JOURNAL OF AIRCRAFT

Vol. 38, No. 5, September– October 2001

Fluid-Structure Interactions of a Round Parachute:

Modeling and Simulation Techniques

Keith R. Stein¤ and Richard J. Benney†

U.S. Army Soldier and Biological Chemical Command, Natick, Massachusetts 01760
Tayfun E. Tezduyar‡
Rice University, Houston, Texas 77005
and
John W. Leonard§ and Michael L. Accorsi¶
University of Connecticut, Storrs, Connecticut 06269

A parallel computational technique is presented for carrying out three-dimensional simulations of parachute
 uid-structure interactions, and this technique is applied to simulations of airdrop performance and control
Downloaded by STANFORD UNIVERSITY on August 5, 2013 | http://arc.aiaa.org | DOI: 10.2514/2.2864

phenomena in terminal descent. The technique uses a stabilized space-time formulation of the time-dependent,
three-dimensional Navier– Stokes equations of incompressible  ows for the  uid dynamics part. Turbulent features
of the  ow are accounted for by using a zero-equation turbulence model. A Ž nite element formulation derived
from the principle of virtual work is used for the parachute structural dynamics. The parachute is represented as
a cable-membrane tension structure. Coupling of the  uid dynamics with the structural dynamics is implemented
over the  uid-structure interface, which is the parachute canopy surface. Large deformations of the structure
require that the  uid dynamics mesh is updated at every time step, and this is accomplished with an automatic
mesh-moving method. The parachute used in the application presented here is a standard U.S. Army personnel
parachute.

Introduction element formulation derived from the principle of virtual work

F LUID-STRUCTURE interactions (FSI) are involved at all is used for the structural dynamics (SD).10;11 The coupling be-
stagesof airdropsystems performance,includinginitialdeploy- tween the FD and the SD is enforced over the  uid-structure in-
ment, during in ation, terminal descent (or gliding/maneuveringfor terface, which is the canopy surface. Large deformations of the
steerable parachutes), and soft landing (i.e., retraction for round structure are handled by updating the FD mesh with an automatic
parachutes,  ared landing for ram-air parachutes). The interac- mesh-moving scheme and remeshing as needed. The DSD/SST
tion between the parachute system and the air ow around it is procedure is well suited for problems involving spatial domains
dominant in most parachute operations, and thus the ability to changing with time, such as those encountered during parachute
simulate parachute FSI is recognized within the parachute re- FSI.12 This formulation has been well tested and applied to a
search community as a serious challenge.1 6 In this paper a de- large variety of  uid dynamics problems involving moving bound-
scription is given of current efforts to develop a general-purpose aries and interfaces. In the space-time formulation the Ž nite ele-
computer model that can accurately predict three-dimensional FSI ment interpolation functions vary both spatially and temporally,
for various parachute systems under different performance stages. and this automatically takes into account changes in the spatial
Here, the focus is on the FSI performance during the termi- domain.
nal descent stage to include riser control performance. Issues in- In recent years the DSD/SST procedure has been applied to a
volved in performing simulations with the current model will variety of FSI problems. Preliminary DSD/SST simulations were
be presented, including the Ž nite element formulations, coupling successfully performed to simulate FSI behaviors for  ow prob-
methods, mesh moving methods, and implementation on parallel lems involving moving cylinders and aerofoils.13;14 Later, the ap-
supercomputers. proach was applied to simulate the FSI response of a  exible
The parallel computational technique presented here targets pipe to internal  ow15 and to two-phase FSI  ow problems in-
three-dimensional simulations of parachute FSI, with application cluding interior ballistics.16 Recently, the approach has been used
to airdrop performance and control phenomena in terminal descent. to predict the FSI response for the in ation of an axisymmetric
The technique is based on the deforming-spatial-domain/stabilized cable-membrane parachute structure,17 to predict the steady-state
space-time (DSD/SST) formulation7;8 of the time-dependent,three- descent characteristics for a ram-air parachute system,4 and to pre-
dimensional Navier– Stokes equations of incompressible  ows for dict steady-state characteristics for a fully in ated T-10 parachute
the  uid dynamics (FD) part. Turbulent features of the  ow are ac- system under controlled conditions (i.e., pinned at the payload
counted for by using a Smagorinsky turbulence model.9 A Ž nite and subjected to a uniform freestream  owŽ eld).3 Finally, a se-
ries of FSI simulations18 and concurrent wind-tunnel experiments19
have been performed for a set of cross-parachute models as
a Ž rst step toward validation of this parachute FSI simulation
Received 20 February 2000; revision received 25 October 2000; accepted capability.
for publication 10 November 2000. Copyright ° c 2001 by the American For the FSI problems presented, special attention is given to the
Institute of Aeronautics and Astronautics, Inc. All rights reserved. transfer of couplinginformationbetween “compatible”and “incom-
¤ Aerospace Engineer, Soldier Systems Center.
† patible” FD and SD interface meshes (i.e., parachute canopy sur-
Aerospace Engineer, Soldier Systems Center. Senior Member AIAA.
‡
James F. Barbour Professor in Engineering and Chairman, Mechanical face meshes). For compatible meshes the FD and SD have nodally
Engineering and Materials Science. equivalent interface meshes, and the transfer of coupling informa-
§ Professor, Department of Civil and Environmental Engineering. tion is straightforward. For incompatible meshes coupling infor-
¶
Professor, Department of Civil and Environmental Engineering. Member mation must be transferred through more complicated projection
AIAA. strategies.4;20
800
 STEIN ET AL. 801

Governing Equations cables and concentrated masses, and time-variant cable lengths (for
Fluid Dynamics control line pulls, reeŽ ng, etc.). The interested reader is directed to
Benney et al.11;23 for further details.
Let Ät ½ IR n sd and .0; T / be the spatial and temporal domains,
respectively, where n sd is the number of space dimensions, and let
0t denote the boundary of Ät . The subscript t implies the time Mesh-Moving Strategy
dependence of the spatial domain. The spatial and temporal coordi- To handle changes in the spatial domain caused by parachute
nates are denoted by x D .x; y; z/ and t 2 .0; T /. The Navier– Stokes canopy deformations,a suitable mesh-moving scheme is necessary.
equations of incompressible  ows are EfŽ cient special purpose mesh-moving algorithms can be designed
³ ´ for problems where parachute motion is somewhat predictable. For
@u example, Kalro and Tezduyar4 utilized an algebraic scheme to han-
½ C u ¢ ru C f r ¢¾ D0 on Ät 8t 2 .0; T / (1) dle the FSI-induced motions of a ram-air parachute system. A more
@t
general purpose mesh-moving scheme is needed for problems with
r ¢u D 0 on Ät 8t 2 .0; T / (2) arbitrary and complex deformations. We use an automatic mesh-
moving scheme for the problems addressed.In this scheme the  uid
Here ½, u, f, and ¾ are the density, velocity, external body force, mesh is treated as a linearly elastic pseudosolid that deforms as dic-
and stress tensor, respectively. For the problems under considera- tated by the motion of the surface boundaries of the  uid domain.24
tion, the  uid is assumed Newtonian, and the dynamic viscosity is This scheme introduces an additional computationalcost associated
modiŽ ed locally using a Smagorinsky turbulence model.9 Dirichlet with the mesh-moving equations, but is well suited for handling the
and Neumann-type boundary conditions are prescribed on .0t /g complex geometries and arbitrary motions for this class of prob-
and .0t /h , respectively, where .0t / g and .0t /h are complementary lems. The solution for the pseudosolid has no physical signiŽ cance
Downloaded by STANFORD UNIVERSITY on August 5, 2013 | http://arc.aiaa.org | DOI: 10.2514/2.2864

subsets of the boundary 0t . The initial condition on the velocity is in itself. Its sole purpose is to deform the FD mesh to handle, with
divergence free. minimal mesh distortion,motions and deformationsin mesh bound-
aries and interfaces.
Structural Dynamics
Let Äst ½ IR n xd be the spatial domain bounded by 0ts , where Fluid-Structure Coupling
n x d D 2 for membranes and n xd D 1 for cables. The boundary 0ts is The  uid-structure coupling occurs at the FSI interface, which is
composed of .0ts /g and .0ts /h . Here, the superscript s corresponds in this case the parachute canopy surface. We use an iterative cou-
to the cable-membrane structure. The equations of motion for the pling approach, with individual systems of equations being solved
structural system are for the  uid and the structure. Coupling is achieved through the
³ ´ transfer of FSI information between the  uid and structure within a
d2 y nonlinear iteration loop, with multiple nonlinear iterations improv-
½s fs r ¢¾s D 0 on Äst (3) ing the convergence of the coupled system. Displacements from
dt 2
the SD solution are treated as Dirichlet boundary conditions in the
mesh-moving scheme. Displacement rates from the SD solution are
where y, ½ s , f s , and ¾ s are the displacement, material density, ex-
treated as Dirichlet boundary conditions in the FD solver. In return,
ternal body forces, and Cauchy stress tensor, respectively. For the
parachute surface tractions from the  uid are used as distributed
problems under consideration, we assume large displacements and
forces in the SD solver. For the applications presented in this paper,
rotationsbut small strains. Thus, constitutiverelationshipsare based
we transfer only the pressure contribution from the FD solution to
on Hookean materials with linear-elastic properties.
the SD solver. Implementation of this iterative coupling is much
more straightforwardthan direct coupling approaches.The separate
Finite Element Formulations
treatment of the  uid and structural solvers allows us to use the two
Fluid Dynamics as subroutines in an FSI code. These reasons, along with the draw-
To handle the time-variant spatial domains encountered in backs of direct coupling approaches, have led most researchers in
parachute problems, we employ the DSD/SST Ž nite element this Ž eld to focus on iterative coupling approaches.3;4;25 27
formulation.7;8 In this formulation the Ž nite element interpolation FSI information can be passed between the FD and SD solvers
polynomials are functions of both space and time, and the stabilized using compatibleor incompatiblemeshes. Compatible meshes refer
variationalformulation of the problem is written over the associated to the cases where the  uid-structure interface is represented by a
space-time domain. This stabilized formulation automaticallytakes nodally equivalent FD surface mesh and SD canopy mesh. Transfer
into account deformations in the spatial domain and protects the of FSI information in this case is straightforward as a result of
computation against numerical oscillations. This method has been the one-to-one mapping between the interface nodes. Incompatible
applied to a large number of problems with moving boundaries meshes refer to cases where the interface is represented by a FD
and interfaces. The DSD/SST method used in this paper is based surfacemesh and a SD canopy mesh that are different.This approach
on the streamline-upwind/Petrov– Galerkin (SUPG)21 and pressure- has the beneŽ t of allowing for individual FD and SD meshes to be
stabilizing/Petrov– Galerkin (PSPG)22 stabilization techniques.The designed to take advantage of the strengthsof each solver. However,
SUPG method is one of the most widely used stabilization meth- incompatiblemeshes require a more complicatedprojectionscheme
ods. The PSPG formulation assures numerical stability while al- for transfer of information across the  uid-structure interface, such
lowing the use of equal-order interpolation functions for velocity as a least-squares projection.4;20 The least-squares projection, for
and pressure. These stabilizationtechniques also prevent numerical transfer of FSI information between incompatible meshes, can be
oscillationsand instabilitiesfor  ows involvinghigh Reynolds num- written as
bers and boundary layers, without introducing excessive numerical Z
dissipation. ±d ¢ .d f ds / d0 (4)
0INT
Structural Dynamics
A semidiscrete Ž nite element formulation for the SD equations where 0INT is the  uid-structure interface, d f and ds correspond
of motion is obtained from the principle of virtual work. Finite dis- to the shared variables between the  uid and the structure (i.e.,
placements of the structure are taken into account by using a total displacements, displacement rates, and tractions), and ±d is a test
Lagrangiandescriptionof the problem. In addition to membrane and function associated with d f or ds (depending on which direction
cable elements,a varietyof parachute-speciŽc featureshave been in- information is being transferred). For a parachute application, these
corporatedinto the SD solver to includetruss and concentrated-mass two approaches are depicted in Fig. 1 with the  uid mesh (center)
elements (representing the payload), drag force approximations for and compatible (left) and incompatible (right) SD meshes.
 802 STEIN ET AL.

Fig. 1 Parachute meshes: compatible (left) vs incompatible (right).

Implementation
The FD and SD solvers are implemented using a message-passing
paradigm and has been ported to a variety of architectures. Com-
putations for the examples presented are carried out on a CRAY Fig. 2 Compatible and incompatible T-10 SD meshes.
T3E-1200. The interested reader is directed to Kalro and Tezduyar4
for further details of the parallel implementation.
Downloaded by STANFORD UNIVERSITY on August 5, 2013 | http://arc.aiaa.org | DOI: 10.2514/2.2864

Numerical Examples: Three-Dimensional FSI

for a T-10 Parachute System
The Army’s T-10 personnel parachute system is a “ at extended
skirt canopy” composed of a 35-ft (10.67 m) diam Dc canopy and 30
suspension lines each 29.4 ft (8.96 m) long. The canopy is called a
 at extendedskirt canopy because in its constructed(or nonstressed)
conŽ guration it is composed of a main circular section with a cir-
cular vent at the apex and an inverted  at ring section, which lies
under the main section and is connected to the main section at the
outer radius. The lines connect to two con uence points (which ap-
proximate the connection points for a personnel harness assembly).
The suspension lines continue as 30 gore-to-gore reinforcements
through the parachute canopy and meet at the apex. For the T-10 the
vent diameter is 0:1Dc , and the width of the skirt is also 0:1Dc .
In the following sections we describe the three-dimensional FSI
strategy for numerical examples involving the T-10 parachute sys-
tem. First, we will compare the numerical results corresponding to
compatible and incompatible meshes. Second, we will demonstrate
the simulation technique for a T-10 parachute/payload system dur-
ing terminal descent. Finally, we will demonstrate the capability to
perform “line pulls” during an FSI simulation.

Compatible vs Incompatible Meshes

SD Problem Setup
The simulations involve two SD models for the T-10. The Ž rst SD
mesh (compatiblecase) consists of 9,183 nodes; 17,490 three-noded
membrane elements for the canopy surface; and 1,920 two-noded Fig. 3 In ated T-10: SD geometry and stresses.
cable elements for the suspension lines and canopy reinforcements.
This SD mesh results in 27,543 equations. FD Problem Setup
The second SD mesh consists of 3,573 nodes; 780 nine-noded A three-dimensional mesh with tetrahedral elements was gener-
(i.e., biquadratic) membrane elements for the canopy surface; and ated for the FD solution using as the surface mesh the three-noded
1,170 two-noded cable elementsfor the suspensionlines and canopy membrane mesh for the T-10 canopy in its in ated conŽ guration.
reinforcements. This SD mesh results in 10,713 equations. The Canopy surface nodes were multiply deŽ ned, with one node for both
parachute system is represented by linearly elastic materials, with the upper and lower surfaces. The mesh consists of 133,097 nodes
thickness and material properties representative of a T-10. Figure 2 and 783,910 tetrahedral elements. The  ow simulations were car-
shows “blown-out” views for the compatible and incompatible SD ried out at a Reynold’s number of 5 £ 106 [which approximates the
meshes for the main canopy reinforcements (cables), the main terminal descent velocity of 22.0 ft/s (6.7 m/s)]. Here, the Reynold’s
canopy (membranes), the extended skirt (membranes), and the number is deŽ ned based on the constructed diameter of the canopy
extended skirt reinforcements and suspension lines (cables). Dc . For this case the boundary conditions are deŽ ned as follows:
The unstructured mesh is allowed to in ate when the canopy is 1) The parachute canopy surface is treated as a zero-porosity
subjectedto a prescribeddifferentialpressureof 0.5 lb/ft2 (23.94 Pa). material and is assigned no-slip conditions.
For this stand-alone structural simulation every node in the SD 2) The in ow boundary below the parachute is assigned a
mesh is unconstrained, with the exception of the two con uence prescribed velocity condition of 22.0 ft/s (6.7 m/s).
point nodes, which are pinned. The fully in ated static conŽ gu- 3) The side boundaries are assigned free-slip conditions.
ration for the T-10 model obtained under the prescribed pressure 4) The out ow boundaryabovethe parachuteis assignedtraction-
loading is shown in Fig. 3. Maximum principal stresses for the free conditions.
parachute canopy (membrane) are superimposed on the surface, Initial unsteady  ow solutions were obtained for the Ž xed canopy
with low stresses along the canopy radial reinforcements and high conŽ guration using a stabilized semidiscrete formulation.22 The
stresses in the canopy midgore regions. semidiscrete formulation, which is less cost intensive than the
 STEIN ET AL. 803

DSD/SST formulation, is adequate for the stand-alone simulations (Ref. 28) and include the payload drag, suspension line drag, and
because there is no time dependence in the spatial domain (i.e., no FSI effects.
deformations of the canopy). Figure 4 shows a “snapshot” in time
of the computed velocities and pressures for the  ow about the T-10 FSI Simulation
canopy. After this  ow is developed, several time steps were com- FSI simulations were conducted for both the compatible and in-
puted, still with the Ž xed canopy but by using the DSD/SST proce- compatible SD meshes. The FSI simulations were initiated using
dure, to obtain the starting FD conditions for the FSI simulation.For the fully in ated static conŽ guration for the compatible mesh as the
the DSD/SST procedurethe FD system results in 958,686 equations. initial condition for the SD model and the fully developed  owŽ eld
The computed drag coefŽ cient for the stand-alone FD simulation, about the Ž xed conŽ guration as the initial condition for the  ow-
based on the total constructedarea of the canopy, was approximately Ž eld. All SD nodes were prescribed to have no initial velocities or
0:72. This is without accounting for the suspension line or payload accelerations. The two payload nodes in the SD model were fully
drag effects. The stand-alone simulation neglects FSI effects and constrained. To begin the two FSI simulations with the same initial
thus is for a different canopy geometry than the expected FSI geom- conditions, static equilibrium displacements from the compatible
etry. With the stated approximations in the stand-alone simulation, SD mesh were projected onto the incompatible SD mesh with a
the computed drag compares quite acceptibly with the experimental least-squares method.
values for the T-10 parachute system which range from 0.78 to 0.87 Coupled simulations were carried out with a nondimensional
time-step size of 0:005, which equates to a dimensional time step
of 0:0032 s. The aerodynamic drag force acting on the canopy was
calculated at each time step. Figure 5 shows the time histories for
the nondimensional drag force. Good comparison is seen between
the compatible and incompatiblemesh simulations,with the incom-
Downloaded by STANFORD UNIVERSITY on August 5, 2013 | http://arc.aiaa.org | DOI: 10.2514/2.2864

patible mesh capturing higher-order modes of oscillation because

of the higher-order biquadratic elements in the SD model. These
oscillations are caused part by the snap through of individual gores
as the parachute settles during the initial stages of the FSI simu-
lation. Here, snap through refers to a local change in concavity of
the canopy surface, which is resisted by compressive forces. For
the problem presented there is no membrane wrinkling model im-
plemented, and the triangular elements are unable to snap through
for the given mesh resolution, whereas the higher-ordermembranes
can experience the snap through. Realistic wrinkling models are be-
ing developed,29 which effectively eliminate structural compressive
stresses.

Freefalling T-10 Parachute System

SD Problem Setup
For this example, we modify the T-10 SD model to include a
payload and a set of four risers connecting the payload to the sus-
pension lines. We use the incompatible mesh from the preceding
example as the base mesh for the parachute system. Each riser is
represented by Ž ve two-noded cable elements. The front two ris-
Fig. 4 In ated T-10: Initial FD mesh and  owŽ eld. ers each attach to seven suspension lines, and the back risers each

Fig. 5 Drag histories for both the compatible and incompatible simulations.
 804 STEIN ET AL.

Table 1a T-10 parachute system: material properties FD Problem Setup

Cables The canopy surface unstructured mesh with triangular elements

was generated by Ž rst generatinga mesh for the  at canopy and then
Suspension Radial Trusses
projecting the displacementsfrom the SD equilibrium solution onto
Material group lines reinforcements Risers Payload
the  at mesh. This mesh is used to representthe T-10 canopyas an in-
Elements 300 870 20 6 terior surface in the FD mesh. A three-dimensionalmesh with tetra-
Area, ft2 0.0001 0.0001 0.0001 0.0001 hedral elements was generated, with 149,253 nodes and 888,344
Density, slugs/ft3 6.0 6.0 2.0 6.0 elements. The unsteady  ow solutions to be used as starting condi-
Young’s modulus, 4.32 £ 106 4.32 £ 106 2.16 £ 107 4.65 £ 109 tions for the FSI simulation were obtained for the Ž xed canopy con-
lb/ft2 Ž guration at Reynolds number of 5 £ 106 using a semidiscrete for-
mulation. This solution was used as the initial condition for the FSI
simulation, which were carried out using the DSD/SST procedure.
Table 1b T-10 parachute system: material properties FSI Simulation
Membranes Concentrated masses For the preceding example we simulated the FSI behavior for
the  ow about a T-10 parachute, where the payload was Ž xed in
Material group Canopy Payload masses
Elements 780 4
Thickness, ft 0.0001 ——
Density (mass) 6.0 (1.94)
Slugs/ft3 (slugs)
Downloaded by STANFORD UNIVERSITY on August 5, 2013 | http://arc.aiaa.org | DOI: 10.2514/2.2864

Young’s modulus, lb/ft2 2.0 £ 105 ——

Poisson ratio 0.3 ——

Fig. 7 FD mesh during FSI simulation.

Fig. 6 SD mesh for T-10 with risers and payload.

attach to eight suspension lines. Additionally, each front and back

riser attaches to the payload. The payload is represented by a stiff
tetrahedral truss system and four concentrated mass elements. The
six two-noded truss elements give the payload structure rigidity, and
the four concentratedmasses approximatethe mass of a typicalpara-
trooper. This payload modeling technique has been demonstrated
for three-dimensionalstand-aloneSD simulations.23 The blown-out
depiction of the SD mesh for the T-10 parachute system with risers
and payload is shown in Fig. 6.
The SD model is broken into six distinct material groups: one
membrane group, three cable groups, a truss group, and a con-
centrated mass group. The membrane group deŽ nes the parachute
canopy. We have distinct cable groups for the suspension lines, the
canopy radial reinforcements, and the risers. The truss and con-
centrated mass groups deŽ ne the payload. The deŽ nitions for the
different material groups are given in Tables 1a and 1b.
A stand-alone damped dynamic simulation was conducted for
the T-10 parachute model to in ate the canopy under a prescribed
differential pressure of 0.5 lb/ft2 (23.94 Pa). For the stand-alone
simulation the four payload node points were Ž xed in space. This
equilibriumsolution is used as the initial conditionfor the SD solver
in the subsequent FSI simulation. Fig. 8 SD mesh during FSI simulation.
 STEIN ET AL. 805
Downloaded by STANFORD UNIVERSITY on August 5, 2013 | http://arc.aiaa.org | DOI: 10.2514/2.2864

Fig. 9 Drag force.

Fig. 10 Net vertical riser force.

space. For this case, where the parachute is allowed to freefall, the displacements(i.e., the FD mesh moves globally with the parachute
parachute structure will fall (or rise) globally at some velocity rela- canopy).
tive to the in ow velocity. Thus, the predicted terminal descent for 3) Interior mesh points are updated based on the pseudosolid
the parachute system will be the terminal SD velocity plus the in- automatic mesh-moving strategy.
 ow velocity. This combined Lagrangian– Eulerian description for The FSI simulation is initiated using the equilibrium solution for
the problem was selected to simplify the setup for the initial condi- the SD solver and the initial condition generated for the FD solver.
tions of the FSI problem. At the onset of the simulation, the payload nodes in the SD model
To handle the freefalling parachute, we implement our automatic are no longer Ž xed in space, and the parachute SD model is fully
mesh-moving scheme as follows: unconstrained. Motion of the structure is driven by the external
1) The parachute canopy surface is assigned a prescribed dis- forces (i.e., gravity, line drag, payload drag), the internal forces, and
placement condition, with the prescribed displacements coming the FD-induced pressures on the canopy. Likewise, the FD solu-
from the SD solution. tion is driven by the prescribed in ow condition and the structural
2) The outer mesh boundaries are assigned prescribed displace- displacements and displacement rates on the canopy surface. The
ment conditions,with displacementsequal to the averageSD canopy initial condition for the FSI simulation is not in a coupled  uid and
 806 STEIN ET AL.
Downloaded by STANFORD UNIVERSITY on August 5, 2013 | http://arc.aiaa.org | DOI: 10.2514/2.2864

Fig. 11 Soft-landing drag force.

Fig. 12 Soft-landing net vertical riser force.

structural equilibrium state. Therefore, the parachute structure ex- directions. In addition, the Ž gures show the vertical position of the
periences a large amount of settling during the initial stages of the payload.
FSI simulation. Figures 7 and 8 show the behavior of the FD and Figure 9 shows the time history of the aerodynamic drag force
SD meshes at four instants during the FSI simulation, with 0:63 s acting on the parachute canopy. The dashed line corresponds to the
of real time between each instant. In Fig. 7 we show, for the FD total gravitationalforce acting on the parachutesystem (i.e., canopy,
mesh, the deforming canopy surface and a section in a Ž xed cutting suspensionlines, risers,and payload weights). As expected,the drag
plane. The combined Lagrangian– Eulerian reference frame for the force oscillates about the weight of the parachute system. Figure 10
FSI simulation is evident by the fact that the canopy rises relative shows the time history of the net vertical tension forces that the risers
to the Ž xed cutting plane in this sequence of pictures. In Fig. 8 we exert on the payload. The force contributionscaused by the payload
show the deforming T-10 structuralmodel for the same four instants. drag approximations are very small (i.e., less than one pound), and,
Severe deformations in the parachute canopy and suspension lines therefore, the net vertical riser force effectively balances the 250-lb
are clearly seen. Also evident are motions of the payload in lateral (1112 N) payload, as seen in Fig. 10.
 STEIN ET AL. 807
Downloaded by STANFORD UNIVERSITY on August 5, 2013 | http://arc.aiaa.org | DOI: 10.2514/2.2864

Fig. 13 Soft-landing payload velocity.

T-10 Control Line Pull Acknowledgments

Being able to represent time-variant cable lengths during an FSI This work was sponsored in part by U.S. Air Force of Sci-
simulation is necessary to represent several important parachute entiŽ c Research (Contract F49620-98-1-0214), by NASA-JSC
operations. These operations include control line “pulls” and “re- (Grant NAG9-1059) and by the Army High Performance Comput-
leases” in parachute maneuvering, riser pulls during soft landing, ing Research Center under the auspices of the U.S. Department
control line pulls for maneuvers and  ared landings of ram-air of the Army, Army Research Laboratory cooperative agreement
parachutes, and dis-reeŽ ng. Various control line operations have DAAH04-95-2-0003 and Contract DAAH04-95-C-0008. The con-
been reported for SD stand-alone simulations.23 tent does not necessarily re ect the position or the policy of the
A “soft landing” was simulated at the end of the T-10 FSI simula- government, and no ofŽ cial endorsement should be inferred.
tion presented in the preceding section. This was accomplished by
smoothlydecreasingin time the naturallengthsof the cable elements
References
1
Peterson, C. W., Strickland, J. H., and Higuchi, H., “The Fluid Dynamics
deŽ ning the four risers. During the simulation, over a 0:5-s interval, of Parachute In ation,” Annual Review of Fluid Mechanics, Vol. 28, 1996,
the risers were each decreased from 3:0 ft (0.91 m) to approximately pp. 361– 387.
2:0 ft (0.61 m). Figure 11 shows the computed drag throughout the 2 Benney, R. J., and Stein, K. R., “A Computational Fluid Structure Inter-
simulation with the line pull starting between 3:5 and 4:5 s. As ex- action Model for Parachute In ation,” Journal of Aircraft, Vol. 33, No. 4,
pected,higherdrag is predictedfor the line-pullcase. However, there 1996, pp. 730– 736.
3 Stein, K., Benney, R., Kalro, V., Tezduyar, T., Leonard, J., and Accorsi,
is a delayed response in the aerodynamic drag on the canopy caused
by the line pull, which results from the elasticityin the the parachute M., “Parachute Fluid-Structure Interactions: 3-D Computation,” Computer
structure and the structural response response to the line pull. Methods in Applied Mechanics and Engineering, Vol. 190, 2000, pp. 373–
386.
Figure 12 shows the net vertical force in the four risers prior to 4
Kalro, V., and Tezduyar, T., “A Parallel Finite Element Methodology
and throughout the soft landing simulation. Here, the response to for 3D Computation of Fluid-Structure Interactions in Airdrop Systems,”
the line pull is immediate, as expected. Figure 13 shows the vertical Proceedings of the 4th Japan-US Symposium on Finite Element Methods in
payload velocitythroughoutthe simulation with and without the line Large-Scale Computational Fluid Dynamics, Funabashi, Japan, 1998.
pull. The reduction in the vertical payload velocity caused by the 5 Ibos, C., Lacroix, C., Goy, A., and Bordenave, P., “Fluid-Structure Sim-

soft landing is evident in this Ž gure. Additional simulations can be ulation of 3D Ram Air Parachute with Sinpa Software,” Proceedings of the
carried out to provide further information on possible soft landings Confederation of European Aerospace Societies/AIAA 15th Aerodynamic
and for line pulls of different magnitudes and durations. Decelerator Systems Technology Conference, AIAA 99-1714, AIAA, Re-
ston, VA, 1999, pp. 88– 98.
6 Mosseev, Y., “The Decelerator Pitch-Dependent Performances Predic-

tion Based on 3D Aeroelastic Analysis,” Proceedings of the Confederation

Conclusions
of European Aerospace Societies/AIAA 15th Aerodynamic Decelerator Sys-
Parachute operations involve highly deformable structures inter- tems Technology Conference, AIAA 99-1717, AIAA, Reston, VA, 1999, pp.
acting with complex  owŽ elds. Being able to predict parachute be- 129– 137.
havior requires modeling of highly nonlinear  uid-structure inter- 7 Tezduyar, T. E., Behr, M., and Liou, J., “A New Strategy for Finite

actions. The simulation capability for this complex application has Element Computations Involving Moving Boundaries and Interfaces—the
been attainableonly recently,with developmentof advancedcompu- Deforming-Spatial-Domain/Space-Time Procedure: I. The Concept and the
tational methods and availability of powerful computers. In this pa- Preliminary Tests,” Computer Methods in Applied Mechanics and Engineer-
ing, Vol. 94, 1992, pp. 339– 351.
per we presenteda method for carryingout simulationsfor parachute 8 Tezduyar, T. E., Behr, M., Mittal, S., and Liou, J., “A New Strategy for Fi-
 uid-structure interactions that can be applied to a broad range of nite Element Computations Involving Moving Boundaries and Interfaces—
parachute systems. The method includes the  uid and structural dy- the Deforming-Spatial-Domain/Space-Time Procedure: II. Computation of
namics models and the couplingstrategy developedfor this purpose. Free-Surface Flows, Two-Liquid Flows, and Flows with Drifting Cylinders,”
Several test applications for a T-10 parachute were also presented Computer Methods in Applied Mechanics and Engineering, Vol. 94, 1992,
to demonstrate this computational capability. pp. 353– 371.
 808 STEIN ET AL.

9 Smagorinsky, J., “General Circulation Experiments with the Primitive federation of European Aerospace Societies/AIAA 15th Aerodynamic Decel-
Equations,” Monthly Weather Review, Vol. 91, 1963, pp. 99– 165. erator Systems Technology Conference, AIAA 99-1737, AIAA, Reston, VA,
10 Benney, R. J., and Leonard, J. W., “A 3-D Finite Element Struc- 1999, pp. 269– 278.
tural Parachute Model,” Proceedings of the 13th AIAA Aerodynamic De- 20 Maman, N., and Farhat, C., “Matching Fluid and Structure Meshes for

celerator Systems Technology Conference, AIAA 95-1563, 1995, pp. 165– Aeroelastic Computations:A Parallel Approach,” Computers and Structures,
175. Vol. 54, 1995, pp. 779– 785.
11 Benney, R. J., Stein, K. R., Leonard, J. W., and Accorsi, M. L., “Cur- 21 Brooks, A. N., and Hughes, T. J. R., “Streamline Upwind/Petrov –

rent 3-D Structural Dynamic Finite Element Modeling Capabilities,” Pro- Galerkin Formulationsfor ConvectionDominated Flows with Particular Em-
ceedings of the 14th AIAA Aerodynamic Decelerator Systems Technology phasis on the Incompressible Navier– Stokes Equations,”Computer Methods
Conference, AIAA 97-1506, AIAA, Reston, VA, 1997, pp. 285– 303. in Applied Mechanics and Engineering, Vol. 32, 1982, pp. 199– 259.
12 22
Kalro, V., Aliabadi, S., Garrard, W., Tezduyar, T., Mittal, S., and Stein, Tezduyar, T. E., “Stabilized Finite Element Formulations for Incom-
K., “Parallel Finite Element Simulationof Large Ram-Air Parachutes,” Inter- pressible Flow Computations,” Advances in Applied Mechanics, Vol. 28,
national Journal for Numerical Methods in Fluids, Vol. 24, 1997, pp. 1353– 1991, pp. 1 – 44.
1369. 23 Benney, R., Stein, K., Zhang, W., Accorsi, M., and Leonard, J., “Con-
13 Mittal, S., and Tezduyar, T. E., “A Finite Element Study of Incompress- trollable Airdrop Simulations Utilizing a 3-D Structural Dynamics Model,”
ible Flows Past Oscillating Cylinders and Airfoils,” International Journal Proceedings of the Confederation of European Aerospace Societies/AIAA
for Numerical Methods in Fluids, Vol. 15, 1992, pp. 1073– 1118. 15th Aerodynamic Decelerator Systems Technology Conference, AIAA 99-
14 Mittal, S., and Tezduyar, T. E., “Massively Parallel Finite Element 1727, AIAA, Reston, VA, 1999, pp. 182– 192.
Computation of Incompressible Flows Involving Fluid-Body Interactions,” 24 Tezduyar, T. E., Behr, M., Mittal, S., and Johnson, A. A., “Compu-

Computer Methods in Applied Mechanics and Engineering, Vol. 112, 1994, tation of Unsteady Incompressible Flows with the Stabilized Finite Ele-
pp. 253– 282. ment Methods-Space-Time Formulations, Iterative Strategies and Massively
15 Mittal, S., and Tezduyar, T. E., “Parallel Finite Element Simulation Parallel Implementations,” New Methods in Transient Analysis, edited by
of 3D Incompressible Flows—Fluid-Structure Interactions,” International P. Smolinski, W. K. Liu, G. Hulbert, and K. Tamma, AMD-Vol. 143, Amer-
Downloaded by STANFORD UNIVERSITY on August 5, 2013 | http://arc.aiaa.org | DOI: 10.2514/2.2864

Journal for Numerical Methods in Fluids, Vol. 21, 1995, pp. 933– 953. ican Society of Mechanical Engineers, New York, 1992, pp. 7– 24.
16 Wren, G., Ray, S., Aliabadi, S., and Tezduyar, T., “Simulation of Flow 25 Farhat, C., Lesoinne, M., and Maman, N., “Mixed Explicit/Implicit

Problems with Moving Mechanical Components, Fluid-Structure Interac- Time Integration of Coupled Aeroelastic Problems: Three-Field Formula-
tions and Two-Fluid Interfaces,” International Journal for Numerical Meth- tion, Geometric Conservation,and Distributed Solution,”InternationalJour-
ods in Fluids, Vol. 24, 1997, pp. 1433– 1448. nal for Numerical Methods in Fluids, Vol. 21, 1995, pp. 807– 835.
17 Stein, K. R., Benney, R. J., Kalro, V., Johnson, A. A., and Tezduyar, 26 Farhat, C., and Lesoinne, M., “Higher-Order Staggered and Subiteration

T. E., “Parallel Computation of Parachute Fluid-StructureInteractions,” Pro- Free Algorithms for Coupled Dynamic Aeroelasticity Problems,” Proceed-
ceedings of the 14th AIAA Aerodynamic Decelerator Systems Technology ings of the 36th AIAA Aerospace Sciences Meeting, AIAA, Reston, VA, 1998.
Conference, AIAA 97-1505, AIAA, Reston, VA, 1997, pp. 277– 284. 27 L öhner, R., Yang, C., Cebral, J., Baum, J. D., Luo, H., Pelessone, D.,
18 Stein, K., Benney, R., Kalro, V., Tezduyar, T., Bretl, T., and Potvin, and Charman, C., “Fluid-Structure Interaction Using a Loose Coupling Al-
J., “Fluid-Structure Interaction Simulations of a Cross Parachute: Compar- gorithm and Adaptive Unstructured Grids,” Proceedings of the 36th AIAA
isons of Numerical Predictions with Wind Tunnel Data,” Proceedings of the Aerospace Sciences Meeting, AIAA, Washington, DC, 1995.
Confederation of European Aerospace Societies/AIAA 15th Aerodynamic 28 Ewing, E. G., Bixby, H. W., and Knacke, T. W., “Recovery Systems

Decelerator Systems Technology Conference, AIAA 99-1725, AIAA, Re- Design Guide,” U.S. Air Force Rept. AFFDL-TR-78-151, 1978.
ston, VA, 1999, pp. 172– 181. 29 Accorsi, M., Lu, K., Leonard, J., Benney, R., and Stein, K., “Issues
19 Brocato, B., Esteve, L., Garcia, D., Manglano, C., Peek, G., Potvin, in Parachute Structural Modeling: Damping and Wrinkling,” Proceedings
J., Benney, R., Stein, K., Alamat, R., and Little, J., “Experimental Study of the Confederation of European Aerospace Societies/AIAA 15th Aerody-
of Fluid-Structure Interactions on a Cross Parachute: Comparison of Wind namic Decelerator Systems Technology Conference, AIAA 99-1729, AIAA,
Tunnel Data and Drop Data with CFD Predictions,” Proceedings of the Con- Reston, VA, 1999, pp. 193– 204.