Better Vision Boosts Inhaler Output

Manufacturer upgrades its machine vision system to slash variability of orifice measurements.

William Leventon

With a new pair of eyes watching a once-troublesome production process, an inhaler maker is getting more accurate
inspections and fewer false rejects.

At the U.S. facility of Bespak PLC, a UK-based manufacturer of drug-delivery devices, a machine vision system
measures the dimensions of orifices in injection-molded plastic inserts that are assembled into inhaler boxes. To meet
customer specifications, the vision system is supposed to reject orifices that don’t fall within a ±0.002-inch-diameter
tolerance. The system also checks to make sure that the orifices are no more than 20 percent oval (i.e., that there is
no more than a 20 percent difference between any two diameter measurements of the same orifice).

But Bespak’s eight-year-old vision system wasn’t up to the job. The system couldn’t pass the most basic validation
tests or meet the company’s repeatability and reliability testing requirements. “Our internal requirement was no more
than 20 percent variability in performance, and our ultimate goal was to achieve less than 10 percent,” explains Jim
Gallion, manager of manufacturing engineering at Bespak’s U.S. facility, located in Apex, North Carolina. “With that
system we were struggling to get below 30 to 40 percent variability, which . . . was leading to an excessive amount of
falsely rejected products.”

A Better Alternative

So Gallion and his colleagues investigated other alternatives. After evaluating several products, they settled on a
Checkpoint 900 vision system from Cognex Corp. (Natick, Massachusetts, USA). The vision system includes a
Pentium computer with a Checkpoint 900 PC plug-in card, two Sony CCD cameras, and a remote-controlled Fostec
fiber-optic light source.

In Bespak’s plant, inhalers are assembled by an automated machine. The machine takes inserts from feeder bowls
and places them two at a time on a rotary indexing table. The table indexes the parts into the view of the two vision
cameras, which are mounted side by side. Motion stops while each camera captures an image of the part in its line of
view. Extreme magnification is necessary to get high-quality images of the tiny orifices, which measure approximately
0.02 inch in diameter.

Image data goes to the Checkpoint card, which uses PatMax geometric pattern matching software to determine the
exact position of the orifices. This is no easy task, due in part to the size of the orifices and in part to machine
vibration. “Even slight vibration can shake the field of view and cause the pattern’s position to vary,” explains Joe
Grove, a Bespak manufacturing engineer. But with its rapid-reset CCD cameras and high-speed processor, the vision
system can capture high-resolution images of a vibrating target, Grove reports.

When the system finds an orifice, it checks the diameter using Checkpoint’s Edge and Arc Gauge tools. These tools
acquire 72 radial points, which are used to calculate an average diameter. The system also uses the radial points to
determine whether the shape of the orifice meets the customer’s “ovality” specifications. The entire inspection
process takes less than a second.

Light Concerns

During inspection, the Fostec light illuminates the orifice from the inside. If light intensity diminishes, “the vision system
is not able to detect the crisp, well-defined edge of the orifice, which affects the accuracy and reliability of the results,”
Grove says.

To maintain adequate light intensity, Grove came up with a data-carrying loop. A key component of the loop is
Checkpoint’s Light Meter tool, which assigns a “light value” to each image the system acquires. This value is
transferred from Checkpoint software to Visual Basic through bridging software called Checkpoint Communication, or
CPComm. The transfer is necessary because, unlike the Checkpoint software, Visual Basic can be used to
communicate with the Fostec light. If the light value is too low, the system sends a Visual Basic message to the light
source to boost intensity. Then the light meter checks the intensity again. If it’s still too low, another Visual Basic
message is dispatched to the light. The process is repeated until the proper intensity level is reached.
“This is a pretty high-level feedback loop that solves the major problem with vision, which is lighting variation,” Gallion
says.

When the vision system spots a defect, it’s logged in a database. Defective parts are dropped in a reject bin. Parts
that pass inspection are placed on an offload conveyor, which carries them to a packaging station.

During the inspection process, a special graphical user interface appears on a 17-inch monitor built into the assembly
machine. Developed by Bespak engineers, the interface displays the parts being inspected and keeps a running tally
of the number of rejects. By hiding Checkpoint’s more complex features, the interface makes it easier for plant
personnel to operate, test, and calibrate the system.

Passing the Test

Since it was installed and validated last year, the new vision system has been inspecting over 20,000 inserts a day.
“The Checkpoint system [has] proved capable of measuring the inserts within strict tolerances,” Gallion reports. “And
it passed the original validation tests and many others with flying colors.”

According to Bespak, the system has been very reliable, producing fewer false rejects and less scrap than its
predecessor. The system has also reduced variability to well below Bespak’s target of 10 percent. Even with tighter-
than-normal tolerances, variability now ranges between 3 and 4 percent. According to Gallion, variability that low “is
completely unheard of for automatic vision inspection at these speeds.”

Encouraged by this experience, Bespak has much bigger plans for the Checkpoint vision system. “We will use this
everywhere,” says Gallion, adding that his company plans to use vision data to make process adjustments that will
prevent the manufacture of faulty parts. “This is what we see the world going to.”

This case history appeared in European Medical Device Manufacturer.