A camera (inset) is mounted at the top of this A-frame structure, which is part of the demonstration unit for computer-aided fabric evaluation. Behind the A-frame is a closeup of the fabric that summer student Jessica Bayliss is inspecting.



THE TEXTILE industry in the United States is huge--it employs more people than any other manufacturing sector and accounts for the most consumer sales among durable and nondurable goods. The U.S. leads the world in growing cotton and is very strong in the production of many kinds of fibers and finished cloth. But most sewing of finished apparel has shifted overseas, and overseas competitors are cutting into our predominance in fiber production and weaving. Over the last ten years, 400,000 American jobs have been lost to this intense overseas competition.
At first glance, a CRADA (Cooperative Research and Development Agreement) between the American textile industry and the Department of Energy's national laboratories might seem an odd match. It was born during a 1992 DOE-sponsored workshop on critical industries, when representatives from textile firms and DOE laboratories realized how much they had to offer each other. The laboratories have developed technologies in energy, environmental cleanup, and national defense that can help the U.S. textile industry improve its competitiveness in worldwide markets and create jobs as well by increasing quality, reducing costs, improving responsiveness and production times and by reducing the environmental impacts of manufacturing operations. Through its involvement in this CRADA, the DOE can obtain more information about market demands and needs in areas of communications, networked information, imaging, sensing, tagging and tracking, and environmental cleanup.
This partnership, known as AMTEX, began in 1993 as a collaborative program among DOE and eleven of its national laboratories and several textile research organizations. The entire textile industry is represented, including the retail, sewn products, textile manufacturing, and fiber production sectors.
Lawrence Livermore National Laboratory's contribution to the AMTEX partnership falls into five areas:
  • Demand-Activated Manufacturing Architecture--For this project, Livermore is developing the first-of-its-kind client authentication software, which is discussed below.
  • Computer-Aided Fabric Evaluation--Livermore-developed sniper detection technology is being applied to this project, which is discussed below.
  • Electronically Embedded Fingerprint--Livermore worked on developing miniature electronic devices for permanent identification and inventory management. Manufacturers will be able to read and write information about size, color, style, process history, country of manufacture, etc. on a device the size of a grain of rice. This "fingerprint" may replace the bar code, which is not permanent and is limited in the amount of information it can store. A larger version of the fingerprint was originally developed by Livermore for permanent identification of weapons and other treaty-controlled items.
  • Textile Resource Conservation--Livermore is developing and testing new processes to conserve the huge amounts of water used to dye and finish fabrics. Using less water in processing also means less downstream water treatment.
  • Rapid Cutting of Textiles--Using its expertise in lasers, Livermore has developed a very-short-pulse, solid-state laser to replace the blade that has been used for decades to cut garment components from piles of fabric. The laser is faster than the blade and could be used for rapid custom cutting. The system is particularly useful for cutting very strong fabrics like Kevlar, which is used for sails and bullet-proof vests. Unfortunately, at present the laser cutting system is too expensive for small apparel manufacturing companies, which make up the bulk of apparel manufacturers in the U.S.

Demand-Activated Manufacturing Architecture

The two devils of the apparel retail market are undersupply and oversupply, both of which are dictated by consumer demand. Undersupply results in empty shelves and fewer sales, and oversupply in price markdowns, wasted resources, and lost profits. In either case, the retailer's bottom line is adversely affected. With an eye to improving the entire U.S. textile marketplace, AMTEX envisions a secure, Internet-based information system to link all sectors of the textile supply chain. The key to this project is securely and selectively communicating demand information from retail companies back to apparel makers, textile manufacturers, and fiber producers. From analysis of point-of-sale data and sophisticated simulations of the entire industry, decision makers will be able to bring the right products to market at the right time at a competitive cost. The underlying assumption is that the entire textile pipeline has to operate more efficiently if there are to be significant gains for each of the four sectors. The figure below gives an indication of the size of the industry and the volume of data the industry generates.



The amount of secure "demand" data to be transmitted over the Livermore-designed TEXTNET data transfer system will be enormous. Ten thousand retail companies with 100,000 stores, each with 25,000 to 1.2 million stockkeeping units, generate 20 billion apparel and household textile purchases annually worth about $200 billion. That translates into electronic communication between apparel and textile manufacturers about the demand for 100 million industry stockkeeping units per year.


Several pairs of industry partners have been electronically transferring purchase orders and advance ship notices for several years over proprietary networks. These are typically "push" transfers in which, for example, a retailer provides sales information about a particular brand of pants to its manufacturer. The provider of the information initiates the transfer to the client and determines what data will be transferred and when.
Livermore is working with Idaho National Engineering Laboratory to develop TEXNET, which incorporates a demand-activated or "pull" transfer of data initiated by the client. For example, apparel manufacturers will have "trading partner agreements" with a multitude of retailers and will regularly request sales and other information from them. Perhaps weekly, Brand A pants manufacturer will request sales information about Brand A pants, and Brand B pants manufacturer will do the same about theirs. Users will be assured that the information is provided not just securely but selectively as well--Brand A manufacturer must not receive information about sales of Brand B pants. Such a system is currently impossible because there is no method for authenticating the identity of a client, i.e., for a retailer to electronically assure that the requester of information about Brand A pants is really Brand A manufacturer and not someone else.
Livermore is developing software for client authentication, which will make possible secure, selective transfer of data among multiple users. With the advanced encryption codes incorporated in this software, a data provider will be able to electronically assess a client's identity and validate its trading partner agreement before releasing client-specific information. The major challenge in designing the system has been the development of tools for implementing necessary security mechanisms.
Livermore is designing the whole TEXNET data transfer system, of which client authentication and management of trading partner agreements are parts. Because a client is receiving only a portion of the total data available, the bandwidth required for data transfer can be narrower than if all the data were to be transferred. Storage and processing requirements for the transfer system are also relatively small because only selected pieces of data are transferring at a time.
TEXNET is designed for the real-time exchange of virtually any type of information and may be customized by its users. Using the Internet, information can securely flow in any direction in the textile pipeline.

Computer-Aided Fabric Evaluation

Several years ago, Livermore developed a sensor that can track the path of a bullet as it flies through the air. This sniper detector can read the bullet's unique signals two hundred times a second from any direction and then track the bullet's path back to its source.
This same real-time, image-processing technology is being used for computer-aided fabric evaluation. Instead of looking at a bullet flying through the air, a camera and computer look for flaws in patterns being printed on fabric or in knitted fabrics. The system can not only detect defects in printing or knitting but can also immediately classify the defect and signal the machine operator that a problem exists and how to correct it.
Quick detection and correction of flaws will bring huge savings to textile manufacturers. For example, high-speed, fabric-printing machinery can handle both narrow and wide fabrics and may print hundreds of yards of material in just a few minutes. Lightly glued to a mat to provide an unmoving, stable surface, the greige fabric (pronounced "gray" and meaning undyed, unfinished fabric) passes under a series of roller screens, each of which prints a different color onto the fabric. A screen could slip slightly out of alignment causing a "misfit" where the dye is shifted on the fabric, a screen could become clogged with dye creating an unprinted area, or lint or thread could adhere to a print screen, also resulting in an unprinted area.
The proprietary system developed at Livermore mounts a high-speed, line-scan camera, or series of cameras depending on the fabric's width, over the fabric immediately after the last screen. (See photos on p. 20.) As the first yards of printed fabric roll beneath the cameras, the computer to which the camera is attached digitizes and dynamically learns the printed pattern by creating a model of the repeated pattern. The computer, powerful enough to handle the huge amounts of data that the process generates, can then inspect the printing process on line.
As it detects and diagnoses flaws, the computer accumulates a history of defects, which the computer draws upon in its reporting to the machine operator. For example, a one-time flaw might not initiate an alarm, but a repetition of small flaws or a large flaw would. Based on the array of defects that are known to occur, the computer's program determines what the problem is and what the operator's response should be. Depending on the particular setup, the operator is notified of the problem by bells, lights, or a printout on a computer monitor.
As the manufacturing facility's central computer accumulates data from the dedicated computers at the various fabric printers, quality control should improve considerably. Managers will have a record of problems with specific printing machines and with particular operators.
Livermore's system to detect flaws in knitted fabrics is very similar to that for detecting printing flaws. The flaws that crop up in the knitting process are different from those that appear in printing, but they can be dealt with by the computer in the same way.
By the time Livermore's work on computer-aided fabric evaluation is complete, pilot facilities will have been installed at the mills of several fabric manufacturers.

Key Words: AMTEX, computer-aided fabric evaluation, demand-activated manufacturing architecture, technology transfer, textiles, TEXTNET.

For further information about Demand-Activated Manufacturing Architecture contact Lansing (Chip) Hatfield (510) 422-8567 (hatfield1@llnl.gov).
For further information about Computer-Aided Fabric Evaluation contact Jose E. Hernandez (510) 423-2160 (hernandez5@llnl.gov).


Return to July 1996 Science & Technology Review