An automotive original equipment manufacturer (OEM) engineer recently added a welded nut to a stamped metal frame rail produced by Ice Industries’ Deerfield Manufacturing subsidiary in Mason, Ohio. Soon after the customer made the change, Deerfield designed and built a welder to add the nut to the stamped part after the stamping operations are completed. If the customer were to receive a frame rail without a nut, Ice Industries would be subject to substantial penalties including 100% containment and inspection of all product until root cause corrective action has been approved by the customer and implemented by Ice—not to mention the blemish on its quality scores.
The traditional approach would have been to have an operator inspect each part for the presence of the nut as the parts are packed. “The problem with manual inspection is that it is never 100% accurate,” said David Randall, Manufacturing Engineering Manager for Deerfield Manufacturing. “People have a tendency to get tired, bored, and start thinking of other things near the end of their shift. Our philosophy is to eliminate even a single bad part.”
Deerfield management decided to take a look at machine vision. “We have often found the costs of machine vision systems to be prohibitive,” Randall said. “Then one of our engineers found the Cognex Checker® vision sensor on the Internet. Besides the low cost, we liked the fact that Checker is so easy to use. We ran through a demo on their website and in just a couple of minutes programmed Checker to inspect a sample part.” Deerfield Manufacturing contacted a local Cognex distributor, Crescent Electric Supply Company (CESCO). The distributor came in and reviewed the part and provided recommendations on how to perform the inspection operation.
Ice Industries Engineer Phil Kautzman programmed the vision sensor by first positioning the camera in the station so that the area to be inspected was visible on the screen. He connected a laptop computer to the camera and loaded the setup software which shows the camera image on the screen. Kautzman drew a square around the nut to call out the feature being inspected. He then replaced the part with one that did not have the nut present. Checker’s One-Click Setup feature automatically created a sensor, learned the feature’s attributes and set the appropriate pass/fail criteria.
“We adjusted the vision sensor with the intention that it will not pass a single part without a nut,” Randall said. “Since the fine tuning of the inspection parameters were completed, Checker has detected every bad part. It’s also important to note that Checker has performed without any downtime in the harsh stamping environment. When we decided to implement Checker, we brought the customer in to review it. His only questions were about how he could implement similar machine vision systems in his own operation. We will be expanding our use of machine vision in the future to ensure that we continue to exceed our customers’ expectations.”