Determinate 3D sensitivities using small displacements method in CAD

[Desiv5]

Hello everybody,

I am usually doing tolerance analysis "only" in 2D, mostly for mechanisms. As I receive more and more requests for tolerance analysis in 3D (frame structures, etc.), I try to apply a method to "manually" do this, so not using dedicated 3D tolerance analysis softwares, which are (for me) more or less black boxes...
So I've read about the method of "small displacements" - using CAD to move different assembly components in the 3 main directions by small amounts and measure the influence on analysed parameter, to be able to determine the sensitivity of the modified contributor. I don't seem to find any practical guide or example for this method and my question is also, how do I have to "put in equation" all the determined sensitivities for calculating the result in 3D? Of course, I know that, for each contributing parameter, contribution = sensitivity x tolerance_interval, but how can I write the "main function" for my 3D analysed parameter? It would be great to have a documentation, description or guide to apply this method.

Thanks in advance for your answers.

Conrad

 

[3DDave]

Look at the Ad-cats for some information. It is the method used in the former CE-TOL and TI-TOL software (same software, just changed hands) and is currently offered for PTC software. See

http://adcats.et.byu.edu/home.php

Basically you need to roll your own and it won't be hard/easy. You need to have some way to record the measurements and then a way to individually change each dimension and record the resulting measurements. If the system lets you step through the dimensions like an array, then it should be easy, but often they don't, in which case it will be harder.

Most of the dimensions will show zero sensitivity; a tweak to the width or length of a block won't change its thickness.

The harder trick is to assign statistical distributions to the dimensions and use the sensitivities to scale the distributions and then use statistics math to combine the distributions into the final distribution for each of the measurements.