Frame FEA How to?

[Erick Scarpone]

We found multiple post on this topic but haven't found the answer to our problem.
We have the geometry for the chassis, suspension points etc etc etc.
We done the calculus of the moment of inertia (Pilot, Engine Frame and Wheel uprights, brake, A-Arms assembly, were using the info of the torque of engine) basically that are all the loads were going to consider at this time.
We know were and how to put this loads in different condition: Brake, Acceleration and Curve handling.
Our big problem is were to constraint the frame for our FEA.
I think in putting them in all the suspension joints with the chassis
My partner thinks that the constraints are different for every case,
and we are even thinking of not using any restraint so we can see how it all deforms!!!
So as you can see were in a bit of a problem!!!
This is are first time doing the FEA in the Frame, we have some FEA experience very simple so we know how to interpret the numbers.

If you could tell us how you constraint your frame we would really apritiatesd!!!
Thanks

[Erick Scarpone]

We found multiple post on this topic but haven't found the answer to our problem.
We have the geometry for the chassis, suspension points etc etc etc.
We done the calculus of the moment of inertia (Pilot, Engine Frame and Wheel uprights, brake, A-Arms assembly, were using the info of the torque of engine) basically that are all the loads were going to consider at this time.
We know were and how to put this loads in different condition: Brake, Acceleration and Curve handling.
Our big problem is were to constraint the frame for our FEA.
I think in putting them in all the suspension joints with the chassis
My partner thinks that the constraints are different for every case,
and we are even thinking of not using any restraint so we can see how it all deforms!!!
So as you can see were in a bit of a problem!!!
This is are first time doing the FEA in the Frame, we have some FEA experience very simple so we know how to interpret the numbers.

If you could tell us how you constraint your frame we would really apritiatesd!!!
Thanks

[Carbaholic]

Different People will give you different answers to that question because everyone does it just a little bit different. There is a balance to be looked at. On the one hand, for the best results, you would want to create an accurate representation of everything on the car, and apply the constraints for each load case through the tires and the suspension members, allowing the tires to flex and the suspension to travel under the load case and give you very accurate force paths. (this would be extremely complicated and would take forever) On the other hand, you can get quick results by simply applying a 0 displacement constraint to some of the suspension mounting points. In this case you end up asking yourself if your load paths are really as accurate as they should be.

Since I'm learning about all of this and honestly don't know the best solution, I'm going to get a few quick solutions and then get as accurate as time permits and then I hope to be able to compare my FEA to drive time measurements and see how close the different methods get me vs. how much time they take.

Eric

[Kamil S]

im not big on FEA, but I have done a few CFD models and I agree with carbaholic on the method.
Start with a really simple model, make your nodes neither too accurate not too coarse, maybe just something that your processor can handle within a 15 minute limit or so. After the results come out, take a look at the overall trend of the analysis, and look for problem areas (where you have a lot of transitions and discontinuities).

If you are making your own code, or if you are using a code with multimesh capabilities, then refine the grid just at these areas, and keep the grid wide enough at continous parts of your frame. Again, not sure how this goes in chassis frames but I would suppose you need to refine your mesh at all joints/connections, and keep them wide at through the tube lengths.

If you can't do the above for any reason, then I would suggest running a simulation at a feasible grid. Take the results of that, and apply them as initial conditions to start another simulation with a finer grid. Just iterate this process until you are satisfied.

There are a ton of other ways to beat around the time it takes for processing, but it all depends on what you want. If you're looking for steady state results only, then accuracy isnt really an issue in my opinion, as long as the analysis method is stable. If you're looking into transient effects and time dependent results, then I'd recommend to take your time developing them. Just remember that any time you put in analysis takes away testing time (so to speak), and real test results are always better.

[Erick Scarpone]

Were definitely looking for a steady state results (It's our first time doing it to the frame so were going step by step), we done some mesh work going from a raw mesh to really complex one and timed it. Done some Time vs Results graph.
Those were not the frame, were working right know in Cosmos Professional our problem is not actually the mesh or putting the loads is what to constrict, like carbaholic says he constrict some of the suspension points.
But witch suspension points? why some? why none? Why not all? Does anyone compare there results to physical test.
We can't find answers yet!!! We have been stuck now over a week with this.
Some Background:
Were trying to do a fair good space frame for next year's cars, seeing how much it will deform with the loads that we applied. Since we don't have telemetric or DAS were working with vehicle dynamic and the times in the different FSAE results.
Thanks to all that you response your all down in my thank you letter inside the thesis soon to be done......I hope!!!

If some has something more to say about constraints it will be very much appreciated!!!

[Big Bird]

Hi Erick,

A very good question you ask here, and certainly a point of contention.

The base issue is that there are multiple points where the tyre loads are being input into the chassis - I can think of 26 points off the top of my head for the car I was most heavily involved with. (If you wanted no expense spared, F1-level-or-above simulation of these load inputs, that would require a testing rig with 26 attachment points). And even for a given dynamic condition, (say, cornering 1g) the proportion of loads at each of these points depends on such issues as suspension component deflection, spring rates, toe bases, blah blah blah.

Without wishing to send you screaming towards a career in basket weaving, I'm only making the point that there is no one single method or metric that will give you an absolute performace comparison between different vehicles.

So you have to develop a reasoned strategy within your team to give a good estimation of results. A series of FEA analyses is certainly helpful, but at some stage in the process you need to validate the FEA results to give credibility to your results. (Note I said at some point - not at every point).

If you could be bothered, I'd suggest something like the below three-point strategy.

1). Firstly, a good estimation of structural efficiency can be achieved with a FEA normal modes analysis, (i.e. natural frequency analysis). Auto industry standard is to do this unconstrained, i.e. with the model floating in free space. Any constraints will artificially stiffen the chassis between those points.

Natural frequency is proportional to stiffness, and inversely proportional to mass. Since we want high stiffness and low mass for our chassis, natural frequency data gives one metric to compare overall structural efficiency. Also, if you animate your results, the way the chassis twists and bends gives you some idea of strong and weak points in the design.

Note that the first six modes you will get out of your results will be rigid body modes - nonsense modes in this case, and a by-product of how the program solves the problem. Your first sensible mode should be the seventh result, which should be around 30-50 Hz depending on your chassis design. Maybe more if you have done well.

2) The "simple" torsion test. Whilst not 100% accurate in determining overall chassis torsional stiffness, (whatever that is!!), you primarily use this stage to validate the accuracy of your FEA analyses to the judges.

Pick some stong points in the chassis somewhere near the front and rear axle lines. Stick two long, stiff steel bars through these points. Restrain three points, load the fourth and measure the deflection. Whilst not 100% accurate, this will give you a ballpark figure of your chassis torsional stiffness.

Repeat the above in FEA, and refine your model until your FEA results are somewhere close to your measured results. I've heard industry heads quote figures that FEA data accurate to 10% is good enough. Don't forget to model your steel bars as flexible beams, and not rigid links - flex of the bars influences the measured deflection in the real world test.

Maybe physically test just the loading beam and FE model that as well - so you know the contribution of beam flex in the overall variance. Should be minimal.

3) Do a "more accurate" FEA representation where the load inputs come through modelled suspension links. The most recent one I've seen involves modelled A-arms and pushrods attaching to the relevant points on the car, and attached to the tyre contact patch. Torsion applied at the contact patches, results measured. This is a more accurate representation of how flex works in the real world - although I don't think anyone has a perfect answer of how to perfectly replicate it.

If you had too much time on your hands you could try to build a physical rig that replicates step 3 - but I think the variances in modelling the rig versus building it could start to overwhelm the chassis variances, and who knows where you would be then.



I only suggest the above as it gives an efficiency metric (stiffness per weight) via the normal modes, a "working" estimate via the modelled suspension links, and a validation of your FE via the physical torsion test. No one of these methods is 100% accurate on its own, but the combination gives a good comparison of how one chassis performs compared to another.

Have fun, good luck

[Erick Scarpone]

Big Bird thank you so much, we were sure of doing the torsional test, (simple and it will give us a good way of seeing if we are off the track in our FEA once we do a physical test)
So I read tip #1 and #3 and in my understanding we should not put any constraints?
Just putting all the loads in their suspension points (We have 22) and see what happens?
We basically took out all the forces out of the chassis and do a equivalent force applied to it.
Like tires Friction, Inertia, Acceleration, Weight, Blah Blah decomposed in (x,y,z) and found the equivalent force on every coordinate system but as it acts on the frame.

P.D. Thanks so much we have a member of our team going to RIT next week for he internship he will sure pass over there, I think he wrote you guys recently