Does anyone have opinions on upgrading to Solidworks Premium vs. purchasing the NEIWorks Add-in for plain old static loading FEA applications? Xpress isn't cutting it and I'm on the fence as to which upgrade to make.
Thanks for your input!
SIMULATION has :- better integration into SolidWorks
- minor bugs- mesher more "robust"
- Simulation constrains / contacts more "stable" and in the 99% more efficient
- better POSTPROCESSING
NEiWorks has :- solutor more advanced into composite material- compatibility with MSC - NX Nastran reports (need in aeronautic certification)- NEiWorks cost a little less
Watch this video from NEI: http://www.nenastran.com/videos/NEiFusion_SpringModel.html
We have attempted to reproduce this simulation in SolidWorks Simulation with the same model of spring. Below are the comparative statistics after running the same simulation in SolidWorks Simulation: Nei Fusion SolidWorks DOF 348 391 363 279Mesh 30 sec. 38 sec.Calculation 3 min. 17 sec.We calculated a displacement of 0.4868 inches, which if you calculate the spring rate as he does in the video comparing that to theoretical one, we get a much closer result than Nastran. When calculating displacement, this is the unknown parameter so it should be the most accurate because it is the basis for strain and stress which are calculated from this. Spring Rate (note: the original video actually uses this many digits, obviously forgetting about significant digits) Theory: 2.060517191 lbf/inNEiFusion: 1.996000016 lbf/in, 3.1% error from theorySW Simulation: 2.054231717 lbf/in, 0.3% error from theory
can you post your SW test spring file?
I'm trying to do what you have done with SW, but SW stop before to complete the analysis.
I see that the diameter of your spring is different from that of the video, as well as the elastic modulus. The test is not reliable if you do not use the same values. If I set Large Displacements, where I find the option to set the analysis with Small Displacements ?
Everything is the same. The video author creates a 0.5 in circular sketch for defining the helix, height 1" and 8 revolutions starting at 180°, and the spring wire diameter is defined as 0.035 in, the same as for the model I attached. The pertinent material properties are the same: 29e6 for modulus of elasticity and 0.32 for Poisson's ratio, same as in the study attached. At the end of the video, the author makes a chnage to the wire diameter making it 0.05", which in my attached model there is a "Thick" configuration but no Simulation study has been created for this config; one would simply have to duplicate the original study, choose to copy it to the new config and then run.
Small displacements is the assumption for a linear static analysis, so there is no required setting. The same was done essentially in the video for NEiFusion. SW Simulation warns you at the end of the calculation that it has detected large deformation, but by selecting "No" in that dialog box it will give you results based on the small displacement assumption. NEiFusion does not provide any warning, which is somewhat troubling especially to a novice user who might takes these results as is. Though I am not faulting the author of the video, he is simply following the same assumption that the theoretical solution takes.
I have selected a number of FEA packages for two different companies. i came out of the aerospace industry and into a startup in 2000 so when choosing an FEA package, i naturally looked at NASTRAN, SDRC I-DEAS and ANSYS. we ran many test cases and not always, but often COSMOS would yield different results than NASTRAN and ANSYS. I had also been a COSMOS user since 1993 but ultimately we ended up with both ANSYS and NASTRAN.
Then for my own company, I opted for ANSYS and have been very pleased as the Workbench interface is really great and my parts do not end up being saddled w/FEA data. My preference has always been to decouple the FEA and CAD software yet have seamless model transfer.
Before you buy, take a look at ANSYS Workbench as well. Make sure that current needs and future needs can be handled by the package you buy or adding modules. i can tell you this, its a big nut to spend and once you do it is damn hard to go back to another software as you will have lots of time and and money invested. Don't forget to budget fot the 20% yearly maintenance. For me, i pay more in FEA maintenance each year than the cost of SW professional.
as a comparison, i ran the model in Workbench ANSYS 12.0 with both a course and a fine mesh
for run 1, ansys defaulted to swept bricks which is my preference over tets (not sure why NEI would use tets when you can sweep it with high order bricks. must be a default element?)
brick elements: 9180
large deflection "on"
mesh time 10 sec
total solve time:60 sec
deflection: .48446 in (large deflection off=.48201)
k= 1lbf/.48446in = 2.0644 within .1% of theory.
tet elements: 79917
total solve time:223 sec
deflection: .48386 in
k= 1lbf/.48386 in = 2.0678 within .3% of theory
in my experience, it is best to use bricks when possible.
the results from NEI are horribly far from theory and from ANSYS and SW. i am not sure that the theory is exactly correct for large deflections such as this spring. regardless, one can always solve a linear static problem using a nonlinear solver and get the same result, it will just take longer to solve.
the large discrepancy for the NEI solution bugged me so i re-ran the model with gravity on to see if i could get the same number as NEI. got .48529 which is still way off. very wierd.
I watched the vido better and I saw that he types 0.035 although the dimension show 0.04. I looked at the material and actually is the same but with different exponential. I apologize for my superficiality observing the video. I'm glad however that you have demonstrated the Simulation superiority and all these data will also be listed on the SolidWorks Italian forum.
Wanted to thank everyone for their comments and suggestions. Very helpful. I'm playing with a trial version of Solidworks Premium this week and also downloaded 2010 to check out the enhancements to Simulation Xpress. I've only scratched the surface with both.
I am more than open to anyone's thoughts on alternate packages. I stumbled upon NEI Works while searching for reviews on the various Solidworks FEA options. I am aware of several other packages, Mechanica and ANSYS but don't have experience with either. I priced out the upgrade to Solidwoks Premium vs. NEI Works vs. adding Mechanica to my Pro/e seat. It was easy to rule out Mechanica since I haven't had a client request Pro/e in almost 2 years and it wasn't any cheaper than the other 2 options.
What I'm really trying to get a grasp on is the comparative "value" in the various options. I haven't had the need for anything beyond Solidworks Xpress until recently and I don't see myself becoming an FEA "guru" but it would be nice to have a package with a few more bells and whistles. If Xpress is a Moped and ANSYS is a Porsche, I'm looking for a Camry and might settle for Kia Rio. At present, I need to be able to analyze static loads in simple assemblies. Of course I'd like it to be accurate but I can live with a few percent here or there...
If anyone has thoughts on low to mid-range FEA packages, other than NEI or Solidworks Premium, I'd love to hear your thoughts. I'm interested in tracking the discussion on higher-end packages but it frankly doesn't make sense for me to spend that much when I'll likely never need that much power.
Thanks again for all your thoughts.
I have to say I have very little experience with Solidworks Simulation. So far, I've been happy with the very low number of meshing errors I get. The solvers certainly seem well optimized, and but for a few stupid cases, make decent use of multiple CPU's.
However it seems with Simulation the user interface is but a hack. The linear static simulation (which comes with Solidworks Premium) seems pretty workable. It seems that as the number of users decreases, so does the quality, and it is evident to me that development has little or not testing---only the developers are running the same test case after test case and don't realize what parts of the UI are disfunctional or broken. Simulation Premium is full of stupid problems which mostly seem trivial and only aggrevate the user, making the setup time the most time consuming part. Ideally, you'd set up your studies quickly and easily so you can go do something else while it crunches the numbers.
A few things off the top of my head (you can look at my other posts for recent problems and complaints):
1. You can define parts as remote masses. That means they impose loads on what you're analyzing, but don't get analyzed themselves. That's great. Except that it only works on parts, because subassemblies are "dissolved" as far as Simulation is concerned. Have a stack of parts hanging off a beam and you just care about the beam's deflection? Well then you either place the mass in manually, or create a part based on your assembly that has similar mass properties, because it will take forever to go part by part and have them rreated as remote masses.
2. When analyzing assemblies, I used to just create a new part that was a single solid body of the whole assembly. That's one way to "get around it" for Xpress. Now with the full-blown version I figured I'd try working on whole assemblies, since then the changes I make to parts are already in the models. Well, the UI efficiency does not scale well with number of parts, and after about 40 or so it became unmanageable---even if you omitted most of them from the simulation; everything would simply take too long to respond. It forces you to suppress the parts in the model and save "mouse lag" time. That's fine if you don't actually need them, but it sucks if you do. And why is "suppressing" a part in the simulation not identical to suppressing the model as far as the simulation is concerned? Who knows.
3. I just ran a linear dynamic sinusoidal simulation (vibration test). I defined an excitation spectrum by listing the response magnitude at different frequencies. Let's not even get into how ass-backward the curve editor is, or the fact that once you save your curve in your database, you can't save it again with the same name (won't even let you overwrite). 12 minutes into the solve (which is almost 30 minutes including meshing), I get an error that says my excitation curve doesn't cover enough frequencies. It turns out the study be default sweeps up to 10 kHz or something and my curve only went up to 700 Hz. There's a missing IF statement if I've ever seen one.
Sure, you get a subscription, you can create a dozen tickets an hour, like I am doing. Then you are working for Solidworks for free, and you better pay for another 3 years of subscriptions in the hope that they actually implement changes to what caused you problems.
I'm going to sign up for a class and hope that all this is because I'm a stubborn fool who wants results yesterday. If not, I will immediately start evaluating alternatives. Evaluation in itself is a daunting task if you don't have a prior case that caused you some trouble to really push things. I wish you luck (if it isn't too late already).
If you sty linear static then Simulation should work great for you. You can dabble into nonlinearities (materials, large deformation, contact) however you will quickly find that there are limitations. My strong preference is for Abaqus for difficult nonlinear problems. There is a much higher cost however I believe that I more than make up for the cost by increasing efficiency. Below are some links to threads that might be useful. I hope this helps.
as you know, this problem is much more complicated than a simple beam or even a tapered beam as the individual leafs can slide wrt to each other.
i ran this model with large deflection and no seperation contact at the leaf interfaces.
in addition to the necessary constraints, i also added a symmetry bc to the ends of the springs
for simplicity, i neglected the spring clamp and used the entire spring length even thought his is not how the spring is actually used.
for a -y unit load placed at the "eye" stress =475 psi and deflection = .0075"
solve time was 82 sec.
can you verify w/SW Sim?
yup, it sure does matter.
Retrieving data ...