Using Computational Fluid Dynamicrigid Body Dynamic Results to Generate Aerodynamic Models for Projectile Flight Simulation

8/13/2019 Using Computational Fluid Dynamicrigid Body Dynamic Results to Generate Aerodynamic Models for Projectile Flig

1/13

1067

Using computational fluid dynamic/rigidbody dynamic results to generate aerodynamicmodels for projectile flight simulation

M Costello1and J Sahu21School of Aerospace Engineering, Georgia Institute of Technology, Atlanta, Georgia, USA2Weapons and Materials Research Directorate, US Army Research Laboratory, Maryland, USA

The manuscript was received on 29 October 2007 and was accepted after revision for publication on 23 June 2008.

DOI: 10.1243/09544100JAERO304

Abstract: A method to efficiently generate a complete aerodynamic description for projectileflight dynamic modelling is described. At the core of the method is an unsteady, time accuratecomputational fluid dynamic simulation that is tightly coupled to a rigid projectile flight dynamicsimulation. A set of short time snippets of simulated projectile motion at different Mach numbersis computed and employed as baseline data. For each time snippet, aerodynamic forces andmoments and the full rigid body state vector of the projectile are known.With time synchronizedair loads and state vector information, aerodynamic coefficients can be estimated with a simplefitting procedure. By inspecting the condition number of the fitting matrix, it is straightforward toassess the suitability of the time history data to predict a selected set of aerodynamic coefficients.The technique it is exercised on an exemplar fin-stabilized projectile with good results.

Keywords: projectile, flight simulation, computational fluid dynamics

1 INTRODUCTION

Four basic methods to predict aerodynamic forcesand moments on a projectile in atmospheric flight arecommonly used in practice: empirical methods, windtunnel testing, computational fluid dynamic (CFD)simulation, and spark range testing. Empirical meth-ods have been found very useful in conceptual designof projectiles where rapid and inexpensive estimatesof aerodynamic coefficients are needed. These tech-niques aerodynamically describe the projectile with aset of geometric properties (diameter, number of fins,nose type, nose radius, etc.) and catalog aerodynamiccoefficients of many different projectiles as a functionof these features.Thisdata is fit to multi-variable equa-tions to create generic models for aerodynamic coeffi-cients as a function of these basic projectile geometricproperties. The database of aerodynamic coefficientsas a function of projectile features is typically obtainedfrom wind tunnel or spark range tests. This approach

Corresponding author: School of Aerospace Engineering, Geor-

gia Institute of Technology, Atlanta, Georgia 30332, USA. email:[email protected]

to projectile aerodynamic coefficient estimation isused in several software packages including MissileDATCOM, PRODAS, and AP98 [16]. The advantage ofthis technique is that it is a general method applica-ble to any projectile. However, it is the least accuratemethod of the four methods mentioned above, par-ticularly for new configurations that fall outside therealm of projectiles used to form the basic aerody-namic database. Wind tunnel testing is often usedduring projectile development programs to convergeon fine details of the aerodynamic design of the shell[7, 8]. In wind tunnel testing, a specific projectile ismounted in a wind tunnel at various angles of attack

with aerodynamic forces and moments measured atvarious Mach numbers using a sting balance. Windtunnel testing has the obvious advantage of beingbased on direct measurement of aerodynamic forcesand moments on the projectile. It is also relatively easyto change the wind tunnel model to allow detailedparametric effects to be investigated. The main dis-advantage to wind tunnel testing is that it requires a

wind tunnel and as such is modestly expensive. Fur-thermore, dynamic derivatives such as pitch and roll

damping as well as Magnus force and moment coef-ficients are difficult to obtain in a wind tunnel and

JAERO304 IMechE 2008 Proc. IMechE Vol. 222 Part G: J. Aerospace Engineering


2/13

1068 M Costello and J Sahu

require a complex physical wind tunnel model. Overthe past couple of decades, tremendous strides havebeen made in the application of CFD to predictionof aerodynamic loads on air vehicles, including pro-

jectiles. These methods are increasingly being used

throughout the weapon development cycle includ-ing early in a program to create relatively low costestimates of aerodynamic characteristics and laterin a program to supplement and reduce expensiveexperimental testing. In CFD simulation, the fun-damental fluid dynamic equations are numericallysolved for a specific configuration. The most sophis-ticated computer codes are capable of unsteady timeaccurate computations using the NavierStokes equa-tions. Examples of these tools include, for example,CFD++, Fluent, and Overflow-D. CFD is computa-tionally expensive, requires powerful computers toobtain results in a reasonably timely manner, andrequires dedicated engineering specialists to drivethese tools [924]. Spark range aerodynamic testinghas long been considered the gold standard for pro-

jectile aerodynamic coefficient estimation. It is themostaccuratemethod for obtainingaerodynamic dataon a specific projectile configuration. In spark rangeaerodynamic testing, a projectile is fired through anenclosed building. At a discrete number of points dur-ing the flight of the projectile (


3/13

Aerodynamic models for projectile flight simulation 1069

Fig. 1 Reference frame and position definitions

Fig. 2 Projectile orientation definitions

equations are integrated forward in time where aero-dynamic forces and moments that drive motion of

the projectile are computed using the CFD algorithm.The RBD projectile model allows for three translationdegrees of freedom and three rotation degrees of free-dom. As shown in Figs 1 and 2, the I frame is attachedtothe ground while the B frame isfixedto the projectile

with theIBaxis pointing out the nose of the projectileand theJBandKBunit vectors forming a right handedtriad. The projectile state vector is comprised of theinertial position components of the projectile masscenter (x,y, z), the standard aerospace sequence Eulerangles (, ,), the body frame components of theprojectile mass center velocity (u, v, w), and the body

frame components of the projectile angular velocityvector (p, q, r).

Both the translational and rotational dynamicequations are expressed in the projectile body refer-ence frame. The standard rigid projectile, body frameequationsof motionare given byequations (1)through(4) [35]

xyz

=

cc ssc cs csc+ sscs sss+ cc css sc

s sc cc

uvw

(1)

=1 st ct0 c s

0 s/c c/c

pqr

(2)

u

vw

= X/m

Y/mZ/m

0

r q

r 0 pq p 0

u

vw

(3)

pqr

= [I]1

LMN

0 r qr 0 p

q p 0

[I]

pqr

(4)

In equations (1) and (2), the shorthand notations

=sin(), c

=cos(), and t

=tan(). Note that

the total applied force components (X, Y,Z) andmoment components (L,M,N) contain contributionsfrom weight and aerodynamics.The aerodynamic por-tion of the applied loads in equations (3) and (4) iscomputed using the CFD simulation and passed to theRBD simulation.

2.2 CFD solution technique

On the other hand, the CFD flow equations areintegrated forward in time where the motion ofthe projectile that drives flow dynamics are com-puted using the RBD algorithm. The complete setof three-dimensional time-dependent NavierStokesequations is solved in a time-accurate manner for sim-ulation of free flight. The commercially available code,CFD++ is used for the time-accurate unsteady CFDsimulations [36,37]. The basic numerical frameworkin the code contains unified-grid, unified-physics, andunified-computing features. The three-dimensionaltime-dependent RANS equations are solved using thefollowing finite volume equation

t

VWdV+ (F G)dA=

VHdV (5)



4/13


whereWis the vector of conservative variables, FandGarethe inviscid and viscous flux vectors, respectively,His the vector of source terms, Vis the cell volume,and A is the surface area of the cell face. A second-order discretization is used for the flow variables

and the turbulent viscosity equation. The turbulenceclosure is based on topology-parameter-free formu-lations and turbulence modelling thus, becomes acritical element in the calculation of turbulent flowsthat are of interest here. Two-equation higher-orderRANS turbulence models are used for the compu-tation of turbulent flows. These models are ideallysuited to unstructured book-keeping and massivelyparallel processing due to their independence fromconstraints related to the placement of boundariesand/or zonalinterfaces. Higher order turbulencemod-els are generally more accurate and are widely used.

A widely used turbulence model for practical appli-cations is the two-equation k model [38] shownbelow

d(k)

dt =

+ t

k

k

+ Pk (6)d()

dt =

+ t

+(C1Pk C2+ E)T1t (7)

where k is the turbulence kinetic energy, is theturbulence dissipation rate, and t is the turbulence

eddy viscosity which is a function of k and ; Pkis a production term, E is a source term, and Ttis a realizable time scale. Even more sophisticatedturbulence modelling such as the large eddy simula-tion (LES) exists but are prohibitively expensive forpractical problems. Recently, hybrid models knownas hybrid RANS/LES [39] have been developed thatcombine the best features of both RANS and LESmethods based on local gird resolution. The RANSpart is still based on the k model and is used forthe most of the flow domain including the turbulentboundary layer region and LES is used in the wake,for example. The hybrid RANS/LES model is gener-ally more suitable for computation of unsteady flowfields especially at transonic and subsonic speeds.For the computation of supersonic flows that are ofinterest in this research, the two-equation k modeldescribed above is adequate. It has been success-fully used with good results on a number of pro-

jectile aerodynamic applications [19, 24, 34]. Theseturbulence equations are solved all the way to the

wall and generally require fine meshes near the wallsurface.

A dual time-stepping approach is used to integratethe flow equations to achieve the desired time-

accuracy. The first is an outer or global (and physical)time step that corresponds to the time discretization

of the physical time variation term. This time step canbe chosen directly by the user and is typically set toa value to represent one-hundredth of the period ofoscillation expected or forced in the transient flow. Itis also applied to every cell and is not spatially varying.

An artificial or inner or local time variation term isadded to the basic physical equations. This time stepand corresponding inner-iteration strategy is chosento help satisfy the physical transient equations to thedesired degree. For the inner iterations, the time stepis allowed to vary spatially. Also, relaxation with multi-grid (algebraic) acceleration is employed to reduce theresidues of the physical transient equations. It is foundthat an order of magnitude reduction in the residues isusually sufficient to produce a goodtransientiteration.

2.3 CFD/RBD coupling and initial conditionsThe projectile in the coupled CFD/RBD simulationalong with its grid moves and rotates as the pro-

jectile flies downrange. Grid velocity is assigned toeach mesh point. This general capability can be tai-lored for many specific situations. For example, thegrid point velocities can be specified to correspondto a spinning projectile. In this case, the grid speedsare assigned as if the grid is attached to the pro-

jectile and spinning with it. Similarly, to account forRBDs, the grid point velocities can be set as if thegrid is attached to the rigid body with six degrees

of freedom. As shown in Fig. 2, the six degrees offreedom comprises of the inertial position compo-nents of the projectile mass center (x,y, z) and thethree standard Euler angles (, ,), roll angle, pitchangle, and yaw angle, respectively. For the RBDs, thecoupling refers to the interaction between the aero-dynamic forces/moments and the dynamic responseof the projectile/body to these forces and moments.The forces and moments are computed every CFDtime step and transferred to a six-degree-of-freedommodule which computes the bodys response to theforces and moments. The response is converted intotranslational and rotational accelerations thatareinte-grated to obtain translational and rotational velocitiesand integrated once more to obtain linear positionand angular orientation. From the dynamic response,the grid point locations and grid point velocities areset.

In order to properly initialize the CFD simula-tion, two modes of operation for the CFD codeare utilized, namely, an uncoupled and a coupledmode. The uncoupled mode is used to initializethe CFD flow solution while the coupled mode rep-resents the final time accurate coupled CFD/RBDsolution. In the uncoupled mode, the RBDs are spec-

ified. The uncoupled mode begins with a computa-tion performed in steady state mode with the grid

Proc. IMechE Vol. 222 Part G: J. Aerospace Engineering JAERO304 IMechE 2008


5/13


velocities prescribed to account for the proper ini-tial position (x0,y0, z0), orientation (0, 0,0), andtranslational velocity (u0, v0, w0) components of thecomplete set of initial conditions to be prescribed.

After the steady state solution is converged, the ini-

tial spin rate (p0) is included and a new quasi-steadystate solution is obtained using time-accurate CFD.

A sufficient number of time steps are performed sothat the angular orientation for the spin axis cor-responds to the prescribed initial conditions. Thisquasi-steadystate flow solution is thestarting point forthe time-accurate coupled solution. For the coupledsolution, the mesh is translated back to the desiredinitial position (x0,y0, z0) and the remaining angu-lar velocity initial conditions (q0, r0) are then added.In the coupled mode, the aerodynamic forces andmoments are passed to the RBD simulation whichpropagates the rigid state of the projectile forward intime.

3 FLIGHT DYNAMIC PROJECTILE AERODYNAMICMODEL

The applied loads in equations (3) and (4) con-tain contributions from projectile weight and bodyaerodynamic forces and moments as shown below

XY

Z

=W

ssc

cc

8V2D2

CX0+CX2(v2 + w2)/V2

CNAv/V pD2V

CYPAw/V

CNAw/V+ pD2V

CYPAv/V

(8)

LMN

= 8 V2D3

CLDD+pD2V

CLP

CMAwV+qD2VCMQ+ pD2VCNPA vVCMA v

V+ rD

2VCMQ+ pD

2VCNPA

w

V

(9)

The terms containingCYPAconstitute the Magnus airload acting at the Magnus center of pressure while theterms containingCX0, CX2, CNAdefine the steady loadacting at the center of pressure. The externally appliedmoment about the projectile mass center is composedof an unsteady aerodynamic momentalongwith termsdue to the fact that the center of pressure and center ofMagnus are not located at the mass center. The terms

involvingCMAaccounts forthe center of pressure beinglocated off the mass center while the terms involving

CNPAaccounts for the center of Magnus being locatedoff the mass center. The aerodynamic coefficients areall a function of local Mach number which are typicallyhandled through a table look-up scheme in projec-tile flight simulation codes. The aerodynamic model

presented in equations (8) and (9) is the standardaerodynamic expansion for symmetric projectiles.

4 AERODYNAMIC COEFFICIENT ESTIMATION

The time accurate coupled CFD/RBD simulationprovides a full flow solution including the aerody-namic portion of the total applied force and moment(X, Y,Z, L,M,N) along with the full state of the rigidprojectile (x,y, z,, , , u, v, w, p, q, r) at every timestep in the solution for each time snippet. Given a setofn short time histories (snippets) that each contain m

time points yields a total ofh=m n time history datapoints for use in estimating the aerodynamic coef-ficients: CX0, CX2, CNA, CYPA, CLDD, CLP, CMA, CMQ, CNPA.Note that for fin-stabilized projectile configurations,the Magnus force and moment are usually suffi-ciently small so thatCYPAand CNPAare set to zero andremoved from the fitting procedure to be describedbelow.

Equations (8) and (9) represent the applied air loadson the projectile expressed in the projectile bodyframe. Computation of the aerodynamic coefficientsis aided by transforming these equations to the instan-taneous aerodynamic angle of attack reference framethat rotates the projectile body frame about theIBaxisby the angle= tan1(w/v).

8V2D2

1 0 00 c s

0 s c

XY

Z

W

ssc

cc

=

CX0+ CX2(v2 + w2)/V2

CNA

v2 + w2

V

pD

2V

v2 + w2

V CYPA

(10)

8

V2D3

1 0 00 c s

0 s c

LM

N

=

CLDD+pD2V

CLP

(vq+ wr)D2

v2 + w2V CMQ+pD

2V

v2 + w2

V CNPA

(vr wq)D2

v2 + w2VCMQ

v2 + w2V

CMA

(11)

Each time history data point provides a total of six

equations given by the components of equations (10)and (11). The first component of equation (10) is



6/13


gathered together for all time history data points toform equation (12). Likewise, the second and thirdcomponents of equation (10) generate equations (13)and (14), respectively, while the first component ofequation (11) constructs equation (15). Finally, the

second and third components of equation (11) aregathered together to form equation (16). Subscripts onthe projectile state vector and aerodynamic force andmoment components represent the time history datapoint.

1 (v21+ w21)/V21...

...

1 (v2h+ w2h)/V2h

CX0

CX2

= 8

V21D2(X1+Wsin 1)

...

8V2h D

2 (Xh+ Wsin h)

(12)

v21+ w21/V1...

v2h+ w2h/Vh

(CNA)

=

8V21D

2(Y1cos 1+Z1sin 1

Wsin 1cos 1)...

8V2h D

2(Yhcos h+Zhsin h

Wsin hcos h)

(13)

p1D

v21+ w212V21

...

phD

v2h+ w2h2V2h

(CYPA)

=

8V21D

2(Y1sin 1+Z1cos 1

Wcos 1cos 1)...

8V2h D

2(Yhsin h+Zhcos h

Wcos hcos h)

(14)

1 p1D

2V1...

...

1 phD2Vh

CLDD

CLP

=

8L1

V21D3

...

8LhV2h D

3

(15)

0 (v1q1+ w1r1)D

2V1

v21+ w21p1D

v21+ w21

2V21

v21+ w21V1

(v1r1 w1q1)D2V1v21+ w21

0

... ... ...

0 (vhqh+ whrh)D

2Vh

v2h+ w2h

phD

v2h+w2h2V2h

v2h+ w2hVh

(vhrh whqh)D2Vh

v2h+ w2h

0

CMACMQ

CNPA

=

8V21D

3 (M1cos 1+N1sin 1)

8

V21D3(M1sin 1+N1cos 1)

...8

V2h D3(Mhcos h+Nhsin h)

8

V2h D3(Mhsin h+Nhcos h)

(16)

Equations (12) to (16) represent a set of five uncou-pled problems to solve for the different aerodynamic

coefficients. To estimate the aerodynamic coefficientsnear a particular Mach number, a set ofn time accu-rate coupled CFD/RBD simulations are created over arelatively short time period. Since an individual timesnippet is over a shorttime period where the projectilestate variables do not change appreciably, it is criticalthat initial conditions for the different time snippetbe selected in an informed way so that the rank ofeach of the fitting matrices above is maximal. Proper-ties of the fitting matrices above, such as the rank orcondition number, can be used as an indicator of thesuitability of theCFD/RBD simulation data to estimatethe aerodynamic coefficients at the target Mach num-ber. Equation (12) is employed to estimate the zero

yaw drag coefficient (CX0) and the yaw drag coefficient(CX2). To minimize the condition number of this fit-ting matrix, both low and high aerodynamic angle ofattacktime snippets arerequired. Equation (13) is usedto compute the normal force coefficient (CNA) and itrequires time historydata with a nonzero aerodynamicangle of attack. Equation (14) is used to compute theMagnus force coefficient (CYPA) and it requires timehistory data with both low and high roll rate and aero-dynamic angle of attack. Equation (15) is employed toestimate the fin cant roll coefficient (CLDD) along with

the roll damping coefficient (CLP). To minimize thecondition number of this fitting matrix, both low and



7/13


Table 1 Time snippet initial conditions

State Case 1 Case 2 Case 3 Case 4

x(m) 0 0 0 0y(m) 0 0 0 0z(m) 0 0 0 0

(degrees) 0 0 0 0(degrees) 0 0 0 0(degrees) 0 0 0 0V (m/s) 1032 1032 1032 1032v(m/s) 0 0 0 0w(m/s 0 352.5 90 0P(rad/s) 0 0 377 0q(rad/s) 0 0 0 10r(rad/s) 0 0 0 0(degrees) 0 20 5 0

high roll rate time snippets are required. Equation (16)is employed to estimate the pitching moment coeffi-

cient (CMA), the pitch damping coefficient (CMQ), andthe Magnus moment coefficient (CNPA). For successfulestimation of these coefficients, time history data withboth low and high roll rate and aerodynamic angleof attack as well as low and high aerodynamic angleof attack are required. To meet the requirements forsuccessful estimation of all five sets of aerodynamiccoefficients, four time snippets are used all with differ-ent initial conditions. Table 1 lists the four cases withlaunch conditions. Notice that the set of time snippetscontain a diverse set of initial conditions: zero aero-dynamic angle of attack and angular rates; high angle

of attack and zero angular rates; low angle of attack,high roll rate with other angular rates zero; zero angleof attack, high pitch rate with other angular rates zero.

For flight dynamic simulation, aerodynamic coef-ficients are required at a set of Mach numbers thatcovers the intended spectrum of flight conditionsfor the round. If aerodynamic coefficients are esti-mated at k different Mach numbers then a total ofl=k nCFD/RBD time snippets must be generatedto construct the entireaerodynamicdatabase for flightsimulation purposes.

5 RESULTS

In order to exercise the method developed above, ageneric finned projectile is considered. A sketch ofthe projectile is shown in Fig. 3. The projectile hasthe following geometric and mass properties: length=0.1259m, reference diameter=0.013194 m, mass=0.0484 kg, mass center location from base=0.0686 m,roll inertia= 0.74e 06kgm2, pitch inertia= 0.484e04kgm2.

As part of a validation of the coupled Navier-Stokesand six-degree-of-freedom method, time-accurate

unsteady numerical computations were performed topredict the flow field, aerodynamic coefficients, and

Fig. 3 Generic finned projectile

Fig. 4 Unstructured mesh near the finned body

the flight paths of this fin-stabilized projectile at an ini-tial supersonic speed, M=3. Full three-dimensionalcomputations were performed and no symmetry wasused.

An unstructured computational mesh was gener-ated for the generic finned projectile (Fig. 4). Ingeneral, most of the grid points are clustered in the

boundary-layer as well as near the afterbody fin andthe wake regions. Three different grids were usedand the total number of grid points varied from 2to 6 million points for the full grid. For the largermeshes, additional grid points were clustered in theboundary-layer as well as near the afterbody fin andthe wake regions. The first spacing away from the

wall was selected to yield a y+ value of 1.0 in eachcase. The projectile configuration has a base cavityand was included in the mesh generation process.The unstructured mesh also included the base cav-ity region that was present in the actual model testedand was generated using the multi-purpose intelli-gent meshing environment grid-generation softwarerecently developed by Metacomp Technologies.

Here, the primary interest is in the validation ofcoupled CFD/RBD techniques for accurate simula-tion of free flight aerodynamics and flight dynamicsof a projectile in supersonic flight. Numerical com-putations were made for the generic finned projectileconfiguration at an initial velocity of 1032 m/s. The ini-tial angle of attack was, =4.9 and initial spin rate

was 2500 rad/s. Figure 5 shows the computed pres-sure contours at a given time or at a given locationin the trajectory. It clearly shows the orientation of the

body at that instant in time and the resulting asym-metric flow field due to the body at angle of attack.



8/13


Fig. 5 Computed pressure contours

The orientation of the projectile of course changes

from one instant in time to another as the projectileflies down range. Figure 6 shows the variation of theEuler pitch angle with distance traveled. As seen in thisfigure, both the amplitude and frequency in the Eulerangle variation are predicted very well by the com-puted results and match extremely well with the datafrom the flight tests. One can also clearly see thatthe amplitude damps out as the projectile flies downrange, i.e. with the increasing x-distance. Althoughnot shown here, similar behavior is observed with theEuler yaw angle and it damps out with the increasing

x-distance. Computed results again compare very well

with measured data from flight tests. As stated earlier,different computational meshes were used to obtainthe numerical results. Grid sizes varied from 2 to 6million total number of points. The effect of the gridsizes on the computed Euler pitch angle is also shownin Fig. 6. The computed results are grid-independent;

the computed pitch angles obtained with 4 and 6million mesh are essentially the same as those resultsobtained with the 2 million point mesh. In all sub-sequent simulations, the 4 million grid point meshhas been used. Additional validation results showing

other state variables and more details can be found inreference [34].

Figures 7 to 12 present projectile state trajectoriesfor each of the four time snippets. Each time snippetis 0.023 s and contains 50 points, leading to an averageoutput time step of 0.0004. The initial conditions foreach of the time snippets is shown in Table 1. Thesefour snippets create time history data at low and highangle of attack, roll rate, and pitch rate needed foraccurate aerodynamic coefficient estimation. Noticethat cases 2 and 3 have notably more drag down dueto the high angle of attack launch conditions. Case 3 islaunched with relatively high roll rate compared to all

Fig. 7 Velocity for the time snippets.

Fig. 6 Effect of mesh size on the Euler pitch angle



9/13


Fig. 8 Aerodynamic angle of attack for the time

snippets

Fig. 9 Roll rate for the time snippets

Fig. 10 Pitch rate for the time snippets

Fig. 11 Euler pitch angle for the time snippets

Fig. 12 Euler yaw angle for the time snippets

other cases. Case 2 generates roll rate toward the endof the time snippet due to high angle of attack roll-pitch coupling. Significant oscillations in Euler pitchangle are created in case 2 with some cross couplingresponse exhibited in Euler yaw angle. Figures 13 to16 plot aerodynamic forces and moments in the localangle of attackreference frame defined above for cases1, 3, and 4 since these cases are the primary cases usedto estimate thecoefficients. For allcases, theaxialforceoscillates from20 to25 N. There exists a slight biasbetween the CFD/RBD and estimated data of about0.5 N for low angle of attack time snippets. For mod-erately high angles of attack (Case 3), the estimateddata also oscillates with a much higher amplitude thanthe CFD/RBD data indicating that CX2 is estimatedlarger than the CFD/RBD suggests. The normal forcetime snippets agree well between the CFD/RBD andestimated data for all time snippets. For the exam-ple finned projectile, side force (FZ) and out-of-plane

moment (MY) are generally small (


10/13


Fig. 13 Estimated (dashed) and CFD/RBD (solid) body

axis axial force (Fx) versus time

Fig. 14 Estimated (dashed) and CFD/RBD (solid) nor-

mal force (Fy) versus time

The CFD/RBD and estimated data agree reasonablywell, but certainly do not overlay one another. Theonly time snippet that creates notable rolling momentis case 3 which is launched with an initial roll rateof 377 rad/s. Notice that the estimated data smoothlygoes through the CFD/RBD data which oscillates ina slightly erratic manner. The in-plane moment (Mz)agrees reasonably well for both the CFD/RBD and esti-mated data. The results shown inFigs4 to15 are typicalfor all Mach numbers. The overall observation fromthe data is that the estimated aerodynamic model fitsthe CFD/RBD data well, with the notable exception ofaxial force where a bias is exhibited.

The example projectile investigated in this paperhas been fired in a spark range at Mach 3.0 withaerodynamic coefficients computed via conventionalaerodynamic range reduction. Table 2 presents a

comparison of aerodynamic coefficients obtainedfrom spark range testing and subsequent coefficients

Fig. 15 Estimated (dashed) and CFD/RBD (solid) body

axis rolling moment (Mx) versus time

Fig. 16 Estimated (dashed) and CFD/RBD (solid) yaw-

ing moment versus time

obtained using the method described here. Noticethat most aerodynamic coefficients such as CX0,CNA,andCMAare in reasonably good agreement with sparkrange reduced data. Axial force yaw drag and rolldampingarebothdifferentbyaround20percentwhilepitch damping is different by around 40 per cent. Withthe exception of CMQ, aerodynamic coefficients arenearly estimated to within the accuracy that can beexpected from a spark range test firing between twosets of firings. The relatively larger errors in CMQ aremore than likely due to the set of initial conditions thatcreate a large condition number for the fitting matrix.

CFD/RBD data was generated at six different Machnumbers ranging from 1.5 to 4.0. The estimationalgorithm discussed above was used to compute acomplete set of aerodynamic coefficients across itsMach range. These results are provided in Table 3.

With the exception of CX2, the steady aerodynamiccoefficients are smooth and follow typical trends for



11/13


Table 2 Comparison of estimated aerodynamic coefficients and estimated coefficients at Mach 3.0

Spark range data CFD/RBD Percent differencespark range reduction PACE estimation between coefficients (%)

Zero yaw axial force coefficient, CX0 0.221 0.238 7.17.7Yaw axial force coefficient,CX2 5.0 5.9 15.018.0

Normal force coefficient derivative,CNA 5.83 5.64 3.23.3Pitching moment coefficient derivative, CMA 12.6 13.82 8.89.7Pitch damping moment coefficient,CMQ 196 134 31.646.3Roll damping moment coefficient,CLP 2.71 3.37 19.624.4

Table 3 Aerodynamic coefficients versus Mach number

Mach no. 1.5000 2.0000 2.5000 3.0000 3.5000 4.0000

CX0 0.4309 0.3413 0.2821 0.2387 0.2051 0.1829CX2 0.2109 5.5363 5.7136 5.9329 5.5131 2.0191CNA 8.0982 7.0441 6.1940 5.6441 5.2608 5.0026CLP 4.4758 4.4114 3.8793 3.3788 2.9415 3.2974CMA 23.7600 18.6200 15.7178 13.8278 12.4043 11.3124CMQ

282.8

277.7

182.3

134.4

112.0

77.8

variation in Mach number. The yaw drag coefficient,CX2, however, is somewhat erratic with a low valueof 0.21 at Mach 1.5 followed by a steady rise untilMach 3.5. Pitch damping decreases with Mach num-ber as would be expected for a fin-stabilized projectilebeyond Mach 1.0. Roll damping steadily increasesuntil Mach 4.0 when in drops off notably.

6 CONCLUSIONS

Using a time-accurate CFD simulation that is tightlycoupled to a RBDs simulation, a method to efficientlygenerate a complete aerodynamic description for pro-

jectile flight dynamic modelling is described. A set ofnshort-time snippets of simulated projectile motion atm differentMach numbersis computed and employedas baseline data. The combined CFD/RBD analysiscomputes time synchronized air loads and projectilestate vector information, leading to a straightforwardfitting procedure to obtain the aerodynamic coeffi-cients. The estimation procedure decouples into fivesub problems that are each solved via linear leastsquares. The method has been applied to an examplesupersonic finned projectile. A comparison of sparkrangeobtained aerodynamic coefficients with the esti-mation method presented here at Mach 3 exhibitsagreement within 10 per cent for CX0, CNA, and CMA;agreement within 20 per cent for CX2 and CLP; andagreement within 40 per cent forCMQ. This techniquereported here provides a promising new means forthe CFD analyst to predict aerodynamic coefficientsfor flight dynamic simulation purposes. It can easilybe extended to flight dynamic modelling of differ-

ent control effectors provided accurate CFD/RBD timesimulation is possible and an aerodynamic coefficient

expansion is defined which includes the effect of thecontrol mechanism.

REFERENCES

1 Sun, J.andCummings, R.Evaluation of missile aerody-namic characteristics using rapid prediction techniques.J. Spacecr. Rockets, 21(6), 513520, 1984.

2 Moore, F.The 2005 version of the Aeroprediction Code

(AP05). In the AIAA Atmospheric Flight MechanicsConference, Providence, Rhode Island, 2004, AIAA paper2004-4715.

3 Sooy, T. and Schmidt, R. Aerodynamic predictions,comparisons, and validations using missile DATCOMand aeroprediction 98. In the AIAA Aerospace SciencesMeeting and Exhibit, Reno, Nevada, 2004, AIAA paper2004-1246.

4 Simon, J. and Blake, W. Missile DATCOM high angleof attack capabilities. In the AIAA Atmospheric FlightMechanics Conference, Portland, Oregon, 1999, AIAApaper 1999-4258.

5 Neely, A. and Auman, I. Missile DATCOM transonic

drag improvements for hemispherical nose shapes. Inthe AIAA Applied Aerodynamics Conference, 2003, AIAApaper 2003-3668.

6 Blake,W. MissileDATCOM 1997status andfutureplans.In the AIAA Applied Aerodynamics Conference, Atlanta,Georgia, 1997, AIAA paper 1997-2280.

7 Dupuis, A. and Berner, C. Wind tunnel tests of along range artillery shell concept. In the AIAA Atmo-spheric Flight Mechanics Conference, Monterey, Califor-nia, 2002, AIAA paper 2002-4416.

8 Berner, C.andDupuis, A.Wind tunnel tests of a grid finprojectile configuration. In the AIAA Aerospace SciencesMeeting, Reno, Nevada, 2001, AIAA paper 2001-0105.

9 Evans, J.Prediction of tubular projectile aerodynamics

using the ZUES Euler code. J. Spacecr. Rockets, 1989,26(5), 314321.



12/13


10 Sturek, W., Nietubicz, C., Sahu, J., and Weinacht, P.Applications of computational fluid dynamics to theaerodynamics of army projectiles. J. Spacecr. Rockets,1994, 31(2), 186199.

11 Nusca,M., Chakravarthy, S.,and Goldberg,U. Computa-tional fluid dynamics capability for the solid-fuel ramjet

projectile.J. Propuls. Power, 6(3), 1990, 256262.12 Silton, S. Navier-Stokes Computations for a spinning

projectile from subsonic to supersonic speeds.J. Spacecr.Rockets, 2005, 42(2), 223231.

13 DeSpirito, J., Vaughn, M.,andWashington, D.Numeri-cal investigation of canard-controlledmissilewith planargrid fins.J. Spacecr. Rockets, 2003, 40(3), 363370.

14 Sahu, J. Numerical computations of transonic criticalaerodynamic behavior.AIAA J., 1990, 28(5), 807816.

15 Weinacht, P. Navier-Stokes prediction of the individ-ual components of the pitch damping sum. J. Spacecr.Rockets, 1998, 35(5), 598605.

16 Guidos, B., Weinacht, P.,and Dolling, D. Navier-Stokes

computations for pointed, spherical, and flat tippedshells at Mach 3. J. Spacecr. Rockets, 1992, 29(3),305311.

17 Sahu,J. and Nietubicz, C. J.Application of chimera tech-nique to projectiles in relativemotion.J. Spacecr. Rockets,1995, 32(5), 795800.

18 Park, S. and Kwon, J. Navier-Stokes computations of sta-bility derivatives for symmetric projectiles. In the AIAA

Aerospace Sciences Meeting, Reno, Nevada, 2004, AIAApaper 2004-0014.

19 Sahu, J.Numerical simulations of supersonic flow overan elliptic-section projectile with jet-interaction. In the

AIAA Aerodynamics Conference, St. Louis, MO, 2427June 2002, AIAA paper 2002-3260.

20 Qin, N., Ludlow, K., Shaw, S., Edwards, J.,and Dupuis,A. Calculation of pitch damping for a flared projectile.J. Spacecr. Rockets, 1997, 34(4), 566568.

21 Weinacht, P.Coupled CFD/GN&C modeling for a smartmaterial canard actuator. In the AIAAAtmospheric FlightMechanics Conference, Providence, Rhode Island, 2004,

AIAA paper 2004-4712.22 Park, S., Kim, Y., and Kwon, J. Prediction of dynamic

damping coefficients using unsteady dual time steppingmethod. In the AIAA Aerospace Sciences Meeting, Reno,Nevada, 2002, AIAA paper 2002-0715.

23 DeSpirito, J.and Heavey, K.CFD computation of mag-nus moment and roll-damping moment of a spinning

projectile. In the AIAA Atmospheric Flight MechanicsConference, Providence, Rhode Island, 2004, AIAA paper2004-4713.

24 Sahu, J.and Heavey, K. R. Unsteady CFD modeling ofmicro-adaptive flow control for an axisymmetric body.Int. J. Comput. Fluid Dyn., 2006, 20(5), 271278.

25 Garon,K.,Abate, G.,and Hathaway,W. Free-flight testingof a generic missile with MEMs protuberances. In the

AIAA Aerospace Sciences Meeting, Reno, Nevada, 2003,AIAA paper 2003-1242.

26 Kruggel, B. High angle of attack freeflight missile testing.In the AIAA Aerospace Sciences Meeting, Reno, Nevada,1999, AIAA paper 1999-0435.

27 Danberg, J.,Sigal, A.,and Clemins,I.Aerodynamic char-

acteristics of a family of cone-cylinder-flare projectiles.J. Spacecr. Rockets, 27(4), 1990.

28 Dupuis, A. Free-flight aerodynamic characteristics ofa practice bomb at subsonic and transonic velocities.In the AIAA Atmospheric Flight Mechanics Conference,Monterey, California, 2002, AIAA paper 2002-4414.

29 Abate, G., Duckerschein, R., and Hathaway, W. Sub-sonic/transonic free-flight tests of a generic missile with

grid fins. In the AIAA Aerospace Sciences Meeting, Reno,Nevada, 2000, AIAA paper 2000-0937.

30 Chapman, G. andKirk, D.A method for extracting aero-dynamic coefficients from free-flight data.AIAA J., 1970,8(4), 753758.

31 Abate, G. and Klomfass, A. Affect upon aeroballisticparameter identification from flight data errors. In the

AIAA Aerospace Sciences Meeting, Reno, Nevada, 2005.32 Abate, G.andKlomfass, A.A new method for obtaining

aeroballistic parameters from flight data. In the Aerobal-listic Range Association Meeting, Freiburg, Germany,2004.

33 Weinacht, P., Sturek, W.,andSchiff, L.Projectile perfor-

mance, stability, and free-flight motion prediction usingcomputational fluid dynamics.J. Spacecr. Rockets, 2004,41(20), 257263.

34 Sahu, J. Time-accurate numerical prediction of free-flight aerodynamics of a finned projectile. In the AIAA

Atmospheric Flight Mechanics Conference, San Fran-cisco, California, 2005, AIAA paper 2005-5817.

35 Etkin, B. Dynamics of atmospheric flight, 1972 (JohnWiley & Sons, USA).

36 Peroomian, O., Chakravarthy, S.,and Goldberg, U. Agrid-transparent methodology for CFD. In the 35thAerospace Sciences Meeting and Exhibit, Reno, Nevada,1997, AIAA paper 97-07245.

37 Peroomian, O., Chakravarthy, S., Palaniswamy, S.,and

Goldberg, U.Convergence acceleration for unified-gridformulation using preconditioned implicit relaxation. Inthe 36th Aerospace Sciences Meeting and Exhibit, Reno,Nevada, 1215 January 1998, AIAA paper 98-0116.

38 Goldberg, U. C., Peroomian, O., andChakravarthy, S.Awall-distance-free K-E model with enhanced near-walltreatment. ASME J. Fluids Eng., 1998, 120, 457462.

39 Batten, P., Goldberg, U.,andChakravarthy, S.Sub-gridturbulence modeling for unsteady flow with acousticresonance. In Proceedings of the 38th AIAA AerospaceSciences Meeting, Reno, Nevada, January 2000, AIAApaper 2000-0473.

APPENDIX

Notation

CLDD fin cant aerodynamic coefficientCLP roll damping aerodynamic coefficientCMQ pitch damping moment aerodynamic

coefficientCNA normal force due to angle of attack

aerodynamic coefficientCX0 zero yaw drag aerodynamic

coefficientCX2 yaw drag aerodynamic coefficient

CYPA Magnus force aerodynamiccoefficient



13/13


D projectile diameterFx, Fy, Fz total applied force components in

body reference frameMx,My,Mz total applied moment components

about mass center in body reference

framep, q, r components of angular velocity

vector in body reference frameu, v, w components of velocity vector of

mass center in body reference frame

V magnitude of relative aerodynamicvelocity vector of mass center

W projectile weight(=mg)x,y, z components of position vector of

mass center in an inertial reference

frame

aerodynamic angle of attack air density, , Euler roll, pitch, and yaw angles


Documents

Using Computational Fluid Dynamicrigid Body Dynamic Results to Generate Aerodynamic Models for Projectile Flight Simulation