Convergence (FEA Part 5)

Bite-sized Virtual course


Duration- 3 hours

Prerequisite – Meshing

Is my model converged? Can I trust the numbers?

It is assumed that the model has been correctly built, that there are no poor constraints, loads and overall approach is good. We are just left with the problem of trying to determine if the stress numbers are good.

There is no single output that states the answer is good. Besides, the software doesn’t know what your objective is. The ‘goodness’ of an answer requires the user to review a variety of information ‘in the round’ in the context of the objective.

A simple ‘plate with a hole’ model and a simple ‘L’ shaped bracket are used to show that converged answers can be achieved from a model with poor constraints and that the actual answer is not an exact number.

The Single Pass and Multipass Adaptive methods are compared.


• What are we trying to find out?
• St Venant’s principle, what can we get away with?
• Understanding the log file content
• What is the RMS stress error estimate?
• Multipass Adaptive, good and bad convergence
• Single Pass Adaptive, good and bad convergence
• Disadvantages of system hardwired measures
• User created measures in convergence
• Converging on user defined measures.
• Showing the effect of refining a mesh
• Stress errors
• Model animation
• Qualitative mesh and fringe critique
• Using iso-surfaces and formatting fringes
• Graphing stresses along a curve
• Developing a list of checks