Capability Study on a New Machining Process - Issues

M

mayor

All,

I am doing a capability study on a new machining process and have run into some issues. I'll start with a little background info:
- the process is thinning plastic sheets (19"x19") to a targeted thickness = 0.035"
- we take thickness measurements at each corner and halfway between the corners, resulting in 8 measurements per sheet.
- when subgrouping, I use a subgroup size of 8 because this would represent one sheet. Is this OK?
- I have data on 43 sheets.
- Issue one: controll chart (X-bar) indicates the process is out of control.
- Issue two: non-normal distribution.
- I am using Minitab for analysis.
I plan to investigate the process further to see what is causing the issues mentioned but I did run Cpk just to see what I would get. I was surprised that it was ~1.5. I don't think a process can be out of control and be capible at the same time, correct? Am I going about the capibility study the right way? Should I be doing something different?

Thanks.
 

bobdoering

Stop X-bar/R Madness!!
Trusted Information Resource
Re: Capability Study - Issues

All,

I am doing a capability study on a new machining process and have run into some issues. I'll start with a little background info:
- the process is thinning plastic sheets (19"x19") to a targeted thickness = 0.035"
- we take thickness measurements at each corner and halfway between the corners, resulting in 8 measurements per sheet.
- when subgrouping, I use a subgroup size of 8 because this would represent one sheet. Is this OK?
- I have data on 43 sheets.
- Issue one: controll chart (X-bar) indicates the process is out of control.
- Issue two: non-normal distribution.
- I am using Minitab for analysis.
I plan to investigate the process further to see what is causing the issues mentioned but I did run Cpk just to see what I would get. I was surprised that it was ~1.5. I don't think a process can be out of control and be capible at the same time, correct? Am I going about the capibility study the right way? Should I be doing something different?

Thanks.

Might want to share the data - hopefully point 1 is the same point on all data sets, etc., for accurate analysis.

For normal distribution calculations, and X-bar-R charts - there is not relation between capability and control.

If it is non-normal, what distribution does it fit?
 
M

mayor

Re: Capability Study - Issues

All points are the same for each sheet. Example: upper left corner always location 1, upper right always location 3, etc...
I'm not sure what type of distribution I have. I looked at transforming the data using Box-Cox and got a lambda value of -5. I was not sure about this approach given the non-normal dist and being out of control. I have looked at the data and it does look like I have an issue with my sheet holding vacuum table. I noticed that location 1 is always low and location 4 is always high. I have attached the data.
 

Attachments

  • Elsmar.xlsx
    17.1 KB · Views: 328

bobdoering

Stop X-bar/R Madness!!
Trusted Information Resource
Re: Capability Study - Issues

All points are the same for each sheet. Example: upper left corner always location 1, upper right always location 3, etc...
I'm not sure what type of distribution I have. I looked at transforming the data using Box-Cox and got a lambda value of -5. I was not sure about this approach given the non-normal dist and being out of control. I have looked at the data and it does look like I have an issue with my sheet holding vacuum table. I noticed that location 1 is always low and location 4 is always high. I have attached the data.

Thank you. I am sure several people will offer to look at the data and respond - but it may take a little time.

One suggestion right off the bat is to do run charts on each location individually and see what capability you have.
 

Bev D

Heretical Statistician
Leader
Super Moderator
yes you can be "capable" and out of control. Purists will say that fi you are not in statistical control then you are not stable adn tehrefore not predictable. but in most of these cases there are underlying subgrouping issues and/or systemic yet cycles or trends which can resutl in 'out of control' conditions yet you still have a stable and capable process.

One question: what formula did you use for Cpk?

I plotted your data on an XBar R chart and a Multi-vari. You do have some systematic differences in thickness based on position. notably position 1 is almost always the thinnest, 4& 6 the thickest. SOyou dont' have a random distribution within the sheet or the subgroup. within sheet is your largest component of variation...the difficulty with this from a 'vanilla' control chart standpoint is that the within sheet variation is not driven by the same source as the between sheet variation: for vanilla control charts, the within subgroup variation msut have the same basic sources as the between. (this is a rough way of stating it...)

What we can say is that you definitley had changes in the within sheet variation and at a different time ahd shifts in the sheet to sheet variation.

I have attached my first pass analysis
 

Attachments

  • Elsmar Sheet Capability.xlsx
    39.5 KB · Views: 436
M

mayor

I see the same thing and have indicated our vacuum table with results that match the data. Looks like I will be flattening the table. I use Minitab, so the Cpk formula used is min(CPU,CPL), which uses the within sigma. Spec limit is +/-0.0005".

I also looked at the control chart (I-MR) for each location and they are out of control as well, so there is variation sheet to sheet, but I agree that within is larger.

So is it correct to say that the process is currently not capable? Hopefully after correcting the table flatness the process will be improved and be capable.

Thanks for the feed back.
 
J

janedoe

I am not able to open the data file, but you are describing several (nearly) normal distributions collected on one chart. I don't doubt your 'out of control' conclusion, but I suggest you look at the data vales (from each location) independently; test for normality at each location. You may have 8 locations...each one stable...but as a group...they look out of control and possible one (or more) locations is, indeed, out of control; unstable.

Whit a subgroup size of 8, I would expect the data to be normal. (That's the beauty of th XR chart....individuals data from a non-normal distribution will have averages that are normal).
 
B

brahmaiah

All,

I am doing a capability study on a new machining process and have run into some issues. I'll start with a little background info:
- the process is thinning plastic sheets (19"x19") to a targeted thickness = 0.035"
- we take thickness measurements at each corner and halfway between the corners, resulting in 8 measurements per sheet.
- when subgrouping, I use a subgroup size of 8 because this would represent one sheet. Is this OK?
- I have data on 43 sheets.
- Issue one: controll chart (X-bar) indicates the process is out of control.
- Issue two: non-normal distribution.
- I am using Minitab for analysis.
I plan to investigate the process further to see what is causing the issues mentioned but I did run Cpk just to see what I would get. I was surprised that it was ~1.5. I don't think a process can be out of control and be capible at the same time, correct? Am I going about the capibility study the right way? Should I be doing something different?

Thanks.
Your sub-grouping method is wrong.Firstly all eight measurements of one sheet should fall within the range of half the tolerance for thickness.You should take the mean of eight readings as the size of the plate measured.Thus the mean size is one sample. you have to take a minimum of 3 sheets as one group of samples for Cpk calculations.About 15 to 25 groups of such measurements make one set of readings suitable for Cpk calculation.More groups give more accurate estimate of Cpk value.One important note is you cannot calculate Cpk with out-of-control plots.you may delete out of control points from calculations if only one or two such readings are found.But the right thing to do when you see out of control points is to investigate the causes and eliminate those causes immediately.You can calculate Cpk only after eliminating assignable causes of out of control points.
If you are getting your Cpk as 1.5 inspite of out of control points it is unusual.Investigate causes.
Remember the main purpose of SPC is not a postmatem of calculting Cpk.Its purpose is to eliminate assignable(special) causes as and when you encounter them during the production operation.
V.J.Brahmaiah.
 
Last edited by a moderator:
Top Bottom