Accuracy of Results

bert_fourie

I did an analysis of a roughly 12m long assembly. The von Mises stress results came in a lot higher in certain localised areas compared to my traditional FEA solution Simsolid, whereas the overall deflection was virtually the same. To be sure I then ran the same analysis in Simscale whose results corresponded closely to Simsolid. I tried changing connection types and added in some corner chamfers but that did not change the results much. The overall distorted shape looks correct and the deflection corresponds closely to the other FEA's. I wonder if this is a function of the size of the structure and the algorithms used that the local stresses may not be calculated correctly on such a big structure. I know there was an earlier comparison run on a smaller object reported in the forum. Would you please comment? Happy to share the file. 

bert_fourie

I found the answer, it is in the way the constraint works which stiffens the areas in question, hence they show more stress. In particular, I have a "fixed" plate which does not allow any deflection to happen. Somebody else had the same issue with a table leg where the "fixed" external constraint influenced the result. By changing the geometry of the "fixed" plate to more accurately simulate the local deflection I was able to get a closer answer. A learning curve here.

sebastian_glanzner

It seems that the default behavior of the simulation is that all surfaces in contact are bonded together. 

I would like to have more control over these contacts.

Perhaps we could set them all to a sliding behavior, which could result in more deflection of the assembly.

bert_fourie

Yes I agree. I tried different contact mates for clevis pin connections using revolute mates, but that resulted in stiff joints. I then tried cylindrical mates and for one joint got it to work reasonably well, but the other joints would for some reason not obey the movement limits I set.

It would be a huge advantage to be able to use customized settings on the different mates to simulate connections with friction and limits at play.

chris_gromek_onshape

Greetings Bert!

Could you post a screenshot to help illustrate the issue?  As of now, the default behavior should be that any pair of contacting parts is fused (fastened) where they touch - unless otherwise mated.  If mated, the corresponding DOF will be honored over the interaction region.

I agree that mates should be enriched for simulation with friction and displacement/rotation limits, but that will not happen until we add more sophisticated solvers.  Until then, mate limits do not inform the simulation boundary conditions, so the linear static results may not be accurate unless the Assembly is fully constrained to begin.

Though the solver is robust to Assemblies in static equilibrium, but not fully constrained, a slight imbalance during the solve can drive toward unstable results.  Enhancing simulation with load stepping or rigid body dynamics could help mitigate this, but we are not ready to implement those technologies at this time.

Best,
Chris

bert_fourie

The overall deflection is calculated correctly, and the stresses are accurate, except at the lugs themselves where the joints are modelled as stiff.

bert_fourie

bert_fourie

This is an extension piece for a vertical drilling rig, with the structure and the drilling machine hanging from a crane and a torque is applied around the drill string. I modelled the drill string as a fixed with the structure rotating around it, which accurately reflects reality as hangs from the crane and is fixed in location by the drill string. The round piece on top of the frame is on the cfg and simulates the mass of the drilling apparatus, for gravitational force application. The rear plate at the far end is fixed. I made the rear lugs longer to simulate more flex in them to try and improve the accuracy of the local stresses.

bert_fourie

This gives you the idea of what we are looking at, a similar setup.

bert_fourie

In this case, I modeled the rear clevis pins as cylindrical mates and set limits as to extent of movement. I was not able to set a relationship between the side faces.

chris_gromek_onshape

WOW!

Without the pictures, it is difficult to appreciate the scale of this thing.  How did you model all of that plate steel?  Are they separate parts?  One solid part?  Or did you manage to do it in a sheet metal model?

-Chris

bert_fourie

In this case, I was really only interested in the extension tube part, the drill base itself I modelled to approximate the structural stiffness of the drill frame. Normally I model each plate as a separate piece so that I can generate dxf files for profile cutting. For FEA it depends on whether or not I want to study the welds, if not I solidify pieces that are welded together. Otherwise, you have to deal with a massive number of connections/mates to specify. It also simplifies meshing and reduces runtime. I do try to run assemblies rather than individual parts as it speeds up my workflow and I can study the interaction between parts.

The issue at stake here was that the client wanted to position the drill unit way further out from the crane than they would usually do and they needed sign-off on that. By the way, they did the first hole with the system today.

The torque itself was not the issue although it resulted in nice graphics - obviously the further out you are the less the counter force. However, if the full torque was applied and the drill string not yet fully captured in the hole, then the torque produces sideways movement of the structure and the risk is that the rear lugs could break off. I had them implement mitigation by reducing torque at the initial stage of drilling.

bert_fourie

chris_gromek_onshape

Bert,

Thank you for the clear explanation - I have a much better understanding of the system.  Congratulations on the first use of the machinery and your mitigation strategy, very well done!

From the analysis side of things, I can say that our adaptive scheme will refine on hot-spots regardless of how you constructed or mated, but the model will be overall more refined as large parts are subdivided into smaller assemblies.  In this case, I completely understand the hassle of having to mate each and every section.  One thing you could try is splitting the geometry into parts in the Part Studio and then Grouping them in the Assembly.  As a Group, we will still treat each of the parts individually, and you won't have the trouble of assigning Mates to each pair.

Having said that, you and I did discuss a future capability whereby you identify certain faces as important and we add additional fidelity/precision to that segment of the mesh.  For anyone else reading this thread, I'd be interested to know if that sort of feature would be helpful for you too, and what you think would be the least obtrusive way to identify those regions of interest.

Best,
Chris