Opening Design Limits to Accomodate Process Capability



I have several parts that do not fall within the stated design specifications and will never fall within these specifications due to machine capability not being adequate.

Basically the machines we currently use are considered the best money can buy, the previous manufactured part/s are ‘fit-for-purpose’ (God I hate that statement) and are on the final assembly.

What we would like to do is to assign concessioned tolerances to the ‘fit-for-purpose’ parts for future orders. We have a great deal of manipulation with regards to design changes and have full authority in this situation.

The question is, how do we apply a new tolerance inline with machine capability? The current idea is to open the tolerances using +/- 4 standard deviations, now personally I’m not too convinced about taking this route and would like your feedback and proposed routes of applying a new tolerance.

We understand the machines process capability and have measured this capability over the past year or so.

Your help in this situation is somewhat needed!




Since you already know that the design tolerance is tighter than is required for the use, my first question would be what tolerance is REALLY needed for the function? Use this tolerance regardless of your process capability.

If the tolerance needed is +/- 4 sigma then by all means set it there and watch your process closely. However if the needed tolerance is greater than +/-4 sigma you have a little more wiggle room in your process without producing bad parts.



Thanks for the reply JRKH.

I understand your view with regards to the actual tolerance required, but this tolerance has been instigated to facilitate a full interchangeable product where each panel should be able to be removed and replaced with another at any given day anywhere in the world.

Although this panel has been manufactured outside these stated limits, it has been proven to be acceptable in this instance. We will continue to ‘strive’ to manufacture the product to the stated specification limits but will have limits that have been proven ‘fit-for-purpose’ that will be used as an ongoing concession.

The production staff would like these limits to be calculated simply by manufacturing the part as bad as possible but ensuring the part fits the assembly, and using those extreme limits as the concession limits. Personally I would like to use some other method, preferable statistical, to establish these limits based upon the actual performance of the machine and/ or process capability.

Any suggestions?

I don’t want much?


Atul Khandekar

James is right in asking what tolerance is REALLY needed for the function? It looks like your designers have to do detailed Tolerance Modeling and Sensitivity Analysis.
- Atul.


Shades of Motorola, this is one of the typical problems that launched the six sigma process. Design to manufacture. Use the tools in the order in which they are intended and you will get amazing results. ( Note, and you do not have to be a black belt)

Geoff Cotton

Quite Involved in Discussions

If you know at what point the product fails to be functional and you have 12 months of data you should be able to determine if there is correlation between size and functionallity.

If you take the data and plot it, calculate the regression and 95% confidence bands etc. you will be able to calculate your 'realistic tolerance'. (If you have got Minitab this is very easy)

Once you've got that recalculate your process capability based on the new spec.

When I do this at our plant the product/process designers go really quiet for some reason.:rolleyes:

Hope this helps
Top Bottom