Checking stopper coatings for defects

Camera-based system automatically checks pharmaceutical container stoppers for coating, position, and defects.

Th 0401feat1 1

Camera-based system automatically checks pharmaceutical container stoppers for coating, position, and defects.

By Lawrence J. Curran,Contributing Editor

An automated vision system has replaced the manual inspection of elastomeric closures or stoppers for pharmaceutical containers. It provides "a great leap forward in achieving continuous improvement in the quality of products delivered to customers," according to West Pharmaceutical Services Inc. (Lionville, PA), which uses the system integrated by Applied Machine Vision Inc. (AMV; Pittsburgh, PA). It also has reduced labor costs and eliminated the subjective judgments associated with human inspectors. The installation and validation of a vision inspection system at West's Jersey Shore, PA, plant represents the company's first application of vision inspection technology to its elastomeric closures, although vision inspection systems have long been used in the company's metals and plastics manufacturing facilities.

Th 0401feat1 1
Click here to enlarge image

Vision inspection system control is directed by an SLC-500 programmable logic controller (top left). The bright spot lighting (bottom center) is activated as parts pass under it on a conveyer and are simultaneously imaged by the camera, mounted horizontally opposite the lens optics and above and to the right of the illumination. The hopper feeds stoppers into a centrifugal feeder (center right).

The vision system inspects 13-mm-diameter Teflon-treated serum stoppers. Teflon film is an important product attribute for certain West Pharmaceutical customers, which use the Teflon stoppers for several products. When the stoppers are used to cover containers, the Teflon film creates a barrier between the elastomeric stopper and the packaged drug, greatly reducing the possibility of contamination caused by drug contact with the stopper.

"The main task of the vision system is to check that the [Teflon film] barrier is in place and in the proper location," says Donald Hill, manager of the engineering department at Jersey Shore. The Teflon film is applied during the molding cycle. Before inspection, the parts are trimmed, washed, and dried. The vision system inspects each stopper to make sure the Teflon coating completely covers the plug end. The system also inspects for other defects, such as mold marks, scratches, and off-center trimming.

After an operator loads stoppers into a bowl-feeding hopper, the vision inspection process is entirely automated. The stoppers are transferred from the hopper to a bowl, where they are rotated—a step that orients the parts plug-side up and then aligns them in a single file. The parts are then conveyed to the inspection station, where a camera takes a digital image that is matched against database standards that have been scanned into a computer.

Captured images of the parts are analyzed by the vision system, and the results are subjected to the user's acceptance criteria. Accepted parts are conveyed into a bin; rejected parts are sorted out. Accepted parts are either packed for shipment to the customer or processed as ready-to-sterilize products, according to user specifications.

SELECTING A PARTNER
When West Pharmaceutical considered automating the inspection of the elastomeric stoppers in 1998, the company surveyed the field of vision-system suppliers, seeking a partner "who could provide the horsepower we needed," says Hill, who was in charge of validating whichever system was selected. Plant manager Dave Bergerstock says the field was soon narrowed from five to three candidates. Some large companies seeking the job sought substantial development fees and high prices for the finished vision system. "But we were looking for a combination of reasonable price, [vision system] experience, the ability to cooperate with us, and the company's survivability potential" over the long term, he says. Based on those criteria, AMV won the job.

Th Acf3c50
Click here to enlarge image

FIGURE 1. System PC is located below the programmable logic controller. Stoppers are fed from the hopper into the bowl feeder, where they are oriented to have their plug side up and also are arranged in a single file. They proceed onto the conveyor belt and move past the inspection station indicated by the camera and the proprietary lighting and lens apparatus. The computer controls the inspection process, vision display, and user interface. Rejected parts are routed into the reject bin, while accepted parts move along two channels to two "good-parts" bins. One bin is serviced (full bag removed and new bag installed) while the other bin is being filled, enabling continuous machine operation.

AMV president Mario Mancini believes that his company's solution got a boost when West Pharmaceutical representatives witnessed a prototype of the inspection system at work in AMV's laboratory. Since being selected, AMV has installed two identical systems in West's Jersey Shore facility. Bergerstock reports that the company "took a year to assess, install, and validate the equipment. The two systems have been used in production since January."

PARTITIONING THE SYSTEM
To start the inspection process, the stoppers are manually fed into a hopper; they then drop into a centrifugal feeder (see Fig. 1). West recommended TL Feeding Systems Inc. (Minneapolis, MN), which tooled a centrifugal bowl for parts presentation. The feeder bowl rotates the parts, orients them for inspection plug-side up, and arranges them in a single file.

The parts move onto the main conveyor belt to the vision-inspection station. They are then illuminated by a Schott-Fostec LLC (Auburn, NY) DCR-II variable light source with a fiberoptic ringlight attachment (see image on p. 25). Supplied by JAI AS (Glostrup, Denmark), the progressive-scan CV-M10 camera with asynchronous reset is linked to an ITI-PCVision frame grabber board from Coreco Imaging Inc. (St. Laurent, Quebec, Canada).

The application software is built around the AMVision tool kit, which performs all the image-analysis functions (see Fig. 2). The software runs on a 300-MHz Windows NT-based Vectra PC from Hewlett-Packard Co. (Palo Alto, CA), which is equipped with a touchscreen monitor for the user interface and the system display. System control is performed by an SLC-500 programmable logic controller (PLC) furnished by Rockwell Automation/Allen Bradley (Milwaukee, WI).

DEVELOPMENT CHALLENGES
The handling of an elastic part coupled with the inspection of a transparent surface made for a challenging project, according to Joe Bach, AMV's director of engineering. "Transferring an almost-liquid part from a rotating stainless-steel plate to a conveyor belt, queuing enough parts to maintain a rate of five parts per second, and singulating them to provide a reasonable spacing for reject and bin selection gates were complicated tasks," he says.

Th Acf3c54
Click here to enlarge image

FIGURE 2. PC screen view allows a user to browse through the image log. The vision system can save the images of the last 40 inspected parts and the last 17 rejects and then reload and inspect them in off-line mode. The reason for a reject or failure is displayed along with a date and time stamp as the operator scrolls through the reject list. The large PASS/FAIL message shows the results of the last inspected image. The design and implementation of the diverter gates for directing inspected parts provided another challenge. "At a fixed part rate, increasing parts spacing meant increasing line speed, leading to some interesting choices between diverter gate speed, size, and pivot point selection," he adds.

Tracking the parts from inspection to reject or to one of the packaging bins was also a difficult job. AMV provided a part-tracking algorithm implemented in the PLC, which detected various parts conditions. They included bad parts failing to arrive at the reject bin, good parts failing to arrive at the good-parts bin diverter, good parts failing to arrive at the selected good-parts bin, and too many parts positioned in any logical conveyor path (parts backup). "Part tracking ensures that the function of the vision system (to accept no bad parts and to minimize rejection of good parts) is not subverted by uncontrollable mechanical anomalies, such as parts jumping on the conveyor," Bach points out.

SHINE THE LIGHT
As in all machine-vision systems, the design of the lighting and optics was crucial. AMV developed a custom light reflector attached to a standard fiberoptic ringlight to balance the illumination of surface inspection of a transparent coating as well as other characteristics of the part. The AMVision software analyzes the image for 12 distinct features of the part, including a complete surface inspection.

To achieve the requirements for surface inspection, 100% of the part's pixels were required to be analyzed with a resolution of approximately 0.001 sq mm/pixel across a 13-mm-diameter part. The cycle time for image acquisition, analysis, and data logging was 120 ms. Says AMV's Mancini, "We took an approach that classified the anomalies detected by the vision system in a way that minimized the false rejects of parts and made sure there were no false accepts, keeping the yields acceptable."

Bergerstock says that West's biggest challenge in adopting the AMV system was "passing vision-inspection-system knowledge on to the maintenance and production people and getting the system going."

West's Hill directed the system's installation and validation and says his main concern during that time revolved around the transition from one system to another. "Any time you stop an existing production line, make drastic changes, and then do a cold start, there can be problems," he stresses. "We gained some real-time experience with it during the phase-in process before shutting down the existing system; we didn't cut it in cold."

Bergerstock concludes that an automated vision system was chosen "because our customers are raising the quality bar. We're increasing the quality of the outbound product by taking the subjectivity of the human eye out of the process, giving our customer almost a 99.9% guarantee of getting a perfect product to the market place."

COMPANY INFORMATION

Applied Machine Vision Inc.
Pittsburgh, PA 15209
Web: www.amvision.com

Coreco Imaging Inc.
St. Laurent, Quebec, Canada
Web: www.imaging.com

Hewlett-Packard Co.
Palo Alto, CA 94303
Web: www.hp.com

JAI AS
DK-2600 Glostrup, Denmark
Web: www.jai.com

Rockwell Automation/Allen Bradley
Milwaukee, WI 53201
Web: www.ab.com

TL Feeding Systems Inc.
Minneapolis, MN 55369-1540
Web: www.mgsmachine. com/tlfeed.htm

More in Factory