Stability in MSA

[Rameshwar25]

Per MSA manual, 4th edition, Stability study is to be condcuted by drawing average and range charts and points beyond UCL and LCL will indicate that measurement system is sable or not. While using control charts for monitoring of production process, general practise is to first calculate trial control limits and then use them to determine presence of any special cause.
My questiuon here is how trial control limits will be calculated for stability?

 

[Ashland78]

I would refer to the customer blueprint first, then your organizations blueprint. Focusing on key characteristics most importantly first.

 

[Bev D]

I think you may be confusing two different uses of the term stability. In the context of a statistical control chart statistical stability is required before calculating a capability index such as Cpk or Ppk. In this case statistical stability refers to product variation not to measurement variation.

In the context of an MSA, stability refers to teh ability of the measurement system to maintain good repeatability and accuracy over time and/or changing environmental conditions.

Statistical stability as assessed by a control chart is NOT required for an MSA.

 

[Rameshwar25]

I am not confused. I know objective of control charts in SPC and MSA both.
You have hurriedly replied without referring to MSA manula. Pl refer Page 86 of MSA manual and read last sentence on that page.

 

[Bev D]

Since I haven’t worked in the automotive industry in 30 years why do t you quote the paragraph you are referring to? I will also default to @Miner to answer

 

[Pudge 72]

MSA 4th Edition (2010) Page 86 last sentence: Analysis of the Control Charts indicates that the measurement process is stable since there are no obvious special cause effects visible.

First of all - let's not be ignorant to each other.
Bev was trying to help and provide feedback to your question, and she took time out of her day to try and help clarify a situation.
I don't think a paragraph and a half is a "hurried" reply.
Second, the context you referred to still tells us nothing about a requirement for stability, or rules / objectives.
Third - agrees with Bev. Let Miner set you straight because it doesn't sound like you want to listen to any of us anyway.

 

[Bev D]

I scraped a copy of the 4th edition: the section that is being referred to is a discussion regarding the effect of a lack of discrimination (NDC) on variables control charts. If there is a lack of discrimination then any process control chart will be unable to adequately detect process variation. This comes Wheeler’s and Lyday’s book on MSA. As far as I can tell there is no requirement to create a control chart to check this. It merely explains why low discrimination systems cannot or should not be used for SPC charts. But again I will rely on @Miner to provide the expert input as he is quite familiar with MSA 4th edition.

 

[Bev D]

There is also a comment earlier in the referenced section “measurement issues” that discusses the need to improve the ‘stability’ of your system such that there are only ‘common causes’ in the measurement error and that ‘special causes’ are to be eliminated. While these are terms often associated with control charts it is not necessary - or proscribed or suggested by the MSA manual that I found - to construct some type of special control chart to detect assignable causes or demonstrate the presence of only special causes in the measurement system. Standard assessment techniques can readily do this.

 

[Bev D]

To close out my thoughts (I promise )

The initial process control chart is created after the MSA is complete. Trial limits are created to determine if there any assignable causes present. (And if there simply aren’t enough parts produced yet…) The control chart will have both part and measurement system error. There may be part and measurement system assignable causes that are present. (We may have missed something in the initial MSA) once the assignable causes have been detected, identified, and removed new limits are calculated for the statistically stable process.

Of course we must be aware that both the part variation and the measurement system variation may be statistically stable but too large. This is why we perform the MSA first as laid out in any credible MSA guide including the MSA 4th edition automotive guide….

 

[Miner]

Per MSA manual, 4th edition, Stability study is to be condcuted by drawing average and range charts and points beyond UCL and LCL will indicate that measurement system is sable or not. While using control charts for monitoring of production process, general practise is to first calculate trial control limits and then use them to determine presence of any special cause.
My questiuon here is how trial control limits will be calculated for stability?

Apologies to all those who were citing my name for input. I was traveling over the weekend and was not checking in here.

@Rameshwar25 I am not entirely sure what you are asking. Are you asking about the formula used to calculate limits or the minimum number of samples? Let's start by using the concept of phase 1 and phase 2 control charts.

1. Phase 1 - The formula used would be the same formula as used for phase 2 control charts. Phase 1 is mainly investigatory and consists of using temporary control limits to determine whether the process is stable enough to calculate more permanent control limits. The phase 1 control limits are only temporary in the sense that the smaller number of subgroups used gives more uncertainty about actual control limit values. However, as you collect the data from more subgroups and calculate phase 2 control limits the actual difference will tend to be quite small UNLESS the process was unstable to begin with.
2. Phase 2 - Again, the formula used to calculate phase 2 control limits is the same formula as used in phase 1. The difference is that you have collected more data/subgroups and have higher confidence that the process is stable. Whereas. phase 1 was almost entirely composed of short-term variation, phase 2 has a greater opportunity to see the effects of long-term variation.

Now if your question pertains to the minimum number of samples required, AIAG would be consistent with the requirements shown in the SPC manual, namely 100 data points, which they tend to state as 20 subgroups of 5. However, Dr. Wheeler would argue that you can calculate reasonable phase 1 control limits with a much smaller number of subgroups. You can search my posts on this forum as I have referenced this several times.

One last note: In most cases, a stability study is not necessary unless you are making measurements that will be affected by changes in temperature (i.e., precision machining), or by warmup of electronic equipment (rare with newer equipment). I recommend performing a risk analysis first to determine whether a stability study is even necessary.

If I missed the intent of your post, please clarify it and I will respond more promptly now that I am back.