Why Doesn’t SPC Work? Part 2

To err is human, but to really mess things up you need a statistician.

Have you ever met people who “do” statistical process control (SPC) only to get some screwy-looking control chart, and then text: OMG I H8 SPC! (If you don’t understand that, ask your nine-year-old child or grandchild.)

Last month we saw how it is not a failure of SPC, but rather an EBKAC (error between keyboard and chair). As I wrote in my last article, “Why Doesn't SPC Work?” perhaps they are not doing the measurement system analysis first, or perhaps autocorrelation in a continuous process. But you batch process folks are not off the hook, which is what this month’s article is about.

A batch process makes a bunch of stuff all at once, like baking a batch of cookies. The batter was mixed at one time, the cookies were put onto the pan and into the oven at the same time, so we expect that all the cookies in that batch are pretty much the same. (If you have ever baked cookies, or run a batch process, I’ll bet you can identify at least three sources of variability within this batch—don’t tell anyone yet though. We already know you're smart since you are reading my article, and you would be stepping on my punchline.)

…

Want to continue?

By logging in you agree to receive communication from Quality Digest. Privacy Policy.

Create a FREE account

Forgot My Password

Comments

Whay Doesn't SPC Work

Either the machine which controls the dimension has a self-correcting feature or, more likely, the operator is adjusting the process based on the last control chart point. This is shown because there is less variation between subgroups than expected.

We had this happen when a customer called to complain of out-of-round on a turned part. Whoever heard of out-of-round on a turned part? - a ground part, yes; but not a turned part. We checked and sure enough, they were right. What happened was that the feed mechanism vibrated out of alignment with the machine. But what has this to do with SPC?

Well, this part was a critical fuel injection housing and had about 14 features that had to be charted. Because the part was so critical, we had one of our better operators on the job. He was extremely conscientious, so much so that he was continuously tweaking the machine to keep everything in the middle. If he'd left it to run the way it wanted to, it would have been obvious that a problem existed. As it was, he controlled it too much.

Thanks for responding

Thanks for responding Richard!

The two reasons you list in your first paragraph would probably not result in the pattern we saw on the second chart, though like Bruce above, you have correctly identified the problem is in the sources of variation.

Automatic control is one of those areas that is tricky, and shows up in different ways on a control chart. But, just as with operator adjustment, a process that is in control is only going to get worse as people or a control system adjust it (this is the lesson of the "Funnel Experiment"). Adjusting the process based on the last point results in rapidly increasing the variation in the means, not in decreasing it. When we used to do the Funnel Experiment, the team that used the control method of "adjust by the amount you were off target in the opposite direction" usually ended up having to walk out of the room and into the hall to do a drop to hit the target on the floor int he middle of the room!

The key on this one was coming up with a reason why the within-sample variation was so large as compared to the between-sample variation. See the next article for the answer!

The Second Control Chart Comments

The data indicates the process is demonstrating a stable and predictable pattern of variation, aka "In Control". The range chart indicates the process variation is being dominated by the within subgroup variation. There is no statistically significant differences between sample groups as seen in the Averages being within the control limits on the averages chart. The averages are "hugging" the process average line indicating further that the bulk of variation is within the sample groups/batches not between them.
Assuming the sampling strategy is the same as was used for the first graph, further investigation can (and should) be confined to identifying the sources of variation within any given sample/batch of data. You could look at an individual moving range chart of the samples to see if there are any patterns within each batch over time or gather additional samples from a batch to see if variation could be traced to time after a batch is mixed, location of samples within the oven, differences between ovens (if there are several) etc. You can continue to modify and refine your sub-grouping of sample data until you can identify a statistically significant difference in the averages. You can use graphical Components of Variation techniques and ANOVA to quantify the sources of variation.

Hi Bruce, and thanks for

Hi Bruce, and thanks for reading and responding!

No, the second chart is out of control - it is showing too little variation in the mean based on what we have seen in the past. (This is one way that the term "control" misleads people - it is out of control because it is unpredictably lacking in variability in the means.) You do correctly identify that the source of variation is different within and between sample points. A simple oneway random effects ANOVA on the data would show something weird, but would NOT indicate a significant effect. Check out the next article for more on that!