Six Sigma – iSixSigma Forums General Forums Methodology Multi-head weighing system MSA

Viewing 2 posts - 1 through 2 (of 2 total)
• Author
Posts
• #53498

MBBinWI
Participant

Looking to perform an MSA on a multi-head weighing system. While each head performs an individual weighing operation, it is the combination of multiple heads via computer control that provides the output “deposit.”

I can see doing the study by individual head (24 of them), or as a system for the “deposit.”

Anyone have any experience performing an MSA on such equipment?

0
#190395

Lucas
Participant

I would think you should prove that each head as an observer is equivalent. Each head has a transducer (maybe the same load cell model?) and therefore it may make a difference which head recieves which weight. Also, I think you should prove the “processing” which amounts to summation is valid. Get the system operating range, the head operating range, and the resolution and accuracy from the manufacturer or system designer as built. Compare this to the intended operating range and the expected operating resolution. Consider that systems are often not linear in ability to measure over the entire range of operation.

I would select at least five different weights (one near the lower limit, one below the lower limit by an amount equal to the resolution you need to discern a low end defect, one mid-range, one upper limit, and one above the upper limit by the amount of resolutin desired to discern an upper limit excursion) and test them all repeatedly over time on each head. It may sound odd to define the over time factor, but I have experience where unexpected variation over hours or days or weeks will give different results so I suggest checking this in some way at some point before “blessing” the system.

If you, for instance had 5 heads, you could place each of the 5 weights on them: weight 1 on head 1, weight 2 on head 2, weight 5 on head 5. This would give total weight Y1. Then place weight 1 on head 2, weight 2 on head 3, . . . weight 5 on head 1. This would give total weight Y2. Y1 can be compared to Y2, etc. to verify the system summation ability to process the result from multiple sensors.

Save the 5 reference weights and set them onto each of the 5 heads randomly occassionally to get Ycontrol values for system monitoring. The weights may be traceable masters, or just controls.

Consider the factors (each head, the system processing, time, operators, temperature, level, tare to zero, etc.) for your system and the resulting risk and liklihood of ocurrence. Speak with the manufacture (design and build) engineers about what you are doing and if possible share you numbers with them to develop good will; at least someone will learn and everyone will benefit.

Hope this helps and was not too obvious. :)

0
Viewing 2 posts - 1 through 2 (of 2 total)

You must be logged in to reply to this topic.