When is a system really “broken”?
Remember the service call I mentioned in the previous post? The customer had some vision systems that were “not working”, and even wanted to consider new equipment!
This may sound like too broad a statement, but generally, machine vision systems already operating in an automation environment actually rarely malfunction. Nor does their performance “get worse over time”. Certainly components like lights and cameras and processors and other peripherals will fail (or in the case of illumination, degrade). But the simple truth is that virtually any change in the performance of the machine vision system is the result of some external force. (Ok; I understand that any time the system’s not doing its job one might call it “broken”, but there’s a further point here.)
The most common of these changes affecting performance is a change in the imaging: cameras being moved, lenses de-focused, light intensity changing, and so on. I’ve had customers dramatically change a camera position with respect to dedicated illumination and then wonder why the inspection failed since the camera was “still looking at the same area”. Changes in the part being inspected also have a dramatic effect on performance, particularly when those potential variations were not known in advance and accommodated for in the imaging and processing.
Very often, the end-user maintenance and debug process for a machine vision system producing excessive false rejects or missing actual defects is a poorly organized and usually not documented procedure involving random things like focusing the camera, adjusting the F-stop, changing thresholds, even replacing components. Maybe these actions might actually help in some circumstances. What I’d like to emphasize however, is the need for a better “root cause analysis” of the problem. The question that needs to be answered is “Why did system performance degrade?”. Look first to the obvious and most likely candidates: changes in imaging and/or part features or presentation. With a clear understanding of the real problem, the resolution is often easier than expected, and may only require something like minor adjustment of inspection parameters.
An awareness of the impact of external changes on a vision system could possibly eliminate or reduce problems that otherwise might balloon into emergencies. If changes to a part are planned, incorporate appropriate changes to the inspection system just as one would plan changes to other aspects of the automation. Where appropriate, plan and document preventative maintenance for lighting and optical components so that degradation of illumination or dirty lenses won’t affect day to day inspection performance.
David…
I recently responded to a service call. The complaint was that the machine vision systems were no longer correctly doing the desired inspection, were missing detectable defects, and were creating a high number of false rejects. Sound familiar?
I was intrigued that the customer wanted us to not only fix their existing vision systems, but also wanted “recommendations for getting in some better equipment”. The equipment was only 4 or 5 years old, but the customer had been to a recent trade show and had heard that machine vision technology was changing so quickly that new components would certainly solve the problem.
Ultimately, I learned that the parts being inspected had changed. Features which had been opaque were now translucent, geometric sizes of other features were different from original. We made some adjustments to inspection algorithms, added some image processing/filtering as appropriate, and the systems were quickly operational again with no false accepts and minimal (< 0.01%) false reject rates. We took some extra time to provide some enhancement to the operator interface which made machine maintenance easier for the customer.
Sounds like a nice success story, but…for me the service call’s not the real story here. There are so many critical “sub-plots” to this story that I barely know where to begin. However, there are two important “back stories” I’ll cover in the next couple of posts.
David…
Welcome to the new Aptúra Machine Vision Solutions web log! Everyone knows how much I like to talk about machine vision, so this will be an interesting endeavor. I look forward to your comments and expect a rich dialog on a variety of topics.
What will 2009 be like for those of us who work in machine vision and related industries? I was considering having an “I survived the housing market collapse, financial institution crash, lending market crisis, automotive industry failure, and political corruption of 2008” t-shirt printed for the New Year’s Eve party. It seems like not much else can go wrong; but in this economic climate if something else can happen it probably will. Ultimately though, I am by nature an optimist, and there may be some silver lining to these persistent storm clouds.
It has been my experience as a business owner over the past two recessions that the industrial automation market perhaps does not thrive in, but does survive an economic downturn. The anecdotal argument is that end-user companies will tend to reduce production, sell-off inventory, conserve cash, and consider ways to improve productivity and competitive position in the marketplace. I don’t know how true this is as an economic theory but the latter two can be considered particular strengths of machine vision technology when properly specified and integrated. Hopefully potential end-users of machine vision will share in that sentiment as the new year matures.
In any case, I wish you a very Happy New Year, and share in everyone’s hope for a prosperous 2009.
David…
