# Questions about 100 % measurement

Discussion in 'Sampling, Standards and Inspection' started by Airwolf, Nov 7, 2019.

1. ### AirwolfMember

Joined:
Sep 20, 2019
Messages:
8
1
Trophy Points:
2
First, I am sorry if this comes across as confused - I am trying to figure this out.

We have an automatic measurement assembled machine, in which a robot arm moves an item along a series of steps. After turning, the item is moved to a steam cleaner, then it goes to one measuring device, then another measuring device. Then the computer decides if everything is okay: accept or discard.

So, this production of an item has 100 % measurement, with no sampling needed. My question is: what happens if any measurements is lost due to some incident. Let's say, as an example, that there is only data present to support that 90 % of the items has been measured and accepted. Is there any way to infer that the rest of the items are accepted?

The obvious option is to measure the whole lot again - but is there an alternative? I guess I can choose to do sampling, but then it would not be 100 % verification any more.

I hope I have been able to make myself understandable.

Nicholas

2. ### Bev DModeratorStaff Member

Joined:
Jul 30, 2015
Messages:
543
602
Trophy Points:
92
Location:
Maine
It really depends on waht is being measured adn the severity of the defect. Plus any Customer requirements.

Depending on the characteristic and the process you may be able to use statistical probability to determine the likelihood of the units that have no data being defective. We would need more details to give any definitive advice.

Airwolf and yodon like this.
3. ### Golfman25Well-Known Member

Joined:
Nov 6, 2015
Messages:
732
361
Trophy Points:
62
The real question is why would you loose some data and can you fix that?

Airwolf and yodon like this.
4. ### AirwolfMember

Joined:
Sep 20, 2019
Messages:
8
1
Trophy Points:
2
Yes, why would we lose data. I know I am looking for a solution to a problem that I do not have.

However, management has decided to come up with a sort of plan - in case something goes wrong and data is gone missing

Here is my train of thought:
So, the assembled machine is manufacturing 5000 units by turning them, and then measuring all 5000. So, let us say that 10 %
of the measurement data goes 'missing'. That means that 500 units doesn't have measurements. These would be impossible to
find, as they are mixed with the rest. It also means that 4500 units DO have measurements, and those 90 % are *almost* 100 %.
But they are not 100 %. If I view the 4500 units as a sample, then it is a really large sample! So, can I use this argument to make
the case that I am really, really close to having 100 % verification?

Nicholas

5. ### Bev DModeratorStaff Member

Joined:
Jul 30, 2015
Messages:
543
602
Trophy Points:
92
Location:
Maine
Uh, no.
If you are required to have 100% verification then you must have 100% verification. This is typically required for nonconformances with very high severity failure modes.

My original questions must be answered for us to give you any real advice.

Joined:
Nov 6, 2015
Messages:
732