On-line Loom Weft Density Control System Design
Atieh Almasi Zefrehyee, Mohammad Amani Tehran, PhD, Hooshang Norsaty, PhD, Marzie Abhajani, PhD
Amirkabir University of Technology, Tehran IRAN
Correspondence to:
Mohammad Amani Tehran email: [email protected]
ABSTRACT
Controlling the weft density in the course of weaving
is of great importance in technical textiles
production. In this study, a closed loop control
system was designed and implemented in a weaving
machine. Moreover, a PID algorithm was used in the
control system. The control action starts by taking a
picture from the fabric near the cloth fell. A machine
vision system was prepared to calculate the number
of weft yarns per inch. Based on the measured weft
density, the take-up roller speed was adjusted by
changing the supply voltage to the take-up motor.
Because of the variable delay between the weft
density measurement and cloth fell, the normal PID
control system was not able to provide a robust
action, and thus, a Smith predictor was utilized to
modify the control behavior. The results of set point
tracing and regulatory control action indicated the
satisfactory performance of our novel system in
various tests.
fabric property. They used a negative sley driven
mechanism in a shuttle loom. The sley was pushed
toward the fell position by two springs. Stemheim
and Grosberg [6,7] determined the effect of sley
motion on the beat-up force and replaced the
mechanically driven sley mechanism by a
microprocessor controlled hydraulic driven sley. Eren
and Porat [8] developed two-input and two-output
control systems, and demonstrated that their control
system could properly respond to the shift in cloth
fell position and make the starting mark less visible.
In addition, some researchers have focused on
increasing fabric evenness by yarn tension control.
Jeddi et al. [9] and Ordoukhany [10] applied a PID
control system and kept warp tension constant, which
led to an increase in weaving machine efficiency and
improvement of fabric quality. Dayik et al. [11]
developed a gene expression programming control
system to decrease the variation of warp tension
during the weaving cycle. They claimed that warp
yarn breakages were reduced by using such a system.
Nosraty et al. [12,13] designed a PID control system
to control the tension of weft yarn in a nozzle air-jet
weaving loom. Their results indicated that fabric
production was more even under the controlled
tension of weft yarn as compared to the uncontrolled
tension. Kuo, Su and Chen [14] developed a neural
network control system to identify the optimal angle
for a beat-up mechanism. Their study resulted in an
improvement in the system transient response
eliminating the steady-state errors.
INTRODUCTION
One of the important properties of woven fabric is
weft and warp density. Within recent decades,
numerous technological advances in weaving looms
have improved woven fabric quality but many
inspections during the weaving process are still being
carried out manually, and sometimes fabrics are
produced with defects. Therefore, utilizing the
weaving machines with a weft density control system
could be considered as an improvement. Some
attempts have been made to reduce weft density
variation and produce more uniform fabrics.
However, most researchers have attempted to prevent
starting mark fault occurring when a loom is restarted
after a stoppage [1]. Greenwood and Cowhig [2-3]
and Greenwood and Vaughan [4] have published a
series of articles in which theoretical and
experimental work were combined to clarify the
influential factors on pickspacing. They mentioned
that cloth fell position and beat-up force affected the
fabric pickspacing. In another study, Greenwood and
McLoughlin [5] developed a mechanical control
system to eliminate the effect of fell position on the
Journal of Engineered Fibers and Fabrics
Volume 10, Issue 4 – 2015
Since the image processing technique was employed
in the current study to monitor the weft density and
the corresponding discrepancy, conventional image
processing methods for weft density measurement are
briefly described. Lin [15] measured the weft density
based on the co-occurrence matrix algorithm. The
results indicated that such a method was only suitable
for single colour plain weave fabric and required long
complex mathematical operations. A Fourier
transform algorithm is also able to inspect the
28
http://www.jeffjournal.org
frequency components and periodic patterns in
images [16-20]. A line profile algorithm is another
method of fabric weft density measurement, which is
able to measure the weft density in various weave
types [20,21]. In addition, other researchers have
applied non-conventional methods for measuring the
weft density [22-25].
tension variation, fluctuation of supply electrical
power and defects in mechanical components such as
the undesirable performance of gear. This paper
attempts to describe a weft density feedback control
system based on the PID control algorithm which
applies a digital image processing technique in the
weft density measuring system.
Although various control systems have been used in
weaving machines to improve fabric quality, an
effective control system has not yet been applied to
minimize the weft density variation.
ELEMENTS OF A WEFT DENSITY CONTROL
SYSTEM
Elements of this control system are shown in Figure
1. It consists of a measuring system, comparison
element, controller and actuator. These elements are
described briefly as follows:
In the weaving process, several conditions cause the
weft density to vary from the set point such as warp
FIGURE 1. A schematic view of the closed loop weft density control system.
In this system, the fabric density must be measured in
real time and compared with the reference value
which is determined in the fabric design
specification. Comparison results are transmitted to
the controller by a signal, results are transmitted to
the controller via a signal, which calculates and
produces the output signal according to an input
signal, and sends it to the actuator. Eventually, the
actuator, depending on the input signal, changes the
take-up roller speed properly. In such a control
system, the acceptable tolerance is ±1 pick per inch.
Control Unit
In this work, a PID controller was employed in the
weft density on-line control system. The PID
controller is a combination of proportional, integral
and derivative control actions. The output of the
controller is defined by Eq. (1).
t
∫
output = u (t ) = k p e(t ) + ki e(t )dt + k d
0
(1)
where u(t) and e(t) are controller output and error
signal, respectively. kp, ki and kd are proportional,
integral and derivative gains, respectively. The error
signal is produced by comparing the measured fabric
weft density with a set point value as defined in Eq.
(2).
Measuring System
The digital image processing technique has been
applied to a measuring system. When the weaving
process is affected by a disturbance and set point
variation, the result is an unevenness in the fabric.
Fourier transform has been considered as an
important algorithm for even weft density evaluation
in previous research. However, our experiments have
indicated that this method is not able to trace weft
density in uneven fabric and is only able to calculate
the frequency of a periodic event. Therefore, the line
profile algorithm was developed and used. This line
profile algorithm is able to measure weft density in
even and uneven fabrics. The details of such a
method will be explained in the machine vision
design section.
Journal of Engineered Fibers and Fabrics
Volume 10, Issue 4 – 2015
de(t )
dt
e(t ) = wds − wd
(2)
where wds represents the set point value and wd is
the fabric weft density evaluated by the measuring
system.
Actuator
The actuator gets an order from the controller and
makes changes to the take-up roller speed to
compensate the error. Take-up roller speed is the
29
http://www.jeffjournal.org
most influential factor on weft density such that the
slightest change in the take-up roller speed could
cause the weft density to deviate from the set point.
take an appropriate image from the subject, and
second, use of a reasonable program for data
processing.
The actuator consists of a motor Y/Δ 380/220 3PH
AC, 220 volt/50 HZ, 0.75 KW, 1450 rpm, an inverter
HYNDAI model N50-022SF and gear chain to drive
the take-up roller. The inverter changes the motor
rotation speed by controlling the frequency of the
electrical power supplied to the motor.
For on-line measurement of weft density during
fabric production, a Dino-Lite camera model AM311
was installed on the weaving machine 2.5 cm away
from the cloth fell position as shown in Figure 2.
Likewise, frontal lighting was added into the system
to improve image quality.
SYSTEM DESIGN
The weft density control system was installed on a
NUOVO PIGNONE SMIT weaving machine,
specifications of which are described in Table I.
The camera was adjusted to take one picture every
0.5 seconds with 600 * 800 pixels in RGB format.
This specific time interval was chosen according to
the weft insertion rate and acceptable tolerances.
Based on machine specifications, two picks were
normally inserted every 0.5 second.
TABLE I. Weaving machine specification.
Machine
type
Weft inserting
mechanism
TP400
flexible rapier
Weft insertion
rate
(picks/min)
240
Read Width
(mm)
The image data were simultaneously transferred to
the PC via a USB port. The gray line profile
algorithm was also used to calculate weft density
using MATLAB software. Data processing was done
in real time. This apparatus could measure fabric weft
density between 9-42 picks/inch for a cotton yarn of
30 Ne.
2400
Machine Vision Design
Basically, an on-line image processing system should
possess two special characteristics; first, an ability to
FIGURE 2. Fabric weft density measuring system.
The measuring program converted the RGB image to
a gray scale image and used some process to increase
the contrast making it more suitable for weft density
measurement. Figure 3 shows a sample of the gray
scale fabric image before and after processing. To
calculate fabric weft density, a graph of fabric image
intensity was plotted. In this graph, the x-axis
included pixel positions and the y-axis encompassed
the average of pixel intensities lying along the picture
in the weft direction as illustrated in Figure 4(a).
Journal of Engineered Fibers and Fabrics
Volume 10, Issue 4 – 2015
As shown in Figure 3, lightened areas represent yarn
presence in the fabric image and the darker regions
are the fabric background. According to the grayscale
image definition, the local maxima in the graph of
fabric image intensity indicate the yarn presence in
the image and the number of these spikes is equal to
the fabric weft density. The existing noise in the
graph were identified as spikes, and thus, eliminated.
30
http://www.jeffjournal.org
To reduce the noise in the fabric image intensity
graph, the moving average smoothing method was
used. In this method, a new value of intensity for
each point was computed by averaging the pixel
intensity and its adjacent pixels. In the measurement
algorithm, a 10-pixel moving average was used.
Figure 4(a) and Figure 4(b) show the average
intensity graph before and after noise reduction.
(a)
(a)
(b)
FIGURE 4. Fabric image intensity: (a) before and (b) after noise
reduction.
(b)
In addition, in order to separate weft yarn (local
maximum) from the remaining noise in the fabric
image intensity graph, local maxima amplitude was
calculated. If the amplitude was less than a quarter of
the maximum amplitude in the graph, the local
maximum would be considered as noise, otherwise it
would be counted as a weft yarn.
FIGURE 3. Fabric image: (a) initial image, (b) gray scale image
after processing.
Journal of Engineered Fibers and Fabrics
Volume 10, Issue 4 – 2015
31
http://www.jeffjournal.org
Control Unit Design
The on-line weft density control program was written
in MATLAB. When the monitored weft density did
not match the set point value, the error signal was
generated and the controller produced a
compensation command accordingly. The output of
the PID controller was an 8-bit coded integer. This
digital number had to be converted to an analogue
value. A special electrical circuit was designed to
convert the controller binary output to the analogue
voltage on a scale from 0 to 10 volts. This electrical
circuit interfaced with a PC via a parallel port. The
generated voltage was fed into the inverter to regulate
the take-up motor speed.
FIGURE 5. The comparison of measured and simulated model
outputs.
Designing the PID controller included the following
steps:
Tuning the Controller
In designing a PID controller, one of the most
important steps is to establish a proper value for
controller gains. The Ziegler-Nichols tuning method
[26] was used to calculate controller gains. This
method employs an experimental technique for
setting the PID gains.
Designing a Simulator
Tuning the weft density control system involved
fabric production with different weft densities. This
process was costly and time consuming. Therefore, a
computerized simulator was designed in which
virtual data replaced the weaving machine and the
measuring system operation in the control loop.
According to the Ziegler-Nichols method, ki and kd
gains were initially set to zero. kp was increased until
it reached the ultimate gain in which the output of the
loop started to oscillate. Both ultimate gain and
oscillation period were employed to set the gains by
using the mathematical relation as mentioned in
Table II.
First, the simulator was applied to adjust the gains of
the PID controller and then the controller was
implemented on the real weaving loom.
The simulator algorithm was programmed in
MATLAB. To ensure the accurate function of the
simulator, the supplied voltage into the weaving
machine and the input value of voltage to the
simulator were manually changed in one step. Figure
5 represents the outputs of the simulator in
comparison with the measured value on the real
loom.
In the first step, the controller gains were obtained by
using the simulator and the control system was tested
virtually at various set points. The results showed that
in the case of set points beyond 30 picks/inch, the
control system responses were not proper and
oscillated around the steady state value as shown in
Figure 6. This was the result of the time delay
occurring in the weft density control system. Time
delay was the result of fabric transfer from cloth fell
to the measuring point implying that the controller
was only sensitive to the retrospective actions and
decided based on the previous events. To modify the
time delay effect, the Smith predictor, an algorithm
for assessing the stability in the control system
affected by long dead time as proposed by O. J. M.
Smith in 1957 [27], was used.
TABLE II. Mathematical equations for PID controller tuning.
*Ku is the ultimate gain and Pu is the oscillation period.
Journal of Engineered Fibers and Fabrics
Volume 10, Issue 4 – 2015
32
http://www.jeffjournal.org
A basic construction of the Smith predictor is
illustrated in Figure 7. The strategy of the Smith
predictor comprises an ordinary PID control system
and a dead time compensating algorithm in the inner
loop which estimates the process variation in the
absence of disturbance. The algorithm consists of
two terms; the first term predicts the process behavior
in the absence of the disturbance and the second
predicts the process behavior in the absence of both
disturbance and dead time.
A new simulator was designed and attached to the
previous control loop as the inner loop (predictor
unit). This new simulator was able to predict the
weaving process in both presence and absence of
dead time. After attaching the predictor to the PID
control loop, the controller was tuned again based on
the Ziegler-Nichols method. The result of tuning is
reported in Table III. These data were extracted
according to the mathematical relation mentioned in
Table II. As illustrated in Figures 8 and 9, the control
system showed more moderated response to the input
change. Figure 8 shows the virtual output of the
control system that was estimated by the simulator
and Figure 9 shows real loom output.
FIGURE 6. Output of the simulated model for ordinary PID
controller in response to 3 picks/inch variation from the set point.
FIGURE
7.
The
Scheme
of
Smith
predictor
[27].
TABLE III. PID controller gain after inserting Smith predictor.
Implementation of the Weft Density On-Line Control
System
To develop the weft density on-line control system,
appropriate changes were made to the weaving loom.
Mechanical motion transfer to the take-up roller from
the main motor was replaced by an AC motor to
provide the ability to control the take-up roller speed
automatically, and the image processing equipment
was installed on the weaving loom. The previous PID
controller and inner loop (predictor unit) were
applied to the real loom. The response of the control
system to the input change with time was
investigated on the real loom. The results showed that
the control system did not work properly on the real
loom, and thus, the system stability requirement
could not be satisfied. Therefore, three PID
FIGURE 8. Output of the simulated model after inserting Smith
predictor to the control system.
Journal of Engineered Fibers and Fabrics
Volume 10, Issue 4 – 2015
33
http://www.jeffjournal.org
gains were manipulated and the response of the
system to the input change was considered. The
control system output showed that reducing the
derivative gain to half of its initial value provided
much more stability in the control system.
In order to improve the stability of the control system
and moderate the variation of the weft density, a
linear filter, as described in Eq. (3), was used.
yi = γ × xi + (1 − γ ) × xi −1
(3)
FIGURE 9. Response of the control system to variation of take-up
roller speed.
where γ is the experimentally obtained gain equal to
0.6, xi and xi-1 represent the current value and the
previous value of the weft density, respectively, and
yi denotes the new value weft density. This filter was
operated on the measuring system and the Smith
predictor outputs.
FUNCTION EVALUATION OF THE WEFT
DENSITY CONTROL SYSTEM
In this step, the function of the weft density control
system was tested and the following experiments
were performed to determine the ability of the control
system in error compensation.
FIGURE 10. Response of the control system to variation of takeup roller speed. WD is the measured weft density, WDF is the
predictor output in the absence of disturbance, and WDT is the
predictor output in the absence of both disturbance and time delay.
Response of the Control System to Variation of the
Take-up Roller Speed
The most effective factor on the weft density is the
take-up roller linear speed. Hence, the control system
is expected to correct the take-up roller linear speed
variation that could be caused by various
disturbances. Therefore, the take-up roller speed was
varied manually producing fabric with different weft
densities from the set point. The control process was
paused and the supplied voltage into the take-up
motor was changed. After one minute, the control
system was run once again. As shown in Figure 9,
the control system was able to compensate the step
changes in the weft density and return it to the set
point value of 25 picks/inch. Furthermore, as can be
seen in Figure 10, both terms related to the output of
the Smith predictor were added to Figure 9.
Journal of Engineered Fibers and Fabrics
Volume 10, Issue 4 – 2015
Fluctuation in Electric Power Feed to Take-up
Motor and the Response of the Control System
Variation of the electric power feed to the take-up
motor from the set value led to the variation of the
take-up roller linear speed. To change the output of
the electric circuit voltage, the controller output was
multiplied by a constant value. This was similar to
fluctuation in the feed electric power to the take-up
motor. Figure 11 illustrates a sample of the control
system response to the variation of take-up roller
speed when the set point value was 21 picks/inch.
This figure shows that the control system could cope
with this kind of disturbance.
34
http://www.jeffjournal.org
image processing technique in a PID control system
on the weaving loom was investigated. A machine
vision system was developed for on-line weft density
measurement based on the computational image
processing routine. Take-up roller speed was
considered as the manipulated element to adjust the
weft density. Moreover, a simulator was designed
and implemented in the control loop to virtually tune
the PID controller gains. The results indicated that
there was a dead time between the fabric formation
and weft density measurement. To surmount the dead
time problem, the PID control loop was modified by
the Smith predictor. The control system elements
were installed on a real weaving machine and tested.
The results of set point tracing and regulatory control
in various conditions demonstrated an acceptable
performance of the control system. The system
adjusts the weaving machine function in proper time
and fabric with uneven weft density is produced in
acceptable distance. Such distance depends on set
point values.
FIGURE 11. Response of the control system after multiplying by
1.2 in the controller output.
Response of the Control System to the
Inconsistency of Weaving Machine Components
Synchronizing between the weaving machine and the
take-up motor is a necessary prerequisite for
production of an even fabric. The system
performance against the inconsistency of the weaving
machine and take-up motor was investigated.
Accordingly, the weft insertion rate was changed
from its programmed value of 240 picks/min to 195
picks/min in the control system. The control system
compensated for the disturbance effect by adjusting
the take-up motor speed. Figure 12 shows the
response of the control system when the set point
value was 21 picks/inch.
REFERENCES
[1] Islam, S. A., "Features Identification of Set
Marks in Weaving," Canadian Textile
Journal 2001, 118, 28-31.
[2] Greenwood, K.; Cowhig, W. T., "The position
of the cloth fell in power looms, part I," The
Journal of the Textile Institute 1956, 47,
T241-T254.
[3] Greenwood, K.; Cowhig, W. T., "The position
of the cloth fell in power looms, part II," The
Journal of the Textile Institute 1956, 47,
T255-T273.
[4] Greenwood, K.; Vaughan, G. N., "The position
of the cloth fell in power looms, part III," The
Journal of the Textile Institute 1956, 47,
T274-T286.
[5] Greenwood, K.; McLoughlin, W. T., "The
design and operation of a loom with negative
beat-up," The Journal of the Textile Institute
1965, 56, T31-T328.
[6] Stemheim, A.; Grosberg P., "Weaving with
constant beat-up force," The Journal of the
Textile Institute 1991, 82, 317-323.
[7] Stemheim, A.; Grosberg P., "The effect of sley
motion on beat-up force," The Journal of the
Textile Institute 1991, 82, 325-331.
[8] Eren, R.; Porat, I.; Greenwood, K., "A
mechatronics approach to the control of
pickspacing," Mechatronics 1995, 5, 165182.
FIGURE 12. Response of the control system to the change of weft
insertion rate.
Figures 9-12 show that the weft density control
system could compensate for the deviation of weft
density from the set point originated by different
sources.
CONCLUSION
By expanding the application of technical fabrics in
all industrial fields and considering the necessity of
evenness in these fabric types as well as enhancing
industrial production rate, the application of the
Journal of Engineered Fibers and Fabrics
Volume 10, Issue 4 – 2015
35
http://www.jeffjournal.org
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
Jeddi, A. A. A.; Nosrati, H.; Ordoukhany, D.;
Rashidian, M., "A Comparative Study of
Electronically and Mechanically Controlled
Warp Yarn Let-off Systems," Indian Journal
of Fiber and Textile Research 1999, 24, 258263.
Ordoukhany, D., Analyses the Relationship
between Warp Beam Rotation and Let-off
Tension with Electrical and Mechanical
Controller in Weaving Loom (MS Thesis)
1996, Department of Textile Engineering,
AmirKabir University of Technology,
Tehran, Iran.
Dayik, M.; Kayacan, M. C.; Calis, H.; Cakmak,
E., "Control of warp tension during weaving
procedure using evaluation programming,"
The Journal of the Textile Institute 2006, 97,
313-324.
Nosraty, H.; Jeddi, A. A. A.; Kabganian, M.;
Bakhtiarinejad, F., "Influence of Controlled
Weft Yarn Tension of a Single Nozzle Air-Jet
Loom on the Physical Properties of the
Fabric," Textile Research Journal 2006, 76,
637-645.
Nosraty, H., Characterization of Weft Yarn
Tension Variation for Air-Jet Loom and
Determination of the Effect of Controlled
Weft Tension on Physical Properties of
Fabric (Ph.D. Thesis) 2000, Department of
Textile Engineering, AmirKabir University of
Technology, Tehran, Iran.
Kuo, C. F. J.; Su, T. L.; Chen, C. H., "Dynamic
modelling and control of a beat-up
mechanism," Polymer-Plastics Technology
and Engineering 2008, 47, 367-375.
Lin, J. J., "Applying a Co-occurrence Matrix to
Automatic Inspection of Weaving Density for
Woven Fabrics," Textile Research Journal
2002, 72, 486-490.
Ravandi, S. A. H.; Toriumi, K., "Fourier
Transform Analysis of Plain Weave Fabric
Appearance," Textile Research Journal 1995,
65, 676-683.
Bugao, X., "Identifying Fabric Structure with
Fast Fourier Transform Techniques," Textile
Research Journal 1996, 66, 496-506.
Chan, C. H., "Fabric defect detection by Fourier
analysis," IEEE Transactions Industry
Applications 2000, 36, 1267-1276.
Tunak, M.; Linka, A.; Volf, P., "Automatic
Assessing and Monitoring of Weaving
Density," Fibers and Polymers 2009, 10,
830-836.
Journal of Engineered Fibers and Fabrics
Volume 10, Issue 4 – 2015
[20] Cardamon, J. M.; Damert, W. C.; Phillips, J. G.;
Marmer; W. N., "Digital Image Analysis for
Fabric Assessment," Textile Research
Journal 2002, 72, 906-916.
[21] Jeong, Y. J.; Jang, J., "Applying Image
Analysis to Automatic Inspection of Fabric
Density for Woven Fabrics," Fibers and
Polymers 2005, 6, 156-161.
[22] Sodomka, L.; Komrska, J., "Laser Diffraction
Measurement of Parameters of Fine-Mesh
Woven Textile," Textile Research Journal
1991, vol. 61, pp. 232-236,.
[23] Ribolzi, S.; Merecklr, J.; Gresser, J.; Exbrayt, P.
E., "Real-Time Fault Detection on Textile
Using Opto-electronic Processing," Textile
Research Journal 1993, 63, 61-71.
[24] Islam, A. T. M. S.; Bandara, M. P. U., "Yarn
Spacing Measurement in Woven Fabrics with
Specific Reference to Start-up Marks," The
Journal of the Textile Institute 1996, 87, 107119.
[25] Kianiha, A.; Sheikhzadeh, M.; Semnani, D.;
Dabirian, F., "An investigation into the weft
insertion periodic errors in fabric using an
image analysis method," The Journal of the
Textile Institute 2010, 101, 457-462.
[26] Bolton, W.; Control Engineering; Longman
Scientific and Technical: 1992
[27] Wolfgang, A.; Practical Process Control for
Engineers and Technicians; ELSEVIER:
2005.
AUTHORS’ ADDRESSES
Atieh Almasi Zefrehyee
Mohammad Amani Tehran, PhD
Hooshang Norsaty, PhD
Marzie Abhajani, PhD
Amirkabir University of Technology
Hafez Street
Tehran 0098
IRAN
36
http://www.jeffjournal.org