SimScale CAE Forum

FE Simulation - Status: Error


#1

Link to Project/Failing run

The linked run fails with “Error”, and no further information?

Inspecting the log – only 1500 lines, but not the start, are available – and I don’t see anything I recognise as an error message.

Can someone take a look and suggest where I should start looking?

Thanks, Buk.


Restore simulation settings to those from an earlier successful run?
#2

Hi @mhere, firstly I’m not an FEA guy, but I would certainly look first at the geometry. The long thin faces are causing a very odd mesh. Hopefully, someone more FEA orientated could comment further.

Best,
Darren


#3

I agree with Darren that the issue might be the geometry as it looks very odd and as if there are plenty of overlapping surfaces.

If you could fix the geometry and just re-upload it into your existing project, that would be a good first step into the right direction :+1:

Cheers!

Jousef


#4

[quote=“jousefm, post:3, topic:82456”]
the issue might be the geometry as it looks very odd
[/q

Might? Is that based on anything you saw in the log? Or just:

Because according to freeCAD both solids are “No folds on surface”; “No self-intersections”; “No invalid indices”; “Nodegenerations”; “No non-manifolds”; “No duplicated points”; “No duplicated faces”; “No flipped normals”.

(I started with a .stl file and it took me most of 5 days to recover the two solids :frowning: )

I’d rather know (for sure) that the problem is due to the geometry; and have a good idea of what I need to fix, before I start the whole process over?


#5

Hi @mhere!

I did not extract that from the solver log this was more intuitively because your geometry is split into hundreds of different faces which looks odd as Darren mentioned. What I tried is to upload it to Onshape and re-upload it as an IGES format which did not work so far. Could you try exporting it as another format from freeCAD and see if you can achieve a change?

Best,

Jousef


#6

Hi @jousefm and @mhere
i’ve also take the geometry to Onshape a clean it up. but the error is still happening with the same simulation set up so i don’t think your geometry is the problem. i think the error is due to your boundary condition. in main time i haven’t grasp the whole concept of the simulation (what you trying to do) in your perspective. so i tweak some parameters in the way i have understand the problem on simulation designer and it work fine for me you can get it here.

just to give you a hands up my simulation run out of maximum run time, increase that and rerun it. and not to confuse the whole thing i have use displacement boundary condition since i don’t know the value of your rotational force.

Regards
Anwar


#7

Hi @Anware!

Looks good! How did you clean it up? In the past uploading and exporting it as another model worked (at least for some models). Your setup is slightly different from the one @mhere shared with us so it would be nice to know what the purpose of the simulation is.

Cheers,

Jousef


#8

Thanks Anwar. I’ll work my way through your version and try to work out what you’ve changed.

As for the purpose of the simulation. It is to try to validate the simulation process by reproducing this effect seen when the real component is subjected to 200N.m of torque:

In real life, the axle is connected to an electric motor, and the purpose of the torque arm is to prevent that axle from rotating.

In the picture you can see that

  • the left arm of the TA has shifted (up) relative to the right arm.

  • the first spline in the TA has sheared.

  • the TA loop has spread allowing the axle the freedom to rotate effectively unconstrained.

In the model, the axle needs to be free to rotate about the Z-axis, and be constrained in al other freedoms.

In theory, the ends of both arms should be ridgedly fixed to the frame; in reality, the material & section of the arm is insufficient to prevent it from distorting as seen. (Despite that for this test an additional component was bolted to the back of the two arms; with almost no effect (Predicable!))

If I can recreate this failure to some level of accuracy, then I can use the same methodology to test different materials and sections and get a feel for how they will fail before committing to have a part manufactured and destruction tested.

This is about the simplest version; many of the other designs are rather more complicated; but I am finding that simulating even this simple one motion model is near impossible. :frowning:


#9

The problem with replacing the remote force with a remote displacement, is that you are defining a rotation, by a mathematical function, when the purpose is to prevent it; and to see the effects when that attempt fails.

That is; when you test a beam under load, you don’t predefine how far it will bend; you apply the load and want the simulation to tell you how far.

(Note: I’m well aware that my applying a force to one end of the axle and fixing the other end 15mm away, means I’m testing the axle’s resistance to torque instead of the TA :slight_smile: but without a TORQUE BC, I don’t see how else to do this.)

The axle component needs to be allowed (but not prescribed) to rotate in the z-axis and constrained in all other freedoms.

The z=axis torque applied to the axle needs to be transmitted via physical contact between the splines, to the TA, and what I’m trying to study is the effect that force has upon the TA.

(Also, why did you add a “fictitious clearance” to the physical contact?)


#10

Hi @mhere,

Great to see you around the forum with another very interesting project!

Simulating this case is not simple at all, as it includes complicated facts such as:

  • Great displacements and rotations
  • Non linear material behavior, that is, plasticity and fracture
  • Non rigid boundary conditions
  • Physical contacts

One can simplify the problem by ruling out some of this aspects, but then you won’t get the same picture.

I think the best application of simulation here would be to predict the resistance of the component, that is, to establish a failure criteria and predict when it will occur, even if the picture won’t show the real-life deformation behavior.

About the CAD model, it is actually a bad model. The problem is that the surface is triangulated and adjacent triangles vary too much in size, which results in a mesh with abrupt changes in element size. This leads to numerical and modelling errors. I think this is due to the exported file format. If you could please provide the geometry in IGES format I think we could have a better starting point to start building the simulation.


#11

Actually you can think of it both ways, and both approaches are valid. The only difference is you determine one variable and measure the other one.

When you test a beam, with an hydraulic press for example, what you are determining is the deformation through fluid pumping (you target a maximum displacement and only stop at component failure), and reaction force is measured indirectly through fluid pressure or a load cell. Same applies for a bolt load and torque measurement, and any test rig for that matter.

The case with simulation is that you can do both ways, and both will yield valid results. It is up to you to find the best approach for each case. And in this case at hand, prescribing the rotation is actually the way to go, because of the same reason you gave:


#12

Sounds like your saying that my simple re-world example is too complicated for your system to simulate?

FWIW: https://www.simscale.com/projects/mhere/splinedtorquearmandaxle-via_freecad_-_iges-

But it has 506 open shells and look like rubbish. (I started with trying IGES format when I first came here, but was advised to switch to STEP.)


#13

Yes, for this, and any system for that matter. Turns out reality is far more complex to replicate than it seem.

What CAD software are you using? Maybe a drawing to recreate it in a better one (such as OnShape)?


#14

The problem with prescribing the rotation is that I have no way of knowing a priori how far it will rotate; it is entirely dependent upon the torque input, the material specified and its (various) cross-section(s).

To be of any use at all, the simulation would need to allow me to vary each of those three parameters, whilst holding the other two steady.

Eg.

  • Aluminium; 5mm thick; will withstand XXX N.m before transitioning from elastic to plastic deformation and/or shear failure.

  • Mild steel case hardened to 20 microns; needs to be X mm thick to withstand 250N.m of torque without strain hardening and spalling.

  • If space is limited such that we can only accommodate 3 mm thickness in the TA; what tensile strength is required to resist 250N.m of torque.

But that is only scratching the surface [sic] of the scenarios I would like to test.

Eg. Assume I determined that a particular profile of Al 6061-T6 could withstand 200 N.m of torque with a 2:1 safety factor.

Al has the unfortunate property of being subject to early cyclic fatigue failure.

The motors used on (many) e-bikes also do regenerative braking, so the TAs are subjected to repetitive clock-wise and counter-clockwise torque loading that - due to engineering tolerances – subject the TA faces to repeated impact loadings that exceed the nominal torque ratings of the motors by a calculable factor.

Given the manufacturing tolerances for both parts, I can calculate how much unrestrained rotation will occur as the torque load transitions from positive to negative and vice versa. From that, the amount of angular acceleration can be determined and when combined with the area of the axle edge, the impact forces the TA faces must withstand can be calculated.

How many cycles will any given temper of Al withstand before the cycle fatigue induces sufficient enbrittlement that shear failure occurs?

One of the earlier simulations I tried here used a multi-layered spring steel TA. The deign is such that as the axle nut is torqued, the ‘wings’ of the multi-layers close tight around the axle flats to exclude manufacturing tolerances and preclude torque direction transitions from having room to accelerate.

By using spring steel, the idea is to use the materials inherent flex to absorb (convert to heat through friction), and feedback the torque transition loads. Ie. Use material science and Hooke’s Law to resist torque loads, rather than an immoveable object.

In the past, I have led the investigation of a part failure of (eg) a towing bracket welded to the chassis of a heavy truck – that through simulation (Dassault Systèmes SIMULIA + CAD et al) – determined that the failures were due to (welding) heat induced stresses well away from the weld seam.

The simulation of that followed the fracture process from the initial cracks right through to the side of the bracket opening out – several centimetres – to the point that the shackle & eye bolt could detach.

I know such simulations are possible – though it was not my job to construct or run the simulations.

Maybe my experience of (the results from) that expensive commercial system have given me too great expectations of what it is possible (for me?) to hope to achieve with a community account on a system based on OSS software – which can be immensely accomplished and powerful, but is notoriously difficult to setup and use correctly.

I thought that the splined axle was simple enough that I would be able to gain experience of setting up and running sims here; but if that is too complicated,I stand no chance of simulating the more promising, but complex solutions to the problem I set out to solve.


#15

Hi @mhere,

Thanks for the detail explanation of your idea. I think that we can build a simulation to fulfill your engineering needs, and then you can learn the techniques and apply them to different design alternatives.

If you wish, let us start by having a correct geometrical model. Can you please provide a drawing or CAD model?

Then we can move to the matter of materials modelling. Plastic deformations are not hard to do in SimScale, but fracture propagation is not currently implemented, although the underlying solver is very capable.

I propose you to turn your problem into a study case and generate a tutorial, where we can study and discuss step by step the modelling assumptions and techniques. What do you think about that?


#16

Here are the dimensions:

I can, if we can find a suitable interchange format? (ANSYS SpaceClaim .scdoc format or anything that freeCAD can export?)

That said, the ones I have are themselves derived from the original that was sent to me in .stl mesh format, so the geometry has been reduced to shells. That’s why the first version I uploaded here was so heavily faceted,

For my second attempt, where I managed some partially successful runs, I loaded the original .stl mesh into FreeCAD and used its mesh routines to analyse, repair and simplify that original mesh before converting it back solids.

(My results look very similar to those@rszoeke produced )

I’m still evaluating the various (free) CAD software available to me trying to find one that I like. I’ve discounted half a dozen or more – tinkercad, blender, sketchup, sculptris, solidedge, onshape – as they seem to be more aimed people who want to finger-paint on mobile/tablet touchscreens rather than technical drawing. (I’m definitely more draughtsman than artist.)

For the most part, I’m finding them clumsy and/or laborious; and in some cases, combined with really crass UI.

Sounds good. I’m certainly up for that if it fits with your needs. Tell me what I need to do?


#17

Hi @mhere,

I’m glad you like the idea!

Here is a link to the model I created in OnShape based on your drawing.

Also this is the project I created to start from scratch. I think it is better to continue the discussion in that project thread.