CONTROL CHART RULES

[Nics Santos]

Hi.l Maam/Sir. I have question, anyone who can help me? In control chart rules. How to select a rule in a process and why? What are the basis? Thank u
Example: out of 8 control chart rules, me as a process owner what rule shall I select? And what is/are the basis of the selection?
Thank you in advance

 

[Miner]

You are asking a great question as many will use all of the extended rules without question and increase their risk of false alarms. You should only use the rules that are appropriate for your process.

Since there are a number of different lists of rules that are numbered differently, I will use Nelson's Rules as a reference.

Nelson's Rule

1. One point is more than 3 standard deviations from the mean - This is the most important rule and should ALWAYS be used. This is the original rule formulated by Shewhart.
2. Nine or more points in a row are on the same side of the mean - This is usually a good rule to use if your process tends to shift such as when changing a batch of raw materials.
3. Six or more points in a row are continually increasing (or deceasing) - This rule is entirely dependent on your process. Some processes are subject to trends but many are not. If your process can trend, consider using it, but most processes do not need this rule.
4. Fourteen or more points in a row alternate in direction, increasing then decreasing (in a sawtooth pattern) - This rule is a diagnostic used only when setting up your chart for the first time. It is used as a check for certain mistakes in subgrouping such as mixtures in process streams. Do not use this after you have verified that you have rational subgroups.
5. Two or three out of three points in a row are more than 2 standard deviations from the mean and on the same side - Only use this rule if you need slightly tighter control over the process than is provided by rule 1. This increases your risk of a false alarm.
6. Four or five out of five points in a row are more than 1 standard deviations from the mean and on the same side - Only use this rule if you need even tighter control over the process than is provided by rules 1 and 5. This increases your risk of a false alarm even more.
7. Fifteen points in a row are all within 1 standard deviation of the mean on either side of the mean - This rule is a diagnostic used only when setting up your chart for the first time. It is used as a check for rational subgrouping. Do not use this after you have verified that you have rational subgroups.
8. Eight points in a row exist, but none within 1 standard deviation of the mean, and the points are in both directions from the mean - This rule is a diagnostic used only when setting up your chart for the first time. It is used as a check for certain mistakes in subgrouping such as mixtures in process streams. Do not use this after you have verified that you have rational subgroups.

One last comment. If you have a measurement system with a larger R&R, see Dr. Wheeler's articles on Class 1, 2 and 3 monitors. He writes about how using rules 2, 5 and 6 above can help you overcome some of the deficiencies of Class 3 monitors.

 

[Bev D]

I assume you mean select in a software package? If you are manually applying th rules there is no reason to select.

Always select one point outside of the limits. Don’t ever not use that one as it is the most effective and is the seminal rule.
The second rule of 2 out of 3 points within 1 sigma of the same limit works pretty well.

The others occur less frequently than these two

 

[Bev D]

GREAT Response Miner!

As always the physics of your process will also influence which rules are more effective (other than rule 1).

 

[Nics Santos]

Thank you so much! It is a big help.
So it is really depends on the control and principle of a process right? If a process is wirebonding for example, which rules are the most appropriate? Wire pull strength test specifically. With Lower Standard Limit of 0.2N. For this reliability test, the higher the Force is better as long as the result will not be lower than 0.2N is okay. Which rule do you think sir?

 

[Nics Santos]

What I mean Sir is choosing of appropriate rule. Yeah, rule 1 shall be used. Thank you

 

[Nics Santos]

So it is not true that the rules to be applied is depend on the Sigma control? Some says that if your process is 6 sigma control you should apply all the rules. Or if your process is 3 sigma the rules must be 4.

 

[Bev D]

In most cases if you don’t know much about SPC ‘select’ all of the rules.
The ‘sigma’ level has NOTHING to do with which rules you choose.
What do you mean by: “if your process is 3 sigma the rules must be 4”?
I am concerned by your statement regarding the ‘standard’ level: control limits have NOTHING to do with the specifications (or “standards”)

Bond (or wire) pull in semiconductor assembly is a non-homogenous process as the pull strength will vary by wire lot, bonder set-up and pad location (which affects wire length). You cannot simply blindly apply the rules, sample size or sampling frequency or the type of control chart.

The type of chart and the rational subgrouping scheme you choose are far more important than which rules you use. At your level of understanding you should use them all and keep referring to Miner’s explanation of the rules in order to learn.

Start reading Donald Wheeler’s articles on SPC. You can find them at SPCPRESS.com. You will want to start with rational subgrouping and the myths of SPC. read all of the articles. SPC is a deep and complex science. It is not enough to take some data and throw it into a software program…that is malpractice.

Wheeler, Donald J., “Myths About Shewhart’s Control Charts”, SPC Tool Kit column, Quality Digest, September, 1996 http://www.qualitydigest.com/sep96/spctool.html

Wheeler, Donald, J., “Foundations of Shewhart’s Charts”, SPC Tool Kit column, Quality Digest, October, 1996 http://www.qualitydigest.com/oct96/spctool.html

Wheeler, Donald, “What is a Rational Subgroup?”, Quality Digest, October 1997 SPCTool

Wheeler, Donald, “Rational Subgrouping”, Quality Digest, June 2015 http://www.qualitydigest.com/inside/quality-insidercolumn/060115-rational-subgrouping.html

Wheeler, Donald, “Rational Sampling”, Quality Digest, July 2015 Rational Sampling

 

[Semoi]

It sounds stupid, but the key point of continuous process improvement is that you continuously improve the process. This also applies to the selected SPC rules. Thus, you are probably wrong, if you believe that you just have to select the SPC rules once and then you are done.

 

[Miner]

The problem with using all of the rules is that you increase the risk of a false alarm with each rule that you add. This is the same problem as the familywise error rate seen with making multiple comparisons.