User Feedback Comparison Between CETOL-6sigma versus 3DCS for doing Tolerance Analysis Studies 1

[Al James at General Atomics]

I need to put together a "non-scientific" user comparison of CETOL-6sigma versus 3DCS, based upon feedback from the customer base. I really need to get data for both software packages, in order to make it a complete and fair comparison. It should also take into account the same CAD software package; such as CREO, Solidworks, or NX to be a true "Apples to Apples and not Apples to Oranges" type comparison.

In particular, I'm trying to get CETOL-6sigma vs 3DCS comparative information as follows:
a. How long does it take to get up-to-speed on the software?
b. How long does it take to re-learn the software after 2-3 months away from usage?
c. How do users feel about the “User Friendliness” of the software tool?

Please advise me where else this information may already exist.

Thanks,
Al
Principal Mechanical Engineer and Program Manager
General Atomics, Electromagnetics Systems Group

 

[3DDave]

I used VSA, which looks suspiciously like 3DCS. At the time we compared it to CE-TOL; that was a while ago.

There was only a short learning curve to VSA software - it was rather easy to use.

CE-TOL at the time depended on the underlying CAD package to regenerate with variations to determine the sensitivities, while VSA creates an internal vector representation. For CE-TOL, the regeneration/recalculation method meant that using the same part twice in an assembly caused that particular variation to be identical across all the places the part was assembled. I expect they fixed that problem, but it would be interesting to check.

In fact, between initial training and finally getting access to the software there was a several month interval. Not so much getting back on the bike, but having been shown how to ride the bike and then no practice - it was a problem because I forgot one step. The VSA support was helpful and I was on my way once that step was identified. I spent about 10 months with the software after that and had no further problems. It's been a decade since then and I am confident I would not have trouble restarting. Lesson: Don't run through an introduction and then shelve the engineer and expect a solid restart.

The internal chain for VSA was to create points and related vectors, create a C program that represented the variations, run the program and harvest the individual outcomes.

The benefit to using vectors was that VSA allowed adaptation to any computable outcomes - if you had a program that took measurements in then VSA could send those measurements to that program and take results from it, converting any analysis into a variation analysis that was based on simulated manufacturing variations - I expect it would be slow, but since the alternative is never, it's a step up.

The down-side was that VSA is a statistical simulator; various runs will result in slight variations in output. The up-side is that this is exactly what one expects out of the factory. CE-TOL was based on idealizations that a real factory will not produce.

At the time, I used the VSA version that was integrated with Pro/Engineer, which meant building the model exactly the way it was to be analyzed. For users who like to model one way and create dimensions after they think about manufacturing that was a problem, but I think the later versions were CAD agnostic and allowed users to do what they wanted; this means depending on a similar level of discipline to that of FEA users.

One thing that is hugely important either way is that without factory variation capture, the analysis with any tool is just a feel-good exercise. One can allocate tolerances in a myriad of ways and the usual concept is to eliminate large contributors, but in the absence of cost and performance information it means little.

The only friendliness issue I really dealt with was the factory. I expect you'll have better luck there, but I tired of the factory hiding its manufacturing errors with grinders at final assembly. The most irritating was them using open setup welding with "two to four passes" as the guide to repeatability. They expected VSA to predict weld distortion.

One area that neither program will help with is picking the right problem ahead of time. It takes more than a little imagination to do that right. There were fit-up problems with standards compliant extrusions that no one thought to validate. The problem on the factory floor was not having a guy big enough to shove a piece of angle into a corner hard enough to make up for the Aluminum Association tolerance. No tolerance analysis software would calculate the force required. 0.500 thick flange is tough to muscle into place. This is a common failure in Dimensioning and Tolerancing - so little stress analysis seems to be done on variation stacks. And, as I indicated before, plugging in actual factory variations, not just designer guesses, are the only way to predict what delivered product performance will be like.

 

[Al James at General Atomics]

Dave, Thank you for the helpful feedback aforementioned.

 

[Al James at General Atomics]

3DDave, Do you know anyone else with similar experience using both CETOL-6 Sigma and 3DCS, such that I can get a much better comparison between these two software programs?

Thanks,
Al

 

[3DDave]

Unfortunately I do not.

Perhaps Tom Rhodes is still at it:

https://www.eng-tips.com/viewthread.cfm?qid=166356

 

[Al James at General Atomics]

Does anyone else on Eng-Tips know someone with similar experience using both CETOL-6 Sigma and 3DCS, such that I can get a much better comparison between these two software programs? I'll even settle for use of one or the other software program to get started.