Professional Documents
Culture Documents
Multidisciplinary Modeling and Design of A Space System
Multidisciplinary Modeling and Design of A Space System
of a Space System
John T. Tester, David G. Robinson
Department of Aeronautical Engineering
Wright-Patterson AFB, OH 45433
446
CH3051-0/91/0000-0446 $1.00 0 1991 IEEE
Authorized licensed use limited to: Tennessee Technological University. Downloaded on April 07,2022 at 15:32:44 UTC from IEEE Xplore. Restrictions apply.
below. Each of the subsystem models are briefly dis- Satellite Rigid Body Rotation Model- The satellite’s
cussed i n the following. rotation about its center of mass was modeled inde-
pendently of the vibrations and the orbital model.
Structural Model - A structure was necessary to con- The bond graph, located lower-center in the figure be-
nect the optical subsystem to the controllers. This low, has no linear, translational mass inertance; only
structrire was a flexible beam tetrahedron assembly. rotational mass inertia is necessary for the applica-
The structure had mass and inertia, as did the op- tion of momentum wheel controllers.
tical system and the momentum wheel controllers.
Torque inputs to the frame were transmitted to the Satellite Orbital Model - The satellite orbit(a1 inodel
optical system to keep it aligned with the missile im- is used t o provide the position of the satellite i n in-
age; unfortunantly, these inputs also produce struc- ertial space. The only force acting on the satellite
tural vibrations in the beams. These vibrations dis- is that of gravity (Note: the torque input from the
place the optical system, thus misaligning the image momentum wheel controllers changes the orientation
on the optical axis. The vibration was modeled us- of the satellite, not the orbit position). The model
ing modal analysis techniques and implemented using requires an initial condition for the orbital altitude:
hond graphs. The structural vibrations/modal bond anywhere from low-earth to geosynchronoiis orbit.
graph is the dominant portion of the upper-right por- The bond graph is located above t,he hallistic mis-
tion of the figure. sile portion of the bond graph.
Opt.ical Rlodcl - Much of the total system perfor- Controls Model- The satellite attitude was cont3rolled
iiiaiicc i l l formation was obtained through the optical via three momentum wheel controllers. h4omentuln
sii1,systein model. The purpose of this system was wheels provide the torque input to the system to pro-
to focus Ilie light of the missile image onto an image duce rotation of the satellite about its physical, prin-
plane, where a. light-sensitive detector plate was lo- cipal axes. This motion allows the satellite’s telescope
cat.ed. ‘The optical simulation utilized exact ray trac- to remain centered on the target image during the
ing t.echniques [l], and is a FORTRAN subroutine missile launch. The momentum wheels were modeled
wit.hin t.hc bond graph (labeled O P T in the figure). as three rigid rotating masses, attached to each of the
A t,arget missile is represented as a moving point satellite’s base spars. The controls model is not a sin-
source of light. A perfect image occurs when all the gle bond graph; instead, it is represented by the threc
(.raced rays converge to a single point on the detector. torque inputs and connections off each 0-junct.ion i n
I f tlerociis or off-axis abberations occur, the rays will the rigid body rotation model.
not, convcrgc, yielding a scatter of points on the two-
tlimcnsional detector plate. The positions of these System Model- The subsystem models were iiite-
point,s at, a given instant of time are captured in the grated into a single system model. This system in-
simiilation for the calculation of MTE. cluded not only the satellite’s subsystems, but that
A second aspect of the optical model was the mea- of its orbit and the missile ballistic niodel as well.
surcment. of the light intensity sensed on the detector The development of the system model had a num-
plat,e. The light ima.ge intensity, used to estimate ILI, ber of benefits. First, the development process itself
\t‘iis niat.lieinatically calculated in a part of the opti- increased communication flow a.mong tea.m members.
cal motlcl sepa.ra.t,efrom the ray tracing portion. IVR A common language was availa.ble which incrcasetl er-
iisctl bot.li ray tra.ce and image intensity results in its ficiency; engineers were discussing int.eract,ioil i n tlis-
c;ilciilat ion. cipline neutral terms (power a.nd flow rat.lier t # l i a i i
loads, temperature, velocit,y or current). This L’C-
Alissilc Orhit.al Model - The missile dynamics model suiting system model a.llowed the engineers t.o t.al;e
siiiiiilat.cs t.he launch ofa two stage Titan I11 mis- an objective view of importance of their subsyst,eni
sile. T h e simulation starts at launch and ends at modeling and design efforts on the overall s y s h i i de-
second s h g e burnout as only the boost portion of sign effort. ‘Extraneous’ modeling and ana.lysis were
t,he missile launch is of interest. This model is based greatly reduced. Bond graphs allows the system en-
on t,he assumption of only two forces acting on the gineer to literally see where these manipula.tions take
missile: gra.vity and thrust and uses bond graphs to place and investigate the subsystems intera.ct,ions i n
iniplement free body dynamics principles. The mis- a logical manner. Synergistic effects are captured t,o
sile honti graph is in the lower-left portion of figure an extent not possible when modeling at the siihsys-
11el o\\, . tem level and the result is that opt,imizat,ion can be
accomplished a t the system level.
447
Authorized licensed use limited to: Tennessee Technological University. Downloaded on April 07,2022 at 15:32:44 UTC from IEEE Xplore. Restrictions apply.
ameter. These variables were optimal for the system
I Design Variables I Optimum I performance at values close to their minimums.
I Orbit radius I 9.723.840 meters I The above activity concluded one complete design
Settlin,g time I 1 second iteration for the satellite. Clearly, more refinement
Beam diameter I 5 centimeters is in order, following along the same procedure of
Primary mirror diameter 1.5352 meters simulation, data collection, optimization and results
Performance Measures O D t i m u m analysis. Many assumptions were made during the
research that were not mentioned in this article. For
AlTE -.39158
example, all performance measures were equally con-
ILI -.8003774
sidered in the optimization process; yet, in an orbital
IVR 2.0269969 system, mass is clearly of top priority early in the
SA4 -.7119192 design process.
Obviously, more detailed analysis and modeling
will have t o be accomplished, but a strong framework
111. O P T I M I Z A T I O N has been established to support future efforts.
A methodical approach was needed to arrive a t the
optimal values for the system design. A design of ex- V. A C K N O W L E D G E M E N T S
periemen ts (DOE) technique involving response sur- The authors would like t o thank the students a n d fac-
face methodology (RSM) was selected. The system ulty of the AFIT System Engineering graduate pro-
compr~tersimulation had 4 design variables. These gram who also participated in this research: Capt,ain
were determined after examining the original 23 vari- Brian Cassiday, Captain Steve Gaines, Captain L a r r y
ables, 4 performance measures, and the interaction Gatschet, Captain Mario Moya, Dr. Brad Lielist, a n d
matrices produced from an interpretive structural Dr. Curt Spenny.
modeling analysis [3]. The RSM central composite
design required 25 variable value settings for the data References
collection process. This data was fitted t o a sec-
ond order, multiresponse, empirical math model us-
[l] Briggs, Hugh C. and others. Integrated Con-
ing A NOVA techiniques. Lastly, the empirical model t rol/St m ct ures/Op t ics D y n a m a c Pe rjo rm n n ce
was simultaneously optimized over the design space Modeling of a Segmented Reflector Telescope.
through the use of a statistical distance function [4]. Prepublished report. Jet Propulsion Laboratory,
The bond graph system model allowed the system- Pasadena, CA. No date.
a t i c change of variables values. The basic system
model i n ENPORT stayed the same; only the values [2] Cassiday, B. K., Gaines 11, S. O., Gat,schet,
w i t h i n its framework were altered for each simulation L. L., Moya, M. N., and Tester, J . T. M i / / -
run. tidisciplinary Modeling and Desagn os a S p c e
Structure. MS Thesis, AFIT/GSE/ENY/SlD-l.
IV. R E S U L T S School of Engineering, Air Force Institute of
The optimal design settings and performance re- Technology (AU), Wright-Patterson AFB OH,
sults are listed in the above table. The performance December 1991.
results are listed in unitless values, since the opti-
miza t,ion process used standardized design variable [3] Hill, Douglas J . and Warfield, John N. Unzfied
Program Planning IEEE Trans. Systems, M a n ,
settings and outputs.
and Cybernetics, Vol SMC-2 No 5 November
The relatively low earth orbit reflects the need for
1972
the shortmestpractical distance between the missile
a n d the sat,ellit,e. This need is due to the maximiza- [4] Khuri, A. I., and Conlon, M., “Simultaneous 01,-
t,ion of light incident upon the detector plate. The timization of Multiple Responses Represented by
earlier regression analysis proved the time response Polynomial Regression Functions,’’ Technomet-
was a significant factor only in the MTE. Similarly, rics, Vol 23, No. 4, Nov 1981.
the beam diameter was a significant factor in only
the syst.em mass, which was also minimized. These [5] ROSENCODE Associates, Inc., The ENPORT
two isolated design variables were therefore driven to Quick Guide. Copyright (c) 1987, by ROSEN-
t,lieir minimum values. The only two design variables CODE Associates, Inc.
which truely coupled in the optimization process were
t,lierefore the orbit radius and the primary mirror di-
448
Authorized licensed use limited to: Tennessee Technological University. Downloaded on April 07,2022 at 15:32:44 UTC from IEEE Xplore. Restrictions apply.
i I
I I
449
Authorized licensed use limited to: Tennessee Technological University. Downloaded on April 07,2022 at 15:32:44 UTC from IEEE Xplore. Restrictions apply.