A major pharmaceutical manufacturer's packaging line presents a challenging vision system application because of the need to select exactly the right label among dozens of possible alternatives needed to meet differing languages and regulatory requirements around the world. The vision system used in the past rejected 25 percent of good packages.
When the pharmaceutical manufacturer installed a new packaging line, they upgraded the vision system to Cognex In-Sight 5100 and 5600 vision systems, which use optical character recognition/optical character verification (OCR/OCV) to read characters on the label and compare them to the correct sequence. This approach reduced the false reject rate to 0.5 percent, making a substantial contribution, improving the first pass yield (Quality) and consequently the overall equipment effectiveness* (OEE) by 200 percent.
Previous vision system had high false rejects
The packaging line used previously in this application was a linear machine with hot stamp pressure-sensitive printing and an image comparison vision system. This vision system had difficulty matching ideal to actual images which led to a high false reject rate. This high false reject rate played a major role in the line's disappointing OEE. The vision system also took a considerable period of time to set up for new labels.
The pharmaceutical manufacturer began looking at alternatives for improving the line. The company developed a rotary labeling machine using thermal transfer printing that offers higher availability and performance. For help in improving the vision system, the company called upon Wilfredo Jiménez and Jesus Otero from WJ Automation & Integration Corp. in San Juan, Puerto Rico (WJAI).
Selecting the vision sensors
"When I first got involved in the project, the company was working with a machine builder to design two rotary label machines," Jiménez says. "They needed to inspect every aspect of a very complex label with near 100 percent accuracy. My contact told me about the problems with the previous system and asked if we could help. We accepted the challenge. The first step was finding a vision sensor that could accurately identify and verify the many different labels. We picked Cognex In-Sight 5100 and 5600 vision sensors because they are fast, accurate, easy to program and compact enough to fit easily within the available envelope of the machine.
"One of the problems with the previous system was that it used a separate camera, frame grabber and computer, which were connected together by cables," Jiménez continues. "The interconnections generated a lot of noise, which was one reason for the poor accuracy of the vision system. The In-Sight 5100 and 5600 avoid this problem because the complete vision system is self-contained within a single housing. Both Cognex vision systems are very fast, providing a 100 millisecond inspection cycle which is well under the cycle time of the machine."
The In-Sight 5100 vision systems acquire up to 60 640 x 480 pixel full frames per second with high quality 8-bit images. The In-Sight 5600 acquires 640 x 480 pixel images at very high speeds.
Programming the vision sensors
For this customer, information can appear on the right side, left side or on both sides of the label,
depending on the specific label that is used. The left side of the label normally has a maximum of three lines of text with a maximum of 15 characters on each line. The right side of the label normally contains the Pharmacode, also known as Pharmaceutical Binary Code, a 1D bar code standard used in the pharmaceutical industry as a packing control system. Some labels also have 2D bar codes which could be on either side of the label. A 1D bar code that defines the lot also appears across the bottom of each label. This means that a relatively larger field of view of 4 inches wide and 2-1/2 inches tall must be read to fully verify the label. The labels are also glossy, so they have a tendency to create reflections that can detract from image quality. Another inspection required was to inspect that the bottom side of bottles are properly printed.
WJAI addressed this challenge by dividing the field of view into three sections—left, right and bottle bottom—and addressing each section with a separate vision system. They used a Cognex In-Sight 5600 for the left side, an In-Sight 5100 for the right and another InSight 5100 for bottle bottom side. The left vision system has a field-of-view of 1-inch wide x 2-inches high, the right system has a field-of-view 1.5-inches wide x 2-inches high and the bottom sensor has a field-of-view 2-inches wide x 2-inch high. They used a diffused light and carefully experimented with the lighting angle to minimize reflections.
WJAI used Cognex In-Sight Explorer software to program the cameras to read the labels. On the left hand camera, Cognex OCR/OCV tools were used to read the three lines of text and compare them to the expected result. Cognex OCR/OCV tools handle low-contrast characters as well as confusing or unevenly spaced characters. Cognex IDMax code reading tools read the bar codes. IDMax handles degradation in code appearance to deliver consistently high read rates. The machine production rate could be tripled from its current 100 bottles per minute without increasing the vision system speed.
Integrating the vision sensors with the packaging line
WJAI configured the vision sensors to communicate with the Allen-Bradley programmable logic controller (PLC) and human machine interface (HMI) software from Rockwell Automation that run the machine. When the operator begins running a new batch of labels, he or she first selects the type of label in the HMI which, in turn, informs the vision system of the text and bar codes that should be on the label and their position. This integration eliminates the need for setting up the vision system when changing from one type of label to another. The PLC also sends a signal that the label is in position ready to be inspected. After inspecting the label, the vision systems provide the results to the PLC and HMI. Bottles that fail the inspection automatically go into a bad parts bin for manual inspection and repair. The operator can see images from the camera in real time on the control window, which makes it easy to resolve any issues.
The new vision system took only one week to validate compared to three months for the previous system. It exceeded the pharmaceutical manufacturer's expectations by providing a false reject rate of only 0.5 percent. It also helped to improve the performance (standard vs actual) of the packaging line by smoothing the progress of changing over from one type of label to the next. These improvements, in first pass yield and availability played a major role in the dramatic improvement in OEE, which vaulted the pharmaceutical manufacturer to among the highest performers in the industry.
*The overall equipment effectiveness (OEE) is the ratio of the fully productive time to the planned production time and can be calculated as a product of its three contributing factors: OEE = Availability X Performance X Quality. This type of calculation makes the OEE a severe test. For example, if all three contributing factors are 90.0 percent, the OEE would be only 72.9 percent. The generally accepted world-class goals for the three factors are: Availability - 90.0 percent, Performance - 95.0 percent, and Quality - 99.9 percent, which yields an OEE of 85.0 percent. Worldwide studies indicate that the average OEE rate in manufacturing plants today is around 60 percent.