Thread: drilling holes in the chassis

Serg, don't put too much faith in a FEA plot unless you know exactly the situation it has been produced for and what it was intended to show.

I use FEA a lot, mainly to back-up my hand calculations (if both agree, I'm confident in the result and move on, if not I start checking both for errors) but also for geometry that is too complex to calculate by hand.
But I always back up the plots with hand calculations, get equivalence and then start adding complexities into the model.

I also greatly simplify FEA plots to save on computer time. This still makes the results useful to me, but they can mean absolutely nothing to another engineer who just sees that plot in isolation without a whole paragraph explaning what it is and isn't meant to show.

Representative loads are another one. If you don't know exactly how much load something will see (preliminary work) but you are using a plot to highlight the weakest link in the structure. In a case like this the stress distribution means everything (if one member is showing 3x the stress of another etc), but the actual numbers aren't relevant.

It's also important to understand the limitation as all but the highest cost FEA programs are linear. Meaning the results are only good for small displacements and double the force will give you double the deflection.
If you have a large deflection the shape of the structure can change enough that the original undeflected shape isn't a good starting point any longer.
To get a good result in this case you would have to pre-calculate the deformed shape, deform your model to this and then run the FEA simulation. This is an iterative process that can suck up whole days of hard work. It's got to be pretty special to command that much work over a simplified run that will give a good-enough result.

The last point I'd make would be linear FEA is only any good at less than the material yeild strength. It can't handle any material yeilding as this changes the stress distribution within the material and changes the result completely.
Basically you can't accurately predict failure loads, but you can use to to confirm what is safe.