Defects that reduce the expected performance of the fabric or that are easily seen and not accepted by a prospective buyer when they occur in a prominent position in a product made of fabric are called 'fabric defect'.
These errors occur due to mechanical reasons or the yarn used during the weaving of the fabric. The raw fabrics produced in the fabric weaving enterprises are examined at the fabric quality control tables after the weaving process. Despite the developments in weaving technologies, fabric defects still constitute a significant cost. Fabric defects are controlled by experienced quality control personnel in textile weaving enterprises. .
According to the information obtained and learned from the literature, an experienced quality control staff can detect only 60-70% of the errors. The fabric, which is produced with 2-96% efficiency on the quality control table with a width of about 99 meters, can only be controlled by wrapping it at a speed of 30 m/min. The quality evaluation of fabrics is not objective and a statistical evaluation cannot be made about fabric defects.
AUTOMATIC FABRIC QUALITY CONTROL METHODS
Fabric quality control systems can consist of three process steps:
• Feature selection
• To compare
Automatic on-line fabric inspection systems use the following methods to detect fabric defects:
• Advanced Image Analysis
• Neural Networks
• Wavelet Packet Model (Best Wavelet Packet Model)
• Fuzzy Logic Method
There are three main approaches to identify errors:
1. Feature tree
2. Numerical identification
3. Technical structure
The feature tree approach describes the appearance of defects in manual fabric inspection.
The numerical identification approach specifies the length and width of the error. Control methods are divided into categories according to their active and passive nature. The active method needs the sensor's illumination source.
Automatic fabric quality control methods are based on light reflection, laser light or video image processing. These automated systems operate a simple photocell scanning or capacitance measuring device on the surface of the fabric. Charge Coupled Device CCD cameras, which are widely used in woven fabrics, have a resolution of 2.048 and 4.096 pixels.
Such systems consist of CCD camera, analog signal to digital signal converter, image (picture) storage unit, digital to analog converter for input to monitor, central processing unit (CPU). They are classified in standard image filters (eg low-pass filter) and the samples are characterized using Fourier Transforms. Many algorithms depend on processing time; Therefore, special purpose powerful computational hardware is needed for fabric control.
IMAGE ANALYSIS METHOD
The main purpose of fabric defect control is to extract only the information that is of interest to the researcher from the very detailed information contained in an image and to evaluate the information. As a fast and effective method, image analysis systems can be applied on various textile products for various measurement and control purposes. Image analysis is a term used to describe the operations performed on images for a specific purpose. These purposes are for example; It may be to put an image in a form that it can be carried more easily or to place it in a computer memory, or it may be extracting only the information that a viewer is interested in. Commonly used image processing operations; It consists of filtering, sampling, classification, coding, feature extraction, pattern recognition and motion prediction. If the data obtained during digital scanning of a fabric differs when compared to scanning of a defect-free fabric under the same conditions, the presence of an error will be clearly identified. However, a wide variety of patterns, weave structures, colors and different types of defects on the fabric will complicate this process. Therefore, the whole process should be simplified step by step. First of all, the colored structure on the fabric should be reduced to a single color with a gray scale image, then other processing steps should be applied on this image.
There are two basic stages of error detection by image analysis. In the first stage, which we can also call the learning stage, the practical limits of the relevant parameters of the fabric in question are calculated on the basis of a fabric that has no faults, and a classification is made for each feature.
In the second stage, which is the inspection stage, it is checked by comparing whether the relevant features are within the limits given in the previous stage. The magnitude of any error is determined by the rate of deviation from predetermined limits. A typical image processing system generally consists of three basic parts. These;
1. Image acquisition unit: This unit consists of a TV camera, A/D and D/A converters and a digital memory. A digital memory has a gray scale of 8 bits (256 digits) and a capacity of 1024 to 1024 pixels.
2. Image processing unit: This unit mainly consists of a microcomputer system, and the interface circuit in the image store is connected to this system together with the disk drive, printer and monitor. The main part of the process takes place in this unit. The process consists of three simple parts:
• Image enhancement,
• Image analysis,
• Image encoding (digitizing)
3. Monitor (Display Unit)
FUZZY LOGIC METHOD ( Fuzzy Logic ) :
In 2001, Choi et al. used the fuzzy logic method in their study based on the wide variety of fabric defects and the uncertainty in their definitions. Fuzzy logic is defined as a mathematical order established for the expression of uncertainties and working with uncertainties. In the system where fuzzy logic rules are used, membership functions for these rules are adapted to the neural network approach. In the research, several fuzzy rule assumptions were made in order to determine the regions without defects, regions with fish (in the weft and warp direction), neps and composite errors.
A created fuzzy rule base has been replaced by many crisp-classicical rules. In classical system-based complex systems, a large number of rules had to be applied to make the right decisions, and fuzzy rule approaches provided the opportunity to create reliable rules by using a narrower rule base. A 4-step process module was created in the study with the idea of determining fabric defects with fuzzy logic system. These:
2-fuzzy rule base,
4-They are defuzzification processes.
At the end of these four stages, the error types in the faulty area in the test fabric image could be determined. In the first stage, the most prominent features of the errors are determined and after the determination of the determining features, their membership functions had to be formed, that is, they had to be blurred. Input parameters were transformed into fuzzy sets created based on knowledge and experience. The measurements on the plain fabric image containing neps and fish defect are taken as variables and these values are used to form the fuzzy set. Before fuzzing, classical linguistic explanations are used to create a membership function, and three field levels as small, medium and large are created in the fuzzy logic system to identify errors.
1-Small area error free
2-medium size area neps or fish defect
The 3-large area indicates the presence of a composite error.
terms used to express
In the second stage, a fuzzy rule base was created and error types were defined with fuzzy rules. In order to reduce the probabilities to a certain number in the formation of the fuzzy logic rule base, some assumptions were made in a way that would not affect the result. By using these rules, the fuzzy logic rule base, which is the basis of fuzzy logic, can be determined. In the third stage, fuzzy decision making process was carried out based on the fuzzy logic rules. Finally, it was necessary to perform the clarification process in which the obtained fuzzy values were converted into odd numbers as a result of a fuzzy set operation. As a result; The position, number and type of the error, which is the output membership function showing the error against the input membership functions, were found. Fabric defects could be classified in the research in which Choi friends used the fuzzy logic method. It has been observed that this method, in which fuzzy logic rules are used in fabric defect control, is similar to the ability of humans to recognize faults, and that the method gives better results than classical methods. The neural-fuzzy system was used in the research conducted by Huang and Chen in 2001 to classify fabric defects. This system is created from fuzzy logic technique and back propagation learning algorithm of neural networks. As neural networks and fuzzy logic systems complement each other, an increase in classification ability has been observed. Blurring is a non-linear mapping operation. The neural fuzzy system was operated by separating the uncertain data and overlapping boundary points between the classes. In the study, plain fabrics, eight of which are faulty and one of which is faultless, were used. These errors are; weft break, warp break, double warp, double weft, hole, thin weft tape, oil stain and cobweb. The images of the fabrics were taken at a resolution of 512 x 512 pixels and three classes were formed by proportioning the dimensions of the defects in the vertical and horizontal directions. Another feature used for classification, apart from error sizes, is image density. For example; It has been observed that the double warp yarn has a higher intensity than the warp break. The image intensity was expressed as the mean and standard deviation of the gray-scale distribution, and the gray-scale mean and standard deviation values of the faulty area were taken as the defining feature. Input parameters are fuzzy using triangle membership functions. The input values used as the defining feature are between 0 and 1 numerical values. Output values in the test phase were calculated according to the data from the samples in the learning phase.
CYLOPS CONTROL METHOD
As Barco's automatic control system on the loom, it identifies warp and weft defects with the help of a moving camera system integrated into the weaving machine. When a warp or weft error occurs, the system stops the loom, the warning lamp lights up, shows its location on the loom with the help of the microprocessor, and the loom is stopped until the "error has been corrected" warning by the weaver. All defect information regarding the Barco Weave Master monitor system is sent to the fabric quality database. Thus, defect maps (maps) and various types of quality reports are created. The control system on the loom is linked to the microprocessor of the weaving machine. Each fault time is recorded with history. Thus, the error in the fabric roll and more detailed information are included in the quality reports. The Cyclops control system is easy to connect to the Barco weavemaster. The Barco monitoring system can be included in the Cyclops automatic on-machine monitoring system. By using TCP/IP protocol, the HUB and the Bench and Barco monitoring system are connected to each other via Ethernet. The fabric is divided into qualities according to the number and concentration of defects determined, and the raw fabric passes through the control and can be sent to other processes.
Control unit on the bench,
from the camera and lighting unit,
and motion picture acquisition
consists of a title.
During the determination (testing/learning) phase, the image of the fabric in a certain number of positions of the fabric is taken and transferred to the image processing unit. Here, special algorithms are used to analyze the weave of the fabric and deviations from the standard are determined as errors. Scanning range is automatically adjusted according to fabric position and width with automatic detection of fabric borders. Illumination and camera placements are adjusted by the calibration of the software module according to the optical characteristics of the fabrics. The structure of the fabric is determined automatically by calculating the algorithm parameters for optimal defect detection. The Cyclops scan head contains a camera and lighting system. The system detects 70% of fabric defects.
• 100% detection of continuous errors such as drawing-in, carding error, warp breakage.
• 100% detection of weft errors in all widths such as weft run-out, double weft.
• By stopping the loom, detecting and preventing fault formation immediately and preventing faulty fabric production.
• Performing a control independent of human errors.
• High fabric quality, less secondary fabric formation.
• Less workload in raw fabric control section.
• Automatic control of fabric on the loom.
• Support with Weave Master Machine monitoring monitor.
• Scanning speed: 18cm\sec.
• Ease of installation.
• Maintenance free
Qualimaster supports raw and finished fabric controls. Apart from these, it determines the error maps for each fabric ball cut and determines the corresponding error scores. Qualimaster prints labels at the end of each ream or cut of fabric. Together with the Weavemaster production monitoring system, weaving plans are combined with Qualimaster's online control on the loom. Depending on the control panel and the cutting tool, this terminal creates a Windows-based graphical user interface for error input. Error codes appear on the screen as buttons. The controller records the error by touching the button corresponding to the error. Special buttons are possible for ongoing faults, cuts and repairs. Optionally, different error codes can be entered from the keyboard.
An error map can be created, the score of which can be calculated for each fabric ball.
Interfaces • Allows barcode reader for each fabric roll definitions • Error marking apparatus • Scales for recording the fabric ball weight • Software for piece labels and defect maps. Each entry is automatically connected to the weft counter for fabric control on the loom. They provide the formation of the defect map during weaving. Qualimaster fabric control software includes extensive reports from the warp preparation to the yarn supplier depending on the weaving.
Qualimaster Report Types
Type Defect Analysis Different fabric types can be compared in terms of defect rates and scores. Ten major errors for each type are shown by Pareto analysis.
Inspector Performance Report At the beginning of each shift, the inspector fabric report marks the starting point. This report shows the checked length and by which operator the errors were made. Managers can compare several controllers for the same fabric type over a long period of time. This type of report can be used by the quality control manager to identify inspectors who need more training for each type and run slower than average. Qualimaster optimizes dividing large balls into smaller lots. If the control and cutting are done in two separate steps, Qualimaster determines the cutting place of the fabric from the defect map and the scoring criteria. According to customer requests, minimum/maximum fabric length and error points are specified.
Elbit Vision I-Tex Control System
It can detect errors at 300 m/min control speed. First of all, the size and coordinate locations of the error are determined and the location of the error is recorded on the error map, and the digital image of the error is stored. The cost of this system depends on the working speed and fabric width.
Loom-tex system elements
• All en control with video scanner
• Combined with dual lighting module
• Process created for each Machine
• For the monitoring method, the central computer can work up to 390 cm width of the workbench.
Identifiable error types:
- Weft slack
- Double warp
- Drafting error
- Double scarf
- Warp break
- Comb track
- Posture marks
- Warp end
- Oil stains
- Dense-sparse weft
- Bad edge
- Thick place
- Weft stacking
Operates the I-Tex 1000 video control system; automatically detects errors, saves them, locates them and grades them later. This system can detect even 0,5 mm errors. Denim fabrics, raw fabrics and monochromatic dyed fabrics can be checked. It can be successfully applied in the fields of non-woven, coating, carbon web, composite materials, metal lamination, plastic, paper.
USTER FABRICSCAN CONTROL SYSTEM
It uses the latest neural network technology. Depending on the fabric type and characteristics, errors with a speed of 90 m/min and a minimum size of 0.3 mm can be detected. It works in the fabric width range of 110 – 440 cm. The error classification system Uster Fabriclass easily identifies the following profiles. Depending on the size and nature of the error, error types are categorically divided into their numbers and types. According to specific requirements, disturbing and non-irritating error classes are created throughout production. Error position, error measure, and error type functions are used in error detection. Other measurements taken are fabric length and width. Fabrics can be raw, finished (dyed), plain, twill and satin. Error classification can be done in two ways as Demerit Error Scoring and Uster Fabriclass. Summary report, machine report and shift report can be produced for process analysis. The system can display the error picture.
Two or more special types of video CCD cameras are used for scanning the fabric, depending on the fabric width. The first phase takes a few minutes as a learning process. The size and location of the errors are recorded and divided into three classes:
3-Surface defect) and its degree
determines. The error classification system created using the feature tree approach includes 100 different types of errors. In image processing, the image image scanned by the video camera is converted into a digital image by the analog to digital converter. The two-dimensional array can be analyzed using image processing techniques to reveal errors. Each error is examined with the help of a combination of standard image processing algorithms and filters. The error type can be determined with the help of numerical definition approach from error classification. The techniques developed for error detection are very numerous and classified according to their approach. Each error is saved as an image. It can be viewed to a large extent and can be printed out with the error report. The fabric is passed between two composite lighting modules that allow reflected and transmitted light to pass. The choice of lighting type depends on the fabric density. With the control of a new commodity, the learning phase, in which the system registers the normal appearance of the fabric and moves on to the internal learning phase, takes place in the first meter of the fabric. The internal learning phase takes about 1 minute and must be performed for each commodity in advance. All parts of the same good can be controlled with the same standard. In summary, it can be said that the error detection rate will increase with the following 3 measures:
• Controlled operation with high camera resolution and fast imaging operation
• A neural network-based, learning-capable system to detect defects in fabric
• A simple distinction of defect characteristics based on the difference in length and color of the defect Zellweger Uster's automatic fabric inspection system performs the following functions:
• Learns the characteristic features and certain faults of the fabric.
• It determines the location of the faults in the fabric.
• Marking. • It records, that is, memorizes it. It automatically analyzes and classifies error information.
SYSTEM CHECK REPORTS
Apart from the measured length and width of the fabric, it also gives information about the number of errors in the fabric roll and per 100 meters. Depending on the FABRICLASS and the fault type, the exact position and size of the fault are displayed, as well as the classification of the faults. If an error is found particularly interesting, one click on it is all it takes to show the exact form of the error.
It gives a quick view of the defect in the fabric. In particular, the frequency, size and position of the fault can be seen with a quick glance. Along with the pattern of errors, this report and the next report aid in process optimization.
Error Type Report
The user can see if a certain type of error occurs too often. In addition, the design of the report is designed to show whether a particular type of error occurs at a frequency higher than the average frequency of the good.
It gives information about the frequency of errors in each class. The list shows annoying and trivial errors by color differences and lengths. The frequency of non-significant defects becomes significant when the typical appearance of the fabric increases until it changes.