MSA - Bias Study - Is there any easy method to follow and apply?

[Manueltrico]

There was a recent production line transfer into our plant,

I'm in charge of calculating all the MSA studies, the customer wants Bias studies, the information I have found for it, it's very confusing, is there any easy method to follow and apply?

Manuel

 

[Atul Khandekar]

Bias Method

Manuel,

Welcome to the Cove!
You will find the methods for conducting all MSA studies in the MSA manual. There are also some examples.

As for bias study, you need to have a 'master' or reference part of a known value. This value could be arrived at by measuring part using a much better (details in the msa manual) measuring instrument than the one under study. This becomes your reference value. You measure this part using the instrument under study, at least 10 times. Calculate average of these 10 readings. The difference between the Average and Reference value is the bias.

Hope this helps.
-Atul.

 

[Manueltrico]

Thank You Atul,

I have followed the instructions you suggested me, but now my problem is with the bias percentage.

the formula I have is = 100 x (bias/process variation)

I tried calculating the process variation by determining the sample standard deviation, and the population std dev. but still I'm not satisfied with the results



Thanks in advance for your help

 

[Atul Khandekar]

but still I'm not satisfied with the results

In what way?

 

[Manueltrico]

In my process,

-The standard deviation is: 0.0128648
-Bias is: .053

Formula to calculate the Bias percentage is: 100 x (bias/process variation)

Note: I understand process variation as: Standard deviation x 6

100 x (0.053 / (0.0128648 x 6)) = 68%

According to the MSA book I have, bias percentage should not be over 10%.

am I misunderstanding.

I don't believe my bias percentage is 68% when my Precition to tolerance ratio at the MSA is 0.11


What am I doing wrong, or is it right?

 

[AlbertPaglinawan]

Manueltrico, how many different testpoints do you have and how many times did you measure each testpoints?