Professional Documents
Culture Documents
turbine
Srithar Rajoo
a
, Alessandro Romagnoli
b
, Ricardo F. Martinez-Botas
b,
*
a
Transportation Research Alliance, Universiti Teknologi Malaysia, 81310 UTM Skudai Johor, Malaysia
b
Department of Mechanical Engineering, Imperial College London, Exhibition Road, London SW7 2AZ, UK
a r t i c l e i n f o
Article history:
Received 10 May 2011
Received in revised form
13 December 2011
Accepted 15 December 2011
Available online 17 January 2012
Keywords:
Turbochargers
Aerodynamic efciency
Unsteady ows
Turbines
a b s t r a c t
This paper discusses the details of unsteady experimentation and analysis of a twin-entry variable
geometry turbine for an automotive turbocharger. The turbine in this study is the product of design
progression from a commercial nozzleless unit to a single-entry variable geometry and consequently to
a twin-entry unit. The main features of the turbine were kept similar across all congurations for
equivalent comparison basis. The unsteady curves of the twin-entry turbine exhibited the conventional
looping characteristics representing lling and emptying effects, which was also the case for the
nozzleless and single-entry nozzled turbine. The swallowing capacity of the twin-entry turbine, during
full admission testing, was recorded to be inconsistent between the two entries, in particular they were
at different pressure ratio levels e the shroud end entry was in most cases more pressurized compared
to the hub end entry, as much as 13%. Contrarily, during out-of-phase testing the swallowing capacity of
both the turbine entries was found to be similar. The cycle-averaged efciency of the nozzled turbine
either twin or single-entry was found to depart signicantly from the equivalent quasi-steady, in
comparison to the nozzleless single-entry turbine, this was as much as 32%.
2011 Elsevier Ltd. All rights reserved.
1. Introduction
A turbocharger turbine operates under unsteady conditions due
to the pulsating nature of the exhaust gases. In consequence, twin-
entry turbines are generally designed and used for better energy
extraction from the pulsating exhaust gases. Twin-entry turbine
allows the pulsationinthe exhaust gas tobe sustained, byseparating
the banks of the manifold till it reaches the rotor/nozzle inlet. In
doing this, the higher isentropic energy in the pulse can made
available for the turbine to be extracted. However, pulsating ow
creates unique operating characteristics in a twin-entry turbine, in
comparison to a single-entry e this is further amplied with the
variable geometry congurations. The staggered pulsation in both
the entries, due to sequences in the exhaust valve opening, means
the turbine is operating in a combination mode of mostly unequal
and partial entry conditions. Thus, there is a need for experimental
worktounderstandthe unsteady-state performance of a twin-entry
variable geometry turbine in various operating conditions, in
comparisontothe single-entryandnozzleless unit. This is necessary
for the understanding of turbine behaviour under various operating
conditions of full, unequal and partial admissions.
Wallace and Blair [1] and Benson and Scrimshaw [2] presented
the earliest systematic study on the unsteady performance of
a radial turbine. These were followed by Wallace et al. [3], Miya-
shita et al. [4], Benson [5] and Kosuge et al. [6]. All these investi-
gations were limited by the inability of the instruments to
instantaneously measure all the relevant performance parameters.
Thus, only the static pressure was measured instantaneously, while
the mass ow rate, temperature, speed and torque were measured
as time-mean values. This remains similar for the investigation
presented by Capobianco et al. [7,8] and Capobianco and Gambar-
otta [9]. Research work by Capobianco et al. [7,8] concentrated on
establishing correlations between the unsteady parameter and the
equivalent steady values. However, as only the pressure measured
instantaneously, all comparative parameters were assessed
through quasi-steady approach. Winterbone et al. [10] and Win-
terbone and Pearson [11] presented a good review of the experi-
mental techniques and understanding of the pulsating ow
performance of a turbocharger turbine.
Dale and Watson [12] and Dale [13] were the earliest to present
an unsteady performance data for a turbine with all the parameters
measured instantaneously (except instantaneous temperature). A
radial twin-entry nozzleless turbine was used to investigate the
* Corresponding author. Tel.: 44 20 7594 7241; fax: 44 20 7823 8845.
E-mail addresses: srithar@fkm.utm.my (S. Rajoo), a.romagnoli@imperial.ac.uk
(A. Romagnoli), r.botas@imperial.ac.uk (R.F. Martinez-Botas).
Contents lists available at SciVerse ScienceDirect
Energy
j ournal homepage: www. el sevi er. com/ l ocat e/ energy
0360-5442/$ e see front matter 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.energy.2011.12.017
Energy 38 (2012) 176e189
performance characteristics under steady, unsteady, equal admis-
sion and partial admission conditions. Dale [13] presented results for
pulse frequency of 40 and 60 Hz, turbine speed 300e700 RPS,
a mean total inlet temperature of 400 K and a maximum instan-
taneous pressure ratio of 1.8. The turbine was coupled to an eddy-
current dynamometer with a loading capacity of 12 kW. Dale and
Watson [12] measured the instantaneous mass ow rate at the
turbine inlet with a hotwire anemometry and the instantaneous
inlet static pressure with a fast response pressure transducer. The
torque was deduced from the rotor rotational speed and accelera-
tion. Almost all researchers investigating the unsteady performance
of a radial/mixed ow turbine have used these measurement
techniques ever since. Based on the instantaneous measurement,
Dale and Watson [12] presented the typical hysteresis loop for
radial turbine unsteady performance curves. Nikpour [14] con-
ducted a study similar to Dale [13] but with a nozzleless turbine. A
hydraulic dynamometer was used to load the turbine with higher
maximum speed compared to Dale [13]. A hysteresis loop was also
recorded for the unsteady performance curve, but with greater
deviation from the equivalent steady, compared to Dale [13].
Baines et al. [15] further developed the work by Dale and Wat-
son [12] with the study of a twin-entry nozzleless radial turbine
under pulsating ow conditions. The measuring instruments
remain largely similar to Dale and Watson [12]. The turbine
performances were evaluated for pulse frequency of 20, 40 and
60 Hz, turbine speed of 300e500 RPS, a mean total inlet temper-
ature of 400 K and a maximum instantaneous pressure ratio of 1.8.
Baines et al. [15] also documented the out-of-phase pulsating ow
results, where dual-loops in the instantaneous efciency trace were
found. These were attributed to the effect of reverse ow from the
open valve limb to the close valve limb in an out-of-phase testing.
In overall, the similar hysteresis loops in the unsteady performance
curves were found as other researchers. The experimental facility in
the current study is an improved version from Dale and Watson
[12] and Baines et al. [15].
The more recent unsteady turbine performance evaluation was
given by Marelli and Capobianco [16], where they have utilized
largely similar measurement techniques as in the current study.
However, their investigation was focused on single-entry radial
turbine under waste gated conditions. In broad perspective, Marelli
and Capobianco [16] showed the signature unsteady curves and
quantied the turbine efciency depreciation in comparison to
quasi-steady assumptions.
This paper addresses the turbine unsteady performance under
thecombinationof twin-entryandvariablegeometrycongurations,
which has not been shown previously. The presented experimental
data and performance characteristics will serve as an essential vali-
dation tools for computation models to address the accurate turbo-
charger component assessments methods in an engine design
process.
2. Experimental facility
Fig. 1 illustrates the turbine congurations used for the current
study, which was designed in progression from nozzleless single-
entry to nozzled twin-entry. The turbine was designed based on
conventional turbomachinery techniques as described in Japikse
and Baines [17]. Table 1 gives the basic features of the nozzless and
the nozzled turbine in the current study, while the further design
details can be seen in Romagnoli et al. [18]. The experimental
facility available in Imperial College London is a simulated recip-
rocating engine test-bed for turbocharger testing. The facility has
the capability of conducting steady and unsteady ow testing of
single and twin-entry turbines. A schematic diagram of the turbine
test-rig is shown in Fig. 2, which includes the inlet and outlet
measurement locations for the turbine stage. The test-rig is
equipped with an eddy-current dynamometer, which enables
turbine testing within a large velocity ratio range [6,7]. The test-rig
is supplied by screw-type compressors, capable to delivering air up
to 1.2 kg/s mass ow rate at a maximum pressure of 5 bars (abso-
lute). The two separated streams of airow in each limb pass
through a rotary air pulse generator, which consists of two rotating
chopper plates as shown in Fig. 3. The chopper plates consist of
unique cut-out, which were originally designed by Dale and
Watson [12] to experimentally simulate the exhaust gas pulsation
of an engine as shown in Fig. 3. Avariable speed D.C. motor controls
the rotating frequency of the chopper plates, hence the frequency of
the pulsation. The phase of the entry pulses in both the limbs can be
varied either to be in-phase or out-of-phase by changing the rela-
tive position of the two chopper plates. The simulated exhaust gas
pulsation of the test-rig has been the centre of numerous industrial
related turbocharger turbine investigations, which reiterates its
relevancy as an accurate representative of an engine exhaust gas.
Some of the most recent ndings are documented by Costall et al.
(2011) [19] and Copeland et al. (2012) [20e22]. The focus of this
study, as with all the other related work in the past, is on charac-
terizing the turbine behaviour under a given pulse ow. The
understanding derived from the characterization can be an essen-
tial input towards the formulation of a generic assessment tool in
engine design process, for any forms of pulsating exhaust gases.
Nomenclature
A Area, mm
2
C Absolute velocity, m/s
D Rotor Diameter, m
MFP Mass ow parameter
N Turbine speed, rps
P Pressure, Pa
PR Pressure ratio
r radius, mm
T Temperature, K
Tr Torque, Nm
U Rotor tip velocity, m/s
_ m Mass ow rate, Kg/s
_
W Power, kW
h Efciency
q Phase Angle
j Azimuth Angle
C
is
isentropic velocity, m/s
Subscript
actual actual conditions
in inner limb (entry)
ins instantaneous conditions
isent isentropic
out outer limb (entry)
SE single-entry
TE twin-entry
tot total (inner outer limb entry)
ts total-to-static
0 Total conditions
1 Turbine stage inlet
5 Turbine stage exit
S. Rajoo et al. / Energy 38 (2012) 176e189 177
3. Unsteady testing and analysis procedures
The conditions for the unsteady test points are set up to corre-
spond to the steady-state velocity ratio at peak efciency for a given
speed. This is achieved by comparing the energy averaged velocity
ratio at unsteady ow condition, with the steady-state velocity
ratio at peak efciency. Once the desired condition is achieved and
the system stabilizes, measurements of pressures, temperatures,
mass owrate, speed and torque are acquired instantaneously with
reference to chopper plate frequency. A pulse signal produced once
per revolution by the chopper plate is used as trigger to concur-
rently measure all the parameters.
The instantaneously acquired data points are processed in few
stages before it can be used to calculate the performance parame-
ters. Hotwire anemometer used to measure mass ow rate is set to
traverse and measure at 36 locations in the pipe cross section.
Meanwhile the data logging system is set to acquire 50 cycles of
data for each location. Thus, a total of 1800 cycles are recorded for
every test point during unsteady experiments. For all but the hot-
wire measurement, the 1800 cycles are ensemble averaged to
reduce any random non-cyclic uctuation in the measurement.
Meanwhile the 36 points hotwire measurements are integrated as
specied in BS 1042:1983 and only 50 cycles are ensemble aver-
aged. The hotwire measurement is then corrected for temperature
before the nal mass ow rate can be deduced. The instantaneous
temperature of the ow is calculated based on the isentropic
compression assumption. The ensemble averaged signals are then
ltered with Finite Impulse Response (FIR) to further attenuate
noises in the reading.
All but the speed signals are recorded with an analogue-digital
card with constant sampling rate of 20,000 samples per second.
The speed signal is logged through a counter card, thus the
recorded number of samples per cycle are not constant and not in
match with the analogue readings. For the purpose of instanta-
neous point-by-point analysis, the speed signal is resample at
a constant 20,000 samples per second with spline interpolation.
The speed signal is then derived to determine the uctuating
torque of the rotor. Consequently the uctuating torque is used to
calculate the total power of the turbine. The difference in the
measurement locations of the actual and isentropic properties, as
shown in Fig. 2, requires a degree of phase shifting to enable
evaluation on common time frame. This is achieved by phase
shifting all the measurement to a common reference location
using bulk sonic ow velocity as described by Szymko et al. [23]
(see Fig. 2).
Fig. 1. Turbine volutes used for the current study.
Table 1
Nozzleless vs. nozzled turbine volute dimensions [18].
Nozzleless Nozzled
Volute tongue position 50
30
T
01
ins
_
P
01
ins
(1)
MFP
ins;TE
_ m
ins;tot
,
_ m
ins;in
_ m
ins;tot
T
01
ins;in
_ m
ins;out
_ m
ins;tot
T
01ins;out
0:5
_
P
01
ins;in
P
01
ins;out
_ (2)
The instantaneous pressure ratio is given in Eq. (3). In a twin-
entry turbine, the pressure ratio can be dened using an area
averaged pressure ratio between the two entries as given in Eq. (4).
PR
ins;SE
P
01
ins
P
5
ins
(3)
PR
ins;TE
1
2
_
P
01
ins;in
P
5
ins
P
01
ins;out
P
5
ins
_
(4)
4.2. Efciency vs. velocity ratio
The turbine instantaneous total-to-static efciency for a single-
entry turbine is given in Eq. (5). In the twin-entry cases instead,
since power is an extensive quantity, the instantaneous total-to-
static efciency can be calculated considering the sum of the
instantaneous isentropic power of the entries, as given in Eq. (6).
h
ts
ins;SE
_
W
actual
ins
_
W
isent
ins
(5)
h
ts
ins;TE
_
W
actual
ins
_
_
W
isent;in
_
W
isent;out
_
ins
(6)
The instantaneous velocity ratio is the ratio between the rotor
tip speed and the inlet isentropic velocity. The formulation of the
instantaneous velocity ratios for the single and twin-entry turbine
is given in Eq. (7) and Eq. (8) respectively.
_
U
C
is
_
ins;SE
pDN
ins
2
_
W
isent
ins
_ m
ins
(7)
_
U
C
is
_
ins;TE
pDN
ins
2
_
_
W
isent;in
_
W
isent;out
_
ins
_ m
ins;tot
(8)
5. Nozzleless and single-entry nozzled turbine: for
comparisons
The twin-entry turbine in the current study is the product of
progressive design stages from the commercial nozzleless followed
by single-entry nozzled turbine. Thus it would be benecial to look
into the unsteady characteristics of the preceding turbine congu-
rations, which will serve as a comparison basis to the twin-entry
variable geometry turbine performance in the following section.
Fig. 4 shows the plots of the turbine swallowing capacity with
nozzled single-entryandnozzleless volutes, for 40Hz and 60Hz ow
and 80% equivalent speed. The nozzled turbine loops shown are for
50
and 65
and 50
out-of-phase
conditions for various ow frequencies and speed are given in
Fig. 6. The inner and outer limb notations in the gure refer to the
individual inlet of the twin-entry turbine, shown in Fig. 2. The inlet
conditions are monitored throughout the experimental process to
ensure a consistent phase between the limbs, as this affects the post
processing for performance analysis.
Figs. 7 and 8 show the swallowing capacity of the twin-entry
turbine during in-phase full admission test at 50% and 80% equiv-
alent speed conditions respectively. In both the gures, results for
40 Hz and 60 Hz pulsating owconditions are shown, as well as the
equivalent quasi-steady curves for the partial admissions. The
swallowing capacity of the individual limbs is shown and the
overall swallowing capacity of the turbine calculated with mass
averaged inlet total temperatures and pressures. The common
looping curves are observed in all the cases, with some degree of
encapsulation around the quasi-steady curves; due to continuous
lling and emptying of the volume during pulsating ow condi-
tions. It can be seen that the outer limb swallowing capacity tends
towards steady characteristics during the 50% speed, but falls below
the equivalent quasi-steady. However during 80% speed condition,
- 2
- 1.6
- 1.2
- 0.8
- 0.4
0
0.4
0.8
1.2
1.6
2
0 90 180 270 360
Phase Angle (Deg)
40 Hz
60 Hz
Nozzleless
80%Speed
- 2
- 1.6
- 1.2
- 0.8
- 0.4
0
0.4
0.8
1.2
1.6
2
0 90 180 270 360
Phase Angle (Deg)
40deg
50deg
65deg
70deg
60Hz, 80%Speed
- 2
- 1.6
- 1.2
- 0.8
- 0.4
0
0.4
0.8
1.2
1.6
2
0 90 180 270 360
Phase Angle (Deg)
40deg
50deg
65deg
70deg
40Hz, 80%Speed
60deg
60Hz 60deg
a
b
c
E
f
f
i
c
i
e
n
c
y
(
t
-
s
)
E
f
f
i
c
i
e
n
c
y
(
t
-
s
)
E
f
f
i
c
i
e
n
c
y
(
t
-
s
)
Fig. 5. Comparison of the efciency curves between the nozzled single-entry and
nozzleless turbine for different nozzle vane angles, different ow frequencies and 80%
equivalent speed condition (nozzleless results from Szymko [24]).
Table 2
Energy weighed cycle-averaged values of velocity ratio and efciency for single-
entry nozzled and nozzleless turbine, 80% equivalent speed conditions (nozzleless
results from Szymko, [24]).
Frequency & conguration U/C
is
h
cycle-avg
h
quasi-steady
40 Hz 40deg 0.65 0.604 (2.8%) 0.576
40 Hz 50deg 0.662 0.607 (3.1%) 0.638
40 Hz 60deg 0.652 0.613 (12.6%) 0.739
40 Hz 65deg 0.645 0.603 (15.4%) 0.757
40 Hz 70deg 0.619 0.588 (14.6%) 0.734
40 Hz nozzleless 0.624 0.626 (4.6%) 0.672
60 Hz 40deg 0.676 0.657 (7.3%) 0.584
60 Hz 50deg 0.684 0.69 (4.2%) 0.648
60 Hz 60deg 0.682 0.686 (7.1%) 0.757
60 Hz 65deg 0.662 0.64 (13.2%) 0.772
60 Hz 70deg 0.62 0.556 (19.4%) 0.75
60 Hz nozzleless 0.653 0.691 (0.8%) 0.683
S. Rajoo et al. / Energy 38 (2012) 176e189 181
100
110
120
130
140
150
160
170
180
190
0 45 90 135 180 225 270 315 360
)
a
P
k
(
e
r
u
s
s
e
r
P
c
i
t
a
t
S
t
e
l
n
I
Phase Angle (deg)
Outer Limb
Inner Limb
40 Hz 50 % Speed 180 Out - of - Phase Admission
100
110
120
130
140
150
160
170
180
190
0 45 90 135 180 225 270 315 360
)
a
P
k
(
e
r
u
s
s
e
r
P
c
i
t
a
t
S
t
e
l
n
I
Phase Angle (deg)
Outer Limb
Inner Limb
60 Hz 50 % Speed 180 Out - of - Phase Admission
100
120
140
160
180
200
220
240
0 45 90 135 180 225 270 315 360
)
a
P
k
(
e
r
u
s
s
e
r
P
c
i
t
a
t
S
t
e
l
n
I
Phase Angle (deg)
Outer Limb
Inner Limb
40 Hz 80 % Speed In - Phase Full Admission
100
120
140
160
180
200
220
240
0 45 90 135 180 225 270 315 360
)
a
P
k
(
e
r
u
s
s
e
r
P
c
i
t
a
t
S
t
e
l
n
I
Phase Angle (deg)
Outer Limb
Inner Limb
60 Hz 80 % Speed In - Phase Full Admission
Fig. 6. Twin-entry turbine typical inlet static pressure proles for in-phase and 180
out-of-phase admission: given for individual entries, inner and outer limbs.
Fig. 7. Swallowing capacity of twin-entry turbine at 50% equivalent speed and in-
phase full admission: given for individual entries, inner and outer limbs as well as
mass averaged combination.
Fig. 8. Swallowing capacity of twin-entry turbine at 80% equivalent speed and in-
phase full admission: given for individual entries, inner and outer limbs as well as
mass averaged combination.
S. Rajoo et al. / Energy 38 (2012) 176e189 182
it shows clear looping characteristics around the steady curve.
Comparing the swallowing capacity between 50% and 80% speeds,
it can be seen that the turbine swallows more compared to the
equivalent steady at higher speed conditions. The cycle-averaged
unsteady mass ow parameter at 40 Hz 80% speed is around 38%
higher than the cycle-averaged steady condition. As for 60 Hz 80%
condition the number is approximately 60%.
Another observation from Figs. 7 and 8 is that the pressure ratio
range during a cycle in pulsating ow emulates the steady curves
more in the 80% compared to the 50% speed. Interesting as it is,
further scrutiny in comparing the two speed cases, one can notice
that for the 80% speed, the swallowing capacity exhibits plateauing
characteristics during the lling period, which is pronounced in the
40 Hz ow. This suggests the turbine is in a choking state, broadly
speaking beyond the pressure ratio of 1.8, which in some ways
explains the situation of the turbine experiencing high pressure
ratio and the strong lling and emptying behaviour in the 80%
speed cases. Furthermore, comparing the individual limbs under
unsteady ow partial admission, the swallowing capacity of the
outer limb is consistently at the higher pressure ratio, this is more
apparent in the 50% speed conditions. However this is not observed
in the steady owpartial admission operation. So the question here
is why the swallowing capacity of the individual limb under partial
admission is different between each other only in the unsteady
ow operation.
The swallowing capacity of a turbine is inuenced by factors
such as, the volute and rotor inlet area, centrifugal forces on the
ow, boundary layer growth and possible ow separations. The
experiment results suggest that one or a combination of these
factors is different between the limbs in the unsteady ow opera-
tion. Even though the current experimental results do not allow
a complete understanding of such a complex behaviour, one can
indicate that the ow has insufcient time to stabilize in the
unsteady ow, thus causing a departure from the steady as well as
a difference between limbs. In this respect, it can be said that ow
near the rotor shroud (fed mainly by the outer limb) has a higher
blockage, which is the result of an increase owcurvature and may
lead to a ow separation. Higher blockage leads to higher pressure
ratio operation as per the experimental observations. Overall, the
unsteady swallowing capacity curves at different conditions in
Figs. 7 and 8 show that the twin-entry turbine exhibits more of
a lling and emptying characteristics, consistent with the observed
characteristics of the single-entry nozzled turbine (Fig. 4).
Fig. 9 shows the swallowing capacity of the twin-entry turbine
under unequal 180
out-of-phase admission: given for individual entries, inner and outer limbs as well as
mass averaged combination.
S. Rajoo et al. / Energy 38 (2012) 176e189 183
6.3. Cycle averaging and comparison to quasi-steady
Given the uncertainty associated with the meaning of the
instantaneous turbine efciency, an alternative method to evaluate
energy conversion is by cycle averaging the parameters. The main
advantage of the cycle averaging method is that it is not affected by
phase shifting. The cycle-averaged efciency represents the ratio of
energy extracted by the turbine per pulse cycle divided by isen-
tropic energy owing into the system, as given in Eq. (9).
h
ts
energyavg
_
q
0
_
h
ts
_
Wt
isent
dt
_
q
0
_
Wt
isent
dt
_
q
0
_
_
Wt
act
dt
_
q
0
_
Wt
isent
dt
h
ts
cycleavg
(9)
The velocity ratio is calculated with energy weighed averaging
as shown in Eq. (10),
1
_
U
C
is
_
energyavg
_
q
0
_
U
C
is
t
,
_
Wt
isent
_
dt
_
q
0
_
Wt
isent
dt
(10)
The calculated energy weighed cycle-averaged velocity ratio
is then used to read the corresponding efciency from the steady
map for that given speed. The cycle-averaged efciency is then
compared with the equivalent quasi-steady efciency. By doing this
an appropriate comparison between parameters obtained at
different conditions can therefore be performed.
Table 3 reports a comparison between the cycle-averaged and
the quasi-steady efciency obtained for 60
out-of-phase admissions: given for individual entries, inner and outer limbs.
1
In a twin-entry conguration
_
W
isent
_
W
isent;in
_
W
isent;out
, while the velocity
ratio U/C
is
is equal to that provided in Eq. (8).
S. Rajoo et al. / Energy 38 (2012) 176e189 184
0.3
0.35
0.4
0.45
0.5
0.55
0.6
0.65
0.7
0.75
0.8
40 50 60 70 80
y
c
n
e
i
c
i
f
f
E
Frequency (Hz)
0.3
0.35
0.4
0.45
0.5
0.55
0.6
0.65
0.7
0.75
0.8
0.4 0.45 0.5 0.55 0.6 0.65
y
c
n
e
i
c
i
f
f
E
U/C
is
Out-of-Phase Flow
Cycle avrg eff - 27.9 revs/K^0.5 Quasi-steady - 27.9 revs/K^0.5
Cycle avrg eff - 43.0 revs/K^0.5 Quasi-steady - 43.0 revs/K^0.5
50% Equiv Speed 50% Equiv Speed
80% Equiv Speed 80% Equiv Speed
Fig. 12. Comparison cycle-averaged vs. quasi-steady efciency for the out-of-phase ow at 50% and 80 equivalent speed for 60
vane angle.
Table 3
Comparison of energy weighted cycle-averaged and quasi-steady efciency.
60
vane angle
Dh h
cycleavg
=h
quasisteady
1 100% Dh h
cycleavg
=h
quasisteady
1 100%
50% equivalent speed 80% equivalent speed
U/C
is
h
cycle-avg
h
quasi-steady
h U/C
is
h
cycle-avg
h
quasi-steady
h
In-phase
40 Hz 0.542 0.472 0.631 25.4% 0.640 0.835 0.784 6.5%
60 Hz 0.503 0.456 0.611 25.4% 0.596 0.616 0.768 19.8%
80 Hz 0.533 0.631 0.626 0.8% 0.629 0.823 0.781 5.4%
Out-of-phase
40Hz 0.587 0.434 0.647 32.9% 0.595 0.631 0.768 17.8%
60Hz 0.464 0.415 0.586 29.2% 0.587 0.520 0.763 31.8%
80Hz 0.567 0.526 0.641 17.9% 0.577 0.560 0.757 26.0%
0.3
0.4
0.5
0.6
0.7
0.8
0.9
40 50 60 70 80
y
c
n
e
i
c
i
f
f
E
Frequency (Hz)
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0.4 0.45 0.5 0.55 0.6 0.65 0.7
y
c
n
e
i
c
i
f
f
E
U/C
is
In-Phase Flow
Cycle avrg eff - 27.9 revs/K^0.5 Quasi-steady - 27.9 revs/K^0.5
Cycle avrg eff - 43.0 revs/K^0.5 Quasi-steady - 43.0 revs/K^0.5
50% Equiv Speed 50% E q uiv S p eed
80% Equiv Speed
80% Equiv Speed
Fig. 11. Comparison cycle-averaged vs. quasi-steady efciency for the in-phase ow at 50% and 80% equivalent speed for 60
vane angle.
S. Rajoo et al. / Energy 38 (2012) 176e189 185
was observed in the cycle-averaged efciency variation. In the out-
of-phase owconditions, the cycle-averaged efciency shows large
discrepancies from the quasi-steady e at 50% equivalent speed
a difference of 32.9%, 29.2% and 17.9% was measured at 40 Hz, 60 Hz
and 80 Hz respectively. A similar drop was also found at 80%
equivalent speed e the cycle-averaged efciency deviates from the
quasi-steady by 17.8%, 31.8% and 26.0% for the 40 Hz, 60 Hz and
80 Hz cases respectively.
Another interesting aspect to taken into account is the effect of
the pulsed owphases (in-phase or out-of-phase) on the efciency.
From Table 3 it can be seen that the out-of-phase condition is
detrimental to the overall efciency above all at high speed. At 80%
equivalent speed, the cycle-averaged efciency drops from 0.835,
0.616, and 0.823 down to 0.631, 0.520, and 0.560 for 40 Hz, 60 Hz
and 80 Hz respectively, from the in-phase to out-of-phase ow
condition e this corresponds to a drop of 32.2%, 18.4%, and 46.9%
respectively. However, the same decit is not observed at lower
speed where the efciency drop is much lesser. At 50% equivalent
speed, the efciency drops from 0.472, 0.456 and 0.631 to 0.434,
0.415 and 0.526 (corresponding to a drop of 8.75%, 9.8% and 19.96%)
for 40 Hz, 60 Hz and 80 Hz respectively, from the in-phase to out-
of-phase ow condition.
In order to further understand the correlation between the
cycle-averaged and the quasi-steady assumption, the ratios
between these efciencies has been calculated for both in-phase
and out-phase ow at 50% and 80% equivalent speeds. The
parameter called efciency ratio is plotted in Fig. 13 against pulse
frequency (40 Hz, 60 Hz, 80 Hz). As noted from Fig. 13, at 80%
equivalent speed, the efciency ratio for the in-phase owdoes not
vary substantially with frequency, which goes in favour of the
quasi-steady assumption. The dip observed at 60 Hz (efciency
ratio z0.8) can be considered as the result of the transition through
a region where the quasi-steady assumption no longer applies. In
the out-of-phase ow conditions instead the efciency ratio
remains below unity for both 50% and 80% equivalent speeds.
In order to understand the efciency trends, the unsteady time-
averagedpower andmass owwere calculatedandcompared tothe
quasi-steady values e these are shown in Fig. 14 for 80% speed. The
quasi-steady values are obtained from the steady curves at the
equivalent unsteady isentropic energy averaged velocity ratio.
The gure showshat the quasi-steady average values generally over
predict the corresponding unsteady values since the power and
mass ow ratio remain below unity. The mass owand power vary
in a consistent manner for both in-phase and out-of-phase ow
conditions e a decreasing value in the ratio observed with
increasing frequency. Despite the similarities existing between the
in-phase and out-of-phase power and mass ow ratios, Fig. 14
shows that in out-of-phase ow condition, the mass ow ratio
(and hence the isentropic power ratio) is approximately equal to
unity for almost any frequency, thus showing that the quasi-steady
assumption is adequate for a full unsteady calculation. The same
does not occur for the in-phase conditions for which a dip at 60 Hz
can be observed in the mass ow and isentropic power ratio. As
alreadyobservedbefore(Fig. 10) sucha dipcouldbe attributedtothe
increasingly unsteady correlations between the wave action and
lling and emptying effects in the nozzled turbine system. On the
efciency side, the trend observed for the ratios between the cycle-
averaged and the quasi-steady efciencies (Figs. 11 and 12) are re-
ected in the actual and isentropic power ratios e the isentropic
power ratios remains consistently above the actual power ratio.
Comparing the cycle-averaged efciency with the quasi-steady
assumption based on full admission conditions is not conclusive
when dealing with twin-entry turbines. It is a known fact that the
twin-entry turbine is meant to work in out-of-phase ow condi-
tions in most cases. The incoming pulses from each bank of
manifolds occur at staggered intervals, thus the turbine works in
partial admission conditions in most cases compared to full admis-
sion. Therefore, in order to evaluate the quasi-steady assumption in
the out-of-phase ow conditions, it is appropriate to refer to the
partial admission maps instead of the full admission. Table 4 shows
the quasi-steady efciency calculated using the full and partial
admission conditions. The cycle-averaged efciencies and the ratio
0.5
40 50 60 70 80
0.6
0.7
0.8
0.9
1
1.1
o
i
t
a
r
y
d
a
e
t
s
-
i
s
a
u
Q
/
y
d
a
e
t
s
n
U
Frequency (Hz)
Ratio between Cycle-averaged and Quasi-steady efficiency
In Phase 50% Equiv Speed
In Phase 80% Equiv spped
Out of phase 50% Equiv Speed
Out of phase 80% Equiv Speed
Fig. 13. Ratio between cycle-averaged and quasi-steady efciency for the in-phase and
out-of-phase ow at 50% and 80 equivalent speed.
0
1
2
40 60 80
o
i
t
a
r
y
d
a
e
t
s
-
i
s
a
u
Q
/
y
d
a
e
t
s
n
U
Frequency (Hz)
In Phase Flow
Wisnt ratio Wact ratio Mass flow ratio Wis ratio
0
1
2
40 60 80
o
i
t
a
r
y
d
a
e
t
s
-
i
s
a
u
Q
/
y
d
a
e
t
s
n
U
Frequency (Hz)
Out of Phase Flow
Fig. 14. Ratio cycle-averaged and quasi-steady power and mass ow rate for the in-phase and out-of-phase ow at 80% equivalent speed and 60
vane angle.
S. Rajoo et al. / Energy 38 (2012) 176e189 186
between the cycle-averaged and the quasi-steady efciencies are
shown in Fig. 15. From Fig. 15 it can be seen that the ratio between
the cycle-averaged and the quasi-steady assumption based on the
partial admission condition is much closer to unity compared to the
cases where the full admission map is used (see Fig. 13). At 80%
equivalent speed, the quasi-steady efciency goes from0.768, 0.763
and 0.757 to 0.530, 0.533 and 0.534 for the 40 Hz, 60 Hz and 80 Hz
cases respectively, considering the full to partial admission condi-
tion e such a drop leads to values of the efciency ratio (in partial
admission) to be 1.03, 0.85 and 0.92 for 40 Hz, 60 Hz and 80 Hz
respectively. At 50% equivalent speed, the ratio between the cycle-
averaged and quasi-steady efciency is approximately 10% higher
than that calculated considering the full admission curve, which
results in an improvement in the evaluation of the quasi-steady
assumption. In summary, the quasi-steady efciency shifts from
0.647, 0.586 and 0.641 to 0.607, 0.607 and 6.08 for the 40 Hz, 60 Hz
and 80 Hz cases respectively, considering the full admission to the
partial admission case.
Table 5 shows the velocity ratio, cycle-averaged efciency at 50%
equivalent speed for different vane angle settings (40
, 60
, 70
)
and ow frequencies (40 Hz, 60 Hz and 80 Hz) e these are plotted
in Fig. 16. The values are energy weighted average as shown in
Eq. (5) and Eq. (6). It can be noticed that the trend of the cycle-
averaged efciency for different vane angles does not seem to
follow a well-dened pattern. For the in-phase ow, the cycle-
averaged efciency shows about 12 percentage points drop from
40
to 60
vane angle
and in-phase ow 40 Hz and 60 Hz, the cycle-averaged efciency is
similar to that calculated for the 60
and
70
, while the 40
vane angle 60
vane angle 70
vane angle
U/C
is
h
cycle-avg
U/C
is
h
cycle-avg
U/C
is
h
cycle-avg
In-phase
40 Hz 0.593 0.593 0.542 0.472 0.51 0.470
60 Hz 0.548 0.590 0.503 0.456 0.541 0.463
80 Hz 0.519 0.347 0.533 0.631 0.487 0.288
Out-of-phase
40 Hz 0.556 0.425 0.587 0.434 0.472 0.261
60 Hz 0.524 0.344 0.464 0.412 0.478 0.311
80 Hz 0.587 0.433 0.567 0.526 0.485 0.349
Table 4
Comparison between cycle-averaged and quasi-steady efciency considering full and partial admission assumption for the quasi-steady value.
Out-of-phase ow (quasi-steady assumption based on partial admission condition)
50% equivalent speed 80% equivalent speed
U/C
is
h
cycle-avg
h
qs, full
h
ratio
h
qs,partial
h
ratio
U/C
is
h
cycle-avg
h
qs,full
h
ratio
h
qs, partial
h
ratio
40 Hz 0.587 0.434 0.647 0.67 0.530 0.81 0.595 0.631 0.768 0.82 0.607 1.03
60 Hz 0.464 0.415 0.586 0.70 0.533 0.77 0.587 0.520 0.763 0.68 0.607 0.85
80 Hz 0.567 0.526 0.641 0.82 0.534 0.98 0.577 0.560 0.757 0.73 0.608 0.92
0.5
0.6
0.7
0.8
0.9
1
1.1
4 0 5 0 6 0 7 0 8 0
o
i
t
a
r
y
d
a
e
t
s
-
i
s
a
u
Q
/
y
d
a
e
t
s
n
U
Frequency (Hz)
revs/K^0.5
revs/K^0.5
revs/K^0.5
revs/K^0.5
Full Admission
50% Equiv Speed
Full Admission
80% Equiv Speed
Partial Admission
50% Equiv Speed
Partial Admission
80% Equiv Speed
Fig. 15. Ratio of cycle-averaged and quasi-steady efciency for the out-of-phase owat
50% and 80% equivalent speed, obtained using the full and partial admission quasi-
steady value.
S. Rajoo et al. / Energy 38 (2012) 176e189 187
turbine. The experimental results build up well with previously
published works and they form a benchmark for future works.
Currently on-going work within the Imperial College London
turbocharger research group are focussing on 1D and 3D simula-
tion in the aim to discover more physical insights of the single and
multiple entries turbine behaviour under pulsating ow condi-
tions. Some of the initial ndings of the simulation work are
documented by Costall et al. [19], Copeland et al. [21] and Chiong
et al. [28]. The extensive unsteady experimental results of the
work in this paper enable the accurate validation of computational
tools in turbocharger turbine performance predictions, as shown
by Chiong et al. [28]. This is benecial for turbocharger and engine
designers, since most available turbine model such as shown by
Xiande Fang et al. [29] and Xiande Fang & Yu Xu [30], are still
based on steady-state ow.
7. Summaries
The current paper discusses performance analysis of a twin-
entry variable geometry turbine tested under pulsating ow
conditions. The presented experimental results are obtained from
a high precision pulsating ow turbocharger test-rig at Imperial
College London. The twin-entry turbine was designed in progres-
sion from single-entry nozzleless and nozzled turbines, while
maintaining the main geometrical features in all cases. These were
used as basis of comparison to gauge the operating variation and
the associated twin-entry turbine performance, under pulsating
ow conditions. Some initial results of the nozzleless and nozzled
single-entry turbine are also shown as to allow readers to follow
through the design progression. Following are the main ndings
from the experimental investigation of this paper.
For the optimum vane angle setting (60
) and in-phase ow
condition, the overall ow capacity is larger than the equivalent
quasi-steady. The encapsulation of the quasi-steady curve is only
partially achieved and it occurs mainly in the lowpressure region of
the maps for low speeds and frequencies. A more appreciable level
of encapsulation with the quasi-steady curve is achieved for the
ow capacity owning to each limb; a similar shape for the hyster-
esis loop could be found for the two entries even though the outer
entry (shroud-side) was found to operate at higher pressure ratios
than the inner entry (hub-side). As for the out-of-phase conditions,
the pulsating nature of the ow at staggered intervals, leads to
a higher rate of lling and empty, this could be observed in the
occurrence of large hysteresis loop amplitude.
The quasi-steady assumption for the turbine efciency under in-
phase ow condition found to be only partially true. At 50%
equivalent speed and lowfrequency the cycle-averaged efciency is
approximately 25% lower than the quasi-steady whereas at high
frequency the cycle-averaged and the quasi-steady efciency are
almost coincident (0.8% difference). Meanwhile at 80% equivalent
speed, the quasi-steady assumption is satised at low and high
frequencies (40 Hz and 80 Hz), whereas a transition region from
quasi-steady to fully unsteady was observed at 60 Hz.
The quasi-steady assumption for the turbine efciency under
out-of-phase ow condition was found to be unsatised e
a difference from18% to more than 30% was measured for both 50%
and 80% equivalent speed. The full admission quasi-steady ef-
ciency is not fully representative of the out-of-phase owcondition
in the turbine, which at each instant in time is more likely to act as
in partial admission conditions. At 80% equivalent speed, the ratio
between the cycle-averaged and the quasi-steady efciency devi-
ates only by few percentage points (1.03, 0.85 and 0.92 at 40 Hz,
60 Hz and 80 Hz respectively) while at 50% equivalent speed, an
improvement of almost 10 percentage points could be measured.
The ratio between the cycle-averaged and the quasi-steady
assumption passed from 0.67, 0.70 and 0.82 for the full admission
to 0.81, 0.77 and 0.98 for partial admission, at 40 Hz, 60 Hz and
80 Hz respectively.
The effects of vane angle on the cycle-averaged efciency
provided different response depending on the pulse ow. In-phase
ow showed that at lower frequencies the close vane position (60
and 70
, 60
and 70