SolidWorks Analysis - Accurate?

[Manolis]

I started playing around with the FEA in SolidWorks tonight (I built a wheel for my car, and decided to test it). Does anybody know how accurate it is? Hopefully it's a ways off, because if not, i've got some work to do http://fsae.com/groupee_common/emoticons/icon_smile.gif

[ben]

How much FE have you done before?

In my experience the results of FE analysis depend dramatically on how the user specifies the geometry, mesh and boundary conditions.

If a commercial package is spitting out unbelievable numbers I would say it's more likely to be your modelling that's wrong.

Ben

University of Birmingham
www.ubracing.co.uk (http://www.ubracing.co.uk)

[dancin stu]

A computer will always give the right answer, you just have to ask it the right questionhttp://fsae.com/groupee_common/emoticons/icon_wink.gif

[Denny Trimble]

Are you using "cosmos express" or the full COSMOS/Works? You should be using Works.

As for "funny results", a lot of first time users say "My part's going to fail" because the program spits out a deformed plot that's 500x or 1000x, when in reality, deflection is quite low.

Also, if you have any doubts, do some simple verification studies. Cantilevered beam bending, buckling (can be difficult to set up boundary conditions correctly, but it's critical!), torsion, etc. Compare the results with your hand calcs.

I recommend this book (http://www.amazon.com/exec/obidos/tg/detail/-/156690160X/103-2788874-7335866?v=glance) as a practical guide to using FEA. My school's copy has lived on my desk for a few months now http://fsae.com/groupee_common/emoticons/icon_smile.gif

University of Washington Formula SAE ('98, '99, '03, '04)

[Erik C]

We are supposed to never start an FE model until all hand calculations are complete. The hand calcs should be a simplified version of what you will be modeling. The results of which should be used as a reality check for the FE output.

As far as FE modeling itself, Remember, in FE, a rigid constraint is just that, rigid. For your wheel model the constraints (lugs) will actually flex slightly to allow a more even distribution of load. Try using degree of freedom spring elements.

For the loads, if you expect large displacements you MUST run the model in a nonlinear fashion. One example I have run across is on a flat pressure bulkhead. When run in a linear static fashion, the maximum Hoffman F.I. was predicted to be 5.5 with a deflection of 4". When the same model was run using a nonlinear analysis, the max F.I. decreased to .75 and the deflection to a more reasonable 1.2". Having done the hand cranks prior to the analysis allowed me to quickly see that the linear static model was unreasonable, and that a nonlinear method should be used. (The hand cranks; using formulas derived from Roark's and Bruhn, were within 10% of the predicted FE results)

[Manolis]

I've never done any FE modeling before, this is my first crack at it. This was done with 'cosmos express' - what does the full version do differently?

I'll go to the library today and see if we've got access to that book! Are there any other recommendations for books on the subject? Simple is good - I'm a first year, and I haven't seen any of this before.

Thanks!
Manolis

Wheel Pics (http://fsae.com/eve/forums?a=tpc&s=763607348&f=125607348&m=5726044615)

[Denny Trimble]

Manolis,
Your picture link doesn't seem to work.

Cosmos Express will give you von mises stress plots and factor of safety, but that's it. You can't get deflections out of it, or other types of stress plots. I'd say it has about 10% of the power of COSMOS/Works. Being able to create and probe plots of "x-direction displacement", and discrete tone stress plots, do buckling, thermal, and frequency studies, are all very valuable. Not to mention mesh controls (local refinement), shell elements, assembly contact control (bonded or not). I can go on.

Check to see if your school computers have COSMOS/Works installed- go to Tools/Addins and look for a COSMOS/Works box.

The student edition of SolidWorks is $200, and it comes with COSMOS/Works, COSMOS/Motion (dynamics), and COSMOS/FloWorks (CFD).

University of Washington Formula SAE ('98, '99, '03, '04)

[Manolis]

Wheel Pics (http://www.rx7club.com/showthread.php?s=&threadid=250771)

[Sam Zimmerman]

I just attended a conference where the COSMOS representative was giving a "how to" for COSMOS Express. His statement was that, even with correct modeling of boundry conditions, COSMOS Express is only meant to give the user enough information to narrow down the number of prototypes that need to be built and tested. It is not meant to model a part in such a way that multiple prototypes are no longer needed.

Sam Zimmerman
Vandals Racing (http://www.uidaho.edu/~racing)

[Erik C]

Manolis-if this is your first year in school, I would worry more about understanding the mechanics that go into calculating stresses and not on FE. Very few people who graduate with a ME or AEM degree will use FEM right out of school so a good foundation is important. or at least that is my .02...

As far as using a model as a qualitative tool rather then a quantitative tool. I couldn't agree more. Using FE to compare different concepts is much easier and far less time consuming then relying on FE to calculate real world stresses.

[jerktaculus]

hey .. solidworks is hit or miss. The program uses h-elements - which are very simple - so to guage how accurate your solution is - try and reduce global element size until the solution stops changing (max local stresses & global max displacement) - then you'll have a reasonable idea that the solution is converged.

Like solidworks/cosmosworks always does - it provides a number of extra features to constrain the parts and apply loads - work with the most simple cases so you know the amplitude is right - i.e. don't apply loads as moments and try to avoid constraining entire faces (surfaces) - since all 6 degrees of freedom will automatically be constrained - even if not implied - and this can easily overconstrain the model.

That's all I got for now - good luck.

Jeremy Koudelka
University of Toronto FSAE

[Jackson]

I have a buddy who uses Cosmos/Works and I use Algor.

Both computers analyzed the same part (our upright/spindle/bearing assembly drawn in SW) with the same stresses and mine were a bit lower. (at least 20%)

Cosmos also doesn't allow you to select "half rounds" like the top half of a bearing seat etc. It only seems to allow you to use the whole face. Anyway, the full COSMOS version was putting stress in places that made absolutely no sense. Like stressing the inside of holes when the whole flat around it was fixed etc (and no they weren't stress concentrators)

The other thing that bothered me about it is that it took about half the time to run his program as mine. Our computers are about the same, but he ran a 60,000 node program in something like 7 minutes. My computer would take almost a half hour for something like that running algor. That makes me wonder if the elements are too simple.

Anyway, I think as long as you have a good idea about what you are modeling a first crack at a part with FEA isn't a bad idea. If you have any design experience at all, you should be able to tell if the program is giving you a bunch of $hit. You still have to do the statics analysis of the part and how the loads and torques track through everything.

Nice looking wheel, but it looks hard to make (unless you bought them...) or wanted to machine LOTS of ally or make a die. and they are 19"???? this can't be for FSAE? They don't look terribly strong. I just got done FEAing our wheels and by far the lowest safety factor was in turning. i.e. the thrust couple between the ground and the spindle going transversely across the car in a corner. You say they are good to 11.5g? on a 21" tire we are talking about say 1000lb front weight of car on outer tire*11.5g=11500lb couple on 10.5" or 10,000 ft-lb? I'm calling Bull$hit on that on bud, just from experience http://fsae.com/groupee_common/emoticons/icon_smile.gif Try fixing a small spot on the bottom of the wheel contact patch and pushing 11500 lb split evenly into the bolt holes. Your wheel will cry as it crushes itself.

Considering your RX7 will never see more than 2g, you could design for 3 and call it good.

Best of luck and sorry for the long rant.

Brian
Washington State FSAE

[Jackson]

By the way Manolis;

I was reading through the other thread and...

Magnesium wheels would not be as strong as aluminum. The strongest magnesium alloy is about half as strong as the strongest aluminum alloy. (something like 7075T651)

IIRC the stronges Mag alloy that I have seen has a yield strength of something like 30ksi. This is a bit less than 6061T6 aluminum at 35-40ksi.

Brian

[Joseph]

Sounds like a fun project

I have been working on a wheel center for a three piece wheel. As far as material is concerned I am using 7075 T6. As compared to magnesium it is about 1/3 stiffer and 1/3 heavier so pound for pound it is about the same. See www.matweb.com (http://www.matweb.com). Since I am designing for stiffness the modulus of elasticity is my main concern, and not the ultimate tensile strength. However the great thing about 7075 T6 is that as long as my stresses are below 23 Ksi it should last for 500,000,000 cycles according to matweb, that's close enough to infinite for me. So I have designed for the worst case load condition and included a safety factor and still the stresses are below 23 Ksi. I wish I could help with the FEA but I am using Pro-e. To the best of my knowledge Pro-e does a very good job. Every time I have used hand calculations Pro-e has been within 1%. If you want more Pro-e will give you more. You can define your own mesh and polynomial order along with lots of other stuff, you could spend a week running one analysis if you wanted.

Joseph Kliewer
University of Oklahoma

[Denny Trimble]

Jackson,
For COSMOS, you can use the "split line" feature in solidworks to cut holes into "halves" for more accurate restraints, and to create regions on surfaces to apply loads (example: brake pad region on rotor surface).

Also, the speed of the COSMOS solver was revolutionary when it first came out. Algor doesn't seem to have the same matrix tricks up its sleeve, though I wish COSMOS/Works came with the ability to do beam, shell, and solid elements in one model as Algor does.

And, the default COSMOS solid mesh element is a tet-10, with nodes at the ends and midpoints of the tet element. It doesn't use any shortcuts.

University of Washington Formula SAE ('98, '99, '03, '04)

[eon]

FEA - is a great indicator rather than a great tool.

FEA modelling should only be one item on the design path - a good basic grounding in geometry is just as important ( more so in my book).

I still see examples people designing chassis (for e.g.) without triangulation and saying that FEA said it "was good"!? One of the chief benefits of Formula SAE that I see for up and coming engineers is a chance to understand at a "physical" level what they are trying to achieve in designing a system and not just to sit down with a cad package and learn engineering in a sterile fashion as seem's to be the trend these days.

Before you believe the next FEA analysis think about the variables you can't factor in like:

Localised stress inherint in the material from manufacture; of the material itself - rolled alloy sections for e.g have a different grain structure/hardness gradient thru they're core.

Build stresses (post tensioned stresses) created in the manufacture of said items? i.e. machining can change a material as easily as welding thou much more subtly.

Environmental issues - heat, solvents etc

and finally you should consider finish of the item: i.e. heat treated, sand blasted etc.

Just some thoughts on keeping FEA in its place.

Steev in the 'gong

www.uow.edu.au/~sselby/skdesign/ (http://www.uow.edu.au/~sselby/skdesign/)

"Tell them an they'll forget
Show them an they'll remember
Let them do it and they'll understand"

Founding member of "DNA" the National Dislexics Association.

You think you've got it hard !? .. try spreading vegemite on rice bubbles!!

"Do not meddle in the affairs of dragons - because you are crunchy and taste good with ketchup."

"There are only 10 types of people in this world: Those who understand binary and those who don't."