(/

COfPONNT PART NOTICE

THIS PAPER IS A COMPONENT PART OF THE FOLLOWING CMPILATION REPORT:

TI1LE: Transactions of the Army Conference on Applied Mathematics and

Comtputing (2nd) Held at Washington, DC on 22-25 May 1984.

To ORDER THE COMPLETE CCOMPILATION REPORT, USE AD-A154 047

THE CO"PONENT PART IS PROVIDED HERE TO ALLOW USERS ACCESS TO INDIVIDUALLY

AUTHORED SECTIONS OF PROCEEDING, ANNALS, SYMPOSIA, ETC. HOWEVER, THE COMPONENT
SHOULD BE CONSIDERED WITHIN THE CONTEXT OF THE OVERALL COUPILATION REPORT AND
NOT AS A STAND-ALONE TECHNICAL REPORT, •
a.

THE FOLLOWING COUPONENT PART NUMBERS COMPRISE THE COMPILATION REPORT,

AD#: P004 902 thru P004 958 AD#:.
• ,AD#.- AID#.-
: AD#-.+ AD#:-. ..

TI S GFRA&I m
DTIC TAB
Unannouncead
J1Istif ocatio

B_ (EL E C YE
* ~~~~Distribution/ ~U 9~5
*Availability CodeasUG1
Avail and/or'
Athi doci~ent bci sben OPPiovedDit Sda.
fny: public releaseo and sale; its A
'jiitribution is unlimited.

" 8. DT~rFOM4O5I: DTIC-1ID

iN\
In ¸N

A METtIODOLOGY FOR TIME DEVELOPMENT OF FIRE CONTROL EQUATIONS

FOR GUNS AND ROCKETS FIRED FROM AIRCRAFT ".

o• Harold J. Breaux

Computer Techniques & Analysis Dranch

Division
Systems Engineering & Concepts Aiialysis
United States Army Ballistic Research Laboratory
Aberdeen Proving Ground, Maryland 21005

ABSTRACT

\ ) The accurate firing of unguided projectiles (bullets or rilckets) from aircraft leads to a requirement
for rapid computation of the launch vector needed to assure the projectiles striking a gIvein target. The
computation of this laying vector and fuse time is the function of tile on-board fire control system, Tile
fire control system includes sensors which measure target range and velocity, aircraft attitude, polit ion
and velocity, sad atmospheric couditious. These meaqurements are fed to an on-board fire control
computer which in real time, typically at W0 Il, must compute anew the laying vector :Ipproprinte for the
rapidly varying variables which influence the ballistic trajectory. Six-degree-of-freedom models, which are
normally used in laboratory ballistic modeling and simulation, are computationally too slow and otherwise
cumbersome to be Implemented for real time fre control. A methodology for developing an alternative-
simplified, yet very accurate, model is described in detail,

1. INTRODUCTION

The fundamental objective of fire control ballitics is the provision of a means for aiming and fuze
setting of a rocket or projectile so as to best assure placement of accurate and effective Ore on a selected
target. This objective was accomplished historically by the publication, and use in the field, of the
classical "firing table', a book of numbers tabulated so as to provide an easy means for determining
azimuth and elevation settings for guns and rockets. As computer technology has evolved, the firing table
has been relegated to manual backup and the field computer computes gun and rocket settingi in real
time. Modern weapon systems such as computer controlled air defense guns, and helicopters and tanks
which fire on the move encounter a fre control problem characterized by very rapidly changing variables
which drive the ballistics. As a result, a wholly new level of difficulty and sophistication is placed on the
problem of fire control ballistic prediction.

The fundamental "generic" models used extensively for prediction of launch and exterior ballistic
performance have come to be known as Six-Degree-of-Freedom (SDF) models. Such models embody the
equations of motion for both translational and rotational displacement of a projectile or rocket. The
development, refinement, maintenance and modilications for new technology, of this type model, has long
been performed in Defense Departme:t Laboratories, and in particular, Army laboratories such as BRL.
See, e.g,, Lieske and McCoyl and Bernell 8. The SDF model is a natural extension of the three-degree-
of-freedom (TDF) model, long a mainstay of ballistic fire control prediction. This latter model represents
only the translational aspects of projectile motion and is of an order of magnitude less diflicult in
computational labor, Howe-ver, TDF models may, in some circumstances, be insufficiently accurate
becawaae they do not model the Interaction of translational motion with the aerodynamic ellects associated
with yaw and pitch along with the yaw and pitch interaction with spin,
I R. F. Lieske and R. L. Mc Coy, I Equatiom of Motion of ARigid Projectile," UiL. Report No. 1244, Marc, 1064.
2 0. Barnett, 'Trajectory Equationu for a Six.Degree.o.-Fretdom-.Misuie liiMg a Fuixcd.Plae Coordinate System,' Twbniral Reprt.
-301, Picatiuty Arsemal, Dover, New Jersey, Juste 1, "

729 "".
.. .. . . . . .. . . . . . . .. . . . . . ..... .A.-
~ ~~............. A .... .- . . . . •.

* U, "%
• *" •J•"•
4" ,• •.F
"", ' •.•,' ',••.••,,'°,•v -•'."-•,".-" ",'.%. NO
, ' . '"'.""', . ' " " ','',,,,% .- " .. ,, ' ,.,' .' ,'.
 Artillery projectiles, like the classical spinning top In mechanics, have very high frequency mgtiun
associated with spin, precession and nutation. A fundamental computational requirensent or numerica
Integration, (the technique used for solving thens models) is that oscillatory variables need to be sampled
at several points within each oscillation to maintain accuracy and stability. The result of this is that sm~all
steps in time are required in the forward marching process of integration. Accordingly, complete SDP
calculations for spinning artillery or automatic cannon projectiles is time consuming, relatively expensive,
and is done sparingly on laboratory computers. An alternadive to the SDY' and TDF models was developed
by Lieske and Reiter z which haj been found extremely useful for firing table computation and
implementation in some ground bawed Ire control computers. Their successful approach was to dcev'lup a
modified TDF model, MTDF, which incorporated an explicit estimate of the "yaw of repose" *nil its
effects into the translation equations withonat having to integrate the high frequency motion of the full
SDF models. This model has resulted. in greatly reducing the computational burden referred to abnve
while incurring only a slight loss in atcurcy. iIs now a standard model in the repertoire of bulli, tic P.
mathematical tools widely used in the United States and the NATO defense community. 6*

Hlelicopter fire control requires the ure of ballistic models as described above. The SDF model ii
needed for rocket, and the MTDF (and SDF ) model are needed for automatic cannon, Unfortunately, for
reasons detailed herein, these models, can only serve an intermediate role, albeit an important one, in the
process of d&veloping an on board-real time ballistic prediction model. A fundamental problem in attack
aircraft fire control is that of maintaining the 'timeliness of the ballistic solution. In a turning, elitsbing
maneuver, the aircraft velocity components, the geometrical relations between target and cannon/rocket
and other variables- all change rapidly. Note also that these models really solve the "Inverse'" of the fire
control problem. In fre control one specifies terminal conditions such am target location with resp,,t to
the launch platform. The solution desired consists of the departure attitude needed to strike the target
and the time of flight. Accordingly, one would need to solve the problem Iteratively, I.e., guess at a trial
solution, and continually readjust the departure angles until the desired terminal conditions are satislied.
Meanwhile, if the aircraft Is in a maneuver, the problem has changed because the variables driving the
ballistics have changed since the iterations were initiated. The consequence of this is that, a ballistic
solution that is not computed instantaneo.ly, (or nearly so), is old and obsolete before the muni*ion can
be fired. Despite the recent and continuing revolution in computer technology, the embedded comlputers
In aircraft ire control systems are small in memory capacity, and are not fast enough to iteratively solve
the models described above at the required frequency.

Modern attack helicopters such as the Cobra and Apache are armed with as automatic cannon and
a family of 2.76 inch or Hydra 70 rockets. By the lip of a switch, the pilot can select the munition he
wishes to fire. Accordingly, the model(s) embedded in the on board Are control computer must be able to
key on this switching process and compute the solution for the munition selected, In fact, it is possible,
" ~and sometimes desirable, to be able to Oire both the gun anot rorkets simultaneously, The requirement for
representing ten or more rocket types, each having differing weights, measures, aerodynamikm, staging and
fuzing cbaracteristics- all add to the need for developing a common general model. Furthermore, Army
aircraft such as the Cobra and Apache have differing modes of articulating rocket pods. It is highly
desirable that the rocket ballistics be developed independently of the method of pod articulation. See
,* Appendix F. The logical strategy evolves for developing a procedure which makes the general model
applicable to a specific munition by eelective retrieval of a pool of constants (bued on switch position),
Some constants of the model, however, may differ from one aircraft type to another. See Appendix E. This
necessity for providiog a capability for many munitions tends to reduce the storage capacity available for
the collection of instructions related to the model itself. The requirement for an accurate hllistic
solution, for such a family of munitions, varying types of aircraft, air defense guns, tanks, or other moving
gun platforms, that can be computed cyclically, in real time, during an engagement, leads to the need
o. for the methodology described herein.
ItIF. Like . ad M. L. Reiter, 'Eqeatioas of ttules for a Mediled PNOit Mum.,' BmAJitle Raefeah Laboratores Report No.
1314, March IM5.77

730 -

%...., .
 '.4

2. GENERAL DESCRIPTION OF THE MATHEMATICAL METHODOLOGY

A perspective for defining the mathematical problem can be obtained by examination of Figure (2.
1). Therein an attack helicopter in flight is depicted engaging a ground target. While the scenario
b!

HYDRA 70 FIRE CONTROL (INDIRECT FIRE)

0.0 APACII 04 COBRA

sunwwitiows't~r•t WE AliRC POIN •co

-, i

to's

MEmTiei i VVc

Figure (2-1). Typical attack helicopter scenario for rocket firing.

depicts the use of a Hydra-70 rocket, the employment cf an automatic cannon would be similar. The
telescopic sight unit, (TSU), is continually maintaining the hue of sight, (LOS), which can be viewed as a
vector connecting a reference point on the aircraft to a selected point on the target. On board sensors
continually measure the vector velocity and attitude orientation of the aircraft along with the LOS range, "
i# , between target and aircraft. For moving targets, sensors coupled with mathematical "ilters"
continually estimate the vector velocity of the target. Target motion, along with other considerations,
determine components z. and ij which along with x, deline where the projectile should be in "one time of
flight". Environmental factors such as wind, air density and temperature are also provided by sensors on
the aircraft. The temperature of the munition is also known, either by use of a magazine thermometer or
by assuming that the munition hu the same temperature as the environment. Required information on
downwash due to the rotation of the helicopter blades is obtained by means described in Appendix D.
"Components of the gravity vector are made available by reference to orientation of coordinate axes that
are aligned with the local gravity vector.

The above described Information Is fed, at typically 50 Hz, to the Fire Control Computer, (FCC).
The primary function of the FCC is to compute the angular settings (the attitude of departure) and for

Y 731]
,' *- .*- • , ., . .. , .
- ,, .,...- . .,,,..- .',' ."
•.*. ,.. ., ., ,. •', *•...- . *...._.%.
•'. . '.. . .. ... ,. , . .. , " ,. ..., . .. ... , . , . ,* .. , .'. /. .... * . . ... . . .. •.
t,* ,* *4 %
% A* . ~.bA ~ . .-' -.. ~sA ~~ A dI ' ~ iA *qpI. I
 some munitions, ruse time, which best assures accurate placement of Are on the target. The objective is
thus seen to be the development of a collection of formulae, to Implement In the FCC, which
accomplishes tbis task. Toward this end It is useful to review two procedures, used previously, that can
serve as building blocks toward development of a more generalized procedure.

Global Fitting Approach

A perspective on the global fitting spproscA can be obtained from the work of Chandler, Baker and
Dinjar at the US Army Redstone Arsenal ad C. Masaitis and HI. Breaux at the Ballistic Research
Laboratory. That work Is reviewed (and references listed) by Breaux 1. The objective of that work was to
find u alternative to the SDF models for use In fre control with ground based missiles such a the
Redstone, Jupiter, Pershing and Lance. For these stationary-at-launch, ground systems, certain
complexities unique to the moving platform are not present and the computational speed factor is not u
critical. A computational cycle time of seconds, or even tens of seconds, is tolerable. However, at the
longer ranges of these systems, other complexities enter such u those associated with earth rotation and
curvature. Nevertheless, a basic procedure employed in that work remains as the cornerstone of current
methodology. That concept is depicted in Figure (2-2) below and consists of approximating one model

! ,

MIN/ OWN ff
BALLISTICS INREAL

till Flut
MEAL

W AC

iK

Figure (2-2). Triangle of Approximations for Development of Fire Control Ballistics, .'
Rdereace to umpubliahed work anDR•L rehiod to sttsch beti~copten I cited i. the Ackmowledgmeut. "

S4. Breaux, 'Tlb. Computatios of Frirnin Table. for Guided Miauiles," DRL Report No..1545, Noemniber ISIS. •

732'°
I11
I..

! ,..V

*"+•i +[ I + " - i " + i. . .+ • i. .. .. + .. . .. . | .. .+ .. . Cq.... C...

C ..., . . .... . . s ~ C ~ t * *I
 with another model. Firt a "truth" model must be developed-both &aa means for assessing actual
ballistic performance and then to make possible the development and accuracy assesment related to the
fre control model. This model must represent all the munitions of interest-- fired with launch conditions
representing the aircraft environment, Speed of computation is not a significant factor here and this
model is iormally implemented In a laboratory computer.

The global fitting approach consists of basically five phaAcs.

(A) Development of the SDF Model.

(B) Specification of a candidate fire control equation (FCE)
model to be fitted.
(C) Development of a data base of trajectory calculations which
span the expected range of all variables (and mixtures)
in the planned deployment scenarios.
(D) Fitting of the model to obtain values for all coefficients.
(E) Computer validation of the model.-

Phae (A) generally consists of obtaining all weights and measures, physical characteristics, and
aerodynamic functions that are appropriate to the munition(s) of interest and processing the data into a
form acceptable to the SDF program, Special features of the munition ballistics or launch dynamics may
require modification of the SDF program, This Is a very critical phase in that everything done hereafter
hinges on the adequacy and accuracy of this model. Phase (0) is the most difficult aspect of the problem
and is the core of the methodology to which a large part of this paper is devoted. Phase (C) consists of
designing a matrix of conditions and use of the SDF program to compute the corresponding trajectories.
The fitting process, Phase (D), is itself a part of the methodologly. However, the procedure designed by
"" Brcauz 4, will be used herein to form a hybridized methodology which will be described. Phase (E) is
* designed to test the execution of the model in a form similar to it's field employment and simultaneously
compare its predicted results with the SDF "truth model". 6"

Phase (E) should not be confused with field validation. In field validation the results of the effort, r,.
namely the ballistic FCE's, are programmed into the Fire Control Computer and actual live tests
conducted. Poor performance In this phase can frequently occur due to improper programming of the
FCE's, poor sensor performance, or a poorly designed or specified SDF model, If the latter Is found to be
true, the SDF model must be fixed and the process described above munt be rtpesled4,
Closed Form Solution Approach

A solution approach has been employed by Norwood 6, for airborne gunfire, that makes use of the
"Method of Siaceci". In this approach, the computational problem is reduced to a collection of Integrals.
Norwood's approach was made more practical by Benokratie4, who obtained closed form expressions for
the Siacci integrals by approximating the projectile drag by three connected line segments, This approach
has the benefit of "permanence" in that the approximate solution is there for all time-for all projectiles
that satisfy the fiat fire assumption and which can be adequately approximated by a TDF model. The
drag curve must also permit adequate approximation by the three line segments. The disadvantage is
that this technique does not Incorporate ballistic effects that arise from factors that only the SDV model
Now that the methodoloay and related solftwaro ha been developed, reprocessing for a revised or sew SDP data bane can be dose

6 Norwood, Jobs M., 'A Review of the Method of Silmi Trajectory Coiputatios for Airborns ausire,' ARL-TM-71-27, Applied
Research Laboratories, The University of Texas at Austin, Austin, Tas, August 1071.

$ 1Denokiaitis, Vitalism, 'A Closed Form Siaccd Fusctios Method for Trajectory Calculations,' Jourial Of Ballistics, V 6, No.1,
1982. pp. 1326.1347.

733
.%.'

• ~~~~~~~~~~~~........
... ....... .. .. ,.,. ,•......... , ..... ',• ......... •• 2•.ft ..

't 1 .*..* . f Ii *, ft...... II. I m i.

 could predict. Ballistie ehecs due to aerodynamic *drift" and "jump" and secondary drag effects related
to initial yaw and yaw rate must be added to the model by extraneous methods which require essentially
the same effort as the global fittien approach. This method can represent the free light phase of a rocact
but not the boost phase. Extension of such a model to include more complex ballisties generally leads to
the need for some form of fitting and in turn necessitates all the stages of the global fitting approach.

The methodology developed herein can be viewed as a hybrid method which combines features of .
both the global fitting technique and the closed form approximation. It cam also be viewed Li a
generalization and extension of the earlier work done by the uualer. The key aspect of the work ".
described therein was the use of step-wise reresslon (least Squares) to facilitate the process of model
building. This process can be be exhibited by simple example. Assume that a process, whose outcome is
represented by w, is dependent on the variables x,ys. Assume that a very accurate model exists that
•" permits the speciOcatio• of ,x,y and a and then computles w. However, the •onmoutation is so expensive
and tinme consuming that it can only be done sparingly and at facilities remote fr,,w, where the need exists.
This suggests the need and ponibility of developing an approximating model which is easier to evaluate.
In attempting such an effort consider the following two approachos to Slobal fitting.

The Empirical Approach to Global Fitting

H1ere no speeifie information on the underlying mathemintical model is assumed known other than the
dependence of w on x, y and a.A candidate linear model Is specified by the formula

WN- so+ a,#+ lft+ 5j+ 4 821+ 34

eggs+ sy

+ 5?3YI+ ass 0+ 498 + (2.1)

* Numerous types of linear modeb could be employed, however, a polynomial model is assumed for
illustration. By providing a data bas of w versus xy and a for an appropriate mix and range of the
variables and the above model to a stepwise regression procedure as described by Breauz T', or Hocking 9
one might arrive at a suitable model. The result of such a regression procedure Is the sequential listing of
submodels, each differing from the previous submodel, by one term taken from the above collection of
terms. The first submodel is the one term stone which best approximates w. The second model contains
two terms, including the one term model plus a second term which when added to the first provides the
best approximation. The process continues in this fashion but simultaneously seeks to weed out terms 'S
that are no longer useful due to cross correlations between clusters of variables that have been introduced
into the model. Variations In strategy and computational sequencing are discussed at length by Hlocking'. ,...
By examining the program output which Includes the progression of the variance of residuals in w, the
correlation coefficient, the 't" values on the regression coeffleients, etc., and performing some
experimentation on model definition, one can generally build a suitable model by this process.

The Physical Approach to Global Fitting

In the physical approach one first recognizes that the problem at hand embodies a field of knowledge to
include a collection of literature. Such a literature indicates that the process leading to the outcome w
bhs a "closed form' result, dependent on x,y, and a, it certain idealizing assumptions are made. One ,.
such outcome might be
,%

SBreaux, H. J., '90 Stepwlse Multiple Linmi Repesislm,' ORL Report No. 1349, Aquoat 1T7,
Breaux, 11.J,, OA Modifctmioa of Efroymsom's Techmique for Btepwin Resteulos Asalysh,' 0om. of the ACM, 1058, Vol. 11,

Harkims, R. R., 'The Asalyais &adSelectleu of Variable inLimea Regruslos,' iometrics, V. St, March 2576, pp 1.49 ,-0

734

JL.
 1%~

w -As + B, ezxp(a X v/). (2-•)

This ,might suggest that a more intelligently defined model woIld proceed from Elq. (2-2) as a bn.,e. iror
this simple model one might use non-linear least squares and try to fit A,, B, and a as free parnrmeterm.
When the model is complex, containing many variables mnd terms, it is generally better to linenrize the
model and proceed as follows: In Eq. (2-2), for example, expand non linear terms in a Taylor series

1w -as + a[ /l+ds(ZN/z) 2 +a,(1/)+

+ s XY (+-.+

As before, this candidate model and the data base is processed by the stepwise regression Inress. The
resultin• model, developed by the latter procedure, will, in general, be superior to one developed I)y Jll',
"iempirical process. Any insight, born of experience and knowledge of the process, helps townrd better
dining the candidate model, used as a point of departure, when thereafter employing :a fitting process.

The (ire control problem addressed herein is analogous to the above simple example with one very
111izportlmit difference. The example concerned itself with one process and one approxinmating imodn(, wtv
needed. Hlence, either of the two approaches, if found to lead to an adequate model, could be viewed as
!
A successful. The attack helicopter fire control problem, by contrast, can be viewed as representing ten or
more related processes. If the problem is addressed by the empirical approach described above, Intuition
and experieiwe indicates that the result will be as many separate models as their are rocket/munition

Tilt. approach taken will be one of examining all components of the launch ald Ilight proaem.s. the
boost nndi free flight coupling, and practical idealizations which provide a physical bsis for simplilivil
rlosed form solutions leading to an Intelligent definition of the candidate model. These compomnents. will
be p•ucessed by the stel)wise regression procedure by use of a data base of calculations froi, nn Si)t
programi to determine their adequacy. The approach is thus seen to be a hybridized one which inchldest
obtaining closed form approximations and then employs the features of the global fitting appronch as
described above.
.4

Note: Due to its length, only Sections 1. and 2. of this paper have been included ia the Proceedings.
t*nfoiriation on the availability of the complete document can be obtained by writing to:

Director
- USA Ballistic Research Laboratory
"ATTN: DRXHR-TSD
Aberdeen Proving Ground, Md. 21006

This aspect wan found to be critically Important in the currently one.5iog Fire Control Systems latecration Frocralu for (70t.rm,
. The toagnitude of effort, &ad resulting time required, to provide for u many u nine dillerent rocket types, forced anu
eatly freeze"
N In tLte balic strurtuie of the model long before work wan completed. The Lhatic model was them progianuned, vhlidAt-d for •periflc
socket typ,•, and then coellcieuts for the remaining members of the famlhy were added al they were obtained. This permiitted
N timely progress on the overall program,

735

.N .4 NN- . . . . . . . . . . . . . . . .

'.~ ~
,~N**.,.,.,.... ,.....,, ..... .,~.N~'N ,.".......t,. .. ... ,.,..,,.....
.. )1..'.., A
4. i N.
.,,.,.
i. .
....
. t . i N i v
.,,•....
.N. t l .A
k Nl
'•.,..',.
4k 4 I•
,
•t N
''
k
..
of."
[ • llqt
pi tl k t "## lt 4
Nt 1ilt•q. t N . t