With the advent of global sourcing, the need for effective quality measurements is more important than ever, and there is a need for a comprehensive, consistent way to establish the quality of goods, for which automated fabric inspection is one such solution, stress R Guruprasad and B K Behera.

Fabric inspection has proven to be one of the most difficult of all textile processes to automate. It has taken decades for computer and scanning technology to develop to the extent that practical, consistent and reasonably user-friendly systems could be produced. Automatic inspection systems are designed to increase the accuracy, consistency and speed of the detection of defects in the manufacturing process of fabrics in order to reduce labour costs, improve product quality and increase manufacturing efficiency.

Today's automated fabric inspection systems are based on adaptive, neural networks. They can learn. So instead of going through complex programming routines, the users are able to simply scan a short length of good quality fabric to show the inspection system what to expect. This coupled with specialised computer processors that have the computing power of several hundred Pentium chips makes these systems viable. They are designed to find and catalogue defects in a wide variety of fabrics including greige fabrics, sheeting, apparel fabrics, upholstery fabrics, industrial fabrics, tyrecord, finished fabrics, piece-dyed fabrics and denim.

Manual vs automatic inspection
It has been stated that the price of the textile fabric is reduced by 45% to 65% due to defects. Wastage reduction through accurate and early detection of defects is an important aspect of quality assurance. With visual fabric inspection, a trained person inspects all types of fabrics, find and correctly identify all defects and divide them into the corresponding classes. The highest level of concentration is maintained only for a period of 20 to 30 minutes. After that, a person will tire continuously. Moreover, the highest concentration will only be achieved if the fabric is "Interesting" enough. Even in a well-run operation, the reproducibility of a visual inspection will rarely be over 50%. With automated inspection, the results are reliable, reproducible and free from the subjective deficiencies of the manual fabric inspection (Table 1).

Operating principle
The cloth passes over a two-part illumination module (Figure 1) that is designed for an inspection in either reflected or transmitted light. The choice of the illumination type is dependent on the density of the fabric, the special types of defects or the textile process stage in which the inspection is carried out. Above the light source, there are 3 to 6 or, in special cases, up to 8 CCD high-resolution line scan cameras depending on the inspection width. The cameras scan the fabric continuously for deviations.

In this process, the fabric is inspected with the resolution that is achieved by an inspection person at a distance of one metre to the fabric. The inspection system is operated from an operating terminal, where article-specific inspection parameters are set and the necessary piece data are entered or read in with a bar code reader. The various reports are also called up via, the operating terminal. The complex image processing is achieved with printed circuit boards that have been developed specifically for this application. After the inspection, the detected defects can be displayed on the screen for a quick and easy visual analysis.

On-line and off-line integration There are basically two forms of physically integrating automatic fabric inspection systems in the production process. On the one hand, there is online integration in which the inspection system is integrated in an existing production machine and on the other; there is offline integration with a built-in fabric transport. Each type of integration has its advantages and disadvantages. Advantage with online systems is the simple system and hence a cheaper machine frame, minimum floor space and the very low operating requirements, because the operator of the machine also operates the inspection system. The disadvantage is that the production machine determines the inspection speed so that it is not always possible to take full advantage of maximum throughput speed of the inspection system. With offline integration, maximum inspection speed can be fully utilised. The disadvantages are the additional machine frame with its own fabric transport drive and the higher personnel requirement.

Uster FABRISCAN
The company's current system, Fabriscan, can inspect fabric at speeds up to 120 metres per minute (Off-line) and can detect defects down to a resolution of 0.3 millimetres (Figure 2). The inspection speed of an on-line system is approximately 30 metres per minute. It can handle fabric widths from 110 to 440 centimetres. The Uster Fabriscan is available for transmitted and reflective light. Some fabric faults can be better recognised in transmitted light, other faults can be better in reflective light. Oil spots can only be seen in reflective light whereas start marks can only be seen in transmitted light.

The Fabriscan marks the fabric faults by paper tag labels and by ink. In addition to fabric grading, the Fabriscan allows the possibility to determine which faults one wish to declare, pass or repair. Basic weave constructions, cotton, cotton blends, wool and filament yarns, plain greige, denim fabrics and single piece dyed, uni-coloured fabrics all can be inspected. However, it cannot measure colour shade variations, dobby and Jacquard designs, pile fabrics (velvets and terry), knitted goods.

 What makes Fabriscan unique is that it classifies defects in a matrix called Uster Fabriclass, which is similar to the well-known Uster Classimat system for yarns. Fabriclass has two axes. On the y-axis is the contrast of the defect and on the x-axis is the length of the defect. This allows the system to tell the difference between disturbing defects versus non-disturbing defects and makes over-detection virtually nonexistent, according to the company. Data on defects can also be stored in a relational database, allowing users to generate any type of report that they need. Cut optimisation software is included to improve first quality fabric yield. The cost for Fabriscan starts at $200,000. Zellweger Uster estimates that the system has a payback of about 12 to 24 months, based on labour savings, cut optimisation, and improved flagging accuracy to customers.

Elbit Vision System's I-TEX
EVS, which introduced its I-TEX system at the ITMA'91 in Hannover, Germany, is the most established player in automatic optical inspection. The I-TEX system is capable of inspection speeds up to 300 metres per minute and can handle fabric widths up to 5 metres. The system's proprietary software algorithms have been designed to imitate the human visual system. It learns the normal pattern of the fabric and detects changes. These changes in the pattern are then analysed by multiple detection algorithms to separate real defects from random but normal variations in the fabric. Once a defect is detected, the x and y location, as well as the size of the defect, are recorded in a defect map. In addition, a digital image of the defect is saved for later review of the system operator.

They currently market four product lines designed to address the quality monitoring needs of different sectors within the fabric manufacturing industry: I-Tex, for the visual inspection and quality monitoring of woven and knitted fabrics; PRIN-TEX, for the detection of printing defects on fabric; Broken Filaments Analyzer, for the detection of filament defects in glass fabrics; And Shade Variation Analyzer, for the detection of shade inconsistencies in dyed fabric. They adapted their core visual interpretation technologies for other applications, such as those in the nonwoven fabric and printing industries, which have visual inspection and quality monitoring needs similar to those of the fabric industry (Figure 3). Potential applications in the nonwoven industry include air filtration media, diapers, surgical dressings and other nonwoven based products.

The I-TEX system cost is dependent on a number of factors such as the fabric application, desired speed and fabric width. The system sells for between $ 1,00,000 and $ 6,50,000. Payback for the system is generally between six months and 2 years, according to EVS.

Barco Vision's Cyclops
What makes the BarcoVision Cyclops system different from the EVS and Zellweger Uster offerings is that it has a traveling scanning head and can be deployed on the weaving machine itself. I-Tex and Fabriscan both inspect fabric in full width either at the batcher for greige fabrics or at the exit end of a finishing machine. Therefore, Cyclops can prevent the production of off-quality fabric by stopping the weaving process if it detects a serious or running defect. Examples of defects that would prompt Cyclops to stop a loom are running warp defects, recurring filling defects and a high concentration of local defects. Whenever the system stops off a loom, the weaver is notified. The specific type of defect and its position are displayed on the loom terminal. After resolving the defect's cause, the weaver makes a declaration on the loom terminal so Cyclops will release the loom for further production (Figure 4). Cyclops is designed to be used with Barco's QualiMaster system, all defect information, pick and time stamped, is sent to a fabric quality database.

The Cyclops scanning head includes a camera and illumination system. The camera is based on CMOS technology. The illumination system has been specifically designed to achieve optimal detection of defects in woven structures. The measuring head travels at a scanning speed of 18 centimetre per second. Proprietary algorithms run on a combination of in-house designed processing hardware and an industrial PC to carry out the image processing. The embedded software is the heart of Cyclops and runs on special purpose hardware designed by Barco. The major features of the software include: Calibration of the camera and illumination, tuning of image processing algorithms for warp/weft density and weave, boundary detection and JPEG-encoded image storage of fabric defects.

 The Cyclops scanners cost $ 5,000 each and can be used for fabric widths up to 280 centimetres. For double panel looms (up to maximum 560 cm), Cyclops comes with a double camera based image acquisition head. There is also a one-time Vision software license of $ 25,000. To estimate the payback period, Barco took the following benefits into account: Reduced manual inspection after weaving, labour savings due to less inspectors, reduction work-in-progress and reduction in off-quality fabric. Thus, the investment for an installation of 100 looms is about $ 5,25,000. The company estimates the annual savings for a typical application to be $ 2,00,000, giving a payback of roughly 2.6 years.

Future of automatic inspection systems
With the advent of global sourcing, the need for effective quality measurements is more important than ever. There is a need for a comprehensive, consistent way to establish the quality of goods and automated fabric inspection is one such solution. It should be noted that it takes years to train a good human inspector, and these automated systems can be installed and "Trained" in a matter of weeks. In addition, the digital maps that automated inspection systems provide, which reliably pinpoint defects, may well be required by the cutters, particularly if advances in single-ply laser cutters continue. Even if a company opts not to reduce the available work force, automatic inspection is still beneficial in terms of the quantity of cloth it grades and decreased amount of defects that are charged back from customers. Many of companies from developing countries like India and China have already opted for these automated systems. The ability of a mill to certify that its fabrics have undergone automated fabric inspection could very well become a requirement for certain applications or markets.

References
1. http://www.uster.com, accessed on 23/06/2008.
2. http://www.Barco.com,accessed on 23/06/2008.
3. http://www.evs-sm.com,accessed on 23/06/2008.
4. http://www.techexchange.com/thelibrary/FabricScan.html, accessed on 28/06/2008.
5. Arnold L Knoll: Automatic Fabric Inspection, Textile Institute and Industry, January 1975
6. Automatic Fault Detection, Textile Horizons, June 1988.
7. D L Munden and L Norton-Wayne: Textile Asia, October 1988.
8. Rudolf Meier et al: International Textile Bulletin, 3/1999.
 9. R N Ulrich Rosler: Melliand International, E292, 8/1992.
10. H Sari-Sarraf and J S Goddard: IEEE Transactions on Industry and Applications, Vol 35, No: 6, December 1999.

Note: For detailed version of this article please refer the print version of The Indian Textile Journal June 2009 issue.

R Guruprasad
Department of Textile Technology,
Indian Institute of Technology (IIT), New Delhi.

B K Behera.
Department of Textile Technology,
Indian Institute of Technology (IIT), New Delhi.