Why is my adaptive solution w/ Simulation not converging?

[Jieve]

Hello,

I’m modeling the contact of a bolted joint using “pucks” to model the bolt head and nut and a thick washer between the head of the bolt and part. A preload force or 27000N is applied to the outer faces of each of the pucks. Originally, the lower plate (10mm thick) had a counter-bore between the clamped part and the plate (end opposite to the bolt head) for a dowel sleeve and I wanted to see if the counter-bored section would deform too much under bolt preload. I have now changed the hole back to a simple clearance hole for a 10mm bolt and have run an h-adaptive study to 5 iterations and 99% accuracy, but cannot get the model to converge (max stress keeps at least doubling, last max was almost 30,000 MPa). I’m using a very very fine mesh control around the contact edges, have used a small radius of 0.25mm at all contact edges, and yet at the contact I get patches/specs of red that look something like the spots on a leopard, with areas of green interspersed between. I’m using node-surface contact (only because the 5x adaptive solution took over 2 days). These huge discrepancies between nearby elements didn’t make much sense to me, and the stress at some of the red patches just keeps increasing to infinity. I’m using a purely elastic model for the materials, as based on hand-calcs I didn’t expect yielding. Is this an artifact of node-surface contact or something else? Should I have used an elastic/plastic model?

I have read a number of times on the eng-tips forum that solidworks simulation doesn’t handle contact well, is this true, and if so, why not?

I am a JPEG of the Von Mises distribution in the puck/washer/base plate/part.

Thanks

 

[EngAddict]

You are likely getting a singularity due the applied load, and as you noticed it will continue to infinity hence it wont converge. This is one of the precautions the operator must be aware of when using adaptive meshing. Personally I don't have the need for it.

The stress you are seeing is a peak stress, it wont have any effect the performance of your assembly. These are only really taken into consideration when assessing fatigue loading. Notice how the stress slightly away from the singularity is quite reasonable? This is where you want to probe your actual stress results. For things like fillet welds there are guides as to how far from the stress intensifiers you should go to take a measurement but for things like this it is mainly up the the analysts experience.

Basically IMO what you modeled is probably overkill. You could use a bolted connector or a split face against a virtual wall if you want to allow rotation at the bolt interface. But in most cases you could probably simplify it with a boundary condition directly on the face or edge. This simplification could be justified when you will notice the peak stress dissipates quickly and rely on hand calcs for the bolt loading and bearing pressures. Just take care not to over stiffen the model. Not worth running simulations for 5 days when you only want the bending in the weldment and the rest can be done by hand.

Just my 2c - hope it helps.