You are on page 1of 12

Original article

Textile Research Journal


0(00) 1–12

Toward three-dimensional modeling ! The Author(s) 2018


Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
of the interaction between the air flow DOI: 10.1177/0040517518758006
journals.sagepub.com/home/trj
and a clamped–free yarn inside the main
nozzle of an air jet loom

Akil Osman1, Lucas Delcour1, Ine Hertens1, Jan Vierendeels1,2


and Joris Degroote1,2

Abstract
In air jet looms, the weft yarn is transported from the prewinder to the reed by means of an air flow. In this work, the
motion of a yarn inside a main nozzle during the first stage of an insertion process is modeled and analyzed. In this stage,
the weft yarn is clamped at one side and free at the other side. Therefore, the deformation waves of a clamped–free yarn
are modeled. A three-dimensional, two-way, fluid–structure interaction simulation has been performed in which the yarn
is represented as a flexible cylinder and the arbitrary Lagrangian–Eulerian technique is employed. The results of the
simulation have been compared quantitatively and qualitatively with experiments. It was, however, not possible to match
the initial position and stress state of the yarn in the simulations to that in the experiments. This causes large differences
between the simulated and measured yarn positions and wave characteristics, especially at the beginning. The agreement
between experimental and simulated wave characteristics notably improves as time progresses, but substantial differ-
ences remain. Analyzing the overall motion of the yarn inside the main nozzle shows that the mixing region, where the
shocks are located, can be considered as an excitation point. In this point, the aerodynamic normal forces are high if the
yarn is not located on the axis of the main nozzle. All deformation waves start from the mixing region and propagate
along the yarn.

Keywords
weft yarn, insertion, main nozzle, fluid–structure interaction, clamped–free, dynamic mesh, deformation waves

The air jet loom is widespread in the textile industry turbulence. The normal forces deform the weft yarn
due to its high productivity. In air jet looms, the weft and these deformations propagate inside the main
yarn is transported from the prewinder to the reed by nozzle, affecting both the yarn and the air flow.
means of an air flow. In the first stage of the insertion, Although modeling the motion of yarns in air flows
the weft yarn is fixed or clamped by a magnetic pin and has attracted many researchers’ interest, most of the
an air flow is built up in the main nozzle. The second published works ignored the interactions between air
stage starts by releasing the weft yarn, which then accel- flows and yarns. For example, in Uno,1 Uno et al.,2
erates by means of aerodynamic forces. In the reed, the
weft yarn speed is kept constant by relay nozzles. When
the weft yarn reaches the end of the reed, it is brought 1
Department of Flow, Heat and Combustion Mechanics, Faculty of
to a halt and cut off. Subsequently, the reed is moved so Engineering and Architecture, Ghent University, Belgium
2
as to push the weft against the cloth. Flanders Make, Belgium
The main nozzle is mainly designed to generate axial
Corresponding author:
forces on the weft yarn. However, due to the effect of Lucas Franky Delcour, Universiteit Gent, Sint-Pietersnieuwstraat 41
gravity, the weft yarn deviates from the axis of the Gent, 9000 Belgium.
nozzle, giving rise to normal forces due to shocks and Email: lucas.delcour@ugent.be
2 Textile Research Journal 0(00)

Adanur and Mohamed,3,4 Turel et al.,5 Adanur and a cotton fabric and found good agreement between their
Turel,6 Celik et al.7 and Nosraty et al.,8 no air flow results and experiments. Zhu and Chin21 modeled the
simulations were carried out. The air flow quantities motion of flexible elastic filaments in a viscous pulsatile
were obtained based on simplified equations or experi- flow using the immersed boundary method. Battocchio
ments. These works were aimed at calculating the yarn et al.22 suggested a two-dimensional multi-body dynam-
speed or the yarn tension or to study the effects of yarn ics model to simulate the dynamics of a long polymeric
motion on the performance of air jet looms. Other fiber in a high-speed turbulent flow. They used the
research focused on developing numerical models Articulated-Body algorithm to model the fiber motion,
to model yarn motion, for example, the one- which consisted of a chain of rigid rods. Pei et al.23 mod-
dimensional model suggested by Vangheluwe et al.,9 eled the dynamics of a ramie yarn hair by performing
De Meulemeester et al.10 and Yamamoto and two-way FSI simulations. They used the same procedure
Matsuoka,11 or the three-dimensional (3D) model as presented by Pei and Yu,19,20 but considered the air
developed by De Meulemeester et al.12 However, the flow to be compressible. Wu et al.24 carried out two-
air flow–yarn interactions were also not taken into dimensional, two-way FSI simulations to analyze
account. Nevertheless, the works1–12 helped one to deformations of a weft yarn during the start-up stage
understand the dynamic behavior of yarns, led to in an air jet loom. They employed the ALE technique
improved performance of air jet looms and provided and took into account the pressure effects, but ignored
numerical models to predict the motion of yarns. the viscous forces. They compared the results of the
The development of computational fluid dynamics simulations with photos taken by a high-speed camera
(CFD) and computational structure dynamics provided during experiments.
powerful tools to carry out air flow–yarn interaction This paper continues the work done by Osman
simulations. For example, Cai and Oxenham13 per- et al.25 on modeling and analyzing the dynamic behav-
formed one-way fluid–structure interaction (FSI) simu- ior of yarns during the insertion process of an air jet
lations of air flow–fiber in the fiber transfer channel of a loom. In that work, the motion of a weft yarn during
rotor spinning machine. They calculated the air flow one complete insertion was modeled by means of FSI
with a CFD software package and transferred the result- simulations of air flow–yarn. The fluid model was two-
ant velocity and pressure to a fiber movement model. dimensional axisymmetric, the structure model was 3D
They showed the effectiveness of using numerical tools and the ALE technique was not used. The fluid mesh
to study the dynamics of fibers in textile processes. Zeng was static and the effect of the yarn motion on the air
et al.14 performed two-way and two-dimensional FSI flow was modeled by adding a source term. The idea of
simulations of a fiber in high-speed air flow. They rep- this work is to study the motion of a weft yarn during
resented the fiber as beads connected by massless rods; the insertion process by means of completely 3D two-
the air flow was laminar and incompressible. The effect way FSI simulations with the ALE technique. In our
of the fiber motion on the air flow was taken into previous work, some simplifications were made to carry
account by a source term. They applied their models out the FSI simulations. However, performing the two-
to simulate fiber motion in the nozzle of an air jet spin- way FSI simulations in a 3D configuration, as pre-
ning machine. However, they mentioned that the pro- sented in this paper, increases the cost and duration
posed procedure was computationally very expensive. of the simulations. In this research, only the motion
Tang and Advani15 and Kondora and Asendrych16 of a yarn during the first stage of the insertion is mod-
modeled the motion of fibers by representing the fiber eled and analyzed, with the yarn clamped at one end
as a chain of spheres connected by ball and socket joints. and free at the other end. This avoids complications
Zeng et al.17 and Sun et al.18 modeled fiber whipping related to the axial motion of the yarn. As the forces
motion (bending deformations) during the process of on the yarn are strongly dependent on its position,
melt blowing by performing two-way FSI simulations. being able to simulate the yarn motion during the
They derived a structure equation for a flexible fiber, clamped part of the insertion is an important first
represented as connected beads, and coupled it with a step into simulating the entire insertion process.
flow solver. Pei and Yu19,20 modeled the motion of a Furthermore, this research aims at providing insight
fabric in the vortex spinning nozzle. They performed into the problems and possibilities of using 3D FSI
two-way FSI simulations using an arbitrary simulations for weft yarn insertion. The results of the
Lagrangian–Eulerian (ALE) technique in which the simulations are compared quantitatively and qualita-
mesh is dynamic. The air flow was incompressible and tively to experiments. Some of the works mentioned
turbulent and the domains of the simulations were two- above, for example Pei and Yu,19,20 also performed
dimensional and planar. They studied the effects of some two-way FSI simulations with the ALE technique.
geometrical parameters of the nozzle, the inlet pressure However, the domains of the simulations were two-
and the yarn delivery speed on the dynamic behavior of dimensional in those works to avoid the complexity
Osman et al. 3

of 3D two-way FSI simulations with moving grids.


Consequently, the presented work is a significant
improvement compared to the state of the art. It
should be noted that a smooth mono-filament yarn is
used for the current work as this reduces uncertainty on
the surface and structural properties of the yarn. The
proposed simulation methodology could be applied to
any type of yarn, provided that a good structural model
is available for the yarn and that one can model the
influence of the yarn’s surface structure.

Experimental details
The numerical simulations are set up according to a
case that has been investigated experimentally. In the
experiment, the deformations of a clamped–free yarn Figure 1. Measured pressure profile at the inlet of the main
were recorded by a high-speed camera in a main nozzle during the experiment. The values are relative to the
nozzle made of glass. The camera type is Kodak Ekta atmospheric pressure.
Pro EM. The image capture rate was adjusted to 10,000
frames per second to capture clearly the motion of the pressure at zero time (the beginning of the recordings)
yarn. The yarn is a nylon 6.6 mono-filament and its corresponds to the holding flow pressure.
properties, according to measurements, are as follows: A meridional section of the geometry of the main
the diameter is 0.72 mm, the yarn linear density is 464 nozzle, used in the experiment, is depicted in Figure 3
tex and the Young’s modulus (E) is 2.5 GPa. The yarn with the main dimensions indicated.
is clamped at the yarn inlet of the main nozzle and is
free at the exit of the tube.
At the beginning of the experiment, the main nozzle Simulation setup
is supplied with a pressure of 0.14 bar (all pressures are
Flow solver
relative to the atmospheric pressure), creating a holding
flow that keeps the yarn more or less on the axis of the The computational fluid domain, which is 3D, consists
main nozzle. Afterwards, the supply pressure is of the main nozzle geometry and two additional outer
increased to 5 bar. There is a tube between the com- zones, as shown in Figure 4(a). The yarn is represented
pressed air tank and the main nozzle, but the pressure as a cylinder with diameter equal to the yarn diameter
at the inlet of the main nozzle is measured during the of 0.72 mm. The cylinder extends from the yarn inlet to
experiment. The pressure was sampled at a frequency of the end of the main nozzle tube. The yarn is clamped at
10 kHz using a piezo-resistive pressure transducer with a the yarn inlet, as shown in Figure 4(a). At the exit of the
least count of 0.046 bar (Meggitt model 8530B, 200 psia tube, the yarn is free. The mesh is structured, as can be
range). Figure 1 shows the measured pressure profile at seen in Figure 4(b), and contains about 367,000 cells.
the inlet of the main nozzle. The initial pressure corres- The mesh is gradually refined toward the mixing region
ponds to the pressure of the holding flow, equal to 0.14 to more accurately capture the shocks. A somewhat
bar. After 5.4 ms, a valve is opened and the pressure higher mesh density has also been incorporated at the
increases sharply to 5.35 bar. This value is higher than tube outlet. The coordinate system is shown in
the inlet pressure, which is equal to 5 bar. This higher Figure 4(a); the origin is located at the yarn inlet.
value is due to pressure waves traveling between the Figure 5 displays the static pressure as calculated on
tank and the inlet of the main nozzle. From 18 to the yarn if it is fixed on the axis with an inlet pressure of
57 ms, the pressure varies between 4 and 5 bar. 5 bar. Two different meshes are used with 367,000 cells
Afterwards, the pressure reduces back to 0.14 bar. and 735,000 cells, respectively.
A schematic diagram of the experimental setup is From Figure 5 it can be seen that the mixing region
provided in Figure 2. The camera records the motion is most sensitive to mesh refinement. However, per-
of the yarn in a vertical plane. The motion is recorded, forming calculations on the finer mesh would drastic-
separately, in two parts of the nozzle tube as the high- ally increase the computational time. Since the
speed camera cannot capture the entire tube at once deviation remains quite limited, it is not expected to
with sufficient resolution. The holding flow is already affect the general conclusions.
present at the beginning of the video recordings. This As the mesh is dynamic, the ALE formulation of the
can be observed in Figure 1 in which the measured flow equations is solved (Donea et al.26 and Sanches
4 Textile Research Journal 0(00)

Figure 2. Experimental setup.

Figure 3. Meridional view of the geometry of the main nozzle. The red, black and blue lines represent the nozzle body, the needle
and the tube, respectively. (Color online only.)

Figure 4. (a) Meridional view of the computational fluid domain with the boundary conditions. (b) Details of the mesh.
Osman et al. 5

and Coda27). The air flow is compressible and turbu- interface over the cells from the surrounding fluid
lent; the shear stress transport k–! model (k–! SST) domain. The Laplacian (diffusion) technique is used
(Menter28) is used as the turbulence model. in this work for the extension of the grid displacements.
The 3D transient simulations of the air flow have This technique is based on the Laplace equation to cal-
been performed using Ansys Fluent 17.0, in which the culate the new positions of the grid nodes.
coupled scheme is used as the solution method for pres- The boundary conditions are set as follows: the inlet
sure and velocity. The second-order upwind scheme is pressure is set equal to the measured pressure profile
used for the convective terms in the density, momentum shown in Figure 1. The outlet pressure is set to atmos-
and energy equations. The least-squares cell-based pheric pressure. The wall of the needle, tube and yarn
method is used for the gradients. The second-order are set to no-slip walls.
implicit scheme is used for the time discretization. As mentioned above, at the beginning of the experi-
Several methods can be used to handle the dynamic ment the holding flow is already present. Therefore, at
mesh. The smoothing method is applicable to a struc- the beginning of the FSI simulation the initial values of
tured and an unstructured mesh and is employed in this the air flow quantities correspond to the values
work. The principle of the smoothing method is to dis- obtained with 0.14 bar inlet pressure. The yarn’s initial
tribute the displacements of the fluid–structure position is on the axis of the main nozzle. The conver-
gence tolerances are set to 106 for all flow quantities.
The time step used for the flow simulations equals
5  106 s and 3400 time steps are carried out.

Structure solver
The structural computational domain is shown in
Figure 6. It consists of the yarn as a cylinder and the
tube of the nozzle (which is called the fictive tube, as
will be explained in the next paragraph). The cylinder’s
length is 0.28 m. The structured mesh of the cylinder
contains 4800 elements of type C3D20R. C stands for
continuum elements, which are used for solid analysis,
3D means 3D, 20 is the number of nodes and R stands
for reduced-integration, which reduces the number of
integration points required to calculate the element
matrices. The properties of the cylinder are set to
those of the yarn, which are linear density 464 tex,
Figure 5. Static pressure on yarn, obtained from simulation Young’s modulus E ¼ 2.5 GPa and Poisson’s ratio
with yarn fixed on the axis. Mesh 2 contains approximately  ¼ 0:39. Neither displacements nor rotations are
double the number of cells of mesh 1.
allowed at the fixed end of the yarn.

Figure 6. Computational structural domain.


6 Textile Research Journal 0(00)

The fictive tube, shown in Figure 6, is not a meshed iterations were sufficient to reach the imposed conver-
body, but an analytical rigid body. The idea of this gence tolerance.
body is to prevent the yarn from moving out of the
main nozzle. Moreover, the radial dimensions of this
fictive tube are 15% smaller than the corresponding Results and discussion
ones from the tube in the fluid domain for dynamic
mesh considerations. If the yarn came into contact
Comparison with the experiment
with the actual tube, the fluid mesh between the yarn To validate the results of the simulations, they are com-
and the tube would be completely compressed and the pared to the experiments. It should be mentioned that
simulation would crash. Therefore, the yarn motion is the initial position of the yarn is not the same in the
constrained to the fictive tube with smaller diameters, simulations as it is in the experiments. Although it is
which prevents failure of the fluid dynamic mesh possible to impose an initial deformation in the simu-
motion. Frictionless contact is set between the yarn lations, obtaining the initial yarn position from the
and the fictive tube in the tangential direction and video recordings is not feasible since only part of the
hard contact in the normal direction. yarn could be recorded at once. Furthermore, extract-
The equation governing the motion of a structure ing the position of the yarn in three dimensions would
(the weft yarn in this case) is require synchronized footage from a second camera
angle. In addition, inaccuracies in the extracted yarn
@2 u~ position, small deviations in the nozzle geometry or
s  r   s ¼ fs ð1Þ small variations in local yarn properties can result in
@t2
a notably different yarn position after some time.
where s is the density of the weft yarn, u~ is the Moreover, the stress components along the yarn
deformation of the weft yarn, fs is the body force cannot be obtained from the video footage, prohibiting
per unit volume and  s is the stress tensor. Equation a complete description of the yarn’s initial state.
(1) is discretized with the finite element method. The Figure 7 shows the positions of the yarn in three
Hilber–Hughes–Taylor implicit time integration frames for both the simulation and the experiment. In
method is used. The geometrical nonlinearity is the simulations, the initial position of the yarn is on the
taken into account. In addition to the fluid loads, axis of the main nozzle. From Figure 7(a) it can be seen
which are pressure and viscous traction, gravity is that the yarn in the first frame of the simulation is still
also included in the structure calculations. The 3D positioned almost perfectly on the axis of the main
structural implicit dynamics have been calculated nozzle. The video recordings show that from 0 to
using Abaqus 6.14 (Simulia Inc., Providence, RI, 5 ms the yarn position remains basically unaltered.
USA). The time step that is used in the structural This implies that from 0 to 5 ms, the air flow simply
simulation is the same as the one used in the flow holds the yarn in the same position. Therefore, it can be
simulation (t ¼5  106 s). stated that the response of the yarn to the air flow is the
same in the experiment and in the simulation during the
first 5 ms. The difference is just in the initial position of
Coupling algorithm
the yarn.
The FSI simulation has been performed with two sep- The different initial position of the yarn in the simu-
arate codes. The air flow calculations are strongly or lation with regard to the experiment affects the subse-
implicitly coupled with the structural calculations and quent positions. Figure 7(b) and (c) display the
the interface Quasi-Newton technique with an approxi- deformations observed during the experiment and the
mation of the inverse of the Jacobian from a least- simulation at a later time. These deformations run
squares model (IQN-ILS) has been used (Degroote along the yarn. Comparing exactly the position of the
et al.29). With an implicit coupling algorithm, the cou- yarn in Figure 7(b) and (c) is not meaningful because of
pling equilibrium conditions on the fluid–structure several reasons. First, the initial position is not the
interface are satisfied in each time step. These equilib- same in the simulations and the experiments. Second,
rium conditions are the kinematic equilibrium v~ ¼ d~ u
dt local variations in the real yarn properties cannot be
and the dynamic equilibrium  f  n~f ¼  s  n~s , where taken into account in the simulations but can influence
v~ is the air flow velocity, n~ is the unit normal pointing the yarn position.
outwards and the subscripts f and s refer to the fluid Figure 8 shows three frames of the yarn position in
and the structure, respectively. the second half of the tube. Figure 8(a) shows the pos-
The relative convergence tolerance is set to 106 with ition of the yarn at 5 ms. Also in this part of the tube,
a maximum number of 20 coupling iterations allowed. the position of the yarn has not changed from the
However, after a few time steps, three coupling beginning of the video recordings up to 5 ms.
Osman et al. 7

Figure 7. Frames of the yarn during experiments and the corresponding frames of the simulations at time: (a) 5 ms; (b) 12 ms; (c)
15 ms. The domain shown starts from the needle tip to 130 mm downstream. Images are vertically enlarged by a factor 2. The tube
border is highlighted in blue and the yarn (edge) is highlighted in green. (Color online only.)

Figure 8. Frames of the yarn during experiments and the corresponding frames of the simulations at time: (a) 5 ms; (b) 12 ms; (c)
15 ms. The length of the shown domain equals 125 mm; the right-hand side of the domain shown is at the tube exit. Images are
vertically enlarged by a factor 2. The tube border is highlighted in blue and the yarn (edge) is highlighted in green. (Color online only.)
8 Textile Research Journal 0(00)

Figure 9. Wave characteristics and measurement. Point A from the shown wave is followed to calculate the wave speed.

Table 1. Measured waves from the experiment Table 2. Calculated waves from the simulation. The percentage
values are the deviations from the corresponding experimental
Wave Frame time [ms] c [m/s]  [mm] values
1 7.8–10.4 15.3 70.0 Wave Frame time [ms] c [m/s]  [mm]
2 12.8–14.7 31.5 100.0
1 8.5–9.9 39.5 (+61.2%) 48.0 (48.8%)
3 15.4–16.6 33.3 80.0
2 10.6–11.8 38.3 (+17.7%) 91.0 (9.9%)
3 14.6–16.2 35.6 (+6.4%) 60.0 (33.3%)

Increasing the inlet pressure of the nozzle forces the


yarn to move back to the axis of the main nozzle, as measuring and calculating the waves from the video
shown in Figure 8(b). Then, the deformation waves recordings is quite challenging.
move along the yarn, as can be seen in Figure 8(c). The difference in mode shapes between the simula-
Comparing the mode shapes of the deformation tion and the experiment, which are seen in Figures 7
waves in Figures 7 and 8 has not given a clear conclu- and 8, are due to the differences in the wave lengths and
sion about the results of the simulations. Therefore, the speeds. As the differences decrease with time, it can be
wave length and speed are considered. postulated that the different initial position of the yarn
The characteristics of the waves are measured and in the simulation and the experiment causes these dif-
calculated, as shown in Figure 9. A point from a wave ferences. In fact, at the beginning of the simulation, the
is chosen to be followed; that point has to be either a yarn is positioned on the axis of the main nozzle, while
crest or a trough. For example, in Figure 9 point A from the initial position of the yarn in the experiment cannot
the chosen wave is followed and the wave speed is calcu- be imposed or controlled and is only approximately
lated according to c ¼ xA =t with xA ¼ xA,frame2  positioned on the axis by the holding flow. Based on
xA,frame1 and t ¼ tframe2  tframe1 . The wave length () is the discussion above it can be stated that the different
taken directly from the frame, as shown in Figure 9. In initial position causes large differences in wave charac-
the same way, three waves are measured from the video teristics for the first deformation wave. For subsequent
recordings and the results are listed in Table 1. deformation waves, the obtained wave speeds are
Table 2 lists the obtained values of the wave charac- within acceptable accuracy (between 6.4% and 17.7%
teristics from the simulations and the differences com- difference) of the experimental values.
pared to those from the experiment. For the first wave
the difference is 61% for the wave speed, but for the
Yarn motion
second and the third wave the differences decrease to
6% for the wave speed. It should be mentioned that the The overall motion of the yarn is analyzed based only
experimental values are only a crude estimate, as on the results of the simulation. Although the
Osman et al. 9

Figure 10. Coordinates of three points from the yarn centerline during the simulations: (a) y-coordinates; (b) z-coordinates. (Color
online only.)

correspondence between experiment and simulation


when comparing the yarn position is rather poor, the
simulations can still provide valuable clues about the
mechanisms behind the yarn deformation.
Three points from the yarn centerline are chosen. By
plotting the coordinates of the three points, it can be
observed how the deformations propagate along the
yarn. The selected points are located in the mixing
zone (see Figure 4), in the middle of the tube and at
the yarn’s free end. Initially, the points are positioned
on the axis at respectively 35.3, 158.4 and 283.0 mm
from the yarn inlet. Figure 10 shows the y- and z-coord-
inates of the three points during the simulations.
Figure 10(a) shows that, in the first 7 ms of the simu-
lation, the yarn falls toward the tube bottom due to
gravity. Moreover, Figure 10 indicates that, for exam-
ple, at 5 ms, the y-coordinates of the three points are
almost the same and the same holds for the z-coordin-
ates. Figure 11 displays frames of the yarn centerline in
the xy- and xz-planes. Note that the scale in
Figure 11(b) is smaller than the scale in Figure 11(a).
In Figure 11(a) it can be seen that the yarn falls due to
gravity, but deformation waves with low amplitudes are
also present. As the yarn falls, it does not experience a Figure 11. Frames of the yarn centerline: (a) y-coordinates;
(b) z-coordinates.
symmetrical flow anymore and, thus, normal forces are
generated due to pressure differences. All deformations
start in the mixing zone and propagate downstream. Figure 12 shows contours of static pressure in the
After 10 ms of simulation time, the amplitudes of the yz-plane at an axial coordinate equal to the coordinate
waves increase along z and y directions, as seen in of the point in the mixing region (red color line in
Figure 10. The wave amplitudes increase due to an Figure 10). The static pressure around the yarn in
increase of the inlet pressure (see Figure 1), which Figure 12 can be divided into two regions: a region
increases the aerodynamic forces on the yarn wall. with high values of static pressure and another with
10 Textile Research Journal 0(00)

Table 3. Speed of the waves shown in Figure 13 [m/s]

c12 c23 c34

In the xy-plane 41.6 34.9 40.7


In the xz-plane 22.3 30.0 36.5

Figure 12. Contours of static pressure in the yz-plane at an


axial coordinate located in the mixing zone, at time: (a) 11.5 ms;
(b) 12 ms.

Figure 14. (a) Yarn centerline position in the xy-plane. (b) Y-


component of viscous force along the yarn centerline. (c) Y-
component of pressure force along the yarn centerline.

The results of the FSI simulation, shown in


Figure 13. Frames of the yarn centerline in the xy- and xz-
Figures 11 and 13, indicate that the deformation
planes. The highlighted numbers show the location of the same waves start from the mixing region where the shocks
crest of a wave in the four frames. are located. There, the yarn is subjected to strong
normal pressure forces if, locally, it is not perfectly
aligned with the axis.
low values. The yarn is pushed toward the region of low Figure 14 shows two frames of the yarn position and
pressure. the associated y-components of the viscous and pres-
Figure 13 shows four other frames of the yarn cen- sure force. Figure 14(b) shows that the contributions of
terline at a later time than the frames in Figure 11. It the viscous force in the vertical direction are small. The
can be seen that the amplitudes of these waves are y-component of the force depends mainly on the pres-
higher than the ones in Figure 11. The numbers that sure, as shown in Figure 14(c). It can be seen that there
are highlighted in Figure 13 show the same crest of a are large contributions in the mixing region where the
wave. This wave or crest propagates inside the tube. shocks are located.
Therefore, the speed of the wave can be calculated Figure 15 shows the contours of the Mach number.
based on this crest. The calculated wave speeds are It can be seen how the Mach number changes around
listed in Table 3. the mixing region due to the effects of shocks,
Osman et al. 11

Figure 15. Contours of Mach number in the xy-plane at 13 ms: (a) around the mixing region; (b) between x ¼ 175 and 190 mm.

expansion and compression waves. This is not the case


Declaration of conflicting interests
in another part of the tube, for example in Figure 15(b).
The authors declared no potential conflicts of interest with
It can be concluded that if the yarn is not on the axis
respect to the research, authorship, and/or publication of this
of the main nozzle, it is subjected to normal forces
article.
generated by the effects of the shocks. The normal
forces deform the yarn and the deformation waves
Funding
run along the yarn. Therefore, the location of the
The authors disclosed receipt of the following financial sup-
shock can be considered as an excitation point.
port for the research, authorship, and/or publication of this
The deformation waves can be linked to the pressure
article: This work was financially sponsored by Aleppo
force in this excitation point. Stronger shocks result in University of Syria, by a Starting Grant from the Special
higher pressure forces, which result in larger Research Fund of Ghent University and by the Flemish gov-
deformations. ernment agency for Innovation by Science and Technology
(IWT), project number 120519.
Conclusion
References
In this paper, the FSI simulation of air flow–yarn inside 1. Uno M. A study on an air-jet loom with sub-streams
a main nozzle has been presented. The domains of the added, part 1: deriving the equation of motion for weft.
simulation are 3D and the ALE technique has been J Text Mach Soc Jpn 1972; 18: 37–44.
employed. The first stage of an insertion process has 2. Uno M, Shiomi A and Kise H. A study on an air-jet loom
been modeled. In this stage, the yarn is clamped on with sub-streams added, part 2: analysis of various weav-
one side and it is free on the other side. The weft ing factors by the equation of motion of weft. J Text Mach
yarn is represented as a cylinder in the simulations. Soc Jpn 1972; 18: 86–92.
The deformation waves that run along the yarn have 3. Adanur S and Mohamed MH. Analysis of yarn motion in
single-nozzle air-jet filling insertion, part I: theoretical
been obtained and analyzed.
models for yarn motion. J Text Inst 1992; 83: 45–55.
The results of the simulation have been compared
4. Adanur S and Mohamed MH. Analysis of yarn tension in
with video recordings of the yarn motion during the air-jet filling insertion. Text Res J 1991; 61: 259–266.
experiments. The initial state of the yarn is not the 5. Turel T, Bakhtiyarov S and Adanur S. Effects of air and
same in the simulation and the experiment. Therefore, yarn characteristics in air-jet filling insertion, part I: air
large differences between the simulation and the experi- velocity and air pressure measurements. Text Res J 2004;
ment have been observed. However, as time progresses 74: 592–597.
and the influence of the initial state diminishes, accept- 6. Adanur S and Turel T. Effects of air and yarn character-
able wave speeds (between 6.4% and 17.7% difference istics in air-jet filling insertion, part II: yarn velocity meas-
with experiments) are retrieved. urements with a profiled reed. Text Res J 2004; 74:
By analyzing the global motion of the yarn it could 657–661.
7. Celik N, Babaarslan O and Bandara MPU. A math-
be concluded that the deformation waves start from the
ematical model for numerical simulation of weft insertion
mixing region and propagate along the yarn. These
on an air-jet weaving machine. Text Res J 2004; 74:
deformations result from the presence of the shocks, 236–240.
which cause a force away from the centerline if the 8. Nosraty H, Jeddi AAA and Mousalloo Y. Simulation ana-
yarn is not perfectly aligned with it. The contribution lysis of weft yarn motion in single nozzle air-jet loom to
of the viscous forces to the radial motion is negligible study the effective parameters. Indian J Fibre Text Res
compared to the effect of the pressure forces. 2008; 33: 45–51.
12 Textile Research Journal 0(00)

9. Vangheluwe L, Sleeckx B and Kiekens P. Numerical simu- some nozzle structure parameters. J Eng Fibers Fabr
lation model for optimisation of weft insertion on project- 2011; 6: 16–29.
ile and rapier looms. Mechatronics 1995; 5: 183–195. 21. Zhu L and Chin RCY. Simulation of elastic filaments
10. De Meulemeester S, Githaiga J, Van Langenhove L, et al. interacting with a viscous pulsatile flow. Comput Meth
Simulation of the dynamic yarn behavior on airjet looms. Appl Mech Eng 2008; 197: 2265–2274.
Text Res J 2005; 75: 724–730. 22. Battocchio F, Sutcliffe MPF and Teschner F. Dynamic
11. Yamamoto S and Matsuoka T. A method for dynamic simulation of long polymeric fibres immersed in a turbu-
simulation of rigid and flexible fibers in a flow field. lent air flow. In: Eberhard P and Ziegler P (eds) proceed-
J Chem Phys 1993; 98: 644–650. ings of the IMSD2012 - the 2nd joint international
12. De Meulemeester S, Puissant P and Van Langenhove L. conference on multibody system dynamics, Stuttgart,
Three-dimensional simulation of the dynamic yarn Germany, 29 May–1 June 2012, http://www2.eng.cam.a-
behaviour on air-jet looms. Text Res J 2009; 79: c.uk/~mpfs/papers/BST2012FibreTurbulentIMSD.pdf
1706–1714. 23. Pei Z, Chen G, Mao L, et al. Dynamics of the ramie yarn
13. Cai Y and Oxenham W. Computer modeling of fiber hair in the nozzle of the jetwind process and effects of
movements in high-speed airflow. RJTA 2005; 9: 77–85. some nozzle parameters. J Nat Fibers 2015; 12: 430–443.
14. Zeng YC, Yang JP and Yu CW. Mixed Euler-Lagrange 24. Wu Z, Chen S, Liu Y, et al. Air-flow characteristics and
approach to modeling fiber motion in high speed air flow. yarn whipping during start-up stage of air-jet weft inser-
Appl Math Model 2005; 29: 253–261. tion. Text Res J 2016; 86: 1988–1999.
15. Tang W and Advani SG. Dynamic simulation of long 25. Osman A, Malengier B, De Meulemeester S, et al.
flexible fibers in shear flow. CMES 2005; 8: 165–176. Simulation of air flow-yarn interaction inside the main
16. Kondora G and Asendrych D. Modelling the dynamics nozzle of an air jet loom. Text Res J. Epub ahead of
of flexibles and rigid fibres. Chem Process Eng Inz 2013; print 8 March 2017. DOI: 10.1177/0040517517697646.
34: 87–100. 26. Donea J, Giuliani S and Halleux JP. An arbitrary
17. Zeng YC, Sun YF and Wang XH. Numerical approach Lagrangian-Eulerian finite element method for transient
to modeling fiber motion during melt blowing. J Appl dynamic fluid-structure interactions. Comput Meth Appl
Polym Sci 2011; 119: 2112–2123. Mech Eng 1982; 33: 689–723.
18. Sun Y, Zeng Y and Wang X. Three-dimensional model of 27. Sanches RAK and Coda HB. On fluid-shell coupling
using an arbitrary Lagrangian-Eulerian fluid solver
whipping motion in the processing of microfibers. Ind
coupled to a positional Lagrangian shell solver. Appl
Eng Chem Res 2011; 50: 1099–1109.
Math Model 2014; 38: 3401–3418.
19. Pei Z and Yu C. Numerical study on the effect of nozzle
28. Menter FR. 2-equation eddy-viscosity turbulence models
pressure and yarn delivery speed on the fiber motion in
for engineering applications. AIAA 1994; 32: 1598–1605.
the nozzle of Murata vortex spinning. J Fluids Struct
29. Degroote J, Bathe K-J and Vierendeels J. Performance of
2011; 27: 121–133.
a new partitioned procedure versus a monolithic proced-
20. Pei Z and Yu C. Investigation on the dynamic behavior
ure in fluid–structure interaction. Comput Struct 2009; 87:
of the fiber in the vortex spinning nozzle and effects of
793–801.

You might also like