Case study: Kraft Foods saves time by upgrading to image-based ID readers
Allergen management is playing an increasingly important role in the packaged foods industry. Manufacturers are taking increasing care to avoid labeling mix-ups that sometimes lead to expensive recalls and potential liability concerns. In order to address these concerns, Kraft Foods Canada scans each label after it has been attached on the packaging line to ensure that it matches the package contents. But the laser scanner type ID readers used in the past to inspect the company’s barbeque sauce products were subject to read failures, especially when changing over a new product with the label in a different location. A technician had to rush over to adjust the position of the ID reader to re-center the label in its field of vision.
Kraft solved the problem by switching to image-based DataMan 300 ID readers from Cognex (www.cognex.com) that can read any label within their 5 inch by 5 inch field of view without requiring adjustment. “The outstanding performance of the DataMan 300 ID readers saves the technical team a considerable amount of time while protecting customers by ensuring the accuracy of our labels,” says Dave Fortin, technician for Kraft Foods Canada, Saint-Laurent, Québec.
Kraft Foods produces biscuits, candy, beverages, cheese, grocery products and convenient meals under the Cadbury, Jacobs, Kraft, LU, Maxwell House, Milka, Nabisco, Oreo, Oscar Mayer, Philadelphia, Tang and Trident brands in approximately 170 countries. The company had 2011 revenue of $54.4 billion. Kraft is the leading packaged goods company in Canada, tracing it roots to the vision of James Lewis Kraft of Stevensville, Ontario. Kraft didn’t like the poor economics he saw as a sales clerk in a country general so he set out to do something about it, in the process creating much of the foundation of modern food merchandising.
Difficult code reading challenge
The barbeque sauce product line at the Saint-Laurent plant produces 30 different stock keeping units (SKU) at a rate of up to 265 bottles per minute. Ensuring that each individual package has the correct label is critical because some of the products have ingredients such as mustard and egg that certain customers may be allergic to. When the line is changed over to produce a different SKU number, the proper labels are manually loaded into the filling machine.
However, the possibility exists that the person operating the machine might load the wrong labels or that a few wrong labels might be accidentally mixed in with the correct labels. To address this concern, Kraft originally used laser-based ID scanners to read the 1-D code on each label as it passed by on the line and send the results to the programmable logic controller (PLC) that runs the machine. The PLC compared the code to the proper value and if the code was wrong the package was ejected from the line.
The problem with the laser-based scanners is that they are only capable of reading codes located within a small field of view. The label design is market driven so codes may be positioned at any position depending on the designer’s decision. As a result, when the labels are changed the code may be in a different position. This required that the position of the laser scanners be adjusted whenever the product line changed to a new SKU number, taking a considerable amount of the technical staff‘s time. Yet even when the laser scanners were positioned perfectly they still often failed to read the code.
So the technical team was frequently called out to make adjustments to the 1-D code readers and often struggled to determine why they generated no read failures. “My team was frequently called upon to adjust the code readers,” Fortin says. “These calls made it difficult for us to fulfill my other responsibilities. A considerable amount of time was also required on the part of the production team to inspect each of the packages that could not be read by the scanner to make sure it was correct.”
“I suggested to Kraft that they consider image-based code reading technology,” says Mike Palmieri, senior technical sales representative for Cadence Automation, a Cognex integrator located in Ste-Thérèse, Québec. The basic idea behind image-based technology is that the reader captures an image and uses a series of algorithms to process the image to make it easier to read. A typical algorithm searches the entire image for the code and identifies the position and orientation of the code for easy reading. Other algorithms handle degradations in code quality due to differences in material types and surfaces.
A key advantage of the image-based approach is that it not only reads 1-D codes and provides higher read rates, but also reads 2-D matrix codes. Because these two-dimensional codes like the Data Matrix can hold a much larger volume of data, they provide a considerable amount of redundancy that is used for error correction. So the code can often be read even when it is damaged.