Pre-Control Chart - Limits are set at 2/3 of the specification

[Fluidfont]

I'm an industrial engineering major working on a case study for my statistical quality control class. I was able to obtain a case from my friend's work where they are utilizing control charts to control a process. However, I cannot seem to find any information about this chart. Here is a link to it.

http://students.washington.edu - Link was: /cyw/chart.jpg

I am Sorry that I am unable to disclose specific information. But basically on the chart there are 8 same processes done on a single machine, but each has a different target value. You can think of punching holes of different diameters simultaneously into a board if you want. This chart is plotting the deviation of each sample from the target value over time. The upper and lower lines you see are just the lowest and highest values traced. I was told that this was a multi-variable control chart, but I cannot seem to find information on it. It seems like a group chart combined with a running location chart or something. Please let me know what information you can find about this type of control chart.

 

[Geoff Cotton]

Welcome to the cove.

Looks like a "zero mean" chart to me. We use them but only to record one characteristic i.e a spigot diameter.

If I had to guess I'd say they are ploting eight different features (holes) and their deviation from the mean.

 

[ben sortin]

It is a "pre control chart." The limits are set at 2/3 of the specification. The recordings of the eight features are uniquely identified by color and symbol. The lines are plotted at the extremes of the subgroup. This identifies which characteristic has the largest deviation from the target in both directions (high/low). I would react to any point outside the limits and any one symbol which ends up being plotted on the same line seven times in a row. Keep an eye on any one characteristic that jumps from one line to the other. Not bad for tool wear monitoring on an eight station screw machine (Acme Gridley circa 1978).

 

[Darius]

I tink both are right (Geoff and Ben)

As Geoff said, the Min and Max lines on the chart is a Group Chart and with the data centered on Zero looks like a target chart (so a Group - Target Chart).

Normally this kind of chart only Max and Min are charted (not control limits and not the other points).

The spec limits are charted on the chart, that's wrong for SPC, unless a precontrol chart, that's is why Ben is right.

So it looks like a mixture to me.

For group charts: There is too litle about them on the net

Innovative Control Charting - Practical SPC Solutions for Today Manufacturing Environment- By Stephen A. Wise and Douglas C. Fair. Editorial ASQ

 

[bpritts]

Looks like some of you folks remember more from your CQE than I do!

Brad

 

[Fluidfont]

Well, They said they use this control chart because it is able to give a full overview of the process. It is too much for them to make an individual X bar with whatever chart.

Ben- Each plot is not an individual feature. They are all identical processes with different nominal targets. Yes, they do spin also. Lets keep that on the DL.

Darius- I was thinking target and group combined also. What is gained from doing this? Other than reducing the waste from having to keep 8 individual measurements.

 

[Darius]

It may also be a Group short run target chart if different variations are expected from the different locations.

I agree that the most of it is reduce the number of chart but Group Charts are used for evaluating the consistency of the characteristics, parameters (from location to location) or process streams. Are recommended also when monitoring processes where x and y or x,y,z axes are simultaneously monitored. On similar but independent processes produccing simular a product not only cuts down the number of control charts but also helps to focus on processes whose output begins to change significatly compared to the other.

Interpretation:
One should look for consecutive repeating locations appearing in the MAX or MIN (four or more is a run and indicate the presence of an assignable cause), or check for the percentage of the presence of the attribute on the MIN or MAX (also with the ranges)

It’s important that the number of locations should be held to 5 or less, if more the likelihood of missing a trend is higher.

:thedeal: