Time to get in shape

This is the time for New Year Resolutions and many of us, encouraged perhaps by TV commercials, have resolved to improve our fitness and generally “get in shape.” I think that same philosophy could be applied to machine vision, so here’s my seven step plan for vision system fitness.

Step 1 Audit – check to see how well the system is running. Is it performing the way it did when new (do you have data to tell?) Do a manual review on 50 consecutive parts and verify that the system is making the correct decision. (I know 50 may not be enough, but I’m assuming you don’t have all day.) Run a few bad parts through and check that they fail.

Step 2 – Backup everything. Save a copy of the latest inspection program (often these get changed on the system but no one ever saves a copy.) Before starting, document the positions of everything, especially the lens aperture. Grab a screenshot of the image and note the greyscale values if possible. Compare this with the master image you captured when you first set up the system. (What? No master? Tsk tsk.)

Step 3 – Clean every thing. Clean the lens (carefully) clean the lights, and clean the trigger. (You might want to make this a “5S” activity.) But, be sure nothing gets moved. Check for damage, especially to cables, and make sure they are still clipped out of harms way.

Step 4 – Clean up the old files on the PC or in the camera. Don’t forget to defragment the hard drive.

Step 5 - Check the timing. Can the inspection program, in its latest form, execute within the part rate? Most machine vision software has some kind of profile tool that will tell you how long the tools take to execute (I especially like the tools that come with the NI Vision Builder product.). Run it and check you can get a result out before the next acquisition.

Step 6 – If the system uses fluorescent lamps, and they’re more than a year old, replace them. Ditto for halogen. If you’re using LED lighting (always recommended,) consider replacement after 4 to 5 years of continuous service.

Step 7 – Check the calibration. If you’re doing any form of gauging, calibration should be part of your Quality Procedures, but if not check that the pixels-to-real world units factor is still valid. (It’s always possible that the camera got moved or the part is being presented slightly differently.)

That’s all it takes, seven steps to a reliable machine vision system. If anyone has additional tips they’d like to share I’ll be happy to publish them here.