given by the following relationship:

Once the overall characteristics of if the radius to the casing   R1S − R1h
is reduced, the peripheral speed of the blade (U1S = R1S. Ω)
Lax
(
= 2 K M *1S + K 2 ) 1 − R 2−R R
1S 1h
 (11)
decreases. However, as the debiting section decreases the R2  2  R2
retention of the flow rate leads to an increase in C1S. Thus,
W1S is influenced by two contradictory effects: An optimum where, K1 = 0.28 and K2 = 0.8. Thus, for typical values of
of R1S exists to minimize the number of relative mach inputs. R1h, R1S and R2, the ratio Lax / R2 varies between 0.62 and

110
0.77 when the number of mach varies between 0.6 and 0.9. The second level is a further study for the study of internal
The number of mach of the input and the exit of the impeller. flow. It begins with the creation of a structured mesh using the
CFX-TurboGrid software (Figure 7).
W1 W2
M *1 = ;M 2=
*
(12)
 RT1  RT2 5. RESULTS OF THE SIMULATION WITH CFX AND
RNG MODEL K.Ε.
The objective of this study:
-Check the execution impact due to the change of some Table 3 shows some execution parameters.
geometric parameters.
-Check the execution impact due to the change in the Table 3. Compressor performance results. Original impeller
number of blades.
Figure 5 shows the different steps taken during the three- input power (w) 407336.0000
dimensional simulation of the turbo centrifugal compressor intake flow coefficient 0.0123
impeller. outlet flow coefficient 0.0898
total pressure ratio 1.9948
total temperature ratio 1.2354
4. DOMAIN OF COMPUTATION AND GENERATION isentropic efficiency 92.7412
OF THE MESH polytropic efficiency 93.4136

The diversity of mesh generation methods can result in We consider here the case of meshes generated with
many ways of characterizing a computing grid. We consider ANSYS Workbench software "TurboGrid". A first choice is
here the case of meshes generated with ANSYS Workbench the set of topology H / J / C / L-Grid. The mesh data are as
software "TurboGrid". A first choice is the set of topology H / follows:
J / C / L-Grid, for the study of turbomachines the axisymmetric
flow hypothesis allows the simplification of the single-channel Table 4. Mesh data. Topology H / J / C / L-Grid
calculation domain interaubages. The different boundaries for
modeling a channel of the centrifugal compressor impeller are Characteristics Total nodes Total element
mesh 33143 28948
shown schematically in Figure 6.
Using ANSYS CFX -post software, the most important
results for the simulation of the original man turbo geometry
have been deduced. Table 4 shows the compilation of some
execution parameters.

5.1 Figures produced for the CFX

The CFX is a software that allows quick and easy

visualization of simulation results, gives various graphs of
various parameters in different plans. In this section they will
be shown to the graphs obtained in the simulation of the man-
turbo compressor.
The gas, an irrotational fluid, enters the impeller, tends to
create a vortex of the opposite direction of rotation (Figure 8).
This takes a region of little pressure on one side of the wheel's
blade, known as the suction side; while on the other it is of
Figure 6. Mesh of the original impeller. Blade to blade greater pressure, known as pressure side. This effect at the
leading edge and edge of fuit in Figures 9 (a, b). The Z-axis
coincides with the axis of the wheel and its direction is in the
direction of the entry of the wheel. As in modeling the
direction of rotation, it is negative in the Z axis, the edge of the
attack means that the rotation goes from right to left. Due to
the format of the impeller, the direction of rotation in the edge
of the leak is opposite, that is to say, it goes from left to right.
It can be visualized clearly in Figures 9 (a, b) and the side
of the blade aspiration, where the value of pressure is less
throughout the leading edge and edge of leak, and in the dawn
opposition side the value of the pressure is greater in all this
region.
A point can be seen in Figure 9 b where the pressure is
greater. Possible explanation for this is the formation of the
mats in the outlet of the impeller making the fluid to return to
Figure 7. Centrifugal impeller; Structured mesh the impeller, being created a region of stagnation raising the
pressure.

111
 Figure 8. Direction of rotation z

a) Pressure in the leading edge b) Pressure in the edge of leaks

Figure 9. Pressure in the leading edge and the trailing edge

Table 5. RNG results k -ε original impeller parameter

Parameter Entrance Leading edge edge of leaks Exit

P static (pa) 1677490.0000 1675170.000 2485910.000 2593270.000
P total (pa) 1688770.0000 1673950.000 3402740.000 3368670.000
P t (rot) (pa) 1676890.0000 1664860.000 1601420.000 1596370.000
T static (K) 339.9550 339.4260 386.9860 391.9690
T total (K) 341.8260 341.7370 423.8920 422.9850
Tt (rot) (K) 341.1140 341.0750 341.1350 341.1240
Enthalpy (J/kg) 43153.4000 43114.3000 43174.2000 43163.4000
Entropy (J/kg) -670.4250 -668.4220 -657.1680 -656.3220
Mach abs 0.1644 0.1906 0.6838 0.6228
Mach rel 0.3431 0.3320 0.1554 0.2424
u (m/s) 134.5690 134.3160 310.3630 333.5010
Cm(m/s) 52.3589 57.7330 27.8787 26.3787
Cu (m/s) -46.7831 -73.1338 -271.8680 -243.0850
C (m/s) 84.7199 105.7770 274.0240 244.9330
Wu (m/s) 87.7860 61.1842 38.4967 90.4162
W (m/s) 103.6210 88.5834 48.6757 95.3097
Flow Angle: Alpha 75.5287 -60.4172 77.4171 39.2016
Flow Angle: Beta -75.1775 -94.4162 -57.0229 -74.9403

Table 5 shows RNG results k –ε original man turbo wheel where vortexes and mats occur, leading to the reduction of the
parameter. gas bill area so that the speed next to the wall is almost zero.
CFX produces vectors of this velocity for 20, 50 and 80% It can be clearly visualized the separation mat, explaining the
of the normal blade size, shown in Figures 10 to 12. These high values of mach in the output of the impeller.
figures allow to remove some conclusions, namely: These last three figures clearly explain the angle variation
• The speed reaches maximum values in the upper blade part along the edge of the leak.
of the impeller, next to the crankcase. The losses inside the wheel can be visualized by the graphs
• A stagnation region exists, already detected before the and curves of the entropy produced by the CFX. Entropy is a
results. The pressure in the region next to the wheel output, in thermodynamic property in which it is necessary to make an
which the speed reaches minimum values. This effect is analysis with respect to a value of the reference to know where
greater in the upper blade part, near the crankcase. its greatest variation occurs. This analysis helps to avoid where
• As commented in all this work, it is exactly in this region the biggest losses occur.

112
 Figure 13 shows the average variation of entropy
throughout the wheel. As it was so expected, entropy increases
throughout the impeller, a time that true compression is not
isentropic. Integrating this result in all the impeller, returning
each point the value of the temperature and the mass, the CFX
calculates the loss of overall efficiency. We can observe an
increase in the entropy of the whole impeller. Represented on
velocity figures. It is obvious that it is obtained at the CFX
extract is the profile of the entropy by the whole impeller in
positions of 20, 50 and 80% of the normal blade size. The
figures produced will be shown to all to expose in detail where
the greatest losses occur (Figures 14 to 16).
In Figure 14 a sample present in a region with a large loss
Figure 10. Velocity at span of 20℅, blade to blade of impeller output, possibly due to the formation of the mat
and also the sliding of the blade, the entropy variation in all
the walls of the impeller is due to a phenomenon turbulence.
Figure 15 sample that the variation of the entropy was more
than what is in the previous figure. This means that the losses
for the vortex and the mat if concentrated more in the
crankcase area after the exit to the impeller.
Figure 16 shows the profile of the entropy in the impeller
exit.
Figure 17 shows that the largest losses occur in the side of
the blade pressure of the impeller and also in the next region
to the crankcase, which corresponds to a normal blade size of
80%. These results are logical with the figures of the pressure,
the number of absolute mach and the speed.
These are the results of the original impeller.
To optimize the impeller of this compressor and reduce
Figure 11. Velocity at span of 50℅, blade to blade turbulence losses, we have changed some parameters of the
original man turbo impeller.

Figure 12. Velocity at the span of 80℅, blade to blade

Figure 14. Entropy 20℅ of span. blade to blade

Figure 13. Entropy throughout the original impeller Figure 15. Entropy 50℅ of span. blade to blade

113
 5.3 First optimization and parametric studies CFD

In this article the results to change the value of the axial

length of the impeller will be displayed increased in more than
10 mm (Lax +), although to have if kept without changing, all
the geometric parameters, had the change of the mesh because
the format of the impeller has changed.
The data of the mesh are the following Table 6:

Table 6. Mesh data, impeller (Lax +)

Characteristics Total nodes Total element

mesh 48288 43060

Table 7 shows some execution parameters.

Figure 16. Entropy 80℅ of span. blade to blade
Table 7. Geometric parameters, impeller (Lax +)
input power (w) 410718.0000
intake flow coefficient 0.0123
outlet flow coefficient 0.0913
total pressure ratio 2.0041
total temperature ratio 1.2374
isentropic efficiency 92.6353
polytropic efficiency 93.3213

The change in geometry of the shape to be increased in the

distance between the inlet and the outlet of the impeller in the
meridian plane means that the bending of the impeller has
been decreased.
It is so expected that this light results in increased friction
losses and reduced losses due to turbulence phenomena mainly
combines the formation of mats and internal recycling.
Figure 17. Entropy at the exit of the original impeller Because the distance between the entrance and the exit will be
greater, the format of blade theoretically, will be able to adapt
5.2 Parametric study and optimization of the impeller better to the turbulent of the fluid.
Analyzing the two results it is verified that the modification
To show the validity of the correlations and models of the geometry provided as a decrease in efficiency, that was
proposed in the study modifications have been made to the for 92.63%. But the pressure reason is increased for 2.00, as
geometry of the compressor impeller. The main modified the pressure reason is used for the calculation of efficiency,
parameters correspond to the impeller: the axial length, the temperature and power. Figure 18 shows the variation of the
number of blades. average entropy throughout the impeller.
In Table 8 the mean values of some parameters in different
positions of the domain are shown.

Table 8. Modified impeller parameter (Lax +)

Parameter Entrance Leading edge edge of leaks Exit

Pstatic (pa) 1675030.000 1673870.000 2482330.000 2590750.000
Ptotal (pa) 1686780.000 1671170.000 3414630.000 3380430.000
Ptotal (rot) (pa) 1675010.000 1661780.000 1597570.000 1592950.000
Tstatic (k) 339.9550 339.4260 386.9860 391.9690
Ttotal (k) 341.8260 341.7370 423.8920 422.9850
Ttotal (rot) (k) 341.1180 341.0820 341.1530 341.1320
Enthalpy (J/kg) 43157.1000 43121.2000 43192.4000 43170.8000
Entropy (J/kg) -670.0810 -667.8520 -656.4290 -655.6880
Mach abs 0.1657 0.1993 0.6897 0.6283
Mach rel 0.3430 0.3336 0.1522 0.2386
u (m/s) 134.4650 134.3950 310.3170 333.5010
Cm(m/s) 53.1702 58.9625 28.3206 26.8398
Cu (m/s) -47.2546 -75.9532 -273.2510 -244.5820
C (m/s) 85.4879 108.1840 275.4330 246.4840
Wu (m/s) 87.2106 58.4433 37.0672 88.9182
W (m/s) 103.8120 87.5921 47.9239 94.0918
Flow Angle: Alpha -76.2488 -60.4858 61.1413 29.9624
Flow Angle: Beta -75.6264 -95.1815 -56.4439 -74.7556

114
 Figure 22. Chain Lines Speed. of impeller Lax +

Figure 18. Entropy throughout the modified Lax + impeller

Figure 23. Entropy 80, blade to blade

Comparing Figure 13 to Figure 18 we observe that the

Figure 19. Velocity at span of 20℅, blade to blade variation of entropy would be less than that in the original
geometry man turbo. Clearly that this analysis is only
qualitative, a period that it is necessary to determine the whole
value of the losses to integrate in all the impeller the reason
writes the variation of the entropy and the temperature. But in
the original geometry, the entropy leaves a value of -669.25 J
/ kg. K and goes from stops -656.25 J / kg. K. For the case with
the modified Lax + geometry that the whole variation of the
entropy is clearly less, therefore goes from -668.75 J / kg. K to
-655.75 / kg. K.
Comparing Table 8 and Table 5, we observe that the
increased value of absolute machs of 0.6838 in the edge of
leaks for 0.6897. Soon, the hypothesis that losses due to
turbulence decrease is verified with these results. The evidence
of this can be seen in Figures 19 to 23.
Figure 20. Velocity at span of 50℅, blade to blade Figures 19 to 23 that the effect of the mat separation in the
losses of the impeller is less than that in the original geometry.
He can observe that the value of the velocity is lesse for the
value in the original case (Figure 21). As it was so expected,
there is a variation of entropy in the region where the mats are
formed. This was not clearly seen in Figure 23 of the Lax +
geometry. Because a small variation of the entropy in the
original case.
If the result shows a decrease in efficiency concludes that
the reduction in losses due to turbulence phenomena exceeds
the increase in friction losses, although not having to get the
reason for the expected pressure. Another explanation for
decreased efficiency is the increase in the number of mach, the
region that this leads to a decrease in efficiency.
Table 9 shows the result of Lax + optimized geometry,
compared to the original geometry.
Figure 21. Velocity at the span of 80℅, blade to blade

115
 Table 9. Original impeller settings and modified impeller. In Table 12, the average values of some parameters in
(Lax +) different positions of the domain are shown.
Table 13 shows comparison of modified geometry Lax -
parameters
original geometry lax
percentage
results with original geometry.
geometry +
RP 1.9948 2.0041 0.466 % Expected Result:
RT 1.2354 1.2374 0.161% • Reduced friction.
η is (%) 92.7412 92.6353 -0.114 % • Increased turbulence.
η pol (%) 93.4136 93.3213 -0.098 %
Power (w) 407336.0000 410718.0000 0.830 %
Table 11. Geometric parameters, impeller (Lax -)
The data of the mesh are the following Table 10: input power (w) 404426.0000
intake flow coefficient 0.0123
Table 10. Mesh data , impeller (Lax -) outlet flow coefficient 0.0909
total pressure ratio 1.9917
Characteristics Total nodes Total element total temperature ratio 1.2337
mesh 44848 39956 isentropic efficiency 93.2141
polytropic efficiency 93.8421
Table 11 shows some execution parameters.

Table 12. Modified impeller parameter (Lax -)

Parameter Entrance Leading edge edge of leaks Exit

Pstatic (pa) 1675930.000 1675480.000 2490080.000 2594920.000
Ptotal (pa) 1687200.000 1670620.000 3394890.000 3360380.000
Ptotal (rot) (pa) 1675510.000 1661760.000 1607470.0000 1601360.000
Tstatic (k) 339.9490 339.6430 386.7000 391.5810
Ttotal (k) 341.8230 341.7180 422.5150 421.6930
T total (rot) (k) 341.1210 341.0990 341.2120 341.1850
Enthalpy (J/kg) 43159.8000 43138.1000 43251.8000 43224.4000
Entropy (J/kg) -670.1630 -667.8090 -658.0100 -657.0340
Mach abs 0.1657 0.1907 0.6792 0.6190
Mach rel 0.3445 0.3319 0.1622 0.2495
u (m/s) 134.7340 134.4320 310.4010 333.5010
Cm(m/s) 53.3209 56.3735 28.2137 26.8281
Cu (m/s) -47.6795 -72.4668 -270.9310 -239.7550
C (m/s) 86.1794 104.1970 273.2000 241.7420
Wu (m/s) 87.0550 61.9670 39.4713 93.7461
W (m/s) 103.9120 87.3359 49.6590 98.8533
Flow Angle: Alpha 75.4021 -61.6495 77.1147 26.9968
Flow Angle: Beta 75.3416 -92.4634 -57.2096 -75.4411

5.4 Second optimization and parametric studies CFD The distance between the entry and the exit is less, the
format of blade theoretically, will be able to increase the
Change the value of the axial length of the turbo man turbulence of the fluid.
impeller will be displayed in less than 10 mm Lax-. Although Comparing Table 12 and Table 5 it is observed that the
to have if kept without changing the geometrical parameter, value of absolute mach decreases by 0.6838 in the edge of leak
had the change of the mesh because the format of the impeller for 0.6792. One possible explanation for the reduction in mach
has changed. number that occurs, an increase in turbulence losses is that the
The reason for pressure is decreasing for 1.9917, the transfer of rotor energy to the fluid is less because of little
pressure reason is used for the calculation of efficiency, passage through the entire impeller. The evidence of this can
temperature and power. We verify that these results are much be seen in Figures 25 to 28 a moment that velocitys are in this
better than the previous case analyzed. case also more.
Figure 24 shows the variation of the entropy throughout the
impeller. Comparing Figure 24 to Figure 13 we observe that Table 13. Original impeller settings and modified impeller.
the variation of entropy in this case would be more than what (Lax -)
in the original geometry. It is clear that this analysis is only
qualitative, a period that it is necessary to determine the full parameters
original geometry
percentage
value of losses to integrate throughout the impeller. But in the geometry lax-
original geometry the entropy leaves a value of -669.25J / kg. RP 1.9948 1.9917 -0.155%
K and it goes up to -656.25 J / kg. K. For the case with the RT 1.2354 1.2337 -0.137%
modified geometry Lax- that all the variation of the entropy is η is (%) 92.7412 93.2141 0.509%
clearly more, goes thus of -669 J / kg. K up to -657 J / kg. K. η pol (%) 93.4136 93.8421 0.458%
Power (w) 407336.0000 404426.0000 -0.714%

116
 Figure 28. Entropy 80℅ span. blade to blade

Figures 25 and 28 the effect of the mat in the losses of the

Figure 24. Entropy throughout the modified Lax- impeller
impeller is more than what in the original geometry. He can
observe that the maximum value of the Velocity is increased
from 160 to 170 m / s. in the previous case, this velocity has
reached a maximum value of 170 m / s. a more energy transfer
occurs for the fluid, which it does with that to the influence of
the losses for turbulence is more.
Figure 28 shows the entropy in the size of 80% span.
Expected Result:
• Reduced friction.
• Increased turbulence.

5.5 Final optimization of the impeller

5.5.1 Description of the final geometry

The same original impeller will be displayed but with 20
Figure 25. Velocity at span of 20℅, blade to blade blades, although to have if kept all the geometrical parameters,
had the change of the mesh because the number of blades has
increased.
The data of the mesh (Table 14) are the following:

Table 14. Mesh data, impeller (20 blades)

Characteristics Total nodes Total element

mesh 34797 30388

Table 15 shows some execution parameters.

Table 15. Geometric parameters, impeller (20 blades)

input power (w) 451067.0000

intake flow coefficient 0.0138
Figure 26. Velocity at span of 50℅, blade to blade outlet flow coefficient 0.0998
total pressure ratio 1.9926
total temperature ratio 1.2344
isentropic efficiency 92.9961
polytropic efficiency 93.6449

The change in geometry increases the number of blades

probably takes an increase in friction losses and greater losses
for the gas incident. Therefore, it is expected that this slight
change in friction loss and loss reduction increases due to the
formation of mats and the internal recycling and reduction of
the slip of the impeller output. If the result shows an increase
in efficiency of 92.74% in the original geometry for 92.99%,
concludes that the reduction of losses due to turbulence
phenomena and slip surpasses the increase in friction losses. It
can be seen that the pressure reason is decreasing for 1.9926.
Figure 27. Velocity at the span of 80℅, blade to blade Figure 29 shows the variation of the average entropy

117
throughout the final impeller. original geometry and all the simulations. The sufficiently
Comparing the value of the absolute mach of the base of interesting data to be observed are the average value of
case with this one, one observes a reduction of 0.6838 of the absolute speed in the output of the impeller given in the form
edge of fuit for 0.6827, with this fact one proves the reduction of Tables 5 and 16, in the original impeller the velocity is
of the losses for the turbulence, thus the reduction of the 244.93m / s, whereas in this case it is 243.54 m / s. therefore
velocity of the exit impeller exists. This always reinforces the the absolute speed is decreased by comparing to that of the
hypothesis of the main cause of improvement in efficiency is original geometry.
because the compressor no longer meets in the clutter. We can Differently from the cases of change of impeller geometry,
see from Figures 30 to 33 the evidence of this is that the in which the fall of the velocity more was accentuated, in this
velocitys are in this case less. case the fall was probably less being this the explanation of
Figures 30 To 33 that the effect of the mat in the impeller that increases of the efficiency.
losses is less than what in the original geometry. It can be seen In Figure 33 also it is possible to visualize a variation of the
that the maximum value of the velocity has fallen from 130 to entropy in the region where the mats are formed. If the result
120 m / s in 80% of the blade size. shows an increase in efficiency concludes that the reduction of
In Table 16, the average values of some parameters in losses for turbulence and slip surpasses the increase in friction
different positions of the domain are shown loss. It is observed that, the number of mach has come close to
Differently from the cases of change of impeller geometry, 0.72. apparently being a good result comparison with the
in which the fall of the velocity more was accentuated, in this original geometry and all the simulations. The sufficiently
case the fall was probably less being this the explanation of interesting data to be observed are the average value of
that increases of the efficiency. absolute speed in the output of the impeller given in the form
In Figure 33 also it is possible to visualize a variation of the of Tables 4 and 14, in the original impeller the velocity is
entropy in the region where the mats are formed. If the result 244.93m / s, whereas in this case it is 243.54 m / s. therefore
shows an increase in efficiency concludes that the reduction of the absolute speed is decreased by comparing to that of the
losses for turbulence and slip surpasses the increase in friction original geometry.
loss. It is observed that, the number of mach has come close to Table 17 shows the comparison of results with original
0.72. apparently being a good result comparison with the geometry.

Table 16. Modified impeller parameter (20 blades)

Parameter Entrance Leading edge edge of leaks Exit

Pstatic (pa) 1663710.000 1665490.000 2477020.000 2584510.000
Ptotal (pa) 1683560.000 1673100.000 3388360.000 3354640.000
Ptotal (rot) (pa) 1670950.000 1662780.000 1599870.000 1594230.000
Tstatic (k) 339.5320 339.1900 386.5860 391.6020
Ttotal (k) 341.8300 341.7000 422.7470 421.9460
T total (rot) (k) 341.0670 341.0270 341.1240 341.1240
Enthalpy (J/kg) 43106.2000 43065.6000 43162.6000 43163.1000
Entropy (J/kg) -669.5160 -668.1970 -656.9120 -655.9330
Mach abs 0.1837 0.2016 0.6827 0.6213
Mach rel 0.3515 0.3439 0.1642 0.2472
u (m/s) 134.5440 134.3140 310.3640 333.5010
Cm(m/s) 58.7362 58.3039 30.9744 28.4071
Cu (m/s) 48.2326 -69.5816 -271.4370 -241.3880
C (m/s) 91.0018 104.8130 274.0860 243.5460
Wu (m/s) 86.3112 64.7341 38.9288 92.1133
W (m/s) 106.2320 90.6905 50.9149 97.7586
Flow Angle: Alpha -76.7953 -64.2838 69.6999 40.4411
Flow Angle: Beta -76.1466 -91.8430 -54.7798 -73.4779

Figure 30. Velocity at span of 20℅, blade to blade

Figure 29. The entropy throughout the final impeller

118
 For the nominal performance, we applied the criteria
suggested by the literature, the main differences between the
original impeller and the optimized one being the value of the
slowdown. To check the aerodynamic profile of the centrifugal
compressor impeller blades, we must change the impeller
dimensions and focus on changing the axial length, but by
changing the numbers of the blades, we obtain good results.

REFERENCES

[1] Aungier, R.H. (2000). Centrifugal compressors: A

strategy for aerodynamics design and analysis. ASME,
Figure 31. Velocity at span of 50℅, blade to blade New York, USA. https://doi.org/10.1115/1.800938
[2] Xu, C., Amano, R.S. (2008). Computational analysis of
swept compressor rotor blades. International Journal for
Computational Methods in Engineering Science and
Mechanics, 9(6): 374-382.
https://doi.org/10.1080/15502280802365840
[3] Xu, C., Amano, R.S. (2010). Computational analysis of
scroll tongue shape to compressor performance by using
different turbulence models. International Journal for
Computational Methods in Engineering Science and
Mechanics, 11(2): 85-99.
https://doi.org/10.1080/15502280903563517
[4] Xu, C., Yang, H.Q., Jiang, Y.D., Yi, Z.W. (2019). The
development of an integrally geared centrifugal
compressor. International Journal of Fluid Mechanics &
Figure 32. Velocity at the span of 80℅, blade to blade Thermal Sciences, 5(1): 1-9.
https://doi.org/10.11648/j.ijfmts.20190501.11
[5] Xu, C. (2007). Design experience and considerations for
centrifugal compressor development. Journal of
Aerospace Engineering, 221(2): 273-287.
https://doi.org/10.1243/09544100JAERO103
[6] Xu, C., Amano, R.S. (2012). Empirical design
considerations for industrial centrifugal compressors.
International Journal of Rotating Machinery, 2012.
https://doi.org/10.1155/2012/184061
[7] Xu, C., Amano, R.S. (2009). Development of a low flow
coefficient single stage centrifugal compressor.
International Journal for Computational Methods in
Engineering Science and Mechanics, 10(4): 282-289.
https://doi.org/10.1080/15502280902939460
[8] Xu, C., Amano, R.S. (2009). The development of a
Figure 33. Entropy 80℅ span, blade to blade
centrifugal compressor impeller. International Journal
for Computational Methods in Engineering Science and
Table 17. Original impeller settings and modified impeller.
Mechanics, 10(4): 290-301.
(20 blades)
https://doi.org/10.1080/15502280903023165
[9] Xu, C., Amano, R.S. (2010). Study of the flow in a
original Geometry
parameters
geometry 20blades
percentage centrifugal compressor. International Journal of Fluid
RP 1.9948 1.9926 -0.110 % Machinery and System, 3(3): 260-270.
RT 1.2354 1.2344 -0.080 % https://doi.org/10.5293/IJFMS.2010.3.3.260
η is (%) 92.7412 92.9961 0.274 % [10] Turner, M.G., Merchant, A., Bruna, D. (2006). A
η pol (%) 93.4136 93.6449 0.247 % turbomachinery design tool for teaching concepts for
Power (w) 407336.0000 451067.0000 10.735% axial-flow fans, compressor, and turbines. ASME Turbo
Expo 2006: Power for Land, Sea, 1: 937-952.
https://doi.org/10.1115/GT2006-90105
6. CONCLUSIONS [11] Xu, C., Amano, R.S. (2009). On the development of
turbomachine blade aerodynamic design system.
The main limitation of the approach presented in this article International Journal for Computational Methods in
is aerodynamic performance. For this, we used an approach Engineering Science and Mech, 10(3): 186-196.
based on the reduction of the sources internal to the impeller https://doi.org/10.1080/15502280902795052
(it is the object of a selection of the aerodynamic parameters [12] Cohen, H., Rogers, G.F.C., Saravanamutto, H.I.H. (1996)
according to the acoustic computations), and using the Gas Turbine Theory. Pearson, New York.
software ANSYS CFX and the method k - ε. [13] Rangwala, A.S. (2005). Turbo-Machinery Dynamics.

119
 Design and Operation, McGraw-Hill, USA. model of compressor stability. ASME Turbo Expo 2006,
[14] Niazi, S., Stein, A., Sankar, L.N. (2000). Numerical 6: 1157-1168. https://doi.org/10.1115/GT2006-90040
studies of stall and surge alleviation in a high-speed [29] Denton, J.D., Dawes, W.N. (1998). Computational fluid
transonic fan rotor. AIAA, pp. 2000-0225. dynamics for turbomachinery design. Journal of
https://doi.org/10.2514/6.2000-225 Mechanical Engineering Science, 213(2): 107-124.
[15] Aghaei Tog, R., Mesgharpoor Tousi, A., Soltani, M. https://doi.org/10.1243/0954406991522211
(2007). Design and CFD analysis of centrifugal [30] Dickmann, H.P., Wimmel, T.S., Szwedowicz, J.,
compressor for a microgasturbine. Aircraft Engineering Filsinger, D., Roduner, C.H. (2006). Unsteady flow in a
and Aerospace Technology, 79(2): 137-143. turbocharger centrifugal compressor, three-dimensional
https://doi.org/10.1108/00022660710732680 computational fluid dynamics simulation and numerical
[16] Xu, C., Muller, M. (2005). Development and design of a and experimental analysis of impeller blade vibration.
centrifugal compressor volute. Int. J. Rotating Machinery, Journal of Turbomachinery, 128(3): 455-465.
3: 190-196. https://doi.org/10.1155/IJRM.2005.190 https://doi.org/10.1115/1.2183317
[17] Yutaka, O., Takashi, G., Eisuke, O. Effect of tapered [31] Xu, C., Amano, R.S. (2012). Aerodynamic and structure
diffuser vane on the flow field and noise of a centrifugal considerations in centrifugal compressor design-blade
compressor. Journal of Thermal Science, 16(4): 301-308. lean effects. Turbo Expo, pp. 557-568.
https://doi.org/10.1007/s11630-007-0301-1 https://doi.org/10.1115/GT2012-68207
[18] Ling, J., Wong, K.C., Armfeild, S. (2007). Numerical [32] Japikse, D. (1996). Centrifugal Compressor Design.
investigation of a small gas turbine compressor. 16th Concepts, ETI Inc. Norwich, Vermont, 3rd edition.
Australian Fluid Mechanics Conference, Crown Plaza,
Gold Coast, Australia, 2-7.
[19] Shu, X.W., Gu, C.G., Xiao, J., Gao, C. (2008). NOMENCLATURE
Centrifugal compressor blade optimization based on
uniform design and genetic algorithm. Frontiers of 𝑚 mass flow rate kg/s
Energy and Power Engineering, p. 453. P pressure Pa
https://doi.org/10.1007/s11708-008-0083-5 𝜖 dissipation rate coefficient 𝑚2/𝑠3
[20] Tang, J., Turunen-Saaresti, T., Larjola, J. (2008). Use of Vi the velocity in the i th direction
partially shrouded impeller in a small centrifugal xi the coordinate in the i th direction
compressor. Journal of Thermal Science, 17(1): 21-27. is the air density
https://doi.org/10.1007/s11630-008-0021-1 i a tensor indicating 1, 2, 3
[21] Shaaban, S. (2015). Design optimization of a centrifugal E is the total energy of the air.
compressor vaneless diffuser. International Journal of K is the air thermal conductivity.
Refrigeration, 60: 142-154. Ji is the diffusion flux of j th species in the i th
https://doi.org/10.1016/j.ijrefrig.2015.06.020 direction.
[22] Kim, C., Lee, H., Yang, J., Son, C., Hwang, Y. (2016).
Study on the performance of a centrifugal compressor Abbreviations
considering running tip clearance. International Journal
of Refrigeration, 65: 92-102. 3D Three dimension
https://doi.org/10.1016/j.ijrefrig.2015.11.008 A/D analog to digital
[23] Pischinger, S., SchQnfelder, C., Bornscheuer, W., Kindl, CFD-RC Computational Fluid Dynamics Research
H., Wiartalla, A. (2001). Integrated air supply and Cooperation, name of a commercial code.
humidification concepts for fuel cell systems. SAE
International, Warrendale, PA. Geographic data
https://doi.org/10.4271/2001-01-0233
[24] Xu, C., Amano, R.S. (2008). Computational analysis of The main geometric characteristics required to initiate a: 1D
swept compressor rotor blades. International Journal for model on the wheel.
Computational Methods in Engineering Science and The rays at the different stations of the floor: R1h and R1S,
Mechanics, 9(6): 374-382. R2.
https://doi.org/10.1080/15502280802365840 The height of the blades: b2.
[25] Xu, C., Amano, R.S. (2017). Effects of asymmetric radial - The axial length: Lax.
clearance on performance of a centrifugal compressor. - The wedging angles of the blades: βp1h, and βp2, βp1s.
Journal of Energy Resources Technology, 140(5). - The number of ZPR blades.
https://doi.org/10.1115/1.4038387
[26] Xu, C., Amano, R.S. (2017). Centrifugal compressor Predicted aerodynamic data
performance improvements through impeller splitter
location. Journal of Energy Resources Technology, The following information in each of the stations of the
140(5). https://doi.org/10.1115/1.4037813 floor:
[27] Gonzalez, J., Fernandez, J., Blanco, E., Santolaria, C. - The performances: ηis and PR.
(2002). Numerical simulation of dynamic effects due to - Turbulence model K-ε.
impeller-volute interaction in a centrifugal pump. Journal - Thermodynamic quantities: PT, TT.
of Fluids Engineering, 124(2): 348-355. - Absolute and relative speeds: V and W.
https://doi.org/10.1115/1.1457452 - Absolute and relative flow angles: α and β.
[28] Chima, R.V. (2006). A three-dimensional unsteady CFD - Pressure ratio.

120