International Journal of Heat and Technology

Vol. 36, No. 2, June, 2018, pp. 391-396

Journal homepage: http://iieta.org/Journals/IJHT

A quasi-one-dimensional model for the centrifugal compressors performance simulations

Wenhai Du1, Yucheng Li1, Longfei Li1, Giulio Lorenzini2*
1
Beijing Institute of Petrochemical Technology, Department of Mechanical Engineering, Qingyuan North Road 19, Daxing
District, Beijing 102617, China
2
University of Parma, Department of Engineering and Architecture, Parco Area delle Scienze 181/A, Parma 43124, Italy

Corresponding Author Email: giulio.lorenzini@unipr.it

https://doi.org/10.18280/ijht.360202 ABSTRACT

Received: 14 December 2017 This paper presents a quasi-one-dimensional numerical tool to simulate the performance of the
Accepted: 24 April 2018 centrifugal compressors. The current model is especially useful since it could offer the reliable
prediction for the centrifugal compressor performance only based on the simple geometries.
Keywords: An adapted version of the Euler equations solved at mid-span by a time-marching, finite-
centrifugal compressor, quasi-one- volume method, is applied in the model. The inviscid effect, the viscous effect and the
dimensional, numerical simulation, loss geometry variation effect in the centrifugal compressor are expressed by the source terms in
models the Euler Equations. In the study, different loss sources in the centrifugal compressor are
analyzed and estimated by empirical correlations. Two different centrifugal compressors are
applied to validate the current model and the numerical simulations are compared to
experimental data. The results suggest that the model provides a valuable tool for evaluating
the centrifugal compressor performance during the preliminary design and optimization
process.

1. INTRODUCTION form. The numerical model is applied to two different

centrifugal compressors and the simulation results are
The centrifugal compressors have been widely applied in compared to the experimental data. The results illustrate that
various domains such as aerospace, automotive industry, oil the model with the tuning coefficient in empirical correlations,
and gas industry, and power generation because of high has the capabilities to correctly capture the performance of the
efficiency, compact size and wide flow range. The high- centrifugal compressor.
performance centrifugal compressor is more appealing to the
turbocharger and micro gas turbine industries due to the fact
that the existing centrifugal compressor has no ability to meet 2. MODEL EQUATIONS
the required performance target in terms of efficiency and flow
range. Therefore it is necessary to develop a fast, robust and Usual quasi-one-dimensional models for axial compressors
general simulation tool to offer reliable predictions for radial only allow reasonable deflections of the fluid, because only the
machines. First of all, this tool could be used to perform axial speed is taken into account. Strong tangential speeds lead
parametric studies to demonstrate the influences of the to undervalued total variables and distort the solution. A fourth
geometry changes on the centrifugal compressor performance equation dealing with the tangential momentum conservation
both design and off-design conditions, especially during the was then added to the three classical quasi-1D Euler equations
preliminary design and optimization processes [1]. In addition, expressing the conservation of mass, axial momentum and
it has the ability to predict the overall performance of a total energy. The cylindrical system of equations was replaced
combination system where centrifugal compressor is a by a curvilinear one in order to extend the application field to
component of such a system [2-3]. radial flows. The equations are discretized at mid-span and the
Most of the centrifugal compressor models are based upon flow is assumed to axisymmetric. Interactions between the
space-marching techniques. They stack duct and bladed working fluid and the centrifugal compressor are introduced
elements, connected through physical variables and into the equations by the source terms. The blade force source
relationships. The impeller analysis often relies on the jet- terms are applied to consider the energy from the blades to the
wake model [4-5], but it has proved to be less accurate than working fluid, while the flow viscidity influences are
the meanline computations using correlations [6]. In this represented by the friction source term. The empirical
paper, the current program resulting from the extension of an correlations provide the information to evaluate the velocity
axial compressor and turbine model [7-8], is developed to triangles at mid-span and these triangles yield the necessary
model the stationary, transient and dynamic simulation information for the assessment of the source terms.
operation of a centrifugal compressor.
The tool is able to compute the flow through a centrifugal 2.1 General formulation
compressor, with detail at the blade row level. It relies upon a
much customized version of the Euler equations, expressed in The quasi-1D Euler equations are applied in the curvilinear
curvilinear, still quasi-one-dimensional and conservative (m, n, ) coordinate system. The upper and lower boundaries

391
of the control volume  shown on figure 1 are solid walls. The out the detailed parametric study on compressor performance
only “porous” boundaries of  are therefore the annular and correlated the efficiency of the single stage centrifugal
surfaces 1 and 2. compressor with four major parameters: inlet specific speed,
impeller tip diameter, inducer tip relative Mach number, and
exit discharge Mach number [13]. For centrifugal compressor
dm
impellers, this loss coefficient includes several contributions:

 tot =  inc +  bl +  sf +  df +  lk +  rc +  cl (4)

r 

• inc : incidence losses

• bl : losses due to the blade loading
1 2 • sf : skin friction losses inside the blade channel
dx • df : disc friction losses
• lk : leakage losses
• rc : recirculation losses
Figure 1. Elementary control volume • cl : clearance losses
In vector, non-dimensional, conservative form, these The incidence loss is evaluated by the correlation proposed
equations are written as by Conrad, which is represented as following:
U F
+ =Q (1) Hinc = finc wui2 2 (5)
t m

  Vm S 
where finc is the incidence factor and wui is the tangential
 S 
 V S   
U= m  
F=
( p +  Vm2 ) S 
(2)
component of impeller inlet relative velocity.
The blade loading loss is developed due to the negative
  V S  
    Vm V S  gradient of velocity in the boundary layer and the empirical
 e S    h S 
correlation is computed by Coppage [12].

The system of equations (1) is able to describe the evolution Hbl = 0.05D2f U 22 (6)
of the working fluid through any type of machine, axial, radial
or mixed-flow if the appropriate source terms are supplied to The diffusion factor D f in the equation is similar to the one
the model. The source terms Q could be divided into three of Lieblein [14].
distinct contributions, categorized according to their physical The skin friction loss results from the action of shearing
meaning. forces on the walls of the impeller due to the turbulent fluid
friction. The loss could be evaluated by the equation of Jansen.
Q = Qb + Q f + Qg (3)
Lb 2
H sf = 5.6C f V2 m (7)
Namely, Qb , Q f and Q g represent the blade force source Dhyd
term, friction force source term and geometrical source term
respectively. The detailed source terms have been applied for where C f is the skin friction coefficient, Lb is the impeller
axial compressors and turbines [8].
flow length and Dhyd is impeller average hydraulic diameter.
2.2 Method of the velocity triangles The disc friction loss mechanism occurs as a result of
adhesive forces between the rotating disk and the fluid in the
The predictive capabilities of the method are highly surrounding enclosure. The disc friction enthalpy loss used is
depending on the source terms, which are brought by the the one of Daily and Nece; it was selected by Oh et al. [15].
estimation of the velocity triangles for the compressor blade
rows at each time step. The flow variables from the current  r22 U 23
hdf = f df (8)
iteration and the blade geometrical data are used as inputs for 4m
the empirical correlations. The total pressure loss coefficients
and the slip factor model are evaluated through open literature With
correlations in order to determine the aerodynamic parameters
in the preliminary design process. The modular conception of 1 + 2
the computer code allows the introduction of tuning = (9)
coefficients in order to particularize the loss and slip factor
2
models for a given hardware to produce a reliable performance
 2.67
 Re0.5 , Re df  3  10
prediction for the centrifugal compressor. This has already 5

been done for axial compressors [7].  df

Empirical equations are established and studied by many f df = (10)
 0.0622 , Re  3  105
previous researchers including Stodola [9], Cordier [10],  Re0.2 df

Rodgers [11], and Wiesner [12]. Rodgers and Sapiro carried  df

392
 U 2 r2 There exist a variety of slip factor correlations in the
Re df = (11) literature. A recent and remarkable article presents a unified
2
correlation that is shown to supersede the correlations of
Stodola, Stanitz, Wiesner or Busemann as following [15-16].
The leakage loss comes from the leakage fluid through the
compressor seals to the regions of low pressure and is
1
proposed by Aungier as following [4].  = 1− (21)
 r1 
1 −  Z bl
 cl U cl U 2
1+ 5 
m r2 
hlk = (12)
2m 2 cos  2

With In the case of vaned diffusers, the deviation angle is

computed through adapted axial-flow compressor
2pcl correlations, with a reference deviation at minimum-loss
U cl = 0.816 (13) conditions, and then an incidence contribution [8].
2

m( r2 V 2 − r1V 1 ) 3. EQUATION SOLVER

pcl = (14)
Z bl r hv L
The calculation domain is discretized in the streamwise
r1 + r2
r = (15) direction at mid-span of the blade rows. The computation
2 domain is modified to take into account the curvilinear frame
hv1 + hv 2 of reference. The blade rows are generally divided internally
hv = (16)
2 into control volumes not only in the non-bladed regions but
also inside of each blade row, in order to obtain moderate flow
mcl = 2 Zbl  L U cl (17) deflections for each elementary cell. In each control volume,
the integral form of the Euler equations is written and the
The recirculation loss is associated with the recirculation of conservative variables are expressed at the cell center. A
air from the diffuser to the impeller and is set up by Oh et al. quadratic reconstruction scheme with a hybrid limiter and a
[15]; Roe’s flux splitting are used to extrapolate the variables and to
calculate the fluxes at the cell interfaces [17].
(
hrc = 8 10−5 sinh 3.5 23 D2f U 22 ) (18)
The real mesh counts more than three hundred mesh cells
and is less readable. It was found that the large deflections
imposed inside the blade zones to the working fluid had to be
The clearance loss is considered when the flow escapes distributed over an important number of inner blade cells to
from impelling by the clearance between the blades and the keep the truncation errors at an acceptable level. The rest of
outer casing of the centrifugal compressor. It could be the machine was meshed accordingly to obtain a homogeneous
expressed by Rodgers as following: 1D spatial discretization.
A time-marching method allows the flow to evolve towards
s 2 a stationary solution. Both explicit and implicit schemes are
H cl = 0.1 U2 (19) implemented. The explicit scheme is a fourth-order accurate
b2
Runge-Kutta. The implicit scheme is based on an Euler single-
step forward in time differentiation. For a stationary
The outlet tangential speed is therefore fully defined if the
simulation, a large CFL number can be applied to speed up the
slip factor can be deduced from the compressor geometry. The
calculation. The annulus boundary layer model proposed by
slip factor is deduced from figure 2 and is defined as
Stratford is implemented [7]. The boundary conditions applied
following:
at the inlet and the outlet of the compressor are displayed in
Table 1:
Vslip
 = 1− (20)
U2 Table 1. Imposed boundary conditions

Vslip
Inlet Outlet
input Pin0 , Tin0 ,  Pout

V2  2
2 4. NUMERICAL SIMULATION
Vm 2

The present preliminary design tool was validated by two

centrifugal compressor rigs from the open literature. The first
V 2 test case comes from the Came’s high speed centrifugal
V 2  compressor rig [18], and the second centrifugal compressor is
the CC3 impeller scaled up by McKain [19]. The centrifugal
compressor of Came has high rotation speed and high pressure
Figure 2. Impeller discharge velocity triangles

393
ratio, while the CC3 impeller is an intermediate pressure ratio.
Both of the two centrifugal compressors have the subsonic
inlet condition and represent the specific application in the gas
turbines and turbochargers.
Came’s centrifugal compressor has a rotating speed of
40000 RPM, the mass flow rate of 1.81kg/s and the total
pressure ratio of 6.5 at the design point. The assumed stage
isentropic efficiency is about 78% and the dimensionless
specific speed is 0.53. The CC3 impeller is designed to obtain
the stage total pressure ratio of 4.0 at a design mass flow rate
of 4.54 kg/s and a rotation speed of 21,789 RPM. The stage
isentropic efficiency is about 83% and the dimensionless
specific speed is 0.6. Both of the two centrifugal compressors
are designed with backswept trailing edges and splitter blades.
The Came’s impeller has 17 full blades and 17 splitter blades,
while the CC3 impeller has 15 full blades and 15 splitter
blades. Both of two centrifugal compressors have the vane
diffuser. The detailed design parameters of the two centrifugal
compressors are listed in the Table 2. Figure 4. Comparison the isentropic efficiency vs. mass flow
of Came’s compressor at 100% RPM
Table 2. Design parameters for the two centrifugal
compressors Figures 3 and 4 show that the total pressure ratio and
isentropic efficiency with the optimized correlations are
Parameter Came CC3 compared with the standard correlations results at 100%
Rotation speed N(rpm) 40,000 21,789 rotation speed. In figure 3, it can be found that the total
Mass flow rate m(kg/s) 1.81 4.54 pressure ratio with the ‘standard correlations’ is overestimated
Total pressure ratio Ptr 6.5 4.1 at reduced mass flow and high mass flow region, which
Isentropic  indicates that the slip factor has been overestimated. From
0.78 0.83
efficiency figure 4, the adiabatic efficiency is also overestimated with the
Blade number
Zf/Zs 17/17 15/15
standard correlations, showing only a constant offset that can
full/splitter easily be compensated by tuning the recirculation loss
Inlet hub radius R1h(mm) 30.48 41.4 coefficient. During the optimization process, the slip factor
Exit blade angle 2b (deg) 30.0 50.0 and pressure loss coefficients have been optimized according
to the experiment results. With the optimized coefficients, the
numerical results are quite satisfactory at 100% rotation speed
The Came’s centrifugal compressor is firstly applied to
in figures 3 and 4.
validate the numerical model. A parametric identification
procedure will be conducted to demonstrate the generalization
capabilities of the current tool. The objective of this procedure
is to optimize the total pressure ratio and adiabatic efficiency
according to the experimental results on the 100% nominal
speed data. It is supposed that the optimized loss and deviation
models will be able to provide a good prediction for the entire
range of mass flows and rotation speeds. The loss coefficient
and the slip factor coefficient are included in the optimization
process in order to satisfy the experimental results.

Figure 5. Total pressure ratio vs. mass flow with the

optimized correlations for Came’s compressor

For the Came’s centrifugal compressor, the predicted

performance maps with the optimized correlations show the
good trends comparing to the experiment results at 75%, 90%,
95% and 100% design rotating speed. The figures 5 and 6
show the numerical and experimental results, the total pressure
Figure 3. Comparison the total pressure ratio vs. mass flow ratio as well as for the isentropic efficiency.
of Came’s compressor at 100% RPM

394
 Finally, figures 7 and 8 present the comparison between the
measured results and numerical simulations for the total
pressure ratio and adiabatic efficiency respectively for the
CC3 centrifugal compressor. Although CC3 centrifugal
compressor has an intermediate pressure ratio, it can be found
that the similar numerical predictions are obtained with the
optimized correlations, which demonstrates the generalities of
the current tool.

Figure 8. Isentropic efficiency vs. mass flow with the

optimized correlations for CC3 compressor

The total pressure loss and slip factor correlations described

in the paper cover a large part of the physical phenomena
occurring during the stationary operation of a centrifugal
compressor. Meanwhile, the open literature correlations
introducing empirical information into the simulation can be
easily exchanged for a manufacturer’s own correlations, or
Figure 6. Isentropic efficiency vs. mass flow with the tuned by means of coefficients to be optimized. The Came’s
optimized correlations for Came’s compressor centrifugal compressor and CC3 centrifugal compressor are
selected as the validation test case. The total pressure loss and
slip factor coefficient are optimized according to the
5. CONCLUSIONS experiment performances at 100% RPM of Came’s centrifugal
compressor, the predicted performances of the two centrifugal
A quasi-one dimensional model for centrifugal compressor compressors during the entire rotation speeds are quite
has been presented. The simulation tool is developed from the satisfying. It indicates that the current code is useful during the
extension of the previous axial compressor, thus the equations preliminary design process of the centrifugal compressor.
have been transposed from cylindrical to curvilinear
coordinates for the centrifugal compressor in order to enhance
the simulation accuracy. The current tool uses only the ACKNOWLEDGMENT
machine geometry and the functional parameters such as the
rotation speed, the inlet total pressure and temperature, or the This research has been supported by Beijing Natural
outlet static pressure, to simulate the flow taking place inside Science Foundation (3182009).
the machine flowpath.

REFERENCES

[1] Whitfield A, Baines NC. (1990). Design of radial

turbomachines. Longman Scientific & Technical.
[2] Du W, Léonard O. (2011). A quasi-one-dimensional
CFD model for multistage compressors. ASME 2011
Turbo Expo: Turbine Technical Conference and
Exposition 17: 7-20. https://doi.org/10.1115/gt2011-
45497
[3] Wenhai DU, Zhu JQ, Léonard O. (2012). Dynamic
simulations of post-stall performance in multistage axial
compressors. Journal of Thermal Science 21(4): 311-
321. https://doi.org/10.1007/s11630-012-0549-y
[4] Aungier RH. (2000). Centrifugal compressors : a strategy
for aerodynamic design and analysis. ASME Press.
[5] Djedai H, Mdouki R, Mansouri Z, Aouissi M. (2017).
Numerical investigation of three-dimensional separation
control in an axial compressor cascade. International
Figure 7. Total pressure ratio vs. mass flow with the Journal of Heat & Technology 35(3): 657-662.
optimized correlations for CC3 compressor https://doi.org/10.18280/ijht.350325

395
[6] Frigne P, Van DBR. (1978). One dimensional design of ordinates_for_a_High_Performance_41_Pressure_Ratio
centrifugal compressors taking into account flow _Centrifugal_Compressor
separation in the impeller. VKI Technical Note 129, pp.
28-42.
[7] Adam O, Léonard O. (2005). A Quasi-one-dimensional NOMENCLATURE
model for axial compressors. 17th ISABE Conference,
Münich, Germany. b2 Outlet blade height [m]
[8] Léonard O, Adam O. (2008). A quasi-one-dimensional
e specific internal energy [KJ/Kg]
cfd model for multistage turbomachines. Journal of
Df dimensionless diffusion factor
Thermal Science 17(1): 7-20.
https://doi.org/10.1007/s11630-008-0007-z F conservative flux vector
[9] Lou F, Fabian JC, Key NL. (2018). A new approach for h specific enthalpy [KJ/Kg]
centrifugal impeller preliminary design for aero-thermal L length [m]
analysis. Journal of Turbomachinery 140: 1-10. p static pressure [Pa]
https://doi.org/10.1115/1.4038876
[10] Cordier O. (1955). Similarity considerations in Q source term vector
turbomachines. Verlag. Dusseldorf, Germany. VDI qm mass flow [Kg/s]
Reports 3. https://doi.org/10.1115/cec1955-0104 r radius [m]
[11] Rodgers C. (1964). Typical performance characteristics s dimensionless solidity
of gas turbine radial compressors. Journal of Engineering
for Gas Turbines & Power 86(2): 161. S flowpath cross-sectional area [m2]
https://doi.org/10.1115/1.3677568 t time [s]
[12] Tian Y, Hu A. (2018). Study on critical speed of rotation U conservative variable vector
in the multistage high speed centrifugal pumps rotors. U tangential rotation speed [m/s]
International Journal of Heat & Technology 36(1): 31- V absolute velocity [m/s]
39. https://doi.org/10.18280/ijht.360105
[13] Velásquez EIG. (2017). Determination of a suitable set Vol blade row volume [m3]
of loss models for centrifugal compressor performance W relative velocity [m/s]
prediction. Chinese Journal of Aeronautics 30(5): 1644- Zbl blade number
1650. https://doi.org/10.1016 /j.cja.2017.08.002
[14] Lieblein S. (1965). experimental flow in two- Greek symbols
dimensional cascades. NASA, SP-36 183-226.
https://doi.org/10.1115/1.801926.ch5  absolute flow angle [deg]
[15] Oh HW, Yoon ES, Chung MK. (1997). An optimum set  relative flow angle [deg]
of loss models for performance prediction of centrifugal
𝑤
̅ pressure loss coefficient
compressors. Proceedings of the Institution of
Mechanical Engineers Part A Journal of Power & Energy
 density [kg/m3]
211(4): 331-338. σ slip factor
https://doi.org/10.1243/0957650971537231 ε blade tip clearance [m]
[16] Wiesner FJ. (1967). A review of slip factors for υ Kinematic viscosity coefficient [m2/s]
centrifugal impellers. Journal of Engineering for Gas
Turbines & Power 89(4): 558-566. Subscripts
https://doi.org/10.1115/1.3616734
[17] Hirsch Ch. (1988). Numerical computation of internal m meridional
and external flows. Fundamentals of Numerical x axial
Discretization, John Wiley & Sons, Chichester, UK. n normal directions/components
[18] Came PM. (1978). Development, application and r radial
experimental evaluation of a design procedure for  tangential
centrifugal compressors. Proceedings of the Institution of t throat condition
Mechanical Engineers 192(1): 49-67. 1 blade leading edge
https://doi.org/10.1243/pime_proc_1978_192_051_02 2 blade trailing edge
[19] Mckain TF, Holbrook GJ. (1997). Coordinates for a high
performance 4:1 pressure ratio centrifugal compressor.
Superscripts
NASA Contract Report 204134.
https://www.researchgate.net/publication/24321070_Co
 total condition

396