Você está na página 1de 25

2010-01-1536

Steady and Unsteady Modeling of Turbocharger Compressors


for Automotive Engines
Fabio Bozza, Vincenzo De Bellis
University of Naples “Federico II”, Italy
Copyright © 2010 SAE International

ABSTRACT
Turbocharging technique will play a fundamental role in the near future not only to improve
automotive engine performance, but also to reduce fuel consumption and exhaust emissions both in
Spark Ignition and Compression Ignition engines. To this end, one-dimensional (1D) modelling is
usually employed to compute the engine-turbocharger matching, to select the boost level in
different operating conditions and to estimate low end torque level and transient response.
However, 1D modeling of a turbocharged engine requires the availability of the turbine and
compressor characteristic maps. This leads to some typical drawbacks:

• performance maps of the turbocharger device are usually limited to a reduced number of
rotational speeds, pressure ratios and mass flow rates. Extrapolation of maps’ data is
commonly required;

• performance maps are experimentally derived on stationary test benches, while the
turbocharger, coupled to an internal combustion engine, usually operates under unsteady
conditions;
• during low speed, high load engine operation a close-to-stall compressor operation usually
occurs. In this case the steady map cannot provide the required information to realize an
accurate analysis of the whole turbocharged engine.
To overcome the above problems, in the present paper two different numerical procedures are
developed: a steady approach is firstly followed to the aim of accurately reproduce the
experimentally derived compressor characteristic maps. The steady procedure describes main
phenomena and losses arising within the stationary and rotating channels constituting the
compressor device. It is utilized to directly compute the related steady map, starting from the
specification of a reduced set of geometrical data. An optimization process is also presented to
identify a number of tuning constants included in the various loss correlations.

Then, a recently proposed and more refined procedure is compared to the previous one. The latter is
based on the solution of the 1D unsteady flow within the compressor stationary and rotating
channels. The refined methodology is capable of describing the unsteady behaviour of the
compressor and to handle typical unstable operating regimes (compressor surging).

Page 1 of 25
Both the steady and unsteady procedures are applied to the simulation of three different
turbochargers and the numerical results show a good agreement with experimentally derived
performance maps.

In addition, the comparison between the steady and 1D procedure results highlights the important
role played by the unsteady phenomena on the overall turbocharger operation.

The proposed methodology can successfully support the design process and transient analysis of
turbocharged internal combustion engines.

INTRODUCTION
Current literature is producing more and more numerous contributions dealing with internal
combustion engine turbocharging [1,2]. Advanced technologies, such as variable geometry and
multi-stage turbocharging, high and low pressure loop EGR systems, assisted turbocharging and
turbocompounding, are emerging to control emissions and maintain fuel economy in both diesel and
gasoline engines.

Complex layouts of the turbocharging system have been in fact recently proposed for heavy-duty
diesel engines [3]. As known, high Exhaust Gas Recirculation (EGR) rates are required to meet
future emissions demands. This also requires a considerably higher boost pressure to maintain the
air excess ratio [4]. A precise managing of the boost level is also needed to extend the operating
domain of Homogeneous Charge Compression Ignition (HCCI) engines and to control the
autoignition time [5].

More recently, manufacturers are devoting a growing attention to the development of modern
turbocharging systems for Spark-Ignition (SI) engines, too. The revival of turbocharged SI engine is
based on the need for fuel economy improvement [6,7]. This can be accomplished by reducing the
engine size and limiting the throttle usage and pumping losses [8,9]. In order to reach the
performance level of larger engines, a turbocharging system is employed, also realizing a higher
low-speed torque, a better vehicle drivability and fun-to-drive (engine downsizing).

During the development of a new turbocharged engine, the computation of the engine and
turbocharger matching conditions is a very common task. This is usually realized through the
employment of 1D codes [10], which can also provide the boundary conditions for more refined 3D
CFD analyses [11]. These procedures require the specification of the turbine and compressor
characteristic maps, provided by the manufacturers. Anyway, in many cases, a limited operating
domain of the turbocharger device is really available, and various extrapolation techniques of maps’
data are commonly included within commercial 1D codes [12,13,14].

In addition, due to various phenomena, the actual compressor behavior may differ from the one
expressed by the characteristic map. First of all, performance data refer to a steady operation of the
device, while the compressor, due to the pressure pulsations caused by the engine [15,16], always
operates under unsteady conditions. Due to packaging of the inlet duct, the incoming air may be not
uniformly distributed across the inlet section of the compressor wheel, determining a variation of
the flow direction at the inlet. Moreover, heat transfer through the casing and heat radiated by the
turbine and the engine may affect the map based values of the adiabiatic efficiency.

Due to the above considerations, an improvement of the classical map-based quasi-steady matching
calculation is required [17]. In the present paper the approach followed in [27] is extended. Both a
Page 2 of 25
steady and an unsteady 1D simulation procedure is developed to characterize the behavior of small
centrifugal compressors, starting from a direct measurements of a reduced set of geometrical data.
In both cases, the whole compressor is schematized as a series of five pipes: the inlet duct, the
impeller, the diffuser, the volute and the outlet duct. The impeller itself is considered as a number of
rotating channels in parallel. In particular, the geometry of the blade-to-blade duct is specified
through the assignment of a limited number of geometrical wheel data, which define the duct
orientation in the space along its curvilinear abscissa.

In order to reproduce the steady experimental map, a direct modeling of main phenomena and
losses arising within the stationary and rotating channels is required. Incidence and blade losses
model are proposed by Leonard and Adam [18], Köning et al. [30], Swain [31] and Japikse [28].
Slip effect is studied by a large number of researchers, such as Wiesner [22], von Backström
[32,33], Paeng and Chung [35], Caridad and Kenyery [34]. They correlate slip factor to impeller
geometry (exit blade angle, number of blades, inlet to exit radii ratio) basing on experimental data.
Qiu et al. [24] propose a new slip factor model, where an explicit dependence on the flow
coefficient is introduced. Dubitsky and Japikse [26] considers jet-wake mixing phenomena and
furnishes a simple friction coefficient correlation for the evaluation of the losses inside a vaneless
diffuser. Gu et at. [29] analyze the losses inside the volute and the volute/diffuser interaction at the
impeller exit. Japikse [45] defines a sudden enlargement loss model to account for volute losses,
mainly at low flow rates.

Additional studies are carried out to also analyze unsteady [41] and/or near-to-stall operating
conditions [36,37,38,39,40,42] using both numerical techniques and experimental investigations.
They include non-linear lumped parameter models to predict surging occurrence and to develop
advanced surging control systems.

In the developed model, proper correlations accounting for slip effect, incidence losses, shock
losses, friction losses and heat transfer are selected from the cited literature. Secondary loss
contributions are considered [23], too. An optimization process is also presented to identify
uncertainties in a number of tuning constants included in the various loss correlations. The steady
procedure is applied to characterize the behavior of three different compressors and to validate the
proposed model. Then, the unsteady procedure is utilized, with the previously identified tuning
constants, to describe soft-surge phenomena, which may arise particularly at low engine speed and
high load.

In the following paragraphs, the geometrical procedure for the compressor schematization is firstly
described in detail. Then, the 1D flow equations referred to both a stationary or a rotating frame are
reported, together with the boundary conditions which handle the transition between the absolute
and the relative motion. The results will show a good agreement with experimentally derived
performance maps and highlight, in particular operating conditions, the importance of a fully
unsteady simulation of the device.

Page 3 of 25
1D COMPRESSOR MODEL
GEOMETRICAL MODEL – The compressor is schematized as follows (Fig. 1):

4
5
V
3
r3 4
2
r2 1
s
1 0
r1
0 Fig. 1 – Main compressor stations

• a short inlet pipe, (from section 0 to 1); the flow direction α1 at the inducer inlet is specified
as an input data; the end section of the inlet pipe is characterized by a sudden area
contraction from Ωinl=πD12/4 to Ω1=π(D1s2- D1h2)/4, taking into account the impeller eye
obstruction. This will cause some inlet losses, as explained in the following;

• the impeller (from section 1 to 2), consisting of Z rotating pipes, Z being the number of
blades. The profile of the blade-to-blade duct is assigned in terms of inlet and outlet blade
angles ϕ1c and ϕ2c and duct length. In order to characterize the flow area, the distribution of
the equivalent diameter along its curvilinear abscissa, s, is also assigned. The latter is
estimated perpendicularly to the blade profile. In addition, the radius distribution from r1 to
r2 is prescribed to compute the local centrifugal forces acting on the fluid;

• the vaneless diffuser (from section 2 to 3). It is considered as a diverging pipe of constant
width in the radial direction. The pressure recovery is predicted by the solution of the
momentum equation related to both the radial and the tangential flow components, as
discussed in the following;

• the volute (from section 3 to 4), schematized as a constant equivalent area duct, collecting
the flow contributions coming from the Z impeller channels. An equivalent length is
considered as half the circumference at the radius r3. The volute ends with an outlet cone of
assigned dimensions (D4, L4);

• the outlet pipe (from section 4 to 5), of length L45 and constant area D4, ending with a
throttle valve, V, controlling the delivered mass flow rate.

Page 4 of 25
r x
φ2h D2
b
φ2s D1h
D1s
D3
xs γ2s
φ1s γ2h
∆x
φ1h

Fig. 2 – Main geometrical characteristics of a turbocharger compressor

Table 1 reports the geometrical data required by the procedure to completely characterize the
geometry of the compressor. Data refer to 3 different small and medium-size turbocharger
compressors (A, B and C). The description of characteristic profiles is based on the hypothesis that
the hub, shroud and meanline meridional curves are elliptical (Fig. 3a). Moreover, the local blade
angle ϕ is assigned as a function of the of local rotation angle γ. In particular, a cubical polynomial
reconstruction along the meridional coordinate, m, is assigned: γ = γ(m). Meridional profiles can be
drawn starting from R1h, R1s, R2 radii, b and ∆x length. Meanline inlet radius is assumed equal to the
geometrical average between hub and shroud radii eq. (1). This assumption ensures that meanline
divides blade to blade duct into two flow pipes delivering the same mass flow rate (Fig. 3a). The 3D
blade geometry is reproduced simply rotating each point of the meridional profile by the local γ
angle. The local blade angle is finally derived by the relationship reported in eq.(2) (Fig. 3b).

Compressor A B C
Inlet Pipe
D0 - L01 Diameter - length of the inlet pipe, mm 47.2 – 28.8 44.8 – 70.3 44.2 – 51
Impeller Wheel
Z Blade number 8 12 10
∆x Axial length of the wheel, mm 16.2 14.5 16.2
D1h - D1s Hub - shroud diameter at the wheel inlet, mm 13.1 – 36 10.6 – 34.1 9 – 30
D2 Wheel outlet diameter, mm 51 46 40
ϕ1h - ϕ1s Hub - shroud inlet blade angles, deg 30° - 53° 25° - 58° 30° - 47°
ϕ2h - ϕ2s Hub - shroud outlet blade angles, deg 43° - 40° 45° - 42° 49° - 39°
γ2h - γ2s Hub - shroud blade rotation angles, deg 67° - 58° 47° - 41° 60° - 54°
b Meridional blade height at the outlet, mm 3.5 3.1 3.6
xs Start splitter blade, mm 5.0 5.4 6.8
sp Blade thickness, mm 0.5 0.6 0.6
hgap Recirculation gap, mm 0.3 0.3 0.3
Vaneless Diffuser
D3 Diffuser outlet diameter, mm 80 91.1 75.4
Volute
D4 - L4 Diameter and length of the outlet cone, mm 53 – 162.2 33.3 - 120 33.9 - 120
Outlet Pipe
L45 outlet pipe length, mm 300 300 300

Table 1 – Geometrical data required by the procedure

Page 5 of 25
r
ϕ
b rdγ
Shroud
dm
Meanline blade
R2
Hub r
R1s
R1m
R1h γ
X

∆x
r
(a) (b)

Fig. 3 – Geometrical schematization of a turbocharger compressor

R1h + R1s dγ tgϕ


R1m = (1) = (2)
2 dm r

Basing on the above procedure, a good reconstruction of the 3D blade profile can be accomplished
and reproduced in a CAD software. As an example, Fig. 4 shows a comparison between the actual
impeller geometry and the CAD reproduced ones, basing only on data listed in Table 1, for the three
compressor A, B and C. To obtain the geometrical information of blade to blade duct modeled in
the 1D code, hydraulic diameter and cross section area also need to be evaluated. Cross section is
drawn starting from the intersection between 3D hub and shroud profiles of two consecutive blades
and a conical surface with a vertex V located on x axis. This conical surface is identified by the θel
angle and a generating line passing through the center of meridional profiles (Fig. 5a). The
intersection between the cone and the wheel duct is a quasi-parallelogram.

The parallelogram is then projected on the plane perpendicular to meanline profile in x-r plane (blue
and red sections, Fig 5b). A further projection is finally carried out on the plane perpendicular to
meanline profile in x-z plane (red and green sections, Fig. 5c). The numerical procedure takes into
account the decrease of flow section area due to blade thickness, sp, and splitter blades. The position
along x axis corresponding to splitter blade leading edge is identified by the parameter xs in Table 1.
Besides, the geometrical module furnishes the equivalent diameters di, ds, dh, defined as:

4Ω 4Ω 4Ω
di = ds = dh = (3)
Pi Ps Ph

being Pi, Ps, Ph, the impeller, the case (shroud) and the hub wetted perimeter, respectively. These
characteristic dimensions are employed to evaluate friction and heat transfer terms in the flow
equations.

Page 6 of 25
Compressor A Compressor B Compressor C

Fig. 4 – 3D reproduced impeller geometry

z
ϕ
θel
θel θ
meanline

V X X r

(a) (b) (c)

Fig. 5 – Schematization required to derive the geometry of the 1D blade to blade pipe.

UNSTEADY FLOW MODEL – The unsteady 1D flow in the stationary ducts is described by the
following equations:

Ct + [F(C)]x = SC
Page 7 of 25
 
 ρcα 
 ρ   ρc   
      
F(C) = ρc 2 + p 
f c (4)
C =  ρc  SC = −ρc 2  α + 2 

ρE   ρcH    d c 
 C  C    
q
 ρcH c α − 4 
 d 

The terms ρ, c, p, EC=cvT+c2/2, HC=cpT0= cpT+c2/2 in the system (4) respectively represent
density, absolute velocity, static pressure, total energy and enthalpy per unit mass. The source term,
SC, accounts for the duct area variation, α = 1/Ω dΩ/dx, the friction coefficient, f, and the rate of
heat exchange, q.

Impeller Flow Model - The system (4) is solved for the inlet pipe, the volute and the outlet duct,
while, within the impeller, the following balance equations apply [18]:

Wt + [F(W )]s = SW

 
 
 ρwα 
 ρ   ρw   
      2 f w  2 f s 2 cw 
W =  ρw  F(W ) = ρw 2 + p  SW = −ρw 2  α + i + ρc − ρu 2 δ  (5)
 d w  d c
ρE   ρwH    i  s 
 W  W    
 ρwH α − 4ρ qh + qs  − ρwu 2 δ 
 W d  
 h ds 

In the above equations, w is the relative velocity, u=ωr, is the tangential velocity (blade speed) and
EW=cvT+w2/2, HW=cpT0w=cpT+w2/2 are the total energy and enthalpy in the relative motion. The
source term SW includes additional contributions arising as a consequence of pipe rotation. They are
related to the centrifugal forces acting on the fluid particle and are computed as a function of the
radius variation along a streamline: δ = 1/r dr/ds.

Coriolis forces, indeed, always being perpendicular to the relative velocity, do not contribute to the
shaft work and are not included in the system (5). Friction forces are subdivided in two
contributions, the first related to the impeller wetted diameter di and computed as a function of the
relative velocity w, and the second concerning the case wetted diameter ds, and computed as a
function of the component cw of the absolute velocity along the meanline (cw = w - u sin ϕ). In this
case, the friction coefficient fs is evaluated with respect to the total absolute velocity c, computed
imposing the velocity triangle. Also heat exchange is subdivided in two terms, the first depending
on impeller hub wetted diameter, dh, and evaluated according to w, and the second arising from heat
exchange on the case, ds, and correlated to c.

Diffuser Flow Model - Along a flow streamline in the vaneless diffuser, a variation of both the
radial and the tangential velocity components, cr, cu, occurs, determining a 2D flow. The system of
equations solved for the vaneless diffuser is hence:

Dt + [F(D)]r = S D

Page 8 of 25
 ρc r α 
 f c 
 ρ   ρcr   ρcr2 α − ρcu2 δ + ρc 2 r 
ρc  ρc 2 + p   b c 
     
D =  r  F(D) =  r  S D = − f 2 cu
ρcr cu α + ρcr cu δ + ρc 
(6)
ρcu   ρcr cu   b c 
 E D  ρcr H D   c 
ρcr H D α − 2ρq − ρcr cu2 δ − f ρcu c 2 u 
 b b c 

being ED=cvT+cr2/2 and HD=cpT0=cpT+cr2/2. With this definition of ED and HD, the energy balance
equation shows an additional term deriving from tangential component of friction force. In this
case, along the radial direction: δ = 1/r.

In the above unsteady procedure, the numerical solution of the systems (3), (4) and (5) is realized
through a 2nd order accurate upwind scheme with a TVD flux correction technique [19]. Temporal
integration follows a Runge-Kutta 4th order method.

STEADY FLOW MODEL - In the steady procedure, temporal derivatives in eqs. (3), (4) and (5)
are obviously neglected. Computation starts with a fixed value of mass flow rate at compressor inlet
duct. Flow equations are discretized in finite differences and numerically integrated with an
iterative procedure.

BOUNDARY CONDITIONS
The boundary conditions in the nodes of adjacent ducts are specified by applying a classical quasi-
steady pipe-to-pipe junction problem. Mass and energy equations at the junction are solved. In the
unsteady procedure, in order to preserve the characteristics of wave propagation phenomena, the
latter are coupled to the compatibility relations, properly modified to include the additional source
terms. The momentum equation is indeed substituted by a total pressure loss relation between
adjacent sections of the connected pipes. Exactly the same pressure loss terms are introduced in the
steady and unsteady procedures. Under the assumption of isentropic flow, pressure loss terms are
ignored. As an example, concerning the “isentropic” boundary condition between the inlet pipe
(section 1) and the inducer throat (1t), the relative velocity, w1, in the last section of the inlet pipe is
first computed from the velocity triangle. Then, the associated Mach number, Mw1, and total
pressure, p0w1, are estimated from the static pressure, p1, and temperature, T1, as:

γ
w1  γ − 1 2  γ −1
M w1 = p0 w1 = p1 1 + M w1  (7)
γRT1  2 

In the absence of incidence losses, the total pressure p0w1 is imposed unchanged in the inducer
throat section (p0w1t = p0w1) and employed to derive the static pressure p1t. The latter may result
higher or lower than p1 depending on the local flow rate.

Page 9 of 25
1t

1 1

Fig. 6 – Variable geometry inlet inside the impeller [20].

Referring to the Fig. 6, in fact, the inlet process inside the inducer has been described as a passage
through a fictitious variable geometry nozzle or diffuser. At high flow rate, the relative flow is
required to expand up to the throat section, while the opposite situation occurs at low flow rate. At
the design point (zero incidence), indeed, the pressures p1 and p1t are the same.

Concerning the boundary condition between the impeller and the diffuser, the outlet velocity
triangle is drawn to compute the absolute velocity c2, mach number Mc2 and total pressure p02. In
the absence of losses, the total pressure p02 is imposed unchanged in the first diffuser section.

SLIP EFFECT, FLOW LOSSES AND HEAT TRANSFER


In order to predict the actual performances of a centrifugal compressor both the ‘slip-effect’ and
various flow losses, taking place in different stations of the device, should be considered. Flow
losses can be grouped in distributed and incidence losses and are included in the present model
under the form of simple correlations. Due to the lack of their generality, some tuning constants are
also introduced.

SLIP EFFECT – Even under ideal (frictionless) conditions, the flow leaving the impeller is said to
‘slip’. The presence of a finite number of blades determines an increase of the flow angle ϕ’2,
respect to the blade angle ϕ2c. A consequent reduction of the tangential component c2u occurs (Fig.
7), determining a decrease of the energy transfer, quantified by the slip factor, σ:

vslip
σ = 1− vslip = c2u − c2' u (8)
u2

ϕ 2'
ϕ 2c

w2 c2 '
w2' c2
u2
c2u'
c2u

Fig. 7 – Outlet velocity triangles and slip effect

Among the several correlations found in the literature [21,22], it was here employed the one
proposed in [24]:
Page 10 of 25
 Fs Φ cos ϕ 2 c Fs 2 Φ 2  dϕ  FΦ 2 s 2 sin ϕ 2c  dρb  
σ = 1 − c sl  2 2 −   +    (9)
 Z 4 cos ϕ 2c  dm  2 4ρ 2b2  dm  2 

csl, s2, Φ2, b2, F, being, respectively, a tuning multiplier, span at impeller exit, flow coefficient,
meridional blade to blade duct height and shape factor. F is defined as:

π π  sp
F = 1 − 2 sin   sin  + ϕ2c  cos ϕ2c − (10)
Z Z  s2 cos ϕ2c

Due to the uncertainty in the slip correlation, csl is linearly varied according to impeller peripheral
speed:

(
csl = cslip + cslipu u2 − u2ref 1000 ) (11)

cslip , cslipu being two calibration constants, and u2ref being a reference blade velocity evaluated at the
rotational speed of 200000 rpm.

FRICTION LOSSES – Friction losses are due to the presence of viscous forces in the boundary
layer and are directly taken into account in the flow equations through the definition of a friction
coefficient, depending on the local absolute or relative velocity based Reynolds Number. Well
known Poiseuille, Blasius or Nikuradse expressions are employed in both inlet, volute and outlet
pipe. Concerning impeller, drag coefficient is evaluated separately for rotating surfaces and fixed
ones and it is corrected according to [46]. In the diffuser, the approach proposed in [26] is followed,
where the friction coefficient is computed as:

0.2
 1.8e5 
f = x fat 0.01 
 Re L  (12)
cL ρ
Lr = (r − r2 )
c
Re L = r (log spiral length)
µ cr

xfat being a tuning constant. The same multiplier is applied to Poiseuille, Blasius or Nikuradse
expressions.

HEAT TRANSFER – The knowledge of the friction coefficient allows to also estimate the heat
transfer coefficient, by applying the Reynold’s Analogy. Different expressions are reported in the
literature and the one here employed has been derived by Gnielinski [25]. In this way, the heat
transfer from the compressed gas is easily modeled once the wall temperature has been assigned.
The latter, in steady flow, is specified as a weighted mean (weight = 0.8) between the gas
temperature and the external ambient temperature. An additional constant multiplier, xhw, is
included in the expression of the heat transfer coefficient.

INLET AND INCIDENCE LOSSES – First of all, the absolute velocity c1 is computed in the end
section of the inlet pipe. The flow is assumed to non-isentropically expand from section Ωinl to Ω1.
Part of the produced kinetic energy, namely the aliquot crid (c12-cinl2)/2, is in fact assumed to be
dissipated at constant pressure, as a consequence of the impeller eye obstruction. The term crid is a
tuning constant of the model. The absolute and peripheral velocities at the impeller inlet so define
the flow angle ϕ1 characterizing the direction of the approaching flow. When the flow angle ϕ1
Page 11 of 25
differs from the blade angle ϕ1c, it is assumed that the kinetic energy associated with the flow
component normal to the blade direction w1n (Fig. 8) is completely dissipated at a constant pressure.

Radial direction
Blade direction ϕ1
φ1c φ1 > φ 1c : Low Flow Rate

u1
Flow direction c1 w1
w1n

Flow direction φ Radial direction


1
φ1c φ1 < φ1c : High Flow Rate

u1
Blade direction

c1 w1
w1n

Fig. 8 – Velocity triangles in off-design condition, for incidence losses calculation

Looking at the Fig. 8, the velocity component w1n can be easily determined as [28]:

 tan ϕ1c 
Low Flow: ϕ1 > ϕ1c w1n = u11 −  (13)
 tan ϕ1 

 1 1 
High Flow: ϕ1 < ϕ1c w1n = u1 −  (14)
 tan ϕ1 tan ϕ1c 

Under this hypothesis, the following relation between the total pressure at the exit section of the
inlet pipe p0w1 and the one at the inducer throat, p0w1t, can be written:

γ
w1n  γ − 1 2  γ −1
M w1n = cinc p0 w1t = p0 w1 1 − M w1n  (15)
γRT01  2 

cinc being the tuning constant affecting the total incidence loss.

RECIRCULATION FLOW – A recirculating flow from the impeller oulet occurs in gap between
the impeller and the case. It alters the total temperature at the inlet and the related work exchange.
The quantity hgap defined in table 1 defines the characteristic dimension of the recirculating passage.
The flow through this gap is computed as a function of the instantaneous pressure difference
downstream and upstream the rotor wheel.

ADDITIONAL IMPELLER LOSSES – Secondary loss mechanisms are considered in the model,
too. They are briefly listed in the following. A more detailed description can be found in [23].

Blade Loading Losses – A high blade loading level may produce flow separation and increased
losses. Impeller blade loading is evaluated according to the average relative velocity difference at

Page 12 of 25
mid-passage between pressure side and suction side, ∆w. It is computed from standard irrotational
flow relations, assuming an ideal or optimum blade loading style:

4πR2 c2u
∆w = (16)
ZLB

LB being the meanline 3D length. Although blade loading loss is distributed along the blade to blade
channel, a total pressure loss is only imposed at the impeller outlet. A blade loading related Mach
number is evaluated as:

∆w 2 6
M bl = cbl (17)
γRT02

cbl being a tuning constant.

Leakage Losses - Flow induced by blades tip clearance determines secondary flows and leakage
losses. Leakage flow rate, m& lk , can be computed as:

m& R2c2u 2∆plk


m& lk = ρ2vlk Zhgap Lm ∆plk = vlk = .816 (18)
ZRmbm Lm ρ2

where, Rm = (R1m + R2 ) / 2 , bm = (b1 + b2 ) / 2 , and Lm, m& , ∆plk, being the meridional meanline
length, the total flow rate, and pressure difference across the blade. A leakage loss related Mach
number is defined as:

2∆plk m& lk
ρ2 m&
M lk = clk (19)
γRT02

clk being a tuning constant.

Supercritical Mach Related Losses - Losses due to local sonic flow on the blade surface are
evaluated according to the local value of sonic velocity, wcr, at the blade mid-passage. A
supercritical loss related Mach number, Mscr, is computed by:

w + w2 + ∆w  w  M w1 − M w1cr wmax
wmax = 1 M w1cr = M w1 cr  M scr = cscr (20)
2  wmax  .4 γRT02

cscr being a tuning constant.


The above mechanisms define the overall pressure loss at impeller outlet with respect to the ‘ideal’
level reached after applying the sole slip effect, p02,sl:
γ
 γ −1 2
p02 = p02, sl 1 −
2
(
M bl + M lk2 + M scr )
2 p0 w2 p0 w1 − p1

p0 w1 ρ1 w12 2 
 γ −1
(21)

Page 13 of 25
Compressibility effects are also accounted for in eq.(21), as advised in [44].
VOLUTE LOSSES - The kinetic energy associated with the radial velocity component, c3r, at
volute inlet is assumed to be completely dissipated. The one associated to the tangential velocity
component, c3u, is indeed assumed to be partly lost at low flow rate according to [45]. The latter can
be expressed by applying a sudden enlargement mixing process at the volute inlet:
 Ω 4in 
c3u ,d = cvol c3u 1 −  (22)
 Ω 4extgϕ3

cvol being a tuning constant and Ω4in,Ω4ex and ϕ3, being inlet and outlet volute cross sections and
flow angle, respectively. Consequently, the pressure loss at the volute inlet can be estimated as:

γ
c32u , d + c32r  γ − 1 2  γ −1
M c 3d = ; p04 = p03 1 − M c 3d  (23)
γRT03  2 

TUNING OF THE LOSS COEFFICIENTS


The described steady and unsteady procedures are completely defined once the whole set of
geometrical data listed in table 1 has been accurately specified, together with the values of the 10
tuning constants: cslip, cslipu, xfat, xhw, crid, cinc, cbl, clk, cscr, and cvol. Due to difficulties to directly
measure the required geometrical data, some of them are only known with an uncertainty degree. In
other words, a possible error is introduced in the specification of some geometrical data, namely:
• hub and shroud inlet blade angle: ϕ1h,actual = ϕ1h + dϕ1 ϕ1s,actual = ϕ1s + dϕ1
• hub and shroud outlet blade angles: ϕ2h,actual = ϕ2h + dϕ2 ϕ2s,actual = ϕ2s + dϕ2
• blade height at the outlet bactual = b + db
• blade thickness sp,actual = sp + dsp
• recirculation gap hgap,actual = hgap + dhgap
The 5 uncertainties dϕ1, dϕ2, db, dsp, and dhgap can be hence considered as additional tuning
constant for the whole procedure. To identify the 5 geometrical uncertainties and the 10 unknown
constants in the loss correlations, an optimization procedure is carried out to minimize the overall
error between the experimental performance map and computed data. In particular, the following
objective function is built:

( )( )
Fobj ( x1 , x2 ,..., x15 ) = 1 + 2εβ 1 + ε η (1 + 0.5εT )

N map βimap − βicalc N map ηimap − ηicalc map


N map Tout calc
,i − Tout ,i (24)
εβ = ∑ εη = ∑ εT = ∑
i =1 βimap i =1 ηimap i =1 Timap

In these relations, βmap and βcalc are the map derived and computed total-to-total pressure ratio,
while η indicates the adiabatic efficiencies and Tout the total outlet temperatures. Pressure ratio,
efficiency and outlet temperature are linked together through the following relationship:

Page 14 of 25
 k −1 
 
Tamb  β k − 1
 
η=   β=
pout
(25)
Tout − Tamb pamb

These expressions, applied to map’s data, are inversely utilized to derive an experimental value of
the outlet temperature. In the model, indeed, the latter is directly computed at the beginning of the
outlet pipe (station 4) and the adiabatic efficiency is consequently estimated. It is worth-nothing to
stress that, due to the presence of the heat transfer, the above expression doesn’t really represent the
compressor efficiency. Anyway, efficiency levels usually reported on the compressor maps are
experimentally evaluated through the measurement of the inlet and outlet total temperature and
pressure. The three errors included in the objective function are, for this reason, consistent with the
available experimental data. They are estimated and summed up (i=1,Nmap) in correspondence with
each rotational speed and mass flow rate reported in the experimental map. In the objective function
definition, a greater weight is given to the pressure ratio and efficiency errors. The whole procedure
is applied with reference to the steady procedure to the three compressor of table 1.

Fig. 9 - Logic development of the optimization procedure

The logic development of the optimization process is shown in Fig. 9. The 15 independents
variables are iteratively modified by the optimizer and are written within the input files of the three
compressors A, B and C. Then the steady procedure, named ‘compstaz’, is executed and the output
file is processed to extract the three errors defining the objective function. The latter is finally
minimized with a genetic algorithm. Fig. 10 show the history chart of the optimization loop.
Starting from very high values, the objective function is rapidly reduced until it reach a constant
minimum level.

Page 15 of 25
Fig. 10 – History chart of the optimization procedure, B compressor

COMPUTATION OF THE STEADY MAPS


Table 2 summarizes the results of the optimization process. The identified values of the geometrical
uncertainties are of few degrees for the flow angles and around 0.1 mm for the other data. These
values are definitely included within the measurement errors of the related data.

Default Compressor
values A B C
Measurement errors
dϕ1 0 -0.8° -5.2° -2.80°
dϕ2 0 -3.0° -9.4° 1.7°
db, mm 0 -0.060 -0.141 -0.061
dsp, mm 0 -0.014 -0.025 -0.008
dhgap, mm 0 -0.081 -0.144 -0.149
Loss constants
cslip, cslipu 1, 0 1.62, 2.40 1.76, 2.49 0.93, 0.52
xfat 1 0.62 2.00 1.98
xhw 1 2.0 1.046 0.222
crid 0.5 0.685 0.625 0.655
cinc 1 1.06 0.65 0.19
cvol 0.5 0.425 0.235 0.095
cbl 1 0.01 0.54 1.90
clk 1 1.50 1.52 1.52
cscr 1 0.675 0.91 1.13
Results
εβ 3.37% 1.18% 2.36%
εη 6.11% 2.48% 4.95%
εT 1.06% 0.61% 1.18%
Fobj 84.29 15.25 61.91
Table 2 – Results of the optimization procedure
Page 16 of 25
Apart from the slip coefficient parameters, optimal values of the tuning constants do not greatly
differ from the ‘default’ unity or literature advised levels. In addition, differences among the three
compressor are in many cases reduced. This indicates the capability of the model in really capturing
the geometry related behavior of the 3 compressors. The global pressure ratio, efficiency and outlet
temperature errors are always limited within few percent. The numerical results, displayed in Fig.
11, show an excellent agreement with the experimental pressure ratio data for the three compressor
A, B and C, as well. Computed iso-speed lines (parameterized with compressor rotational speed X
10-3) follows with good accuracy the test points. Fig. 11 also reports the comparison of the total
temperature at the compressor discharge and the adiabatic efficiency. The agreement is very
satisfactory especially in the low and medium speed range. Only at the highest rotational speeds a
slightly less satisfactory matching is observable, especially on the outlet temperature, probably due
to the greater influence of the heat transfer. In these conditions, moreover, an operation closer to
sonic conditions at the compressor inlet occur, and some of the correlations employed may fail.
4.2 4.2 4.2
Exp Exp Exp Tref = 283 K
Calc Calc Calc Pref = 1.013 bar
3.8 3.8 3.8
220 240
Tref = 293 K Tref = 298 K
210
Pressure Ratio (total_to_total)

Pressure Ratio (total_to_total)

Pressure Ratio (total_to_total)


3.4 Pref = 1 bar 3.4 3.4
Pref = 1 bar
200 220
3.0 3.0 3.0
200 220
180
2.6 2.6 2.6 212
180 193
160
2.2 2.2 160 2.2
174
140
140 155
1.8 1.8 1.8
120 120 135
116
1.4 90 1.4 100 1.4 97
80

1.0 1.0 1.0


0 100 200 300 400 500 600 700 0 100 200 300 400 500 600 700 0 100 200 300 400 500 600 700
Mass Flow Rate, kg/h Mass Flow Rate, kg/h Mass Flow Rate, kg/h
550 550 550
220
Exp Exp Exp
Calc Calc Calc
210
240
500 500 500 220
200 Tref = 298 K
Tref = 293 K 220
Pref = 1 bar 212
Pref = 1 bar
Total Temperature, K

Total Temperature, K

Total Temperature, K

180
450 450 200 450 193
180
160
160 174
400 140 400 400
140 155
120 135
120
350 350 350 116
90 100
80 97 Tref = 283 K
Pref = 1.013 bar

300 300 300


0 100 200 300 400 500 600 700 0 100 200 300 400 500 600 700 0 100 200 300 400 500 600 700
Mass Flow Rate, kg/h Mass Flow Rate, kg/h Mass Flow Rate, kg/s
1.0 1.0 1.0

Exp
Calc
0.9 200 0.9 0.9
120 140 160 180
210 160 180 200
90 140 135
220 120 220 155
0.8 0.8 0.8 116 174 193
100 97
240

0.7 0.7 0.7


Eta_ad

Eta_ad

Eta_ad

80 212

0.6 0.6 0.6 220

0.5 0.5 0.5

Exp Exp
0.4 0.4 Calc
0.4
Tref = 293 K Calc Tref = 293 K Tref = 293 K
Pref = 1 bar Pref = 1 bar Pref = 1 bar
0.3 0.3 0.3
0 100 200 300 400 500 600 700 0 100 200 300 400 500 600 700 0 100 200 300 400 500 600 700
Mass Flow Rate, kg/h Mass Flow Rate, kg/h Mass Flow Rate, kg/h

Compressor A Compressor B Compressor C


Fig. 11 – Comparison of the computed maps with the experimental data with the steady procedure
Page 17 of 25
The results shown authorize to conclude that a good reproduction of the steady operating condition
has been obtained in whole compressor operating range, and the model can be successfully
employed to compute additional iso-speed lines, as shown in Fig. 12. In this way, the need to ‘un-
physically’ extrapolate maps’ data, usually realized in engine-turbocharger matching calculations, is
no more required. Fig. 12 also shows that the steady procedure, at low flow rates, is able to estimate
a wider compressor operating range with respect to experimental data points. It is interesting to note
the latter is always characterized by an increasing branch of the iso-speeds, determining, as known,
an unstable compressor behavior, and stall and/or surging occurrence. As expected, experimental
data point are mostly located on the decreasing iso-speed branch, and this indicates that the
procedure correctly predicts the extension of the stable operating domain, and the related stall limit.
The steady procedure may be also employed to compute other operating maps, with a different
upstream pressure and temperature, and composition. This may be the case in two-stage
turbocharging systems, with long-route EGR devices, where the compressor operates with a
variable mixture of air and recycled burned gases. While upstream pressure and temperature
variations are usually accounted for through the definition of a proper corrected mass flow
m& corr = m& T Tref p p ref rate and speed ncorr = n T Tref , gas composition effects are usually
neglected. Indeed, they can be easily evaluated in the proposed model, as shown in Fig. 13.
4.2 4.2
Exp Stall Line k=1.4, R=287 J/kg/K
Calc k=1.3, R=293 J/kg/K
3.8 230 3.8 230

Tref = 293 K 220 Tref = 293 K 220


Pressure Ratio (total_to_total)

Pressure Ratio (total_to_total)

3.4 Pref = 1 bar 210 3.4 Pref = 1 bar 210


200 200
3.0 3.0
190 190
180 180
2.6 2.6
170 170
160 160
2.2 2.2
150 150
140 140
1.8 130 1.8 130
120 120
110 110
100 100
1.4 90 1.4 90

50 50
1.0 1.0
0 100 200 300 400 500 600 700 0 100 200 300 400 500 600 700
Mass Flow Rate, kg/h Mass Flow Rate, kg/h

Fig. 12 – Extended map. A Compressor Fig. 13 – Influence of gas composition

EXAMPLE OF THE UNSTEADY OPERATION


The previously described approach is very useful for a fast map prediction, before carrying out
coupled engine-turbocharger calculations. In this case, as already pointed out, the intrinsically
unsteady engine operation induces unsteadiness on the compressor behavior, too. The latter, as
shown in [27], may induce a small delay in the compressor pulses, affecting the engine filling
conditions. In addition, as already pointed out, under low speed, high load engine conditions, an
unstable compressor operation can be obtained as a consequence of a close-to-stall average
compressor matching point.

In order to verify the capabilities of the model in capturing the described phenomena, the unsteady
model version is employed. A close-to-stall compressor operation was imposed by specifying a
constant and very high compressor speed (220.000 rpm) together with a high throttling of the valve
V located at the end of the outlet pipe in Fig. 1. Figure 14a shows the parabolic profile expressing
Page 18 of 25
the characteristic line of the throttle valve. Its intersection with the 220000 rpm iso-speed line is
located very close to the computed stall margin (box shaped point in Fig. 14a). In these conditions,
the unsteady procedure doesn’t converge on a stable point. Almost instantaneously, inlet and outlet
mass flow rates and outlet pressure begin to rapidly vary during time (Fig. 14b) reaching a periodic
behavior. Difference between inlet and outlet mass flow rates clearly highlights the presence of
mass (and energy) storage, mainly located within the volute volume. The final periodic solution,
plotted in Fig. 14a, can be interpreted as a soft-surging loop, whose frequency depends on the outlet
pipe volume and valve closure. In the analyzed case, mass flow oscillations always remain positive
and no backflow is occurring. However, this may be not the case for an increased closure of the
outlet valve. Unfortunately, in this last situation, the developed model fails because included loss
correlations are only valid for a positive flow. Work is under progress to also model backflows and
deep surging loops. The results presented allow to conclude that the unsteady behavior appearing in
the unstable operating region seems to be well captured by the model. It is worth-nothing to
underline that, similarly, the procedure is well suited to handle the unsteadiness related to upstream
and downstream pressure pulsations arising as a consequence of intake valve opening and closing,
or depending on rotational speed oscillations, due to an instantaneous torque unbalance with the
turbine.
4.2
Exp
Steady Model
3.8 Surging Loop 3.7 900
Outlet Pressure
Tref = 293 K 3.6
220 800
Pressure Ratio (total_to_total)

3.4 Pref = 1 bar 210


Outlet Total Pressure, bar

3.5

Mass Flow Rate, kg/h


External Valve 200 700
3.0 Characteristic
3.4
180 600
Mass Flow
2.6 3.3 Rate
500
160 3.2
2.2
400
140 3.1
1.8 Outlet 300
120 3
Inlet
1.4 90 2.9 200

0.3 0.35 0.4 0.45 0.5 0.55 0.6


1.0 Time, ms

0 100 200 300 400 500 600 700


Mass Flow Rate, kg/h (b)
(a)

Fig. 14 – Development of a surging loop on the compressor map (a) and during time (b).
Unsteady Model.

CONCLUSION
A one-dimensional model of a centrifugal compressor was described in detail with particular
reference to the geometrical description of the complex 3D shaped impeller wheel. Steady and
unsteady flow equations, boundary conditions and correlations for the simulation of various flow
losses and slip effect were illustrated. The model, basing on a reduced set of geometrical data easily
measured on the device, was applied to the simulation of a 3 backswept vanes compressors,
constituting typical turbocharger groups for internal combustion engines. The identification of
tuning constants and uncertain geometrical data was realized with the help of an optimization code.
The reduced variation of tuning constants indicates that the model reached a good generality level.

Page 19 of 25
Nevertheless, some aspects, especially concerning the modeling of the slip phenomenon, still
require a further deepening.

Once tuned, the numerical procedure was proven to furnish the steady performance maps of the 3
devices all over their operating region. Pressure ratio, outlet temperature and adiabatic efficiency
showed a very good agreement with the test data. The stable operating limit of the A compressor
was also estimated with good accuracy.

If utilized to compute the engine-turbocharger matching conditions, the proposed methodology was
proven to overcome some of the limitations related to the employment of the steady performance
map. The direct calculation of the performance map in fact allows avoiding the need of map’s
extrapolation and takes into account actual working fluid properties and composition.

The unsteady version of the model was then applied to simulate a close-to-stall compressor
operation. This allowed to describe the unstable operating regime, where soft-surge phenomena
may arise, particularly at low engine speed and high load. The ongoing inclusion of backflows in
the model will give the possibility to also handle deep surge phenomena.

The presented results authorize to further develop the research activity with the objective to include
the procedure within a whole 1D model of a turbocharged engine, for the simulation of both steady
and transient engine operation.

REFERENCES
1. N. Watson and M.S. Janota, “Turbocharging the Internal Combustion Engine”, John Wiley,
New York, 1982.
2. J.B. Heywood, “Internal Combustion Engine Fundamentals”, McGraw-Hill Int. editions, 1988
3. Z. Zhang, “Experimental Study on the Three-phase Sequential Turbocharging System with Two
Unequal Size Turbochargers”, SAE Paper 2008-01-1698, 2008.
4. N. Winkler, H. E. Angstrom, “Simulations and Measurements of a Two-Stage Turbocharged
Heavy-Duty Diesel Engine including EGR in Transient Operation”, SAE Paper 2008-01-0539,
2008.
5. R. Hasegawa, H. Yanagihara, “Hcci Combustion in DI Diesel Engine”, SAE Paper 2003-01-
0745, 2003
6. J. Stokes, T. H. Lake and R. J. Osborne, “A Gasoline Engine Concept for Improved Fuel
Economy –The Lean Boost System”, SAE paper 2000-01-2902.
7. Shahed M.R., “Engine Downsizing and Boosting for CO2 Emission Reduction”, presented at
International Vehicle Technology Symposium, March 11-13, 2003.
8. Police G., Diana S., Giglio V., Iorio B., Rispoli N., “Downsizing of SI Engines by Turbo-
Charging”, ESDA2006-95215, Proceedings of ESDA2006, 8th Biennial ASME Conference on
Engineering Systems Design and Analysis, July 4-7, 2006, Torino, Italy.
9. P. Leduc, B. Dubar, A. Ranini, and G. Monnier, “Downsizing of Gasoline Engines: an Efficient
Way to Reduce CO2 Emissions”, Oil and Gas Science and technology - Rev. IFP, Vol. 58
(2003), No. 1, pp. 115-127.
10. F. Bozza, A. Gimelli, L. Strazzullo, E. Torella, C. Cascone, “Steady-state and Transient
Operation Simulation of a ‘Downsized’ Turbocharged SI Engine”, SAE Paper 2007-01-0381
11. Fontana G., Galloni E., “Knock Resistance in a Small Turbocharged Spark-Ignition Engine”,
International Session – The Sustainable Mobility Challenge – of 61st ATI 2006 Congress, SAE
n. 2006-01-2995.
12. GT-Power, User’s Manual and Tutorial, GT-SUITETM Version 6.0, 2003, Gamma Techn.
Page 20 of 25
13. Wave, User’s Manual, 2002, Ricardo
14. Boost, User’s Manual, 2003, AVL
15. S. Guilain, “Modeling and Measurement of the Transient Response of a Turbocharged SI
Engine” SAE Paper 2005-01-0691,2005
16. M. Capobianco, S. Marelli, “Turbocharger Turbine Performance under steady and unsteady
flow: test bed analysis and correlation criteria”, 8th International Conference on Turbochargers
and Turbocharging, Inst.Mech.Engrs., London, 2006.
17. J. Macek, O. Vitek “Simulation of Pulsating Flow Unsteady Operation of a Turbocharger Radial
Turbine” SAE Paper 2008-01-0295, 2008.
18. Leonard O., Adam O., “A Quasi-One Dimensional CFD Model for Multistage Turbomachines”,
Journal of Thermal Science, vol. 17, No.1 (2008). Springer.
19. Harten, A., "On a Class of High Resolution Total Variation Stable Finite Difference Schemes",
Journal of Computational Physics, Vol. 21, 21-23, 1984.
20. Pelton R., J., ”One-Dimensional Radial Flow Turbomachinery Performance Modeling”, Master
Thesis, Science Department of Mechanical Engineering, Brigham Young University, 2007
21. Dixon, S. L., “Fluid Mechanics and Thermodynamics of Turbomachinery”, Fourth Edition,
Buttherworth-Heinemann, 1998, ISBN 0-7506-7059-2
22. Wiesner F., J., “A review of slip factors for centrifugal compressors”, Jrl. of Eng. Power, ASME
Transaction, 89, 558-72, 1967.
23. R. H. Aungier, “Mean Streamline Aerodynamic Performance Analysis of Centrifugal
Compressors”, Transaction of ASME 366 / Vol. 117, JULY 1995
24. Xuwen Qiu, Chanaka Mallikarachchi, and Mark Anderson “A New Slip Factor Model for Axial
and Radial Impellers” , ASME Paper GT2007-27064, 2007
25. V. Gnielinski “New equations for heat and mass transfer in turbulent pipe and channel”, Int.
Chemical engineering, 1976.
26. O. Dubitsky, D. Japikse, “Vaneless Diffuser Advanced Model”, ASME Transaction, Journal of
Turbomachinery, 011020, Vol. 130, 2008
27. Bozza F., Gimelli A.,“Unsteady 1D Simulation of a Turbocharger Compressor”, SAE paper
2009-01-0308, 2009. Also in SAE International Journal of Engines, 2, pp. 189-198, Print ISSN:
1946-3936, Online ISSN: 1946-3944, October 2009.
28. D. Japikse, “Turbomachinery Performance Modeling”, SAE Paper 2009-01-0307, Cliff Garrett
Turbomachinery and Applications Engineering Award, 2008.
29. F. Gu, A. Engeda, M. Cave, J. Di Liberti, “A Numerical Investigation on the Volute/Diffuser
Interaction due to the Axial Distortion at the Impeller Exit”, Transactions of ASME 475 / Vol
123, SEPTEMBER 2001.
30. W. M. König, D. K. Hennecke, L. Fottner, “Improved Blade Profile Loss and Deviation Angle
Models for Advanced Transonic Compressor Bladings: Part I – A Model for Subsonic Flow”,
Transactions of ASME 73 / Vol 118, JANUARY 1996.
31. E. Swain, “Improving a one-dimensional centrifugal compressor performance prediction
method”, Proc. IMech 653 / Vol. 219 Part A: J. Power and Energy, 2005.
32. T. W. von Backström, “A Unified Correlation for Slip Factor in Centrifugal Impellers”,
Transactions of ASME 1 / Vol 128, JANUARY 2006.
33. T. W. von Backström, “A compact equation for the prediction of eddy-induced slip in
centrifugal impellers”, Proc. IMech 911 / Vol. 220 Part A: J. Power and Energy, 2006.
34. J. A. Caridad, F. Kenyery, “Slip Factor for Centrifugal Impellers Under Single and Two-Phase
Flow Conditions”, Transactions of ASME 317 / Vol 127, MARCH 2005.
35. K. S. Paeng, M.K. Chung, “A new slip factor for centrifugal impellers”, Proc. Instn Mech Engrs
645 / Vol. 215 Part A: J. Power and Energy, 2001.

Page 21 of 25
36. G. Arnulfi, P. Giannattasio, C. Giusto, A.F. Massardo, D. Micheli, P. Pinamonti, “Multistage
Centrifugal Compressor Surge Analysis: Part I – Experimental Investigation”, Transactions of
ASME 305 / Vol 121, 1999.
37. G. Arnulfi, P. Giannattasio, C. Giusto, A.F. Massardo, D. Micheli, P. Pinamonti, “Multistage
Centrifugal Compressor Surge Analysis: Part II – Numerical Simulation and Dynamic Control
Parameters Evaluation”, Transactions of ASME 312 / Vol 121, APRIL 1999.
38. G. Arnulfi, P. Giannattasio, D. Micheli, P. Pinamonti, “An Innovative Device for Passive
Control of Surge in Industrial Compressor Systems”, ASME Transactions 473/Vol 123, 2001.
39. G. Arnulfi, F. Blanchini, P. Giannattasio, D. Micheli, P. Pinamonti, “Extensive study on the
control of centrifugal compressor surge”, Proc. Instn Mech Engrs 289 / Vol. 220 Part A: J.
Power and Energy, 2005
40. S. Mizuki, Y. Asaga, Y. Ono, H. Tsujita, “Investigation of Surge Behavior in a Micro
Centrifugal Compressor”, Journal of Thermal Science, 97 / Vol 15, No. 2, 2006.
41. J. Macek, O. Vitek, ”Simulation of Pulsating Flow Unsteady Operation of a Turbocharger
Radial Turbine”, SAE paper 2008-01-0295, 2008.
42. S. Cros, X. Carbonneau, “Computational study of the aerodynamic impact of stall margin
improvements in a high tip speed fan”, 8th European Conference on Turbomachinery, 23-27
March 2009, Graz, Austria
43. J. Andersen, F. Lindström, F. Westin, “Surge Definitions for Radial Compressors in
Automotive Turbochargers”, SAE paper 2008-01-0296, 2008.
44. R.H. Aungier, “Centrifugal Compressors a Strategy for Aerodynamic Design and Analysis”,
ASME Press, New York, 2000.
45. D. Japikse, “Advanced diffusion levels in turbocharger compressors and component matching”.
In IMechE Conference on Turbocharging and Turbochargers, 1982, paper C45/82, p. 143
(Mechanical Engineering Publications, London).
46. W. Traupel, “Die Theorie der Stroömung durch Radialmaschinen”, Braun, Karlsruhe, 1962

ACKNOWLEDGEMENTS
To AVL List Gmbh – Graz, Austria, for the financial support given in the course of a cooperation
with the DIME – University of Naples.

NOMENCLATURE
b Meridional blade height, vaneless diffuser thickness
C Vector of the conservative variables for stationary ducts
c Absolute velocity
cbl Tuning constant for blade loading losses
cinc Tuning constant for incidence losses at the impeller inlet
clk Tuning constant for leakage losses
crid Tuning constant for impeller eye obstruction losses
cscr Tuning constant for supercritical Mach losses
csl,cslip,cslipu Tuning constants for slip effect
cvol Tuning constant for incidence losses at the volute inlet
cw Absolute velocity component along the meanline in the impeller
D Vector of the conservative variables for vaneless diffuser
d Equivalent diameter
D Diameter
E Total energy per unit mass
Page 22 of 25
F Flux terms vector
F Shape factor
f Friction coefficient
H Total enthalpy per unit mass
hgap Recirculation gap
Lr Log spiral length
LB Meanline length
M Mach number
m Meridional curvilinear abscissa
m& Mass flow rate
p, ∆p Pressure, pressure difference
P Perimeter
q Rate of heat exchange
R Gas constant
r, R Radius
Re Reynolds number
S Vector of the source terms
s Impeller curvilinear abscissa
sp Blade thickness
s2 Span length at impeller outlet
T Temperature
u Tangential blade velocity
V Throttle valve
v Velocity
w, ∆w Relative velocity, relative velocity difference
W Vector of the conservative variables for rotating ducts
xfat Friction term multiplier
xhw Heat transfer term multiplier
Z Number of impeller blades

GREEKS
α Duct area variation term
β Compression ratio
γ Specific heats ratio, rotation angle
δ Radius variation term along a streamline
ε Specific error
η Efficiency
θ Constructive blade angle in meridional plane
θel Geometrical construction ellipse angle
µ Dynamic viscosity
ρ Density
σ Slip factor
ϕ’ Actual relative flow angle compared with meridional direction, positive if opposed to
rotational direction
ϕc, ϕ Constructive blade angle
Φ Flow coefficient
ω Angular velocity
Page 23 of 25
Ω Area of a duct section

SUBSCRIPTS
0 Total conditions, Intake duct inlet
1 Intake duct outlet, Inducer inlet
1t Inducer throat section
2 Impeller outlet, Diffuser inlet
3 Diffuser outlet, Volute inlet
4 Volute outlet
amb Ambient conditions
bl Blade loading
C Referred to stationary ducts
cr Referred to critic or sonic conditions
corr Corrected value according to reference one
D Referred to the vaneless diffuser
d Referred to dissipated quantity
ex Outlet
h Hub
i Impeller
in Inlet
inl Inlet duct
lk Leakage
m Meridional component, mean value
max Theoretical value at impeller mid-passage
n Velocity component normal to the blade direction
out Discharge conditions
r Radial velocity component, spatial derivative for vaneless diffuser
ref Reference value
s Spatial derivative for rotating ducts, Shroud
scr Supercritical Mach conditions
sl, slip Referred to slip
t Temporal derivative
u Tangential velocity component
x Spatial derivative for stationary ducts
W Referred to rotating ducts
w Referred to the relative motion

SUPERSCRIPTS
’ Effective flow condition (after slip)
cal Calculated value
map Experimental map data

CONTACT
Fabio Bozza (Full Professor), Vincenzo De Bellis (PhD Student)
(DIME). Università di Napoli “Federico II”.
Page 24 of 25
Via Claudio 21, 80125 Napoli (Italy).
Tel. +39 081 7683274 – 3264 - Fax: +39 081 2394165
E-mail: fabio.bozza@unina.it, vincenzo_debellis@virgilio.it

DEFINITIONS, ACRONYMS, ABBREVIATIONS


1D, 2D, 3D: One-, Two-, Three-Dimensional
CFD: Computational Fluid-Dynamic
EGR: Exhaust Gas Recirculation
HCCI: Homogeneous Charge Compression Ignition
ICE: Internal Combustion Engine
SI: Spark Ignition
TVD: Total Variation Diminishing

Page 25 of 25

Você também pode gostar