Proceedings of GT2006

ASME Turbo Expo 2006: Power for Land, Sea

Proceedings and Air
of GT2006
ASME Turbo Expo 2006: Power for Land, Sea andSpain
May 8-11, 2006, Barcelona, Air
May 8-11, 2006, Barcelona, Spain

GT2006-91088

GT2006-91088

COMPARISON OF TRANSIENT MODELING TECHNIQUES FOR A MICRO TURBINE

ENGINE

Craig R. Davison A.M. Birk

Department of Mechanical and

Materials Engineering
Queen’s University
Kingston, Ontario, Canada

ABSTRACT program to simulate transient performance but did not consider

A large number of papers have been published on transient the effect of gas dynamics or heat soak. Others, such as Al-
modeling of large industrial and military gas turbines. Few, Hinai and Feliachi [4], used control theory to model micro
however, have examined micro turbines. The decrease in size turbines in power systems, which provide little physical insight
affects the relative rates of change of shaft speed, gas dynamics into the engine behaviour.
and heat soak. This paper compares the modeled transient It is important to examine the dynamics of a micro engine
effects of a micro turbojet engine comprised of a single stage of because a reduction in engine size changes the relative rates of
radial compression and a single stage of axial expansion, with a change of shaft speed, heating and gas dynamics, within the
diameter of 12cm. The model was validated with experimental engine. For larger engines it is expected that the shaft transient
data. model can run with larger time steps than the gas dynamic
Several forms of the model were produced starting with the models, but this changes with engine size because the
shaft and fuel transients. Conservation of mass, and then differential equations depend on the diameter to different
energy, was subsequently added for the compressor, combustor powers. Using the full dimensionless groups for engine work
and turbine, and a large inlet plenum that was part of the and mass flow, and the geometric relationships for polar
experimental apparatus. Heat soak to the engine body was moment of inertia and volume, the following set of
incorporated into both the shaft and energy models. Heat soak relationships were determined:
was considered in the compressor, combustor and turbine. W& ∝ d 2
Since the engine diameter appears in the differential
equations to different powers, the relative rates of change vary m& ∝ d 2
with diameter. The rate of change of shaft speed is very
strongly influenced. The responses of the different transient N ∝ d −1
effects are compared. The relative solution times are also V ∝ d3
discussed, since the relative size of the required time steps
changes when compared to a large engine. m ∝ d3
INTRODUCTION I ∝ d5
Much work has been done modeling the transient and Substituting these into the transient equations for shaft
dynamic responses of large gas turbine engines. Military speed, gas pressure, gas temperature and component
combat aircraft can spend up to 70% of their time operating in a temperature (Given by equations (6), (9), (12) and (23)
transient state[1] and have received much attention. Large respectively) yielded relationship (1), for change in shaft speed,
power generating turbines have also been modeled, such as the (2), for change in pressure, (3), for change in gas temperature,
work by Kim et al. [2]. Micro turbines have received less and (4), for change in engine component temperature.
attention. Chiang et al. [3] used a commercial modeling Applying the relations to the transient conservation of

1 Copyright © 2006 by ASME

Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 06/19/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 momentum equation yields relationship (5) and which is the PR Pressure ratio
rate of change of reduced mass flow. This indicates that, Pr Prandtl number
relative to each other, the effects of gas dynamics are the same R Ideal gas constant
regardless of engine size. These relationships are good as first Re Reynolds number
order approximations and give an indication of how different T Temperature
sized engines will behave comparatively. It is clear that the rate TR Temperature ratio
of shaft acceleration will be much higher for a micro turbine, V Component volume
compared to the gas dynamics and thermal heat soak, than for a &
W Power
larger engine. d Diameter
dΩ 1 h Enthalpy
∝ 2 (1)
dt d m Mass
m
& Mass flow
p Pressure
∂P 1 (2)
∝ t Time
∂t d γ Ratio of specific heats
µ Viscosity
∂T 1 ρ Density
∝ (3) Ω Rotational speed (rad/s)
∂t d
Subscripts
dTComponent 1 C Compressor
∝ (4)
dt d L Length
T Turbine
∂m& red 1 f Fuel
∝ (5) g Gas
∂t d l Losses
To investigate these relationships, transient models were o Stagnation (total) condition
produced of a small turbo jet engine, the ArtesJet JG-100. The r Radius
engine is 0.23m (9.1 inches) long and has a 0.113m (4.45 red Reduced conditions
inches) maximum diameter. Its maximum operating point is rel Relative to design conditions
defined below:
PR = 3.25
& = 0.27 kg s (0.60 lb/s)
m Station Numbering
F = 95N (21 lb) 0 Ambient
1 Inlet plenum inlet
N = 126kRPM
2 Compressor inlet
The transient modeling techniques examined and compared
3 Compressor outlet
included the following:
4 Combustor inlet
1. Shaft dynamics
5 Turbine inlet
2. Gas dynamics – mass conservation
6 Turbine outlet
3. Gas dynamics – energy conservation
ss quasi-steady state station
4. Thermal transients – heat transfer to engine body
MODELING TECHNIQUES
NOMENCLATURE
A Area
Shaft Transient
Cp Constant pressure specific heat
Shaft transient models account for the energy imbalance in
EGT Exhaust gas temperature
the shaft that causes the change in rotational speed. To
F Thrust
accurately simulate the engine they must also incorporate the
&
Q Heat flow transient behaviour of the control system and fuel flow. The
H Convective heat transfer coefficient shaft acceleration is given by equation (6). If the turbine is
I Moment of inertia producing more or less power than is being consumed the
L Length engine will accelerate or decelerate. The polar moment of
N Rotational speed (RPM) inertia was calculated based on the measured geometry of the
Nu Nusselt number shaft, compressor wheel and turbine rotor.

2 Copyright © 2006 by ASME

Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 06/19/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Bleed Air

Fuel

1 2ss Inlet 2 3ss 3 4 5ss 5 Turb. & 6ss Turb. & 6

Cmp. Combst.
Plenum Cmp. Combst. Prop.
Volume Volume Prop.
Inlet Volume Nozzle
Volume
Plenum
Station 0
(Ambient Conditions)

Figure 1: JG-100 transient engine schematic

dΩ W& C + W&T + W& l To account for the accumulation, volumes were included in
= (6)
the model, as shown in Figure 1. The transient conservation
dt ΩI Rotating Components equations for a component were based on its exit conditions.
This is known as the inter-component volume method. The
The fuel flow was modeled with equation (7). The If was steady state results were based on the inlet conditions to the
assumed to be a constant, representing the inertia of the fuel component and steady state calculations, such as performance
system. The change in flow rate was forced by the difference maps. The components were assumed to have no volume and,
between the fuel set point, based on the control voltage applied, therefore, no storage capacity and the exit values are referred to
and fuel flow. The control voltage was calculated with a PID as quasi-steady-state. Immediately following them is the
control system algorithm, and the control settings were volume of the component, which does nothing but store, in this
identical to those used on the experimental engine. To further case, the mass.
mimic the test bed, the speed and control voltage were digitized The conservation of mass in the volume is given by
to match the resolution of the 8 bit data acquisition and control equation (8) [5]. Applying the ideal gas assumption and the
board. inter-component volume method then yields equation (9).
Further, the small Mach number assumption allows the
dm& f m& f set − m& f
= (7) stagnation pressure and temperature to be used in place of the
dt If static values, which will not significantly affect the models’
results [6]. This was important since it allowed a direct
The transient solver yielded the rotational speed and the solution from the information provided by the transient solver.
fuel flow, which, combined with the inlet conditions, defined If the static values were used, the solution would become
the engine operation. A modified version of the aero-thermal iterative and significantly slow down the calculations.
steady-state solution then calculated the engine operating point
{Davison, 2004 79 /id}. ∂ρ m& in − m& out
= (8)
This procedure required only two transient equations, but ∂t V
needed an iterative solution of the steady state equations for
every solution of the transient equations. However, if the ∂p (m& ss − m& )RTss
transient equations could define the operation of each = (9)
component the need for an iterative steady state solution would ∂t V
be eliminated, resulting in more rapid solutions for the same
time step size. With the small Mach number assumption equation (9)
yielded the differential equation (10) for the inlet plenum.
Gas dynamic transients Similar equations were applied to each component and solved
The next stage of modeling incorporated the conservation for the outlet pressure of each volume. Since only the
of mass. During steady state operation, the mass flow into a conservation of mass equations were being applied in this case,
component equals the flow out, but under transient conditions the transient and steady state stagnation pressures and
the density in a component changes and mass either temperatures were equivalent. This version will be referred to
accumulates or is ejected, resulting in a different inlet and as the mass model.
outlet flow rate.

3 Copyright © 2006 by ASME

Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 06/19/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 ∂po 2 (m& 2 ss − m& 2 )R2To 2 ss (10)
applied to the combustor. This provided some idea of the
= effects of heat soak, but did not accurately model the heat
∂t VInPlen transfer, which not only required a more applicable convective
equation, but should account for radiation as well.
The second gas dynamic transient implemented was energy
conservation. The transient energy equation is given by
equation (11), which, after applying ideal gas assumptions and
Nu = 0.0201B Re 0.8 (14)
inter-component volume method, yields equation (12) [7]. The
energy equation for the inlet plenum, after applying the low m& L
Mach number assumption, is given by (13). This version will
Re = (15)
µA
be referred to as the energy model.

∂ (ρh − p ) (m& h )in − (m& h )out Stator Rotor

= (11) Stator Outer Rotor Outer
∂t V Casing (2) Casing (7)

∂T RTss (γ ss m& ss Tss + Tm& − γm& T − Tm& ss ) (12)

=
∂t p ssV Stator
Rotor Blade
(6)
Blade (3)
∂To 2 R2To 2 ss (γ o 2 ss m& 2 ss To 2 ss + Tm& 2 − γ o 2 m& 2To 2 − To 2 m& 2 ss )
= (13) Rotor
∂t p o 2 ssV InPlen
Rim (5)
Stator
The conservation of momentum equations were not Hub (1) Rotor
incorporated into the model. Incorporating the momentum
Diaphragm
equations could allow the simulation of compressor surge, but
(4)
this was beyond the scope of this work, and the time scales
with the mass and energy equations were already well below Figure 2: Heat transfer components of turbine rotor and
what would practically be measured on a micro engine. As the stator
engine size decreases the transient time scale decreases
More detailed equations were available for the turbine. It
requiring higher speed equipment and higher cost for data
was broken into seven sections as shown in Figure 2. The
acquisition. However, the cost of the engine is also decreasing
sections of the stator outer casing, stator hub and rotor rim
with size, reducing the incentive for expensive investment.
exposed to the flow were similar to a flat plate, and the
appropriate Nusselt number is given by equation (16) for
Thermal Transients
laminar flow and by (17) for mixed flow [10].
Thermal transient analysis accounts for the heat transfer
Pilidis and Maccallum [11] present equations for turbine
between the gas flowing through the engine and the engine
component Nusselt numbers. The stator and rotor blade
body. This changes both the temperature and pressure of the
Nusselt numbers were calculated with equation (18), which is
gas at the outlet of the component and provides the temperature
specifically for turbine blades. The rotor outer casing was
of the engine. The thermal transient equations were added to
treated as a cylinder in which another cylinder is rotating and
both the shaft and the energy models. The convective
the Nusselt number was calculated with equation (19). The
correlations and transient heat transfer equations were the same
inner diameter refers to the turbine tips and the outer to the
in both models.
inner surface of the rotor outer casing. Finally, the wheel
Thermal effects on gas turbine transient behaviour are
diaphragm was treated as a rotating face, and the Nusselt
often ignored due to lack of information on the heat transfer
number is given by equation (20).
within the engine [8]. For this engine, nothing could be found
for micro engines or radial compressors. The correlations
chosen, however, provided at least an order of magnitude Nu L = 0.664 Re1L/ 2 Pr1 / 3 (16)
result, and allowed insight into the importance of heat soak on
the engine behaviour.
Stamatis et al. [9] suggest equation (14), with equation (15)
( )
Nu L = 0.037 Re 4L / 5 − 871 Pr1 / 3 (17)
for Reynolds number for the compressor. Although they
applied it separately to the rotor and housing, for the JG-100 Nu = 0.235 Re 0.64 (18)
the value of the constant B was unknown, so both were
grouped together into one equation. The same equation was

4 Copyright © 2006 by ASME

Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 06/19/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

  (d d ) 
Nu = 0.0151 + 2.3 outer inner (d outer d inner ) Re 0.8 Pr 0.33 (19)
0.45
The addition of the gas dynamic terms had no effect on the
L result at this time scale. Therefore, the shaft and thermal
 
results were compared to the experimental results and the
additional models subsequently compared to the shaft to
Nu r = 0.0253 Re 0r .8 (20) determine the effect of adding the additional dynamic
equations.
The heat transfer changed the temperature at the outlet of Figure 3 compares the shaft speed for the engine as it
the component. The enthalpy transferred to the engine body accelerated from 118.2kRPM to 126kRPM. The model slightly
was subtracted from the steady state enthalpy, calculated from overestimated the speed through a large portion of the
performance maps or combustion. In the compressor and acceleration, but the correlation was good. The correspondence
turbine the change in temperature also affects the pressure ratio between modeled and experimental compressor delivery
at the exit of the component. The pressure ratio from the pressure (CDP) was very similar to that for the shaft speed.
performance map was modified by equation (21), to obtain the Figure 4 compares the EGT. The correlation in this case
actual pressure ratio [9]. The pressure and temperature ratios was not as good as for shaft speed and CDP, in part due to the
from the map refer to the conditions if no cooling was taking heat soak of the engine, this will be discussed later. Another
place. The component work was calculated using the efficiency contributing factor was the transient response time of the
from the performance map, and the outlet enthalpy was thermocouple, which was not modeled. This partly explains
obtained from equation (22). When the thermal transients were why the modeled temperature was higher than that measured on
included with the energy model the resulting exit enthalpy was the engine. The higher modeled fuel flow, shown in Figure 5,
used to calculate the quasi-steady state temperature used in the also contributes to the higher EGT.
energy transient equation (12). Figure 6 compares the modeled and experimental shaft
speed as the engine decelerated from 126.0kRPM to
102.5kRPM. In this case the match was very good. The
PR = TR x difference in the simulation and run 1 was less than that
ln PRMap ln TRMap (21) between run 1 and 2. A comparison of CDP again revealed a
x= correspondence very similar to that for the shaft. The EGT
1 − Q& W&
Heat Component comparison is given in Figure 7 and, as with acceleration, the
measured temperature lagged behind the modeled. The
modeled fuel flow in this case, in Figure 8, is very close to the
(Q& Heat − W& Component ) experimental.
ho out = ho in + (22)
m& The shaft transient model predicted the trends and behavior
of the engine during acceleration and deceleration. The most
The transient equations produced the temperatures of the likely cause of the differences was the fuel control system
engine components. One equation was required for each model, which was the least understood and involved a
component transferring heat, resulting in a set of 9 thermal combination of motor, pump, piping and electronic controls.
transient equations of the form of equation (23). The area, in
the equation, was that available for heat transfer, and the mass 128
was the total amount of the component being heated. The 127

model assumed that the conduction through the component was 126

fast compared to the convection, and, therefore, the component 125

temperature was uniform. 124

N (kRPM)

123

dTComponent HAComponent (TGas − TComponent ) 122

= (23)
121

dt CpComponent mComponent 120

119
118
VALIDATION 0 5 10 15 20 25 30 35 40

An experimental program was carried out to validate the Time (s)

engine model and details on the test rig are given by Davison Set point Op. Engine 1 Op. Engine 2 Shaft Thermal

and Birk [12]. The shaft transient formed the basis of all the
other transient solutions, and the addition of gas dynamic Figure 3: Comparison of shaft speed for 2 experimental
transients had little effect on the overall engine response. The engine runs with shaft and thermal transient models, for
addition of thermal transients had a slight effect on the shaft acceleration from 118.2 to 126.0 kRPM
model, and continued to affect the result over relatively long
time duration.

5 Copyright © 2006 by ASME

Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 06/19/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 1.06 1.02

1.05 1.00

1.04 0.98
To6/T o6 t=0

To6/T o6 t=0
1.03 0.96

1.02 0.94

1.01 0.92

1.00 0.90

0.99 0.88
0 5 10 15 20 25 30 35 40 0 5 10 15 20
Time (s) Time (s)

Op. Engine 1 Op. Engine 2 Shaft Thermal Op. Engine 1 Op. Engine 2 Shaft Thermal

Figure 4: Comparison of To6 for 2 experimental engine runs Figure 7: Comparison of To6 for 2 experimental engine runs
with shaft and thermal transient model, for acceleration with shaft transient model, for deceleration from 126.0 to
from 118.2 to 126.0 kRPM 102.5 kRPM
1.19 1.05
1.17 1.00
1.15 0.95
1.13
0.90
mfuel /mfuel t=0
mfuel /m fuel t=0

1.11
0.85
1.09
0.80
1.07
0.75
1.05
1.03 0.70

1.01 0.65

0.99 0.60
0 5 10 15 20 25 30 35 40 0 5 10 15 20
Time (s) Time (s)

Op. Engine 1 Op Engine 2 Shaft Thermal Op. Engine 1 Op Engine 2 Shaft Thermal

Figure 5: Comparison of fuel consumption for 2 Figure 8: Comparison of fuel consumption for 2
experimental engine runs with shaft and thermal transient experimental engine runs with shaft transient model, for
model, for acceleration from 118.2 to 126.0 kRPM deceleration from 126.0 to 102.5 kRPM
130
COMPARISON OF MODELS
125
The optimal time step size for each technique is given in
120 Table 1. The optimal time step was chosen based on the error
compared to the previous time step and the time required to run
N (kRPM)

115 the simulation. The time steps were varied by a factor of 10.
110
Table 1 is based on a test case defined as follows:
Nstart=118.2 kRPM
105 Nend=126.0 kRPM (becomes set point at t=0.99s)
Tamb=293.0 K
100
0 5 10 15 20
Pamb=101.325 kPa
Time (s) AH=0
Set point Op. Engine 1 Op. Engine 2 Shaft Thermal Processor: P3 - 800MHz

Figure 6: Comparison of shaft speed for 2 experimental The error was defined as the difference in time to reach
engine runs with shaft transient model, for deceleration 120 kRPM for the particular step size compared to the case
from 126.0 to 102.5 kRPM with the smallest step. For the shaft transient this resulted in
Figure 9. A 1ms step size appeared to be optimal. The run
time increased very sharply when moving to a 0.1ms step size,
and the error increased sharply when moving up to the 10ms

6 Copyright © 2006 by ASME

Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 06/19/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 step size. Similar trends were seen for the other models. Table the solution time and made convergence more difficult.
1 compares each of the models required step size, errors and However, the effects of heat soak did alter the operating point.
running time. The energy-thermal model was interesting. It Figure 3 compares the shaft speed from the shaft and
had a 0.1ms step size chosen, not because of a significant thermal models. The difference was minor, and if this was the
decrease in error, but due to a slight decrease in run time only parameter of interest the extra run time and complexity
compared to 1ms step size. would not be justified. The CDP had a greater change and the
Due to the small step size required for the shaft transient, EGT, in Figure 4, was affected to an even greater extent. In
the mass transient provided the fastest solution. Although the this case the engine casing was absorbing heat as the engine
same numbers of steps were required, the mass transient accelerated, resulting in a lower EGT. The effect of heat soak
solution provided the outlet pressure of each component; brought the EGT closer to the experimental result, but the
therefore, a solution could be obtained directly from the effect of the thermocouple transient still played an, un-
information provided by the transient results. The shaft modeled, role.
transient, however, required an iterative solution for each time In addition to the gas path operating point, the thermal
step to determine the operating point. simulation provided engine component temperatures. Figure
1.0 0.10 10 shows the compressor temperature. As expected, the
0.9 0.09 component temperature lagged behind the average gas
temperature, but eventually caught up. Figure 11 is the turbine
(s from reaching 120 kRPM)
0.8 0.08
rotor components’ temperatures. The blade temperature
(Fraction of slowest)

0.7 0.07
Relative Run Time

0.6 0.06 approached the gas temperature the fastest. This was expected,
since the blades are similar to fins in the flow, allowing for
Error

0.5 0.05
0.4 0.04 enhanced heat transfer.
0.3 0.03 Figure 12 examines the later part of the deceleration from
0.2 0.02 126.0 to 102.5kRPM. It shows the turbine rotor temperatures
0.1 0.01 and the fuel flow, and demonstrates the effect of the fuel
0.0 0.00 control system on the temperature. Each time the fuel flow
0.1 1 10 100 decreased, the gas temperature reacted in a corresponding
Step Size (ms)
fashion. The turbine blades, with the high heat transfer
Run Time Error coefficient and relatively low mass, tracked the gas
temperature, but lagged slightly behind and damped out the
Figure 9: Error and relative run time for standard test case effect. The other components showed no effect, at this scale,
with decreasing step size for shaft transient model from the changes in gas temperature, due to the increased
damping created by the greater mass and reduced heat transfer
effectiveness. This detail may be of little practical value for
Table 1: Comparison of transient models running test case this engine, but does provide interesting insights into engine
comparing run times and error in engine operating time to response and the potential thermal stresses in the components.
reach 120kRPM
384

Relative Relative Error

Time 382
Run Run compared to
Model Step
Time to Time to 10x smaller 380
Size
Temperature (K)

Shaft Mass step size 378

Shaft 1.0 12.3 1 ms 10.2 ms
Thermal 2.1 26.3 1 ms 2.0 ms 376

Mass 1/12.3 1.0 1 ms 12.7 ms 374

Energy 1/9.6 1.3 1 ms 12.4 ms
372
Energy-
4.1 50.3 0.1 ms 0.1 ms
Thermal 370
0 10 20 30 40 50 60
Time (s)
Comparison of Thermal and Shaft Models
The addition of thermal transient equations allowed the Tg T compressor

temperature of the engine components to be calculated, and

more accurately represented the temperature and pressure of the Figure 10: Compressor average gas and component
gas stream at the component exits. The disadvantage was a temperatures for acceleration from 118.2 to 126.0 kRPM
more complex system of differential equations. This increased

7 Copyright © 2006 by ASME

Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 06/19/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 1120 model. However, if we consider the time immediately
following the opening of the hole the perspective changes.
1110
Figure 15 again shows the shaft speed from each of the
1100 models, but only for 9ms after the hole opening. At this scale
Temperature (K)

the shaft model immediately started to linearly decrease speed,

1090
while the models including the gas dynamics showed a gradual
1080 increase in the angular deceleration. Also, at this scale, the
thermal model differed very little from the base energy model
1070
due to the slower response of the heat soak.
1060 This is also seen in Figure 16 that compares the change in
EGT. The shaft model was a step change. As soon as the air
1050
0 10 20 30 40 50 60 flow through the combustor and turbine was reduced, due to
Time (s) the air leakage, the temperature spiked. The mass model
Tg T blade T inner casing T outer casing
overshot the temperature, presumably because it accounted for
mass storage but not energy. Once the energy storage was also
Figure 11: Turbine rotor average gas and component accounted for, the temperature rise was rounded out. All the
temperatures for acceleration from 118.2 to 126.0 kRPM models that did not account for heat transfer were converging
at the end of the 9ms. The thermal model was again seen to
930 3.155 diverge from the energy model as time passed and the effect of
929 3.145 the heat transfer increased.
928 3.135
Figure 17 plots the CDP and again shows the smoothing
effect of the gas dynamic models, and the step in the shaft
Fuel Flow (kg/s)
Temperature (K)

927 3.125
model. Since the CDP is not strongly dependent on the energy
926 3.115 conservation, the mass simulation did not show any unusual
925 3.105 spikes. Again the shaft, mass and energy models converge,
924 3.095
despite taking different paths to get there. The thermal model
once more diverged as time passed. However, Figure 14 shows
923 3.085
the thermal model started to re-converge, but after seconds
922 3.075 instead of milliseconds.
30 35 40 45 50 55 60
Instead of plotting parameters with time it is instructive to
Time (s)
plot the running line on the compressor performance map, as in
Tg T blade T rim T diaphragm T outer casing Fuel
Figure 18. This plot shows the first 9 milliseconds after the
opening, and the sharp changes in the shaft model and the
Figure 12: Turbine rotor average gas and component
smoothed paths of the gas dynamic models are apparent.
temperatures for deceleration from 126.0 to 102.5 kRPM
126.0

Comparison of Mass, Energy and Shaft Models

A comparison of engine operation during normal 125.5

acceleration and deceleration demonstrated practically no

N (kRPM)

difference between the shaft, mass and energy models. The 125.0
control system prevented any large spikes in fuel flow, so the
engine did not change state rapidly enough that accumulation in
124.5
the control volumes significantly affected the operating point.
Plots similar to Figure 3 to ***will not be presented, since at
that scale the curves from the mass and energy models were 124.0
coincident to the shaft curves. 0 1 2 3 4 5

The mass and energy transients affected the results on a Time (s)

small time scale when a sudden change in engine operation Set point Shaft Mass Energy Energy Thermal
occurred, such as the failure of a seal or plug after the
compressor. To model a similar failure a 5mm diameter hole Figure 13: Comparison of transient models of shaft speed
was inserted 10ms into the simulation. Figure 13 compares the for bleed air through 5mm diameter hole at 0.01s for 5s
various models’ predicted shaft speed, and Figure 14 compares after opening
the predicted CDP. Over the long time period shown in these
figures, the only significant difference was in the thermal

8 Copyright © 2006 by ASME

Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 06/19/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 1.000 1.000

0.995 0.998

0.990 0.996

P o3/P o3 t=0
P 03 /P 03 t=0

0.985 0.994

0.980 0.992

0.975 0.990

0.970 0.988
0 1 2 3 4 5 -1 1 3 5 7 9
Time (s) Time (ms)

Shaft Mass Energy Energy Thermal Shaft Mass Energy Energy Thermal

Figure 14: Comparison of transient models of CDP for Figure 17: Comparison of transient models of CDP for
bleed air through 5mm diameter hole at 0.01s for 5s after bleed air through 5mm diameter hole at 10ms (adjusted to
opening 0ms on graph) for 9ms after opening
126.02 3.249

126.00 3.243

125.98 3.237
N (kRPM)

PR

125.96 3.231

125.94 3.225

125.92 3.219

125.90 3.213
-1 1 3 5 7 9 0.968 0.971 0.974 0.977 0.980 0.983
Time (ms) mrel

Set point Shaft Mass Energy Energy Thermal Shaft Mass Energy Energy Thermal

Figure 15: Comparison of transient models of shaft speed Figure 18: Comparison of transient models on compressor
for bleed air through 5mm diameter hole at 10ms (adjusted performance map for bleed air through 5mm diameter hole
to 0ms on graph) for 9ms after opening at 10ms plotted for 9ms after opening

1.014
DISCUSSION
1.012
The mass transient model was the best choice for most
1.010 uses. For gas path parameters the difference between it and the
1.008 thermal energy model was less than the difference between the
T o6/T o6 t=0

1.006
models and the experimental results, but the addition of heat
soak did bring the solution closer to the experimental results.
1.004
The time required for a solution was the deciding factor with
1.002 the mass model being the fastest.
1.000 If the temperature of the engine itself was required, the
thermal models would be useful. This could, for example, be
0.998
-1 1 3 5 7 9 the case for tip clearance studies. It also provided insight into
Time (ms) the engine behaviour that could not be otherwise easily
Shaft Mass Energy Energy Thermal
obtained. Unless small time scale response was necessary, the
shaft version of the thermal model was superior, since it ran
Figure 16: Comparison of transient models of EGT for twice as fast as the energy thermal model. This was the only
bleed air through 5mm diameter hole at 10ms (adjusted to case where the shaft model was faster than the gas dynamic
0ms on graph) for 9ms after opening solutions. For an engine this size the basic shaft model was
much slower than the gas dynamic models. Its only advantage

9 Copyright © 2006 by ASME

Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 06/19/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 was the relative simplicity of implementation. If a steady state 5. Kim,J.H., Song,T.W., Kim,T.S., and Ro,S.T., Model
model already exists then it is a relatively easy task to convert it Development and Simulation of Transient Behavior of
to a shaft transient model. Heavy Duty Gas Turbines, ASME Turbo Expo 2000, 2000,
If engine behaviour immediately after an abrupt change in Munich Germany, Paper Number: 2000-GT-0548.
engine state is of interest, then the energy model was the best. 6. Sanghi,V., Lakshmanan,B.K., Rajasekaran,R., 2001,
By accounting for energy conservation it corrected some Aerothermal Model for Real-Time Digital Simulation of a
temperature spikes that showed up in the mass model. In Mixed-Flow Turbofan Engine, Journal of Propulsion and
addition, the energy model runs almost as fast as the mass Power 17(3):629-35
model. If simulation time was not an important limitation the 7. RTO Applied Vehicle Technology Panel, 2002,
energy model would be a strong candidate to run the majority Performance Prediction and Simulation of Gas Turbine
of the simulations, since, in other work, it was found to be more Engine Operation, Rep. RTO-TR-044, Research and
stable when modeling degraded engine behaviour. Technology Organisation, NATO, Neuilly sur Seine,
According to Alves and Barbosa [8] shaft transients occur France
in the 1-5Hz (1000-200ms) and gas dynamics in the 5-50Hz 8. Alves,M.A.d.C., Barbosa,J.R., 2003, A step further in gas
(200-20ms) range. As can be seen in Figure 17 and Figure 16 turbine dynamic simulation, Proceedings of the Institution
in less than 10ms the shaft and gas dynamic models were of Mechanical Engineers. Part A, Journal of power and
reconverging for this engine and the abrupt changes had all energy. 217(6):583-92
passed. When modeling small engines it is important to keep 9. Stamatis,A., Mathioudakis,K., Ruiz,J., and Curnock,B.,
these differences in mind, and remember that currently most Real Time Engine Model Implementation for Adaptive
papers published refer to relatively large engines. Control & Performance Monitoring of Large Civil
Turbofans, ASME Turbo Expo 2001, 4-6-2001, New
CONCLUSIONS Orleans, Louisiana, Paper Number: 2001-GT-0362.
The dynamic response of micro turbine engines is different 10. Incropera,F.P., DeWitt,D.P., 1996, Introduction to Heat
than larger ones. The gas dynamic and heat soak transients Transfer, Toronto: John Wiley & Sons,
depend on the inverse of diameter, while shaft acceleration 11. Pilidis,P. and Maccallum,N.R.L., Models for Predicting Tip
depends on the inverse of diameter to the 2nd power. The Clearance Changes in Gas Turbines, 60th Symposium of
effects of this on dynamic models of a micro engine were the AGARD Propulsion and Energetics Panel, 11-10-1982,
examined. Neuilly sur Seine, France, Paper Number: AGARD-CP-
324 17.
ACKNOWLEDGMENTS 12. Davison,C.R. and Birk,A.M., Set up and Operational
The authors would like to thank the Canadian Natural Experience with a Micro-Turbine Engine for Research and
Sciences and Engineering Research Council (NSERC) for Education, ASME Turbo Expo 2004, 14-6-2004, Vienna,
funding this work, and Allied Signal (Honeywell) for a grant to Austria, Paper Number: GT2004-53377.
help initiate it. In addition, we would like to thank NSERC and
the Ontario Graduate Scholarship program for scholarship
support.

REFERENCES
1. Merrington,G., 1989, Fault Diagnosis of Gas Turbine
Engines From Transient Data, Journal of Engineering for
Gas Turbines and Power 111:237-43
2. Kim,J.H., Song,T.W., Kim,T.S., and Ro,S.T., Dynamic
Simulation of Full Start-Up Procedure of Heavy Duty Gas
Turbines, ASME Turbo Expo 2001, 4-6-2001, New
Orleans, Louisiana, Paper Number: 2001-GT-0017.
3. Chiang,H.-W.D., Hsu,C.-N., Lai,A., and Lin,R., An
Investigation of Steady and Dynamic Performance of a
Small Turbojet Engine, ASME Turbo Expo 2002,
Amsterdam, 2002, New York, Paper Number: GT-2002-
30577.
4. Al-Hinai,A., Feliachi,A., Dynamic model of a microturbine
used as a distributed generator, 209-213, 2002,
Piscataway, NJ, USA, IEEE, Proceedings of the Thirty-
Fourth Southeastern Symposium on System Theory,

10 Copyright © 2006 by ASME

Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 06/19/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use