Statistical Process Control and the need for Measurement System Analysis (Gage R&R)

[Sankaran Mahadevan]

Hi guys,

The variation in any Statistical Process Control includes also the measuring equipment variation. If the Process is in Statistical Control and Process Capability index is fairly high say at 2, then is it necessary to do Measurement System Analysis (Gage R &R study). The Control Chart is maintained for all the three shifts and three operators are using the same chart. So in this case I think that Appraiser variation is also taken into account. Kindly comment on this point.

 

[KenK - 2009]

Given your scenario, I would venture the following statements:

1. If the observed process variation is considered sufficient, then there probably is not necessarily a real need to know the contribution of the measurement system to that variation. Certainly it is not bad enough to make the process variation appear unacceptable.

2. Regardless, you will still need to run studies investigating bias, linearity, and stability, since those are location inaccuracies rather than variation issues. You could think you're OK when you're really not.

 

[Al Dyer]

I guess my only comment would be he following:

1- Was a gage R&R done prior to production with three different operaters.

2- Gage R&R is based on using 3 random people to measure the effectiveness of the gage.

3- Gage stability and linearity are based on the gage, not the person using the gage.

4- Always consider the use of the gage. Are gage blocks used as a gage or are calipers verified by a calibrated gage blocks.

Just a personal view and I have been wrong in the past

ASD...

 

[AJLenarz]

Sankaran –

The real question here is…. what are you obligated to? I can’t speak on behalf of the TS world, but if you operate in the QS realm, you must perform “appropriate statistical studies… to analyze the variation present in the results of each type of measuring and test equipment system.”

These studies may consist of bias, linearity, stability, repeatability and reproducibility studies.

The key word here is “appropriate”. Choose the appropriate statistical studies relevant to your measuring system and process and be prepared to explain your choice.

I would tend to concur. With a Cpk of 2, I would question the necessity of a GR&R. The exception might be in a situation with high Cpk requirements.

But i'm just a pup in the world of quality.

 

[Tommy]

So much of the effectiveness of a control chart depends upon the 'rational' subgrouping of data. If the subgroup represents low (short-time) variation, the control limits for the averages will be relatively tight. When viewing the averages chart for such a situation, a large number of out-of-control points signify that the the variation between sample times is greater than the variation within a subgroup. This could be an indication of low measurement variability. Conversely, if no points in the averages chart are out-of-control, then the measurement error may be sufficiently significant disabling the detect meaningful shifts from sample to sample. You must ascertain what variation your subgroup represents.

 

[JuanAgos]

Re: Statistical Process Control

Hello, my name is Juan,

My opinion on the matter is:

To my experience in statistical control of process I consider and e been able to observe in different processes, that the variation of the process is hit directly by the properties statistical that the measurement system has that is used. The analysis of the measurement system looks for:

1. to identify the suitable equipment, (depending the objective of the measurement, or to control the process by means of control graphs or simply to accept or to reject product) for the characteristic that is desired to measure and / or to control, for this task I apply the following principle, my equipment of measurement must cover the following aspects:

a.- He forms: The form of the measurement equipment must be adapted to its use (feasible, adaptable to the atmosphere where it was used).

b.-Exactitude: The measurement equipment must be able to detect but minimum variation of the process, a general rule says to you that your equipment must have a relation of 1/10 discrimination. That is your equipment must be able to detect one tenth part of the tolerance or variation of your process like minimum.

c. It reaches: When one works with electronic equipment, this recommendation is due to take care of that the reach or total rank of measurement of the equipment is a greater 20% to the maximum value of the characteristic to measure, is because when the equipment is exposed to greater measurements to 80% of their reach, they lose in gradual form the exactitude of the equipment

2.- To compare a system of measurement with another system of measurement

3.- To identify when the measurement systems begin to lose the exactitude that have

4 .- To assure that the statistical properties of the systems stay in you limit them allowed, after repair activities have been made or fits to the equipment of measurement

5.- To assure that the users of the equipment, has the ability adapted in the use of the equipment.

To very personal opinion, first it is required to evaluate if the equipment is the adapted one to measure the selected characteristic, before using it and but even though it is tried to use graphs of control in the process, since the capacity which you calculate for your process can not be the true one since you can be using an inadequate equipment, this is very dangerous and much to confusion is lent.

I hope helps this you, for something but detailed on the matter, send me a mail and we can chat more in detail on situations you specify......