FEA for Dynamic Multybody Physics Fatigue 5

[Timms]

He

I am trying to decide What FEA package I should buy. I need to model the rear supsension of a motorcycle. I hope to be able to animate it and determine the stresses the rear swing arm, axel and differential will be subject to. I would like to model it in Solidworks but am wondering if Cosmos would be sufficient or if I need to use a program like Ansys. I have a weee bit of experience with ANSYS but perhaps not enough to save time on my learning curve. If I get Cosmos I would have to learn everthing from scratch. Could anyone give me some direction?

Blessings

 

[GBor]

This isn't a simple analysis. Cosmos, in the right hands, could possibly do it, depending on strain-rate effects and other complex dynamic considerations. In the same cost category, Algor could possibly do it. For "crash" type simulations, LS-Dyna is probably the industry leader, but if you are just trying to model some typical "over-the-road" motion, you may not need to go to that expense. If you have access to a package already, then use it. If you are looking at purchasing a package to do it and are cost-controlled, look at COSMOS, NENastran, and Algor. If this is a one-time thing, look for a consultant.

Garland E. Borowski, PE
Borowski Engineering & Analytical Services, Inc.

http://www.borowskiengineering.com

 

[GregLocock]

You would have to know a great deal about the system to use an FEA package. The usual way to generate those loads would be to model the forces in the suspension during different events in a package like ADAMS or NASTRAN 4D.

If you do know the loads, or are prepared to work from hand calculations, then any FEA package should suffice.

Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[dmacx]

I would recommend NEiNastran since they have what you need, have great support and are not over priced. Their motion software appears very capable though I have never used it myself (just basing this on what they have sent me). ANSYS is also very capable but more pricy. I can not recommend COSMOS.

 

[rb1957]

i guess you could do everything at once, but i'd preferr to break the job down into pieces. i'd use a dynamic package (maybe Adams) to understand the motion of the system, and the loads applied between the elements of the system. then i'd use an FE package (just about any) to get the internal stresses ... you'd get a bunch of snap-shots.

the advantage of this approach is that its tools that i'm more experienced with (and so i have some confidence with them). to do everything together would be at least a different piece of software, which would take alot of work to get to know ! maybe a job shop could help, if you could trust them !

 

[Timms]

Thanks a bunch for all your responses!

Right now I am concerned with over the road motion. I also have a number of other frames that I will have to model. I need an efficient system.

Currently I don't have any software. I would like to use Cosmos because I don't want to have to remodel everything a second time. 3 of you mentioned Nastran. would I have to remodel it if I went with Nastran? I also know that others have modeled similar systems with Adams and then used ansys. I get so many different suggestions that I find it hard to make a decision without just picking one!

Do they all provide good customer support? Can I trust the sales rep. when they say it can be modeled?

Can anyone tell me the difference between Adams and an FEA program. Does Adams find forces without going as far as stress analysis?

Blessings

 

[fkmeyers]

I have dealt with ANSYS, COSMOS, Noran and others and I can tell you Noran (NENastran)has the best support. They really go out of their way to help you.

Most FEA products have translators to NASTRAN and from COSMOS. I know NENastran does through their FEMAP interface.

As for the sales rep. do not listen to them. Talk to the tech support people as they usually know or will find out exactly what info you need. The sales reps are typically well versed on the marketing propaganda. The techs will tell you how it really is since they are usally not trained sales people (they should not be).

I am not an expert on motion software but as I understand most of them break the model down into rigid or reduced flexible bodies and handle them using equations of motion or kinematics.

 

[GregLocock]

ADAMS basically assumes that bodies are rigid , or can be represented by a complex spring mass model. As such it doesn't really generate stresses, it just generates forces at attachment points.

That's not to say you can't do some stress analysis in ADAMS, but that's like cutting butter with a chainsaw.

The way we do it is:


1) define input to wheel hubs by recording test data (preferred)

or

generate input to wheelhubs from road profile in ADAMS or other equivalent package. (non preferred)

2) Take those forces and apply them to the ADAMS model. Write out time histories of forces at each joint and bush.

3) Reduce the forces generated in 2 to a suitable form for fatigue analysis, and UTS analysis.

4) Take those forces and apply them to the FE model of each component.

5) redesign everything and start again.

Step 1 is very difficult and expensive. Another option is to replace them with known loads (eg the 3-2-1 rule)



Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[GBor]

Timms,

I've been accused of being a bit too high on Algor in the past, but in your case, I think this may be the right tool:

Algor did a simulation of an automotive suspension system about 2-1/2 years ago, maybe 3. They have some elements built in that can allow you to simplify the model quite a bit. NENastran is a good product and I'm sure it could do this type of problem as well, but I would check efficiency on both of them. This is where I think Algor has an edge in your situation. Algor's general contact element includes a "bottoming out" point called "contact" (as if the shock bottomed), a "bottoming out" in the other direction (as if the shock was fully extended), and a modulus in between these points (to account for the shock operation).

In both cases, I think your CAD model will be useless. I'm a SolidWorks fan, but for suspension systems in FEA, you will likely want line elements for much of the structure where in CAD, you are going to want to show the pretty cylinder for the suspension.

I'm not sure I would completely agree with Frank's comment about Noran providing the best support...they do a good job, but I think there are good and bad to both support camps. Kinda' like eating pecans...they all look good, but some just don't taste quite right.

I've seen COSMOS do some impact analysis. I wasn't as impressed.

My 2 cents (again, and again, and...if I didn't give them away, I'd be a millionaire)

 

[badlam]

Further to what GregLocock said about ADAMS assuming that the "bodies are rigid" - you may wish to checkout ADAMS/Autoflex to help reduce the cycling times between items 4) and 5).

http://www.mscsoftware.com/assets/1722_ADAM303ZAFLXZLTDAT_R5.pdf

 

[Timms]

BGor,

I will check out Algor as well, thanks.

GregLocock,

Thankyou very much for the breakdown! I am find that step 1 is very difficult. I have been trying to solve the forces of my system using vibration theory combined with dynamic analysis including kinematic calculations but I keep getting forces that seem unrealistic. Either way too high or way too low. I have modeled it as a sprung and unsprung mass system. I have set it up so that the tire represents a spring and a very, very week dashpot as a damper. With this situation my accelerations seem to be too low. Previously I ignore the tire and just accounted for the shock, damper and mass of the body. In this situation my accelerations were too high. Would you happen to know an aproximate damping coeficient for tires? Should I be trying to model them as pure springs? If I could just get past this stage I'm sure things would come together faster.
I hope to set up a lab and test for loads, accelerations and strains. If I was able to obtain accelerations off the parts wouldn't that be sufficient? From there I could just use the masses to determine the forces? Is it possible to hook up load cells and strain gages or even accellerometers to a motorcycle, drive it down a few different roads and record the data? Ultimatly I want to design the prototypes on the software to achieve as good a design as possible before constructing it and then testing it experimentally to make any additional adjustments. If anyone has any suggestions please feel free to comment.


Blessings

 

[GregLocock]

Take a tyre mounted on its wheel. Bounce it on a concrete floor. That, and some very easy maths will tell you its stiffness and the damping. Basically the damping is 0 in context.

Measuring acceleratons won't do. The reason is that you don't really know the masses that are reacting them (crudely). You have to measure forces. One way to do this on a bike is to measure the true state of tension and bending in the rod of the telescopic, with properly designed strain gauges. We don't do it that way, we mount a 6 dof load cell in the wheel, but the strain gauge system should be simpler and more robust. You need to measure all the inputs simulataneously, ie both telescopics, and all the forces from the rear suspension.

However it is my belief that what you are proposing goes far beyond what others in the field are doing. Are you sure that there are no rules of thumb that will get you there?

What events are you trying to survive? In my experience a fairly simple excel model can be relied on to generate fairly accurate (+/- 30%) forces for a kerb impact, or pothole, for example.


Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[Timms]

Greg

Thanks for the tips. I'm sure there probably are some rules of thumb. It's a pretty competitive world out there! Perhaps I'm looking in the wrong places though. I would gladdly accept any helpful rules of thumb that could be offered. I would greatly appreciate an example of that excel model. I'm hoping to get some sort of theoretical feed back that the system will surive harsh road conditions weather it be washboard dirt roads or speeding down the highway and accidently hitting a pothole. We want some analytical assurance that our product is reliable before we start producing.

When you mention telescopic are you referring to the Shock shaft of the rear suspension or the swing arm?

Blessings

 

[GregLocock]

by the telescopic I meant the front forks.





Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[Timms]

GregLocock

Sorry about that. I realized after I read your response again this morning that it was obvious that you were referring to the front forks. Strain gages seem to be a good approach since I shouldn't have to alter the prototype to install them. Thanks a bunch for your willingness to help Greg.

Blessings

 

[Timms]

Gbor

I have looked into Algor. I must say they do have a really good customer support set up. The only thing that I am hesitant on is that their dashpot damping system only seems to support viscous damping. I'm not sure that it has the flexibility that I need. Perhaps I can do worse case scenarios but I'm hoping to get as close to exact as I can. I think NEnastran does have the capability that I want but I think you may be right about their customer support.

Blessings

 

[GBor]

Timms,

My comment was about customer support in general, not specifically about NENastran...I've had some very good experiences with them, I'm just not ready to say that anyone in the customer support realm is "the best".

It seems as through there are other forms of damping available in Algor. I'm pretty sure they have Raleigh damping. I'd have to be sitting behind the computer with it loaded up to go through anything else, but they may still not be what you are after. You can always call them and ask.

Blessing to you and Happy Thanksgiving (for those in the US).

 

[dmacx]

Timms,

As an NENastran user myself I can comment here since you are leaning that way...

NENastran has Raleigh damping, user defined damping matrices, viscous damping, structural damping, frequency dependent damping, modal damping and just about every type of damping around. The reason is that Nastran is the standard in the automotive and aerospace industries for static and dynamics analysis.

As for support, Noran gives you unlimited support (phone and email) and bunch of hours of mentoring which they never track. I know because this year I have used over 20 hours of it. Mentoring is not a training video or web seminar. It is personal, one on one with their tech support staff which I have found very helpful. I personally have been dealing with a guy that seems to know my industry as well as the product very well. When I have problems with the set up of a model I simply call or email them and they walk me through in that same day. My problems are not simple either. They are complex dynamic analyses of spacecraft components. Other companies I have dealt with previously force you to upload a support request and then get back to you after a few days if you are lucky. I know Algor has a lot of web based training and I really can not comment on the quality or benefits of this method, but having taking the Noran dynamics class I can tell you that it was the best course I ever had and I learned more in one week with them than an entire dynamics class in grad school.

Everyone has their favorite products and wants them to succeed. I personally have had only good experiences with the support team at Noran. They seem to always be there for me.

Have a happy Thanksgiving!

 

[gio1]

Abaqus/Explicit allows you to model tyre behaviour with special fluid elements, and can simulate the dynamics and calculate the stresses within the same analysis, without having to worry about calculating the loads and then applying them in individual loadcases in Nastran or other static FE solver.

In Abaqus/Explicit you could define displacement loads for your tyre model (to represent road input) and let the solver work out displacements/velocity/accelerations and stresses at every node in the model. The main problem I guess would be to collect and use appropriate damping data to use both for material hysteresis (tyre) and joints behaviour.

Abaqus support is generally very good

 

[Timms]

Wow!

Thank you all very much for your input! I think I am really starting to narrow things down. I hope you are all having a wonderfull thanksgiving (depending if you are American or not)! I already had mine

Gbor

I appologize for my asumption. After I read your comments a second time I realized that yes you were only hesitant at the impications of them being "the best" since they all have their strengths and weaknesses conceringing costumer support. I did happen to go through a brief web meeting with Algor. I met with a sales rep and my potetial engineering support mentor and he walked me through a simple demonstration. This is when I discovered the problem with damping. Perhaps he wasn't aware of the Raleigh damping but as far as he was concerned viscous damping was the only option.

Blessings