International Journal of Refrigeration 110 (2020) 239–247

Contents lists available at ScienceDirect

International Journal of Refrigeration

journal homepage: www.elsevier.com/locate/ijrefrig

Numerical analysis and correlations for radial and tangential leakage

of gas in scroll compressors
Evandro L.L. Pereira1, Cesar J. Deschamps∗
POLO Research Laboratories for Emerging Technologies in Cooling and Thermophysics, Federal University of Santa Catarina, CEP 88040-900, Florianopolis,
SC, Brazil

a r t i c l e i n f o a b s t r a c t

Article history: Scroll compressors are positive displacement machines of orbital motion that compress a gas by means
Received 8 May 2019 of two interfitting spiral-shaped scroll members. Gas leakage is an important source of inefficiency in
Revised 6 November 2019
scroll compressors and it therefore needs to be estimated. Currently, correlations for leakage required in
Accepted 7 November 2019
simulation models are based on simplified flow conditions. Thus, a high degree of uncertainty is asso-
Available online 13 November 2019
ciated with leakage predictions for the wide range of operating conditions found in actual applications.
Keywords: This paper reports a numerical analysis of radial and tangential leakage in scroll compressors considering
Scroll compressor different gases, operating conditions and geometries expressed as dimensionless parameters. Numerical
Radial leakage models were developed to analyze the aforementioned phenomena in detail, including the geometry of
Tangential leakage the wraps. Based on the predicted values, correlations for the radial and tangential leakage of gas partic-
Compressor modeling ularly suitable for comprehensive simulation models were developed and validated through comparison
with experimental data and other models available in the literature.
© 2019 Elsevier Ltd and IIR. All rights reserved.

Analyse numérique et corrélations des fuites radiales et tangentielles de gaz dans

les compresseurs à spirale

Mots-clés: Compresseur à spirale; Fuite radiale; Fuite tangentielle; Modélisation du compresseur

1. Introduction press the gas is dissipated when the gas leaks from one chamber
to the other. Moreover, hotter gas arriving in a chamber via leak-
The scroll compressor is well known for its efficiency in ap- age is compressed again, consuming extra amounts of energy and
plications involving high refrigeration capacity. An inherent advan- contributing to an increase in the temperature of the gas in that
tage of scroll compressors in comparison with other compression chamber.
mechanisms is the longer time available for the suction and dis- Basically, four modeling approaches for estimating leakage in
charge processes. However, gas leakage through different pathways scroll compressors are reported in the literature: (i) the isentropic
is a major source of thermodynamic irreversibility. For instance, ra- flow model; (ii) the incompressible viscous flow model; (iii) the
dial leakage occurs through the axial clearance between the tip adiabatic compressible viscous flow model (Fanno flow model);
of the involute and the opposite scroll base. Another pathway is and (iv) the quasi-one-dimensional model. The model for isen-
through the radial clearance between the flanks of the two scrolls, tropic flow in a convergent nozzle is the simplest approach and
resulting in tangential leakage. Leakage leads to a reduction in the it accommodates compressibility effects. Viscous and geometric ef-
isentropic efficiency, since a portion of the energy used to com- fects are included through the flow coefficient, and its value is usu-
ally adjusted with reference to experimental data (Kim et al., 1998;
Cho et al., 20 0 0). Considerable errors can be expected in small
∗
Corresponding author. clearances and when entrance losses are important, as in the case
E-mail address: deschamps@polo.ufsc.br (C.J. Deschamps). of radial leakage (Zuk et al., 1972). Ishii et al. (1996) adopted a
1
Present address: Embraco - Research and Development, Joinville, SC, Brazil.

https://doi.org/10.1016/j.ijrefrig.2019.11.002
0140-7007/© 2019 Elsevier Ltd and IIR. All rights reserved.
240 E.L.L. Pereira and C.J. Deschamps / International Journal of Refrigeration 110 (2020) 239–247

Prins and Infante-Ferreira (1998a) presented a comparative

Nomenclature analysis of leakage predictions given by four models based on
the quasi-one-dimensional formulation: (i) isentropic flow model;
a base circle radius (m) (ii) model of Ishii et al. (1996) for incompressible viscous flow;
cp specific heat at constant pressure (J kg−1 K−1 ) (iii) model of Yuan et al. (1992) for compressible viscous flow;
cv specific heat at constant volume (J kg−1 K−1 ) and (iv) model of Anderson (1995) for compressible viscous flow.
C∗ chamber dimensionless curvature (-) The authors used the experimental data of Ishii et al. (1996) and
CL entrance velocity loss coefficient (-) Peveling (1988) to assess the performance of these models and
Dh hydraulic diameter (m) found that none of them was appropriate for all conditions. In
h wrap height of scroll (m) another study, Prins and Infante-Ferreira (1998b) numerically es-
I turbulence intensity (-) timated leakage using the Fanno flow model and data reported by
L turbulence length scale (m) Peveling (1988) to calibrate an expression for the friction factor.
Lf length of channel formed by flank clearance (m) However, only a slight improvement was found in relation to leak-
fc compressor operating frequency (s−1 ) age predictions given by the model of Ishii et al. (1996).
m˙  mass flow rate per unit width (-) More recently, Bell et al. (2013) proposed a method to ob-
M Mach number (-) tain correlations for radial and tangential leakages of refrigerant
N number of compression chambers (-) in scroll compressors. Their method was based on developing an
ph upstream gas pressure (Pa) empirical correction term for the isentropic nozzle model using
pl downstream gas pressure (Pa) a large dataset of numerical results obtained from detailed one-
P potential for leakage (-) dimensional modeling of the leakage flow path. Some discrepan-
q form parameter (-) cies between the correlations and numerical results were verified
r0 radius of the circular orbit (-) by the authors in the case of choked flow, but both correlations
R gas constant (-) were seen to predict greater than 93% of the points within 20%.
Re Reynolds number (-) As can be seen from the literature review, considerable work
Rec critical Reynolds number (-) has been carried out on modeling leakage in scroll compressors,
Rc average curvature radius (m) but there is no clear evidence regarding the most appropriate
t wrap thickness of scroll; time (m; s) model to predict leakage in scroll compressors. The isentropic flow
Th upstream gas temperature (K) model seems to be the most widely adopted, with contraction co-
Tl downstream gas temperature (K) efficients defined via the calibration of experimental data. Incom-
Us characteristic velocity (m s−1 ) pressible and compressible viscous models are also used, and these
require the calibration of friction factors. This paper reports a nu-
Greek letters merical analysis of radial and tangential leakage in scroll compres-
α initial involute angle (-) sors and the development of correlations particularly suitable for
γ specific heat ratio (-) lumped simulation models, considering different refrigerants, op-
δ clearance (m) erating conditions and geometric parameters.
δf flank clearance (m)
δt tip clearance (m)
2. Simulation model
μ fluid dynamic viscosity (Pa s)
 pressure ratio (-)
The governing equations (mass, momentum and energy) for
ρ fluid density (kg m−3 )
compressible turbulent flow were solved for averaged quantities.
ρh upstream gas density (kg m−3 )
The Menter shear stress transport (SST) turbulence model was se-
ρl downstream gas density (kg m−3 )
lected to estimate the turbulent transport following the eddy vis-
ϕ involute angle (-)
cosity, μt , concept, given its accuracy for rapidly strained flows
ϕe final involute angle (-)
(Menter, 1993). The ideal gas equation of state was adopted to re-
late the density with the pressure and temperature. This simpli-
fying hypothesis is convenient for the dimensional analysis of the
model for incompressible viscous flow based on the correlations of problem, greatly reducing the number of simulations and allowing
Darcy-Weisbach to predict leakage. An iterative solution procedure the development of correlations for a wide range of operating con-
is required in their model due to the dependence of the friction ditions, geometries and fluids.
factor on the mass flow rate, characterized by the Reynolds num- The simulation model was developed with the commercial code
ber. The authors fitted and validated the model considering differ- ANSYS Fluent based on the finite volume method. A second-order
ent gases and operating conditions. upwind scheme was adopted to interpolate the flow quantities re-
In the Fanno flow model, the viscous effects are estimated quired at the faces of each cell of the computational grid. Analysis
via the friction factor. As the Mach number of the flow is not of the grid and time discretization was performed to assess trunca-
known, an iterative solution procedure is needed. An analytical so- tion errors and guarantee the solution accuracy. The coupling be-
lution is possible for the radial leakage, but the tangential leak- tween the pressure and velocity fields was achieved with the PISO
age requires a numerical solution because the dimension of the scheme.
clearance varies along the flow (Suefuji et al., 1992; Fan and The viscous wall region was modeled with the enhanced wall
Chen, 1994). One approach adopted is to consider the clearance to treatment combined with a two-layer model, in which the cells
be constant and adjust its length to obtain the same viscous fric- are split into a viscosity-affected region and a fully turbulent re-
tion loss (Yanagisawa and Shimizu, 1985). Finally, the quasi-one- gion. Following the two-layer approach, the SST turbulence model
dimensional model incorporates the changes in flow area, provid- is employed in the fully turbulent region (y+ > 30). On the other
ing a relationship between the pressure drop and the volumetric hand, in the viscosity-affected near-wall region the one-equation
flow rate along the clearance (Yuan et al., 1992; Huang, 1994). This model of Wolfstein (1969) is adopted.
class of model incorporates viscous and inertial effects, requiring a Grid refinement in the viscous wall region (y+ < 30) is essential
numerical solution procedure. to resolve the flow property gradients. In order to properly solve
 E.L.L. Pereira and C.J. Deschamps / International Journal of Refrigeration 110 (2020) 239–247 241

Fig. 1. Solution domain for radial leakage.

this region, two criteria were established: (i) the values of y+ for
grid cells adjacent to the solid walls should be close to 1; and (ii) Fig. 2. Solution domain for tangential leakage.
the viscous wall region should be discretized with at least 10 grid
cells. Moreover, the growth of the cell size near the wall region
The dimensionless curvature of the chamber
was limited to 10%. The grid convergence index (GCI) derived from
the theory of generalized Richardson extrapolation (Roache, 1998) Dh
C∗ = (1)
was adopted to verify grid convergence, i.e., to identify the point Rc
from which further grid refinement did not produce a variation
is a function of the hydraulic diameter, Dh , and the average curva-
greater than 3% in the mass flow rate.
ture radius, Rc :
4V 1
L + Lw,int

2.1. Solution procedure for radial leakage w,ext
Dh = Rc = (2)
As 2π 2
The length of the involute curve that forms the compressor where V and As are the volume and surface area of the compres-
chambers is typically much larger than its thickness, except near sion chamber.
the central region of the scroll. This allows the curvature to be ne- The parameter Rc is the average radius of the most central
glected and the analysis can be simplified to a two-dimensional chamber, i.e., the high-pressure chamber. The hydraulic diameter
flow. Fig. 1 illustrates the solution domain adopted in the two- Dh for a two-dimensional flow (h  a) is equivalent to Dh = 2r0 . All
dimensional modeling of the radial leakage through the tip clear- other considerations made for the radial leakage model (boundary
ance. In contrast to the work of Huang (1994), the opposite side- conditions, mesh refinement, convergence criteria, etc.) were also
walls of the chamber are positioned at an equal distance r0 from adopted in the modeling of the tangential leakage.
the clearance, which corresponds to the radius of the circular orbit.
The computational domain represents half of the chamber height.
The upstream boundary conditions of pressure, ph , and tempera- 3. Dimensional analysis and effect of different parameters on
ture, Th , are average values for the chamber and were imposed on leakage
the mean scroll height (h/2). Preliminary analysis showed low sen-
sitivity of the leakage to typical values of r0 and h. Thus, average Dimensional analysis was carried out to reduce the number of
values were adopted (t/a = 1.5, h/a = 10; where a is the radius of variables in the problem, thus decreasing the number of simula-
the base circle). The dimensionless clearance δt∗ (= δt /t) was the tions to a minimum. In addition to the dimensionless groups re-
parameter varied in the analysis. quired to ensure geometric similarity, the following dimensionless
In addition to the boundary conditions of pressure and temper- groups were defined for dynamic similarity:
ature (ph , Th ), turbulence intensity (I = 3%) and turbulence length m˙  pl p h ρh δ 2
scale (L = 0.07 r0 ) were also prescribed at the inlet. Identical con- M=  = P= (3)
δρh γ RTh ph 107 μ2
ditions for turbulence were imposed at the outlet boundary for the
case of reverse flow. The discretization of the solution domain al- In the above equations, δ is either the tip clearance δ t or flank
lowed the solution of the viscous layer near the wall. The solu- clearance δ f and R is the gas constant. A factor of 107 was used to
tion procedure was assumed to have reached convergence when limit the maximum value of P. The parameter M is the characteris-
the relative change in the mass flow rate over fifty iterations was tic Mach number at the clearance entrance and it is related to the
less than 10−5 . mass flow rate per unit width, m˙  . The parameter  is the ratio
between the downstream pressure, pl , and the upstream pressure,
2.2. Solution procedure for tangential leakage ph . Finally, P can be considered as the potential for leakage, asso-
ciated with concurrent forces due to pressure and viscous friction.
Unlike other models in the literature that simplify the flank ge- The ratio of specific heats, γ , is also required in the compressible
ometry to two circular curves of constant radius, this study con- flow analysis. Physically,  varies between 0 and 1. For typical con-
sidered the actual variation in the curvature. The in-house code ditions of interest, P is within the range of 10−3 to 103 , although
developed to generate the computational mesh restricted the so- values between 1 and 100 are most commonly observed. For re-
lution domain to two adjacent half-chambers, as illustrated in frigerants, γ assumes values between 1.10 and 1.40, with values of
Fig. 2. Thus, the dimensionless clearance δ ∗f (= δ f /Dh ) and curva- around 1.20 for most gases. The simplified equation for the speed
ture C ∗ (= Dh /Rc ) parameters define the flow geometry to be ana- of sound in the parameter M, with constant ratio of specific heats,
lyzed. was adopted for consistency with the hypothesis of ideal gas
242 E.L.L. Pereira and C.J. Deschamps / International Journal of Refrigeration 110 (2020) 239–247

15 est. Thus, the following functional relationships can be written for

R-134a Te = -10oC
o the tip clearance and flank clearance:
R-410a Tc = 45 C  
Mt = f (δt∗ , , P, γ ) M f = f δ ∗f , C ∗ , , P, γ (4)
Potential for leakage - P

Establishing the functional relationships (2) is the main objec-

10 tive of the study reported herein. Through the simulation models
described in Section 2, the influence of each dimensionless pa-
rameter of the problem on the leakage through the tip and flank
clearances was analyzed, considering conditions extending beyond
5 those expected for the flow through clearances in scroll compres-
sors.
Figs. 5(a) and 5(b) show the variation in the Mach number, M,
with the pressure ratio, , in the tip and flank clearances, respec-
tively. Different leakage potentials, P, are considered for δt∗ = δ ∗f =
0 0.01, and γ = 1.10. The dashed line represents the solution associ-
360 450 540 630 720 810 900 990
Involute angle [degrees] ated with the isentropic flow in a convergent nozzle. The discrete
points correspond to the numerical predictions, which were fitted
Fig. 3. Typical values of P during the compression process for R-134a and R-410a. with spline curves. As expected, M increases with a reduction in
 and for some values of P the flow reaches the sonic condition
at the exit of the tip clearance (Fig. 5a), or at the minimum cross
section in the case of the flank clearance (Fig. 5b). After reaching
this flow condition, M remains constant with a reduction in .
The relationship between the viscous and inertial effects can be
analyzed through the variations in M with P or M with clearances
δt∗ and δ ∗f , as shown in Figs. 6 and 7 for  = 0.1. The asymptotic
variations in M with P or M with clearances δt∗ and δ ∗f , delimit the
regions controlled by viscous effects, characterized by the lowest
values of P or of clearances δt∗ and δ ∗f . On the other hand, the
inertial effects are strong at the highest values of P or of clear-
ances δt∗ and δ ∗f . In the central regions of Figs. 6 and 7, M varies
considerably with P and the clearances δt∗ and δ ∗f , with the vis-
cous and inertial effects being important. As the viscous effects de-
crease, due to an increase in the flow potential or the clearances,
the flow behavior approaches the isentropic solution for compress-
ible flow in convergent nozzles. For the flow through the tip clear-
Fig. 4. Typical values of  during the compression process for R-134a and R-410a. ance, the viscous effects are negligible when P > 102 and δt∗ = 0.1
(Fig. 6a) or when δt∗ > 0.05 and P = 103 (Fig. 7a). On the other
hand, the flow in the flank clearance is influenced by the viscous
Figs. 3 and 4 show the typical values of P and  for the refrig- effects even for the maximum values of P and δ ∗f (Figs. 6b and 7b),
erants R-134a and R-410a during the compression process, which although the inertial effects are dominant.
were predicted with a lumped simulation model. The upstream In the case of the flank clearance, the influence of the cham-
and downstream pressures were fixed in the simulations for the ber curvature, C∗ , on M also needs to be considered. Indirectly, the
refrigerants R-134a and R-410a, according to the evaporating and dimensionless curvature of the chamber downstream of the min-
condensing temperatures of −10 °C and 45 °C. imum clearance, C∗ , defines the length Lf of the channel formed
It should be noted that the compression process takes place as by the flank clearance. As can be seen in Fig. 8, as the C∗ value
the shaft rotates between 360° and 990° for the compressor con- increases the viscous effects due to the reduction in the length Lf
sidered in the simulation. The complete compression process re- will decrease. It should be noted that due to the effects of the cur-
quires three orbits of the scroll to move the refrigerant from the vature and changing cross-section of the flow path, the leakage in
low-pressure suction condition to the high-pressure discharge con- the flank clearance δ ∗f is quite a different phenomenon in compar-
dition. The first 360° of rotation corresponds to the suction pro- ison with the leakage in the tip clearance δt∗ .
cess. Hence, the compression process starts after the shaft rotates The effect of the ratio of specific heats, γ , was also investigated.
360° and ends at approximately 990° when the compression cham- In general, it was observed that M decreases almost linearly by up
ber reaches the discharge port. to 10% when γ is varied from 1.10 to 1.40. It is important to note
The behavior of P follows the gas pressure itself, increasing ex- that although M decreases when γ is increased, the mass flow rate
ponentially with the compression process. The pressure ratio  per unit width m˙  increases proportionally with the square root of
undergoes a sudden drop at the beginning of the compression pro- γ , according to Eq. (3).
cess due to the pressure decrease associated with the rapid growth
of the suction chamber volume. This is followed by an almost lin- 4. Correlations for radial and tangential leakages
ear decrease in  until the beginning of the discharge process,
when the pressure in the chamber initially stabilizes and then in- The asymptotic increase in M as the pressure ratio  de-
creases again. As can be noted from the figure, the maximum leak- creases is well represented by the Richards growth function
age occurs near the start of the discharge process, when P is large (Richards, 1959):
and  is small.  
f1 1+q
Since the mass flow rate through the clearances must be deter- M= × − 1 × fγ (5)
mined, the Mach number is the dimensionless parameter of inter-
q 1 + q exp [− f2 × (1 − )]
 E.L.L. Pereira and C.J. Deschamps / International Journal of Refrigeration 110 (2020) 239–247 243

Fig. 5. Variation in M with  for different values of P: (a) tip clearance (δt∗ = 0.01); (b) flank clearance (δ∗f = 0.01 and C∗ = 0.2).

Fig. 6. Variation in M with P for different values of δt∗ and δ∗f : (a) tip clearance ( = 0.1); (b) flank clearance ( = 0.1 and C∗ = 0.2).

Fig. 7. Variation in M with δt∗ and δ∗f for different values of P: (a) tip clearance ( = 0.1); (b) flank clearance ( = 0.1 and C∗ = 0.2).
244 E.L.L. Pereira and C.J. Deschamps / International Journal of Refrigeration 110 (2020) 239–247

Fig. 8. Variation in M with C∗ for (a) different values of δ∗f ( = 0.1; P = 1); (b) different values of P ( = 0.1; δ∗f = 0.005).

Table 1 Table 2
Parameters c1, i , c2, i and c3, i in Eq. (9) for radial leakage. Parameters c1, i, j , c2, i, j and c3, i, j in Eq. (14) for tangential leakage.

P i c1, i c2, i c3, i P i j c1, i, j c2, i, j c3, i, j

≤1 1 0.593 6.83 × 10−3 0.538 ≤1 1 1 −6.5413 × 10−3 −1.6255 × 10−1 1.2519 × 10−2
2 6.223 1.89 × 10−3 0.296 2 2.4500 × 10−2 −7.6984 × 10−2 −5.0886 × 10−1
>1 1 0.876 3.59 × 10−3 0.299 3 −6.9910 × 10−3 −1.6456 × 10−1 8.6602 × 10−3
2 4.178 2.20 × 10−2 0.401 2 1 2.0122 × 10+1 −1.2248 × 10−1 8.0950 × 10+1
2 1.4157 × 10−1 −1.4801 × 10−1 6.0047 × 10−1
3 −1.6422 × 10−1 −1.7633 × 10−1 5.8955 × 10−1
>1 1 1 −9.4039 × 10−3 −1.6734 × 10−1 1.6841 × 10−2
where q (= 0.01) is a form parameter that changes the shape of 2 −8.8514 × 10−3 −1.2546 × 10−1 −3.0428 × 10−1
the curve and 3 −3.5852 × 10−3 −1.6996 × 10−1 4.4348 × 10−3
2 1 1.9877 × 10+1 −1.7071 × 10−1 1.6063 × 10+2
γ − 1.1 −2.2861 × 10−1 −1.0649 × 10−1
fγ = 1 − (6) 2 2.0050
4 3 3.2934 × 10−2 −4.6381 × 10−2 4.0344 × 10−1
is a function that corrects the value of M according to the ratio
of specific heats, γ . On the other hand, f1 and f2 are functions of
the dimensionless parameters that are dependent on the clearance The parameters h1, 1 , h1, 2 , h1, 3 , h2, 1 , h2, 2 and h2, 3 in
under analysis. In the case of the radial leakage, Eqs. (12) and (13) are written in the general form
 
0.53 hi, j (C ∗ ) = c1,i, j ln C ∗ + c2,i, j + c3,i, j (14)
f1 (P, δt∗ ) =  −0.80 (7)
1 + g1 × δt∗ The parameters c1, i, j , c2, i, j and c3, i, j are given in Table 2,
also for two intervals of values for P. The constants in Tables 1
10
f2 (P, δt∗ ) =  ∗ −0.56 (8) and 2 were obtained with the Levenberg–Marquardt curve-fitting
1 + g2 × δt algorithm. The data associated with the numerical simulations and
the correlations implemented in C++ are available for the inter-
where functions g1 and g2 are obtained from ested reader in a “Data in Brief” paper (Pereira et al., 2019).
c2,i − c1,i
gi (P ) = c1,i + (9)
1 + q × P −c3,i 4.1. Validation of the proposed correlations
The parameters c1, i , c2, i and c3, i are given in Table 1, according
to two different ranges of values for the dimensionless parameter The proposed correlations for radial and tangential leakage
P. were validated through comparisons with experimental data.
Functions f1 and f2 assume different forms for tangential leak- Given that few measurements for clearances with geometries
age: similar to those found in scroll compressors are available, ex-
perimental data obtained in general studies on gas flow through
  0.59
f1 C ∗ , P, δ ∗f =  ∗ −0.469 (10) microchannels were also used. It should be mentioned that mea-
δf
1 + g1 × surements of leakage through such small clearances are extremely
difficult due to manufacturing tolerances, clamping of the test sec-
 ∗    0.413
f2 C , P, δ ∗f = g2 × δ ∗f + 1.9 (11) tion, and distortions associated with thermal and pressure loads.
Considering the four works from which we extracted experimental
with functions g1 and g2 being expressed by data for comparisons with the correlation, only two of them
(Shi et al., 2001; Kim et al., 2017) provided some information
g1 (C ∗ , P ) = h1,1 P h1,2 + h1,3 (12)
about the measurement uncertainty. Shi et al. (2001) pointed out
an error less than 2.5% in the measurement of mass flow rate,
h2,1
g2 (C ∗ , P ) = (13) but no information was made available about the assessment of
1 + h2,2 P −h2,3 the clearance dimension. Kim et al. (2017) indicated that the mass
 E.L.L. Pereira and C.J. Deschamps / International Journal of Refrigeration 110 (2020) 239–247 245

Table 3
Test conditions in the experiments of Zuk et al. (1972).

Test 1 Test 2 Test 3

Upstream pressure – ph [kPa] 276 418 671

Upstream temperature – Th [K] 296 300 295
Clearance – δ [μm] 38 38 39

Fig. 10. Radial leakage: Exp. data and correlations: present study;
Bell et al. (2017); Zuk & Smith (1969).

Table 4
Test conditions in the experiments conducted by Shi et al. (2001).

ph [kPa] Knh Mah Reh

Test A 40.0 0.00019 0.455 7477.7

Test B 13.35 0.00053 0.326 1822.8
Fig. 9. Radial leakage: Exp. data and correlations: present study; Test C 6.65 0.00104 0.237 690.6
Bell et al. (2017); Zuk & Smith (1969). Test D 1.32 0.00514 0.072 41.5

flow rate was measured with a bubble meter within an error of

1% and that the radial clearance between the roller and cylinder Suefuji et al. (1992) measured the mass flow rate of R22
was confirmed by Scanning electron microscope to be 19.989 μm. through typical clearances of scroll compressors (clearance δ =
The results of the theoretical model of Zuk and 3.6–21.0 μm, length t = 4.5 mm, width L = 126 mm), by varying
Smith (1969) and correlations of Bell et al. (2013) are also the upstream pressure ph during the tests and keeping the down-
presented in the same figures. The analytical model of Zuk and stream pressure constant (pl = 100 kPa). Three clearances were
Smith (1969) was developed to estimate radial leakage in small tested (Test 1: 3.6 μm; Test 2: 9.0 μm; Test 3: 21.0 μm). How-
gaps between plates by applying the compressible flow for- ever, the authors did not provide information on the gas temper-
mulation of the Reynolds equation with static boundaries. The ature at the entrance. In the study reported herein, the gas was
correlations of Bell et al. (2013) for radial and tangential leakages assumed to be at ambient temperature (Th ≈ 300 K). For the three
of refrigerant were devised via a correction term for the isentropic clearances analyzed, reasonable agreement was observed between
nozzle model based on predictions obtained from one-dimensional the results of the correlation and measurements, as indicated in
modeling of the leakage flow path and considering real gas Fig. 10, with an average relative difference of 19.5%. The analyti-
properties. cal solution of the Reynolds equation (Zuk and Smith, 1969) is also
in good agreement with the measurements, whereas the correla-
4.1.1. Radial leakage tion of Bell et al. (2013) is seen to underpredict the leakage as the
Zuk et al. (1972) reported experimental results for air flow pressure ratio and clearance dimension are decreased.
through a radial diffuser geometry, which can be simplified to the Shi et al. (2001) studied the flow of air in clearances formed by
geometry of flow between parallel plates (clearance δ = 38–39 μm, parallel plates (clearance δ = 1.0 mm, length t = 120 mm, width
length t = 6.4 mm, width L = 9.55 mm). The test conditions are L = 50.5 mm). Experimental data were obtained for Th = 288 K
summarized in Table 3 according to the nomenclature used in this and four values for the upstream flow pressure ph , as indicated
work. The outlet pressure was varied in the tests in order to test in Table 4, resulting in four flow conditions characterized by the
different pressure ratios, . Knudsen number, Kh , Mach number, Mah , and Reynolds number,
Fig. 9 shows that predicted values for the radial leakage given Reh . The clearance dimension is much larger than those found in
by Eq. (4) are in good agreement with the experimental results scroll compressors, but it has a geometric ratio δ /t similar to that
of Zuk et al. (1972) under the three test conditions, with an av- found for the tip clearance. Fig. 11 shows that the predicted values
erage relative difference of 2.4%. On the other hand, the solution provided by the correlation proposed herein are in good agreement
of the Reynolds equation (Zuk and Smith, 1969) overpredicts leak- with the measured values reported by Shi et al. (2001), with an av-
age in all the test conditions because fluid inertia cannot be ne- erage relative difference of 11.7%. The greatest difference between
glected and hence the Reynolds equation fails. The correlation of the results for the two formulations occurs for test condition A,
Bell et al. (2013) is in reasonable agreement with the measure- probably because in this case the flow is in the laminar-turbulent
ments for the test conditions 1 and 2, but overpredicts the leak- transition regime, with a Reynolds number of approximately 7500,
age for the test condition 3. This deficiency can be the result which is more difficult to solve numerically. As can be seen, the so-
of neglecting the energy loss at the clearance entrance. In fact, lution of the Reynolds equation is not suitable in all the test con-
Zuk et al. (1972) solved the Fanno flow equations with and without ditions, with the exception of the test condition D in which the
the presence of minor losses at the clearance entrance and found Reynolds number is quite small. Again, as observed in the compar-
agreement with the measured values only when an entrance loss isons with the experimental data of Zuk et al. (1972), the correla-
coefficient CL = 0.6 was used to correct the stagnation properties tion of Bell et al. (2013) shows a trend to overpredict the leakage
at the entrance. as the mass flow rate, or Reynolds number, is increased.
246 E.L.L. Pereira and C.J. Deschamps / International Journal of Refrigeration 110 (2020) 239–247

relation considering different pressure ratios. As can be seen, leak-

age estimates provided by the correlation are in agreement with
the measured values for the most relevant pressure ratios ( = 0.1
and 0.5), with an average relative difference of 7.9%. For the pres-
sure ratio  = 0.9, there is a discrepancy between the correlation
and the measured values of approximately 45%, which cannot be
easily explained. Nevertheless, it should be noted that the exper-
imental data does not show the expected reduction in the mass
flow rate as the pressure ratio approaches  = 0.9. The correlation
of Bell et al. (2013) shows similar results, although overpredicting
the leakage to a moderate extent.

5. Conclusions

The leakage of gas in scroll compressors is a major source of

thermodynamic inefficiency. Current methods to estimate leakage
Fig. 11. Radial leakage: Exp. data and correlations; present study; are either based on excessively simplified flow conditions or adopt
Bell et al. (2017); Zuk & Smith (1969). more consistent formulations that are computationally too expen-
sive for use in comprehensive simulation models or compressor
optimization. A numerical analysis of radial and tangential leak-
ages in scroll compressors, considering different gases, operating
conditions and geometries, expressed as dimensionless parameters,
was reported in this paper. The results demonstrated that the ef-
fects of the entrance loss in the tip clearance and the curvature
of the scroll in the flank clearance can be significant and there-
fore should not be neglected. Based on the predicted values, cor-
relations for the radial and tangential leakage of gas were devel-
oped and compared with experimental data reported in the litera-
ture. Overall, the proposed correlations showed better agreement
with the experimental data than the results of two other mod-
els also included in the comparisons. The hypothesis of ideal gas
used in the numerical model is convenient for the dimensional
analysis of the problem, greatly reducing the number of simula-
tions. It is well known that refrigerants do not perfectly follow the
ideal gas hypothesis and in principle models for real gases, such as
Fig. 12. Tangential leakage: Exp. data and correlations: present study; those developed by NIST (2007), should be used instead. In fact,
Bell et al. (2017). although the comparison between the correlation and measure-
ments of Suefuji et al. (1992) for R22 showed reasonable agree-
ment, a systematic investigation should be carried out in future
4.1.2. Tangential leakage work, probably via real gas flow simulations, to assess the level of
In the experimental apparatus reported in the literature for the error introduced by the hypothesis of ideal gas.
analysis of tangential leakage a simplified geometry of two non-
concentric cylinders is adopted, similar to the geometry of the ra- Declaration of Competing Interest
dial clearance in rotary compressors. Hence, unlike the clearance
geometry formed in scroll compressors, the upstream and down- The authors declare that they have no known competing finan-
stream chambers in these experiments have identical geometries, cial interests or personal relationships that could have appeared to
with a constant curvature. influence the work reported in this paper.
Yuan et al. (1992) measured the volumetric flow rate of R-12
and R-22 through a simplified geometry of radial clearance δ be- Acknowledgements
tween 4 μm and 13 μm. Subsequently, Fan and Chen (1994) used
the same apparatus to include N2 in the measurements but, unfor- This study was conducted as part of a technical-scientific coop-
tunately, the width of the clearance cross-section was not reported. eration program between the Federal University of Santa Catarina
More recently, Kim et al. (2017) performed a numerical analysis and EMBRACO. The authors are also grateful to the Brazilian gov-
with experimental validation of the leakage through a simplified ernmental agency CNPq (grant 465448/2014-3) and to EMBRAPII
geometry of radial clearance. The experimental test section had a Unit POLO/UFSC.
clearance δ = 20 μm formed by two non-concentric cylinders of
Supplementary material
width L = 20 mm, and N2 was used as the working fluid. dr.. Min
in an email kindly gave us further details of this test section (B.
Supplementary material associated with this article can be
Min, personal communication, March 2, 2018), informing that the
found, in the online version, at doi:10.1016/j.ijrefrig.2019.11.002.
radii of the two non-concentric cylinders were 20 mm and 25 mm
and that the upstream gas temperature Th was 293 K. Therefore, References
the corresponding average radius, Rc , hydraulic diameter, Dh , and
dimensionless curvature, C∗ , are 22.5 mm, 10 mm and 0.44, respec- Anderson Jr, J.D., 1995. Computational Fluid Dynamics - the Basics with Applica-
tively. tions. McGraw-Hill, Singapore.
Bell, I.H., Groll, E.A., Braun, J.E., Horton, W.T., 2013. A computationally efficient hy-
Fig. 12 shows a comparison between the measured values re- brid leakage model for positive displacement compressors and expanders. Int. J.
ported by Kim et al. (2017) and those provided by the present cor- Refrigeration 36, 152–158.
 E.L.L. Pereira and C.J. Deschamps / International Journal of Refrigeration 110 (2020) 239–247 247

Cho, N., Youn, Y., Lee, B., Min, M., 20 0 0. The characteristics of tangential leak- International Compressor Engineering Conference at Purdue, West Lafayette,
age in scroll compressors for air conditioners. In: Proceedings of the Interna- pp. 571–576.
tional Compressor Engineering Conference at Purdue, West Lafayette, pp. 807– Prins, J., Infante-Ferreira, C.A., 1998b. Quasi one-dimensional steady-state models for
814. gas leakage - part II: improvement of the viscous modeling. In: Proceedings of
Fan, Z., Chen, Z., 1994. A calculating method for gas leakage in compressor. In: the International Compressor Engineering Conference at Purdue, West Lafayette,
Proceedings of the International Compressor Engineering Conference at Purdue, pp. 577–582.
West Lafayette, pp. 47–53. Richards, F.J., 1959. A flexible growth function for empirical use. J. Exp. Bot. 10,
Huang, Y., 1994. Leakage calculation through clearances. In: Proceedings of the 290–300.
International Compressor Engineering Conference at Purdue, West Lafayette, Roache, P.J., 1998. Verification and Validation in Computational Science and Engi-
pp. 35–40. neering. Hermosa Publishers, Albuquerque, New Mexico.
Ishii, N., Bird, K., Sano, K., Oono, M., Iwamura, S., Otokura, T., 1996. Refrigerant leak- Shi, W., Miyamoto, M., Katoh, T., Kurima, J., 2001. Chocked flow of low-density gas
age flow evaluation for scroll compressors. In: Proceedings of the International in a narrow parallel-plate channel with adiabatic walls. Int. J. Heat Mass Transf.
Compressor Engineering Conference at Purdue, West Lafayette, pp. 633–638. 44, 2555–2565.
Kim, Y., Seo, K., Park, H., 1998. Modeling on the performance of an inverter driven Suefuji, K., Shiibayashi, M., Tojo, K., 1992. Performance analysis of hermetic scroll
scroll compressor. In: Proceedings of the International Compressor Engineering compressor. In: Proceedings of the International Compressor EngineeringCon-
Conference at Purdue, West Lafayette, pp. 755–760. ference at Purdue, West Lafayette, pp. 75–84.
Kim, G., Min, B., Na, S., Choi, G., Kim, D., 2017. Estimation of leakage through radial Wolfshtein, M., 1969. The velocity and temperature distribution of one-dimensional
clearance during compression process of a rolling piston rotary compressor. J. flow with turbulence augmentation and pressure gradient. Int. J. Heat Mass
Mech. Sci. Technol. 31 (12), 6033–6040. Transf. 12, 301–318.
Menter, F.R., 1993. Zonal two equation k-ω turbulence models for aerodynamic Yanagisawa, T., Shimizu, T., 1985. Leakage losses with a rolling piston type ro-
flows. In: Proceedings of the 24th AIAA Fluid Dynamics Conference, Orlando, tary compressor. I. Radial clearance on the rolling piston. Int. J. Refrigeration
USA, Paper 93-2906. 8, 152–158.
NIST, National Institute of Standards and Technology. Refprop - Reference fluid ther- Yuan, X., Chen, Z., Fan, Z., 1992. Calculating model and experimental investigation of
modynamic and transport properties, version 8, USA, 2007. gas leakage. In: Proceedings of the International Compressor Engineering Con-
Pereira, E.L.L., Braga, V.M., Deschamps, C.J., 2019. Data from the numerical analysis ference at Purdue, pp. 1249–1255.
of radial and tangential leakage of gas in scroll compressors. Data Brief. Zuk, J., Smith, P., 1969. Computer Program for viscous, Isothermal Compressible
Peveling, F.J., 1988. Ein beitrag zur optimierung adiabater schraubenmaschinen in Flow Across a Sealing Dam With Small Tilt Angle. NASA Technical Note, D-5373,
simulationsrechnungen Ph.D. Thesis. VDI, Düsseldorf. Washington, D.C.
Prins, J., Infante-Ferreira, C.A., 1998a. Quasi one-dimensional steady-state models Zuk, J., Ludwig, L.P., Johnson, R.L., 1972. Quasi-one-dimensional Compressible Flow
for gas leakage - part I: comparison and validation. In: Proceedings of the Across Face Seals and Narrow Slots. NASA Technical Note, D-6668, Washington,
D.C.