Sensors scrutinize plastic packages
Two vision sensors use edge-detection and pattern-matching tools to image and analyze plastic containers.
by R. Winn Hardin, Contributing Editor
InMold labeling (IML), a process that forges a label into the wall of a plastic container during injection molding, has become common among injection-molder manufacturers over the past few decades. Although IML often can be expensive to implement, it offers several advantages over conventional glue-on labeling. For example, no extra labeling equipment or assembly steps are required because the labeling occurs during the molding process. In addition, production-line throughput is improved because a separate labeling station is not needed.
A pair of In-Sight 2000 vision sensors acquires and sends images to a pair of dedicated In-Sight 2000 controllers. The images check for label placement, raised features on the package, and position of a lot-code box.
On the downside, the labels are placed in the mold prior to injection by a robotic mechanism and held in place by electrostatic charge. If this charge is low, the label can move during the injection process. Also, inadequate injection pressure and material variations can produce surface defects that would degrade label and lid quality and the placement of machine-readable lot codes.
Precise Technology Inc. (North Versailles, PA), a contract injection molder serving the packaging, consumer/industrial, and health-care markets, recently faced these problems at its Swedesboro, NJ, facility. The plant, also called Precise IML, is reportedly the largest IML packaging facility in the United States. Among other products, the plant produces seven different lines of baby-wipe packages for one of the world's largest household-goods manufacturers.
Precise Technology turned to vision-systems integrator Serview Inc. (Bristol, PA) for assistance. Serview designed a packaging inspection system with two Cognex Corp. (Natick, MA) In-Sight 2000 vision sensors and controllers and custom and laser lighting on each of four manufacturing lines triggered by photoelectric sensors from SICK AG (Waldkirch, Germany; see Fig. 1). The In-Sight 2000s image a 10-in. region of interest that includes the package's label, sidewalls, and lid. Pass/fail information is fed through an Allen-Bradley (Milwaukee, WI) DeviceNet industrial network to an SLC 500 controller, which in turn controls a reject mechanism downstream from the inspection zone.
"A key advantage of using IML is product safety," explains Jim Bruynell, maintenance manager at Precise Technology. "Because the label becomes part of the container during molding, it can't be peeled off. In addition, the entire container can be recycled, because the label and the tub both are made of polypropylene," he adds.
Despite safety and process advantages, IML poses some challenges with respect to correct label placement. While improper label registration is a major problem, other problems can occur, as well. For example, labels that have been misloaded into the molder may be applied backwards, and, because the plant produces containers for more than seven different styles of baby wipes, occasionally the wrong label may be applied. In addition, Precise Technology needed to ensure that a blank white box printed on the label, where a lot code is later applied at the end-customer's facility, is in the correct position.
Serview director of operations and integration for the Precise Technologies project, Steven LeBlanc, says the In-Sight 2000 sensor was selected because of its user-friendly operator interface and the spreadsheet-like programming feature. He adds, "The sensors had the right vision tools for the job and were also fairly straightforward to use for operators who do not have previous experience with machine-vision applications. Also, the sensors are less expensive than PC-based systems, making the overall installation more cost-effective for the company."
Wiping out defects
Two In-Sight vision sensors provide full inspection coverage of the left- and right-hand portions of each container. After a container has been molded and the lid installed, the entire assembly is flipped upside down (so the tub bottom faces up) and moves down a conveyor toward the inspection point (see photo on p. 29).
LeBlanc used the 640 × 480-pixel, 8-bit gray-scale In-Sight 2000 vision sensors—each with a 6-in. field of view at an 18-in. standoff—to increase package resolution and cover the right- and left-hand portions of the 10-in. label area. The dedicated In-Sight 2000 processor uses edge-detection and pattern-matching algorithms to verify that the label position is within a ±2-mm tolerance, ensure each label is correct for the container, and inspect the lid hinge to make sure it has been properly snapped into place.
Looking down at a 45° angle, the cameras, which are triggered by SICK AG WT23 adjustable photoelectric sensors across a 24-V LVDS wire, capture an image of each side of the container. The images are analyzed using the In-Sight PatFind pattern-matching tool, which compares each image to a pretrained reference image of the correct label. Once this match has occurred, various edge-detection and inspection tools measure the exact x, y, and theta position of the label to ensure that the lid has been properly assembled. More than 30 containers are inspected per minute.
LeBlanc says he spent some time fine-tuning the application with both software and lighting. Serview, which makes its own fluorescent backlighting arrays, selected a Lasiris SNF structured laser line light from StockerYale Inc. (Salem, NH) to help the vision sensors determine the z-axis coordinate, or height, of the lid's raised features. The laser light paints the package with a straight line of light. By measuring the offsets in the straight line of light, In-Sight's vision-processing software can determine a 3-D z-axis position for each pixel along the line in a repeatable manner, LeBlanc explains. Both the fluorescent and laser lights are always "on" and, therefore, do not need triggering.
Passing the wipes
LeBlanc adds, "The baby-wipe application involves a myriad of inspections, from measuring label position to assembly verification of the container. There were also a variety of part-appearance variations to deal with, such as glare on the part, different colored plastics, and label graphic variations. So, we implemented a highly engineered illumination system involving fluorescent backlighting, fluorescent front lighting, and structured laser lighting techniques. The resulting system produces high-contrast images that work well for the vision sensor's measurement and inspection tools."
If a container fails an inspection, the information is passed on to the plant's Allen-Bradley DeviceNet-based PLC network. Pass/fail signals are sent through discrete I/O lines to an Allen-Bradley Flex I/O collector module. The module is scanned by an Allen-Bradley SLC 500 controller, which sends a control signal to a downstream ejector. When the defective package reaches that point on the conveyor, the ejector removes the package.
Passed containers proceed down the conveyor to a diverter mechanism, which places the package in a stacker. A VGA monitor located above the inspection station shows a live image of the inspections as they occur; green and red graphics on the screen indicate pass or fail, a readout of the total number of parts inspected per shift, and the number of bad parts per shift (see Fig. 2).
FIGURE 2. An optional monitor for the In-Sight system shows acquired images in real time. It also presents various production and inspection statistics related to a specific manufacturing line.
According to Precise Technologies' Bruynell, the vision sensors are inspecting the labels on a 24/7 basis with high accuracy and repeatability. He says, "We know that the labels shipped are the same every time. This allows us to ensure 100% product quality for our customers and ties in with our overall continuous improvement strategy."
To date, the company has implemented the two-vision-sensor platform on four production lines. It currently has plans to tie the inspection-results data into the company's Enterprise Resource Planning system for improved production reporting. According to Serview's LeBlanc, this project was started before the In-Sight modules started shipping with Ethernet capability, so an upgrade or optional Ethernet module may be required if Precise Technology decides to integrate the vision inspection systems with reporting software.
Milwaukee, WI, USA
Natick, MA, USA
Precise Technology Inc.
North Versailles, PA, USA
Bristol, PA, USA
Salem, NH, USA