How to determine process stability for a process whose outputs should be steadily increasing?

[DuncanGibbons]

Hi all,

I am wanting to determine if a process is stable and in control. This process is a metal powder reuse process whereby the outputs are flowability metrics and chemistry. These outputs should be steadily increasingly in a linear fashion up until a set point afterwhich the powder is scrapped. Are there SPC or control chart methods for assessing such processes? I.e. to determine if the process drifts from the standard linear regression line?

Maybe some method of normalizing the data for use with xbar charts? Also if you have any references to such problems that would be great.

Thanks

 

[John Predmore]

If the metric increases in a linear fashion, the increase per unit time should be a constant. So can you control chart on the delta between the current measurement in one time period minus the previous measurement on the same batch?

 

[MOester]

Agreed. If you need numbers to plot on a horizontal chart, and you know the slope of the line it SHOULD be following, just do a coordinate transform on your data to get it horizontal.

 

[Steve Prevette]

I would plot the sequential changes (first derivative) as already suggested. Another thing to consider it treating it like "tool wear" and Bob Doering has quite a few notes on that here.

 

[bobdoering]

Look under X Hi/Lo-R charting for precision machining. That charting is specifically for a function, as you have. Do not waste your time using Sherwhart charts, as they are extremely ineffective for functions - they expect independent, random outputs - and functions do not meet either of those requirements.