I would like to develop a simulation of a tube being squashed flat between two steel rollers. My reading so far suggests that this problem is very nonlinear, and may involve dynamic explicit analysis which is difficult for CalculiX.
The questions I am hoping to answer with this analysis are:
What is the longitudinal force applied to the cylinders during the operation?
How much torque must be applied to the cylinders to pull through and squash the tube?
My experience with FEA includes a single introductory course at university, and mucking around with openFOAM, so I realise I may be jumping into the deep end with this problem.
Indeed, explicit dynamics would be best for this kind of problem and it’s not really robust enough in CalculiX. So you should either use OpenRadioss or implicit method in CalculiX - such problems are often successfully solved with a nonlinear static step.
Thank you for your reply. The example you posted looks like a great starting point.
What makes explicit dynamics a better choice in this instance? Is it simply because computing dynamics implicitly is more computationally expensive than explicit dynamics, or are there also implications for accuracy?
Putting aside the limitations of this procedure in CalculiX, explicit dynamics has two main applications - high-speed dynamics and highly nonlinear quasi-statics (e.g. metal forming). If a problem is static but severely nonlinear (tricky contact conditions, large plastic deformation, and instabilities), implicit solvers will usually struggle to converge. Explicit solvers don’t have this issue, and if you ensure low enough inertia effects, you can get a quasi-static solution with an explicit dynamics solver. However, the main issue is that there’s a conditional stability, and very low increment sizes have to be used to fulfill this requirement. You can use mass scaling to speed up the analyses significantly, though. Such topics are discussed in the context of CalculiX here: https://www.youtube.com/watch?v=-X0Shj7UXE0
Nice challenge, you should make a two seps analysis, use symmetry (a quarter model would be enough if the two rollers are flat), and be very conservative with the mesh, maybe adding more elements in the initial crushing zone and then increase the size along the tube axial direction.