Turning a Machine Vision Losing Streak into a Machine Vision Winning Streak

Vision systems started making their way into manufacturing in the 1980s and passed through the Gartner® Hype Cycle™ from the peak of inflated expectations through the trough of disillusionment and have long been on the plateau of productivity.  However, many systems still try to meet inflated expectations and lead to disillusionment when the vision system is scrapped and written off.  By understanding some basic failure modes and setting proper expectations, your next vision project can be successful and start a winning streak that will help you achieve the elusive plateau of productivity. 

A Problem Well Stated Is a Problem Half Solved

Charles Kettering coined this phrase, and it should be the mantra of every vision system.  The specification is essential to know what the vision system will accomplish and be the gauge for whether the system is meeting expectations. 

The key to creating the specification is to "speak the language of the vision system."  Many times, a collection of product samples is shared, and it's obvious to the human eye what the problems are.  However, the vision system speaks in contrast, pixels, and quantities, so there needs to be a reconciliation between these two approaches.  There are many times when an engineer will want to find a 'splotch,' a 'gel,' a 'bug,' or a ‘streak,' but to a vision system, theseare all ‘contrasting blobs’ with certain geometrical features that need to be mathematically defined.  Similarly, for gauging applications where a measurement needs to be taken, something as simple as defining what the edge of the product is can make or break the inspection.  Taking the time to translate the inspection needs into the language of the vision system and defining these defects in terms of pixels and contrast will help to specify not only the vision hardware, but also the final qualification plan and everything in between. 

Systems Will Fail at the Interfaces

Sir George Buckley, the former CEO of 3M, shared the insight that systems are most likely to fail at the interfaces.  A door is not going to fail in the middle but will fail at the hinges or the doorknob.  This concept is true for vision systems as well.  A vision system will function consistently by itself but will fail when interacting with a larger system of automated production.  Managing those interfaces between the PLC, the mechanical design, the user interface, the operators, and the engineers is essential to the success of the overall system.  There are many times when a vision system is functioning exactly as designed but is failing a lot of parts.  Rather than understand why the parts are failing, it's easier to disable the vision system.  

"The vision system is the number one knowledge creation system on a manufacturing line and can give waste by cause information, process capability data (Cpk), and key process indicators to pinpoint the source of the failures to turn them into successes."

The interface to the system is often overlooked and isn't a focus of the vision industry.  The use of smart cameras that operate like a black box can be a detriment when it takes specialized skills to access the outputs of the system.  Many times, I've been told, ‘The vision system isn't working’ or worse, that it's ‘locked up.’ A good interface will engage users rather than intimidate them – this cannot be understated!  The more that people understand the system and what it's doing, the more successful it will be. 

A Vision System Is the Last Line of Defense and the First Line of Offense

In soccer, the goalkeeper is the last line of defense but also the first person to turn things into offense.  Many times, a vision is simply used to reject defective products, and the reasons for rejection are either unknown or ignored.  The vision system is the number one knowledge creation system on a manufacturing line and can give waste by cause information, process capability data (Cpk), and key process indicators to pinpoint the source of the failures to turn them into successes.

The Bottom Line

To start a winning streak with vision systems, start by understanding how to speak the language of the system to set a good specifications for operation.  Next, define the interfaces to the rest of the manufacturing line, including the operators and engineers.  Finally, understand the system well enough to interpret the vision data points as key process tools to make the products always runto target.  With this, you’ll be winning every time!