Consider a tee bracket subjected to a horizontal force. This is just a simple example, but as the mesh refinement increases the corner stress is shown to be diverging, not converging. How do you know what the actual stress will be?
The corner stress can be calculated easily by hand, but this example shows one of the challenges of using FEA accurately.
The hand calculated stress at the joint is 60MPa. So initially the stress is underestimated, then apparently over estimated, with the expected stress lying somewhere between.
This is a classic FEA problem; the actual corner stress cannot be determined from this FE model. The stresses will not converge because of the re-entrant corner (crack).
If this part was made of glass (or another brittle material) it would break. But because most materials yield, this part would live fine for ductile materials and static loads. However, for a fatigue load case, even metals would fail due to this kind of stress concentration. Some simple redesign would address this issue by adding fillets, for example.
The FEA is correct in highlighting the location of stress concentration, but the magnitude of the stress at the crack is unreliable and dependant on the mesh size used. It takes real skill and experience to know what the actual stress will be!
For complex problems, where the stresses are difficult to derive accurately, it’s best to leave the advanced FEA for stress analysis to an expert analyst.