ISABE-2017-21369

Numerical prediction of stability limit in

centrifugal compressors with vaneless
diffuser
Cravero Carlo
[email protected]
University of Genoa
Dipartimento di Ingegneria Meccanica, Energetica, Gestionale e dei Trasporti
Genoa
Italy

Marsano Davide
[email protected]
University of Genoa
Dipartimento di Ingegneria Meccanica, Energetica, Gestionale e dei Trasporti
Genoa
Italy

ABSTRACT
The compressor operating range is a crucial performance parameter for the design and development of such a
component. During the design phase it is of utmost importance to be able to predict or estimate the limit mass
flow rate, for a given speed line, that will form the surge line and that will influence the overall operating range
and matching capabilities of the compressor. CFD simulations are routinely used for the compressor analysis
and strongly embedded into the design process. Nevertheless, the ability of a CFD model to accurately predict
the surge limit, without the need of a massively expensive full 3D unsteady simulation of the centrifugal
compressor stage, is still an open issue. Moreover, in a design process is very cumbersome to require series of
computationally intensive simulations for a 3D geometry that is still under development. In the present work,
the use of a CFD approach for the numerical prediction of the stability limit of centrifugal compressors is
presented. The proposed approach is applied to centrifugal stages with vaneless diffusers and the obtained surge
lines are compared to available experimental datasets. Different criteria for limiting mass flow are presented and
discussed.
Keywords: centrifugal compressors, surge prediction, CFD

NOMENCLATURE
b blade height
B span percentage of reverse radial velocity at diffuser inlet
ṁ mass flow rate
N rotational speed referred to design speed
r radius
S span percentage over critical flow angle
SP stability parameter
W velocity
y+ non dimensional boundary layer distance from wall

Symbols
α absolute flow angle
β static pressure ratio
γ metal angle

ISABE 2017
2 ISABE 2017-21369

Subscripts
1 impeller inlet
2 impeller outlet
4 diffuser outlet
cor corrected: referred to best efficiency
point
h hub
LE leading edge
t total
TE trailing edge

1.0 INTRODUCTION
The prediction of the unstable condition in a centrifugal compressor is an essential part of the design process. It
is common experience that a compressor often works in an acceptable manner with extensive separate flow
regions. The appearance of such areas in the blade cascades does not necessarily indicate the imminent unstable
operation. In centrifugal compressors, there are many controversial opinions about the causes and conditions for
stall or surge. Some believe that the rotating stall is the precursor of the surge; others say that the rotating stall,
especially in the case of vaneless diffusers, is not important. Part of this diatribe arises because the impeller
often has large separate or stalled flow regions even under operating conditions far from stall or surge. Another
confusion starts from singularly treat of the various compressor components, so in this case the instability is
determined by a single component. The reality is that the components are strongly coupled and therefore the
combined effect is fundamental. In order to better understand the physics of rotating stall, experimental and
numerical analyses have been conducted and published. With vaneless diffusers, experimental investigations
were carried out by Jansen [1], Abdelhamid [2], Frigne and Van Den Braembussche [3]. The stall in the vaned
diffuser has similar characteristics to that of the impeller. The stability of the component is conditioned by the
distortion of the flow along the span at impeller outlet. [4, 5]. Several approaches and parameters have been
studied to detect instability and predict the operating range of a centrifugal compressor. In vaned diffusers, the
semi-vaneless space is the more critical component [6]. In fact, it interacts with channel diffuser through
blockage of the boundary layer on the throat [7]. Greitzer [8] identified the stage dynamic stability limit as the
peak value for total-to-static pressure rise characteristic curve vs mass flow rate. This criterion was extended to
individual components of the centrifugal stage [9, 10] to locate the component that triggers instability. Instead,
compressors with vaneless diffuser, some studies have shown that the flow angle at the inlet diffuser is a
parameter commonly used to detect the lower limit of operating conditions [11, 12]. The CFD approach is
routinely used for compressor design, usually embedded into integrated design/analysis platforms [13, 14], and
is an efficient technology to support turbomachinery teaching concepts [15]. It is therefore important to develop
surge mass flow prediction methods that, using a standard CFD simulation with steady flow and mixing plane
approach, can support the compressor design phase. In a previous work on centrifugal compressors with vaned
diffusers [16] a CFD methodology was developed to predict the minimum mass flow for a given rotational
speed, by using a 3D CFD model to save a considerable amount of computational resources. The results, always
consistent with experimental data with a low error percentage, have shown the applicability of the approach. In
the present work, the same criterion is applied to compressors with vaneless diffuser. The proposed
methodology is based on the stability parameter SP, discussed by Cumpsty [17], following Dean’s original work
[18]. The above criterion applied to compressors with vaneless diffusers did not show the same accuracy,
because the volute causes a circumferential distortion of the pressure that spreads backward to the impeller [19,
20]. Alternative criteria to detect the surge limit from quick CFD steady analysis have been proposed and tested.

2.0 CFD APPROACH

For the numerical approach, a CFD procedure is set up using the ANSYS CFX platform. A simplified 3D
model, called “simple model”, is adopted to simulate a centrifugal compressor stage composed of: straight inlet
duct, single rotor channel and a single vaneless diffuser portion. The volute is not present. A periodicity
condition has been set for rotor and diffuser domains, while the rotor-stator interface is modelled with a mixing
plane. Steady simulations have been adopted with a SST turbulence enclosure and a second order solution
scheme. The following boundary conditions are set: total pressure, total temperature and a medium turbulence
intensity (5%) at the inlet and outlet constant mass flow rate condition (the set of simulation is performed far
from chocking). The simulation set is performed at decreasing mass flow rates near surge. Adiabatic viscous
walls are set for rotor and diffuser boundaries.
C. CRAVERO – D. MARSANO ISABE2017-21369 3

A multiblock structured mesh is generated with grid near wall y+ close to one with about two Mcells in total.
Figure 1 shows a mesh view in a blade-to-blade section.

Figure 1 Mesh details on blade to blade plane

Different rotational speeds have been considered to cover the compressor map. For a given constant speedline,
the mass flow rate is progressively decreased until numerical instability (oscillations in residual monitors)
occurs. The interesting property of the “simple model” is its inherent ability to predict a wider stability range
with respect to the real one; this helps to identify the unstable operating limit as described in the following.

3.0 APPLICATIONS
Two centrifugal compressors with vaneless diffuser have been investigated. To validate the proposed criteria
experimental data are used. The main feature of the compressors are shown in Table 1. Both have backward
rotor blades. C1 is a compressor for turbocharging applications whose data are proprietary and cannot be
provided. C2 is a compressor for propulsion applications from open literature [21]. Several studies have been
carried out on this compressor by Krain et al. [22], Eisenlohr et al. [23] and Hah et al. [24].
Table 1
Compressor dataset

Application r4/r2 b2/r2

C1 turbocharging 1.67 0.064
C2 propulsion 1.47 0.050

For compressor C2 the blade angle at trailing edge γ2 is 52 [deg], while the blade angle at leading edge tip γ1LE is
26.5 [deg]. The leading edge hub radius and tip radius are r1h =30 [mm] and r1LE = 78 [mm], while the impeller
tip radius r2 is 112 [mm]. The exit blade height b2ax is 10.2 [mm].
The first criterion discussed for limiting mass flow is the stability parameter previously developed for
compressor stages with vaned diffuser [16].
3.1 Stability parameter SP
The SP is based on the local performance curve slop to detect the stability limit by the slope change from
negative to positive, according to equation 1:

1 
SP  ( 1 )
 m
A set of simulations for each compressor has been performed with decreasing mass flow rate for three rotational
speed. The limiting mass flow is obtained when the condition SP=0 is reached.
In Figure 2 the obtained distributions for the SP at different corrected speeds N are shown and the obtained
surge lines are reported in Figure 3 and compared to the experimental data.
4 ISABE 2017-21369

1 0,4
0,6
0,5 0,2 0,7
0,8
0

SP
0
SP

0 0,2 0,4 0,6 0,8

-0,5 0 0,2 0,4 0,6 0,8 1
0,5
-0,2
-1 0,83
1
-1,5 ṁcor -0,4 ṁcor

Figure 2 Stability parameters distributions: (left) C1 – (right) C2

The interest is on the mass flow value while the pressure ratio value cannot be accurate due to the lack of the
volute in the CFD model. Figure 3 also show the percentage error of the predicted mass flow rate (histograms);
the error numerical values are shown in Table 2.
The accuracy found on the calculated limit mass flow may already be considered acceptable for a preliminary
design purpose taking into account that the above percentage is computed on small numbers of mass flow rates;
the actual error in kg/s is small.

Figure 3 Surge line prediction using stability parameter criterion: (left) C1 – (right) C2

The maximum error never exceeds 20 %, except for one speedline on C2. It was found that in the first
configuration C1 the errors made are lower than with C2.
Table 2
Percentage errors for the massflow rate on the surge line – stability parameter criterion

Ncor 0.5 0.83 1

C1
Error [%] 13.6 8.9 14.3
Ncor 0.6 0.7 0.8
C2
Error [%] 23.3 17.1 19.2

1
rotor
diffuser
0,5
SP

0
0,222 0,224 0,226 0,228 0,23 0,232 0,234 0,236 0,238 0,24
-0,5 ṁcor

Figure 4 SP for compressor C1 at N=0.5

The SP procedure can also be used to identify which component (rotor or diffuser) contributes more to
instability. As an example, in Figure 4, the local distribution of SP for compressor C1 is reported at the
normalized speed N=0.5. Similar results are obtained at different rotational speeds. The above diagram shows
that the diffuser is the most unstable component having always a positive SP. The rotor, with a negative SP at
higher mass flow, drives the stage into instability at mcor less than 0.236 when SP>0.
C. CRAVERO – D. MARSANO ISABE2017-21369 5

Even if the correct trend in surge line prediction and stability limit analysis can be obtained with the SP
criterion, the overall errors from 15% to 20% are considered too high with respect to 5%-10% detected in case
of vaned diffusers [16]. With the vaneless diffuser there is a stronger coupling with the volute circumferential
pressure distribution that can affect both diffuser and compressor stability with respect to what occurs with
vaned diffusers. For the above reasons other criteria have been considered and discussed in the following
paragraphs.
3.2 Critical flow angle
The attention is focused on the inlet diffuser zone that has been highlighted as one of the most critical parts in a
centrifugal compressor with vaneless diffuser [11, 25]. In fact, the inlet flow angle to the diffuser is a parameter
widely used to predict the vaneless rotating stall. A critical value of this angle, for surge limit, has been
proposed by Senoo-Kobayashi [26]; it depends on the shape of the diffuser (inlet height and radius) as in
equation (2):

( 2 )

In order to understand the relationship between the critical flow angle defined by Eq. 2 and the compressor
stability, the spanwise distributions of the absolute flow angle (circumferentially averaged) at the diffuser inlet
have been investigated. In Figure 5 two flow angle distributions at diffuser inlet are shown for compressor C1 at
N=1 reduced speed and for two corrected mass flow rates (0.65 left – 0.58 right)

Figure 5 Absolute flow angle distributions at diffuser inlet for compressor C1 at N=1; mcor=0.65 (left) and mcor=0.58 (right)

It can be observed that only from a certain value of span percentage (called S in the following) the absolute flow
angle exceeds the critical angle and toward the shroud surface. The value of S increases as the flowrate
decreases. The values of S obtained from the simulation campaign on C1 and C2 compressors at the
experimental surge massflow rate are reported in Table 3.
Table 3
Values of S obtained from CFD at the experimental surge massflow

Ncor 0.5 0.83 1

C1
S [%] 20.6 20.9 23.7
Ncor 0.6 0.7 0.8
C2
S [%] 23.1 22.1 21.6

It can be observed that the value of S corresponding to the surge massflow is always close to 20. The average
value of 22 has been selected and considered the critical value for stability limit (Scritic). The above value has
been used to predict the surge massflow value from the CFD simulation campaign. In Figure 6 the variation of S
with the mass flow rate for compressor C1 at N=0.5 is shown as an example. The critical mass flow rate is
therefore obtained by interpolating the above curve with the fixed value of S critic =22. Very similar trends are
obtained for the other conditions and compressor C2. It is interesting to note the monotonic increase of S by
reducing the flowrate.
6 ISABE 2017-21369

Figure 6 Variation of S with massflow rate - Case C1 at N=0.5

In Table 4 the percentage errors in surge mass flow estimation from the above approach with respect to the
experimental data are reported. The present criterion gives very encouraging results with maximum error around
7%.
Table 4
Percentage error in surge mass flow rate - modified critical flow angle criterion

Ncor 0.5 0.83 1

C1
Error [%] 2.5 1.3 7
Ncor 0.6 0.7 0.8
C2
Error [%] 7.2 0.5 5.4

3.3 Recirculations zone

Another criterion for the surge mass flow rate prediction has been introduced from the analysis of the flow
distribution at the diffuser inlet. When a compressor operates at near surge conditions, a recirculation zone is
present toward the shroud in the first part of the diffuser, mainly due to jet and wake phenomena from the
impeller flowpath [11, 27]. This recirculation zone is responsible for backward flow to the rotor, which causes
instability if its extension exceeds a certain limit. Therefore, it was decided to find the critical value of the
recirculation zone by plotting the radial velocity spanwise distribution at the diffuser inlet. In Figure 7 three
distributions at different flow rates of C1 are reported as an example.

Figure 7 Spanwise distributions for radial velocity at diffuser inlet. Case C1 at N=0.83

It can be observed that the reverse flow from radial velocity increases by reducing the massflow. The percentage
value of the span with reverse radial flow is identified as B. The values of B obtained from a CFD campaign at
the experimental surge mass flow rates are reported in Table 5.
Table 5
Values of span B from CFD at the surge mass flow rate

Ncor 0.5 0.83 1

C1
B [%] 8.7 6.4 7.4
Ncor 0.6 0.7 0.8
C2
B [%] 13.7 11 11.5

The values from Table 5 show that an average value of B for both compressors would be inappropriate; it was
therefore decided to consider an average value of B for each case and assume it as the critical value to detect the
surge mass flow (Bcritic=7.5 for C1 and Bcritic=12.0 for C2). As for the previous criterion based on the critical
flow angle, the CFD campaign has been used to plot the values of B versus mass flow rate.
C. CRAVERO – D. MARSANO ISABE2017-21369 7

In Figure 8 case C1 at N=0.5 is shown as an example. The critical mass flow rate is therefore obtained by
interpolating the above curve with the fixed value of B critic. Very similar trends are found for the other conditions
and compressor C2. It is interesting to note the monotonic increase of B by reducing the flowrate.

Figure 8 Variation of B with massflow rate - Case C1 at N=0.5

In Table 6 the percentage errors in surge mass flow estimation from the above approach with respect to the
experimental data are reported. A good accuracy is reported for case C1 while a percentage error around 10% is
shown for case C2. From the above results, it is evident that the analysis based on the parameter B associated to
the flow recirculation zone needs further investigations and cannot be a general rule without additional analysis
on different compressors. In this case, the geometrical parameters of the stage can have a direct effect.
Table 6
Percentage error in surge mass flow rate – recirculation zone criterion

Ncor 0.5 0.83 1

C1
Error [%] 2.5 2.4 0.2
Ncor 0.6 0.7 0.8
C2
Error [%] 8.5 10.2 9.4

3.4 Diffusion ratio

In this case, the attention is paid to the impeller. It was decided to consider the deceleration of the flow in
relative frame from the rotor inlet to the trailing edge section in terms of relative velocity ratio WTE/W1 .
In Table 7 the values of the above ratio obtained from the CFD simulation campaign at the surge mass flow
from the experimental data are reported.
Table 7
Values of WTE/W1 from CFD at experimental surge mass flow

Ncor 0.5 0.83 1

C1 WTE/W1 0.852 0.857 0.875
Ncor 0.6 0.7 0.8
C2 WTE/W1 0.797 0.807 0.82

It can be observed that, for a given configuration, very similar values of the ratio are obtained in the rotational
speed range. A different number has been used as critical reference value for the above compressors: 0.85 for
case C1 and 0.80 for case C2. In Figure 9 the variation of the diffusion ratio with the mass flow rate for
compressor C1 at N=0.5 (obtained from CFD) is shown as an example. The critical mass flow rate is calculated
by interpolating the above curve with the fixed value of WTE/W1 considered critical. Very similar trends are
obtained for the other conditions and compressor C2. As for the previous analysis and parameters, there is a
monotonic variation with the flowrate.

Figure 9 Variation of diffusion ratio with massflow rate - Case C1 at N=0.5

8 ISABE 2017-21369

In Table 8 the percentage errors in surge mass flow estimation from this method with respect to the
experimental values are reported. Very low errors are obtained in general. Nevertheless, as for the recirculation
zone criterion, the critical values of the diffusion ratio need to be tuned for a given compressor. This aspect
requires further investigations to correlate the critical value of the parameter to the geometrical data of the
compressor.
Table 8
Percentage error in surge mass flow rate – diffusion ratio criterion

Ncor 0.5 0.83 1

C1
Error [%] 1.2 3.8 9.9
Ncor 0.6 0.7 0.8
C2
Error [%] 3.3 0.8 6.6

4.0 CONCLUSIONS
In the paper, different criteria for surge mass flow prediction at the early phase of a centrifugal compressor with
vaneless diffuser design have been presented and discussed. Two different compressor stages have been used
and the results obtained from the surge prediction criteria have been systematically compared to the available
experimental data. The criterion based on the stability parameter developed for stages with vaned diffusers [16]
did not show the same level of accuracy in presence of the vaneless diffuser. The authors, based on the
experimental evidence from open literature, have proposed other criteria. The criterion based on the modified
version of the critical flow angle did show good accuracy and, according to the available data, allowed a single
value for the Scritic parameter for both compressors. The criteria based on the radial velocity reverse flow
extension at the diffuser inlet and the diffuser ratio did show good accuracy but needed the tuning of the critical
parameter value for each compressor. This aspect needs further investigations in order to correlate the critical
parameter value to the geometrical data of the compressor stage.

REFERENCES
[1] JANSEN W., Rotating stall in a vaneless diffuser, ASME Journal of Basic Engineering, Vol. 86, pp. 750-
758, 1964
[2] ABDELHAMID A.N., “Effects of vaneless diffuser geometry on flow instability in centrifugal compression
systems”, Canadian Aeronautics and Space Journal, No.3, pp. 259-266, 1983
[3] FRIGNE P., VAN DEN BRAEMBUSSCHE R., “Distinction between different types of impeller and
diffuser rotating stall in a centrifugal compressor with vaneless diffuser”, ASME Journal of Engineering for Gas
Turbines and Power, Vol. 106, pp. 468 – 474, 1984
[4] KRAIN H., “A study on centrifugal impeller and diffuser flow”, ASME J. Engineering for Power, Vol. 103,
No. 4, pp. 688-697, 1981
[5] LIU Y., LIU B., “Investigation of unsteady impeller-diffuser interaction in a transonic centrifugal
compressor stage”, ASME GT2010-22737, 2010
[6] KENNY D., “A comparison of the predicted and measured performance of high pressure ratio compressor
diffusers”, VKI Lecture Series 50 Advanced Radial Compressors, 1972
[7] RENEAU L., JOHNSTON J.P., KLINE S., “Performance and design of straight 2D diffusers”, J Basic
Engineering, Vol. 89, No 1, pp.1412-150, 1967
[8] GREITZER E.M., “The stability of pumping system–The freeman scholar lecture”, ASME Journal of Fluids
Engineering, Vol. 103, pp.193-241, 1981
[9] DEAN R.C., “The fluid dynamic design of advanced centrifugal compressors”, Lecture Notes Von Karman
Institute, Brussels, March 1974
[10] JAPIKSE D., “A critical evaluation of stall concepts for centrifugal compressors and pumps – Studies in
component performance”, ASME-Conference on stall and surge in compressors and pumps, part 7, 1984
[11] SENOO Y., KINOSHITA Y., “Influence of inlet flow conditions and geometries of centrifugal vaneless
diffusers on critical flow angle for reverse flow”, ASME J. Fluids Engineering, Vol. 99, 1977
C. CRAVERO – D. MARSANO ISABE2017-21369 9

[12] CLARKE C., MARECHALE R., ENGEDA A., CAVE M., “Investigation of centrifugal compressor
vaneless diffuser stability via a local flow angle approach”, Proc IMeche Part A: J Power and Energy, 0(0) 1-8,
2016
[13] CRAVERO C., “A design methodology for radial turbomachinery. Application to turbines and
compressors”, paper FEDSM2002-31335, The 2002 Joint US ASME-European Fluid Engineering Summer
Conference, Montreal, 14-18 July, 2002
[14] BARSI D., COSTA C., CRAVERO C., RICCI G., “Aerodynamic design of a centrifugal compressor stage
using an automatic optimization strategy”, ASME paper GT2014-26465, Proc. of ASME Turbo Expo 2012,
Dusseldorf, 16-20 June, 2014
[15] BRUNA D., CRAVERO C., TURNER M., MERCHANT A., “An educational software suite for teaching
design strategies for multistage axial-flow compressors”, ASME Paper GT2007-27160, Proc. of ASME Turbo
Expo 2007, Montreal, 14-17 May, 2007
[16] CARRETTA M., CRAVERO C., MARSANO D., “Numerical prediction of centrifugal compressor
stability limit”, ASME Paper GT2017-63352, Proc. of ASME Turbo Expo 2017, Charlotte USA, 26-30 June,
2017
[17] CUMPSTY N.A., “Compressor Aerodynamics”, Longman Scientific&Technical, Harlow (Essex), England,
1989
[18] DEAN R.C., “The fluid dynamic design of advanced centrifugal compressors”, Creare report TN-185,
Creare Inc., Hanover (New Hampshire), USA, 1974
[19] SIDERIS M., Van Den Braembussche R., “Evaluation of the flow in a vaneless diffuser followed by a
volute”, Proc 3rd Int. Conf. on Turbocharging and Turbochargers, I.M.E., 1986
[20] Sideris M., VAN DEN BRAEMBUSSCHE R., “Experimental study of the flow in a vaneless diffuser with
circumferential pressure distortion”, VKI TN 157, 1986
[21] KRAIN H., RICHTER F.A., TIEDE V., “Investigations of the flow through a high pressure ratio
centrifugal impeller”, Proc. of ASME Turbo Expo, 2002
[22] KRAIN H., HOFFMAN B., PAK H., “Aerodynamics of a Centrifugal Compressor Impeller with
Transonic Inlet Conditions”, ASME 95-GT-79, 1995
[23] EISENLOHR G., DALBERT P., KRAIN H., PROLL H., RICHTER F.A. ROHNE K.H., “Analysis of the
Transonic Flow at the Inlet of a High Pressure Ratio Centrifugal Impeller”, ASME 98-GT-24, 1998
[24] HAH C., KRAIN H., “Analysis of the Transonic Flow Fields inside a High Pressure Ratio Centrifugal
Compressor at Design and Off Design Conditions”, ASME 99-GT-446, 1999
[25] SENOO Y., NISHI M., “Prediction of flow separation in diffusers by boundary layer calculation”, Trans.
ASME, Journ. Fluids Engineering, 99(2), 1977
[26] SENOO Y., KINOSHITA Y., “Limits of rotating stall and stall in vaneless diffuser of centrifugal
compressors”, ASME P 78 GT 19, 1978
[27] BONAIUTI D., ARNONE A., HAH C., HAYAMI, “Development of Secondary Flow Field in a Low
Solidity Diffuser in a Transonic Centrifugal Compressor Stage”, ASME Turbo Expo 2002: GT2002-30371, 2002