Machine vision--more than a pass/fail decision

Last month, one of our readers called me with a systems-integration problem. He works for a large company that manufactures stepladders. To construct these, parts are first placed on forming machines where they are crimped together at high pressure. About once a month, he said, the parts are not placed in the correct location on the machine. But the crimping machine doesn`t know that. When the operator presses the button, aluminum is crimped left, right, and center, with parts flying everywhere.

Dec 1st, 1996

Machine vision--more than a pass/fail decision

Andy Wilson Editor

andyw@pennwell.com

Last month, one of our readers called me with a systems-integration problem. He works for a large company that manufactures stepladders. To construct these, parts are first placed on forming machines where they are crimped together at high pressure. About once a month, he said, the parts are not placed in the correct location on the machine. But the crimping machine doesn`t know that. When the operator presses the button, aluminum is crimped left, right, and center, with parts flying everywhere.

After the machine is turned off, it takes about a day to clean up the mess. This results in downtime and operator frustration. The answer? Take a picture of the parts before they are crimped, make sure they are located correctly, and send a pass/fail decision to a programmable logic controller (PLC) in less than one second. "And we want to do this for $1500," the reader said.

Quite a challenge. But there was a carrot being dangled (I think he thought I was a supplier). "If it works, we might install 300 in our factory," he said. Simple mathematics made the problem $45,000 more interesting.

After extensive research, I discovered that with the lowest -cost PC, PCI frame grabber, camera, and shareware he could probably test a system for about $1500. That`s the solution a promise of $45,000 delivers.

An effective solution

But would a simple, low-cost image processing solution really prove effective in the long term? Maybe the ladder parts were being placed incorrectly by the operator. Or maybe the machine crashed every month because parts were being delivered from a different outside supplier, a different person was operating the machine, or power to the machine varied at different times of day. If this was the case, the answer is no.

To gain these benefits, our reader might want to place more than a simple pass/fail system on the production line. By integrating another visual inspection onto the production line, for example, dimensions of parts could be inspected before they are placed on the crimping machine. By placing a data-acquisition system in the set-up, power-line fluctuations could be measured simultaneously.

An integrated system

Integrating these systems with networking, graphical-user interfaces, and database-management software would provide our reader with a better picture of why the machine was failing. And, it would result in a more productive inspection system--one where collected data could also be used for inventory tracking, product management, and improving manufacturing operator efficiency.

At present, such systems cost far more than $1500 per station, and management is often reluctant to purchase equipment that appears to have no pay-back period. But if products such as drugs, semiconductors, and canned goods have to be recalled, manufacturers will spend millions in advertising campaigns, mailings and shipping. If lawsuits are involved, then the costs rise further.

Only completely integrated vision systems can address such problems. So while our reader may solve part of his problem for $1500, he should realize that there may be more to machine vision than low-cost pass/fail decisions.

More in Boards & Software