by

advertisement
THE EFFECT OF FLIGHT SIMULATOR MOTION ON MODELLED VESTIBULAR RESPONSE
by
Kathleen M. Misovec
B.S., Massachusetts Institute of Technology, 1984
Submitted in Partial Fulfillment
of the Requirements for the
Degree of
Master of Science in Aeronautics and Astronautics
at the
Massachusetts Institute of Technology
September, 1986
()Massachusetts
Institute of Technology
1986
Signature of Author
Department of Aeronautics and Astronautics
September, 1986
Certified by
Professor Steven R. Bussolari
Thesis Supervisor
Certified by
Professor Harold Y. Wachman Chairman
Departmental Graduate Committee
Department of Aeronautics and Astronautics
I
Room 14-0551
mITLibraries
Document Services
77 Massachusetts Avenue
Cambridge, MA 02139
Ph: 617.253.2800
Email: docs@mit.edu
http://Iibraries.mit.edu/docs
DISCLAIMER OF QUALITY
Due to the condition of the original material, there are unavoidable
flaws in this reproduction, We have made every effort possible to
provide you with the best copy available. If you are dissatisfied with
this product and find it unusable, please contact Document Services as
soon as possible.
Thank you.
Some pages in the original document contain pictures,
graphics, or text that is illegible.
THE EFFECT OF FLIGHT SIMULATOR MOTION ON MODELLED VESTIBULAR RESPONSE
by
Kathleen M. Misovec
Submitted to the Department of Aeronautics and Astronautics on August 30,
1986 in partial fulfillment of the requirements for the Degree of Master of
Science in Aeronautics and Astronautics
ABSTRACT
A study of the effect of flight simulator motion on modelled vestibular
response was conducted. Experiments were performed on a Boeing 727 flight
simulator which used a synergistic motion base. Three different motion
conditions were compared in the experiments. One condition consisted only
of high frequency, low amplitude special effects such as turbulance and
landing gear extension. The second condition, jostle motion, consisted of
motion capability in the lateral and vertical degrees of freedom as well as
special effects. The third condition, full motion, was the six degree of
freedom capability normally used on the flight simulator.
Three flight scenarios, designed to require significant amounts of
pilot control activity, were flown by eighteen subjects. The subjects were
divided into three groups of six, each group performing one scenario. The
first scenario consisted of an engine flame out on take off. The second, an
airwork scenario, consisted of steep turns, approach to stall maneuvers, and
standard rate turns with yaw damper failure.
The third scenario was an ILS
approach and landing in wind shear.
The two primary types of measurements that were analyzed were acceleration
errors and vestibular errors. Acceleration error is defined to be the
difference between the accelerations of the aircraft and the simulator.
Vestibular error is defined to be the difference between the pilot's modelled
vestibular responses in the aircraft and the simulator. Two other sets of
measurements, pilot opinion and pilot performance, were also compared for
the three motion conditions of the experiment.
In general, no significant differences were found between the motion
conditions for any of the measurements. However, for the vestibular error
measurments, the rotational vestibular errors were usually below established
thresholds of perception while many of the translational errors were above
the thresholds of perception.
Thesis supervisor: Steven R. Bussolari, PhD
Title: Assistant Professor of Aeronautics and Astronautics
This thesis is dedicated in memory of Margaret Mary Misovec.
ACKNOWLEDGEMENTS
I would like to thank my father, Dr. Andrew Misovec. His ideas and our
conversations about engineering have always been extremely valuable to my
education. His constant encouragement and insights are greatly appreciated.
Professor Steven Bussolari provided the basic idea for the experiments
and contributed to thcir implementation and analysis. I would like to thank
Dr. Alfred Lee for the analysis of performance data and Ted Demosthenes for
his generous help with running the experiments. Thanks are due to Dr.
Charles Oman for consultations about the vestibular models and to Dr. Alan
Natapoff for consultations about the statistical analysis of the data.
Conversations with Dr. Mohammed Massoumnia have been extremely valuable
to my general education and are greatly appreciated. I would also like to
thank Ed Kneller, Mark Shelhamer and R. Bryan Sullivan for their helpful
thoughts on the technical aspects of the analysis of these experiments.
Thanks also to Dan Merfeld for his ideas. The generous amounts of advice and
help from Sherry Modestino and Ed Kneller in writing and preparing this
thesis are gratefully acknowledged.
I am grateful to Margaret Misovec, Mary Misovec, Paul Misovec, Andrew
C. Misovec, Michael Misovec, James Smith, Patricia Tellier and Rene Tellier
for their support and encouragement. Thanks also to Kevin Ackley, Marilyn
Cieuzo, Suzanne Cox, Wayne Greene, Justin Marble and Jennifer Wiseman.
TABLE OF CONTENTS
CHAPTER 1:
1.1
1.2
1.3
1.4
1.5
1.6
INTRODUCTION . . . . . . . .
. . . . . . . . . . .
. . . .
. .
Types of Motion Systems . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
Types of Washout Systems
Problems with the Optimal Control Washout Design - What Should
Be Optimized? . . . . . . . . . . . . . . . . . . . . . . . .
Problems Associated with Quantifying Simulator Realism . ....
. . . . . . . . . . . . .
This Research - More Basic Questions
Thesis Outline
. . . . . . . . . . . . . . . . . . . . . .
CHAPTER 2: FORMULATION OF THE VESTIBULAR ERROR AS A MOTION FIDELITY
MEASURE . . . . . . . . . . . . . . . . . . . . . . . . .
.
. . . . . .. .. .
2.1 Introduction .. .. '. . . . . . .
..
.. ... . . . .. . .
2.2 Semicircular Canals . . . . . *..
. .
2.2.1 Physical Description of the Semicircular canals
2.2.2 Mathematical Model of the semicircular canals . . .
2.3 Otoliths
. . . . . . . . . . . . . . . . . . . . . . . . .
2.3.1 Physical Description of Otoliths . . . . . . . . . .
2.3.2 Mathematical Models of the Otoliths
. . . . . . . .
2.4 Vestibular Error Measurements . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
CHAPTER 3: EXPERIMENTAL DESIGN AND ANALYSIS MEASUREMENTS
. . . . . . .
3.1 Brief Introduction
. . . . . . . . . . . . . . . . . . . .
3.2. Experiment Description . . . . . . . . . . . . . . . . . . .
3.2.1 Simulator . . . . . . . . . . . . . . . . . . . . . .
3.2.2 Motion Conditions . . . . . . . . . . . . . . . . . .
3.2.3
Flight Scenarios . . . . . . . . . . . . . . . . . . .
3.2.3.1 Familiarization Scenario
. . . . . . . . . . .
3.2.3.2 Engine flame-out on takeoff . . . . . . . . . .
3.2.3.3 Airwork Scenario . . . . . . . . . . . . . . .
3.2.3.4 ILS Approach and Landing Scenario . . . . . . .
3.2.4 Subjects . . . . . . . . . . . . . . . . . . . . . . .
3.2.5 Order Effects . . . . . . . . . . . . . . . . . . . .
3.3 Analysis Measurements. . . . . . . . . . . . . . . . . . . . .
3.3.1 Opinion Measurements. . . . . . . . . . . . . . . . . .
3.3.2 Acceleration Error Measurements
. . . . . . . . . . . .
3.3.3 Vestibular Error Measurements
. . . . . . . . . . . .
3.3.4 Pilot Performance Measurements. . . . . . . . . . . . .
3.3.4.1 Engine Flame-out Scenario . . . . . . . . . . . .
3.3.4.2 Airwork Scenario. . . . . . . . . . . . . . . . .
3.3.4.3 ILS Approach and Landing Scenario . . . . . . . .
3.4 Data Collection . . . . . . . . . . . . . . . . . . . . . . .
3.4.1 Variables . . . . . . . . . . . . . . . . . . . . . . .
3.4.2 Trials . . . . . . . . . . . . *.* . . .. . . .
. . . .
3.4.3 Pertinent Degrees of Freedom for Each Scenario
. . . .
3.4.4 Data Windows . . . . . . . . . . . . . . . . . . . . .
3.4.5 Data Collection Problems . . . . . . . . . . . . . . .
3.4.5.1 Possible Aliasing of Rotational Data . . . . . .
3.4.5.2 Lateral Axis Data . . . . . . . . . . . . . . . .
3.4.5.3 Longitundinal Axis Acceleration Data. . . . . . .
3.4.5.4 Other Problems. . . . . . . . . . . . . . . . . .
3
5
7
13
17
17
21
23
24
24
24
24
25
33
33
33
35
40
40
40
40
40
42
43
43
44
44
46
47
47
47
51
52
53
53
53
54
54
54
57
57
58
59
59
60
60
60
CHAPTER 4: PILOT OPINION RESULTS AND DISCUSSION
4.1 Introduction . . . . . . . . . . . .
4.2 Presentation of Opinion Results . . .
4.3 Discussion of Pilot Opinion Results. .
. . . .
. . . .
. . . .
. . .
.
.
.
. . . .
. . . .
. . . .
. . . . .
. . .
. . .
. . .
. . . .
62
62
62
64
CHAPTER 5: ACCELERATION ERROR RESULTS AND DISCUSSION . . . . . . . . .
70
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . .
70
5.2 Engine Flame-Out Acceleration Error Results
. . . . . . . .
71
5.3 Steep Turu Acceleration Error Results . . . . . . . . . . . .
72
5.4 Stall Acceleration Error Results . . . . . . . . . . . . . .
75
5.5 Rate Turns with Yaw Damper Failure Acceleration Error Results.
78
5.6 ILS Approach and Landing Scenario Results
. . . . . . . . .
80
5.6.1 Approach Segment: 500' - 200'
. . . . . . . . . . . .
80
5.6.2 Landing Segment:
Last 20 to 25 s Before Touchdown .
83
5.7 Discussion. .........
,.
.............
'..
....... . ...... 86
5.7.1 Discussion of Analysis Method. . . . . . . . . . . . . . 86
5.7.2 Discussion of Results. . . . . . . . . . . . . . . . . . 88
CHAPTER 6: VESTIBULAR ERROR RESULTS AND DISCUSSION
. . . . . . . .
6.1 Engine Flame-Out Vestibular Error Results . . . . . . . . .
6.2 Steep Turn Vestibular Error Results . . . . . . . . . . . .
6.3 Stall Vestibular Error Results . . . . . . . . . . . . . .
6.4 Rate Turns with Yaw Damper Failure Vestibular Error Results
6.5 ILS Approach and Landing Vestibular Error Results.
. ....
6.5.1 Approach Segment: 500' - 200'
. . . . . . . . . . .
6.5.2 Landing Segment:
Last 20-25 sec before touchdown
.
6.6 Discussion of Vestibular Error Results
. . . . . . . . . .
CHAPTER 7: PILOT PERFORMANCE RESULTS . . . . . . .
7.1 Engine Flame-Out Performance Results . .
7.2 Airwork Scenario Performance
. . . . . .
7.2.1 Approach-to-Stall . . . . . . . . .
7.2.2 Rate Turns with Yaw Damper Failure
7.3 ILS Approach and Landing Scenario
. . .
7.4 Discussion of Performance Results. . . . .
CHAPTER 8:
8.1
8.2
8.3
8.4
. .
. .
. .
.
. .
. .
.
. . . .
. . . .
. . . .
. . . . .
. . . .
. . . .
. . . . .
CHAPTER 9:
CONCLUSIONS AND RECOMMENDATIONS . .
9.1 Conclusions . . . . . . . . . . . . .
9.2 Recommendations
. . . . . . . . . .
4
. . . .
. . . .
. . . .
. .
. .
. .
.
.
.
.
90
91
92
95
98
101
101
105
109
. . . . 111
. . . .
111
. . . . 112
. . . . 112
. . . . 112
. . . .
112
. . . . 113
PROBLEMS DUE TO ALIASED ROTATIONAL VESTIBULAR RESPONSES
Introduction . . . . . . . . . . . . . . . . . . .
Aliasing General Information . . . . . . . . . . .
The Effect of Undersampling on the Results of This
Rotational Vestibular Error Results . . . . . . . .
.
.
.
. .
.
117
. . . . .
. . . . .
Study
. .
. . . . .
117
117
119
119
. . . .
. . . .
. . . .
.
. .
. .
. .
131
131
132
CHAPTER 1: INTRODUCTION
In addition to being useful as both subjects and tools of
scientific research, flight simulators provide training to aircraft
pilots, and are also used for pilot certification.
Pilot training
and certification encompasses a broad range of flight manuevers
from standard procedures such as cruise to more complicated and
frequently dangerous manuevers such as an engine failure on takeoff.
Performance of such manuevers in a simulator instead of an actual
aircraft is valuable for two reasons.
One is because if a pilot
makes a mistake in an actual aircraft, collision is possible.
The
other is that although simulators may have a higher initial cost
than the actual flight vehicle, the operational costs of simulators
are usually less than those of an actual flight vehicle.
This
makes pilot training and certification less expensive in a simulator
in an actual aircraft.
It is vitally important that the simulation is realistic enough
so that pilot training received in simulated emergency manuevers
is adequate enough to apply to similar situations that may arise
in an actual aircraft.
Currently it is not known how realistic
the simulator must be to achieve adequate "transfer-of-training"
from simulator to aircraft.
So the effort has been to make the
simulation as realistic as possible.
of realism.
There are three general types
Engineering realism may be parameterized by dynamic
characteristics such as accelerations and forces.
Psychological
realism could be achieved by such factors as an actual instrument
panel, or realistic air traffic control simulation and could be
measured by pilot opinion about the fidelity of the simulation.
5
Physiological fidelity pertains to the accurate representation of
sensory signals sent to and processed by the central nervous system.
A natural tradeoff arises between the fidelity of the simulation and the cost associated with the complexity of the system.
The lower fidelity flight training systems might be a cardboard
representation of the instrument panel or relatively simple program
that runs on a microcomputer.
Both realism and cost are increased
as real instruments and an actual cockpit setting, which provide
primarily psychological fidelity, are added to the system.
Physio-
logical as well as psychological fidelity can be increased, at
high financial cost, by including a visual system in the simulation.
A motion system may increase engineering, psychological and physiological realism to the simulation.
In recent years, technological
improvements in motion systems have been mainly software-oriented
so that their cost, which is high primarly due to dynamic requirements of obtaining accelerations, remains relatively high.
Most of the more advanced simulators today have highly complex,
expensive motion systems.
Many motion systems have been installed
and methods of controlling the motion base have been widely
researched because of an implicit assumption that motion capability
in a simulator is critical for realism and therefore necessary for
adequate transfer-of-training from simulator to aircraft.
The
goal of this thesis is to scientifically examine this assumption
by comparing the effects of different levels of motion capability
on various parameters of engineering, psychological and physiological realism.
6
1.1 Types of Motion Systems
There are six possible independent directions or degrees of
freedom in which motion capability is possible in all physical
unconstrained systems.
As shown in Figure 1.1, there are the
three translational degrees of freedom and three rotational degrees
of freedom.
In this thesis, the translational degrees of freedom
will be referred to as "surge" for longitudinal motion; "sway" for
lateral motion, and "heave" for vertical motion.
While the rota-
tional degrees of freedom will be referred to as "pitch",
and
"yaw",
"roll"
IVarer&I
Figure 1.1:
The six possible degrees of freedom
of motion.
There are a variety of different mechanical ways to produce
simulator motion currently in use.
In cascaded systems, such as
that shown in Figure 1.2, each degree of freedom is separately and
independently controlled by a cascade of six motion elements.
Translational degrees of freedom are achieved by allowing a cab to
7
move on a cascade of linear tracks.
This motion is characterized
by the position, velocity and acceleration of the cab along the
tracks. The length of the track limits translational motion capability.
Rotational directional capability is acheieved by suspending
the cab in a motor-driven gimbal.
Three nested gimbals are needed
to obtain complete rotational capability about any axis.
motion is limited by the maximum angular positions,
Rotational
angular velo-
cities, and angular accelerations that the system can undergo.
In other systems, known as synergistic simulators, the actuators work together to achieve motion in a single degree of freedom
(see Figure 1.3).
This can be achieved by having a cab supported
by a platform which is supported by six legs attached to the ground.
The legs are controlled by hydraulic pistons and limitations are
due to the maximum length of the legs and the their maximum rates
of change.
The maximum force that the legs can generate is another
motion capability limiting factor.
The primary advantage of this
"hexapod" system over the cascaded system is that the hardware is
simpler.
Some systems, such as the Vertical Motion Simulator (VMS)
located at NASA Ames, have a motion base system that is a combination of the cascade and the hexapod system.
In the VMS, as shown
in Figure 1.4, linear tracks provide the translational motion capabilitities,
lities.
while a hexapod system provides the rotational capabiAlthough there are only two tracks in this system,
the
cab can rotate about a vertical axis to allow the horizontal track
to be used for both longitudinal and lateral motion (Sullivan,
1985).
8
Another type of system is exemplified by the Large Amplitude
Multi-Mode Aerospace Research Simulator ("LAMARS") located at
Wright-Patterson Air Force Base.
In this system, a 10 meter beam
suspends a large sphere which supports the simulator cab.
beam is capable of lateral and vertical motions.
The
The cab is gim-
balled so that it can move in all three rotational degrees of
freedom.
Centrifuge motion systems are gimballed cabs that are suspended
at the end of a large rotating arm (see Figure 1.5).
The design of
a controller for these motion systems is challenging because the
pilot is in a rotating environment.
Unlike the above systems which
are only capable of half of a g acceleration levels, centrifuge
motion systems can achieve translational accelerations of up to
40g (Ish-Shalom, 1982).
The capability of large sustained acceler-
ation is extremely useful in training military pilots to avoid
blacking out, which happens ocassionally when they are flying high
performance fighter aircraft.
The high g-levels can draw the blood
from their heads and cause them to lose consciousness.
One of the
highest fidelity flight simulators currently in existence are variable stability airplanes that are used as flying simulators.
These
planes are designed so that their natural frequencies can be varied
to simulate different types of aircraft.
9
CDC199I MSPILAT
AI=ZaS P.ATFORM
PITCH
AXIS
YISUAL O3KAT
t"ie
curxia
c-zAagpg
Faci2
A,
977T
AXIST
(WCA -liCA4.
CA4043
Figure 1.2 Example of Cascaded Motion Base System (Sinacori,
10
1977)
Figure 1.3:
Example of Synergistic Motion Base System
(Unscaled Drawing)
11
Figure 1.4 Vertical Motion Simulator (VMS)
(Sullivan,1985)
IVon-Gierke
Figure 1.5: Artist's Conception of ,cntri'uge Motion Base
System (7on-Gierke
19(a
12
1.2 Types of Washout Systems
Designing a simulator motion drive logic system is a challengThe basic task is to compute simulator motion
ing control problem.
base commands on the basis of computed airplane motions without
exceeding the limitations of the motion system, while retaining as
much realism as possible (Figure 1.6) (Ish-Shalom, 1982).
The
term "washout system" is given to the simulator controller because
basically the system undergoes an initial acceleration in some
direction that must be slowly faded or "washed" out so that the
simulator will not reach the limits of its excursion capabilities.
Afterwards, the simulator, which has changed position as a result
of the acceleration, is slowly allowed to return to a more central
position so that it can have enough room for travel in the next
portion of the flight.
Ideally the system should be designed so
that the whole procedure is as realistic as possible to the pilot.
For example when the simulator accelerations are being washed out
and when the simulator returns to a central position, it should do
so in such a way that the motions involved are below the threshold
of perception in the pilot.
There are a variety of washout systems that are currently in
use in industry today.
Examples of the simpler types of washout
systems in use are systems that employ some sort of accelerationmatching strategy.
For example, the simulator acceleration could
be the actual aircraft acceleration multiplied by a constant gain.
The "clipped magnitude" concept, another example of this type of
washout system, matches the accelerations as long as they don't
exceed a certain limit dictated by the capabilities of the
13
simulator.
One of the disadvantages of these systems is that
they are not self-centering.
Other disadvantages
include the facts
that the proportional drive needlessly limits low displacement high
frequency accelerations; while the clipped magnitude drive unnecessarily limits high frequency high amplitude accelerations that
are still within the excursion limits of the simulator.
More complicated washout systems in use today represent the
simulator motion drive logic problem in the frequency domain and
thus are able to employ techniques which take advantage of what
has happened earlier in the simulation.
By extrapolating knowledge
about rates of change, these types of systems are also able to
make and act upon predictions about what will happen next in a
simulated flight.
In other words these simulators have memory. An
example of this type of system is the linear crossfeed washout,
which uses second order high-pass filters with cross coupling
between translational and rotational degress of freedom.
ing the simulator below the perception level of the pilot,
By tiltthis
type of system is also able to take advantage of the component of
gravity to obtain longitudinal and lateral accelerations and thus
minimize undesired accelerations in these directions (see Figure
1.7).
Another type of washout system is the linear quadratic opti-
mal control system.
In this method, after certain assumptions are
made, a quadratic cost functional, which represents the engineering
tradeoffs being made, is minimized. (Sivan et al.,
1981,
Ish-Shalom, 1982)
1982, Sturgeon,
Other examples include a non-linear adap-
tive washout system that optimizes its control parameters in real
time (Parrish,
1976).
Work has been done on non-linear optimal vwdstou+
14
1970, Kosut, 1979). By using mathematical
filters (Friedland et al.,
models of the vestibular system, which is the human balancing and
motion sensing system that consists of rotational and translational
accelerometers located in the inner ear. Ish-Shalom (1982) designed
an optimal motion base system which takes direct advantage of what
is known about human perception.
FLIGHT DISTURBANCES
PILOT'S OUTSIDE VIEW, FLIGHT
IISTRUMEIJT READINGS
PILOT
AIRPLANE 1ICTIQNS
TA' K
AIRPLANE
--- ------ ---- -- -- -- -- -- -- -- -----FLIGHT SIMULATOR
PT
S IMULATIOI
-- LO
CONTROLS
AIRPLANE
I INDISPLAY
COMPUTATIONS
-------------------
SIMULATOR
UNIT
SIMULATOR MOTloi
OTER
PILOT
DISPLAYS
Figure 1.6: Comparison of Aircraft Flight to Simulated Flight
(Ish-Shalom, 1982)
15
AIRCRAFT SURGE
BODY AXIS
PITCH ANGLE (6)
F
LOCAL
HORIZONTAL
F
'I
VERTICAL
COORDINATE SYSTEM
Figure 1.7:
g
V
FORCE VECTORS
This fi-ure shows how tiltinthe pilot can
cause a translational acceleration
due to
a component of gravit,,.
16
1.3 Problems with-the Optimal Control Washout Design
--What
Should
Be Optimized?
The usual procedure of any optimal control problem is to minimize a cost functional, which is the mathematical representation
of the engineering tradeoffs that are being made.
With the Ricatti
equation and some computational power, one can quickly find the
"best" or optimal solution to accomodate these tradeoffs.
challenging engineering problem is not in applying
The
elegant, but
standard, mathematical techniques in order to minimize a cost functional; it is deciding what tradeoffs should be made and how they
should be incorporated into a cost functional.
For the flight
simulator problem, a major tradeoff exists between how much motion
the simulator is allowed and how realistic the simulated flight
can be.
The cost functional would therefore penalize large motion
excursions and it would also penalize motion that would cause the
simulator to seem unrealistic to the pilot.
However, it is not
clear what to use to quantify simulator realism.
1.4 Problems Associated with Ouantifying Simulator Realism
Realism is a vague and complicated notion.
It is helpful to
realize simulators can and do play powerful "tricks" on the pilot
by purposely stimulating various sense organs in a way that will
make him think he is moving when he actually is not.
For example,
if a person sees a low frequency movement in his peripheral field
of view, he experiences a very strong illusion of self-motion called
vection.
Amusement parks frequently take advantage of this phenome-
non in circular movie theatres by using a wide field-of-view to
give the viewers the illusion that they are moving along with the
17
scene on the screen.
A good visual system alone can be very effec-
tive in making a pilot think he is moving when he is actually stationary, and consequently, it is conceivable that the necessity of
motion, in order to achieve realism, may be reduced. On the other
hand, contradictory signals from the visual system and the vestibular system may produce lack of realism and even motion sickness
(Oman, 1982).
Another reason that the relationship between motion capability
and simulator fidelity is not obvious is because it is not clear
how the degrees of freedom are different and which are most important in providing realistic cues to the pilot.
For example, it is
plausible that human beings, because they spend most of their lives
exposed to a constant 9.8 m/sec2 gravitational acceleration, might
process vertical accelerations in a vastly different way than accelerations in the lateral or longitudinal directions.
Although there
is basic knowledge about how the six independent degrees of freedom
interact, this also is a largely unanswered question.
One possibility for quantifying realism is to use cockpit
motions of the simulator as compared with those of the actual aircraft (see Figure 1.8).
This, however, would require an extremely
expensive motion system or an in-flight simulator such one described
in 1.1.
Another possibility is using an "orientation estimate"
which the central nervous system estimate of the motion after processing signals from the vestibular as well as the visual, tactile,
auditory and other sensory systems.
Although research is currently
being done to find out more about the interaction of these systems,
not enough is currently known to apply this method.
18
TAS K
AIRP LANE
lTori
INERTI.\l
SENSORS
VESTIBULARITACTILE
ORIENTATION
ESTIMATOR
EN SO(BRAIh)
OTHER ORIENTATION-
7 Lj--,,
--l!)
ONTROLLE
A!RPLlAME
IFT 10:;
A19PLA!IE
SENSORS
VISUAL.AUDITORY
coc /r1'
SENSORY
MOT IOI
ONSEASUREMENTS
CK
0,lENTA' ioII
ES
TIMAIloilI
COTROL
EFFORT
M EFORYE.ANCE
*7E14-
COMPUTED
IMPLAXE
'-.0T 10:1
IMUO.A TO
Lg
fEATIAL MOTION
JESTI SE
SrT
ACS
E-
~
YESTCULR. :ACILEORIE11TATION
OTHER
STA(
R
ESTIMATOR
(ORAMN) -
FILTER
--TIN
0---
VISUAL, AU01T09Y
5i
.,:ASPOUT
COMPUTA-
:alENTATION
_fQA*.r
AIRPLANL
aCONTROL.LER.
SENSCRS
AT-rG
4 Lar
A
P
W . ?
~
rj
L
TAS..
I-
- -
-- - -
-
''atching Points to Achieve
Figure 1.8: Comparison of ?ossible
Simulator nealism (Ish-Shaloa, 1982)
19
t
Ish-Shalom (1982) used models of the vestibular system to quantify simulator realism for his optimal flight simulator design.
The cost functional in this method is based on the difference between the physiological outputs of the vestibular system of a pilot
in an imaginary reference airplane to those of the pilot in the
simulator.
This difference is defined to be the vestibular error.
Sullivan (1985) implemented an optimal washout system based
on the vestibular error on the VMS and compared it to the motion
control system normally used for this system.
These systems were
compared in terms of performance in a tracking task, Cooper-Harper
handling quality ratings and simulator motion quality ratings given
by pilot subjects.
His experiments show that the use of vestibular
models is a reasonable approach to the design and evaluation of
motion control systems.
Because one of the primary goals of the simulation is to adequately train pilots, two other possible methods to acheive simulator realism are matching pilot control effort and matching pilot
performance for the simulator and the aircraft.
to these methods is that it
A disadvantage
is not at all clear that just because
the pilot is controlling the aircraft the same way or performance
is measured to be the same that the simulation is realistic.
A
variety of different types of motions could cause the pilot to
execute the same control motion while the same performance can be
achieved by vastly different strategies.
More about all of these
possibilities is discussed by Ish-Shalom (1982).
Sullivan (1985)
found that although performance is widely used as a simulator fid-
20
elity measure, it is not necessarily a valid or consistent measure.
1.5
This Research - More Basic Questions
As stated earlier, there is an implicit assumption that simu-
lator motion is critical for realism and thus for the transfer of
training from simulator to aircraft.
Therefore,
much effort has
been put into the problem of how to best design controllers for
this motion in order to make the simulation realistic.
A major
problem with this research has been that it is not clear how to
quantify simulator realism.
Although some work has been done using
scientific knowledge of the human perceptual systems, many washout
systems today are still being designed mainly by intuition.
This research will examine more basic questions than how to
best control the motion of the simulator in order to make the flight
realistic.
The purpose of this thesis is to examine the relation-
ship between flight simulator motion and realism.
This will be
done by examining the effect of different levels of motion capability
on various parameters of engineering,
physiological realism.
psychological,
and
The variables used to represent realism
are vestibular errors and acceleration errors, as well as pilot
opinion and performance measures.
Vestibular error is primarily a
physiological fidelity measure; acceleration error, an engineering
fidelity measure, and opinion, a psychological measure.
Because
these measures are not independent, it will important to look for
consistencies or inconsistencies in what they show.
There is not
really a good reason to classify performance as a realism measure.
21
As mentioned before different control strategies can achieve the
same performance, so similar performance does not necessarily signify realism.
However, because difference in performance could
signify lack of realism, it will be helpful to examine performance
measures in addition more direct realism measurements.
Three different motion conditions were used to represent the
spectrum of motion capability.
For the lowest fidelity motion
condition, the simulator only consisted of special effects motions
like turbulence.
The amplitude of these motions was small compared
to the other conditions.
In the "jostle" motion condition,
simulator was allowed to move in only two directions -and longitudinal translational.
in this condition.
the
lateral
Special effects were also included
The last motion condition allowed motion in
all six degrees of freedom.
Special effects were included in this
condition also.
Experiments, which are described more fully in the next chapter, were performed on a Boeing 727-200 simulator located at NASA
Ames.
Flight scenarios were chosen that were representative of
the training environment.
These manuevers required significant
pilot control activity so that motion platform effects, if they
existed, and could be measured this way, were as detectable as
possible.
The three flight scenarios chosen were engine flameout
on takeoff, an airwork scenario, and an ILS approach and landing
windshear scenario.
The airwork scenario consisted of steep turns,
approach-to-stall manuevers and standard rate turns with a yaw
damper failure.
The effect of motion on pilot vestibular error,
22
opinion and performance is relevant to all types of washout system
design methodologies, not just the optimal control methodology.
1.6
Thesis Outline
This thesis is organized as follows. Chapter 1 is a general
introduction to flight simulator technology and the problems that
are examined in this thesis.
in chapter 2.
Chapter 3 is a detailed description of the experiment
methodology issues.
in chapter 4.
Opinion results are are presented and discussed
Acceleration error results are presented and disc-
ussed in chapter 5.
chapter 6.
The vestibular models are formulated
The vestibular error results are located in
Chapter 7 is a presentation of the performance results.
Chapter 8 is a discussion about the problems that occurred due to
undersampling of the data.
Chapter 9 deals with conclusions and
suggestions for further research.
23
CHAPTER 2: FORMURATION OF THE VESTIBULAR ERROR AS A MOTION FIDELITY
2.1
Introduction
In the absence of visual cues, the vestibular system is the
primary physiological system used to detect motion. The vestibular
system consists of the semicircular canals, which detect rotational
accelerations and the otoliths, which detect translational motion
(see Figures 2.1 and 2.2).
The vestibular system is particularly
good at detecting high frequency motion, while the visual system
perceives strong motion cues from low frequency motions especially
in the peripheral field-of-view. This report focuses solely on the
vestibular system because it is not clear how the two systems interact.
Also because of the limited travel of the simulator, motion
that is simulated is primarily high frequency motion.
This chapter is a description of the vestibular system.
Mathe-
matical models of the semicircular circular canals and otoliths are
presented,
2.2
2.2.1
and the concept of vestibular error is introduced.
Semicircular Canals
Physical Description of the Semicircular canals
The semicircular canals are used to sense rotational motion with
respect to inertial space.
There are three approximately orthogonal
canals which are filled with a fluid called endolymph. When the head
rotates with respect to inertial space the endolymph, due to its
inertia, initially tends to remain fixed. The semicircular canals,
which form approximately two thirds of a circle, have an expanded
24
section at the base called the ampulla, which leads into a common
base called the utricle (see Figures 2.3,
2.4, and 2.5).
The ampulla
is blocked by the cupula. When the water-like endolymph moves relative to the canals, which is the end result of the fluids initial
tendency to stay fixed in inertial space, sensory hair cells embedded
By increasing their firing
in the base of the cupula are bent.
frequency, these hair cells provide information about the motion
to the central nervous system.
The hair cells in a particular
semicircular canal are all polarized for sensing motion in the
same direction. So each canal is used for a particular direction.
Theoretically,
by combinations of the information provided by the
three orthogonal canals, the CNS can detect motion in any direction.
There are two types of hair cells in the canals - type I and type
II
(see Figure 2.6).
Type I cells,
which have an irregular dis-
charge pattern in the afferents and show adaptation,
may be sensitive
to both hairbendings and the rate of change of hairbendings (Hosman
and Van der Vaart, 1978).
Type II hair cells are believed to pri-
marily sense hair bendings.
2.2.2
Mathematical Model of the semicircular canals
The endolymph is acted on by primarily two forces.
The first
is viscous drag, which is the proportional rate of velocity of the
fluid with respect to the walls of the canal. The second is modelled
as a linear elastic restoring force that is caused by the cupula's
tendency to return to its
resting position.
The equations below
represent a basic but incomplete model of a canal and were first
proposed by Steinhausen in the 1930's
25
(Ormsby, 1974 ).
as
I#
m(ec + eci) -
9ec
-V 8ec -k eec
angular deflection of endolymph with respect to canal
angular position of canal with respect to inertial space
m - moment of inertia of endolymph
V - coefficient of viscous drag
k - coefficient of linear restoring force due to displacement
of fluid in the canal
eci
-
-
The system appears to be overdamped and k/V << V/m this equation
can be approximated as:
ec(s)i)-l
c i(s )(s+k/V) (s+V/m)
This model is called the overdamped torsional pendulum model.
Recent
efforts have included a lead term to represent the sensitivity of
type I haircells to the rate of cupula displacement. Based on studies
of the squirrel monkey, Fernandez and Goldberg (1976) estimated
the lead time constant to be 0.049 sec.
Another addition to the
basic torsional pendulum model has been included because adaptation
to prolonged accelerations was noticed.
Adaptation could be due
to processing by the CNS; or it could be attributed to adaptation
in the sensory haircells; or it could be a combination of these
effects.
The adaptation term, which is due to the efforts of Young
and Oman (1969), causes a phase lead and gain attenuation at low
frequencies.
In any case the full transfer function which now
relates spatial orientation to angular acceleration of the head
for one degree of freedom can now be written as:
H(s) -
Gscc
.
Tas
(1+Tas)
26
(l+TLs)
(1+Tls)(l+T2 s)
The output is measured in threshold units so that a response
2
of one is obtained for a sinusoidal input of amplitude 1.450/sec
and frequency of 0.94 rad/sec as found by Hosman and Van der Vaart
(1978), who also obtained T1 -5.9 sec.
The adaptation term has a
value of Ta- 80 sec (Young and Oman, 1969).
T2 has such a high
frequency that is difficult to measure. Theoretical estimates by
Steer (1967)
estimate this value to be 0.005 sec (Hosman and Van
der Vaart, 1978).
Bode plots are shown in Figure 2.7.
In this report, this model is limited by the sampling rate of
the flight simulator computer, which was 30 times per second, so
the highest frequency in the model should be 15 Hz. Therefore, to
prevent undersampling, the
lead term and the high frequency pole,
with frequencies of 20.4 hz and 200 hz rspectively,
inated.
have been elim-
In this report, the following transfer function was used
to represent a semicircular canal:
H(s) - Gscc
.
Tas
(1+Tas)
.
I
(1+Tls)
Gscc - 222.7 sec 2 /rad
- 80 sec
Ta
- 5.9 sec
Tj
This transfer function is used for each of the three degrees
of freedom.
Bode plots are shown in Figure 2.8.
27
M~t~ZU
Figre
.1
Th
stucureoftheinnr
I
ar.(Peeram169
~
0
0
Figure 2.2 Orientation of the seimicircuiir canals and
the otoliths with respect to the head
(Jongkees ,1967)
28
Cristo
Ampulla
Membranous Horizontal
Semicircular Canal
1.1 mm
Utricle
6.4 rmm
0.24 mm
Figure 2.3:
Average dimensions
(Peters,1969)
of Semicircular Canali
x
y
Superior Canals
Utricles
Vectoril
indicate
effective direction of
angular .cceleration
Left
Coa
l
Right
Horizontal Canal
Posterior Canals
Fi&LARE2,41
The effective directions of the semicircular canals
(approximate) (Peters,1969)
29
CELLS
-SENSORY
Figure 2.5: Simplified diagram of ampulla section
of semicircular canal (Peters, 1969)
Type
~fL
I
S
heir mm
Efend.,
n
*
r
LA\.IWA,.W
O*
b
VA
Figure 2.6: Schematic drawing of type i and type
Hair cells
(Wersall et al,
30
1971)
I
SEMICIRCULAR CANAL MODEL FREQUENCY RESPONSE PLOTS
MAGNITUDE
10
2
k
I'Prs
(
10
IC
1s)(rice
(I +TI S)
10
1o0
oL
-3
,o-2
1
100
u)
CrodI)
101
)
10,
100.
so.[
0.
(de9 vee")
-
-So.
-100.'
Id j-3
Figure 2.7:
...........
-2
10
i
100
Ctad
/Sec
102
Frequency Response Plots of Semicircular Canal Model
31
10
10
4
-2
FREQUENCY
SCC MODEL
CI2
0t1
-1
100
RESPONSE
123
100.
50.
-o..
-100.,
220i
16-1
100
io1
102
13
Figure 2.8 Frequency Response Plots of Semicircular Canal
Model Used in this Report.
32
Otoliths
2.3
2.3.1
Physical Description
The otolith organs, also located in the inner ear, are used
to detect translational accelerations. This includes gravity, which
is a translational acceleration of 9.8 m/sec 2 .
There are two oto-
lith organs located in each inner ear. One of them is located in the
utricle, which is the common base of the semicircular canals.
The
other otolith organ is located in the saccule, which is a downward
extension of the utricle.
The basic structure of the otolith con-
sists of a supporting base called the macula (Figure 2.9).
Covering
the macula is a gelatinous layer containing suspended calcite crystals.
When the head is exposed to a specific force, the calcite
crystals have more inertia than the gelatinous layer, and the layer
shears. Sensory haircells embedded in the macula are bent and a
signal is sent to the central nervous system.
The otoliths have
the same two basic types of haircells that are located in the semicircular canals; however,
unlike the canals,
are polarized in different directions.
hairs in one organ
This enables motion sensa-
tion in the three degrees of freedom when there are only organs in
two approximately orthogonal directions.
2.3.2
Mathematical Models of the Otoliths
An otolith organ can be modelled as a overdamped spring-mass-
damper system with the following transfer function, which describes
translational acceleration perception to specific force.
33
H(s) -
K
(l+T3 s)(l+T4 s)
Modifications proposed by Young and Meiry (1968) include a lead
term which takes into account adaptation.
H(s) -
K(1+Tns)
(1+T3 s)(l+T 4 s)
Tn
T3
T4
K
- 13.2 sec
- 5.3 sec
- .67 sec
- gain factor
In this report, approximations were made to prevent undersampling.
The transfer function actually used was:
H(s)
-
Go(l+Tns)
(l+T3 s)
Go
Tn
T3-
-
2.13 sec 2 /m
13.2 sec
5.3 sec
Threshold units were used to give a response of 1 when subjected to accelerations of 0.47 m/sec 2 as discovered by Hosman and
Van der Vaart (1978). Figure 2.10 shows other experimental estimates
of the threshold level of perception.
In this report, all
of freedom are represented by this model.
degrees
Bode plots for one degree
of freedom are shown in Figure 2.11.
Time constants in the otolith model are not as definite as in
the semicircular canals.
This is partly because of difficulties
in the experimental tools used.
For semicircular canal research,
subjective experiments can be performed in a rotating chair with
34
well-trained subjects.
For otolith research, a smoothly running
translational car is needed (Hosman and Van der Vaart, 1978).
2.4
Vestibular Error Measurements
Based on the vestibular models described above, one can cal-
culate a response in threshold units.
Because the simulator under-
goes different accelerations than the actual airplane, the simulator
vestibular response will be different from the vestibular response
that would occur in an actual aircraft (see Figure 2.12).
The dif-
ference in these two responses is defined as the vestibular error.
The input accelerations to the vestibular must be in cockpit body
axis coordinates.
Because it was desired to compare average magnitudes of these
errors and not to compare functionality, a root mean square error
was computed for specific data windows of interest.
will be discussed further in Chapter 3.
35
This measure
STATOCONIm
GEAT:..NOW
'~v
(Peters,
1969)
36
50 33 25 171
100
0.1
0.08
0.06
10
PERIOD, SECONDS
5 3.3
2
1
0.5
0 05
0.25 0.17 0.1
F I iI
I
0.02
.
2
.
0.04
Nd
ooZ
.
--
-
L-
0.01
p
CD
--
0.005
0.008
0.006
3
(06 3 4 1116
0.004
0.003
0.002
04:4p
0.001
0.0007
0.0005
0.0003
0.006 0.01
0.02 0.04 0.06 0.1
Fig ure,.-2 .10
0.2
0.4 0.6 1.0
2
FREQUENCY, Hz
3 4 5 6 8 10
20 40
60
80
Threshold-of-Pe'rception Measurements and a Design Limit ror
Spurious Accelerations
(Taback
--
Table 2.1
SUIARY OF THPESHOLD-OF-PERCEPTI
1983)
ON
(from fig. 2.10)
DATA SOURCES
(-r~gAB
ci, \9
DATA
POINT
SOURCE
1
Chaney 1(4)
Vertical Axis Shiaker
10 males, seated with lap belt restraint and footrests.
Each subject makes 4 determinations at 4 frequencies
2
3
4
Chaney 1(5)
Chen and
Robertson 3
5 males, standing, feet attached to moving platform.
5
Chen and
Robertson 3
Vertical Axis Shaker
Closed room on a
platform with 2-.
axis
horizontal motioon
Closed room on a 25m
pendulum
Goril and
Vertical Axis Shaker
TEST RIG
TEST SUBJECTS/ACCOMMODATIONS
ACCELERATI
DIRECTIONS
TEST SUBJE
Z-Z
(8 measurements for each frequency data point).
6
7
8
9
Snyder 1(8)
Gurney 1(12)
10
11
12
Landsberg 1(18)
Richer and
Meister 1(22)
13
von Bekesy 2
14
Walsh 1(24)
15
16
17
18
19
2 subjects at each frequency, standing.
Z-Z
3 X-X
4 Y-Y
10 subjects sitting.
10 subjects sitting.
5 X-X
6 Y-Y
20 subjects standing.
7 X-X
6 males (air crew) seated in a cockpit simulation.
Vertical circula r are 3 subjects sitting blindfolded (seesaw). .
3.26m radius
3.6m Pendulum
Subjects lay face-down and face-up.
Platform drivenI by
10 subjects in five positions; Standing, X-X, Z-Z;
eccentric mass
lay face-up X-X, Y-Y, Z-Z. (12 is X-X face up only.)
vibrator
2m Horizontal
2 subjects seated
circular arc
1.55m Pendulum
4 to 7 subjects each test point. Lay face-up, facedown, on sides. 8 measurements for each subject.
Z-Z
Z-Z
Z-Z
11, All
12, X-X
y-y
14, X-X
15, Y-Y
16, Z-Z
Walsh 1(26)
Benson
Bionetics
at Miami U.4
Vertical circula r arc 7 males, 4 measurements each, 8 times. Lay face-up.
Vertical hydraul ic
6 males, 4 females in an aircraft ejection seat.
driven oscillato r4m Pendulum
2 males lay face-down, series of measurements at one
frequency.
37
X-X
Z-Z
Y-Y
OTOL ITN
Re Pd
FPGQUgw-E
10.
N'S
- POT'
. .
1-'
Malj It-
1'
10
7
10~1
lou
w
C.raA/,A..
30.
p 1.t
20.
10.
0.
1 0~4
10~1
100
10
14 (rad/sfr)
Figure 2.11:
Otolith Model freouency response plots
38
MDELLED
AIRCRAFT
TRANSLATIONAL
ACCELEXATIONS
MDELLED
AIRCRAFT
MDEL Or
OTOLIN ORCAN
ROTATIONAL
MDEL Or SEMICIACULA"
ACCELERATIONS
CANAL
ORGAN
TAWLATIOUAL
VZSTISULAA
Eacas
t
ROTATIONAL
ACCELEZATION
SIMULATOR
TRANSLATIONAL
ACCELERATIONS
MOEL OF OTULITN
ORGAN
SIMULAroft
ACCgLUT
ONS
i
MODEL OF SEMICIRCULAR
CANAL
OR.AN
Figure 2.12:
Calculation of Vestibular Error
39
CHAPTER 3: EXPERIMENTAL DESIGN AND ANALYSIS MEASUREMENTS
3.1 Brief Introduction
This chapter is a summary of the experimental procedure.
A
description of the experiments is presented first and is divided
into sections about the simulator, motion conditions, flight scenarios, subjects, and order effects.
The second part of this chapter
discusses the actual measurements taken and and the third section
is a discussion of data collection.
3.2.
Experiment Description
3.2.1
Simulator
Experiments were performed with a Boeing 727-200 series flight
simulator on a six degree of freedom synergistic motion base.
uses a nonlinear adaptive washout system.
It
The visual system used
for this study is a computer-generated dusk/twilight scene.
This
simulator meets the requirements for Phase II certification under
Federal Aviation Regulations.
The simulator,
which is designed by
the Singer-Link company, is located at the NASA Ames Research
Center.
3.2.2
Figure 3.1 is a picture of the actual aircraft.
Motion Conditions
Three motion conditions were used.
In the "fixed-base" con-
dition, only special effects motion was allowed.
included runway touchdown bumps,
Special effects
vibrations due to the roughness of
40
*O-rr
i
dw.
Figure 3.1:
The Boeing 727-200 series Aircraft
41
the runway, buffets associated with flap, gear and spoiler extension, turbulence, and Mach and stall buffet.
The special effects
only motion condition is the most severely limited condition in the
study.
The amplitude of these vibrations is small compared to the
motion capabilities in the other conditions.
"Jostle " motion
condition provided only two degrees of freedom: vertical translaThis condition, which also in-
tional and lateral translational.
clude all of the special effects mentioned above, was included to
study the effects of providing mostly translational acceleration
information to the pilot.
The last
motion condition, full motion,
provided full six degree of freedom capabilities as well as the
special effects included in the other conditions. This condition
is the nominal condition for the simulator.
3.2.3
Flight Scenarios
The environmental data base used in the simulations was the
area of the San Francisco International Airport.
takeoff weight was set to 148000 lbs.
The aircraft
Although flights were short
so the fuel burn that would occur is slight, the weight was held
constant throughout the experiments in order to eliminate the
effects of weight changes due to fuel burn which might vary due to
different piloting techniques.
All scenarios, with the exception
of the ILS Approach and Landing Scenario, were conducted in the
standard day (pressure - 1 atm, temperature- 150 C) with no wind
and good visibility.
42
Familiarization Scenario
3.2.3.1
This scenario was flown by all subjects directly before beginning their series of experimental flight scenarios.
The purpose
of this scenario was to allow the pilots to get used to the simulator.
Because the purpose of this study was to compare solely
the effects of motion fidelity levels on a variety of parameters,
allowing the subjects to become familiar with the simulator hopefully reduced the possible learning effects.
(In similar experi-
ments Sullivan (1985) found learning effects on some performance
measures.)
A familiarization scenario run began at the departure end of
runway 28R at the San Francisco International Airport.
The pilots
followed air traffic control vectors around the traffic pattern
for a visual approach to a touch-and-go landing.
The subject then
followed the vectors around the traffic pattern for another visual
approach to a full stop landing.
Subjects then had the choice of
performing this scenario again.
3.2.3.2
Engine flame-out on takeoff
This scenario began with the aircraft ready for takeoff at
the end of runway 28R of the San Francisco International Airport.
The subject was informed in advance that there would be an engine
flame out in one of the off-centerline engines (#1 or #3).
At the
beginning of each run, the experimenter randomly chose which engine
would fail.
Randomization was used to prevent the subject from
making anticipatory control motions.
The time of engine failure
was also varied but always occurred within five seconds following
43
rotation, that is beyond V2.
The engine flameout occured after V2
so that the pilots could not decide to abort the takeoff.
Subjects
were instructed to maintain runway heading and level out at an
altitude of 2000 ft.
3.2.3.3
Airwork Scenario
The Airwork Scenario, run began with the aircraft at 15000 ft
with 250 knots indicated airspeed directly above the San Francisco
International Airport with a heading of 2800.
The pilot was in-
structed to perform two consecutive steep turns 450 bank, one to
the right and one to the left, to make one "s" turn, two approachto-stall manuevers, and then two standard rate turns with the yaw
dampers failed.
3.2.3.4
ILS ADproach and Landing Scenario
Pilots flying the ILS Approach and Landing scenario were initialized at an altitude of 4000 ft,
an airspeed of 220 knots indi-
cated airspeed, and an intercept course of 300 off the localizer
to SFO's runway 28R.
levels.
ft.
The simulator was set for moderate turbulence
There was 600 ft ceiling and unlimited visibility at 500
A windshear, described in Figure 3.2, was introduced at this
altitude.
Pilots were instructed to land the aircraft and were
informed of the presence of windshear.
44
ALTTUDE
* 1000 ft
900
ft
. 800
ft
700
ft
600 ft
500 ft
2
10 kts
FIGURE 3.2:
5 kts
i
i
i
0Okts
WINDSPEED
- 400
ft
- 300
ft
ft
-
200
.
100
ft
2
ft
--- Aid
10 kts
5 kts
(K-+S)
WINDSHEAR MODEL FOR ILS APPROACH AND LANDING SCENARIO
(Arrow-3iindcult direci.on of
oircraf+).
to incm
45
mied4
reIQLi S/ e
3.2.4 SubJects
Eighteen air transport pilots were used in the study.
The subjects were told that the experiment was a study on flight
simulator fidelity.
They were not told before the experiment that
the experimental conditions would vary only the motion platform
capabilities.
Briefing material given to the pilots before the
experiments is located in the Appendices.
The primary subject chose his preferred seat, left or right and
flew as pilot-in-command, and the secondary subject flew secondin-command.
Data was taken on the first subject and then the levels
of command were switched,
the new subject chose his seat and data
was recorded again. Due to schedule changes, the second-in-command
seat was sometimes taken by a non-subject pilot.
To minimize the
effect on exposure to another subject's experiment which might
induce learning effects, subjects who flew in groups of two were
given different flight scnearios to perform.
The eighteen
pilots were randomly divided into groups of
three so that six subjects flew each of the three flight scenarios.
Pilots 4, 6, 7, 12, 16, and 18 performed the ILS Approach and
Landing in wind shear scenario.
Pilots 2,
5,
10,
11,
performed the engine-flame out scenario, and pilots
13, 15
performed the airwork scenario.
14,
and 17
1, 3, 8, 9,
Before performing the
actual experiments, all subjects flew the familiarization scenario
which consisted of
VFR takeoffs, approaches and landings.
46
3.2.5
Order Effects
There are six permutations
motion conditions.
of possible orders to present the
Each subject flew a different order so that
ordering effects would be minimized.
In all motion conditions,
normal procedures for full six degree of freedom were conducted
before the experiment so that the pilots would not notice testing
motion base differences before the experiment ocurred.
motion condition, the subjects filled out a
After each
questionnaire which
was used to provide pilot opinion data.
In the debriefing session,
subjects were able to make addi-
tional comments pertaining to the flight conditions.
session, they were told what the conditions were.
During this
Notes on their
comments are provided in the Appendices.
3.3
Analysis Measurements
The following sections will describe the measurements taken
and examined.
These measurements are categorized into opinion
measurements, acceleration error measurements, vestibular error
measurements,
and performance measurements.
Opinion Measurements
3.3.1
After each motion condition,
questionnaire,
the subject filled
which asked for the subjects comparison of certain
aspects of the condition to corresponding aspects
aircraft.
out an opinion
in an actual
The actual questionnaire is reprinted in Figure 3.3.
There were six questions on the questionnaire.
47
The first
two ques-
tions had to do with the pilots opinion of the workload levels of
the simulator as compared to the actual aircraft.
Workload levels
may interfere with the ability of pilots to perform well especially
in long duration flights or critical phases of flight.
concept is an important one that should be considered in
training.
So this
pilot
Because one of the primary purposes of flight simulators
is for training aircraft pilots, these questions were included in
the study.
The first question asked the pilots to compare the
demand on attention, skill and effort required to control the simulator to what it would have been in an actual aircraft.
question pertained to the entire flight scenario.
The
This
second
question, on the other hand, was restricted to times when the simulator was undergoing configuration changes.
This question also
dealt with their demand on attention, skill or effort.
In the third question the pilots compared the response of the
simulator to control efforts during the entire flight scenario.
This question was aimed at their direct opinion of the simulator
itself; whereas, the first two questions were aimed at their levels
of effort.
The answer range for the simulator response question
was from "much slower than aircraft"
to
"much faster than air-
craft."
The fourth question required the pilots to compare the value
of the simulator to the aircraft for the purpose of training pilots.
The value of the simulator for pilot checking was compared to the
value of the actual aircraft for pilot checking in question five.
The sixth question asked the pilots to rank the simulator in terms
48
of overall realism.
The answers ranged from major deficiencies to
no deficiencies at all.
The opinion results were analyzed two ways.
For the first
method of analysis, for each question and each motion condition,
the mean over all pilots performing the same flight scenario was
These means and the standard deviations associated with
computed.
These plots show the differences on average,
them were plotted.
if there are any, between motion conditions in the pilots opinion
to the specific question asked.
The second method of analysis was an analysis of variance
statistical test used to determine if the differences found were
"statistically significant"
.
The answers to each question were
analyzed separately by an F-test, which assumes random sampling of
the data,
equal condition variances,
the data.
and a normal distribution of
The null hypothesis of these F-tests is that for a parti-
cular question, the mean answer is the same for all three motion
conditions.
The F-test gives a level of significance for rejecting
this hypothesis.
It does not confirm the null hypothesis.
mechanics of the test are to compute the
The
"among groups " mean
square, which represents the variance among motion conditions in
the answer to the particular question,
to the "within groups"
mean square, which represents the unexplained variance in the
answers to this questions.
variances.
The F-ratio is the ratio of these two
Probabilities of obtaining an F-ratio larger that than
the one obtained are computed. If the F-ratio is large,
in other
words the among groups mean square is larger than what would be
estimated by chance
(which is estimated by the within groups mean
49
square),
then the null hypothesis can be rejected at a level of
significance defined as the probability of obtaining a higher Fratio.
SIMULATOR
CONDITION NUMBER
During the entire flight scenario, the demand on attention, skill,
or effort reouired to control the simulator was:
1.
1----------------
-4---------5
2.
much more
than aircraft
very similar
to aircraft
much less
than aircraft
During aircraft configuration changes, the demand on attention,
skill, or effort required to control the simulator wass
1------
much less
than aircraft
------------
very similar
much more
than aircraft
to aircraft
During the entire flight scenario, the resoonse of the simulator
to control inouts was:
1-----------------------------------4--much slower
tian aircraft
4.
eouivalent
much more
than aircraft
to aircraft
The utility of this simulator for allot c.-.cking is:
I
much less
tman aircraft
6.
5
The utility of this simulator for pilot t.-aining iss
much less
than aircraft
5.
---
much faster
than aircraft
very similar
to aircraft
2------------
eouivalent
much more
than aircraft
zo aircraft
The overall realism of the simulation as
comoared
to the aircraft
4----
---
was:
I
major
deficiencies
Figure 3.3:
2---------- -------some minor
deficiencies
no
deficiencies
Pilot Opinion Questionnaire
50
5
Acceleration Error Measurements
3.3.2
Because the simulator had limited travel, simulator accelerations were different from those that would actually occur in the
real aircraft. This difference is defined to be the acceleration
error.
We were interested in comparing the magnitude of this error
for the three motion conditions.
This study does not compare dif-
ferences in the functionality of the data.
Therefore,
degree of freedom of interest in a particular scenario,
for each
a root
mean square error was calculated for a data window of interest
(see equation below).
The particular degrees of freedom and data
windows for each flight manuever are discussed in the data collection section of this chapter.
rmsaj
-
((ai ac(k) - ai sm(k))2
rmsai - rms acceleration error in the ith direction
ai ac(k) - aircraft acceleration in the ith direction
for the kth measurement in the data window.
ai sm(k) - simulator acceleration in the ith direction
for the kth measurement in the data
n = number of measurement points
Two methods were used to analyze the effect of motion base
differences on acceleration error measurements.
specific
performing
In the first, for
degrees of freedom and motion conditons, means over pilots
the same maneuver were computed.
Comparison plots of
the mean rms error values and the standard deviations of these
means were plotted.
The second analysis method of this data was
an F-test for which the null hypothesis was that the mean rms error
value was the same for all three motion conditions.
51
3.3.3 Vestibular Error Measurements
A root mean square for the vestibular error was also calculated
to obtain information about the relative magnitude of the errors
for different motion conditons.
(see equation below).
The root
mean square error is computed for the particular data windows of
interest described in later sections of this chapter.
rmsyi -
J
((Yi ac(k) - yi sm(k))2)
rmsyi - rms vestibular error in the ith direction
Yi ac(k) - aircraft vestibular response in the ith
direction for the kth measurement
in the data window.
yi sm(k) - simulator vestibular error in the ith
direction for the kth measurement
the data window.
n = number of measurement points
The vestibular error was analyzed two different ways.
different degrees of freedom were analyzed separately.
The
The first
analysis method consisted of averaging the rms error for a particular degree of freedom and for a particular motion condition for
over a set of pilots who had performed the same scenario.
Motion
effects were studied by comparison plots of the means and standard
deviations of the means for the three different motion conditions.
The second analysis method used was a statistical analysis of
variance test to examine whether or not the differences associated
with the plots and the data that makes up the plot is statistically
significant. The null hypothesis of these F-test were that the
mean root mean square errors (over the set of pilots performing
the scenario) were the same for different motion conditions.
These
tests assume random data samples, equal condition variances, and
normal distribution of the measurement errors.
52
Similar to the
absolute acceleration error measurements, only certain degrees of
freedom were studied for each particular flight maneuver.
These
are listed by flight maneuver in the data collection section of
this chapter.
3.3.4 Pilot Performance Measurements
This section describes the pilot performance measures examined
for each flight scenario.
3.3.4.1 Engine Flame Out Scenario
The mean aircraft centerline deviation over all pilots performing the scenario was computed and compared for each motion
condition.
This measure was computed for increments during the
data window of interest, which is described later in the chapter.
Centerline deviation was considered an important parameter to study
because loss of power in an off-centerline engine will cause the
plane to yaw as a result of an unbalanced torque about the aircraft's center of gravity.
The other measurement sudied in this scenario was the time to
climb to a safe altitude of 400'
from a speed 120 kias.
This
measure was computed for each subject for each trial for each of
the motion conditions.
3.3.4.2 Airwork Scenario
The mean variance of aircraft pitch and bank for the data
window defined for stall was computed for each run.
These numbers
for each motion condition were averaged over all runs.
53
The mean variance of aircraft pitch and bank for the data
window defined for stall was computed for each run.
The numbers
were averaged and compared for each motion condition over all runs.
3.3.4.3 ILS Approach and Landing Scenario
During the approach segment of this flight, which will be
defined later, mean glideslope and localizer deviations were computed for each run.
The average deviations are compared for all
pilots for each motion condition.
During the landing protion of
this scenario, which will be defined late, the mean sink rate and
lateral deviation are studied.
3.4 Data Collection
The following sections describe the data variables that were
collected, specific degrees of freedom that were analyzed for the
acceleration error and vestibular error measurements, and pertinent
data windows that were studied for each flight manuever.
This
section then describes problems that were associated with the data
collection and data processing.
These problems included under-
sampling the rotational degree of freedom acceleration measurements
as well as lost data.
3.4.1 Variables
This section contains a list, table 3.1, of all of the variables and how often they were collected.
were used in the simulation software.
54
The names associated
The experiment frame count (jtimer)
time into the experiment
(30 x jtimer -
was used to calculate the
Jwindow and jintc
time).
were not used for this study but were included as an aid to data
collection and processing done at NASA Ames.
The array JYSCC con-
tains the outputs of the semicircular canal model for both the
simulator and the aircraft model.
This data was not used at all
because the incorrect model was used (see data collection problems).
Instead these responses were calculated at MIT.
The jyoto array
contains the otolith model outputs in threshold units for both the
simulator and the modelled aircraft.
the vestibular error.
These were used in computing
The modelled aircraft accelerations in all
six degrees of freedom are shown in table 3.1 as are all six simulator platform accelerations.
The other variables in the list are
those that are used for performance measurements or for identifying
data widows.
Some are pilot control variables which are not in-
cluded in this report.
55
TABLE 3.1: DATA VARIABLE LIST
(ALL VARIABLES CALCULATED AT 30 HZ)
COLLECTION RATE (HZ)
w 1
;
&
IW
IE
rW
JW INDOW
r-INTC
+ c00
S r
iJYEcC
+00±
rw w EW
EW
+003
EW
.L .1.r A
EU
-
JY C
J,YrOT 0
J3 YOT 0
FDPA
FDQA
FDRA
JAANA
J.AAYiA
+00±
+0
0
. A :-. A,
EW
CUJ
+AA.
+004
EW
.LAr A =
EU
+005
EW
+000
+000
+0A
+000
+000
J MA '.r A
JMD
A r4-r
JMAZA
-L-7 A
JMDPA
JMDRA
N GC-
NGCCE I
FWE
FHGEL
FtVIA S
FTHETA
FPHI
NPSIGA
FESPOS
FSWPCS
NGCAGC
EPR±
EPRI
EPR
FRUDT
+A
+000
+000
+000
+000
+000
+000
+000
+000
+0.115
+000
+.LA ^I
+000
+000
+000
+000
+000
+000
+000
+000
+000
EW
EW
EWU
EW
*
EWU
EW
EW
EWU
EWU
EWU
EW
EW.
EWU
EWU
EW
EU
EU
EWU
EW
EU
EWU
EWU
EW
EWU
EWU
-15
EXPER IMENT FRAME C 1U.NT (30H)
ACTIVE WINDOWE (INTEGER BYTE ARRAY-4
S
15
%SINCE MIDNIGHT -0 HZ TICK
CLCCK TIME MIT
CC M1DL OUTPLT A/C
TLL
ACCEL
- --15
MIT 5CC M 11D7E L 0I ITPLIT ACPITCH ArcCEL-1
15
NIT SCC MODEL OUTPUT A/C YAW ACCEL
T15IL;
A
1
ACCEL-MIT SCC MODEL OUTPUT MOTION ROLL
MIT SCC MODEL OUTPUT MOTION PITCH ACCEL----- 15
MIT 5CC MODEL OUTPUT MOTION YAW ACCEL T1LITH MODEL OUTPUT A/C LONG ACCELMIT
MIT OTOLITH MODEL OUTPUT A/C LAT ACCEL'
MIT
TLITH MDEL
UTPUT A/C VERT ACCEL
15
15
MIT OTOLITH MODEL OUTPUT MOTION LONG ACCEL- MIT OTOLITH MODEL OUTPUT MOTION LAT ACCEL - - 1
...E15CE- -- 15
1 ACCEL
MIT OTOLITH MODEL M~UTPUT MOTION VERT
A/C AXIS ROLL ACCELERATION ----s
A/C
AXIS
PITCH
ACCELERATION
-
MIT MOTION PITCH ACCEL I
MIT MOTION YAW ACCEL
N--TRUE GROUND SPEED- --
LEFT HY~D BRAKE
GCA G/
TORQUE
VERT VELOCITY
- -
HEIGHT
INDICATED
AIRSPEED
A/C
ANGLE
PITCH
A/C BANK
A/C TRUE
ATRETCHED
STRETCHED
-
AEL
ANGLE (DEG)HEADING
(DEG)-
STICK
WHEEL
RANGE
5E_
E
ENG INE
TOTAL RUDDER
--
--
--
-
TO TD
--
--
-
--
-
-
-
(DEG-
-
--
--
---
-~~
-~~
----
--
-
-
~~
---
--
-
1
1
1
1
-
-
-
--
-15
15
1
15
15
-
-~~
-~~
__
--
-
-
(DEG
-
--
--
--
(DEG-.
--
1
1
1
-.
-
--
--
(FT)--
--
56
- -
1
1
-
-
-
-'
-
-
-
-
-
-
-
--
-
-
-
15
1
1
1
--
-
-
-
- 15
--
-
-
-
-
-
-
-
-
- 15
-
--
-
--
-
-
15
--
-
--
(FTp-
POSITION
POSITION
ENGINE I EP --- -CENIN
--
--
-
-
--
-
-
-
-
-
-
-
(KNTS)--
(D:)
-
-
15
15
15
-
-
-
(FT-
-
-
EL AIrMSAE-ANI
GEMETRIC
15
-
15
-
-
ANG3LE -
COURSE DEVIATION
-
-
-
-
LAT DEVIATION
-
-
-
FORCE+±6-
FT-
DEVIATION
GCA BGAS
-
-
MIT MOTION ROLL ACCEL IN-
AGS
EW
EW
--
A/C AXIS TURN ACCELERATIN
MIT A/C LNG ACCEL INMIT A/C LAT ACCEL IN
-MIT A/C VERT ACCEL INMIT MOTION LONG ACCEL INMIT MOTION LAT ACCEL INMIT MOTION VERT ACCEL IN-
-
-
1
3.4.2 Trials
There were two runs per motion condition per subject for both
the engine flameout scenario.
Because of the length of the airwork scenario, the subjects
flew only one run per motion condition.
This means that for the
steep turns and the rate turns with yaw dampers failed, there is
only one data measurement per subject per motion condition which
is the transition between two turns.
Every other flight manuever
in the study consists of two data measurements per subject per
motion condition.
The ILS Approach and Landing Scenario had two runs per subject
per motion condition.
3.4.3 Pertinent Degrees of Freedom for Each Scenario
Calculations were not computed for every degree of freedom for
every flight manuever. Interest was primarily in the degrees of
freedom in which significant motion occurred.
In the engine flame-
out scenario, interest was mainly in the yaw rotational
and the
lateral translational directions. In the airwork scenario, for the
approach-to-stall manuever, we were interested in roll, pitch, and
longitudinal directions.
For both the steep turn and the rate
turns with yaw damper failure manuevers, we were interested in all
three rotational degrees of freedom as well as the lateral translational degree of freedom.
In the ILS Approach and Landing scen-
ario, all six degrees of freedom were of interest.
Problems des-
cribed in the data collection caused us to lose the lateral axis
57
measurements completely as well as some of the longitudinal absolute
acceleration error measurements.
Data Windows
3.4.4
One of the reasons the flight scenarios were chosen was because
they were typical tasks required in a training or checking environment and they required significant amounts of pilot control effort,
as compared to cruise or more standard aspects of flight.
For
this study, only the more critical portions of the flight, which
required significant control activity, were considered in the motion
condition comparisons.
For the engine flameout scenario, for
example, the portion of flight that occurs when the
pilot has
realized the problem and made most compensation decisions is of
primary interest.
A short time period directly before the flameout
is also of interest because it can have direct effects on the period
after the problem has occured.
For these reasons, a data window
of ten seconds before the engine flameout and ten second afterwards
was chosen as the data window of interest for this particular scenario.
For the two successive steep turns, there was one data window
during the transition between the two turns.
This window ocurred
at ten seconds before and after the wings were level (zero bank
angle).
For each of the two approach-to-stall manuevers, data was
analyzed 10 sec prior to and following the lowest indicated airspeed
attained. For the standard rate turns with failed yaw dampers,
data was taken at the transition point between the turns: again
defined as 10 sec before and after zero bank angle.
58
In the ILS approach and landing scenario, two data windows were
examined per run.
One was between approximately 500'-200', which
The second,the landing segment,
was the approach segment.
approximately 20-25 seconds before touchdown.
was
Numbers, defining
the data windows, are approximate because altitude measurements
were taken only every second.
3.4.5
Data Collection Problems
3.4.5.1 Possible Aliasing of Rotational Degrees of Freedom
For the translational degrees of freedom, the correct vestibular model was included in the simulator software.
experiment,
at the rate of 30 times per second,
responses were calculated.
the vestibular
They were stored at 15 hz.
the amount of magnetic tapes that were needed.
experiment,
'So during the
the high frequency model (TL -
.05
to save on
At the time of the
sec)
in the semi-
circular canal model, was included in the simulator software and
semicircular canal responses were calculated.
in this report because it
were in error.
They were not used
was possible that these calculations
In actuality this data is correct, and analysis is
currently being done on it.
For this report, the semicircular
canal response used for analysis was calculated using the input
angular acceleration data that had been collected at 15 hz., and
using a model which eliminated the high frequency zero (See chpt2.)
The main problem is that the simulator only limited motions
that had a frequency content higher than 15 hz.
semicircular canal calculations
Because the new
that were being done at 15 hz,
59
high frequency motion
(> 7.5 hz) would be aliased.
This means
that they would look like low frequency motions.
Power spectra are shown for the relevant degrees of freedom
for a typical run of each scenario to get an estimate of how much
aliasing had occured. These are presented and discussed in Chpt. 8.
3.4.5.2 Lateral Axis Data
All lateral axis data, both absolute and vestibular response
data, was lost.
This is because of an error made in the simulator
software in converting from center-of-gravity to cockpit coordinates.
3.4.5.3 Longitudinal Axis Acceleration Data
In the ILS Approach and Landing Scenario, longitudinal acceleration error was not calculated.
3.4.5.4 Other Problems
Some runs were not included in the analysis because the data
was on a bad magnetic tape,
In some cases, data files had been
only partially transferred to a format compatible with the MIT
Man-Vehicle Laboratory computer facilities.
Although some of the
problems could be fixed by getting the back-up tape at the Ames
simulator facility and re-transferring the data, there was not a
lot of time to wait for this to be done.
with a few missing runs.
So the analysis continued
These are indicated in the appendices.
In other cases, for example in the ILS approach and landing
scenario, a pilot may not have completed the entire run.
60
So there
are some missing runs for the last 20-25 seconds before touchdown
data window for that run.
61
CHAPTER 4: PILOT OPINION RESULTS AND DISCUSSION
4.1
Introduction
This chapter is a presentation of the results of the opinion
questionnaire.
The opinion questionnaire required the subjects to
compare the simulator to the actual aircraft for various aspects
of the simulation. These included workload levels during different
phases of flight, and simulator response abilities to control input.
The opinion questionnaire also required pilots to rate the value
of the simulation on the utility of pilot training and pilot checking.
Finally, pilots gave their opinions about the overall realism
of the flight.
The analysis consisted of plots of the mean answers to the
questions over all pilots performing the same flight scenario and
F-tests to see if the differences between motion conditions was
significant for these means.
Results are presented in the first
part of the chapter and discussed in the second part of the chapter.
4.2
Presentation of Opinion Results
Raw data for the opinion questionnaire as well as the analysis
of variance tables and debrief session notes can be found in Appendix B.
The opinion questionnaire, which the subjects filled out
after each motion condition, is shown in Figure 3.3.
Plots of the
mean answer to each question (averaged over pilots flying the same
scenario) are shown in Figures 4.1-4.3.
Figure 4.2
shows the
average opinions of pilots who flew the engine flameout scenario;
62
figure 4.3,
shows the average opinions of pilots who flew the air-
work scenario; and figure 4.4, show the average opinions of pilots
who flew the ILS approach and landing scenario.
Figures 4.1-4.3 show that there is not much difference between
motion conditious in the opinions of the pilots to any of the questions asked. Furthermore, the figures show that overall the pilots
seldom answered lower than a "2" or higher than a "4",
most of the
data clusters around the number "3". For the questions on scenario
workload, configuration workload, simulator control response, training utility and checking utility, an answer of "3" indicates that
the simulator is very similar to the actual aircraft.
For the
question on overall simulator realism, an answer of three indicates
that there were some minor deficiencies in the simulation.
Tables 4.1-4.3 show the pilot opinion statistical results.
Analysis of variance tests were obtained to get a statistically
reliable level of significance at which the hypothesis of these Ftests, which is the mean answer to a particular question is the
same for all
three motion conditions,
can be rejected.
Tables 4.1-4.3 show that the null hypothesis can not be rejected at statistically
asked.
significant levels for any of the questions
The level of significance that is termed "statistically
significant " is defined to a probability of 0.05 of obtaining an
F-ratio larger than the one obtained.
The lowest probability in
these tables, which is associated with the answers of the engine
flameout pilots to the simulator response question, is only 0.25
corresponding to an F-ratio of 1.5.
The other values in the table
show even higher probabilities, and the null hypothesis is not
63
rejected at statistically significant levels for any of the questions asked.
The notes on the pilot debriefing session, provided in the
appendix, show some pilots making specific comments about the response or sensitivity in the pitch and roll degrees of freedom.
Pilots also made specific comments about particular motion conditions.
4.3
Discussion of Opinion Results
Taken together the plots and the tables lead to the conclusion
that there is no statisticlly significant difference in the answers
to these questions between motion conditions.
conclusion is debatable, however.
The value of this
It is possible that the ques-
tions were too vague too find differences in the opinions or that
more questions should have been asked.
In the first five categories
(scenario workload, configuration workload, simulator response,
training utility, and checking utility) pilots seem to agree that
the simulation is close to the actual aircraft.
However they also
agree that there are some minor deficiencies in the overall realism
of the simulator, as evidenced by the answers close to "3" in the
realism question.
Because these deficiencies were not discovered
in the other questions, perhaps the questions were too vague or
more questions should have been asked.
The notes on the debrief
session, show that several pilots made comments about specific
degrees of freedom and about specific motion conditions.
Perhaps
questions which compared the simulator control response to the
actual aircraft control response in particular degrees of freedom
should have been asked.
64
CONFIGURATION
scenario workload
WORKLOAD
oxx "=s
TWAmWmSf
nw
isca3-
4
3
mmo.m
SIMULATOR
AW7L"
RESPONSE
TRAINING
UTILITY
3.
To.s a-o
ft"mm
SaG
CHECKING UTILITY
REALISM
,
4.-
.I
,' -3-
,.
K c awl3
WM a.-
v.nm
m
Figure 4.1:
mn
wut
AL 4U
Opinions of pilots who flew the engine flame-out
scenario.
65
.7m
Scenario
Configuration Workload
Workload
I,
no
I
f
f
SIM"
2*
n
I..]
MM Los
fW 416c"r
P"
Simulator Response
Training Utility
T.
'M
is" "Isw
f
I
checking
f
3{
""MAN
" "SCUM
utility
Realism
-0 8
LSCA.
M
Sa
4
{
2
A" is.
Figure 4.Z,
jobs&
OV" MMM
JMFRA
uI.L.
Opinions of pilots who flew the airwork scenario.
66
I,,e.
SCENARIO
CONFIGURATION
WORKLOAD
w"
ma
i
il:'1 3.
SIMULATOR
WORKLOAD
4-
TRAINING
RESPONSE
UTILITY
mae
s.
a-
f
3-
NVA4
mm
-I
-O
P".0Rft
EAISm
Voem
sImo s3
REALISM
CHECKING
UTILITY
4.
i.
f
a-ri
"m3-
I
maA"
R" so
Figure 4.3:
Niel
ww"m
sputa " IMm
ap9mm
Opinion of pilot's who flew the ILS landing scenario.
67
TABLE 4.1:
Engine Flame-Out Pilot Opinion Statistical Results
F-Ratio
Scenario Workload
Probabilty
.5
.62
Configuration Workload
1.22
.32
Simulator Response
1.5
.25
Training Utility
.6
.56
Checking Utility
.21
.81
Overall Realism
.02
.98)
TABLE 4.2: Airwork Scenario Pilot Opinion Statistical Results
F-Ratio
Probabilty
Scenario Workload
.09
.91
Configuration Workload
.46
.64
1.01
.39
Training Utility
.78
.47
Checking Utility
.09
.92
Overall Realism
.05
.95
Simulator Response
68
TABLE 4.3: ILS Approach and Landing Pilot Opinion Statistical Results
F-Ratio
Probabilty
Scenario Workload
.65
.54
Configuration Workload
.38
.69
Simulator Response
.75
.49
Training Utility
.34
.72
Checking Utility
.27
Overall Realism
1.18
69
-
.77
.34
CHAPTER 5: ACCELERATION ERROR RESULTS AND DISCUSSION
5.1 Introduction
This chapter is a presentation and discussion of the results
of the acceleration error analyses.
The acceleration error was
computed by subtracting simulator accelerations from modelled aircraft accelerations for specific degrees of freedom.
Because only
magnitude information about critical portions of the flight was
desired, a root mean square error was computed for each pilot for
the data windows discussed in chapter 3.
The
basic analyses were F-tests and plots of the mean rms
error over all the pilots performing a specific manuever for a
particular degree of freedom.
The null hypothesis of these tests
is that the mean rms errors for different motion conditions are
equal.
The plots are comparisons of the means and standard devia-
tions of the means for each motion condition.
The F-ratios and
corresponding probabilities are printed below the graphs.
The
null hypothesis will be rejected at "statistically significant"
levles if the associated probability is less than or equal to 0.05
The chapter is organized by flight scenarios.
results are presented and first.
Engine flameout
The acceleration error results
from the manuevers of the airwork scenario: steep turns, stalls,
and rate turns with a yaw damper failure, follow.
The ILS approach
and landing acceleration error results are presented last.
The
chapter is concluded with a discussion of the accleration error
results.
70
5.3
Engine Flame-Out Acceleration Error Results
Figure 5.1 shows the mean rms yaw acceleration errors.
The
accelerations are on the order of approximately 0.008-0.012
rad/sec 2 .
There is an upwards slope to the plot which implies
that the rms errors were the lowest for full six degree of freedom
motion.
Jostle, two degree of freedom, motion had the next lowest
mean rms error; while the special effects only motion had the
highest mean rms error, as well as the highest standard deviation.
The associated F-ratio, 2.15, gives a probability of 0.137 that
the null hypothesis is true, which is not low enough to reject it
at a statistically significant level.
Based on the plot and the
F-test, there is an upward trend to the graph, this trend is insignificant based on the F-test.
Mean RMS Yaw
ACCELERATION ERROR
Engine Flame Out
0.011
-
0.013 0.012
0.011 0.01
0.009 -
0.00
-
0.002 0.001
0"
Jostle
motion
full
motion
Figure 5.1:
F
=
2.15,
71
special effects
only motion
P - 0.137
5.4
Steep Turn Results
The three rotational degrees of freedom, which are pitch, roll
and yaw, considered in this maneuver all showed the same shape of
the graph; the lowest mean rms errors were associated with the two
degree of freedom jostle motion, the next lowest mean rms error is
due to the special effects only motion; while the highest mean rms
error is associated with the full six degree of freedom motion
(Figures 5.2, 5.3, and 5.4).
This result is interesting because
one might expect the full motion condition to have lower rms errors
than the other conditions.
For the pitch acceleration error case,
the F-ratio is 3.64 and the associated probability level is
These differences are close to being significant.
differences are not statistically significant
.056.
Although the
for the roll case
with a probability of 0.174, when an F-test is done between the
firzt t'.
r.ti
ndiA.LJtzLL
(ull
and
jostLe), we obtain a probab-
ility level of 0.077 which closer to being significant.
Both the
pitch and roll mean rms acceleration error plots show a fair amount
of difference between the motion conditions.
The yaw acceleration
error case has a probability level of 0.201 and on the plot, the
difference in the means do not appear to be as prominent as the
other two cases.
Considering the orders of magnitude of the errors,
roll acceleration error is found to be the highest overall.
72
mean rms pitch
ACCELERATION ERROR
steep turns
0.015-
0.0130.0130.0120.011 -
0.01 0.008 -
Cd
0.007 -
-4
"o 0.006 -
0.003 0.0020.001 0-
I-
Figure 5.2:
mean
special effects
only motion
jostle
motion
full
motion
F
=
3.64,
P - 0.056
rms roll acceleration
errors
steep turns
.
0.035eq
U
0.03-
Q)
0.025-
0.02 -
0.015 0.01 0.0050-
Figure 5.3:
special effects
only motion
jostle
motion
full
motion
F = 2.00,
73
P
0.174
Mean rms yaw acceleration
error
steep turns
0.01
-
0.009
-
0.0080.007U,
~i-4
0.006
-
0.005
-
0.004
-
0.003 -
I
f
0.002 0.001 -
a
jostle
motion
full
motion
Figure 5.4:
F = 1.82,
74
special effects
only motion
P - 0.201
5.5
Stall Acceleration Error Results
Figures 5.5-5.7 illustrate the results obtained for the stall
flight manuever.
These Figures show mean rms acceleration errors
for the rotational degrees of freedom of pitch and roll and the
longitudinal translational degree of freedom.
The roll
acceleration plot has the lowest mean rms error for
the jostle motion condition and highest for special effects motion.
The standard deviations for the full motion case and the special
effects only motion case are quite large compared to the standard
deviation associated with jostle motion.
The plot shows prominent
differneces; while the F-test implies that the hypothesis that the
means are equal can not be rejected at statistically
levels (with p-.364).
variances,
significant
However the F-test assumes equal population
and this assumption probably does not hold up well for
the following reason.
The population variance of the individual
points is proportional to the variance of the means.
The standard
deviations of the means are not equal in this plot which implies
that the variance of the rms errors are probably not equal.
This
weakens the conclusions that can be drawn from this F-test.
The other degrees of freedom, pitch acceleration error and
longitudinal acceleration error (Figures 5.6 and 5.7)
show large
similarities both in the means and in their standard deviations.
The associated probabilities agree and imply that the null hypothesis can not be rejected at significant levels.
75
MEAN RMS
ROLL ACCELERATION
ERRORS
STALL
C14
co
-r-
0.020.019 0.0180.0170.0160.0150.0140.0130.012 0.011 0.01 0.0090.0080.007 0.0060.0050.0040.003 0.002 0.001 0--
I-
full
motion
Figure 5.5:
MEAN
jodtle
motion
F - 1.05,
special effects
only motion
P
=
0.364
RMS PITCH ACCELERATION
ERROR
STALL
0.015 0.0140.013 0.012 -
f
0.011-
c~OM -
Cd
0.008 0.007 0.00 0.0050.0040.003-
o.0o0 0.001
0 i-
full
motion
Figure 5.6:
76
jostle
motion
F - 0.12,
special effects
only motion
P = 0.886
MEAN
U
STALL
0.6
.
0.58
-
0.57
-
0.56 0.55
4.J
RMS LONGITUDINAL ACC ERRORS
0.54
{
f
0.53
0.52
0.51
0.5
jostle
full
motion
Figure 5.7:
motion
F = 0.21,
77
special effects
only motion
P = 0.816
5.6
Rate Turns with Yaw Damper Failure Results
The mean rms acceleration error for the pitch and yaw direc-
tions (Figures 5.8 and 5.10) have the same shape, with the highest
errors for the full motion condition and the lowest for the special
effects only condition.
For the roll direction, jostle motion has
a slightly higher mean rms error than full motion.
For all of
these directions; however it is noticed that the mean rms acceleration errors for full and jostle motion are very close in value.
None of the differences are found to be statistically significant
according to the results of the associated F-tests.
The roll
acceleration errors are higher, on average than both the pitch and
yaw acceleration errors.
MEAN RMS PITCH ACCLERATION
RATE TURNS WITH YAW DAMPER FAILURE
0.04 -
ERRORS
0.035 0.03-
C*4
0.025 Ca
W,
0.020.015 0.0L -
{
0.0050*
full
motion
Figure 5.8:
jostle
motion
F = 0.87,
78
special effects
only motion
P = 0.443
MEAN
RMS
ROLL ACCELERATION
ERROR
RATE TURNS WITH YAW DAMPER FAILURE
026
024 -
-
022
0.:1
U
0.16 -
C,,
U,
.9-I
-o
CU
04t4 0.12
-
0.1 -
0.08
-
0.06 -
0.04
-
0.02 0-
full
Figure 5.9:
MEAN
0.06 -
special effects
only motion
jostle
motion
motion
F = 0.50,
P
-
0.620
RMS YAW ACCELERATION
RATE TURNS WITH YAW DAMPER FAILURE
ERRORS
0.05 -
C..4
C,,
-r4
{
0.04 -
0.03 -
0.02
-
0.01
-
0
full
motion
Figure 5.10:
jostle
motion
F
79
-
0.57,
special effects
only motion
P - 0.577
5.7
5.7.1
ILS Approach and Landing Scenario Results
Approach Segment: 500' - 200'
Figures 5.11, 5.12, 5.13 show the rotational degrees of freedom
mean rms errors for the different motion conditions.
For all cases
mean rms errors are highest for special effects only motion and
lowest for for full motion.
Although these differences are not
major and the statistical results do not allow rejection of the
null hypothesis at significant levels.
The mean rms acceleration
errors in the roll direction are slightly higher than these errors
in the pitch and yaw directions.
The vertical axis acceleration
error plot (Figure 5.13) also does not show major differences between motion conditions and the hypothesis that mean rms errors
are equal for the different motion conditions can not be rejected
at significant levels.
MEAN
RMS
PITCH ACCELERATION
ERRORS
ULs LANDING 500'-200'
0.040.0300.038 0.0370.036 -
0.03 0.033
0.031 -
full
motion
Figure 5.11:
jostle
motion
F -
1.36,
80
special effects
only motion
P - 0.272
MEAN
ROLL ACCELERATION
ILS LANDING 500'-200'
ERRORS
0.06-
0.059 0.058 -
0.057
-
0.056 0.055 0.054 -
0.053
w
{
-
0.052 0.051 0.050.049 -
*1-4
.t
S
0.048 0.047 0.046 -
0.045
-
0.044
-
0.043
-
0.042 0.041 0.04.
jostle
full
motion
Figure 5.12:
MEAN
0.015
special effects
only motion
motion
F = 0.20,
P - 0.821
RMS YAW ACCELERATION
ERRORS
ILS LANDING 500'-200'
-
0.014 0.013 0.012 0.011 C%4
0.01 0.009 -
0.008 Cd
0.007 *~0.0060
0.0050.0040.003 0.002 0.001-
0 -J-
I
full
motion
Figure 5.13:
jostle
motion
F
81
-
1.18,
-
____
special effects
only motion
P - 0.322
MEAN
RMS VERTICAL AXIS
ACC ERRORS
ILS LANDING 500'-200'
2.93-
C14
U
2no
-
2.89
-
f
a)
U,
U,
I..'
a)
a)
2.88 2.87 2.86
-
-
I
I
full
jostle
motion
motion
Figure 5.14:
F
82
=
0.54,
special effects
only motion
P = 0.588
5.7.2
Landing segment: Last 20 to 25 s Before Touchdown
The mean rms pitch, roll and yaw acceleration errors are shown
in Figures 5.15,
5.16,
and 5.17 respectively.
Both the roll and
the pitch plots show a downward slope implying largest mean rms
errors for the full motion case, and lowest; for the special effects
only motion condition.
This trend is small, and the differences
are not supported by the F-tests.
error plot,
In the mean rms roll acceleration
the means for full and jostle motion are quite close.
The mean rms error for the angular acceleration in the yaw direction
is highest for the jostle motion condition and lowest for the full
motion condition.
The F-test does not support differences in the
means for any of the rotational degrees of freedom.
In these plots,
the average roll acceleration error appears to have higher errors
than the average pitch and yaw acceleration errors.
The vertical axis acceleration error plot (Figure 5.18) shows
little differences in the mean rms errors between motion conditions.
The hypothesis that these quantities are equal is not rejected by
the F-test.
83
MEAN
0.05 0.04
-
0.03
-
0.02
-
PITCH ACCELERATION
RMS
ERRORS
ILS LANDING LAST 20-25 SEC
{
0
C12
co,
w~
0.01
0
f ull
motion
Figure 5.15:
MEAN RMS
0.2
-i
jostle
motion
F - 1.59,
special effects
only motion
P - 0.221
ROLL ACCELERATION
ILS LANDING LAST 20-25 SEC
ERRORS
0.19
:.181
0.17
C:,
0.16
-
0.15
-
0.14r
-
C13
5-4
0.13
0.12-
0.11O.J
full
mo tion
Figure 5.16:
jostle
motion
F - 0.98,
84
special effects
only motion
P = 0.389
MEAN
0.0 4 1
RMS YAW ACCELERATION
ILS LANDING
ERRORS
20-25 SEC
LAST
0.035 0.03C14
Cl)
0.025 -
cc,
0.02 0.015 0.01 0.005 -
jostle
motion
motion
Figure 5.17:
Mean
IsI
3
I
I
I
full
F
1.57,
-
speciAl effects
only motion
P - 0.226
RMS Vertical Axis Accel Errors
IW Landing last 20-25 soc
2.92.972.96C.,
Cl,
2.95-
C')
S.d
f+
2.94-
2.93 2.922.912.9
.1
I
jostle
motion
full
motion
Figure 5.18:
F
-
85
0.40,
I
special effects
only motion
P - 0.674
5.2
Discussion of Acceleration Error Results
Discussion of Analysis Method
The root mean square is purposely used to ignore differences
in the functionality of the data because the purpose of this study
is to obtain very basic differences in the relative levels of these
errors for the motion conditions studied. The more complicated
differences associated with the fact that these errors vary with
time in a functional matter are important and should be studied
further.
The plots shown in the preceding pages are useful because they
are magnitude comparisons of an average root mean square acceleration error for different motion conditions.
For this report this
measure is useful because it is one way to compare realism for the
three motion conditions.
The weaknesses in all of the results derived from the plots
shown in the preceding pages is that acceleration error is an incomplete way to look at realism.
In this study we are looking at
several incomplete measures of realism and trying to obtain a broad
picture by studying various specific aspects.
it is misleading in some ways.
One reason is that
This measure completely ignores
what is known about human motion sensing systems.
The errors may
be below the threshold of perception of the pilot, in which case
they may not be important.
In the next chapter, the question of
perception threshold is examined more thoroughly.
There are also weaknesses in conclusions drawn from the Fratios and probabilities.
The null hypothesis of these tests is
that the means shown in the corresponding plots are equal.
86
The
probability gives a significance level for rejecting this hypoAt probability levels of 0.05,
thesis.
rejected at what is defined
the null hypothesis can be
to be statistically significant
levels.
This implies a 5 % chance that the null hypothesis is true given
the F-ratio, which is an estimate of the ratio of the variance
between the motion conditions as compared to the overall variance,
obtained.
The weakness in the conclusions based on these tests is
that the F-test makes assumptions about the distributions of the
data and we are not sure if these assumptions are valid.
The F-
test assumes that the data are independent samples of a normally
distributed variable.
It assumes that the three motion conditions
have the same variance of the individual measures of the rms value.
The assumption about independence is probably a good one, partly
because subjects probably did not effect the results of other subjects.
Also because the familiarization scenario was flown to
minimize learning effects and the order of presentation of the
motion conditions was randomized to minimize learning effects within
a subject.
The other assumptions of the F-test:
that the data is
normally distributed with the conditions having the same population
variances, is not at all clear.
Because the population variance of the individual points is
proportional to the variance of the means, the standard deviations
in these plots can be used to estimate the validity of the assumption of equal population variances.
Because these assumptions can not be validated,
sions drawn from these numbers is weakened.
the
conclu-
The numbers are still
valuable and worthwhile to look at because they give a rough esti-
87
mate of how significant these differences may be.
Also by consider-
ing the plots, we have a stronger pictorial representation of the
differences that exist between motion conditions.
The F-test is a
secondary weaker measure of the results shown in the plots.
Discussion of Results
There appears to be some differences in motion conditions as
measured by acceleration error.
None of the differences found were
statistically significant at the 0.05 level; however, the plots show
relative differences.
Not all degrees of freedom show marked dif-
ferences and some of the flight scenarios, such as the ILS aprroach
and landing, do not show strong differences in any of the degrees
of freedom.
The direction of the differences appears to be counter-intuitive at times.
It seems as though the lowest errors would be
associated with the full six degree of freedom motion; however,
this is not always the case.
It does not hold up in steep turns
roll, pitch and yaw axes, which have higher mean rms errors for
full motion than for the other motion conditions.
turn with yaw damper failure manuever,
For the rate
the full motion mean rms
acceleration errors are higher than the special effects only motion
errors for all rotational degrees of freedom.
In the vertical
axis for the landing portion of the ILS approach and landing
scenario,
the mean rms error is slightly larger for full six degree
of freedom motion than for the other two.
Perhaps the reason this occurs is because the simulator might
be undergoing low frequency tilts to obtain the necessary accelera-
88
tions for the translational axes.
This could induce higher errors
in the rotational axes for the full motion condition than for the
jostle or special effects conditions.
It would be interesting to
see if these errors induce errors that are above the threshold of
perception of the vestibular system.
89
CHAPTER 6:
VESTIBULAR ERROR RESULTS AND DISCUSSION
This chapter is a presentation and discussion of the results
of the vestibular error analyses.
in Chapter 2,
The vestibular error, as explained
was computed by subtracting modelled simulator ves-
tibular responses from modelled aircraft vestibular responses for
specific degrees of freedom.
As stated in chapter 3,
not every
degree of freedom was studied for every flight manuever.
The
degrees of freedom that were studied for each flight manuever, were
considered to be particularly important.
Because only magnitude
information about critical portions of the flight was desired, a
root mean square error was computed for each plot for the critical
portions of the flight.
The basic analyses were F-tests and plots of the mean rms
errors over all pilots performing a specific manuever for a particular degree of freedom.
The null hypothesis of these tests is
that the mean rms errors for different motion conditions are equal.
The plots are comparisons of the means and standard deviations of
the means for each motion condition.
The F-ratios and correspond-
ing probabilities are printed below the graphs.
The chapter is organized by flight scenarios.
vestibular error results are presented first.
Engine flameout
The vestibular error
results from the manuevers of the airwork scenario:
stalls,
and rate turns with a yaw damper failure,
steep turns,
follow.
The ILS
approach and landing vestibular error results are presented last.
The chapter is concluded with a discussion of these results.
90
Engine Flame-Out Results
6.1
The mean rms vestibular errors in the yaw direction are shown
in Figure 6.1. The plot slopes upwards implying highest errors for
the special effects only condition and lowest for the full motion
condition.
In Chapter 5,
Figure 5.1, which shows the
acceleration errors in the yaw direction,
mean rms
also slopes upwards.
This implies that in this case, the acceleration and vestibular
errors give similar information, at least in terms of the direction
of the trend.
The mean rms error for the full motion condition is
0.96 threshold units, which is below the threshold of perception.
The other conditions have means that are greater than 1, implying
that the error can be perceived by the pilot.
The F-test on the
data does not support differences in the three motion conditions.
Mean RMS Vestibular Yaw Error
Engine Flame Out
1.25
1.21.15
1.05-
0
0j 0.93 0.9-
0.3
full
motion
jostle
motion
special effects
motion only
Figure 6.1:
F - 0.97,
P - 0.392
91
6.2
Steep Turn Vestibular Error Results
Figures 6.2, 6.3, and 6.4 show the mean rms vestibular errors
in the pitch, roll and yaw directions.
The mean rms vestibular
errors in the pitch and yaw directions are well below the threshold of perception for all motion conditions.
The mean rms vesti-
bular error in the roll direction is above the perception level
(1.22 threshold units) for the full motion condition; but the other
two conditions have mean vestibular roll errors below the perception
threshold.
These plots imply that the full six degree of freedom
condition is worse in a perceivable way for roll motion in this
manuever.
The mean rms acceleration error in the roll and pitch directions plots in Chapter 5 (Figures 5.3 and 5.2) show the same shapes
as Figures 6.3 and 6.2, with the highest errors in the full motion
case and the lowest in the jostle motion case.
The shape of the
yaw axis plots are different for acceleration error measurements
than they are for vestibular error measurements.
Figure 5.4 shows
that the mean rms acceleration error in the yaw direction is slightly lower for jostle motion than for special effects only motion. On
the other hand, the numbers associated with Figure 6.4 say that
the mean rms vestibular error in the yaw direction is slightly
higher for jostle motion as compared to special effects.
This is
interesting because in this case the acceleration error is misleading.
Although the differences are very small, and they are not
important for determining differences between motion conditions;
the data is taken from the same set of runs and therefore shows
that the acceleration error can be misleading.
92
It will be interest-
ing to see if any of the other quantities measured show opposite
trends for the vestibular errors compared to the acceleration errors.
The analysis of variance tests done for the vestibular errors
do not allow us to reject the "equal rms error mean" hypothesis at
statistically significant levels (significant implies p - 0.05).
The lowest probability level is for the pitch axis (p - 0.097).
When an F-test is done solely between full motion and jostle motion,
however, the probability obtained (p - 0.011) is significant.
An
F-test performed between full and jostle motion conditions for the
vestibular roll rms errors also enables us to reject (at the 0.028
significance level) the hypothesis that the mean rms errors for
these conditions are equal.
Mean RMS vestibular
pitch
error
steep turns
fl6
0.5 -
.A-I
r-4
0.3
-
0.2
-
0.1
-
I
0
41a
0 ~1
3
full
motion
jostle
motion
Figure 6.2:
F = 2.81,
93
special effects
motiQn only
P = 0.097
Mean Rms
1.5
vestibular roll error
steep turns
-
1.4 -
1.3
-
1.2
-
1.1
-
I-
0
0.9
-
0.7
-
0.7
-
0.5
-
0.4
-
0.3
-
0.2
-
0.1
-
W,
0
full
motion
jostle
motion
special effects
motion only
Figure 6.3:
F - 2.22,
P
Mean rms errors
=
0.148
vestibular yaw
steep turns
0
0.24-
0.22H
0.2
0.18 CD
4-1
0
U,
0.16
-
0.14
-
0.12
-
O.1 0.08
-
0.06
-
0.04
-
0.02
-
.d
0~ I
I
full
motion
jostle
Figure 6.4:
F
motion
94
-
0.90,
special effects
motion only
P - 0.429
6.3
Stall Vestibular Error Results
The mean rms vestibular errors in the roll and pitch directions
are plotted in Figures 6.5 and 6.6 respectively.
The average errors
in these plots are below the threshold of perception for all three
motion conditions - full,
pitch vestibular errors,
jostle,
and special effects only.
For
the highest mean root mean square error is
for the full motion case and the lowest, for the special effects
only motion.
For the roll vestibular errors, the highest mean rms
error is for the full motion condition and the lowest is for the
jostle motion condition.
The mean rms acceleration errors give
different information for the roll direction.
Figure 5.5 show the
mean rms acceleration errors in the roll direction are highest for
the special effects only motion.
The F-tests do not allow rejec-
tion of the null hypothesis at statistically significant levels for
either degree of freedom.
An F-test performed just between the
full and jostle conditions for the mean roll vestibular error gives
a probability level of .24
which is still not significant.
standard deviation shown in this plot are not equal.
The
This raises
questions about the validity of the F-test.
The mean rms vestibular longitudinal errors are shown in Figure
6.7.
These errors are above the threshold of motion perception
for all three motion conditions by a factor of approximately 1.5,
and they appear to be very close in value for all three conditions.
The F-test gives a probability level of 0.914, which does not allow
us to reject the null hypothesis at even close to significant levels.
95
MEAN RMS VESTIBULAR ROLL ERRORS
STALL
0.1 0.090.08 0.0771
0.06 -
4
0.05 0
U,
0.03
-
0.1 -
full
motion
jostle
Figure 6.5:
F
special effects
motion
=
0.82,
motion only
P - 0.450
MEAN RMS VESTIBULAR
STALL
0.01
PITCH ERRORS
0.07
0.06 14
0.050
0.00.031
o.aa
I
-
0.O a-
Imotion
jostle
motion
special effects
motion only
Figure 6.6:
F - 0.96,
P
full
96
=
0.395
MEAN
RMS VESTIBULAR
LONGITUDINAL
ERRORS
STALL
1.6
1.59
-
1.50
-
1.57
-
1.56
1.55
-
1.54
-
-
1.52u
1.51a o 1.4382
1.5o
{
-
f
1.49-
1.43 1.42 -
143
-
1.41
I
full
motion
jostle
Figure 6.7:
F
motion
=
97
0.09,
--
special effects
motion only
P
-
0.914
Rate Turns with Yaw Damper Failure Results
6.4
The mean rms vestibular errors in the pitch, roll and yaw
directions, shown in Figures 6.8 through 6.10 are all well below
For all motion conditions, the aver-
the threshold of perception.
age roll rms vestibular errors are slightly higher than the average
pitch and yaw rms vestibular errors.
According to the results of
the F-tests, the null hypothesis can not be rejected at statistically significant levels.
The plots themselves do not show marked
trends.
The shape of the plot for the mean rms vestibular errors in the
roll direction conform with the corresponding acceleration error
plots in Chapter 5 (see Figures 5.9 and 6.9) although the differences
between full and jostle motion seem more pronounced in the vestibular
plot.
However, the relative level of errors between motion condi-
tions change slightly for the yaw and pitch degrees of freedom.
In the vestibular error plots in Chapter 6,
jostle motion has a
slightly higher mean rms vestibular errors in these directions
than full motion.
ter 5,
Whereas in the acceleration error plots in Chap-
jostle motion has slightly lower mean rms errors than full
motion for the pitch and yaw directions (see Figure 5.8 and 5.10).
These changes are very small.
Although
full and jostle motion (see appendices)
F-tests strictly between
do not statistically
allow
us to reject the hypothesis that the mean rms errors are equal for
different motion conditons for either the pitch degree of freedom
or the yaw direction.
The data for acceleration error information
comes from the same set of runs as the data for vestibular error
98
information, and it gives a slightly different picture of what is
going on.
MEAN RMS VESTIBULAR PITCH ERRORS
RATE TURNS WITH YAW DAMPER FAILURE
0.01
0.0090.0080,
-w
0.0070.006-
~0
0.005 -
C',
0.004-
-w
0.003 0.002
-
0.001 -
0.
I
motion
Figure 6.8:
F
full
MEAN
RMS
=
special effects
motion only
0.94,
P = 0.414
VESTIBULAR ROLL ERRORS
RATE TURNS
0.04
I
jostle
motion
WITH
TAW DAMPER
FAILURE
0.035-
0.03 (n
0.025
-
0.02
-
0.015
-
0.01
-
0
C',
1.1
0.005
a
,
full
motion
jostle
Figure 6.9:
F
motion
99
=
1.01,
special effects
motion only
P = 0.392
MEAN
RMS VESTIBULAR YAW ERRORS
RATE TURNS WITH YAW DAMPER FAILURS
0.01
-
0.0090.008 0.007Vi4
0.006r-4
0.005-
0
0.0030.0020.0010*
jostle
motion
full
motion
Figure 6.10:
100
F
=
0.36,
special effects
motion only
P - 0.703
6.5
6.5.1
ILS Ap2roach and Landing Results
Approach segment:
500' - 200'
The mean rms vestibular error for the rotational degrees of
freedom are presented in Figures 6.11 to 6.13.
These plots all
show errors which are below the threshold of perception for all
three conditions.
The mean rms errors are close to being equal
and the null hypothesis cannot be rejected at statistically significant levels for any of the degrees of freedom.
There are some changes in the magnitude of the average errors
relative to each other between the acceleration error plots and
the vestibular error plots.
These are most noticable in the pitch
direction and can be observed by comparing Figures 5.11 and 6.11.
For the translational degrees of freedom (see Figures 6.15 and
6.16) the vestibular longitudinal errors are below the threshold
perceeption level.
The vertical vestibular errors, on the other
hand are on the order of six times above the level of perception.
The F-tests do not allow us to reject the null hypotheses at
statistically significant levels for both the vertical and the
translational rms errors.
The direction of the trend is different for vestibular error
measurements than for acceleration error measurements for the vertical degree of freedom (see Figures 6.14 and 5.14).
We are unable
to compare the direction for the longitudinal direction because
the acceleration error data was not processed.
101
MEAN RMS VESTIBULAR PITCH ERROR
ILS LANOING 500'-200'
0.5
0.48
0.45
0.44
0.42
2
,-4
0.4
-
0.38
-
0.35
-
0.34
-
0.32
-
0.3
-
0.28
-
f
0.25
0.24
-
0.22
-
0.2
full
jostle
motion
motion
Figure 6.11:
VESTIBULAR
4.J
v-4
0
.J
P - 0.987
F = 0.01,
ROLL MEAN RMS
ERRORS
ILS IANDING 5W'-200
0.55-
0.54
special effects
motion only
-
0.53 0.52 0.51 0.5 0.40 0.48 -
0.47
-
0.46 0.450.44
0.430.420.41
0.4
0.390.30-
0.370360.35
jostle
motion
full
motion
Figure 6.12:
102
F = 0.02,
special effects
motion only
P
=
0.983
MEAN VESTIBULAR YAW RMS ERRORS
ILS LANDiNG 500'-200r
0.1
0.02
CA
4.J
0.08 -
0
0.07
-
*1
0.06 -
0.05 -
--
full
motion
Figure 6.13:
jostle
motion
F - 0.53,
103
special effects
motion only
P
-
0.594
MEAN RMS VESTIBULAR VERTICAL ERRORS
IL9 LANDING 500-200'
6.5
6.406.466.446.42co
6.46.38
6.36 -
f
6.340
6.326.36.28-
6.26
f
-
6.24
-
6.22
6.2
full
motion
jostle
motion
F - 0.48,
Figure 6.14:
MEAN
RMS VESTIBULAR
0.8 -
0.7
0.622
LONGITUDINAL
ERRORS
-
'-I
0
=
I
U,
~LJ
~1-4
P
ItS LANDING 500-200'
I
0.9
special effects
motion only
-
{
TI
0.6-
0.5
I
I
full
motion
jostle
motion
Figure 6.15:
104
F
-
0.63,
special effects
motion only
P
-
0.538
Data Window for the last 20-25 sec before touchdown
6.5.2
The mean rms vestibular errors are below the threshold level
of perception for the pitch and yaw directions;
for the roll direc-
tion, however, the mean rms error is just at the threshold level
of perception
for the full motion condition,
and below the thres-
hold level of perception for the other motion conditions.
The analysis of variance tests berween the three motion conditions are not significant for any of these plots, although they
come close in the pitch degree of freedom with a probablity level
of 0.086.
The only different information given by acceleration error
results in Chapter 5 and vestibular error results in chapter 6
occurs in the yaw degree of freedom.
For this data window, both the translational degrees of freedom
examined have errors that are above the threshold level of perception.
For the longitudinal direction, they are on the order of
two times the perception level; for the vertical direction, they
are on the order of six times the perception level.
The results
of the F-test are not statistically significant for the three motion
conditions examined.
There are slight differneces in the informa-
tion given by acceleration error measurements and vestibular error
measurements in the vertical direction pertaining to jostle and
special effects motion (see Figure 5.18 and 6.19).
The accelera-
tion and vestibular errors for the longitudinal degree of freedom
were not compared.
105
MEAN RMS VESTIBULAR PITCH ERROR
ILS LANOING LAST 20-25 SEC
* 0.3
0.29
0.28
-
0.270.26
0.25 -
{
0.24
0.23 0.22 0.21 0.2
0.19 0.18 0.17 -
~0
0j
0.15
0.14
0.13
0.12
0.11
1
W. I
full
motion
jostle
motion
Figure 6.16:
special effects
motion only
F = 2.69,
P
=
0.086
RMS MEAN VESTIBULAR ROLL ERRORS
I.S LANOING LAST 20-25 SEC
1.2-
1.-
0.9 -
0
0.5 -
'4j
0.7 -
0.6
05
jostle
motion
full.
motion
Figure 6.17:
106
F - 1.80,
special effects
motion only
P = 0.184
MEAN RMS VESTIBULAR YAW ERROR
LS
0.2
LANDING LAST 20--25 SEC
-
0.19 0.18 0.17 0.16 C,,
0.15 -
f
0.14 0.13
'0
0
0.12
0.11
C,,
aJ
4-I
0.1
0.09
0.08
0.07
0.06
0.05 -
full
motion
jostle
motion
Figure 6.18:
107
F = 0.60,
special effects
motion only
P = 0.554
MEAN
RMS VESTIBULAR VERTICAL ERRORS
ILS Landing last 20-25 sec
6.7
6.65
6.6
co
r--4
6.5
-
6.45
-
6.4
-
0
co,
6.35
6.3-
I
I
jostle
motion
full
motion
F
Figure 6.19:
MEAN
I
RMS VESTIBULAR
1.27,
=
special effects
motion only
P = 0.296
LONGITUDINAL
ILS Landing last 20-25 see
ERRORS
3
2.9
2.8
2.7
2.6
U,
4.'
"-4
2.5
2.+
-
-o 2.3
-
'-.4
0
U,
22-
s.d
1.l
1.8
-
1.8 6
1.5
- .
I
full
motion
Figure 6.20:
jostle
motion
F
=
0.92,
108
special effects
motion only
P
=
0.410
6.6
Discussion of Vestibular Error.Results
There are a number of interesting results that were found by
examining the vestibular error.
The first is that there were no
significant differences found between the three motion conditions.
Two differences were found to be significant in a two-way comparison.
The result that the motion conditions were not remarkably different
most of the time is very important because it means that adding
motion capability to a simulator may not be necessry.
In fact, in
some directions, motion capability makes the situation worse;
although in some cases,
made more realistic.
this may be because another direction is
For example,
pitch motion may be sacrificed
in order to achieve longitudinal fidelity.
The next interesting result is that frequently the rotational
errors were below the threshold level of perception for many
the motion conditions examined.
of
This is important because it means
that the simulator is doing its job well in these directions.
Although fewer translational degrees of freedom were examined than
rotational, most of the average translational errors studied were
above the perception level threshold.
The exceptions were the ILS
approach segment mean longitudinal rms errors, which were below
perception level for all motion conditions.
The result that on
the whole rotational errors were less than translational errors in
terms of perception is significant because it
implies
that perhaps
more cross-coupling should be attempted to trade-off some expendable
rotational error for the purpose of achieving less error in the
translational directions.
Although it is important to realize
that aliasing may change the conclusions drawn from rotational
109
vestibular error results.
The next chapter will examine what the
effects of aliasing may have been.
The third result is that the acceleration error and the vestibular error can give contradictory information about which motion
conditions are better than others. This result is not suprising
because the vestibular response is a frequency dependent function.
110
CHAPTER 7: PILOT PERFORMANCE RESULTS
This chapter is a summary of pilot performance results.
Cal-
culations, plots, and statistical tests were done at the NASA Ames
The
facility and are also presented by Lee and Bussolari (1986).
results that are presented here are not complete.
Most of the
Statistical analyses
plots do not show standard deviation bars.
were conducted and differences are said to be "statistically reliable" if a significance level of 0.1 was calculated.
statistical analysis are not presented here.
Data for the
A short summary of
these results are presented for completeness of this study
.
Engine
flameout results are presented first followed by performance results
from the airwork scenario.
The ILS approach and landing performance
measures are presented and discussed last.
7.1
Engine Flame-Out
Figure 7.1 shows the mean aircraft centerline deviation in
feet, averaged over all pilots.
This plot is for ten seconds prior
to and ten seconds following engine flame-out,
where the EPR of the failed engine drops.
which is the point
According to Lee and
Bussolari (1986), no statistically reliable differences were found.
Time to climb out to an altitude of 400' from an speed of 120
KIAS was calculated for each run and the results were averaged
over all
runs and pilots.
ferent motion conditions.
Figure 7.2 shows this plot for the difFull six degree of freedom motion is
higher than the other conditions by a small amount.
Lee and Bussolari (1986),
According to
the difference between the fastest and
111
the slowest average times for the motion conditions is approximately
three seconds which is not statistically significant.
7.2
Airwork Scenario Performance
7.2.1 Aproach-to-Stall Manuevers
Figure 7.3 shows the mean variance (averged over all airwork
pilots) in aircraft attitude during the period ten seconds prior
to and following the point where the lowest indicated airspeed was
attained.
For both pitch and bank angles, the variance is highest
for full motion and lowest for jostle motion.
According to Lee
and Bussolari (1986), no statistically reliable differences were
found in analyses of this data.
7.2.2
Rate Turns with Yaw Damper Failure
Figure 7.4 shows the mean variance of aircraft pitch (7.4a)
and bank (7.4b)
angles during the ten second period prior to and
following zero bank angle.
For pitch angle, the mean variance is
highest for jostle motion and lowest for special effects only motion.
For mean bank variance, which is on the order of a hundred
times mean pitch variance, full motion has the highest mean variance, while jostle and special effects motion are very close in
value.
According to Lee and Bussolari (1986), no reliable dif-
ferences were found.
7.3
ILS Approach and Landing ScenarioFigure 7.5 shows the mean glideslope and localizer deviations
during the period 20 sec after an altitude of 500' was reached.
112
(In other sections of this report the data window is from the altitudes of 500'-200', which usually lasts about 20 sec).
Although
there are small differences in the glideslope and localizer deviations, these differences fall within those expected due to sampling
variation in the data.
According to Lee and Bussolari (1986), none
of these differences were statistically reliable.
Figure 7.6 shows the mean sink rate and lateral deviation
during the landing phase of the flight 20 seconds prior to touchdown.
There appear to be little differences between the motion
conditions.
The mean sink rate is lower for jostle motion than
for the other two conditions.
(1986),
According to Lee and Bussolari
these differences are within those expected due to sampling
variation alone and are not statistically reliable.
Discussion of Pilot Performance Results
No statistically reliable differences between motion conditions were reported for any of the pilot performance results presented.
Based on the results of this study, the conclusion that
pilot performance is not a valid measure of motion fidelity is not
unreasonable.
Sullivan (1985) also reaches the conclusion that per-
formance is not a valid or consistent measure of motion fidelity.
This result makes sense theoretically because similar performance measures may be obtained by different control strategies.
This implies that similar performance to the aircraft does not
necessarily imply realism; although lack of similarity in performance measures could signify lack of realism.
113
100
-
90 L-
80 70
z
-
60 -
50
-
40-
0 6 DOF
U 2 DOF
0 SPECIAL EFFECTS
30
20
a*
10
0
a
-10
-8
Figure 7.1
-6
-4
-2
EF
2
TIME (SECODS)
4
I
I
6
8
Aircraft centerline deviation ortor to and following
engine flameout (EF) as a function of motion platforn condition.
(N-6
pilots)
IN RLIod. 9%
(,4
50
U'
0 6 DOF
50
a
L4
2OF
C SPECIAL EFFECTS
30 -
-- I
7
20
10-
Figure
'.UTimS to climb to altitude following engine
flameout. (N-6) (LEE
ANO
114
MIWsOLARI i otol)
I
10
0 60F
13 2 0OF
-
6
5
-
60
C tSPECK EFFECTS
4C
16
r
50
. (b)
(a)
K-;
3
2
I
40
<
30
20
10
I I
Ftgure'13 Mean variance of aircraft oitch (a) and bank (b) angle .
during aporoacri to stall. (N-6 pilots) ( Le-e ad
iusok". 196)
0
C
w
6
0%or
-
-
S
02 D
C SPECA
600
FFECrs
500
U
(a)
4
2
4
w
2
(b)
300 F
V
200
'F
I I
Ll
157L
Z-Irolr
FIgUrelu: Mean vaiance of aircraft pitch (a) wd bar (b) angle
durrng stanard ttrns with yaw anOprs faile. (N6 pilots) (Lee
And)C
400 >
s
-sMo'Alo.) ivs)
115
100
90
6 w
SM
EFC
SM
75
4
SM
C
0 SPECtAL E'FECTS
60
TT
-
I-
45-
z
4
SM
r
-'4'
30.
15
GLIDESLOPE
I
LOCALIZER
Figure|riean gildesioce and localizer deviation curing tne
Instrument aooroacn maneuver. (T-Oarw l S.D.. N-6 13lots)
(LEE AjjD TU5'5OLARI, 19%(2)
6 DOF
0
2 DO'
0 SPECIAL
12 U
SM
U,
Si.
SM
4
12
EFFECTS
10 -
10
/
V.
6
6
6
2
I-
In
4,9
2
4
/
I;
2
'/
u
I
I
.
Figtre% Landing sink rate and touchdown laterml deviation.
(N-6 piots)
116
116
16
CHAPTER 8:
8.1
PROBLEMS DUE TO ALIASED ROTATIONAL VESTIBULAR RESPONSES
Introduction
This chapter contains a discussion of how using undersampled
rotational acceleration measurements to calculate the semicircular
response affected the results of this thesis.
8.2
Aliasing General Information
The rotational acceleration measurements used to calculate
the semicircular canal response
were sampled at 15 Hz. Theoretical-
ly, any motion that is greater than half this frequency, which is
7.5 hz will be aliased; in other words, it will look like a lower
frequency motion.
This is because the minimum sampling rate needed
to obtain the correct frequency is two times the period (or half
the frequency) (Figure 8.1).
It is known that aliasing magnitude
information is symmetric about the critical sampling frequency.
For our case, frequencies of 8 Hz will be aliased to frequencies
of 7 Hz and the higher frequencies will be correspondingly aliased
to lower frequencies. For example, 10 Hz is aliased to 5 Hz, etc.
Whereas the magnitude information is preserved,
the phase of the
aliased signal is transformed randomly because it is not known
where the signal is being sampled (see Figure 8.1).
The important questions is how the aliasing affect the measurements and the results of this study.
117
I.
Figure 8.1: Example of Aliasing Effects
true waveform
o)
datapoints produced by sampling AA'
O]
datapoints produced by sampling
---.---.-.-.
aliased waveform produced by sampling 'A'
'B'
aliased waveform produced by sampling 'B'
118
8.3
The Effect of Undersami~ling on the Results of This Study
A number of results in this study are based on the rotational
acceleration errors.
The fact that the rotational acceleration
measurements were undersampled does not change the rms acceleration
error calculations for the following reasons.
The root mean square
calculation is essentially an averaging calculation to determine
magnitude of the error.
It is not a frequency dependent calcula-
Undersampling becomes a problem when the data is put through
tion.
a frequency dependent calculation such as the semicircular canal
response calculation because low frequencies are there that
shouldn't be.
No problems are expected in the rms rotational acceleration
error calculations because they are not frequency dependent.
The
results of the rotational acceleration error analysis remain unchanged.
The Effects of Undersampling on Rotational Vestibular Error
8.3.1
Results
On the other hand, semicircular canal error results can be
affected by aliasing because they are a frequency dependent calculation.
The question to consider is how much are the results
affected.
In most situations if aliasing occurs, the total effect
on the response is unpredictable; while the amplitude information
is preserved in the transformation to lower frequencies, the phase
information is transformed by an unpredictable amount (see figure
8.1 ).
The reason this amount is unpredictable is because it is not
known where on the signal the sampling points are taken.
119
In this
situation, we are interested in the effect on the root mean square
vestibular error. Although this function is concerned with the
average magnitude of the error, phase effects become important
because of the subtraction of simulator responses from aircraft
responses.
Phase effects could cause the subtraction can be changed
into an addition.
As a result the mean rms error can be changed.
For aliasing to have a major effect on the calculations of
rotational vestibular error, the power at the frequencies that are
aliased would have to be on the order of the power at the low frequencies.
Thus, we are interested in how much power the signals
have at frequencies greater than 7.5 hz.
If these powers are rel-
atively low, then the error calculation is dominated by the motions
lower than 7.5 hz.
One way to estimate the power of the true acc-
eleration signal is to consider the power spectrums of the aliased
acceleration signal. The power spectrum is only good to half the
sampling frequency.
Although the aliased signal may have greater
powers in the frequency range of 0-7.5 hz than the true signal,
the shape of the aliased power spectrum can still useful.
It can
be useful because we can see if it is a reasonable aliased result
of a true spectrum with low relative power at frequencies beyond
7.5 Hz.
Figures 8.2-8.13
show the aliased power spectrums for sim-
ulator accelerations of typical runs of each flight manuever data
window.
Because of occasional overlap, the ILS approach and land-
ing segments were analyzed together.
the full motion condition.
All of these runs are for
The important question is whether or
120
not these aliased spectrums make the possibility of relatively low
power levels at frequencies greater than 7.5 hz. unlikely.
One interesting thing to notice about most of these power
spectrums is that their power does start decreasing before 7.5 hz.
The amount of decrease seems to be the least in the Ils simulator
roll acceleration plot.
Some of the plots
decrease and at fre-
quencies close to 7.5 hz start to increase again.
If the true
spectum really increases in this way, then these power levels are
high and aliasing could have a major effect on the vestibular
errors.
Figure 8.14
shows that relatively low power levels at
frequencies greater than 7.5 hz could also cause this effect if the
downward slope at frequencies greater than 7.5 hz is steeper than
the downward slope at frequencies lower than 7.5 hz.
This is be-
cause aliasing magnitude effects are symmetric and power is a magnitude computation.
For physical reasons that are explained below,
the the speculation that the signals were aliased in a way similar
to figure 8.14 is more likely than the speculation that the true
power spectrum really increases at frequencies greater than 7.5
hz.
The speculation that there is relatively low energy at the
higher frequencies is supported for the following physical reasons.
Actual aircraft motions of large amplitude at these high frequencies
takes a significants amount of energy because the aircraft is so
massive.
This energy can come from the aircraft itself or from
disturbance forces.
The aircraft is not designed to operate
such high energy vibrations in the cockpit.
with
Disturbance effects
usually do not sustain such high energies for a significant amount
121
of time.
We are concerned with the accelerations in the cockpit,
although the wings tips may be vibrating at high energy levels,
amount that the cockpit is vibrating is usually much lower.
the
These
assumptions are justified in the simulator for similar reasons.
Making the actuators move such a massive simulator at such high
frequencies at large amplitudes would also take a lot of energy.
In addition, the simulator is limited to motions no higher than 15
hz.
This limit is probably set by the computer rate and not by
the highest frequencies for which the simulator and its
actuators
are capable.
Typical power spectrums were only considered for the full
motion case.
The actual aircraft should have spectrums similar to
these because of the reasons stated above.
The power spectrum of
jostle motion for the rotational degrees of freedom should be similar to the corresponding power spectrums of special effects only
motion.
effects.
Jostle motions are confined to heave and sway and special
Special effects motion consists primarily of high fre-
quency low amplitude motions.
still below 7.5 hz.
However these motion are probably
It is reasonable to assume that the power
spectrums for the jostle and special effects only motion conditions
will also have low power levels in the spectrum of 7.5-15 hz. relative to the power levels of motions in the frequencies less than
7.5 hz.
This is because 7.5 hz is still very fast even for special
effects.
In summary aliasing could have affected the rotational vestibular error results.
If so,
the extent and the direction of the
effects is uncertain due to unpredictable shifts in phase.
122
In any
case, aliasing probably did not effect the full motion rotational
vestibular error results much because the powers at frequencies
above 7.5 hz were probably low and consequently calculations were
probably dominated by low frequency motions.
The aliased power
spectrums examined for this condition do not contradict this result.
Although the aliased power spectrum were not computed for the other
motion conditions it is probable that aliasing effects will also
be small for similar reasons.
Calculations are currently being done with the data that includes the lead term of the semicircular canal model.
Aliasing
will not be a factor in these calculations because these semicircular canal responses were calculated at 30 hz.
The effect of
the zero will be to add gain and phase lead at frequencies greater
than 3.18 hz.
to the vestibular response.
The differences between
the rms error results of this study and the results of the new
analysis are uncertain.
123
01.
7.5 Hz
1.0~
Figure 8.2: Power spectrum of simulator yaw acceleration
for typical engine flame-out run.
10"6
10-li
1.010-1(
7.4 Hz
Power spectrum of simulatro roll acceleration
Figure 8.3:
turn run.
steep
for typical
124
101. 1
1t
0-
.10'"
7.5 Hz
Figure 8.4: Power spectrum of simulator pitch accleration
for typical steep turn run.
:1.0'"'
1
I
I
I
t
III
I
I
I
f1
1.
Q8
i.0-(
7.5 Hz
Figure 8.5:
Power spectrum of simulator yaw acceleration
for typical steep turn run.
125
T
I li
II
i
I
I
T-1
1
AufyI~4W~
1.Ol
1.0-6
iO~
.1.
O**
I0
7.5 Hz
Power spectrum of simulator pitch acceleration
Figure 8.6:
maneuver.
for typical stall
I
II
I
1111I
I
It1I
1
1
1
1.0"i
108
1.0"s
IV.(
1 1ii I 1
I I
7.5 Hz
Figure 8.7:
Power spectrum for simulator roll acceleration
for typical stall maneuver.
126
I
I
I
111
I
1
1
1
1
1
1I
I
I
7.5 Hz
Figure 8.8: Power spectrum for simulator roll acceleration
for typical turn with yaw damper failure.
10-4
10
1 1 1
1,1
1
1
1
1
1111J
-
I
I
7.5 Hz
Figure 8.9:
for typical
Power spectrum for simulator yaw acceleration
turn with yaw damper failure.
127
I. 0~
I
I
I I . I I .
I
I
.
I
t I I
i
I
1.0" ~
it,0"~
1>1
.1.
0"~
1.0'
1.0 ~
:1,
()-i)
10"1O
.0"11
1()-I 2
I
I I pI I I
I
I
.
. . I
7.5 Hz
Figure 8.10:
Power spectrum of simulator pitch acceleration
for typical turn with yaw damper failure.
U
1.0-12,
Ir~
JI
U
m
I
U
U
U
U
p
a
10-71
11111111
I
I
1111
III
I
'liii
I
7.5 Hz
Figure 8.11:
Power spectrum of simulator roll acceleration
for typical LLS approach and landing.
128
,.v
7.5 Rz
Figure 8.12: Power spectrum of simulator pitch acceleration
for typical ILS approach and landing .
1l
1j
1i
1
1
1
1
1
1
1
7.5 Hz
Figure 8.13: Power spectrum of simulator yaw acceleration
for typical ILS approach and landing.
129
Hypothetical Aliased
Spectrum Shape
Hypothetical True Spectrum
Shape
0
7.5 Hz
7.5 Hz
log frequency
Figure 8.14:
log frequency
Hypothetical Example of how spectrum may
be aliased to have apparent incresed power
at frequencies near the critical frequency
130
CHAPTER 9:
9.1
CONCLUSIONS AND RECOMMENDATIONS
Conclusions
This study shows that most of the differences found between
motion conditions were not statistically significant for any of the
realism parameters examined.
These include pilot opinion, pilot
performance, acceleration and vestibular errors.
Because the
assumptions of the statistics have not been proven true for this
data, the actual confidence levels found are not assured, they are
only useful as a rough guideline to what is going on.
More impor-
tant is the fact that most of these differences were within reasonable bounds when the analysis consisted of comparison plots for the
means of the realism measurement over the group of pilots performing
the same flight scenario.
The fact that for most cases, major differences were not found
in the motion conditions to any of the realism parameters is very
important because it shows that fidelity is not necessarily gained
by adding more motion capability to the system.
Expensive full
six degree of motion systems may not perform that much better than
less capable systems (at least for the conditions of this experiment).
In fact, full motion can be less realistic than the other
conditions.
For many of the rotational degrees of freedom, the vestibular
error was below the threshold level of perception.
On the other
hand, a lot of the translational degrees of freedom errors were
above the threshold of perception.
This is important and may be
useful in future simulator controller design.
131
However, the validity of conclusions about the rotational
vestibular errors could be weakened by the possibilty aliasing
effects.
It was argued that these effects would be small due to
low power levels of both the simulator and the aircraft at frequencies greater than 7.5 hz.
In addition the semicircular canal
responses were computed without lead term that adds gain and phase
lead at frequencies greater than 3.18 hz.
9.2
Recommendations
1.
The rotational vestibular error analysis be redone with
the seimicircular canal model responses calculated by the simulator
computer.
The lead term is included in this model and aliasing
will not be a problem.
2. More specific opinion questions could be asked of the
pilots, e.g.comparative questions about the response for specific
degrees of freedom for different motion conditions.
3.
As this was a study of quantitative comparisons of the
error levels between motion conditions, the results of a time-series
statistical analysis which considers the functionality of the data
would be useful.
4.
For improving the design of this simulator,
the errors in
the rotational degrees of freedom should be increased in order to
reduce the errors in the translational degrees of freedom.
This
is because rotational errors of were frequently much lower than
the perception level threshold; while translational errors were
frequently much higher.
132
REFERENCES
Fernandez, C. and Goldberg, J.M. "Physiology of Peripheral Neurons
Innervating Otolith Organs of the Squirrel Monkey I, II, III" J.
Neurophys. 39: 970-1008, 1976.
Friedland, B., Ling, C.K., and Hutton, M.F., "Quasi-optimum design
ofcontrol systems for moving basesimulators", NASA CR 1614, 1970.
Hosman, R.J.A.W., and van der Vaart, J.C. "Vestibular Models amd
Thresholds of Motion Perception. Results of Tests in a Flight Simulator.
Delft University of Technology, Report LR-265, 1978.
Ish-Shalom, J. Design of Optimal Motion for Flight Simulators,
Ph.D. Thesis, MIT, 1982.
Jongkees, L.B.W "On the Otoliths: Teir function and the way to
test them, Third symposium on the role of vestibular organs in space
exploration, 1967,NASA SP-152,1968.
Kosut, R.L., "Nonlinear optimal cue-shaping filters for motion
base simulators." J Guidance and Control 2:486-490, 1979.
Lee, A.T., and Bussolari, S.R,"Flight Simulator Requirements For
Airline Transport Pilot Training: An Evaluation of Motion System Design
Alternatives," Proceedings of the IEE Second International Conference
on Simulators, Universtiy of Warwick, UK, Sept. 1986.
Oman, C.M., A Heuristic Mathematical Model for the Dynamics of
Senory Conflict and Motion Sickness,"
Acta Oto-Laryngologica Supplement
392, 1982.
Ormsby, C.C. "Model of Human Dynamic Orientation" Ph.D. Thesis, Massachusetts Institute of Technology, 1974.
Parrish, R.V., and Martin, D.J., "Comparison of a linear and
nonlinear washout for motion simulators utilizing objective and subjective
data from CTOL transport landing approaches", NASA TN D 8157, 1976.
Peters, R.A., "Dynamics of the Vestibular System and Their Relation
to Motion Perception, Spatial Orientation, and Illusions, NASA CR-1309,1969.
Sinacori, J.B., Stapleford, R.L., Jewell, W.F. and Lebman,J.M.,
"Researchers Guide to NASA AMES Flight Simulator for Advanced Aircraft
(FSAA),"NASA CR-2875, August, 1977.
Sivan, R., Ish-Shalom, J. and Huang, J.K. "An Optimal Control Approach
to the Design of Moving Flight Simulators," IEEE Transactions on Systems,
Man, and Cybernetics, Vol. SMC-12, No. 6, November/December, 1982.
Steer, R.W., The Influence of Angular and Linear Acceleration and
Thermal Stimulation on the Semicircular Canal, Sc.D. Thesis, MIT, 1967.
Sturgeon, W.R.,
"Controllers for aircraft motion simulators", J
133
Guidance and Control 4:184-191, 1981.
Sullivan, R.B., The Use of Vestibular Models in Flight Simulator
Motion Washout Systems: An Experimental Evaluation, SM Thesis, MIT, 1985.
Von-Gierke,H.E. and Steinmetz, E. "Motion Devices for Linear and
Angular Oscillation and for Abrupt Acceleration Studies on Human Subjects
(Impact)", Publication 903, National Academy Of Sciences-National
Research Council, Washington, D.C.,1961.
Wersall, J. and D. Bagger-Sjoback "Morphology of the Vestibular
Sense Organs", in H.H. Kornhuber ed. Handbook of Sensory Physiology.,
vol. VI, Vestibular system Partl: Basic Mechanisms, Springer-Verlag,
Berlin, Heidelberg,new York, 1971.
Young, L.R. and Meiry, J.L. A revised dynamic otolith model.
Med 40:606-608, 1968.
Young, L.R. and Oman, C.M. "Model for vestibular adaptation to
horizontal rotation," Aerospace Medicine, 40(10):1076-1080, 1969.
134
Aerospace
SUMMARY OF CONTENTS OF APPENDIX A:
This appendix is a summary of all results pertaining to pilot
opinion.
It includes a list of pilot comments and experimenter
notes on the debrief session.
The raw data answers to the opinion
questionnaire and anova tables are also included.
divided into groups who flew the same scenario.
135
The pilots are
PILOT COMMENTS AND NOTES ON DEBRIEF SESSION
ENGINE FLAME-OUT SCENARIO PILOTS
PILOT 1;
Jostle and special effects motion "felt more realistic"
pertaining to external inputs to aircarft.
Subject reports "wing low" in jostle and special effects
motion.
Felt there was some wind in full motion.
PILOT 3:
Pitch sensitivity greater than aircraft.
Needs to pull too hard on take-off.
Flight director unfamiliar.
PILOT 9;
No perception of altered fidelity of specifically motion.
PILOT 13:
On retrospect, sustained acceleration perceived.
PILOT 15:
For all conditions roll response not as "stiff"
as aircraft.
Tendency to disregard motion cues.
Acceleration/deceleration cues are needed in simulator.
For full motion, pilot reported "swerve to right on first
take-off which may have been late rudder input."
For special effects only motion, on "second take-off at about
170 kts,flap
change seemed to require more immediate
attention to rudder trim change."
AIRWORK SCENARIO PILOTS
Pilot 2:
Eye fatigue on jostle motion condition.
Pilot 10:
Jostle motion best in pitch, but not realistic in roll.
Special Effects motion has too much roll response.
Full motion too little roll response.
Pitch response in familiarization flight is unlike aircraft
on rotation
Jostle motion most like aircraft but easier.
Little or no learning in steep turns.
Minimal learning in stalls
Little bit
of learning in turns with yaw damper failure, but
used techniques learned in training.
136
Special effects only motion not as desirable or realistic.
Pilot 14:
Briefing on dutch roll technique would be helpful.
Pilot 17:
Too much lag especially in elevator.
ILS Approach and Landing Scenario Pilots
Pilot 7:
In full motion felt left cross wind was recurrance of the
left side bias.
Observed trim inputs seemed to take too long to affect stick
force of ADI indication.
Felt mushy on rotation.
Jostle motion: "easiest to fly of all of them."
Overall: Not sure how faithful HSI Sperry flight director is.
Feels that motion may not be necessary. He would not have
said this before.
Pilot 12:
Control forces during configuration changes seemed slightly
excessive.
Control response of simulator seemed faster than aircraft.
Pilot 16:
Lag in controls, mushy on controls.
little motion.
In retrospect, felt a
Pilot 18:
In familiarization scenario, "felt an unusual pitch up, but
maybe its me."
137
flame out subjects opinion data
engine
1
5
6
7
3
3.00 2.42
3 .00 4.00
3 .00 3.00
3.50
3 .00
3.00
3 .00
4.00 4.00
3 .00 4.00
3 .00
3.00
3 .00
2.71
3.00
4.00
4.00
2
3
4
1
,,
,
IVEntI
2
1
Column:
5C~LP ,r~ff~?&.i9'~
4tj
Analysis of Variance
Source
of Variation
Among
Within
Adj Total
Square
18
.2692667
4.891857
20
5.161124
2
Groups
Groups
Table
Mean
Degrees of Sum of
Freedom
Squares
.1346333
.2717699
F
Ratio
0.50
Probability
0.617
The columns used in this analysis are
Column
Column
Column
I label
2 label
3 label
Ericrne.
Column:
1
2
3
4
5
6
7
:scenario
:scenario
:scenario
PLamrwe
4
Mj~cqfN . 3,25 ;A.
- our
6
5
3.00 3.42
3.00 3.00
3.00 3.00
3.58 3.79
3.00 3.00
4.00 3.00
3.00 4.00
workload 6 dof
workload 2 dof
workload 0 dof
2.92
2.00
2.00
3.33
3.00
3.00
4.00
2 , SI
.21
Analysis of Variance Table
Source
of Variation
Degrees of Sum of
Freedom
Squares
Among Groups
Within Groups
2
18
.6946381
5.117486
Adj Total
20
5.812124
.. . . . . . . . . . . . . . .
The columns used in this
Column
Column
Column
4 label
5 label
6 label
Mean
FProbabilit
Square
Ratio
.347319
.2843048
0.318
....................................
analysis are
:configuration workload 6 dof
:configuration workload 2 dof
:configuration
workload 0 dof
138
1.22
ep tG
pe
8
7
Co1tmn:
ff
r-LJ+
6 LAtI
9
3.50
3.00
4.00 4.00 4.00
2.75 3.63 3.50
3.00 4.00 3.00
4.00 4.00 3.00
3.00
3.00 3 .00
3.83
4.00 3.50
1
2
3
4
4.00
5
6
7
3-L
3A9q 3 I1
Oen C r-
Analysis of Variance Table
Source
Among Groups
Within Groups
2
18
.6327714
3.803429
Total
20
4.4362
Adj
The columns used in this
Column
Column
Column
7 label
8 label
9 label
1
2
3
4
5
6
7
10
.3163857
.2113016
1.50
Probability
0.250
analysis are
6 U*T
rLJArME
11
12
3.00 2.50 2.17
4.00 4.00 4.00
2.00 2.00 2.00
2.79 2.92 2.83
3.00 3.00 3.00
3.00 3.00 2.00
4.00 3.00 3.00
r, I f-r I Q 3.11 2.'41
Z7
2,71
. 27
'13
Analysis
Source
of Variation
Among Groups
Within Groups
of
Variance
Degrees of Sum of
Freedom
Squares
2
.
Table
Mean
Square
.5560667
8.309657
18
Adj Total
20
.. . . . . . . . . ..
.2780334
.4616476
10 label
11 label
12 label
:training
:training
:training
F
Ratio
0.60
Probability
0.558
8.865724
..........................................
The columns used in this analysis are
Column
Column
Column
F
Ratio
:simulator response 6 dof
:simulator response 2 dof
:simulator response 0 dof
9r~,36pJE
Column:
Mean
Square
Degrees of Sum of
Freedom
Squares
of Variation
utility
utility
ut.ility
139
6 dof
2 dof
0 dof
I- F0r1 E
Column:
1
2
3
4
5
6
7
13
14
3.42
4.00
2.00
2. 71
3 .00 3.00
2.00 2.00
4.00 3.00
3.42
4.00
2.00
3.00
A"I Vy,
CkT
15
2.67
4.00
2.00
3.00
3.00
2.00
3.00
?,~
Z4~)
Analysis
of Variance Table
Degrees of Sum of
Freedom
Squares
Source
of Variation
Among
Groups
Within Groups
2
18
.2351524
10.15977
Adj
20
10.39492
Total
Mean
Square
.1175762
.5644317
F
Ratio
0.21
Probability
0.814
The columns used in this analysis are
Column
Column
Column
13 label
14 label
15 label
:checking
:checking
:checking
FerYaE (OUIr
ENG(NE
16
Column:
1
2
3
4
5
6
7
)Vk'I-
2.92
3.00
3.00
3.00
4.00
3.00
4.00
17
3.00
4.00
3.00
2.96
4.00
3.00
3.00
utility 6 dof
utility 2 dof
utility 0 dof
18
2.80
3.00
3.00
2.83
4.00
3.00
4.00
'D
0
I~
2
~
Analysis of Variance Table
Source
of Variation
.............
Among
Groups
Degrees 9f Sum of
Mean
F
Probability
Freedom
Squares
Square
Ratio
.........................................
2
9.2666679-03
4.633334E-03
0.02
0.982
Within Groups
18
4.621914
Adj Total
.............
20
4.631181
.........................................
The columns used in this analysis are
Column
Column
Column
16
17
18
label
label
label
:realism
:realism
:realism
140
6 dof
2 dof
0 dof
.256773
pilot opinion data
Airwork scenario
Column:
-
rv i~E
2
2w,37
3 .00
4.00
4.71
4.00
3.00
1 4S
.14
S-4
v I
1 16
00: M e-A IJ
3
2.00
4.00
3. 58
2.00
3 .00
4.00
3.00
3.00
4.00
3.00
3.21
4.00
3.00
1
2
3
4
5
6
,,,
1
Analysis of Variance Table
Source
of Variation
Among
Mean
Degrees of Sum of
Freedom
Squares
Groups
Square
F
Ratio
2
.1068778
5.343889E-02
Within Groups
15
8.8505
.5900334
Adj Total
17
8.957377
0.09
in this analysis
The columns used
Column
Column
Column
1 label
2 label
3 label
:scenario
:scenario
:scenario
Probability
0.914
are
workload 6 dof
workload 2 dof
workload 0 dof
Airwork scenario pilot opinion data
4
Column:
1
2
3
4
5
6
MEW
6
5
3.00 4.00
4.00 4.00
3.00 2.00
3.17 3.33
5.00 3.00
3.00 2.00
2.00
3.00
4.00
3.92
4.00
2.00
3153
;.06
S (5
'
- 11
,L
5nr-txW.
Analysis of Variance Table
Source
of Variation
Degrees of Sum of
Mean
F
Freedom
Square
Ratio
Squares
Among Groups
Within Groups
2
15
.7486778
12.15017
Adj Total
17
12.89884
.3743389
.£100111
The columns used in this analysis are
Column
Column
Column
4 label :configuration workload 6 dof
5 label :configuration workload 2 dof
6 label :configuration workload 0 dof
141
Probability
0.46
0.639
Airwotk scenario pilot opinion data
1
2
3
4
5
6
3.00
3.00
4.00
4.00
4.00
3.00
9
8
7
Cobumn:
3.00
2.75
4.00
4.67
5.00
2.00
3.00
3.00
3.00
3.21
4.00
1.00
,,s
flVl F Ar N
s
SA1) orTaeavi
zt.14ff
*q3
Analysis of Variance Table
Source
Degrees
of Variation
Freedom
Among
Within
Groups
Groups
of Sum of
Squares
2
Adj Total
15
1.792478
13.34208
17
15.13456
Mean
Square
F
Ratio
.8962389
.8894723
1.01
Probability
0.389
The columns used in this analysis are
Column
Column
8 label
Column
9 label
7 label
:siaulator response 6 dof
:simulator response 2 dof
:simulator response 0 dof
Airwork scenario pilot opinion data
Column:
1.
2
3
4
6
3.00
3.50
4.00
4.00
4.00
3.00
3,51
rAvjEAO~
11
10
12
4.00
2.92
2.00
2.58
4.00
3.00
2.00
2.50
4.00
4.00
4.00
2.00
3tol
3.o%
Analysis of Variance Table
Source
of Variation
Degrees of
Sus
Freedon
Squares
of
F
Square
Ratio
Among Groups
Within Groups
2
1
.5
15
9.5578
.6371867
Adj
Total
17
10.5578
The
columns used
Column
Column
Column
10 label
in this analysis are
:training utility
11 label :training utility
12 label
:training
utility
142
Probability
moan
6 dof
2 dof
0 dof
0.78
0.474
Airwork scenario pilot opinion data
Column:
1
2
3
4
5
6
rvi F-A J
15
14
13
2.00 1.00 3.00
3 .00
2.50 2.83
2.00 2.00 1.00
4.00 4.00 2.33
1.00 4.00 4.00
2 .00 2.00 2.00
2.'5S
2,33
ql.
S1) OF Woan
L.jq
Analysis of Variance Table
Probability
F
Mean
Degrees of Sum of
Source
Ratio
Square
Squares
of Variation Freedom
---------------------------------------.. . . . . . . . . . . . .........
0.09 0.917
.1030889
.2061778
2
Among Groups
1.178347
17.6752
15
Within Groups
17.88138
17
Adj Total
................................................
The columns used in this analysis
are
13 label :checking utility 6 dof
14 label :checking utility 2 dof
15 label :checking utility 0 dof
Column
Column
Column
Airwork scenario pilot opinion data
Column:
16
1
2
3
4
5
6
rv
i !, n'
3.00
2.00
4.00
4.88
2.00
4.00
1,1
17
2.00
2.50
4.00
4.13
4.00
3.00
18
4.00
3.83
3.00
3.83
4.00
2.00
3,2./
.33
Analysis of Variance Table
Source
Degrees of Sum of
Mean
F
Probability
of Variation Freedom
Squares
Square
Ratio
.........
..................................................
Among
Groups
2
9.621111E-02
4.810556E-02
0.05
0.951
Within Groups
15
14.22795
.94853
Adj Total
.........
17
14.32416
..................................................
The columns used in this analysis are
Column
Column
Column
16 label
17 label
18 label
:realism
:realism
:realism
6 dof
2 dof
o dof
143
ILS approach and landing pilot
opinion data
Column:
1
2
3
4
5
6
1
2
3.00
4.00
3.33
2.00
3.00
4.00
4.00
4.00
3.75
2.00
4.00
4.00
4.00
4.00
3.50
2.75
3.00
5.00
3
Analysis
Source
of Variation
of Variance
Degrees of Sum of
Squares
Freedom
Among Groups
Within Groups
2
15
.8129333
9.334917
Total
17
10.14785
Adj
Mean
Square
ILS approach and landing pilot
4
1
3.00
2 4.00
3 3.75
4
2.50
5 4.00
6 4.00
0.65
.4064667
.6223278
Probability
0.535
:SCENARIO WORKLOAD 6 DOF MOTION
:SCENARIO WORKLOAD 2 DOF MOTION
:SCENARIO WORKLOAD 0 DOF
1 label
2 label
3 label
Column:
F
Ratio
in this analysis are
The columns used
Column
Column
Column
Table
5
4.00
4.00
3.00
3.00
4.00
3.00
opinion data
6
3.00
3.00
3.00
2.58
4.00
4.00
9,24
:) N;-: A t3
.2q
Analysis of Variance Table
Source
of Variation
Degrees of Sum of
Freedom
Squares
Among Groups
Within Groups
2
15
.2704333
5.312417
Adj
Total
17
5.58285
The
columns
Column
Column
Column
used in this analysis
4 label
5 label
6 label
Mean
Square
.1352167
.3541611
are
:CONFIGURATION WORKLOAD
:CONFIGURATION WORKLOAD
:CONFIGURATION WORKLOAD
144
F
Ratio
6 DOF
2 DOF
0 DOF
0.38
Probability
0.689
ILS approach and landing pilot opinion data
1
2.00
2.00
2.58
3.50
2.00
2.00
2
3
4
5
6
ftv EA PJ
9
8
7
Column:
3.00
2.00
3.00
3.67
2.00
4.00
2.00
3.00
3.75
3.58
2.00
1.00
?..'w
2,1
2,560
- 5
14
1,N
Sb OF MOM
-------------------------.....................
Analysis of Variance Table
Source
of Variation
Degrees of Sum of
Freedom
Squares
Among
Groups
Within Groups
2
15
1.107011
11.00683
Total
17
12.11384
Adj
The columns used in
Column
Column
Column
7 label
8 label
9 label
this analysis
Mean
Square
.5535055
.7337889
F
Ratio
0.75
Probability
0.487
are
:SIMULATOR RESPONSE 6 DOF
:SIMULATOR RESPONSE 2 DOF
:SIMULATOR RESPONSE 0 DOF
ILS approach and landing pilot opinion data
Column:
1
2
3
4
5
6
10
2.00
3.00
4.00
2.67
4.00
4.00
11
4.00
3.00
3.00
2.00
4.00
2.00
12
3.00
3.00
3.50
2.63
4.00
4.00
7
3, S.,36
.43 . 3
1.28
r) OF m
.is
Analysis
Degrees of Sum of
Squares
Freedom
Source
of Variation
Among Groups
Within Groups
The columns
Column
Column
Column
Mean
Square
.4187444
9.274834
2
15
17
.. . . . .
Adj Total
. . .. .
of Variance Table
.2093722
.6183222
0.34
Probability
0.718
9.693578
.............................................
used in this analysis are
10 label
11 label
12 label
F
Ratio
:TRAINING UTILITY 6 DOF
:TRAINING UTILITY 2 DOF
:TRAINING UTILITY 0 DOF
145
ILS approach and landing pilot opinion data
3 .00
3 .00
2.67
2.71
3.00
5.00
1
2
3
4
5
6
15
14
13
Column:
4.00
3.00
3.83
2.67
3.00
5.00
3.00
3.00
3.75
2.83
2.00
5.00
-2Lt
I'll
3,22,
rr-O
Analysis
Source
Degrees of Sum of
Freedom
Squares
of Variation
Among
of Variance Table
Groups
Within Groups
2
15
.4567111
12.80707
Adj Total
17
13.26378
Mean
Square
.2283556
.8538045
F
Ratio
0.27
Probability
0.769
The columns used in this analysis are
Column
Column
Column
13 label
14 label
15 label
:CHECKING UTILITY 6 DOF
:CHECKING UTILITY 2 DOF
:CHECKING UTILITY 0 DOF
ILS approach and landing pilot opinion data
16
Column:
17
18
1
3.00 4.00 3.00
2
3
4
5
6
3.00 3.00 3.00
3.00 3.08 3.00
e),ooftwI
2.71 3.00 2.67
2.00 2.00 2.00
2.00 3.00 1.00
4
,1
Analysis of Variance
Source
of Variation
Among Groups
Within Groups
Adj Total
....
...
Degrees of Sun of
Squares
Freedom
17
........
The columns used in this
Column
Column
Column
16 label
17 label
18 label
Mean
Square
1.018144
6.476167
2
15
Table
.5090722
.4317445
F
Ratio
1.18
Probability
0.335 ,
7.494311
.....................................
analysis are
:REALISM 6 DOF
:REALISM 2 DOF
:REALISM 0 DOF
SUMMARY OF CONTENTS OF APPENDIX B
This appendix contains a copy of the briefing material that
was given to the subjects to read before performing the experiment.
MIT/NASA FLIGHT SIMULATOR EVALUATION EXPERIMENT
PILOT BRIEFING SUMMARY
Pilot
ID -
Crew ID-
--
Date
147
:..
;
ZNTDUCTI0fN
:.ur.ose of t.i-s stucV is to examtigne t he roIe of flight
alot training ana certification.
simulator fIdelity in a r transoort
The ex-er-men-s cescribed belcow are cesignec to icentify potential
simulator ceficiencies and to determine metrlocs to improve the
effectiveness of Flicht simulation.
An imoortant elemyient of this
investication is the evaluation of various levels of simulation by
aircrew zarticioants in a series of realistic flight scenarios.
Each
oilot will
ex:erien-e the simulation at three different levels of
fidelity.
The pilots will be asked to evaluate the simulation durino
the course of the experiment by means of a series of
numerical ratino
scales.
The followina sections descrioe the nature of the fliaht
scenarios and the experimental procedure.
2.0
DE CO=TN 0F SULTOR
The simulator used in this study is manufactured by Singer-Link
and is essentially identical to those used by Delta Airlines in their
pilot training orogram.
it is designed to simulate the Boeing 727-200
series aircraft.
::.0
GENERAL EXPERIMENTAL PROCEDURE
Each aircrew Darticioant will be briefec on the safety features
of the NASA MVSRF B-727 flight simulator and on the experimental
procedure prior to the start of the exoeriment.
The test plan
provides for two aircrew participants for each experimental run.
One
particioant will act as pilot-in-command and provide the simulator
evaluation while the other assists in the aircraft operation. The two
particioants will then exchange roles and the exoeriment will be
reoeated. During the data collection portion of the experiment, the
crewmember acting as pilot in command will occupy the flight deck
position with which he/she has the most recent experience (i.e.
current F/O's will occuoy the right seat, etc.).
The flight portion
of the experiment will begin with a series of familiarization
maneuvers to enable the crewmembers to adjust to the soecific cockpit
layout of the simulator.
Then a different series of maneuvers will be
flown under varying levels of fidelity. After each set of
reoetitions, the pilot will be asked to evaluate the simulation by
means of the Flight Simulator Rating Forms.
4.0
FLIGHIT SCENARIOS
The flight portion of the experiment will consist of one
of the following four scenarios.
In general. Scenario I will
for the familiarization oeriod and one of the remaining three
used for the data taking portion of the exoeriment.
Aircrew
particioants are asked to set uo ane fly tie aircraft as they
148
or more
be used
will be
would
in
rno
.*
rat iora! env 1 rorlmernt.
a- the csarture
Scenario 1 -he siYulator wi I be n- . - Iz
9 14e200 lns.
Takeoff weiJort wiI,
enc of Rwy 28R at SF0.
- eather wil1 be clear with calm wrics.
7- eni'.ct wilI follow ATC
vectors around the traffic Dattern for a vssual a:iroach to a
touch-anC-co lanCing.
The Dilot wil: t ar follow vectors arounc
the traffic pattern for another visual aocroach to a full-stoo
landi rig.
Scenario 2:
The simulator will be initialized at the deoarture
end of Rwy -28R. Weather will be clear with calm winds.
Clearance
will be to fly runway heading, climb and maintain 4000 ft.
lakeoff weight will be :48,000 lbs.
An engine flameout during
takeoff will result in either a rejected takeoff or an engine-out
takeoff with a climb to the clearance altitude on the runway
heading.
Scenario 3: The simulator will be initialized at altitude in
level flight.
Weather will be clear with calm winds.
Aircraft
weight will be 148,000 lbs.
The pilot will perform two steeo 360
The pilot will then perform
degree turns, one left and one right.
two aooroach to stall maneuvers in the clean configuration.
The
oilot will then oerform additional turns as vectored by ATC (there
is the possibility of yaw damper failure during these turns).
Scenario 4:
The simulator is initialized in flight,
configuration, on a vector for the ILS ZSR.
Aircraft
be 148.000 lbs.
The weather will be 600 ft. overcast
visibility
with the oossibility of wind shear on the
Ao:roach will be hand flown with the flight director
will be terminated by a full stoo landing.
5.0
in the clean
weight will
ten mile
aooroach.
coucled and
Si.ULATGR EVALUATI0N
After the aircrew particioant flies the exoerimental scenario
under a soecific level of simulation fidelity, he/she will be asked to
evaluate the simulation according to the Flight Simulator Rating Form.
Because both marticipants will be acting as pilot in command during
the course of the exoeriment, it is necessary to perform the
evaluation in such a way that the responses of one pilot do not
This will be avoided by the -use
influence the resoonses of another.
of a written evaluation form.
In addition, each aircrew particioant
will act as pilot in command on a different scenario from that
performed by the other participant.
The questions in the Flight Simulator Evaluation Form are
desioned to orovide you with a scale by which to rate certain
simulator characteristics in soecified fliaht maneuvers.
Each item
consists of a scale of characteristics, with the numbers 1-5.
For
example:
149
e
DturtC. taxi, ri
tie
actual aircraftC
i.
------
s1aefrr;
-
-
0
- -
tTre
sil!ator
- --- - - -
-
hac,
as comoared
- - -- 4--------------5
Maj 0Or
some minor
no
deficiencies
ceficiencies
deficiencies
to
You are to cnoose
The numoers form a scale between two extremes.
a number that best exoresses your rating of the simulator on the scale
For exanole, if you think that the simulator had no
orovicec.
deficiencies in nosewieel steerinc as comoared to the actual aircraft
during taxi, you woud Cloose 1. If you believe that the simulator
had only minor deficiencies, you might choose ., and so forth.
Once you have selected trse number that best describes your
If the
evaluation, circle the aporooriate number on each scale.
cuestion refers to a maneuver or maneuver segment that you have not
attemoted in the simulator, leave the scale blank and go on to the
The cuestions in the Flight Simulator Rating Form have
next cuestion.
no right
or wrong answers.
Please read each question carefully
and
Do not soend a
mark the resoonse that BEST reflects your evaluation.
lot of time on each one, your FIRST response is usually the best.
-In order to assure that
your identity is protected,
there will
be no
record of your name or affiliation on the Flight'Simulator Rating
Form or on the attached
Aircrew
Participant
150
Background
Questionnaire.
How Ionc
have you been emoloyed by your oresent
airline?
years.
2.
How lona have you served as a pilot in the following categories?
______
General Aviation
Cosymuter/Air
Airline
Taxi
______
years.
____-_
years.
years.
Military:
.anker/Transoort/Bomber _.._._ years.
Fighter
Helicooter
Other
3.
_____.
(soecify)
years.
._____
Indicate the total time you have served in each of the following
crew positions in the B-727 aircraft.
years.
First Officer ._____
Flight
years.
Engineer ._._._
Estimate your total
Other transoort
Other flight
What
hours.
simulator ______-hours.
simulator
is your age?
years.
hours in flight simulators in the past year.
B-727 simulator
6.
Circle the appropriate
Caotain
First Officer
Captain ------
5.
years.
What is your present crew positign?
answer.
A.
B.
4.
years.
____--
._____
____
hours.
years.
151
SUMMARY OF CONTENTS OF APPENDIX C:
This appendix contains the raw data used for the acceleration
error analyses.
The translational degrees of freedom acceleration
errors are in units of m/sec 2 ;
while the rotational degrees of
freedom acceleration errors are in units of rad/sec 2 .
The means
and standard deviation of the means for each motion condition,
over all pilots performing the same manuever are presented in
tables.
The analysis of variance tables are presented with the
data that is associated with them.
scenario.
The data is presented by flight
Engine flameout results are presented first.
The
acceleration error results from the airwork scenario follow, with
steep turns presented first
followed by stall
the results from the rate turn with
presented.
results,
and then
a yaw damper failure are
The results from the approach segment of the ILS
appraoch and landing scenario precedes the results from the landing segment of this scenario.
LIST OF CONTENTS OF APPENDIX C:
Engine flameout yaw acceleration error data and analysis
Steep turn pitch acceleration error data and analysis
Steep turn roll acceleration error data and analysis
Steep turn yaw acceleration error data and analysis
Stall roll acceleration error data and analysis
Stall pitch acceleration error data and analysis
Stall longitudinal acceleration error data and analysis
Rate Turn with Yaw Damper Failure pitch acceleration error data
and analysis
152
Rate Turn with Yaw Damper Failure roll acceleration error data and
analysis
Rate Turn with Yaw Damper Failure yaw acceleration error data and
analysis
ILS Approach and Landing 500' -200' pitch acceleration error data
and analysis
ILS Approach and Landing 500'-200' roll acceleration error data
and analysis
ILS Approach and Landing 500' -200' yaw
acceleration error data
and analysis
ILS Approach and Landing 500'-200' vertical acceleration error
data and analysis
ILS Approach and Landing last 20-25 sec pitch acceleration error
data and
analysis
ILS Approach and Landing last 20-25 sec roll acceleration error
data and analysis
ILS Approach and Landing last 20-25 sec yaw acceleration error
data and analysis
ILS Approach and Landing last 20-25 sec vertical acceleration
error data and
analysis
153
AccE LE1 AT)0tW
ENGINE FLAME-OUT YAW,,ERROR
1
Column:
1
2
a3
4
5
6
7
8
9
10
11
12
.00862
.00932
.00745
.00961
.01197
.00766
.00840
.00917
.00845
.00989
.00769
Col
Col (
Col(
Col
Col (
Col(
2
3
.00922
.00879
.01319
.00977
.00622
.00763
.00882
.01229
.01238
.00834
.01428
.02017
.01320
.01120
.00804
.00911
.00919
.00695
4
.00007
.00009
.00006
.00009
.00014
.00006
.00007
.00008
.00007
.00010
.00006
6
.00009
.00008
.00020
.00041
.00017
.00013
.00006
.00008
.00017
.00010
.00004
.00006
.00008
.00015
.00015
.00007
.00008
.00005
SUBJECT
1)6 DOF RMS ERROR
2)2 DOF RMS ERROR
3)0 DOF RMS ERROR
4)6 DOF MEAN SQUARE ERROR
5)2 DOF MEAN SQUARE ERROR
6)0 DOF MEAN SQUARE ERROR
MEAN
ROW
1:
ROW
2:
ROW
3:
ROW
4:
ROW
5:
ROW
6:
ROW
7:
ROW
8:
ROW
9:
ROW 10:
ROW 11:
ROW 12:
STANDARD DEVIATION OF THE MEAN
COLUMN 1
.00
COLUMN 2
COLUMN 3
5
3
SC00072
00 152
CA
154
3
3
8
8
9
9
13
13
15
15
ERROR ANALYSIS
-1GINEFLAME OUT SCENARIO YAW ACCELERATION
CONT"D
Analysis of Variance Table
-Source
of Variation
Among
Groups
.c
Within Groups
Degrees of Sum of
Squares
Freedom
F
Ratio
Mean
Square
2
3.183238E-05
1. 591619E-05
26
1.925218E-04
2.15
Probability
0.137
7.404685E-06
Adj Total
2.243542E-04
28
The columns used in this analysis are
Column
Column
Column
1 label :6 DOF RMS ERROR
2 label :2 DOF RMS ERROR
3 label :0 DOF RMS ERROR
Analysis of Variance Table
Source
of Variation
Amng
Groups
Degrees of Sum of
Squares
Fraeeo
1
Mean
Square
F
Ratio
Probability
3.100915E-05
3.100915E-05
3.60
Within Groups
17
1.463808E-04
8.610632E-06
Adj Total
18
1.773899E-04
The columns used in this analysis are
Column
1 Label :ColumI 1
Column
3 label :Column 3
155
0.072
steep turn pitch acceleration error analysis
3
2
1
Column:
4
6
5
12. .00793 .00690 .01239 .00006 .00005 .00015
.00002 .00003
.00469 .00517
.2
.00711
.00664
.01557
.00899
4
5
6
1)6
2)2
3)0
4)6
5)2
6)0
Col(
Col(
Col (
Col
Col(
Col(
COLUTN 1
MEAN
-09
Z5
COLU N 2
COLUMN
.00379 .01119
.00394 .00435
.00415 .00442
.00195
dof
dof
dof
dof
dof
dof
.00005
.00004
.00024
.00008
.00001 .00013
.00002 .00002
.00002 .00002
.00000
rms error
rms error
rms error
mean square error
mean square error
mean square error
ROW
ROW
ROW
ROW
ROW
ROW
1:
2:
3:4:
5:
6:
SUBJECT
2
5
10
11
14
17
STANDAR.D DEVIATION OF THE %,AN
00163
O H2-o.
ooi172
3
-60750
Analysis of Variance Table
Source
of Variation
Among
Groups
Degrees of Sum of
Squares
Freedom
Mean
Square
F
Ratio
2
7.185664E-05
3.592832E-05
13
1.283327E-04
3.64
Within Groups
9.871749E-06
Adj Total
15
2.001894E-04
The columns used in this analysis are
Column
Column
Column
1 label :6 dof rms error
2 label :2 dof rms error
3 label :0 dof rms error
156
Probability
0.056
CONTINUED
STEEP TURNS PITCH ACCELERATION ERROR ANALYSIS
Analysis of Variance Table
Source
of Variation
Degrees of Sun of
Squares
Freed3m
Mean
Square
F
Ratio
Groups
1
6.849126E-05
6.849126E-05
Within Groups
9
6.596682E-05
7.329646E-06
Adj Total
10
1.344581E-04
Among
9.34
'he columns used in this analysis are
Column
Column
1 label :6 dof rms error
2 label :2 dof rms error
1 S7
Probability
0.013
A(CELZW102V'
steep turns -rollAerror analysis
1
Column:
1
2
-m3
4
5
6
.03836
.02312
.01495
.00826
.02093
.01833
.01493
.02387 .01360
.01611 .00641
.05423 .01581
.01438 .00962
.02413
4
3
2
.00034
.00022
.00057 .00018
.00026 .00004
.00294 .00025
.00021 .00009
.00058
6
5
.00147
.00053
.00022
.00007
.00044
SUBJECT
1)6
2)2
3)0
4)6
5)2
6)0
Col(
Col(
Col(
Col(
Col
Col(
MEAN
dof
dof
dof
dof
dof
dof
rms error
rms error
rms error
mean square error
mean square error
mean square error
STANDARD DEVIATION OF THE MEAN
COLUMN 1
-
3
COLUM 2j
jCOLUY.N 3
l
0L
9
.ooS6O Z.
.u..
158
ROW
ROW
ROW
ROW
ROW
ROW
1:
2:
3:
4:
5:
6:
2
5
10
11
14
17
STEEP TURNS ROLL ACCELERATION ERROR ANALYSIS CONTINUED
Analysis of Variance Table
Source
of Variation
Degrees of Sum of
Freedom
Squares
Mean
Square
F
Ratio
Probability
----------------------------------------------------------Among
Groups
2
5.041706E-04
2.520853E-04
Within Groups
13
1.636186E-03
1.258605E-04
Adj Total
15
2.140357E-03
2.00
0.174
----------------------------------------------------------The columns used in this analysis are
Column
Column
Column
1 label :6 dof rms error
2 label :2 dof rms error
3 label :0 dof rms error
Analysis of Variance Table
Source
of Variation
Amng
Groups
Degrees of Sum of
Freed
Squares
1
Mean
Square
F
Ratio
Probability
4.917089E-04
4.917089E-04
3.91
Within Groups
9
1.131484E-03
1.257205E-04
Adj Total
10
1.623193E-03
The coluwn
Column
Column
used in this analysis are
1 label :6 dof
2 label :2 dof =serr=
159
0.077
~~~0~~~~
steep turn yaw acceleration analysis
1
Column:
1
2
$I
.
5
6
7
8
.00384 .00202
.00420
.00327 .00263
.00254 .00055
.00625 .00222
.00354 .00208
Col(
Col(
Col
Col(
Col(
Col(
4
3
2
.00437
.00445
.00264
.00051
.00286
6
5
.00001 .00000
.00002
.00001 .00001
.00001 .00000
.00004 .00000
.00001 .00000
1)6 dof rms error
2)2 dof rms error
3)0 dof rms error
4)6 dof mean square error
5)2 dof mean square error
6)0 dof mean square error
.00002
.00002
.00001
.00000
.00001
ROW
ROW
ROW
ROW
ROW
ROW
COLUMN 2
6:
SUBJECT
2
5
10
11
14
17
STANDARD DEVIATION OF THE MEAN
MEAN
COLUN 1
.:
2:
3:
4:
5:
,>
3,006q
.O62
COLUHN 3
GO 219
07OO
-
Analysis of Variance Table
Source
of Variation
Degrees of Sum of
Freedom
Squares
Mean
Square
F
Ratio
Among Groups
2
.C3.515055E-06
7.03011E-06
Within Groups
13
2.513733E-05
1.933641E-06
Adj Total
15
3.216745E-05
1.82
The columns used in this analysis are
Column
Column
Column
1 label :6 dof rms error
2 label :2 dof rms error
3 label :0 dof rms error
160
Probability
0.201
STEEP TURNS YAW ACCELERATION
CONTINUED
ERROR ANALYSIS
Analysis of Variance Table
Source
of Variation
Degrees of Sum of
Freanm
Squares
Mean
Square
F
Ratio
Groups
1
1.271001E-06
1.271001E-06
Within Groups
9
1.723585E-05
1.915095E-06
Adj Total
10
1.850685E-05
Among
0.66
Probability
0.509
The columns used in this analysis are
Column
Column
2 label :2 dof rms error
3 label :0 dof rxs error
Analysis of Variance Table
Source
of Variation
Dgrees of Sum of
Freedom
Sqares
Mean
Square
F
Ratio
Groups
1
7.022605E-06
7.022605E-06
Within Groups
9
1.481441E-05
1. 646046E-06
Adi Total
10
2.183702E-05
Amag
4.27
'Ih columuu used in this analysis are
column
Colum
1 label :6 dof ras error
2 label :2 dof rms error
161
Probability
0.067
STALL ROLL ACCELZRATION ERROR
1
Column:
1
2
3
4
5
6
7
8
9
10
11
12
.00686
.00170
.00203
.00228
.00456
.00259
.04626 .00205
.00235 .00210
.00115 .00448
.00116 .00262
.00093 .00081
.00199 .00291
.00766 .00298
.00368 .00297
2
3
.00646
.00337
.00125
.00071
.03897
.03723
.00121
.00198
.00145
.00312
4
5
6
.00005 .00000 .00004
.00000 .00001 .00001
.00002 .00000
.00001 .00000
.00214 .00000 .00152
.00001 .00000 .00139
.00000 .00002 .00000
.00000 .00001 .00000
.00000 .00000 .00000
.00000 .00001 .00001
.00006 .00001
.00001 .00001
SUBJECT
Coi(
Col(
Col(
Col(
Col (
Col(
1)6 DOF RMS ERRORS
2)2 DOF RMS ERRORS
3)SPECIAL EFFECTS RMS ERRORS
4)6 DOF MEAN SQUARES
5)2 DOF MEAN SQUARE ERRORS
6)SPECIAL EFFECTS ONLY ERRORS
MEAN
COLUMN i
COLUMN 3
,31
ooL43
.0003
005eIR
1:
2:
3:
4:
5:
6:
7:
8:
9:
10:
11:
ROW
12:
STANDARD DEVIATION OF THE MEAN
.O()9
COLUMN 2
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
2
2
5
5
10
10
11
11
14
14
17
17
CONT'D
ERROR ANALYSIS
STALL ROLL ACCELERATION
Analysis of Variance Table
Source
of Variation
Among
Mean
Degrees of Sum of
Squares
Freedom
Groups
F
Ratio
Square
Probability
2.744528E-04
1.372264E-04
2
1.05
Within Groups
29
3.803538E-03
1.311565E-04
Adj Total
31
4.077991E-03
0.364
The columns used in this analysis are
Column
Column
Column
1 label
2 label
3 label
:6 DOF RMS ERRORS
:2 DOF RMS ERRORS
:SPECIAL EFFECTS RMS ERRORS
Analysis of Variance Table
Source
of Variation
Amng
Groups
Degrees of Sum of
Freeda
Squares
Mean
Square
F
Ratio
1
2.579375E-04
2.579375E-04
Within Groups
20
2.072126E-03
1.036063E-04
Mj Total
21
2.330063E-03
2.49
The columns used in this analysis are
Column
2 label :2 DOF M9ERRORS
Column
3 label :SPECAL EFFECIS 1
163
ERCM
Probability
0.130
STALL PITCH ACCLERATION ERROR ANALYSIS
1
Column:
3
4
.00668
.01211
.01281
.00831
.00566
.00949
.00852
4
5
Col(
Col(
Col(
Col(
1
COLUMN 2
COLUMN 3
6
.00008
.00004
.00003
.00005
.00023
.00009
.00952 .00004 .00008
.01126 .00015 .00009 .00013
.01210 .00016 .00003 .00015
.00338 .00007 .00012 .00001
.00399 .00003 .00001 .00002
.00009 .00016
.00007 .00035
.00898
.00941
.00561
.01108
.00316
.01270
.01878
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
1)6 DOF RMS ERROR
2)2 DOF RMS ERROR
3)SPECIAL EFFECTS RMS ERRORS
4)6 DOF MEAN SQUARE ERRORS
5)2 DOF MEAN SQUARE ERRORS
6)SPECIAL EFFECTS MEAN SQUARE ERRORS
Col(
Col(
COLU
3
.00579 .00549 .00915 .00003 .00003
.00623 .00505 .00617 .00004 .00003
.00004
.00664 .00554
.00005
.00700 .00707
.01589 .01386 .01508 .00025 .00019
1
2
5
6
7
8
9
10
11
12
2
MEAN
.c
fq15 -
STANDARD DEVIATION OF THE MEAN
,
8
:oI
I
,
q
.oO193
Analysis of Variance Table
Source
of Variation
Among
Degrees of Sum of
Freedom
Squares
Mean
F
Square
Ratio
Groups
2
3.827692E-06
1.913846E-06
Within Groups
29
4.552619E-04
1.569869E-05
Adj Total
31
4.590896E-04
The columns used in this analysis are
Column
Column
Column
1 label :6 DOF RMS ERROR
2 label :2 DOF RMS ERROR
3 label :SPECIAL EFFECTS RMS ERRORS
164
Probability
0.12
0.886
1:
2:
3:
4:
5:
6:
7:
8:
9:
10:
11:
12:
SUBJECT
2
2
5
5
10
10
11
11
14
14
17
17
STALL
J0NUDIAL A
Colu i:
1
-. 1
2
3
4
5
6
7
8
9
10
11
12
ATIN ER1
2
3
AAIMsIS
4
5
.55224
.55751
.50618
.54139
.58052
.56418
.56041
.54160
.56961 .58183
.58058 .55336
.48542 .49506
.50130 .51191
.54379 .55471
.55300 .52069
.57019 .52063
.53550 .53108
.53495 .57683
.59213 .57537
.53648
.52032
Col(
Col
Col(
Col(
Col(
Col(
1)6 DOF PHS ERRRM
2)2 DF RM M
3) SPECIAL EFFECTS RHS EM
N
ZM EMXS
4)6 DF MEAN
5) 2 DOF MEMN SM
EM
6) SPECIAL EFFECTS MEMN qM
6
.32446 .33853
.33708
.23564
.25131
.29571
.30581
.32512
.28676
.28618
.35062
.28781
.27074
.30497
.31082
.25622
.29310
.33701
.31830
.31405
.29333
EAN
.30621
.24508
.26205
.30770
.27112
.27105
.28205
.33274
.33105
SUBJECT
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ERR
STANDARD DEVIATION
COLUMN 1
COLUMN 2
555.0-7
.
454
COLUMN 3
.
1:
2:
3:
4:
5:
6:
7:
8:
9:
10:
11:
12:
OF THE MEAN
y'
11
.(
3
.15.o
. Analysis of Variance Table
Degra of S=r of
So
Squares
of Variatimi Freedc=
Mean
Sqare
F
Ratio
2
3.468219E-04
1.734109E-04
Within &*a
27
2.280975E-02
8. 448054E-04
Adj Total
29
2.315657E-02
Amxng
Groups
0.21
The clumn
ColuMn
Colum
Colum
used in this analysis are
1 label :6 DOF RI6 ERRMRS
2 label :2 DOF M4 EM
3 label :SPECIAL EFFE.IS MERRC
165
Probability
0.816
2
2
5
5
10
10
11
11
14
14
17
17
yaw damper failure pitch acceleration error analysis
Column:
1
2
3
4
5
6
1
2
.00295 .00290
.00624
n.02193 .02652
.01404 .01370
.04138 .03117
.01400 .03067
3
4
.00304 .00001
.00692
.00774 .00048
.00906 .00020
.02409 .00171
.00020
5
.00001
.00004
.00070
.00019
.00097
.00094
6
.00001
.00005
.00006
.00008
.00058
SUBJECT
Col
Col(
Col(
Col(
Col(
Col (
1)6 dof rms error
2)2 dof rms error
3)special effects rms error
4)6 dof mean square error
5)2 dof mean square error
6) special effects mean square error
MEAN
COLUMN
STANDARD DEVIATION OF THE MEAN
CAS(.
000039i
COLUM 2
COLU N3
-o
166
ROW
ROW
ROW
ROW
ROW
ROW
1:
2:
3:
4:
5:
6:
2
5
10
11
14
17
r
CONTINUED
RATE ~~JRS W/YAW DAY2ER FAILURE PITCH ACCELERATION ERROR ANALYSIS
Analysis of Variance Table
Source
of Variatian
Degrees of S=. of
Squares
Fred
Mean
Square
F
Ratio
Amrng GrCUps
2
2.493415E-04
1.246708E-04
Within qcups
13
1.868517E-03
1.437321E-04
Adj Total
15
2.117859E-03
Probability
0.87 0.443
The cuIs
Col=
Colu
Col
used in th.is analysis are
1 label :6 dof
s ero
2 label :2 dof rs err=r
3 label :special effects r s error
Analysis of Variance Table
Source
of Variation
Among
Groups
Degrees of Sum of
Freedom
Squares
1
Mean
Square
F
Ratio
Probability
1.9076E-04
1.9076E-04
1.63
Within Groups
9
1.051962E-03
1.168847E-04
Adj Total
10
1.242722E-03
The columns used in this analysis are
Column
Column
2 label :2 dof rms error
3 label :special effects rms error
167
0.232
yaw danper failure roll acceleration error
Column:
1
2
.01125 .00579
.06250
14138 .20664
.13432 .17632
.31489 .24784
.13861 .22424
1
2
3
4
5
6
3
.00393
.08255
.10413
.08723
.22288
4
5
6
.00013 .00003 .00002
.00391 .00681
.01999 .04270 .01084
.01804 .03109 .00761
.09916 .06142 .04968
.01921 .05028
SUBJECT
Col
Col(
Col(
Col(
Col(
Col
1)6 dof rms error
2) 2 dof =s error
3)special effects rms error
4)6 dof mean square error
5)2 dof mean square error
6) special effects mean square error
MEAN
ROW
RCW
ROW
ROW
RCW
ROW
1:
2:
3:
4:
5:
6:
STANDARD DEVIATION OF THE MEAN
,4496
L $ 14
COLUMN i
1
COLUMN 2
'3
COLUMN 3
.0014
.-o12,
.- 4
Analysis of Variance Table
Source
of Variation
Degrees of Sum of
Freedom
Squares
Mean
Square
F
Ratio
2
9.070793E-03
4.535397E-03
Within GrC.zps
13
.1190325
Adj Total
15
.1281033
Amnq
Groups
0.50
9.156349E-03
The columns used in this analysis are
Column
Column
Column
1 label :6 dof rms error
2 label :2 dof =ns error
3 label :special effects rms error
Probability
0.620
2
5
10
11
14
17
yaw damper failure
Column:
1
1 .00291
2
3 -. 03324
4 '.03448
5 .08992
6 .03035
7
Col(
Col(
Col(
col(
Col
Col
2
.00234
.01319
.04797
.04204
.06545
.05272
4
-7 2
.aa
6
DEVIATION OF THE MEAN
,
.
5
.00148 .00001 .00001 .00000
.00017 .00040
.01996
.02212 .00110 .00230 .00049
.02323 .00119 .00177 .00054
.04944 .00809 .00428 .00244
.00092 .00278
STANDARD
COLUMN 1
COLU.MN 3
3
1)6 dof rns error
2)2 dof rzs error
3)special effects rms error
4)6 dof mean square error
5)2 dof mean square error
6)special effects mean square error
MEAN
COLUMN 2
yaw accleration error analysis
oi'H32
'T -5
,o
-7(
169
SUBJECT
ROW
1:
ROW
ROW
ROW
2:
3:
4:
2
5
10
11
ROW
5:
14
ROW
6:
17
RATE TURNS W/
YAW DA PER FAILURE YAW ACCELERATION ERROR ANALYSIS
CONTINUED
Aralysis of Variance Table
Source
of Variation
Among
Within
Groups
oups
Adj Total
Degrees of Sum of
Freaedm
Squares
Mean
Square
F
patio
2
7.195278E-04
3.597639E-04
13
8.162361E-03
6.278739E-04
15
8.881889E-03
0.57
Probability
0.577
The coluts used in this analysis are
Column
Column.
Column
1 label :6 dof =s error
2 label :2 dof r= error
3 label :special effects rms error
Analysis of Variance Table
Source
of Variation
Among
Groups
Degrees of Sum of
Freedom
Squares
1
Mean
Square
F
Ratio
Probability
2.184614E-06
2.184614E-06
0.00
Within Groups
9
6.990411E-03
7.767123E-04
Adj Total
10
6.992595E-03
The columns used in this analysis are
Column
Column
1 label :6 dof rms error
2 label :2 dof rms error
170
0.986
AC4EL eFxn orJ
ILS LANDING 500'-200' PITCHAERROR ANALYSIS
Column:
1
1
.02960
.03818
2
3
4
5
.02815
.03643
6
.03399
7
8
9
.03529
10
11
12
.03773
.03240
.03470
.03625
.03265
.03877
13
Col(
Col(
Col(
Col(
Col(
Col(
2
.03489
4
.03237
.00088 .00122
.00146
.03521
.00142 .00124
.03307
.00079 .00109
.04028
.00133 .00162
.03622 .00116
.03272 .03929 .00105 .00107
.03205 .03974 .00125 .00103
.04030
.00120 .00162
.03773 .03322 .00131 .00142
.04195 .03891 .00107 .00176
.03644 .04497 .00150 .00133
1)6 DOF
2)2 DOF
3)0 DOF
4)6 DO?
5)2 DOF
6)0 DO?
6
5
.03303
.04245
.03448
.03188
.00105
.00109
.00180
.00119
.00102
.00131
.00154
.00158
.00110
.00151
.00202
SUBJECT
RMS ERROR
RMS ERROR
ROW
ROW
ROW
ROW
RMS ERROR
MEANS SQUARE ERROR
MEAN SQUARE ERROR
MEAN SQUARE ERROR
1:
2:
3:
4:
ROW 5:
ROW 6:
ROW
7:
ROW
8:
ROW 9:
ROW 10:
ROW 11:
ROW 12:
STANDARD DEVIATION OF THE %EAN
MEAN
COLUMN I
3
o
COLUMN 2
0
1
1
6003
COLUMN 3
j
Analysis of Variance Table
Source
of Variation
Among
Degrees of Sum of
Freedom
Squares
Mean
Square
F
Ratio
Groups
2
3.865778E-05
1.932889E-05
Within Groups
30
4.263946E-04
1.421315E-05
Adj Total
32
4.650524E-04
1.36
The columns used in this analysis are
Column
Column
Column
1 label :6 DO? RMS ERROR
2 label :2 DOF RMS ERROR
3 label : 0 DO? RMS ERROR
171
Probability
0.272
4
4
6
6
7
7
12
12
16
16
18
18
ACc6L9ERATlOrJ
ILS LANDING 500-200' ROLLAERROR
Column:
1
.03424
. 05044
.03691
.04123
.07200
.04502
.04886
.06199
.05411
.05551
.04452
.06417
2
3
.04159
.03660
.03251
.05261
.04500
.05518
.04871
.05854
.06917
4
5
6
.00104 .00173 .00134
.00254
.00106
.04531
.00136 .00205 .00277
.04424
.00170 .00196 .00203
.04070
.00518 .00166 .00304
.00203
.00237
.08034
.00239 .00645 .00343
.04483
.00384 .00201 .00478
.04625
.00293 .00214
.04623 .04021 .00308 .00214 .00162
.07233 .07188 .00198 .00523 .00517
.06164 .08687 .00412 .00380 .00755
SUBJECT
1)6 DOF RMS ERROR
2)2 DOF RMS ERROR
3)0 DOF RMS ERROR
4)6 DOF MEAN SQUARE ERROR
Col(
Co(
Co(
Co(
Co(
5)2
Co(
DOF
ROW
1:
ROW
2:
ROW
3:
ROW
4:
ROW 5:
ROW 6:
ROW
7:
ROW
8:
ROW
9:
ROW 10:
ROW 11:
ROW 12:
MEAN SQUARE ERROR
6)0 DOF MEAN SQUARE ERROR
MEAN
-
4
4
6
6
7
7
12
12
16
16
18
18
STAflARD DEVIATION OF THE MEAN
,05---.0033(----
COL M N 2
. 5 2
COLUMN 300(4q
5,
o 0 4J
Analysis of Variance
Source
of Variation
Among
Groups
Degrees
Freedom
of Sum of
Squares
Table
Mean
Square
F
Ratio
2
7.920194E-05
3.960097E-05
30
5.97566E-03
0.20
WithIn Groups
1.991887E-04
Adj Total
6.054862E-03
32
The columns used in this analysis are
Column
Column
Column
I label :6 DOT RMS ERROR
2 label :2 DOT RMS ERROR
3 label :0 DOT RMS ERROR
17.
Probability
0.821
tJ
A CCSLEZA10
YAWAERROR
500-200'
ILS LANDING
Column:
1
6
5
4
3
2
.00497 .00564 .00002 .00002 .00003
.00619 .00005
.00004
.00498 .00682 .00785 .00002 .00005 .00006
.00810 .00679 .00664 .00007 .00005 .00004
.00680 .00580 .00850 .00005 .00003 .00007
.00664 .00011
.00004
.01029
.00477
.00477
.01135
.00803
.01165
.01427
.00013
.00551
.00814 .01588 .00006
.00738
.00003
.00924
.01043
.01117
.00666
.01223
.00937
.00914
.01701
.01496
.00009
.00011
.00012
.00020
.00007
.00005
.00004
.00015
.00009
.00014
.00025
.00008
.00029
.00022
SUBJECT
Col(
Col(
Col(
Col(
Col(
Col(
COLUMN 1
C.OLMN
2
1)6
2)2
3)0
4)6
DO?
DOIP
DOF
DO?
5)2 DOF
6)0 DO?
MEAN
,oosiz
RMS ERROR
RMS ERROR
RMS ERROR
MEAN SQUARE ERROR
MEAN SQUARE ERROR
MEAN SQUARE ERROR
STA,%DARD DEVIATION
OF
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
1:
2:
3:
4:
5:
6:
7:
8:
9:
10:
11:
12:
4
4
6
6
7
7
12
12
16
16
18
18
THE xEAN
0,o cg
,0 092q
cools
COLUMN 3
(, o IQ4
Analysis of Variance Table
Source
of Variation
Among
Groups
Degrees of Sum of
Squares
Freedom
2
Mean
Square
F
Ratio
Probability
2.47266E-05
1.23633E-05
1.18
Within Groups
30
3.148263E-04
1.049421E-05
Adj Total
32
3.395529E-04
The columns used in this analysis are
Column
Column
Column
1 label :6 DO? RMS ERROR
2 label :2 DOF RMS ERROR
3 label :0 DOF RMS ERROR
173
0.322
I.....
VICAL AXIS AaM =CN ERRR
IIS 500-200'
1
Coiumn:
6
5
4
3
2
1 2.8925 2.8962 2.8721 8.3663 8.3879 8.2490
2.9084 8.2993
8.4586
2 2.8808
3
4
5
6
7
8
9
10
11
12
2.9060 2.8923 2.9402
2.8738 2.9072 2.8959
2.8425 2.8978 2.9272
2.8938
2.9245
2.9261 2.9044 2.9124
2.9150 2.8871 2.8804
2.8796 2.8902
2.8948 2.8972 2.8979
2.8638 2.8878 2.9139
2.9305 2.9144 2.8920
col(
Col(
Col(
col(
col(
col
8.4448 8.3654 8.6446
8.2588 8.4515 8.3861
8.0797 8.3974 8.5687
8.3739
8.5525
8.5621 8.4354 8.4820
8.4971 8.3355 8.2965
8.2923 8.3531
8.3797 8.3940 8.3976
8.2014 8.3395 8.491
8.5876 8.4939 8.3635
1)6 Do ImS mum0
2)2 OF RM EIRKt
3)0 Mir 14m ERIUM
4)6 DOF IMEAN S=M
5)2 MF IMEAN sq =
I6)0 DOF IMEAN sq=
COLUMN 1
A0
5:
12
12
16
10:
16
18
18
11:
12:
DKg"
F*
of SLm of
suarm
Mean
Square
F
Ratio
2
4.654903E-04
2.327451E-04
Within troups
30
1.289757E-02
Adj Ttal
32
Gr4S
7
7
6:
7:
8:
9:
Tble
Analysis of Varianc
A~r
6
6
.o2 I
- 004
..
q5
SCUZ=
of Varistim
2:
3:
4:
SUBJECT
4
4
STANDARD DEVIATION OF THE MAN
ObI,0C)qL-
MEAI
C0LUM2
COLUMN 3 1
1:
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
0.54
4.299191E-04
1.336306E-02
The clumns usd in this analysis are
1 labal :6
2 label :2
3 label :0
ms
Prcbability
0.588
ILS LANDING LAST 20-25 SEC PITCHAERROR ANALYSIS
Column:
1
2
c3
m4
5
6
7
8
9
10
11
12
1
.01985
.03303
.06298
.03053
.03325
.02954
.03059
.02785
.03100
.04927
.06943
.05045
Col(
Col(
Col(
Col (
Col(
Col(
3
2
.01934
.03658
.03782
.02975
.06536
.05849
.03196
.03160
.03675
1)6
2)2
3)0
4)6
5)2
6)0
DOF
DOF
DOF
DOF
DOF
DOF
.01555
.04435
.03889
.01694
.03437
.03629
.02489
.03845
.00109
.00397
.00093
.00111
.00087
.00094
.00078
.00096
.02726 .00243
.02640 .00482
.00255
4
5
6
.00039 .00024
.00197
.00037 .00151
.00134 .00029
.00143 .00118
.00132
.00089 .00062
.00427 .00148
.00342
.00102 .00074
.00100 .00070
.00135
RMS ERROR
RMS ERROR
RMS ERROR
MEAN SQUARE ERROR
MEAN SQUARE ERROR
MEAN SQUARE ERROR
S UBJECT
ROW
ROW
1:
2:
ROW
3:
ROW
ROW
ROW
ROW
ROW
ROW
4:
5:
6:
7:
8:
9:
ROW
10:
ROW
11:
ROW
12:
STANDARD DEVIATION OF THE MEAN
MEAN
COLUMN 1
.0Yo072.
I
COLUMN 2
.
I
COLUMN 3
(4
I
-35
175
4
4
6
6
7
7
12
12
16
16
18
18
ILS LANDING LAST
20-25 sec
pitch acceleration error analysis
continued
I
Analysis of Variance Table
Source
of Variation
Degrees of Sum of
Freedom
Squares
Mean
Square
F
Ratio
2
5.709019E-04
2.85451E-04
WithIn Groups
28
5.013378E-03
1.790492E-04
Adj Total
30
5.58428E-03
Among
Groups
1.59
Probability
0.221
The columns used in this analysis are
Column
Column
Column
1 label :6 DOF RMS ERROR
2 label :2 DOF RMS ERROR
3 label :0 DOF RMS ERROR
Analysis of Variance Table
Source
of Variation
Amang
Groups
Degrees of Sum of
Fred
Squares
1
Mean
Square
F
Ratio
Probability
5.644842E-04
5.644842E-04
3.54
Within Groups
19
3.033859E-03
1.596768E-04
Adj Total
20
3.598344E-03
The coluzmn
Column
Column
used in this analysis are
1 label :6 DOF RS ERRCR
3 label :0 DOF IM ERRCR
176
0.072
ILS LANDING LAST 20-25 SEC ROLL ACCELERATION ERROR ANALYSIS
1
Column:
1
2
2
3
4
.04953
.13852
.24293 .05156
.13817 .09598 .05183
.13128 .16359 .10053 .01723
.14380 .16477 .08359 .02068
.07539
.10392 .00568
.17502 .19781 .13317 .03063
.10942 .22862 .09859 .01197
.06316 .15837
.00399
.15578 .06279 .16026 .02427
.25941 .27782 .10522 .06729
.19099 .13477
.03648
.22707
.22766
3
4
5
6
7
8
9
10
11
12
Col(
Col(
Col(
Col(
Col(
Col(
1)6
2)2
3)0
4)6
5)2
6)0
5
6
.00245
.01919
.05901
.01909 .00921
.02676 .01011
.02715 .00699
.01080
.03913 .01773
.05227 .00972
.02508
.00394 .02568
.07718 .01107
.01816
DOF RMS ERROR
DOF RMS ERROR
DOF RMS ERROR
DOF MEAN SQUARE ERROR
DOF MEAN SQUARE ERROR
DOF MEAN SQUARE ERROR
SUBJECT
ROW
ROW
1:
4
2:
ROW
3:
4
6
ROW
4:
ROW
5:
ROW
6:
ROW
7:
ROW
8:
ROW
9:
ROW 10:
ROW 11:
.0W 12:
6
7
7
12
12
16
16
18
18
STANDARD DEVIATION OF THE MEAN
MEAN
COLUMN 2
(c
COLUMN 3
&Al
,7
179y
93
00
Analysis of Variance Table
source
of Variation
Degrees of Sum of
Freedom
Squares
Mean
Square
F
Groups
2
7.204142E-03
3.602071E-03
Within Groups
28
.1031859
Adj Total
30
.1103901
Among
Probability
Ratio
0.98
The columns used in this analysis are
Column
Column
Column
1 label :6 DOF RMS ERROR
2 label :2 DOF RMS ERROR
3 label :0 DO? RMS ERROR
177
3.685212E-03
0.389
I
ACCE LE ZA1 10 &J
ILS LANDING LAST 20-25 SEC YAWAERROR ANALYSIS
1
Column:
.00962
1
2
(3
4
5
6
7
8
9
10
11
12
.01279
.01958
.02021
.00722
.00872
.03703
.02058
.01021
.01140
.02217
.01908
Col(
Col(
Col(
Col(
Col(
Col(
MEAN
COLUMN 1
COLUMN 2
3
2
.03006
.02841
.01222
.04777
.04312
.02142
.01118
.03106
.01855
1)6
2)2
3)0
4)6
5)2
6)0
DOF
DOF
DOF
DOF
DOF
DOF
.01557
.01986
.01505
.01784
.00866
.00504
.04398
.02715
.00016
.00038
.00041
.00005
.00008
.00137
.00042
.00010
.02208 .00013
.01676 .00049
.00036
4
5
.00009
.00090
.00081
.00015
RMS ERROR
RMS ERROR
RMS ERROR
MEAN SQUARE ERROR
MEAN SQUARE ERROR
MEAN SQUARE ERROR
3
COLUMN 3
178
.00024
.00039
.00023
.00032
.00007
.00003
.00193
.00074
.00228
.00186
.00046
.00012 .00049
.00096 .00028
.00034
STANDARD DEVIATION OF THE MEAN
01
6
ROW
ROW
ROW
ROW
ROW
I.
2:
3:
4:
5:
ROW
ROW
6:
7:
ROW
8:
ROW
9:
ROW
10:
ROW
11:
ROW
12:
SUBJECT
4
4
6
6
7
7
12
12
16
16
18
18
continued
yaw acceleration error analysis
20-25 sec
ILS LANDING LAST
Analysis of Variance Table
Source
of Variation
Among
Degrees of Sum of
Squares
Freedom
2
Groups
Mean
Square
F
Ratio
Probability
3.724191E-04
1. 862095E-04
1.57
Within Groups
28
3.321733E-03
1.186333E-04
Adj Total
30
3.694152E-03
0.226
The columns used in this analysis are
Column
Column
Column
1 label :6 DOF RMS ERROR
2 label :2 DOF RMS ERROR
3 label :0 DOF RMS ERROR
Analysis of Variance Table
sourc
of Variation
Degrees of Sum of
Freedcm
Squares
Mean
Square
F
Ratio
Amrng Groups
1
3.487895E-04
3.487895E-04
Within Groups
19
2.90
2.285952E-03
1.203133E-04
Adj Total
20
2.634741E-03
The columns used in this analysis are
Column
Column
1 label :6 DOF RMERR
2 label :2 DOF RM ERR
179
Probability
0.102
IS
IANDG LAST 20-25 SEC VERICAL AXIS A
1
Colt=:
6
7
8
9
10
11
12
3
8.5928 8.5691
2.9468 8.8372
2.9727
8.6837
8.6772
2.9484 2.9233 2.9457 8.6930 8.5455
2.9521 2.9607 2.9333 8.7150 8.7660
2.9270 2.9508 2.9567 8.5671 8.7070
2.9211 8.5507
2.9242
2.9629 2.9285 2.9268 8.7786 8.5760
2.9082 2.9847 2.9802 8.4577 8.9086
8.7295 8.5187
2.9546 2.9187
2.9489 2.9113 2.9398 8.6962 8.4755
2.9944 2.9641 2.9349 8.9667 8.7860
2.9266 2.9135
8.5652 8.4886
col(
Col(
Col(
Col(
col(
1)6 DOF ImS
,
2)2 DOF ImS
3)0 DOF smS
4)6
M
5)2 DMF MEA
r
I
6)0
mE
COUL"
8.6424
8.6138
SUBJECT
ERRmR
EMMR
SM
sq
sqAmE
STANDARD DEVIATION OF THE MEAN
6
,
93-7
a, 941[3
COLUMN 3
8.8813
ROW
1:
ROW
2:
3:
ROW
4:
ROW
ROW
ROW 6:
7:
ROW
8:
ROW
9:
ROW
ROW 10:
ROW 11:
ROW 12:
'Z.194-11
2
8.6045
8.7421
8.5331
8.5659
ERRmR
MEAN
CTLUMN
6
5
4
2.9314 2.9273
1
2
3
4
5
2
RAION ERMRANALYSIS
49
0-7 8 -
,
0 0,5 9
kArlysis of Varianm Table
sourceZDegres of S=U of
of Variati
Fread
Squam
Anzgq G=Vs
2
Within &=up
28
mean
square
F
Ratio
4.112165E-04
2.056082E-04
1.439289E-02
5.140317E-04
Adj Tctal
The colum
column
Column
column
30
Prcbability
.0148041
used in this analysis are
1 label :6
2 label :2 DOT ME
3 label :0 Dor R EmmR
18)
0.40
0.674
4
4
6
6
7
7
12
12
16
16
18
18
SUMMARY OF CONTENTS OF APPENDIX D:
This appendix contains the raw data used for the vestibular
All data is in threshold units.
error analyses.
A value of "1"
indicates that the theoretically the subject can just percieve
that motion is
detatil.
occuring.
Chapter 2 discusses these units in more
The means and standard deviation of the means for each motion
condition, over all pilots performing the same manuever are presented
in tables.
The analysis of variance tables are presented with the
data that is associated with them.
flight scenario.
The data is presented by
Engine flameout results are presented first.
The
vestibular error results from the airwork scenario follow, with
steep turns presented first
followed by stall
the results from the rate turn with
presented.
results,
and then
a yaw damper failure are
The results from the approach segment of the ILS
appraoch and landing scenario precedes the results from the landing
segment of this scenario.
18]
LIST OF CONTENTS OF APPENDIX D:
Engine flameout yaw vestibular error data and analysis
Steep turn pitch vestibular error data and analysis
Steep turn roll vestibular error data and analysis
Steep turn yaw vestibular error data and analysis
Stall roll vestibular error data and analysis
Stall pitch vestibular error data and analysis
Stall longitudinal vestibular error data and analysis
Rate Turn with Yaw Damper Failure pitch vestibular error data and
analysis
Rate Turn with Yaw Damper Failure roll vestibular error data and
analysis
Rate Turn with Yaw Damper Failure yaw vestibular error data and
analysis
ILS Approach and Landing 500'-200' pitch vestibular error data and
analysis
ILS Approach and Landing 500'-200' roll vestibular error data and
analysis
ILS Approach and Landing 500'-200' yaw
analysis
vestibular error data and
ILS Approach and Landing 500'-200' vertical vestibular error data
and analysis
ILS Approach and Landing last 20-25 sec pitch vestibular error
data and analysis
ILS Approach and Landing last 20-25 sec roll vestibular error data
and analysis
ILS Approach and Landing last 20-25 sec yaw vestibular error data
and analysis
ILS Approach and Landing last 20-25 sec vertical vestibular error
data and analysis
182
ENGINE FLAME-OUT VESTIBULAR YAW ANALYSIS
1
2
3
5
6
7
8
9
10
11
12
3
2
1
Column:
0.57949
0.89982 1.05760
0.84368 0.82787 1.07350
1.36440
1.08920 1.48050 1.48930
1.29710 1.17160 1.21660
0.80807 0.58842 0.84973
0.93228 0.82139 1.04220
1.02230 0.99446
0.96506 1.39720 1.05880
1.22650 1.36840
0.88871 0.88646 0.80002
Col (
Col (
Col(
Col(
Col(
Col(
1)6
2)2
3)0
4)6
5)2
6)0
MEAN
COLUMN
DOF
DOF
DOF
DOF
DOF
DOF
4
0.33581
0.80968 1.11852
0.71180 0.68537
1.18636
1.68247
0.65298
0.86915
1.04510
0.93134
1.50430
0.78981
1.15240
1.86159
2.21801
1.48012
0.72204
1.08618
SUBJECT
ROW
I:
ROW
2:
ROW
3:
ROW
4:
ROW
5:
ROW
6:
ROW
7:
ROW
8:
ROW
9:
ROW 10:
ROW 11:
ROW 12:
STANDARD DEVIATION OF THE MEAN
I2
o
1
6
2.19188
1.37265
0.34624
0.67468
0.98895
1.95217 1.12106
1.87252
0.78581 0.64003
RMS ERROR
RMS ERROR
RMS ERROR
MEAN SQUARE ERROR
MEAN SQUARE ERROR
MEAN SQUARE ERROR
COLUMN 2
COLUMN 3
5
1'0_
183
ZZq
1
3
3
8
8
9
9
13
13
15
15
ENGINE FLAME OUT SCENARIO VESTIBULAR ERROR YAW AXIS ANALYSIS
CONT'D
Analysis of Variance Table
Source
of Variation
Among
Mean
Square
Degrees of Sum of
Freedom
Squares
Groups
Within Groups
26
.116184
1.554139
Adj Total
28
1.670323-
2
F
Ratio
.058092
0.97
Probability
0.392
5.977458E-02
The columns used in this analysis are
Column
Column
Column
1 label :6 DOF RMS ERROR
2 label
3 label
:2 DOF RMS ERROR
:0 DOF RMS ERROR
Analysis of Variance Table
Srce
of Variatin
Am=ng
Groups
Degrees of Sum of
Squares
Free=
1
Mean
Square
.1077512
.1077512
.0463782
Within Groups
17
.7884294
Adj Total
18
.8961806
The columns used in this analysis are
Column
Column
1 label :column I
3 label :Column 3
1w4
F
Ratio
2.32
Probability
0.143
steep turn vestibular pitch error analysis
Column:
1
2
3
4
5
6
.49245 .38053
.31980
.43115 .22085
.29534 .14713
.38876 .17000
.63084 .06295
MEAN
COLUMN 3
.14506
.39215
.63139
.17858
.04149
4
5
.24251 .14480
.10227
.18589 .04877
.08723 .02165
.15113 .02890
.39796 .00396
6
.02104
.15378
.39865
.03189
.00172
1)6 dof rms error
2)2 doff rms error
3)0 doff rms error
4)6 dof mean square error
5)2 doff mean square error
6)0 dof mean square error
Col (
Col
Col
Col(
Col (
Col (
COLUMN 1
COLUMN 2
3
2
1
STANDARD DEVIATION OF THE MEAN
-qq77/
-
. O5 5 (O
-7-7
. eO
.2o
SUBJECT
ROW
I1:
ROW
2:
3:
5
4
11
14
17
ROW
ROW
ROW
ROW
5:
6:
2
10
STEEP TURNS VESTIB3LAR PITCH ERROR ANALYSIS
CONTINUED
Analysis of Variance Table
Mean
Square
Degrees of Sum of
Squares
Freedom
Source
of Variation
F
Ratio
2
.1519925
7.599627E-02
Withln Groups
13
.3520598
2.708152E-02
Adj Total
15
.5040523
Among
Groups
2.81
Probability
0.097
The columns used in this analysis are
Column
Column
Column
1 label :6 dof rms error
2 label :2 dof rms error
3 label :0 dof rms error
Analysis of Variance Table
Source
of Variation
Degrees of Sum of
Freedom
Squares
Among Groups
Within Groups
1
9
.1453176
.1306511
Adj Total
10
.2759686
Mean
Square
The columns used in this analysis are
Column
Column
1 label :6 dof rms error
2 label :2 dof rms error
F
Ratio
10.01
.1453176
1.451679E-02
Probability
0.011
steep turn vestibular roll error analysis
Column:
1
2
c3
4
5
6
7
2
1
1.5665 1.0017
1.0999
1.4720 0.7893
0.7548 0.2432
1.3384 0.6367
0.9889 0.3359
Col(
Col(
Col(
Col(
Col (
Col (
1)6
2)2
3)0
4)6
5)2
6)0
3
0.4642
1.8077
0.8182
0.3627
0.1993
4
6
5
2.4539 1.0034
1.2098
2.1668 0.6230
0.5697 0.0591
1.7913 0.4054
0.9779 0.1128
0.2155
3.2678
0.6695
0.1315
0.0397
SUBJECT
dof
rms error
dof rms error
do f rms error
dof mean square error
dof mean square error
dof mean square error
ROW
1:
ROW
2:
ROW
ROW
3:
2
5
10
4:
11
ROW
5:
6:
14
ROW
MEAN
jCOLUMN
.&
l
STANDARD DEVIATION OF THE MEAN
.1l5 2S
COLUMN I
COLUMN
3
187
1
17
STEEP TURNS VESTIBULAR ROLL ERROR ANALYSIS CONTINUED
Analysis of Variance Table
Source
of Variation
Degrees of Sum of
Squares
Freedom
Among Groups
Within Groups
Adj Total
2
13
.9308698
2.726478
15
3.657348
Mean
Square
F
Ratio
.4654349
.2097291
--------------------------
2.22
------
1 label :6 dof rms error
2 label :2 dof rms error
3 label :0 dof rms error
of Variation
Analysis of Variance Table
Degrees of Sum of
Mean
Freedcn
Squares
Square
Amng Groups
Within Groups
1
9
.7942883
1.070055
Adj Total
10
1.864343
The columns used in this analysis ar
COlumn
C0lUMn
1 label :6 dof rm error
2 label :2 dof rz error
188
.7942883
.118895
0.148
-----------------
The columns used in this analysis are
Column
Column
Column
Probability
tio
6.68
0.028
steep turn Vestibular yaw error analysis
Column:
1
2
23
4
5
6
.24332 .10754
.29683
.19525 .14756
.12155 .02072
.15878 .08873
.24425 .06745
Col (
Col
Col (
Col
Col
Col
dof
dof
dof
dof
dof
dof
.05359
.33986
.15214
.02205
.02722
4
.05920 .01156
.08811
.03812 .02177
.01477 .00043
.02521 .00787
.05966 .00455
rms error
rms error
rms error
mean square error
mean square error
mean square error
,i.92
COLUMN1
COLUMN 3
1)6
2)2
3)0
4)6
5)2
6)0
3
STANDARD DEVIATION OF THE MEAN
MEAN
COLUMN 2
2
1
,
IL.1
. O'9O0(0
11S9
1
6
5
.00287
.11550
.02315
.00049
.00074
SUBJECT
ROW
ROW
ROW
ROW
ROW
ROW
1:
2:
3:
4:
5:
6:
2
5
10
11
14
17
STEEP TURNS VESTIBULAR YAW ERROR ANALYSIS CONTINUED
Analysis of Variance Table
Among
Groups
F
Ratio
Mean
Square
Degrees of Sum of
Squares
Freedom
Source
of Variation
2
1.798314E-02
8.991569E-03
13
.1291844
15
.1471675
0.90
Within Groups
Adj Total
Probabilit
0.429
9.93726E-03
--------------------------------------------The columns used in this analysis are
Column
Column
Column
1 label :6 dof rms error
2 label :2 dof rms error
3 label :0 dof rms error
Analysis of Variance Table
Mean
Square
Degrees of SUM of
Sqares
Freed:U
Vriatio
Sof Variaticn
of
AVGrxxupS
1
F
Ratio
1.380957E-02
1.380957E-02 21
2.17
Within Grc*IpS
9
5.720565E-02
6.356183E-03
jTotal10
7.101522E-02
rae oolun
used in this analySis are
Colu=
Column
PrcabilitY
1.label :6 dof rMS error
2 label :2 dof rms error
190
.7
0.172
STALL VESTIBULAR ROLL ERROR ANALYSIS
1
Column:
2
3
5
4
6
1 .01616 .00820 .00189 .00026 .00007 .00000
2 .01136 .00346 .00046 .00013 .00001 .00000
.00225 .00016
.04747 .01283
3
.00020 .00001
.01418 .00307
4 :
5 .35507 .00525 .19174 .12607 .00003 .03676
6 .00660 .00185 .05918 .00004 .00000 .00350
7 .00411 .01992 .00455 .00002 .00040 .00002
8 .00085 .00464 .00389 .00000 .00002 .00002
9 .00363 .00383 .00185 .00001 .00001 .00000
10 .00305 .00693 .00151 .00001 .00005 .00000
.00666 .00018
11 .08158 .01335
.00022 .00003
12 .01472 .00523
SUBJECT
Col(
Col(
Col(
Col(
Col(
Col(
1)6 DOF RMS ERROR
2)2 DOF RMS ERROR
3)SPECIAL EFFECTS ONLY RMS ERRORS
4)6 DOF MEAN SQUARE ERROR
5)2 DOF MEAN SQUARE ERROR
6)SPECIAL EFFECTS ERROR
STANDARD DEVIATION OF THE
MEAN
COL~UMN 2
COLUIMN 3
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
,ao3(as4-,-
3:
4:
5:
6:
7:
8:
9:
10:
11:
12:
2
2
5
5
10
10
14
14
17
17
EAN
I(4
03
,eigg,6
Analysis of Variance Table
Source
of Variation
Among
Groups
Degrees of Sun of
Freedom
Squares
2
Mean
Square
F
Ratio
Probabilit'y
8.096127E-03
4.048063E-03
0.82
Within Groups
29
.1430462
Adj Total
31
.1511423
4.932626E-03
The columns used in this analysis are
Column
Column
Column
I label :6 DOF RMS ERROR
2 label :2 DOF RMS ERROR
3 label :SPECIAL EFFECTS ONLY RMS ERRORS
191
0.450
CONTINUED
STALL VESTIBULAR ROLL ERROR
Analysis of Variance Table
Source
of Variation
Amng
Degrees of Su of
Fredcm
Squares
Groups
Mean
Square
F
Ratio
1
8.093612E-03
8.093612E-03
Within Groups
20
.1104547
Adj Total
21
.1185483
5.522736E-03
Probability
1.47
0.240
The columns used in this analysis are
Column
Column
1 label :6 DOF FMERRCR
2 label :2 DOF RMS E
Analysis of Variance Table
Source
of Variation
Among
Groups
Degrees of Sum of
Freedom
Squares
1
Mean
Square
F
Ratio
Probability
1.558702E-03
1.558702E-03
0.91
Within Groups
20
3.433945E-02
1.716973E-03
Adj Total
21
3.589815E-02
The columns used in this analysis are
Column
Column
2 label :2 DOF RMS ERROR
3 label :SPECIAL EFFECTS ONLY RMS ERRORS
192
0.352
STALL VESTIBULAR PITCH ERROR ANALYSIS
1
Column:
1
2
3
4
5
6
7
8
9
10
11
12
2
3
4
5
6
.01498 .02235 .00268
.04059 .00719 .00086
.06973 .05662
.03724 .03289
.13065 .04193 .07671
.01753 .01040 .01533
.00022 .00050 .00001
.00165 .00005 .00000
.00486 .00321
.00139 .00108
.01707 .00176 .00588
.00031 .00011 .00024
.04456 .03495 .04442 .00199 .00122 .00197
.00962 .00711 .02285 .00009 .00005 .00052
.03483 .05267 .00426 .00121 .00277 .00002
.00898 .00799 .00207 .00008 .00006 .00000
.10552 .05587
.01113 .00312
.03454 .03298
.00119 .00109
SUBJECT
Col(
Col (
Col (
Col(
Col(
Col(
1)6 DOF RMS ERRORS
2)2 DOF RMS ERRORS
3)SPECIAL EFFECTS ONLY RMS ERRORS
4)6 DOF MEAN SQAURE ERRORS
5)2 DOF MEAN SQUARE ERRORS
6)SPECIAL EFFECTS ONLY MEAN SQUARE ERRORS
ROW
1:
ROW
2:
ROW
3:
ROW
4:
ROW
5:
ROW
6:
ROW
7:
ROW
8:
ROW
9:
ROW 10:
ROW 11:
ROW 12:
STANDARD DEVIATION OF THE MEAN
MEAN
COLUMN1
04
COLU4N 2
o
COLUMN 3
,02.S911
1
.
0
51309
0(.(01
7
'C)
3
Analysis of Variance Table
Source
of Variation
Degrees of Sum of
Squares
Freedom
Mean
Square
F
Ratio
2
1.759314E-03
8.796572E-04
Within Groups
29
2.659635E-02
9.171155E-04
Adj Total
31
2.835566E-02
Among
Groups
0.96
The columns used in this analysis are
Column
Column
Column
1 label :6 DOF RMS ERRORS
2 label :2 DOF RMS ERRORS
3 label :SPECIAL EFFECTS ONLY RMS ERRORS
193
Probability
0.395
2
2
5
5
10
10
11
11
14
14
17
17
Stall VestibUlar IangitudinaEError analysis
Column:
3
2
1
1.5963 1.5935 1.6205
1.5243 1.5758 1.5626
1.3525 1.3774
1.3464 1.3706
1.5613 1.5134 1.5306
1.5416 1.6036 1.5526
1.5374 1.5685 1.5124
1.4484 1.5395 1.4917
1.5558 1.6690 1.6499
1.6059 1.5581 1.6257
1.5095 1.5142
1.5499 1.5223
1
2
3
4
5
6
7
8
9
10
11
12
2.5482 2.5393 2.6259
2.3234 2.4830 2.4416
1.8292 1.8972
1.8127 1.8785
2.4376 2.2903 2.3428
2.3765 2.5714 2.4106
2.3634 2.4602 2.2872
2.0980 2.3700 2.2252
2.4205 2.7855 2.7221
2.5788 2.4276 2.6430
2.2786 2.2928
2.4023 2.3173
1) 6 DOF MEAW RM M
2) 2 DOF MEAN MS EM
3) SPECIAL ETECIS
4)6 DOF MN Sq-V=ERM
5)2 DOF MEAMN SqREEtS
6) SPECIAL EMFECIS MEN SqREE
col(
col(
Col(
col(
Col(
Col(
,(00(
1
COLUmN 2
o
-
1:
2:
3:
4:
SUBJECT
2
2
6:
7:
9:
10:
11:
12:
14
14
17
17
o
Analysis of Variarce Table
Degrsm of Sum of
source
Sqarem
of Variatioi Fr.d
Amon
Groups
Mean
Square
F
Ratio
6.232564E-04
0.09
Within Groups
29
.2005916
Adj Total
31
.2018382
The =1=s
Colum
Column
Colt=
Probability
1.246513E-03
2
used in this analysis are
1 label :6 DO? MO EMU
2 label :2 DO? IM ERERS
3 label :SPECAL UEC>
194
6.916953E-03
5
10
10
11
11
2.9*
COLUMN 3j
5
8:
c
9
.
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
STANDAD DEVIATION OF THE MEAN
MEAN
COLU",
6
5
4
0.914
yaw damper failure vestibular pitch error analysis
1
Column:
6
3
4
5
.00001 .00000 .00000
.00347
.00000
.00065 .00002 .00001
.00103 .00000 .00000
.00542 .00709 .00044 .00003 .00005
.00167 .00515
.00000 .00003
1 .00044
2
3 .4.004 7 3
4 2.00034
5
2
.00029
.00215
.00283
.00027
6
.00000
.00001
.00000
.00000
.00000
SUBJECT
Col(
Col
Col
Col(
Col(
Col(
COLUMN 1
COLUMN 2
COLUMN 3
1)6 dof rms error
2)2 dof rms error
3)0 dof rms error
4)mean square error 6 dof
5)mean square error 2 dof
6)mean square error 0 dof
MEAIN
COZ.$Z
ROW
ROW
ROW
ROW
ROW
ROW
STANDARD DEVIATION OF THE MEAN
.001o01
(-(1
o
1
(O4
I
195
1:
2:
3:
4:
5:
6:
2
5
10
11
14
17
RATE TURNS W/YAW DAMPER FAILURE VESTIBULAR PITCH ERROR ANALYSIS
CONTI=UD
Analysis of Variance Table
Source
of-Variation
Among
Degrees of Sum of
Freedom
Squares
F
Ratio
Probability
9.802042E-06
2
Groups
Mean
Square
4.901021E-06
0.94
13
Within Groups
ri5.187103E-06
7.723438E-05
15
Adj Total
0.414
6.743233E-05
The columns used in this analysis are
Column
Column
Column
1 label :6 dof rms error
2 label :2 dof rms error
3 label :0 dof rms error
Analysis of Variance Table
Source
of Variation
Among
Degrees of Sum of
Freedom
Squares
Mean
Square
F
Ratio
Groups
1
5.360303E-07
5.360303E-07
Within Groups
9
5.998633E-05
6.665148E-06
Adj Total
10
6.052237E-05
0.08
Probability
0.719
-----------m-----------------w----- w---------------M-----------
The columns used in this analysis are
Column
Column
1 label :6 dof rms error
2 label :2 dof rms error
196
yaw damper failure vestibular roll error analysis
1
Column:
2
3
4
5
6
1 .00174 .00062 .00002 .00000 .00000 .00000
2
3 2.03309
4 .00325
5 .04121
6 .01623
.02009
.02132
.03447
.05627
.03665
.00040 .00183
.04276
.00850 .00109 .00045 .00007
.00969 .00001 .00119 .00009
.00413 .00170 .00317 .00002
.00026 .00134
SUBJECT
Col(
Col(
Col(
Col(
Col(
Col (
1)6 dof rms error
2)2 dof rms error
3)0 dof rms error
4)6 dof mean square error
5)2 dof mean square error
6)mean square error 0 dof
STANDARD DEVIATION OF THE MEAN
MEAN
COLUMN1
0191O
COLUMN 2
o
,.
COLUMN 3
,OI.?OZ
,
.
-70
197
ROW
ROW
ROW
ROW
ROW
ROW
1:
2:
3:
4:
2
5
10
11
5:
14
6*:
17
RATE TURNS W/YAW DAMPER FAILURE VESTIBULAR ROLL ERROR ANALYSIS
CONTINUED
.Analysis of Variance Table
Source
of Variation
Degrees of Sum of
Squares
Freedom
Mean
Square
Groups
2
6.483721E-04
3.24186E-04
Within Ggoups
13
4.181815E-03
Among
F
Ratio
Probability
1.01
0.392
3.216781E-04
15
Adj Total
4.830187E-03
The columns used in this analysis are
Column
Column
Column
1 label :6 dof rms error
2 label :2 dof rms error
3 label :0 dof rms error
---------------- m--------m---------- m---------M-----------Analysis of Variance Table
Source
of Variation
Degrees of Sum of
Freedom
Squares
Mean
Square
F
Ratio
------------------------------
Among
------------
Groups
1
6.314917E-04
6.314917E-04
Within Groups
9
2.936414E-03
3.262682E-04
Adj Total
10
1.94
---------
3.567906E-03
m------------------
m------------------
The columns used in this analysis are
Column
Column
Probability
2 label :2 dof rms error
3 label :0 dof rms error
198
0.196
Yaw damper failure vestibular yaw axis error analysis
Column:
1
2
3
4
.00045c 00024 .00001
.00421 .01032
.00761 .00492 .00178
.00084 .00081 .00256
.01186 .01483 .00089
.00000
.00354 .00852
5
.00000
.00002
.00006 .00002
.00000 .00000
.00014 .00022
.00001 .00007
6
.00000
.00011
.00000
.00001
.00000
SUBJECT
Col(
Col(
Col(
Col(
Col(
Col(
1)6
2)2
3)0
4)6
5)2
6)0
MEAN
COLUYM 1
COLUMN 2
COLUMN
3
dp f
dof
dof
dof
dof
dof
rms error
rms error
rms error
mean square error
mean square error
mean square error
ROW
ROW
1:
2:
ROW
ROW
4:
ROW
ROW
STANDARD DEVIATION OF THE MEAN
002 17
5
0Z19
.00135
199
3:
5:
6:
2
5
10
11
14
17
RATE TURNS W/YAW DAMPER FAILURE VESTIBULAR YAW ERROR ANALYSIS CONTINUED
Analysis of Variance Table
Source
of Variation
Groups
Among
Mean
Square
Degrees of Sum of
Squares
Freedom
Probability
F
Ratio
1.726678E-05
2
8.63339E-06
0.703
0.36
Within Groups
13
3.10294E-04
2.386877E-05
Adj Total
15
3.275607E-04
The columns used in this analysis are
Column
Column
Column
1 label :6 dpf rms error
2 label :2 dof rms error
3 label :0 dof rms error
Analysis of Variance Table
Source
of Variation
Degrees of Sum of
Freedam
Squares
Mean
Square
F
Ratio
Among Groups
1
1.446735E-06
1. 446735E-06
Within Groups
9
2.417001E-04
2.685557E-05
Adj Tctal
10
2.431468E-04
0.05
'he columns used in this analysis are
Column
Column
1 label :6 dpf rms error
2 label :2 dof rms error
200
Probability
0.782
Ils landing 500-200'
Column:
1
2
Vestibular Pitch Error Analysis
3
4
5
.246184 .28316 .20340
.2n15
.19930
.05849
.04803
.08018
.39660
.29583
.33196
.28121
.35283
.39329
.36489
.27283
.24172
.40584
.32445 .36454 .15729 .10527
.28370 .38531 .08752 .08049
.29327 .26887 .11020 .08601
.25617 .07908
.31486 .43328 .12449 .09914
.37208 .45187 .15468 .13844
.28528
.13314 .08138
.31961 .27899 .07444 .10215
.42587 .26294 .05843 .18137
.27429 .33849 .16471 .07524
Col(
Col(
CoI(
Col(
Col(
Col(
6
.04137
.03972
.13289
.14846
.07229
.06562
.18773
.20419
.07784
.06914
.11458
1)6 DOF RMS
2)2 DOF RMS
3) ' 0' DOF RMS
4) 6 DOF IMAN SQUARED ERROR
5)2 DOF ?MEA SQUARE ERROR
6)'0' DOF )CAN SQUARE EnOR
RCW
1:
4
RCW
RCW
ROW
ROW
2:
3:
4:
5:
4
RCW
6:
ROW
ROW
ROW
7:
8:
- 9:
10:
RCW
11:
ROW
12:
ROW
NS7:DARD DS"VAtON OF THE
COLL"-
LN
6
6
7
7
12
12
16
16
is
18
IQ.1-
COL.Y.N 3$3
.-3
17
-0 -
.
-.-
-
-
-
-
Analysis of Variance Table
Degrees of Sum of
Squares
Freedom
Source
of Variation
Amorfg
Groups
2
Mean
Square
F
Ratio
1. 252864E-04
6.264322E-05
0.01
Within Groups
30
.1450809
Adj Total
32
.1452062
--------------------
0.987
4.836029E-03
------ -- ---
The columns used in this analysis are
Column
Column
Column
Probability
1 label :6 DOF RMS
2 label :2 DOF RMS
3 label :'O' DOF RMS
201
ils landing 500'-200' vestibular roll error analysis
1
Colt=:
2
3
4
1 .26490 .32970 .22602 .07017
.20452 .12135
2 i.34836
3 .39275 .41215 .44870 .15425
4 .42369 .38310 .48499 .17951
5 .67393 .28868 .43767 .45418
.35305 .12684
6 .35615
7 .52620 .77198 .63102 .27689
8 .68275 .51772 .78715 .46615
9 .59153 .33200
.34991
10 .41113 .37178 .32170 .16903
11 .32790 .76090 .48985 .10752
12 .66327 .45585 .66928 .43993
13
1)6 DOF
2)2 Do?
3)0 DO?
4)6 DO?
Col
Col(
Col(
col(
5
6
.10870 .05109
.04183
.16987 .20133
.14677 .23522
.08334 .19156
.12464
.59595 .39819
.26803 .61961
.11022
.13822 .10349
.57897 .23995
.20780 .44794
SUBJECT
Im ER
RM
Im "ll
zmM sqm
ROW
ROW
ROW
ROW
mmg sq
6)0 DOD ?m sqDEm
Col (
5)2 DOF
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
MEAN
COLUMN 1
1:
2:
3:
4
4
6
4:
6
5:
6:
7:
8:
9:
10:
11:
12:
7
7
12
12
16
16
18
18
STANDARD DEVIATION OF THE
.7S
MEAN
#2"
COLUMN
(-
COLUMN 3_
s-td
--
Analysis of Variance Table
Saurceo
of Variaticn
Amcng
Groups
Dres
of Sum of
FrIN M
Squarin
2
Mean
Square
F
Ratio
Prcbability
9.745206E-04
4.872603E-04
Within Grups
30
.8463396
Mj Total
32
.8473141
The colum used in this analysis are
Colum
CoUM
Colum1
I 1 label :6 DO? a EM
E
2 label :2 DOF M EM
3 label :0 DOF IM EM
3)2
R
0.02
2.821132E-02
0.983
ILS LANDING 500-200'
VESTIBULAR YAW ERROR ANALYSIS
Column:
1
2
3
4
.03849 .03865 .03679 .00148
.04753
.05246
.08157
.06104
.08133
.11629
.08431
.05904
.06788
.07791
.11319
.03577 .00226
.05745 .06593 .00275
.05816 .07370 .00665
.03974 .07098 .00373
.04776 .00661
.13084 .13018 .01352
.08853 .17951 .00711
.05247
.00349
.05537 .06956 .00461
.12470 .11785 .00607
.06834 .11342 .01281
Col(
Col(
Col(
Col(
Col(
Col(
1
.00149 .00135
.00128
.00330 .00435
.00338 .00543
.00158 .00504
.00228
.01712 .01695
.00784 .03222
.00275
.00307 .00484
.01555 .01389
.00467 .01286
ROW
1:
ROW
2:
ROW
3:
ROW
4:
ROW
5:
ROW
6:
ROW
7:
ROw
8:
ROW
9:
ROW 10:
ROW 11:
ROW 12:
1)6 DOF RMS ERROR
2)2 DOF RMS ERROR
3)0 DOF RMS ERROR
4)6 DOF MEAN SQUARED ERROR
5)2 DOF MEAN SQUARED ERROR
6)0 DOF MEAN SQUARED ERROR
FxEAN
COLUMN
6
5
OF
STAnARD DEVIATION
THE
MnAN
SUBJECT
4
4
6
6
7
7
12
12
16
16
18
0-
COLUMN 2
COLUMN 3_
67t 2000
)r5Iq
Analysis of Variance Table
Source
of Variation
Degrees of Sum of
Freedom
Squares
Mean
Square
F
Ratio
7503-------
Among
Groups
2
1.275035E-03
6 .375176E-04
ithfn Groups
30
3.606195E-02
Adj Total
3.733698E-02
32
The columns used in this analysis are
Column
Column
Column
----
0.53
1.202065E-03
1 label
2 label
3 label
:6 DOF RMS ERROR
:2 DOF RMS ERROR
:0 DOF RMS ERROR
203
Probability
0.594
II.S
ANDL
G 500-200'VESTINAR VERITCAL AXIS ACEATION E
Coltu:
1 6.2344
2 6.2298
3 X6.3350
4 6.1356
5 6.0606
6 6.3023
7 6.3923
8 6.2646
9 7.3088
10 6.1945
11 6.1204
12 6.4276
Col(
Col
Col(
Col(
Col(
col(
6.2475 6.1512
6.2393
6.2021 6.5267
6.3423 6.3563
6.2594 6.2886
6.2146
6.1868
6.2502
6.2018
6.1829
6.3755
6.2502
6.2332
6.2191
6.4907
6.3950
6
5
4
3
2
1
38.867 39.031 37.837
38.810
40.132 38.466
37.646 40.224
36.731 39.180
38.928
42.598
40.403
39.547
39.720
39.065
40.861 38.621 38.852
39.245 38.277 38.678
53.419 39.065
38.371 38.462 42.129
37.459 38.228 40.896
41.314 40.647
1)6 dof rms error
2)2 dof rms error
3)0 dof rms eror
4)6 dof mean square error
5)2 dof mean square error
6)0 dof mean square error
ROW
ROW
ROW
ROW
1:
2:
ROW
ROW
ROW
ROW
5:
ROW
ROW
ROW
ROW
COLUM 11
£
/
, A io'7
14-z
z
I
00940,
%COLUMN 2
COLUMN 3
4:
6:
7:
8:
9:
10:
11:
12:
1; AN ARD DEVI TION OF THE MEAN
ALM
-
3:
SUBJECT
4
4
6
6
7
7
12
12
16
16
18
18
STANDlARDl DEVIATION OF THE MEAN
VEANM
Mr
R ANALYSIS
, *
1 02
t6,.1I50
0'
-7
6 OP50I110
I
An I
I
ILS LANDING 500-200'
VESTIBULAR VERTICAL ERROR ANALYSIS
CONTINUED
Analysis of Variarxe Table
Source
of Variation
Degrees of Sim of
Freedom
Sqares
Mean
Square
F
Ratio
Among Groups
2
4.481315E-02
2.240657E-02
ithin Groups
29
1.343928
31
1.388741
Adj Total
Probability
4.634235E-02
The columns used in this analysis are
Column
Column
Column
1 label :6 dof rms error
2 label :2 dof rs error
3 label :0 dof rms error
Analysis of Variance Table
Source
of Variation
Among
Groups
Degrees of SuM of
Freedom
Squares
1
Mean
Square
F
Ratio
Probability
4.178346E-02
4.178346E-02
0.69
Within Groups
20
1.206781
Adj Total
21
1.248564
6.033903E-02
'Ihe columns used in this analysis are
Column
Column
1 label :6 dof rms error
2 label :2 dof rm err=
205
0.415
EIMCATI
IIS LANDIM 500-200' VESTIBLAR INGI'ITJIDAL
1
Colt:
2
3
1 1.0031 0.4518 0.5026
0.7871
2 0.7553
3 -0.5662 0.4322 0.4606
4 0.4045 0.3854 0.6200
5 0.5168 0.6847 0.7421
6 0.9089
1.0692
7 1.3528 0.8227 0.8120
8 0.6325 0.7637 0.5217
9 1.0377 0.4526
10 0.5401 0.5984 1.6875
11 0.6727 1.3619 0.3964
12 1.0548 0.4333
4
ERRCR ANALYSIS
6
5
1.0061 0.2041 0.2526
0.6195
0.5705
0.3205 0.1868 0.2122
0.1636 0.1485 0.3844
0.2670 0.4688 0.5508
0.8261
1.1432
0.6768
0.5833
0.2049
0.3581
0.4525 1.8549
0.6594
0.2721
1.8301
0.4001
1.0767
0.2917
2.8475
0.1571
1. 1126 0.1878
13
SUBJECT
1)6 DOF M ERE=R
m
2)2 DO?
M ERR
DOF
3)0
M
4)6 DO? MEN
5)2 DOF MEAN SqD=
6)0 DOF MEw SUM
Col(
cml(
col(
col(
col(
col(
1'
COLUMN
COLUMN 2
COLUMN 3
1:
4
ROW
2:
3:
4:
5:6:
7:
8:
9:
4
6
6
7
7
12
12
16
16
18
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW 10:
ROW 1U:
ROW 12:
STANDARD DEVIATION OF THE MEAN
. O
2
MEAN
,5g-7 1
(1159(01,
1 'S
ROW
O19
96
-1a
,9
Analysis of Variarnx Table
Scurce
of Variatim
Degrim of SLm of
Squares
Freed=s
man
Square
F
Ratio
Amang Grup
2
.1313013
6.565065E-02
Within
29
3.002143
.1035222
Adj Total
31
3.133445
IhA colu=na used in this analysis are
1 label :6 DO? ME
2 label :2 DO? M ERR
3 label :0 DO? m E
206
Prbability
18
ILS LANDING LAST 20-25 SEC VESTIBULAR PITCH ERROR ANALYSIS
1
Column:
1
2
.10665
.17806
.41999
.22897
.20947
.18884
.22219
.23646
.24244
.23748
.36376
.36405
22
3
4
5
6
7
8
9
10
11
12
.12142
.23232
.19106
.16971
.44732
.24305
.21347
.24067
.19552
3
4
.06547
.20024
.28449
.13075
.18635
.17102
.18292
.29977
.03171
.17639
.05243
.04388
.03566
.04937
.05591
.05878
.14407 .05640
.13613 .13232
.13253
5
6
.01137 .00429
.04010
.01474 .08093
.05397 .01710
.03650 .03473
.02925
.02880 .03346
.20010 .08986
.05907
.04557 .02076
.05792 .01853
.03823
SUBJECT
Col(
Col(
Col(
Col(
Col(
Col(
MEAN
COLUN I.
COLUMN 2
*
COLUMN 3
*
1)6
2)2
3)0
4)6
5)2
6)0
DOF
DOF
DOF
DOF
DOF
DOF
RMS ERROR
RMS ERROR
RMS ERROR
MEAN SQUARE ERROR
MEAN SQUARE ERROR
ROW
ROW
ROW
ROW
ROW
MEAN SQUARE ERROR
ROW
STANDARD DEVIATION OF THE MEAN
.
. /~A
,
207
4aaQ5I
ROW
ROW
ROW
ROW
ROW
ROW
1:
2:
3:
4:
5:
6:
7:
8:
9:
10:
11:
12:
4
4
6
6
7
7
12
12
16
16
18
18
ILS LANDING LAST
20-25 sec
vestibular pitch error analysis
continued
Analysis of Variance Table
Source
of Variation
Mean
Square
Degrees of Sum of
Squares
Freedom
F
Ratio
2
3.624465E-02
1.812232E-02
WithIn Groups
28
.1889603
Adj Total
30
.2252049
Among
Groups
2.69
Probability
0.086
6.748582E-03
The columns used in this analysis are
1 label :6 DOF RMS ERROR
2 label :2 DOF RMS ERROR
3 label :0 DOF R S ERROR
Column
Column
Column
Analysis of Variance Table
Source
of Variation
Among
Groups
Degrees of Sum of
Freedom
Sqares
Mean
Square
F
Ratio
1
3.587835E-02
3.587835E-02
Within Groups
19
.1097514
Adj Total
20
.1456298
6.21
5.776392E-03
The oolumns used in this analysis are
Column
Column
1 label :6 DO? RMSEMCR
3 label :0 DOF IM EMM
208
Probability
0.021
ILS LANDING LAST 20-25 SEC VESTIBULAR ROLL ERROR ANALYSIS
Column:
1
4
3
2
0.25388 0.55113
1.02550 1.38086
1.17510
1.56040 0.81062 0.63444 2.25120
0.95981 0.95156 0.75551 0.92124
0.86984 0.78773 0.42099 0.75662
0.47793 0.21721
0.46606
1.26290 1.11740 0.92298 1.59492
0.87460 1.59530 0.76501 0.76493
0.44014 0.66109
0.19372
0.72128 0.40162 0.83198 0.52024
1.33670 1.96880 0.52688 1.78677
1.39650 0.68742
1.95021
5
6
0.06446 0.30374
1.05165
0.65710 0.40251
0.90547 0.57080
0.62052 0.17723
0.22842
1.24858 0.85189
2.54498 0.58524
0.43704
0.16130 0.69219
3.87617 0.27760
0.47255
SUBJECT
Col(
Col(
Col(
Col(
Col(
Col(
1)6
2)2
3)0
4)6
5)2
6)0
DOF
DOF
DOF
DOF
DOF
DOF
RMS ERROR
RMS ERROR
RMS ERROR
MEAN SQUARE ERROR
MEAN SQUARE ERROR
RMS ERROR
ROW
1:
ROW
2:
ROW
3:
ROW
ROW
4:
5:
ROW
ROW
6:
7:
ROW
8:
,OW
9:
IOW 10:
2OW 11:
)OW 12:
MEAN
COLUMN 1
COLUMN 2
COLUMN 3
STANDARD DEVIATION OF THE MEAN
*/tioo/
0
.,
a
- 6).
209
353
4
4
6
6
7
7
12
12
16
16
18
18
20-25 sec
ILS LANDING LAST
continued
vestibular roll error analysis
Analysis of Variance Table
Source
of Variation
Groups
Among
Within Groups
2
28
.5339456
4.153348
Total
30
4.687293
Adj
The columns used in this
Column
Column
Column
Mean
Square
Degrees of Sum of
Squares
Freedom
F
Ratio
1.80
.2669728
.1483338
Probability
0.184
analysis are
1 label :6 DOF RMS ERROR
2 label :2 DOF RMS ERROR
3 label :0 DOF RMS ERROR
Table
Analysis of Varianc
sorce
of Variation
DegreesOf SumOf
Squares
Freedc
Aming Groups
Within Groups
1
19*
Adj Total
20
.5003579
1.694478
MeanF
Square
2.194836
1 label :6 DOF RM
3 label :0 DOF R
210
EMR
ERRR
Ratio
5.61
.5003579
8.918304E-02
lem columnS used in this analysis are
Column
Column
Probability
0.027
ILS LANDING LAST 25 SEC VESTIBULAR YAW ERROR
Column:
1
z2
3
4
5
6
7
8
9
10
11
12
1
.04849
.06611
.13638 .17584
.15293 .17201
.05373 .05855
.04762
.27062 .26965
.16685 .30929
.07446 .08847
.05498 .06909
.11926 .22268
.13989 .09893
Col(
Col(
Col(
Col(
Col(
Col(
MEAN
COLUMN
1
COLUMN
2
2
1 (2
3
.06302
.08534
.10311
.13490
.04304
.02348
.30618
.21238
.11252
.08567
4
5
.00235
.00437
.01860
.02339
.00289
.00227
.07324
.02784
.00554
.00302
.01422
.01957
.00397
.00728
.03092 .01063
.02959 .01820
.00343 .00185
.00055
.07271 .09375
.09566 .04511
.00783
.00477 .01266
.04959 .00734
.00979
ROW
1:
ROW
2:
ROW
3:
ROW
4:
ROW
5:
ROW
6:
ROW
7:
ROW
8:
ROW
9.:
ROW 10:
ROW 11:
ROW 12:
1)6 DOF RMS ERRORS
2)2
3)0
4)6
5)2
6)0
DOF
DOF
DOF
DOF
6
RMS ERRORS
RMS ERRORS
MEAN SQUARE ERROR
MEAN SQUARE ERROR
DOF MEAN SQUARE ERROR
SUBJECT
4
4
6
6
7
7
12
12
16
16
18
18
STANDARD DEVIATION OF THE MEAN
.
153
M
COLUMN 3
Analysis of Variance Table
Source
of Variation
Among
Degrees of Sum of
Freedom
Squares
Groups
Mean
Square
F
2
8.070948E-03
4.035474E-03
Within Groups
28
.1875894
Adj Total
30
.1956603
0.60
-
The columns used in this analysis are
Column
Column
Column
Probability
Ratio
1 label :6 DO? RMS ERRORS
2 label :2 DO? RMS ERRORS
3 label :0 DO? RMS ERRORS
6.69962E-03
0.554
ILs
ANDM
IAST 20-25 SEC VESTIJIAR V
Column:
3
2
1
6.3200 6.2231
6.6934
6.2210 6.3712
6.4800 6.2999
6.4947 6.5398
6.4438
6.2542 6.3725
6.8827 6.6242
1
43.269
2 6.5779
46.196
3 m6.7967
40.853
4 6.3917
41.861
5 6.4700
40.955
6 6.3996
41.459
7 6.4389
8 6.8293
46.640
9 6.1710 6.5687
38.081
10 6.5727 6.7879 6.2213 43.200
11 6.9811 6.6267 6.2658 48.736
12 6.4602 6.2688
41.735
4
ICAL AXIS ERROR ANALYSIS
6
5
39.943 38.727
44.802
38.700 40.593
41.991 39.689
42.181 42.769
41.522
39.115 40.608
47.371 43.880
43.148
46.076 38.704
43.913 39.260
39.298
SUBJECT
Col(
Col(
Col(
Col(
Col(
Col(
1)6 dof
s eerror
2)2 dof rms error
3)0 dof :ms error
4)6 dof
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
mean square error
5)2 dof mean square error
6)0 dof mean square error
MEANSTANDARD DEVIATION OF
5 Yo70
COLUMN 1
'
COLUMN 2
(o, 94
COLU4tN 3
(,&5
5a
5
1:
2:
3:
4:
5:
6:
7:
8:
9:
10:
11:
12:
E MEAN
4
4
6
6
7
7
12
12
16
16
18
18
ILS LANDING LAST
vestibular vertical error analysis
20-25 sec
continued
Analysis of Variance Table
Source
of Variation
F
Ratio
Mean
Square
Degrees of St= of
Squares
Freedom
AMrq Groups
2
.1152107
5.760533E-02
Within &r=ups
28
1.268216
4.529343E-02
Mj Total
30
1.383427
1.27
Probability
0.296
The columns used in this analysis are
Column
Column
Colun
1 label :6 dof r=s error
2 label :2 dof rms error
3 label :0 dof rm error
Analysis of Variance Table
Source
of Variation
Degrees of Sum of
Freedo
Squares
Amng Groups
Within Grmps
1
19
.1148377
.7944804
Mj Total
20
.9093182
Mean
Square
The columnw used in this analysis are
Column
Column
1 label :6 dof rms error
3 label :0 dof rms error
213
F
Ratio
.1148377
2.75
4.181476E-02
Probability
0.111
VESI1 LAR LCNTIGDIAL AXIS
ILS IANDICL IA9I 20-25 SEC
1
*
3.633
1.9060 2.0005
1
6
5
4
3
2
4.002
0.7267 3.270
0.528
2 1.8083
3 g2.6466 3.0342 0.8544 7.004 9.206 0.730
4 3.4897 1.5169 3.5110 12.178 2.301 12.327
5 2.3892 3.1931 3.0186 5.708 10.196 9.112
4.0032 1.075
16.026
6 1.0368
7 2.5390 2.5699 2.5841 6.446 6.604 6.677
8 1.1548 2.3811 2.5786 1.334 5.670 6.649
0.107 6.328
9 0.3278 2.5156
10 0.6513 0.6441 0.5977 0.424 0.415 0.357
11 1.7182 3.8697 2.8028 2.952 14.975 7.855
9.415 12.728
12 3.0684 3.5677
SL'3JECT
1)6 DOF RM
2)2 DOF IMS EMM~t
3)0 DOF RMS ERFR~
Col(
col(
col
col(
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
ROW
4)6 DF MN SGZRE
5)2 DF MEAN SGUE
6)0 OF MEAN SRE
o
Degrees of S= of
Squares
Frgedm
Amiq Groups
Within GrS
2
28
2.09303
31.86762
Adi Total
30
33.96065
7-7
column
1 label :6 DF
2 label :2 DF
3 label :0 DMF
6:
8:
9:
10:
11:
12:
~6 ~R
-7 4
Table
Mean
sqUare
2w columns used in this analysis are
column
column
4:
7-
Anlysis of Variarc
SOLICI
of Variation
2:
3:
STANDARD DEVIATION OF THE MEAN
MEAN
COLUMN1
COLUMN 2
COLUMN 3.
1:
1.046515
1.138129
F
Ratio
0.92
Prcbability
0.410
4
4
6
6
7
7
12
12
16
.16
18
18
Download