Article

2018, Vol. 47(6) 1038–1059

! The Author(s) 2016

Numerical study of Reprints and permissions:

sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/1528083716661203
parachute inflation journals.sagepub.com/home/jit

process based on
smoothed particle
hydrodynamics
fluid structure
interaction method

Han Cheng, Chen Chen, Xiaohan Liu, Meng Li,

Zhichao Chen and Lin Wei

Abstract
Shortcomings in large deformation calculations by mesh-based numerical methods have
been overcome by smoothed particle hydrodynamics method, which is a new meshless
algorithm based on Lagrange description and has been widely used in simulations as
blasting and impacting, but not been applied in research of engineered fabrics yet. In this
paper, a fluid structure interaction method coupling smoothed particle hydrodynamics
and finite element was proposed, by which the inflating process of a model parachute
was investigated. In the modeling of parachute, the same nodes were shared by beam
elements of reinforced belts and adjacent canopy elements to simulate the elastic con-
straints, while the parachute meshes model was adjusted to satisfy requirement of fluid
structure interaction calculation by loading internal pressure, and surrounding flow field
was described by smoothed particle hydrodynamics particles. Then, the fluid structure
interaction calculating could be realized by contacting algorithm between particles and
mesh nodes. The dynamic processes of expanding structure and flow field were
obtained by this method. According to the analysis of numerical results, the parachute
inflating process could be divided into pre-inflating stage, fully inflating stage and inflated
stage; moreover, the noise occurred in wind tunnel experiment could be explained by
this method. The ‘‘breathing’’ phenomenon and top collapsing of canopy appeared in

Aviation Engineering Institute, Civil Aviation Flight University of China, Guanghan, China

Corresponding author:
Han Cheng, Aviation Engineering Institute, Civil Aviation Flight University of China, Guanghan 618307, China.
Email: [email protected]
Cheng et al. 1039

numerical results, as corresponded to the tunnel experiment. This new method could
be a good supplement in parachute design and research.

Keywords
Inflatable fabrics, meshless method, engineering fabrics, fluid structure interaction,
nonlinear dynamics

Introduction
Parachutes are widely used in aviation lifesaving, aerospace recovery and other
fields because of their light weight, easy folding and high reliability. Parachute
inflating process is always a hot topic in aviation life support research, as
whether the parachute could inflate and deploy with preset instructions will
directly affect the safety of personnel or system. The investigation approaches
are mainly divided into experimental and numerical, in general. The experi-
ments, such as airdropping and wind tunnel test [1], cost huge manpower,
material and financial resources, while the data are hard to collect and mech-
anism of expanding failure is difficult to reveal. The numerical analysis with
advantages of economy and repeatability has become an important approach in
parachute design and research.
Many different models or algorithms have been proposed for inflating investi-
gation [2–10]. Purvis [2] used simplified canopy structure and flow field model to
simulate a two-dimensional parachute inflating. Tezduyar et al. [3] and Stein et al.
[4] realized parachute opening by deforming spatial domain/stabilized space time)
method. Immersed boundary method was used by Kim and Peskin [5] to calculate
single parachute and parachute cluster inflating, but only in low Reynolds number.
Tutt and coworkers [6, 7] and Cheng et al. [8] investigated the parachute opening
process by ALE (Arbitrary Eulerian–Lagrangian) method. Takizawa et al. [9, 10]
applied SSTFSI (stabilized space time fluid structure interaction) method in work
phase calculation of fully expanded parachute. These methods or models provide
some useful references for new investigation. However, it could be found clearly
that most works were based on weak coupling technology, as the coupling infor-
mation could not exchange at one time step, and the inertia forces of fluid were
ignored in structure calculation. The technology of weak coupling was difficult to
calculate the complete process of inflating. Therefore, a good pre-inflated shape
was used in most works to avoid errors caused by weak coupling. Although the
ALE method could avoid this, a large part of calculation time would be consumed
on distortion meshes processing and flow field information updating. Besides, all
previous methods were carried on the mesh-based model, which could only
describe continuous medium as air and water, but not the medium discrete in
macroscopic and continuous in mesoscopic, such as rain, hail and snow.
Therefore, the parachute inflating under crucial environment could not be calcu-
lated by previous methods.
1040 Journal of Industrial Textiles 47(6)

The SPH (smoothed particle hydrodynamics) method as one of meshless numerical

methods has advantages that traditional mesh-based models do not have. The adap-
tion of SPH approximation method prevents the randomness of particles distribution
from affecting the formula, which could be used to deal some large deformation
problems. With continuous improvement and modification, the accuracy and stability
of SPH have satisfied requirements of engineering application [11]. This meshless
method with wide application range could be applied in different scales of discrete
or continuous systems, such as fluid motion, large deformation and impact problems
[12–15], as it has become an important approach in numerical investigation.
The parachute inflating is a typical FSI problem with material and geometric
nonlinearity, as all expanding process of inflatable fabric. Under Lagrangian
description, the air has large displacement and deformation during interaction,
which SPH method was good at solving. Therefore, the canopy was described by
finite elements (FEs) and air flow by SPH particles in this paper. As both the FEs
and particles were based on Lagrangian description, the FSI problem could be
transformed into a contact problem of different materials. In this paper, the inflat-
ing process of a model parachute was calculated by the method above. The
dynamic processes of structure and flow field were obtained.

Mathematical model
Fundamental theory of SPH
In the SPH method, the approximation f ðxÞ of x was expressed by its neighbored
particles in a domain
Z
f ðxÞ ¼ f ðx0 Þðx  x0 Þdx0 ð1Þ

where ðx  x0 Þ is Dirac function, is integral domain, x is particle’s coordinates

and x0 is neighbored particle’s.
And the kernel weight function Wðx  x0 , hÞ was used to replace the Dirac func-
tion in the equation (1)
Z
 
f ðxÞ ¼ f ðx0 ÞWðx  x0 , hÞdx0 ð2Þ

where h is smoothing length. And the kernel weight function should be under the
following conditions
8R
>
> Wðx  x0 , hÞdx0 ¼ 1
>
>
< lim Wðx  x0 , hÞ ¼ ðx  x0 Þ
h!0 ð3Þ
>
> Wðx  x0 , hÞ 4 0 in
>
>
:
Wðx  x0 , hÞ ¼0 outside
Cheng et al. 1041

There are lots of kernel weight functions satisfying conditions (3), among which
the function based on cubic B-spline proposed by Monaghan is the most widely
used one [11] that was used to construct the governing equations here. The three-
dimensional expression is as follows
82 2 1 3
3 < 3  R þ 2R 0  R 5 1
>
WðR, hÞ ¼  16ð2  RÞ3 1  R52 ð4Þ
2h3 >
:
0 R2

where R ¼ kx  x0 k=h.
Therefore, the integral approximation (equation (2)) could be transformed to
the sum of all the particles in support domain, and the discrete form is as follows

  X N
mj
f ðxi Þ ¼ f ðxj ÞWij ð5Þ

j¼1 j

where N is the total number of particles in support domain, mj is mass of particle

and j is density of particle, Wij ¼ Wðxi  xj , hÞ.
Therefore, the derivative function of equation (5) was given by

  X N
mj
r  f ðxi Þ ¼ f ðxj Þ  ri Wij ð6Þ

j¼1 j

 
and rij ¼ xi  xj .
xi xj @Wij
where ri Wij ¼ rij @rij

Governing and discretization equations

The governing functions could be expressed by equations (7) to (9). The Greek
alphabet  and  denote directions. The variable with superscript  or  is vector
while the one with  is tensor, which is different from the expression of mesh-
based method

d @v
¼   ð7Þ
dt @x

dv 1 @r
¼ þ Fext ð8Þ
dt  @x

de r @v
¼ ð9Þ
dt  @x
1042 Journal of Industrial Textiles 47(6)

where Fext is external force of unit mass, r is total stress tensor which consists of
pressure p and viscous shearing stress s

r ¼ p þ s ð10Þ

In the Newtonian fluid, the viscous shearing stress s is directly proportional to
the shearing strain rate e , and the viscosity coefficient  is the ratio

 
@v @v 2
s ¼ e ¼  þ  ðr  vÞ ð11Þ
@x @x 3

Based on equations (5) and (6), the expressions of governing equations used in
SPH algorithm were

di X N
@Wij
¼ mj ðvi  vj Þ  ð12Þ
dt j¼1 @xi
! !
dvi XN
pi pj @Wij X N
i e j e
j @Wij
¼ mj 2 þ 2 þ ij  þ mj 2
i
þ 2 
þ Fiext ð13Þ
dt j¼1
i j @xi j¼1
 i  j @xi

!
dei 1 X
N
pi pj @Wij i  
¼ mj 2 þ 2 þ ij ðvi  vj Þ  þ e e ð14Þ
dt 2 j¼1 i j @xi 2i i j

Here, e
i (the shearing strain rate of the particle i) and ij (artificial viscosity
coefficient) were respectively given by

!
X
N
mj @W ij
XN
m j @W ij 2 XN
m j
e
i ¼ v þ v  vj  ri Wij  ð15Þ
j¼1
j j @xi  j @xi
j¼1 j
3 j¼1 j

8 2
<  cij ij þ  ij
ij ¼ ðvi  vj Þ  ðxi  xj Þ 5 0 ð16Þ
: ij
0 ðvi  vj Þ  ðxi  xj Þ  0

where,  and  are standard constants (the values are about 1.0). The other
parameters are expressed by

hij ðvi  vj Þ  ðxi  xj Þ

ij ¼ ð17Þ
r2ij þ ’2
Cheng et al. 1043

1
cij ¼ ðci  cj Þ ð18Þ
2
1
ij ¼ ði  j Þ ð19Þ
2
1
hij ¼ ðhi  hj Þ ð20Þ
2
where c is acoustic velocity, ’ ¼ 0:1hij .

Coupled SPH particles with FE nodes

FE model was applied by most researchers to describe inflatable fabrics, as it not
only can reflect the mechanical properties but also the dynamic simulation can be
achieved. In addition, this algorithm with high calculates precision can be inte-
grated with other system model [16]. Therefore, the FE model was used to simulate
parachute in this paper.
With Lagrangian description of fabrics and air, the FSI problem here can be
treated as a contact problem between two materials. The penalty function algo-
rithm was used to realize the coupling, as widely applied in many well-known
explicit dynamics codes since this method was easy to program and accurate in
momentum conservation.
The SPH particles were defined as slave nodes, while the nodes of canopy elem-
ents were as master ones (Figure 1).

Figure 1. The projection relationship between canopy elements and SPH particles (here,ns is
slave node, ms is master node, C is contact point, Si are the elements that may be occurred
contact with SPH particle,ci and ciþ1 are the edge vectors of element S i at ms , g is the vector
from mS to nS and s is the projection vector of g on Si ).
1044 Journal of Industrial Textiles 47(6)

In each time step, the nearest master node ms of each slave node ns was searched.
The element contacted with slave node ns was chosen by


ðci  sÞ  ðci  ciþ1 Þ 4 0
ð21Þ
ðci  sÞ  ðs  ciþ1 Þ 4 0

where s ¼ g  ðg  mÞm. m is outer normal unit vector of element Si and could be

obtained by

ci  ciþ1
m ¼   ð22Þ
ci  ciþ1 

If the slave node ns was just on the common edge of adjacent elements, the
projection vector s would be redefined as
 
ci
s ¼ max g  ð23Þ
j ci j

The spatial coordinates of contact point C was defined as

xc ð c , c Þ¼ xI NI ð c , cÞ ð24Þ

where NI is shape function.

Then, the reference element coordinates nc of contact point C was obtained from
the simultaneous equations (Figure 2)
( @xc
@ ð c, cÞ  ½xnS  xc ð c , c Þ ¼0
@xc
ð25Þ
@ ð c, cÞ  ½xnS  xc ð c , c Þ ¼0

The penetration of SPH particle through the canopy element was judged by

l ¼ ni  ½xnS  xc ð c , c Þ 5 0 ð26Þ

@xc @xc
ð c, cÞ  ð c, cÞ
where ni ¼ @ @ .
@xc @xc 
 @ ð c, c Þ  @ ð c c 
, Þ
The SPH particle has penetrated through the canopy element (if l 5 0). The
coupling force could be calculated by equation (27), and the coupling force was
Cheng et al. 1045

Figure 2. The position of contact point C under reference element coordinates.

applied on slave node ns and contact point C, respectively. Finally, the coupling
force fc was taken as the external force in equations (8) and (13)

fc ¼ lki ni ð27Þ

where ki ¼ fsi KLii Ai , fsi is contact stiffness ratio factor, and the value is 0.1, in general,
Ki and Ai are bulk modulus and surface area of shell element, respectively, Li is
maximum diagonal length of shell element.
The whole coupling calculation process is shown in Figure 3. Two methods were
carried separately at the same time. The SPH method was applied with Leapfrog
(LF) scheme [11] in time marching, and the FE method was applied with central
difference scheme [17].

Case study
A flat circular parachute without material permeability was used to investigate. The
flattened canopy structure is shown in Figure 4. The whole canopy was sewed by
eight single gores. The nominal area of this model parachute was 0.78 m2, and each
parachute line’s length was 0.75 m.

Material and constitutive model

The canopy was made in 49201 type polyamide grid silk (Figure 5), the lines and
reinforced belts were made in 3-250A flame-retardant polyamide rope and 10-200A
1046 Journal of Industrial Textiles 47(6)

Figure 3. The coupling calculation process.

SPH: smoothed particle hydrodynamics.

Figure 4. Flattened structure of model parachute.

type belt, respectively. The material parameters of fabrics and air are shown
in Table 1.
In this paper, the fabrics described by FEs were treated as continuum, and the
mechanical properties of yarn layer were homogenized. Therefore, the lines and
reinforced belts were described by one-dimensional linear elastic constitutive equa-
tion as follows

x ¼ E"x ð28Þ

where x is Cauchy stress, "x is strain and E is Young’s modulus.

The canopy presents small strain and large rotation in parachute expanding, and
a large deformation effect of structure is mainly resulted from the large rotation.
Cheng et al. 1047

Figure 5. The 49201 type polyamide grid silk.

Table 1. Material parameters.

Material properties of canopy Density of canopy (kg/m3) 533

Young’s modulus of canopy (Pa) 4.3Eþ8
Thickness of canopy (m) 1E4
Material properties of line Density of line (kg/m3) 462
Young’s modulus of line (Pa) 9.7Eþ10
Diameter of line (m) 3E3
Material properties of reinforced belt Density of reinforced belt (kg/m3) 462
Young’s modulus of reinforced belt (Pa) 9.7Eþ10
Breadth of reinforced belt (m) 1E2
Properties of air Density of air (kg/m3) 1.18
Temperature of air ( ) 25
Velocity of air (m/s) 30
Ambient pressure (Pa) 1.01Eþ5

Therefore, the Saint Venant Kirchhoff constitutive equation was used to describe
the fabric, which had been widely used in parachute dynamic simulation [6–8]

S ¼ traceðEÞI þ 2E ð29Þ

1048 Journal of Industrial Textiles 47(6)

where S is Second Piola-Kirchhoff stress, and  are Lame coefficients and E is

Green strain.
Then, the Cauchy stress r used in governing equations above could be
obtained by

¼J1 F  S  FT ð30Þ

where F is deformation gradient and J is Jacobian determinant.

In addition, the canopy was discretized by three nodes shell elements, and the
fabric material anisotropy was neglected in this paper. As a result, the constitutive
equation (29) in Voigt matrix form was shown by
8 9 2 38 9
< S11 >
> = 1 0 < E11 >
> =
E 6
S22 ¼ 4 1 0 75 E22 ð31Þ
>
: >
; ð1 þ Þð1  2 Þ 12
>
: >
;
S12 0 0 2 E12

where is Poisson ratio.

In this paper, the velocity of air was less than Mach 0.3, which means the air
could be considered as incompressible material, and the stress description is divided
into a deviatoric part and a hydrostatic part. The constitutive equation of air is
shown in equation (32), which must be integrated with ideal gas state equation

dev hyd
ij ¼ ij þ ij ¼ ðvi,j þ vj:i Þ  pij ð32Þ

where  is dynamic viscosity coefficient and p is hydrostatic pressure.

Parachute model
The textile modeling was key and difficulty point in this paper, which was divided
into three steps in this paper.

The canopy modeling: First, the canopy’s folded geometry model was established in
sequence of points, lines and surfaces according to actual sizes (Figure 6(a)), on
which, the mesh model was based. In general, the accuracy of the mesh model was
higher based on quadrilateral elements than on triangular elements, and the quad-
rilateral elements were suitable to describe the model with a good pre-inflated shape
and befitting in numerical calculation with implicit algorithm [2–5, 9, 10]. However,
the triangular elements had better adaptability and were more suitable in solving of
the large displacement and deformation problems [8]. In Figure 6, Model A was
meshed by triangular elements (4656) and Model B by quadrilateral elements
(4640). It was found that the Jacobian ratio of all triangular elements was greater
than 0.8 in Model A, while only 77% elements’ Jacobian ratio was acceptable
(greater than 0.8) in Model B. The elements of poor quality concentrated mainly
Cheng et al. 1049

Figure 6. The geometry model and mesh model: (a) geometry model; (b) Model A and (c)
Model B. Here, the element which Jacobian ratio was less than 0.8 was highlighted by gray color.
The Jacobian ratio could characterize the difference between the element’s actual and ideal shape
and was usually used to evaluate the quality of meshes. The greater ratio, the better quality
elements have.

on the canopy top, of which the modeling quality was very difficult to be improved
by local trimming. The model with low Jacobian ratio was more likely to cause
calculation instability and extend calculation time. Therefore, the Model A was
used in this work.

The reinforced belt and lines modeling: In this paper, the lines and reinforced belts were
meshed by beam elements (616). In order to simulate the elastic constraints gen-
erated from these belts accurately, each beam element shared the same nodes with
adjacent triangular elements (Figure 7). This sharing ensured the canopy and rein-
forced belts have same deformation at same position. In fact the Young’s modulus
of reinforced belts was larger than the canopy’s (Table 1), and therefore, the
deformation of canopy was constrained by those belts in calculation. In
Figure 8, a change in small area could be found in the canopy area with reinforced
belts, due to the elastic constraints during inflating process. The engineering appli-
cation that reinforced belts are used to prevent canopy rupture with large deform-
ation was also verified.

Model adjustment: The SPH/FE coupling method did not need to establish compli-
cated, body fitted meshes or define interaction surfaces, which was same in ALE
method. The parachute elements and SPH particles were allowed to intersperse
with each other. But there must reserve a few SPH particles inside folded canopy
model to ensure the interaction with both sides of SPH particles was at the same
time (similarly, inside of folded canopy model should reserve a few of flow field
meshes in ALE method). Therefore, the folded parachute model obtained by the
first two steps must be adjusted (Figure 9).
1050 Journal of Industrial Textiles 47(6)

Figure 7. The reinforced belt modeling: (a) the actual reinforced belt and (b) the corresponding
model.

Figure 8. The total area changes of canopy during parachute inflating.

In order to solve this problem, Tutt et al. [7] proposed the FH (free hanging)
method. But that method could not guarantee the canopy was completely folded,
and the inflation time would be affected greatly by the model obtained by FH
method. In this paper, the Loading Pressure method was proposed. First, the
Cheng et al. 1051

Figure 9. The relative position relationship of parachute model before adjustment and SPH
particles (view from bottom).
SPH: smoothed particle hydrodynamics.

connection point of parachute lines was fixed. Then, a small amount of pressure
(1000 N) was applied on the inside surfaces of the canopy. The canopy deformation
process under the internal pressure was obtained by explicit calculation (Figure 10).
At 0.04 s, the parachute model could satisfy the requirement of coupling calculation
and be exported, and the explicit calculation was terminated (Figure 11).

Flow field model

Since the air was simulated by SPH particles, and the velocity boundary could not
be defined as finite volume method, the flow field was taken as a reference in this
paper, other than the parachute as a reference in wind tunnel experiment. The
relative motion of parachute was defined to simulate the parachute opening in
wind tunnel. SPH particles (1,458,000) were used to describe the air flow field.
The flow field boundary was defined symmetrical (Figure 12(a)). The velocity con-
dition was applied on the collection point of parachute lines. The parachute elem-
ents and SPH particles were allowed to intersperse with each other, as the whole
model was shown in Figure 12(b).

Wind tunnel experiment in Yu’s work

The corresponding wind tunnel experiment investigation was carried on in a low-
speed wind tunnel with the diameter of 2.5 m [18]. The experimental apparatus are
1052 Journal of Industrial Textiles 47(6)

Figure 10. The deformation process of parachute model.

Figure 11. The relative position relationship of parachute model after adjustment and SPH
particles (view from bottom).
SPH: smoothed particle hydrodynamics.
Cheng et al. 1053

Figure 12. The elements and SPH particles model: (a) air flow field and (b) the whole model.

shown in Figure 13. A high-speed camera was set in front of the wind tunnel, with
its image data were received and stored in computer, and an electric control system
was used to trigger the dowel and collect the data.
In this wind tunnel experiment, the canopy top and guiding parachute were
connected with a line (Figure 13(b)), while the lines connection point and para-
chute box were also connected. Then, the model parachute was folded according
to standard requirements and stored in the parachute box sealed by package
cloth.
When the velocity of blowing air from wind tunnel was stable, the experiment
countdown began. Once it started, the electric control system pulled out the dowel,
and then the model parachute would be pulled out by the guiding out of the
parachute box. Meanwhile the high-speed camera recorded images and sent them
to the computer.

Results and discussion

Analysis of numerical results
The changes of shape and velocity vector are shown in Figure 14.
1054 Journal of Industrial Textiles 47(6)

Figure 13. The wind tunnel experimental apparatus in Yu’s work [18]: (a) schematic figure and
(b) local magnification figure.

According to the changes of canopy shape and flow field, as shown in Figure 14,
the inflating process of model parachute could be divided into pre-inflating stage,
fully inflating stage and inflated stage.

The pre-inflating stage (0–45 ms): First, the canopy bottom began to expand, and
then the canopy top began to expand (Figure 14(a)) as the airflow entered into
the canopy. Meanwhile, the canopy bottom easily formed edge curling due to the
lack of effective constraints, which would exacerbate the instability of flow field.
The canopy top was fully expanded at 45 ms, which was marked as the end of pre-
inflating stage.

The fully inflating stage (45–55 ms): As there was no vent on canopy top, the fluid
inside flew out faster. During this stage, symmetric vortex was forming in external
flow field (Figure 14(b)), while the effective decelerating surface was forming with
the canopy expanding gradually from top to bottom. The inflating was very soon
Cheng et al. 1055

Figure 14. The dynamic change of shape and flow field over time in numerical results: (a)-(e)
represent 20 ms, 45 ms, 55 ms, 75 ms and 113 ms respectively (under the airflow velocity con-
dition of 30 m/s).
1056 Journal of Industrial Textiles 47(6)

due to the small parachute size. The fully inflating stage ended at 55 ms
(Figure 14(c)).

The inflated stage (after 55 ms): After the canopy was inflated, it continued to expand
due to inertial force. And over-expanding would be stopped quickly under the elastic
forces of the reinforced belts and lines, which also caused the collapse of canopy top.
Then, the canopy formed the largest aerodynamic shape when projection diameter
reached maximum value at 75 ms (Figure 14(d)). But the top collapse disappeared
soon due to airflow. Subsequently, the ‘‘breathing’’ phenomenon occurred, and a
stronger symmetric vortex in external flow field was found, which appeared strong
and weak alternately later (Figure 14(e)). Under the canopy and airflow interaction,
complicated changes of structure and the flow field were found near the canopy
bottom, where constant flapping and jittering due to lack of effective constraints
appeared, which explained the noise during wind tunnel experiment.

Comparison of numerical results and experimental results

In Yu’s work, the model parachute opening experiments at various velocities
(20 m/s, 30 m/s and 40 m/s) were carried out. The comparison of projection diam-
eter between numerical and experimental results is shown in Figure 15.

Figure 15. The comparison of projection diameter.

Cheng et al. 1057

Figure 16. The comparison of shape changes (under the airflow velocity condition of 30 m/s).

It was found that the greater the airflow velocity, the earlier the peak value of
projection diameter appeared. The results indicated that shorter parachute inflating
time and earlier peak value appearance could be obtained by increasing airflow
velocity from wind tunnel. Almost all of the results in Figure 15 showed that the
canopies would reach the maximum projection diameter only one time at the first
peak. Then, the projection diameter changed alternately with ‘‘breathing’’ phenom-
enon of canopies.
Longer inflating time was found in experiment than in simulation. The initial
inflation shape was considered as the main reason. An ideal and completely
symmetry model was used in calculation (Figure 10), while the canopy shape
was asymmetric before inflation in experiment. Besides, there was a deflection
angle between the airflow direction and canopy’s central axis due to gravity
when the canopy was pulled out of parachute box and began to inflate
(Figure 16). The gravity was neglected in calculation, and thus, there was no
deflection angle. Furthermore, mechanical vibration and turbulence intensity
resulted in the asymmetric of canopy shape when it began to expand in wind
tunnel. The error between the two results verified that the inflating time and alti-
tude loss could be reduced by a good pre-inflation canopy shape.
Figure 16 is the shape changes under the airflow velocity condition of 30 m/s. It
was found that the shape changes were basically the same. After the canopies were
inflated (the experimental: 245 ms; the numerical: 55 ms), top collapse and max-
imum projection diameter were found on both canopies (the experimental: 272 ms;
the numerical: 75 ms). Then, the canopies restored to inflated shape under the
elastic force of material. The comparison showed that the numerical method
used in this paper could be a good simulation for the dynamic process of the
parachute inflating.

Conclusions
As a meshless numerical algorithm, the SPH method has the advantage that the
traditional mesh-based methods have not, and a very wide prospect of application.
1058 Journal of Industrial Textiles 47(6)

In this paper, the SPH/FE coupling method was used to simulate the dynamic
process of a model parachute’s inflating. The conclusions are as follows:

1. The beam element shared the same nodes with adjacent elements, which could
simulate the elastic constraints from reinforced belts effectively. The numerical
results of different models also verified that the reinforced belts could effectively
control the canopy deformation.
2. The Loading Pressure method was proposed to adjust textile model, and this
new method could ensure the folding of adjusted model was completed and
would satisfy the requirement of FSI calculation.
3. The SPH/FE coupling method could exchange the coupling information imme-
diately at one time step by using contact algorithm, by which the fluid meshes
reconstruction and flow field information updating could be avoided. It is com-
pletely a new method to investigate the working process of parachute
numerically.
4. After inflated, the canopy would appear top collapse and reach the maximum
projection diameter for the first and only time. Then, ‘‘breathing’’ phenomenon
of canopy occurred while constantly flapping and jittering of canopy bottom
appeared, where effective constraints were lacking. That made the noise during
wind tunnel experiment clear.

Acknowledgements
The investigations described in this paper are a part of the research project no. F2015KF03
realized at Civil Aviation Flight Technology and Flight Safety Research Base.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, author-
ship, and/or publication of this article.

Funding
The author(s) received no financial support for the research, authorship, and/or publication
of this article.

References
[1] Ewing EG, Knacke TW and Bixby HW. Recovery systems design guide. Beijing: Aviation
Industry Press, 1988.
[2] Purvis JW. Theoretical analysis of parachute inflation including fluid kinetics. AIAA
Report 1981-1925, 1981. Reston, VA: American Institute of Aeronautics and
Astronautics Inc.
[3] Tezduyar TE, Behr M and Liou J. A new strategy for finite element computations
involving moving boundaries and interfaces: The deforming-spatial-domain/ space-
time Procedure: I. The concept and the preliminary numerical tests. Comput Meth
Appl Mech Eng 1992; 94: 339–351.
Cheng et al. 1059

[4] Stein K, Benney R and Kalro V. Parallel computation of parachute fluid-structure inter-
actions. AIAA Report 1997-1505, 1997. Reston, VA: American Institute of Aeronautics
and Astronautics Inc.
[5] Kim YS and Peskin CS. 3-D Parachute simulation by the immersed boundary method.
Comput Fluids 2009; 38: 1080–1090.
[6] Tutt BA and Taylor AP. The use of LS-DYNA to simulate the inflation of a parachute
canopy. AIAA Report 2005-1608, 2005. Reston, VA: American Institute of Aeronautics
and Astronautics Inc.
[7] Tutt BA, Roland S, Charles RD, et al. Finite mass simulation techniques in LS-DYNA.
AIAA Report 2011-2592, 2011. Reston, VA: American Institute of Aeronautics and
Astronautics Inc.
[8] Cheng H, Yu L, Rong W, et al. A numerical study of parachute inflation based on a
mixed method. Aviation 2005; 16: 115–123.
[9] Takizawa K, Wright S, Moorman C, et al. Fluid structure interaction modeling of
parachute clusters. Int J Numer Meth Fluid 2011; 65: 286–307.
[10] Takizawa K, Spielman T, Moorman C, et al. Fluid structure interaction modeling of
spacecraft parachutes for simulation-based design. J Appl Mech 2012; 79: 1–9.
[11] Liu GR and Liu MB. Smoothed particle hydrodynamics: a meshfree particle method.
Changsha: Hunan University publishing house, 2005.
[12] Essam AB, Hisham E and Souli M. Experimental and numerical investigation for
membrane deployment using SPH and ALE formulations. CMES-Comp Model Eng
2015; 104: 405–424.
[13] Ma L, Bao RH and Guo YM. Waterjet penetration simulation by hybrid code of SPH
and FEA. Int J Impact Eng 2008; 35: 1035–1042.
[14] Shangguan ZN, Han DF and Ma QW. Application of SPH-ALE method in numerical
wave making problem. J Ship Mech 2015; 19: 43–51.
[15] Lee ES, Moulinec C, Xu R, et al. Comparisons of weakly compressible and truly
incompressible algorithms for the SPH mesh free particle method. J Comput Phys
2008; 227: 8417–8436.
[16] Liu JJ, Long J, Liang K, et al. Reverse modeling of complicated folded fabric. J Ind
Text 2016; 46: 417–435.
[17] Souli M, Ouahsine A and Lewin L. ALE formulation for fluid-structure interaction
problems. Comput Meth Appl Mech Eng 2000; 190: 659–675.
[18] Yu L. Parachute emulation and experimental investigation. PhD Thesis, Nanjing
University of Aeronautics and Astronautics, PRC, 2006.