Speeding up Simulations with Global Adaptive Meshing

Predicting the time needed to perform a non-linear and meshing analysis can be very difficult, particularly if it is very non-linear and not something that you’ve done before. Some years ago, in a different job, I quoted what I thought was a small services job, modelling the insertion of a rubber plug into a test tube used in automated blood testing equipment.

The geometry was trivial and could be modelled axi-symetrically. The client even had the necessary inputs for the Mooney model of the elastomer and I was using what I was told was the best non-linear FEA tool on the market. I figured it was a day’s work to model the insertion and provide some basic results to them. How wrong could I have been?

Over a week later I was still struggling with converging the model. As the plug was inserted it folded around the lip of the tube and slid past itself into the tube.

Analysis stuck before completion

At that location there was a very high strain condition and a very small bend radius which led to some horribly distorted/collapsed elements which stopped the analysis. Endless attempts at meshing the initial geometry to give elements that didn’t distort so badly didn’t work out, so I went through a process of running the solve a little way, converting the distorted mesh to geometry, remeshing, mapping the strain as a pre-state and running a little way further. It was a complete nightmare.

Other convergence problems, relating to high-friction large sliding contact conditions, led us to look at changing solver. We had a demonstration from MSC of the Marc non-linear FEA and after seeing the simplicity of the contact set up and the powerful global adaptive meshing capabilities we were sold and moved forward with Marc as our main FEA tool.

So, what is global adaptive meshing? Well, essentially it automates the manual process I went through with the plug/tube problem. We can define a 2D or 3D contact body to globally re-mesh at specific intervals during the simulation but also on detection of large strain changes, element distortion, contact penetration etc. In this way we’re not sacrificing accuracy through having distorted elements in the critical areas where high strain events are taking place.

It lets us easily simulate scenarios like these:

Rolling a billet

Packer seals

Forming

Machining chip formation

And my plug/stopper example? Adding adaptivity to the mesh, a 2 minute job in the gui, results in a successful solution in less than 30s runtime. I wish I’d changed to Marc earlier!

Analysis completed quickly thanks to adaptive meshing

If you have problems with your trickier non-linear problems, like degrading the fidelity of the solution or compromising on aspects you really want to model but that cause convergence issues, maybe you should have a look at Marc too. It’s MSC’s best kept secret.

Marc is available as part of the MSC One token licensing system which also gives you access to many of the other MSC solutions like Nastran, Adams and Apex. For small capacity users this can cost a lot less than you imagine, or indeed pay for your existing solution.

I’ve been working with tricky non-linear problems for over 25 years, so please get in touch if you’d like to investigate how Marc might improve your simulation quality and throughput and we’ll see if we can help.