FEA TOOL

[bernardg]

Hello all,

My company makes permanant magnet dc motors... and we are in the process of buying a FEA software.
For CAD modeling, we use SolidWorks... so we prefer to buy the FEA program that is compatable with the SolidWorks. We enquired about Cosmos and MSC... and I am confused on which is better. Based on your industrial experiences... kindly recommend a FEA (Cosmos, MSC or any other). Any input and suggestion would be very much appreciated.

Out requirements are...
1) Thermal analysis on the motor - to assist in proper material selection and incorporate forced convection to ensure proper heat dissipation.
2) Our applications require that our motors withstand the vibration of 25G peak.
So we need to do a mock vibration analysis and see the amount of excitation, each and every component experiences at the different frequency levels. This would give us an idea of which components are structurally unstable, and which components need extra support.
3) On the Endbells - to study the stress, strain & deformation.... add/remove material - optimize the design... and still have higher margin of safety.

Thanks in advance for all your help!

bernie

 

[Thane]

Depending on how much you want to spend, ANSYS would be a good software package to consider.

My company is currently upgrading from ANSYS Design Space to ANSYS Professional to upgrade our thermal and large deflection analysis capabilities.

Thane

 

[bernardg]

Well, I don't have experience in working with ANSYS... but I have a feeling that in our applications... we wont need such a versatile and accurate tool.

We have an estimate of $10,000. So now, what will you recommend?

Thanks in advance for all your help.

bernie

 

[TheTick]

I work for a manufacturer of hinges. We use SolidWorks for CAD and MSC NASTRAN for FEA.

We had an independent consultant review our FEA needs. He recommended that we use MSC NASTRAN and not COSMOS for the folowing reasons:
•MSC NASTRAN allows multiple element types in a single solution; COSMOS does not. This allows us to combine beam and plate elements in a single model.
•COSMOS nonlinear does not cope well with large movement and large rotation. Load vector adjustment over large movement is much better, too. Perhaps this would be an issue w.r.t. analyzing motors.
•MSC NASTRAN has much better mesh refinement capabilities.

Due to illness, the part of The Tick will be played by... The Tick.

http://www.EsoxRepublic.com

 

[magicme]

i believe ANSYS is out of the question on the pricing issue.
a one year license will cost $10,000 to $20,000 depending on your setup..... and that is not an outright purchase (which i think is near $30,000).

thanks to TheTick for the info on COSMOS.

daveleo

 

[bernardg]

Thanks Thane, Tick & Daveleo for your helpful inputs.

I have a couple of questions. Based on your experience, you can help me understand this better.

(i) Can we do a mock vibration analysis and see the amount of excitation, each and every component experiences at the different frequency levels using MSC?
(ii) Applying the constraints and loads on the endbells... using the MSC, can we visually see the crack initiation and propogation?

Thanks,
bernie

 

[bjpil]

bernardg

MSC/Nastran is very well suited for your vibration needs. But you need to clarify your goals a little better than you have. Are you going to do any correlation or are you looking at only trend analysis? The reason I ask is that dampening will have an effect on your results. If you plan on correlation then you will need to invest in other tools to do it properly.

MSC can also supply a good fatigue module (from N'Code) that interfaces well with the FEA results from Nastran that will predict crack initiation and propogation. This of course depends on the pre- and post- processor you are using. I am not aware of any program that will allow you to visually see crack initiation or propogation and I am not sure why you would need this.

BJP

 

[bernardg]

Hello,

We need the analysis software only during the product development... so we would be looking at only trend analysis. Pretty basic FEA would suffice.

The reason why I am asking if we can visually see the crack in the system is...
We sent our motors to the vibration test... and the endbells ruptured during the test. Before making those zinc castings, we did the FEA on those using COSMOSXpress (package that comes with the SolidWorks)... and it gave us the stress values at the different stress concentration areas. But we didn't know how to correlate that with the structural integrity of the system.... and what should be the maximum stress that the endbells can withstand before failure.
--> How would i correlate the stress with the failure?

Thanks for your help.

BRG

 

[magicme]

i will put down some thoughts and i'm off on vacation.....

corellating FEA with test data means the model has to have exactly the same loads and boundary conditions that are applied during the test..... are you sure of the applied loads and constraining forces and moments? this is not an easy task.

you also said "vibration test" but later sentences sounded like maybe it was a shock test? .... if it was a shock test, was the computer analysis steady state or transient shock or a forced response analysis (like a shock spectrum or DDAM?) the best corellation would probably happen if you did a forced vibratory response analysis alongside a test with (of course) known driving forces. corellating a shock test to analysis is an awful mess and there are a number of (not-simple) approaches.

i recognize that i did not answer your questions, i am hoping to possibly get you to rethink exactly what you did and exactly what you expect to find.....sometimes we engineers don't always do that very well.

daveleo

 

[bjpil]

This is one of those issues that may have many different paths you could take. I think that you might want to take a step back before you purchase a vibration FEA package. If you could answer some of these questions maybe some light could be shed on your issues and a more cost-effective solution may be found to help you.

Depending on the amount of testing you have already done and might still be able to do here are some ideas you might want to investigate. I am sure others will have more ideas as well and you may already have done some of these.

Did your component break at the same spot that your max principle stresses occurred for the different load cases? If not then your FE model needs to be updated before you move on to other FE work.
How comfortable are you with your load history (magnitude and cycle count)?
You could do some correlation using a component fatigue test vs. the specimen fatigue testing by using cycle count/magnitude along with Minor’s Rule and FEA to find a better solution. Strain gage testing would also aid in correlating your FE model.

As daveleo points out, this may not be a simple fatigue case and as you suggest, an investigation to see if a resonance is occurring in its operating range may be needed. Did you notice if your motors hit a resonant frequency while testing? Depending on your resources, timing, experience and size of your motors here are few test you may want to try 1) a sine sweep test – sweep the entire assembly through a frequency range to find your resonant frequencies or 2) a impact modal test – which will also aid in finding resonant frequencies.

The avenue that you are going down is also valid but may take more time right now than you have due to the learning curve. If you choose to get vibration software could I suggest first finding a consultant that does this sort of work and has a proven record. I suggest the consultant because this is what I would have preferred if I had known better. They could help reduce the learning curve and give you a completed model and methodology for the next time that you have a similar problem.

BJP

 

[bernardg]

In a nutshell...
We manufacture permanant magnet dc motors... and our motor need to pass the 10G random vibration test in the 3 perpendicular planes, w/ the shaker table excited at 24 - 1000 Hz. (The motors are not powered up. They are mounted on the fixture that is bolted down to the shaker table. Then the shaker table is excited at the 10G RMS random).

The main systems in the motor assembly are the shell, armature and the endbells. The armature consists of the laminations and windings... and has bearings on both sides. The end bells have the bearing pocket...on to which the armature bearing are rested into.

To keep the armature from bouncing back and forth during the vibration test... we use the preload spring (to take the tolerance stackups) and then loctite the bearing on one of the endbells. Because of the minimal clearance between the bearing and the bearing bore (for the noise quality)... during the course of the vibration test, we have instances where the bearing breaks lose from the loctite and start banging against the endbells. At that point, the preload springs gets crushed into pieces... and then eventually the endbells get ruptured.

ANSWERS TO BJPIL'S QUESTIONS:
(i) YES, the end bells break at the same spot where the max principle stresses occurred for the different load cases.
(ii) We contacted the Tech support at B&K... and after the analysis, they ruled out the option of any resonant condition occuring during the vibration test.

As i mentioned in my previous reply... all we did with the CosmosXpress was to apply the load on the bearing pocket to duplicate a condition where the pocket would be taking the pounding from the armature when the loctite breaks lose... in an effort to simulate the principal stress at the stress concentrated areas.

Say, for a particular armature... the pounding that the bearing pocket takes is 100 lbs-force. Due to the oscillations...the number of cycles = 10,000. With that data, I can find the principal stress at the end of 10000 cycles at any given point. Without doing any experiment... is there a way on the FEA tool to find out the max stress that the endbell can take, before starting to crack at its stress concentrated areas.

Your input & comments would be very much appreciated.

BRG

 

[bjpil]

If you are doing a linear analysis then all you can do is predict the load at which the part will start yielding. If you are doing a non-linear material analysis then you can predict the single load that the part will rupture. If you want to predict cyclic life then you need to do a fatigue analysis. What I suggest you do is run a certain number of samples a 10g until they crack. Then run a certain number of samples at a lesser load until they crack. Create a stress life curve for the component. Figure out how many cycles you want your part to survive and then revise your FEA model to incorporate this stress reduction. Calculate a Weibull number for the change. This all assumes that your part does not see any plastic strain in your testing.

BJP

 

[bernardg]

Hello all,

Thanks for your replies. One last question...

We have decided to go with the MSC software. They have sent me a quote for "MSC.visualNastran 4D for SolidWorks"

Have any of you have prior experience working with this tool? For the requirements that I mentioned previously in this thread... will this one suffice, or do you have a different recommendation?

Kindly send me a prompt reply.

Thanks in advance for your help!

Bernie.

 

[corus]

As you've already worked out the principal stress, and presumably the stress range, then I'm not sure why you need another FEA package. FEA won't tell you when cracking will occur. The normal practice is to assess the stress range against an S-N curve for that particular material, if you have it. There are some codes available that will assess the stress from an FEA package by using the S-N curve but if it's a simple problem there's usually no need.

corus