You are on page 1of 6

Fidelity Assessment of a Jetstream 41 Flight Simulator Model

Lukasz Lachowicz, Mark White, Neil Cameron


Department of Engineering, University of Liverpool
The Quadrangle, Liverpool, L69 3GH
l.lachowicz@liv.ac.uk

Abstract
The increasing reliance on Flight Simulation
Training Devices by civil and military air operators,
along with the continuous development of complex
computer modelling tools such as FLIGHTLAB bring
up the issue of flight simulation fidelity. Regulatory
documents such as the European Aviation Safety
Agency Certification Specifications for Aeroplane
Flight Simulation Training Devices define the limits on
how good is good enough, although some of these
definitions are missing important details from the point
of view of model engineering. This study used a
linearised simulator model of the Jetstream 41 twinengine airliner to investigate the suitability of the
current certification requirement for short period mode
dynamics. It has been revealed that the definition of this
requirement is excessively general and strict on the
accuracy of the models short period natural frequency,
possibly leading to unnecessary complexity and costs
associated with the models development.

Notation
A
B
Mq
Mw
M
Ue
Zw
Z
g
q
w
u
x

SP

SP

State matrix
Control matrix
Pitch damping derivative (1/sec)
Pitch stiffness derivative (rad/secft)
Pitching moment control derivative (1/sec2)
Equilibrium airspeed (ft/sec)
Heave damping derivative (1/sec)
Vertical force control derivative (ft/sec2rad)
Acceleration due to gravity (ft/sec2)
Pitch rate (deg/sec)
Heave velocity (ft/sec)
Control vector
State vector
Angle of attack (deg)
Short period damping ratio
Elevator control input (%)
Pitch angle (deg)
Short period natural frequency (rad/sec)

CAP
CFD
CS-FSTD(A)
EASA
FFS
HUD
HQs
HQR
MTE
TER

Control Anticipation Parameter


(rad/sec2/g)
Computational Fluid Dynamics
Certification Specifications for
Aeroplane Flight Simulation Training
Devices
European Aviation Safety Agency
Full Flight Simulator
Head-Up Display
Handling Qualities
Handling Qualities Rating
Mission Task Element
Transfer Effectiveness Ratio

1. Introduction
The purpose of Flight Simulation Training Devices
(FSTD) is to provide a relatively inexpensive and
flexible alternative to aircraft-based flight training, and
their use is becoming an integral part of training
programmes conducted in the commercial and military
sectors [1].
The fidelity of FSTDs could be interpreted as the
level of similarity between the device and the simulated
aircraft, in terms their functionality and flying qualities.
A high fidelity simulator offers a level or realism that
enables the pilot to seamlessly transfer the skills gained
in such a device to the real flight deck. The
effectiveness of this transfer is commonly expressed as
the Transfer Effectiveness Ratio (TER):
=

(1)

Tc is the amount of time needed for the pilot to reach a


specific training goal by flying solely in the aircraft, Te
is the time to reach the same goal with simulator-based
training, and Xe is the time spent training in the device
[2]. A positive value of TER indicates an adequate level
of fidelity and a potential reduction in the training costs.

This study was concerned with fidelity of the FSTD


flight dynamics model that is required to achieve a
positive TER and meet the current flight simulation
regulations. This was achieved by investigating the
sensitivity of pilots perception to deficiencies in the
linearised model of a fixed-wing aircraft.

2. Simulation Fidelity Standards


In Europe, FSTDs are regulated by the European
Aviation Safety Agency (EASA), with the qualification
requirements for a range of device classes and fidelity
levels defined in the Certification Specifications for
Aeroplane Flight Simulation Training Devices (CSFSTD(A)) [3]. The Full Flight Simulator (FFS) class
devices offer the highest training effectiveness, and
feature the most accurate representation of the flight
dynamics. Development of aerodynamic models is
associated with expensive data-gathering and analysis
techniques (such as parameter identification via flight
testing [4]), and the amount of data that could be
processed in real time simulations is limited by the
computational capacity of the device.
One study revealed that 80% of fidelity should be
achievable with physical modelling but the remaining
20% requires artificial tuning (Padfield, et al., 2005).
Various tuning techniques have been developed such as
one proposed by Lu et. al [5], in which correction
factors are applied to the models stability derivatives
based on the differences between the factual (flight test)
data and the deficient model. The process is defined in
the form of state-space matrices by Equations 2-4,
where is the velocity vector, A is the state matrix, x is
the state vector, B is the control matrix, and u is the
control vector.
= +
=
=

flies a Mission Task Element (MTE), which is a test


course specifically designed to excite the regime of
flight dynamics relevant to the criterion selected from
the handbook, and then awards a Handling Qualities
Rating (HQR) in accord with the relevant rating scale.
Ratings are given in terms of HQ Levels, with Level 1
representing desirable performance attainable with
minimum pilot workload, and Level 3 corresponding to
likely loss of control with heavy workload. The
approach is equally applicable to simulators as well as
real aircraft, which allows a standardised and objective
comparison to be made between the two.
The Control Anticipation Parameter (CAP) is an HQ
criterion used to quantify the predictability of the short
period pitch response. It is defined by Equation 5 as a
function of the short period natural frequency (SP),
airspeed (Ue), and the heave damping derivative (Zw).
The value of Zw is dependent upon the lift curve slope
of the main wing [4]. The CAP HQ chart for Category
B flight phases, which include the cruise conditions, is
shown on Figure 1.
=

(5)

(2)
(3)
(4)

This technique could technically be applied as an


infinite loop, without ever achieving a perfect match
between the model and test data. This leads to the
question of how much accuracy in the values of
stability derivatives is required, such that any further
enhancements would not introduce any perceivable
change to the models flying qualities.

3. Handling Qualities
Standards described in MIL-HDBK-1797 by U.S.
Department of Defence [6] provide the means to
quantify and rate the Handling Qualities (HQ) of a
fixed-wing aircraft in terms of task task-specific
performance, safety, and the workload imposed upon
the pilot. The aircraft is evaluated by a test pilot who

Figure 1: HQ chart for the Control Anticipation Parameter

4. Model Description
The analysis was based on a FLIGHTLAB model of
the British Aerospace Jetstream 41 (F-JS41) built by
Lachowicz [7]. The model was developed from a
generic twin-engine airliner template that was gradually
modified by introducing new design and aerodynamic
data specific to the JS41 (Figure 2). The new data
include:
-

Look-up tables containing the aerofoil


aerodynamic coefficients for the operating

range of Mach numbers and flap deflections,


obtained
from XFOIL
and
existing
experimental data;
Look-up tables containing the aerodynamic
force coefficients acting on the fuselage body
for a range of small attitude angles, obtained
from Computational Fluid Dynamics (CFD)
simulations in Fluent; and
Design data obtained from the original aircraft
documentation.

obtained from FLIGHTLAB after trimming the model


in straight and level cruising flight at a density altitude
of 24,000ft and airspeed of 250kts.

[ ] = [


] [ ] + [ ]

(6)

Figure 4: Simulink short period model

A MATLAB script was used to apply a scaling factor


to the selected derivative, and then drive the Simulink
model by applying a 0.5sec step input of a predefined
magnitude.
Figure 2: British Aerospace Jetstream 41

5. Linearised Model
CS-FSTD(A) defines a requirement on the FFS
models short period pitch response in cruising flight,
stating that the pitch angle () must not deviate by more
than 1.5 from the baseline response. Following on the
idea of fidelity enhancement through fine tuning of the
stability derivatives, this brings up the question of much
accuracy is needed in the individual derivatives such
that the difference between the model and baseline pitch
responses remain within the tolerance limit.
A Simulink model of the F-JS41 state-space short
period approximation, given by Equation 6, was used to
investigate this issue. The advantage of this approach is
that the stability derivatives contained in the state
matrix could be modified directly, and the resulting
differences in response could be determined
automatically. The state and control matrices were

5.1 Off-Line Analysis Results


For elevator control inputs of up to 10% magnitude,
the value of Mq could vary by up to +23/-18%, and Mw
by up to +19/-16% from the baseline. Undesirable short
period oscillations could be introduced into the
response by either reducing M q or increasing Mw. When
varying Mq, the maximum deviation in occurs approx.
0.1s after releasing the control input, which changes the
amount time the input must be applied for to give the
required change in pitch attitude.
Values of Mq and Mw in fixed-wing aircraft
(Equations 7 and 8) are both dominated by the
aerodynamics of the tailplane [4], thus an uncertainty
might be introduced into both by an inaccurate source
of the relevant design data.

Figure 3: Maximum allowable changes in response due to a percentage change in (a) M q and (b) Mw stability derivatives

2
2
2
+

+
=

(7)
(8)

6. Flight Simulator Trial


A drogue tracking MTE was flown in the
HELIFLIGHT simulator at the University of Liverpool
(UoL) (Figure 5) to determine the minimum offset in
Mq and Mw derivatives that would noticeably alter the
models short period response. The refuelling drogue
was attached behind another aircraft travelling at the
same speed, and oscillated in pitch. The test pilot (a
qualified private pilot) was required to follow the centre
of the drogue with the middle of the Head-Up Display.

increments to simulate an error in Mq and Mw


derivatives, given by Equations 7 and 8 respectively.
The reason for not modifying the derivatives directly is
that SP and SP are more familiar parameters that could
be readily interpreted, and which allow the use of the
CAP HQ chart in selection of the test points. The
drawback of this approach is that the two derivatives
are coupled, so that they were both varied
simultaneously. The baseline values of Zw and Ue were
maintained throughout the tests.
The values of SP and SP were initially increased or
decreased by 1rad/s and 0.1 respectively, until the point
at which the pilot perceived a subjective change in the
pitch response characteristics. Smaller increments were
then used to determine the limits with greater accuracy.
The tests were repeated for a different baseline ( SP =
3.0rad/s, SP = 0.67) to investigate the influence of SP
on the results.

6.1 Flight Trial Results


The pilot applied high frequency inputs of up to 10%
magnitude to make continuous pitch attitude
adjustments to follow the refuelling drogue (Figure 7).

Figure 5: University of Liverpool HELIFLIGHT Simulator


Figure 7: Time histories from the drogue tracking MTE

Data points obtained from the flight trial were used to


define regions, in terms of percentage Mq and Mw
magnitude change, within which the pilot was unable to
perceive any change in the flying qualities (Figure 8).
The regions are valid for the two test baselines and
elevator control inputs of up to 10% magnitude. More
data points with an intermediate SP baseline would be
needed to investigate any potential correlation in
geometry of the regions. Numerical results are
presented in Table 1.
Figure 6: Pilot's view of the drogue tracking MTE

Following an initial familiarisation run with the


baseline model (SP = 6.5rad/s, SP = 0.67), the
magnitudes of SP and SP were varied in small

Figure 8: Thresholds of pilot perception for changes in Mq and


Mw derivatives with up to 10% elevator control input

Table 1: Combinations of short period natural frequency and


damping ratios that in a perceivable change in the pitch
response. SP,baseline = 0.67.

A
B
C
D
E
F
G
H

SP,baseline
(rad/s)
6.5
6.5
6.5
6.5
3.0
3.0
3.0
3.0

SP
(rad/s)
7.5
5.5
6.5
6.5
3.5
2.0
3.0
3.0

SP
0.67
0.67
0.50
0.73
0.67
0.67
0.62
0.72

Mq
(%)
+20.0
-20.0
-32.9
+11.6
+30.8
-33.0
+14.3
-14.3

Mw
(%)
+39.1
-32.4
+15.2
-5.5
+43.4
-86.3
-19.5
+19.5

The results are further plotted on the CAP HQ chart


(Figure 9) to analyse them with respect to HQ Levels,
natural frequency, and damping ratio. The perception
thresholds are as in Figure 8, with letters crossreferencing the individual vertices. The solid concentric
regions are equivalent to the maximum combined
deviations in Mq and Mw allowable by the CS-FSTD(A)
criterion. These were determined using the Simulink
short period model for a range of small control inputs.
With 6.5rad/s baseline the pilot was able to perceive
the changes in SP at or before reaching the tolerance
region for 10% input, which is illustrated by points C
and D. In case of the 3rad/s baseline the SP perception
threshold correlates with the 5% input boundary, as
illustrated by points G and H. This suggests that the CSFSTD(A) requirement imposes an adequate limit on the
error in SP at high SP, but also that it is too restrictive
at low SP.

Figure 9: CAP HQ chart illustrating the pilot's perception


thresholds with up to 10% control inputs and theoretical
FSTD(A) tolerance bounds. CAP is given by Equation 5.

The pilot was unable to sense the changes in SP


(CAP 2) before the tolerance regions for 5% control
input or less were breached, as illustrated by points A,
B, E, and F. This suggests that more error in SP, and
thus in the error in Mq and Mw equivalent to SP by
Equations 5 and 6, should be allowed for by the
requirement. A relaxation of the requirement could be
achieved by stretching the tolerance regions for 10%
input in the CAP axis, such that they coincide with
points A, B, E, and F.
It could also be seen that its possible to obtain a
Level 2 and Level 3 rating with 2.5% control inputs at
the high baseline SP. Inputs of such small magnitude
would not realistically be applied by the pilot to control
the attitude. However, the areas of the tolerance regions
are increasing with SP and this suggests that they
would further extend towards Level 2 or Level 3 CAP
boundaries in the upper section of the chart.

7. Conclusion
It has been shown that the CS-FSTD(A) requirement
for the short period mode is too general and does
account for the important and complex influences of the
input magnitude, short period natural frequency, and the
damping ratio. With certain types of deficiencies in the
models stability derivatives the requirement could
mandate enhancements that are unperceivable to the
pilot, and ultimately have no effect on the simulator
training effectiveness. This could result in unnecessary
complexity and financial costs associated with the
model development.
It is suggested that a new short period criterion based
on the aircraft design parameters, such as the stability
derivatives analysed in this study, is developed to

replace the existing one. The methodology established


in this study should assist in achieving this.
The effects of error in the natural frequency on other
handling qualities, such as the pitch bandwidth, will
also need to be investigated.

References
[1] Allerton, D. (2010). Principles of Flight Simulation.
Chichester: John Wiley & Sons, Ltd.
[2] Hays, R. & Singer, M. (1989). Simulation Fidelity in
Training System Design. New York: Springer-Verlag.
[3] EASA. (2012, July 4). Certification Specifications for
Aeroplane Flight Simulation Training Devices
[4] Cook, M. V. (2013). Flight Dynamics Principles (3rd ed.)
Oxford: Butterworth-Heinemann.
[5] Lu, L., Padfield, G, White, M., & Perfect, P. (2011).
Fidelity enhancement of a rotorcraft simulation model
through system identification. The Aeronautical Journal,
453-470.
[6] Department of Defence. (1997). MIL-STR-1797A Flying
Qualities of Piloted Aircraft.
[7] Lachowicz, L. (2015). Development of a Jetstream 41
Flight Simulator Model. Liverpool: University of
Liverpool.

You might also like