Should I Use FEA and What 2

[Philbur]

I am a mechanical engineer with no FEA experience. I am exploring the possibility of using the technique to analyse stresses and deflections caused by externally applied point loading during alternative piston ring manufacturing processes. My interest is for personal use in relation to my hobby of model IC engine design and construction. I have four questions.

1) Is FEA appropriate or, bearing in mind my lack of experience, are there more appropriate techniques.

2) What kind of learning curve can I expect (hours) to get up and running. I've look at a 300 odd page user manual for UNA 7.2,it looks rather intimidating.

3) What free or low cost software and books would be most appropriate.

4) Is plug and play for this relatively simple application a possibility or am I dreaming.

Thanks for any help
Phil Burman
Stavanger, Norway

 

[GregLocock]

1) probably FEA will do it, but for such a simple case I'd have thought a hand calculation might be more appropriate.

2) To produce pretty pictures on a screen: 2 days with a decent package . To produce a validated model of anything you'd trust your life on: up to a year, or never for some people.

3) The demo version of Femap / Nastran for windows is nice to use. It has a 300 node limit, but apart from that it is fully featured. If your problem is symmetrical then you can effectively double that, almost.

4) Well given that if it goes wrong you've only lost money and time, and that comparative results are easier than absolute ones, I'd certainly give it a try if you think you'll use FEA in the future. If you do decide to go with nastran for windows I'd be happy to work in parallel with you, once you've done the tutorials. I'm a bit rusty on this linear stuff (grin).
Cheers

Greg Locock

 

[Philbur]

Thanks for the help Greg.

I've requested the demo your recommended and have also downloaded the full FEMAP Basic version and started work with the tutorials. But of course it's not fully functional, it appears the solver is missing or not usable without the dongle. I've also started studying the engineering theory behind my particular interest so I should soon now if hand calcs will be enough, but I doubt it. Although the component is relatively simply my interest is to analyze the behavior in some depth. My need is to determine the geometry of an uncompressed none circular ring that is circular when compress and which will apply a predefined, uniform distributed load against the cylinder bore. From this I then need to determine what point loads need to be applied to a circular none loaded ring in order to approximate the necessary uncompressed none circular ring geometry. It is clear that I need to do some hand calcs to more fully understand the issues involved before I use an FEA model. As the ring is symmetrical I currently assume that I will only need to model half of the ring. So I have some more questions

1) Do you thing 300 nodes will be sufficient, is it borderline or is it obvious enough etc? After going through what is going to be a very steep learning curve I would like to avoid the shock of realizing I need to spend $x,xxx in order to achieve my goal.
2) If 300 nodes is not enough what can I expect to pay for a suitable application? I spent 20 minutes looking for prices on the EDS site without success. Are they scary a person may sue as a consequence of heart problems induced by unsupervised exposure to their pricing policy? I have requested prices.

It was an interesting coincidence to find myself in and engineering presentation today that included a brief presentation of an FEA of a complex sub-sea pipeline structure subjected to bending and tensile stresses during deployment from the laying vessel. The point being that the experience I gain from a foray into FEA will at least be useful in giving me a more realistic and knowledgeable overview of FEA.

I thank you for your offer of help and realize that everybody’s time is valuable (at least to themselves) so I will take care not to ask questions that I can reasonable expect to find in the documentation. I anticipate most of my questions will relate to how best to apply the techniques to my particular application.

Thanks
Phil

 

[GregLocock]

Ooh, I have given you a bum lead there. The solver in the demo I have is MSC Nastran for Windows, so you need to talk to the Nastran people. It uses FEMAP as a preprocessor, that's where I was confused. If they don't do that any more then I'll do it another way.

Prices for the full package are around $10-20 000 as I remember.

Once you've learned how to do it using a nice GUI pre and post processor you can then look at the data files (deck) and hack them directly. This will give you enough experience to go and model to any detail you like using one of the free solvers.

I think 300 nodes is enough to learn about the system, particularly since I'm increasingly sure that an analytical approach is viable! If Roark doesn't cover this very load case then a reciprocity method might be helpful (don't ask me, my uni notes have been recycled).

If the system is in pure bending, and given your ring's slenderness ratio I suspect this is a decent approximation, then beam elements will do, so you could have 299 beam elements in your model. That is a lot of detail. Cheers

Greg Locock

 

[Philbur]

Thanks again for your further comments. I am assuming the FEMAP demo CD will be a fully implemented version with FEMAP solver, limited to 300 nodes. I am not sure if I need the additional hassle of patching different modules together!!!

I am expecting the Demo may take up to a month to arrive, if ever. Therefore I was considering down loading a 1300 node LISA package see

http://www.freebyte.com/cad/fea.htm#fea.

I’ve searched this site for “LISA” with the only negative response “It does not support good geometric modeling capabilities”. As I assume my area of interest (and capability) will be dog meat for any FEA application, my main concern is a short learning curve and a high level of user friendliness. On this basis would LISA be as good a choice as FEMAP. I don’t want to invest time and effort in LISA if I am going to bin it when FEMAP arrives and slide back down the learning curve. I’ve had a bad experience with CAD.

Thanks Again
Phil

 

[GregLocock]

LISA and Nastran for Windows are as night and day in terms of ease of use. Would you like to contact me via my website

http://www.geocities.com/greglocock/index.html

and I'll get you up and running next week?






Cheers

Greg Locock