You are on page 1of 25

Introduction to Modeling and Simulation

John A. Christian
Aerospace Engineering and Engineering Mechanics
The University of Texas at Austin
Suggested Readings:
NEW SMAD, Chapter 14
ASE 166M Spacecraft Systems Laboratory
The purpose of this class
Learn how to model spacecraft subsystem performance

Draw on experience and knowledge learned in previous
classes

Leads to a few questions
What kind of spacecraft are there?
What are the common spacecraft subsystems?
What do I need to know about modeling and simulation?
What are the basics of spacecraft subsystem sizing?
ASE 166M Spacecraft Systems Laboratory 2
Classes of Spacecraft & Mission Types
Crewed Spacecraft
Crew transportation (Space Shuttle, Orion, Soyuz)
Space stations (ISS, MIR)
Surface systems (habitat, etc)
Robotic Spacecraft
Near Earth
Earth observing (weather, surveillance, Earth science)
Communications (TDRSS)
Scientific (Hubble)
Planetary
Flyby (New Horizons, Voyager, Cassini)
Orbiter (MRO, MESSENGER)
Lander (Phoenix, MERs, MSL)
Probe (Huygens)
ASE 166M Spacecraft Systems Laboratory 3
Spacecraft Subsystems
Common Subsystems
Propulsion
Attitude Determination & Control System (ADCS)
Communications
Command & Data Handling (C&DH)
Power
Structures
Thermal Control System (TCS)

Other Subsystems (depending on spacecraft application)
Thermal Protection System (TPS)
Landing Systems
Environmental Control and Life Support System (ECLSS)
ASE 166M Spacecraft Systems Laboratory 4
We will cover
these subsystems
this semester
Parametric Modeling
This semester we will be developing conceptual design (pre-phase A)
level tools. If you need more fidelity, youll have to call in the subsystem
experts.

Many of the models we develop will be parametric. A parametric model
is usually a series of mathematic relationships that relate a metric of
system performance (e.g. mass, power, cost) to some set of design
variables or parameters.

Parametric models are useful during spacecraft design because:
they usually give you enough insight to make initial design decisions
they dont require as much information as more sophisticated design tools
and analyses
they easily allow for trade studies between multiple subsystems with
competing interests (this is your final project!)
ASE 166M Spacecraft Systems Laboratory 5
All models are wrong
ASE 166M Spacecraft Systems Laboratory 6
All models are wrong, but some models are useful.
--George Box
In 1618-1621 Kepler suggested the following model
for the motion of the Planets about the Sun:

1. The orbit of each planet is in an ellipse with the sun at
one focus
2. The heliocentric radius vector of each planet sweeps
over equal area in equal time
3. The square of the orbital period is proportional to the
cube of the ellipse semimajor axis http://www.thejubileeacademy.org/articles/2006_
solar_system_image.html
Later, Newton would suggest a more accurate model:
3
/
1
3
/
i P
n
P i
i
i P
i P
P P P
r
m m
G m r r F

=
=
= =

Which model is better?
All models are wrong
ASE 166M Spacecraft Systems Laboratory 7
All models are wrong, but some models are useful.
--George Box
m

F

Newton suggested the following model: ( ) r r F

m m
dt
d
= =
But this is nothing more than a mathematical abstraction of the
observed behavior of a physical system. In fact, many years later
Einstein demonstrated that Newtons model is wrong.
Einstein suggested this model:
This model has been shown to be more accurate than Newtons
model.
But is the model perfect?
More importantly, does it matter
if the model is perfect?
( )
|
|
.
|

\
|

= =
2 2
0
1 c v
m
dt
d
m
dt
d r
r F

Wait, are you saying everything I know is wrong?


ASE 166M Spacecraft Systems Laboratory 8
Yes.
But heres the beauty of engineering:
You dont have to be exactly right, you just have to be close enough.
The solution to almost any problem may be made arbitrarily complex. Dont
develop a complicated model when a simple one will get the job done. More
complicated models take more resources to develop, more resources to
validate/verify, and more resources to run.
This brings up a central point in modeling and simulation.
What fidelity model should you use?
How accurate does your result need to be in order to answer the
question?

How good (accurate) is your input data. If only rough values are known
for the inputs, a very sophisticated model may not give you a better
estimate than a simpler model. This is why you typically move from low-
fidelity models early in the life cycle to high-fidelity models later in the life
cycle.
ASE 166M Spacecraft Systems Laboratory 9
More Fidelity Less Fidelity
Where on this spectrum should you be?
Deep thoughts on model fidelity
Assumption: Almost all models involve some form of simplifying
assumptions or approximations to make the problem solvable. The
important question is what errors are you introducing by simplifying the
problem.

Error: Are these errors important and do they effect the results? Knowing
what assumptions to make (and when to make them) is what separates
you from your computer.

Range: Be aware of the range of validity of your model. Again, always
ask yourself: When do the simplifying assumptions used to make the
problem tractable break down?

Results: Engineers are frequently asked to perform analysis using
existing codes and models. Be a smart user. How much do you trust the
results? Always know the range of validity of the model. Always research
the model assumptions and know the models limitations.
ASE 166M Spacecraft Systems Laboratory 10
More deep thoughts on model fidelity
Significant Figures: Be cautious of the number of significant digits you report in
your results. Think of it this way, the number of digits you use represents the
degree to which you believe the outcome of your model.




There are at least two main reasons to limit the number of digits you report:
1. Your input variables are only known to a certain level of precision
2. The approximations made by model introduce errors. Do not report results to
a precision that implies a higher-fidelity model than what you actually used.

Fidelity: Be cautious of disparate fidelity between models used in the same
analysis. The fidelity of the analysis will be limited by your least accurate model.
Can you lower the fidelity of some of your models (thus reducing modeling effort)
without impacting the overall fidelity of the output results?
ASE 166M Spacecraft Systems Laboratory 11
The idea is to be approximately right, rather than exactly wrong.
--John Tukey
An Example
ASE 166M Spacecraft Systems Laboratory 12
Compute the range, r, of a
projectile fired from a
cannon with an initial
velocity, v
0
, of 11 m/s at an
inclination, , of 18 degrees.
Most of you would probably take the following approach:
Assume a flat Earth and that gravity is constant, g
0
, and acts down. Neglect drag.

Under these assumptions, you could derive the following equations for projectile motion:

a
y
= -g
0
v
y
= -g
0
*t + v
0
*sin()

y = -0.5*g
0
*t
2
+ v
0
*sin()*t + y
0
g
0

a
x
= 0
v
x
= v
0
*cos()

x = v
0
*cos()*t + x
0
Experience suggests that some of you might also say that g
0
= 9.80665 m/s
2
. You may go on to
say that the projectile would impact the ground at x = 11.22792 m.
2 m
But this level of accuracy is not
consistent with the assumptions
made in your model!!!
Instead, you should say something like g
0
= 9.81 m/s
2
and x
impact
= 11.2 m.
Simulation
Simulation is the act of subjecting your model to
environments and inputs that are in some way
representative of what the real system will experience


Common types of simulation include:
Computer based simulations
Field tests


ASE 166M Spacecraft Systems Laboratory 13
Multidisciplinary Simulation
What if I want to create an integrated simulation using models from
many different disciplines (e.g. trajectory, power, propulsion, thermal,
cost, risk, etc.)?
There are at least three ways to accomplish this task:
1. Monolithic Code: rewrite all models as individual subroutines (frequently with
approximations such as look-up tables, etc.) and integrate subroutines into a
single code
2. Manually Integrated Models: each model is executed by a local expert and the
results are integrated manually through meetings, telephone calls, emails, etc.
3. Tightly Integrated Models: shell scripts or wrappers are added around existing
(legacy) models. These wrapped models are electronically integrated for
automated execution and data exchange
ASE 166M Spacecraft Systems Laboratory 14
Manually
Integrated
Models
faster, single user
Monolithic
Code
Tightly
Integrated
Models
expert involvement,
more fidelity
Why is Integrated Design Important?
ASE 166M Spacecraft Systems Laboratory 15
Image Ref: NASA Chief Engineers Office.
ASE 166M Spacecraft Systems Laboratory 16
User
Inputs
Mission
Profile
(?V)
Comm.
Mass and
Sizing
Power
Propulsion
Geometry
GN&C
(ADCS)
Thermal
N
2
Diagrams
N
2
Diagrams are also called Design Structure Matrices (DSMs)

DSMs are used to gain an understanding of how information is
passed between different subsystems (or functions) within a
system
Feedback
Feed-forward
N
2
Diagrams
ASE 166M Spacecraft Systems Laboratory 17
Model
output to
upstream
output to
downstream
input from
downstream
input from
upstream
D
A
E
C
B
The presence of a dot signifies that
information is passed from one model
to another at the indicated intersection
D
E
Feedback requires iteration. The two
relevant models must iterate until the
outputs and inputs match.
In this case, for example, on the first pass through the N
2
diagram you must guess the inputs from E to D
(because you have not yet run model E). After you run D, you use the outputs to run E. You will see that the
outputs of E probably do not match your guess. Therefore you rerun D using the updated values of the
outputs from E. The new outputs of D will require you to rerun E, and so on. You must continue to iterate until
the process converges.
Order of model execution
The design is said to be closed when
all iterations have converged and all
the inputs/outputs to all the model are
consistent with each other.
Fixed-Point Iteration
Iteration is frequently performed through fixed-point
iteration (FPI). Iterations in MS Excel are performed
using this method. This is also sometimes called the
method of successive substitution.

It is important to know how this technique works because
this method is not universally applicable and this may
cause MS Excel to crash for some problems.
ASE 166M Spacecraft Systems Laboratory 18
D
E
A fixed-point is defined as follows:


Using this definition of a fixed point,
a FPI routine may be set up:
( )
* *
x g x =
x
y y = x
y = g(x)
x
*
( )
n n
x g x =
+1
Fixed-Point Iteration
ASE 166M Spacecraft Systems Laboratory 19
x
y y = x
y = g(x)
x
0
x
1
x
2
x
y y = x
y = g(x)
x
0
x
1
x
2
x
3
stable fixed point
unstable fixed point
A FPI routine should be stable if -1 < g(x
*
) < 1

Consider the following two examples
y
0
y
1
Fixed-Point Iteration Example
ASE 166M Spacecraft Systems Laboratory 20
Consider the following problem:
x
n+1
= log
10
(x
n
+2)
x
0
= 4.0
Iter Num x
n
x
n+1
=g(x
n
)
0 4.00000 0.77815
1 0.77815 0.44376
2 0.44376 0.38806
3 0.38806 0.37804
4 0.37804 0.37622
5 0.37622 0.37589
6 0.37589 0.37583
7 0.37583 0.37581
8 0.37581 0.37581
9 0.37581 0.37581
This problem converges to within
5 significant digits of the fixed
point after 8 iterations
N
2
Diagrams as FPI
ASE 166M Spacecraft Systems Laboratory 21
A
B
C
( )
n n
x g x =
+1
Recall the simple FPI defined earlier:
Let the variable passed from C to A be x
But we need a guess of x
before we can begin the
analysis.
Lets try closing this
feedback with FPI. Break
the link and call your initial
guess x
0
x
0
After one complete pass through the DSM, we
have a new value of x that is produced by
module C, this will be x
1
Replace x
0
with x
1
and
repeat the FPI process until
the value for x converges to
within a specified tolerance

g(x
0
)

x
1
N
2
Diagram Rescheduling
Rescheduling is the process of rearranging the order of execution of
models within a simulation

The objective of rescheduling is usually to reduce the simulation runtime
ASE 166M Spacecraft Systems Laboratory 22
Mass
and
Sizing
Cust.
Req.
Cost
Risk
Traj.
Risk
Cust.
Req.
Cost
Mass
and
Sizing
Traj.
N
2
Example 1 Launch Vehicle Design
ASE 166M Spacecraft Systems Laboratory 23
T
W
u
W
b
S
Opt.
,r,
T_W
0
W
gross
A
e
, T
vac
,
Isp
vac
V
split
W
gross
, S
ref
MR
upper
r,
T_W
0
W
gross
, S
ref
,
W
prop
, W
land
T
sl
/W
e
W
upper
W
dry_booster
MR
booster
W_S
W
gross
P
p
e
N
2
Diagram in ModelCenter
You may see the intersections
labeled to indicate what variables
are being passed between models
Christian, J.A., Final Project: Launch Vehicle Design, AE6374/Advanced
Design Methods II, Georgia Institute of Technology, Spring 2006.
N
2
Example 2 Mars EDL Design
ASE 166M Spacecraft Systems Laboratory 24
Aerodynamics model computes
pressure distribution and lift/drag
coefficients
Vehicle sizer and optimizer adjusts
vehicle geometry to meet mission
needs
Supersonic inflatable aerodynamic
decelerator (IAD) model dynamically
scales to meet mission needs
-6
-4
-2
0
2
-2
0
2
0
0.5
1
1.5
2
2.5
Crossrange, km
Downrange, km
A
l
t
i
t
u
d
e
,

k
m
Powered descent guidance
algorithm attempts to guide vehicle
to specified landing site

9
7
5a
3
5b

5c
5d
8

6

5e
4

6.5 m
8.4 m
4 m
Christian, J.A., DCarlo, P.A., Otero, R.E., Salmon, J.L.,
and Sanders, J.L. An Entry System for Delivering
Massive Payloads to the Martian Surface,
AE8803/Planetary Entry, Georgia Institute of
Technology, Spring 2007.
ASE 166M Spacecraft Systems Laboratory 25
Questions
Marty Brennan
Office: WRW 415A
marty.brennan@gmail.com

You might also like