General Assembly Tolerance Question

[nds88]

I recently worked on a project that had many components that required high precision tolerancing at interfaces to control a desired outcome. I don't have a ton of GD&T experience and so I got some help from others around me at the time. I often felt like the possibilities of misalignment were endlessly spiraling. Has anyone else experienced this? I thought maybe in Aerospace or R&D this might be common.

What methods did you use to do a stack-up or tolerance analysis? Are there any modernized software methods? I noticed Solidworks has a feature but I found it hard to learn without spending significant time in it. I ended up doing everything I could by hand or excel.

 

[3DDave]

I don't use stack-up analysis. I use tolerance allocation after first determining the useful limits and combine that with choosing designs that are less sensitive to small variations causing large variations.

But, if one is stuck for it, look at VSA (Variation Systems Analysis) as available from DCS.

https://www.3dcs.com/tolerance-analysis-software-and-spc-systems/3dcs-software

I used it when it was a separate company as an integrated Pro/Engineer product. It made the analysis portion very easy. What is tough is for that analysis to be useful, like any other analysis, one needs to know what the variation the producer will create so that one can determine if the rate of acceptable parts is worthwhile. Too often what a manufacturing operation says they can "hold" is far different than what they actually do, wasting time if what they claim is what is analyzed.

 

[GregLocock]

As an example of that, various parts of a suspension are machined to locations within 0.05mm of nominal. We then fit them to a car body where the mounting holes on one side of the car might move by 3mm relative to each other , and across the car I worked on one a long time ago where there was a 12mm slip joint in the floorpan.

We used to use VSA, i think CATIA has something similar, but I don't work with the stamping people any more.

You might wonder how we bolt a subframe, which is typically a two piece clamshell pressing, hence the holes are pretty accurate, into a body shell which is a welded assembly of several stampings. There is one master bolt hole, the one on the same rail has a longitudinal slot, and the two on the other rail have big clearances.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[nds88]

Thanks for the replies, I will look into both VSA and DCS. I'm kind of surprised that with all of the analysis and CAD software out there in the world that there aren't more widely used tolerance programs. I get the feeling that if a design requires so much precision to meet an objective then that probably makes for poor design practice anyway.

 

[3DDave]

It's not much used because analysis doesn't produce a number but a range of probabilities. Even if the software was free there was only a tiny number of engineers I ever worked with who were concerned with the sort of answers VSA produced.

 

[SnTMan]

I get the feeling that if a design requires so much precision to meet an objective then that probably makes for poor design practice anyway.

Now there's a thought

Regards,

Mike

The problem with sloppy work is that the supply FAR EXCEEDS the demand

 

[CWB1]

I cant speak to Solidworks, but Creo and others have the ability to graphically demonstrate/analyze MMC & LMC based on material and other inputs to perform stackups, interference checks, and even FEA/CFD/other analysis at those conditions.

I get the feeling that if a design requires so much precision to meet an objective then that probably makes for poor design practice anyway.

Not necessarily. As engineering tools and manufacturing continue to improve over time we are naturally going to continue packing more into less space using few materials, and need to analyze the fit of everything. Compare anything made today with an equivalent product from 1960 and I'd wager you'll see more the need for a lot more stackup analyses.

 

[hendersdc]

I don't like the built in solidworks tolerance analysis tools because it requires you to have built the solid model the same way you want to dimension it on the drawing. Due to the iterative nature of design, that rarely happens for me so excel spreadsheets are the way to go.

In stackups with more than a few tolerances included you don't want to look at worst case stackup, you need to look at statistical approaches.

If things are still proving difficult consider adding bias features such as reference edges/hard stops/alignment pins/etc that you can shove one part up against on another part and leave clearance in your mounting holes. You can also make use of alignment jigs/fixtures that can allow you to have tight tolerances on only the critical features and relax the rest.

 

[3DDave]

The disadvantage to building models that don't behave according to the dimensioning and tolerancing scheme is they are hugely expensive to troubleshoot when a dimensional change to the drawing is required; often someone else given the task will make a change and then the entire model has to be "reinspected" 100% looking for any of the other, possibly among hundreds of dimensions, for one that changed that should not have. It's much easier for the person who knows what the model is for to just redefine the model to match the desired scheme than to make endless feature shims to make it look like it meets the requirements.

 

[nds88]

I can see the disadvantage to creating a model that controls tolerances, as 3DDave mentions, but I work in a place that is a bit of a revolving door and I had been picking up drawings on an assembly that had been kicked down the road from someone who no longer works here. After sorting through some complicated interfaces of parts with little GDT experience and lots of frustration I said there had to be a more effective way of figuring out if it can work, if it can be held by the machinist, and if it can meet requirements. One of my bigger frustrations was trying to determine an angular alignment based on contact to a surface profile...

 

[3DDave]

Angular changes like that are frustrating to model in a solid modeler. The VCA/3DCS software handles this case easily by affecting the surface contact to effectively change the surface normals.

 

[EdStainless]

In some equipment stackup issues are inevitable. We used to make multi stage pumps, often 60 or more stages in one assembly. Even after a lot of part consolidation that gave us nearly 200 parts on a 24' shaft.

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, consulting work welcomed