Cmk (Machine Capability) - Case study

M

michu

I'd like to know your interpretations for attached capability study results (especially regarding normal probability plot). Presented data was send to me by supplier as a machine capability.

Thanks in advance

Michu
 

Attachments

  • machine capability study.pdf
    47.3 KB · Views: 1,256

Tim Folkerts

Trusted Information Resource
Re: Cmk - case study

The first thing that I notice is that the gage resolution is quite poor. The data readings are all either 0.078 or 0.079 (the majority being 0.079) ; the spec limit is 0.080. In other words, the number of distinct categoriwes (NDC) is 2 ... AND the spec limit would be the next distinct category! With most of the readings withing 1 unit of the spec limit, I would question carefully the ability of the instrument to truly gage the results.

Given this resolution, then the rest of the analysis is a little shaky. The numbers might be good, but I don't have a lot of confidence in them.


Tim F
 
M

michu

Re: Cmk - case study

My first impression about result was exactly the same. What's your opinion about normal probability plot? Is it possible to say, based on presented results, that data were taken from process with normal distribution?
 

bobdoering

Stop X-bar/R Madness!!
Trusted Information Resource
Re: Cmk - case study

My first impression about result was exactly the same. What's your opinion about normal probability plot? Is it possible to say, based on presented results, that data were taken from process with normal distribution?

Probably a better question is: would you expect a normal distribution? What type of process is this? Why is there no lower specification limit? A little more information is needed to really understand the data!
 

Miner

Forum Moderator
Leader
Admin
Re: Cmk - case study

My first impression about result was exactly the same. What's your opinion about normal probability plot? Is it possible to say, based on presented results, that data were taken from process with normal distribution?
The p-value from the Anderson-Darling test for Normality is 0.000, which indicates that the data are NOT Normal. However, this may be the result of inadequate gage resolution as previously discussed. Have you performed an MSA on this gage?
 

bobdoering

Stop X-bar/R Madness!!
Trusted Information Resource
Re: Cmk - case study

The p-value from the Anderson-Darling test for Normality is 0.000, which indicates that the data are NOT Normal. However, this may be the result of inadequate gage resolution as previously discussed. Have you performed an MSA on this gage?

I agree - there does seem to be a lack of resolution to determine exactly what is going on with this distribution. But, if this machine capability is for precision machining (which is the need to understand the actual process) and totally from tool wear, then the expectation would be that the distribution would not be normal. What is also missing from the analysis is the variation in time order (run chart), e.g., are all of the smaller data before, after, or in between the larger data? Big difference in the interpretation! :cool:
 
Top Bottom