Capability Study for a Specific Thickness of Rubber Sheeting

PABSR

Inactive Registered Visitor
Hello to all and thanks for your valuable help!
I work in a rubber roll manufacturing company. One of our customers is asking a Process Capability study performed on a certain quantity of rolls, and for a specific thickness. I understand that Process Capability is used to demonstrate whether a Process is capable, and not to demonstrate the capability to manufacture a determined specific thickness rubber roll. In other words, the process and its several steps can produce rolls to any thickness. If I demonstrate the process is capable at .125" thickness, could it be assumed the process is also capable for let's say 1/16" thickness? I think that yes it could be assumed the process is capable for 1/16" as long as the same steps are taken to produce a .125" and a 1/16" roll. The only thing that changes is the machine set up to a different thickness.
I will really appreciate your comments and help.
Thank you. :)
 

Miner

Forum Moderator
Leader
Admin
Re: Capability study for a specific thickness

This will heavily depend on the process. Certain types of processes may have constant variation regardless of nominal dimension. Other types of processes will have increased variation at larger nominal dimensions. I recommend that you perform a capability study at at least three different dimensions (i.e., bottom, middle, top) of your ranges.

Perform a test for equal variances. If the variances are equal, you are in luck and you can use one capability for all dimensions. If the variances are not equal, plot the variances versus the dimension. If the plot appears linear, perform a linear regression and use the model to predict capability at a given dimension. You may also try using the coefficient of variation (CV = SD / mean) in a similar manner.

If it is not linear, perform two more studies at the quartiles of the range, and perform a curvilinear or nonlinear regression to predict capability at a given dimension.
 
1

12stones

Re: Capability study for a specific thickness

The only way to ensure that the process would be capable at 1/16" would be to perform the same study you did at 0.125" to ensure that it is capable. Depending on your machines, is there any other variability that could come into play with a different machine setup? As Minor stated, it depends on your process, and you can't assume that because they're virtually the same, that they'll produce the same results.
 
A

Al Dyer

Another thing to consider might be the R&R of your measurement system as the Shore Duro reading on the material could show significant variation.

Al...
 

PABSR

Inactive Registered Visitor
Re: Capability study for a specific thickness

To All...my sincere thanks for your help.
Miner, when you say: 'If it is not linear, perform two more studies at the quartiles of the range, and perform a curvilinear or nonlinear regression to predict capability at a given dimension'. Why would it have to be at the quartiles? Would you recommend any book or site where I could learn more about curvilinear regression, variance analysis and related stuff?

12 stones, thank you. I just wanted to know a good way, and economical to alternatively demonstrate capabilty. I think that if the customer wants a study like this, he should pay the cost to perform it. As he is not doing that, then I had thought that we could show them capability for other thicknesses.

Al, the main characteristic is hardness but not measured in oo Duro, but in Compression Deflection. Thickness is the second characteristic. I've performed R&R studies on both but in different thicknesses.
 

Miner

Forum Moderator
Leader
Admin
Re: Capability study for a specific thickness

Miner, when you say: 'If it is not linear, perform two more studies at the quartiles of the range, and perform a curvilinear or nonlinear regression to predict capability at a given dimension'. Why would it have to be at the quartiles? Would you recommend any book or site where I could learn more about curvilinear regression, variance analysis and related stuff?

Al, the main characteristic is hardness but not measured in oo Duro, but in Compression Deflection. Thickness is the second characteristic. I've performed R&R studies on both but in different thicknesses.
It does not have to be precisely at the quartiles. Doing so gives you five evenly spaced data points with which to estimate the relationship.

Here are some online resources:
Curvilinear regression refers to the fitting of polynomials (i.e., y = B0 + B1X^2 and B2X^3). Most statistics software packages perform this type. Nonlinear regression is more complex (e.g., y = e^(B0/x), and fewer packages will do this. Minitab 16 is the first version of Minitab to do so.

Since your characteristic of interest is compression deflection, I suspect that you will find that it is not a linear relationship.
 

bobdoering

Stop X-bar/R Madness!!
Trusted Information Resource
Remember this - you have two variations in converting type processes - left to right across the web and over time down the web. The variation is different, but needs to be considered in overall variation for capability. To sort it out, you need to do capability at separate locations across the width of the sheet. You also have lot to lot variation, depending on raw material variation, and vulcanizing oven variation, etc.

Always cosndier the total variance equation when developing a plan to evaluate capability.
 
Top Bottom