Intro to Measurement System Analysis (MSA) of Continuous Data Part 1: What and Why?
This is the first in a series of articles about MSA. The intent is to share a careers worth of experience with measurement systems and their application in inspection, process control and continual improvement. Calibration, measurement uncertainty and attribute MSA are outside of the scope of this series and is better left to experts in those fields, of whom we have several in the Cove.
What is MSA and why do we need it?
To put it simply, any time we measure a part in order to determine whether it is in specification, whether the process is in a state of statistical control, or whether changing a control factor has a statistically significant effect on a response factor, there is some uncertainty about the true value of the part. Not only does one part vary from the next part, but repeated measurements of the same part will vary.
Why do we see this variation and what causes it? It is important to understand that when we measure a part, we rarely measure the true value of that part. There are many sources of variation that affect the measurement that we obtain using a measurement device, or gage.
These sources of variation include:
This not only affects the measured value, but also the variation seen as shown below:
XMeasured Value = X Actual Value ± X Form ± X Bias ± X Linearity ± X Stability ± X Repeatability ± X Reproducibility ± X Oper.*Part
σ^2Measured = σ^2Actual + σ^2Form + σ^2Bias + σ^2Linearity + σ^2Stability + σ^2Repeatability + σ^2Reproducibility + σ^2Oper.*Part
The end result of this is threefold:
MSA is a methodology for quantify the impact of each of these sources of variation. This allows us to quantify the risk for making Type I and II errors, to understanding the effect of measurement variation on the process capability, as well as the impact on the number of samples required to verify an improvement.
The next article will be:
Intro to Measurement System Analysis (MSA) of Continuous Data Part 2: Bias
What is MSA and why do we need it?
To put it simply, any time we measure a part in order to determine whether it is in specification, whether the process is in a state of statistical control, or whether changing a control factor has a statistically significant effect on a response factor, there is some uncertainty about the true value of the part. Not only does one part vary from the next part, but repeated measurements of the same part will vary.
Why do we see this variation and what causes it? It is important to understand that when we measure a part, we rarely measure the true value of that part. There are many sources of variation that affect the measurement that we obtain using a measurement device, or gage.
These sources of variation include:
- Within part variation in form
- Bias
- Linearity
- Stability
- Repeatability
- Reproducibility
- Operator * Part interaction
This not only affects the measured value, but also the variation seen as shown below:
XMeasured Value = X Actual Value ± X Form ± X Bias ± X Linearity ± X Stability ± X Repeatability ± X Reproducibility ± X Oper.*Part
σ^2Measured = σ^2Actual + σ^2Form + σ^2Bias + σ^2Linearity + σ^2Stability + σ^2Repeatability + σ^2Reproducibility + σ^2Oper.*Part
The end result of this is threefold:
- Any individual measurement has an uncertainty associated with it. When the specification falls within this zone of uncertainty you have two risks:
- Rejecting a good part (Type I error, Alpha risk)
- Accepting a bad part (Type II error; Beta risk)
- The capability of the process relative to the specifications is degraded by the additional variation added by the measurement system (i.e., your capability is actually better than you realize).
- The number of sample parts required to statistically verify the effect of a change in a control factor on a response factor is increased.
MSA is a methodology for quantify the impact of each of these sources of variation. This allows us to quantify the risk for making Type I and II errors, to understanding the effect of measurement variation on the process capability, as well as the impact on the number of samples required to verify an improvement.
The next article will be:
Intro to Measurement System Analysis (MSA) of Continuous Data Part 2: Bias
Total Comments 7
Comments
-
Good start. Thanks.
Please also consider adding this to Elsmar Wiki.
Rgds,
-Atul.Posted 3rd July 2009 at 02:43 AM by Atul Khandekar 
-
Thanks! We all appreciate your time in putting this together! I look forward to Part 2!
Posted 3rd July 2009 at 09:59 AM by Marc
-
Atul,
What did you have in mind regarding the Wiki?Posted 10th July 2009 at 04:17 PM by Miner
-
Thank you! I know many people will benefit from your taking the time and sharing!
Posted 17th July 2009 at 03:11 AM by Marc
-
thank you
Posted 11th September 2009 at 06:39 AM by armali
-
Thanks a lot for your time & effort.
Posted 29th October 2009 at 01:27 PM by MasterBB
-
Thank you.
Posted 20th November 2009 at 06:42 PM by Juan Dude
Total Trackbacks 0
