I suspect most FEM programs do not consider the effects of local yielding and redistribution of stresses.

Dik

What FEA method was used? How do your actual materia properties compare to the ones used in the simulation? Did your simulation indicate "burst" based solely on plastic collapse, or did it consider local failure as well? What was the discretion error of the simulation? What's the error in the pressure measurement in the experiment?

Using Maths to model the real work within 6% seems pretty good to me. That is why code calculations have safety margins.

Details, details.

I was reading a report that indicated the calculated failure pressure that was higher than the actual burst pressure when the samples were tested to rupture. I found that to be surprising. I expected the calculated result to be lower than the actual burst pressure, i.e., I expected the formula to include a safety factor that resulted in a maximum pressure that would be lower than the actual test pressure that would cause the test specimen to burst. I thought the calculated values would be slightly more conservative than the actual failure pressure. Silly me and making an assumption.

Best regards - Al

Well, obviously the test results were wrong :)

The problem with sloppy work is that the supply FAR EXCEEDS the demand

The fabricated vessel has inhomogeneities that are not normally modeled in a FEA model. Weld heat affected zone has variation in hardness and strength, weld defects ( slag, voids , microcracks), base metal discrepancies from the initial mill N+T or other heat treatments. The properties that are input to the program are average properties or minimum properties that are based on a statistical analysis of test data, whihc implies there is a scatter of properties and there does not exist one true pure value for the main properties that define failure.

"...when logic, and proportion, have fallen, sloppy dead..." Grace Slick

Understood, but I thought the mathematical model would be conservative enough that the burst test would fail at a higher pressure than indicated by the model.

Best regards - Al

A metallurgical analysis of the failure might show a weakness not modeled in the FEA model. Also, There is the possibility of numerically generated errors in the FEA model- just because the output has pretty colored graphs and does not have error messages does not mean that the output is correct.

"...when logic, and proportion, have fallen, sloppy dead..." Grace Slick

I would concur with davefitz.

Particularly his 2nd post...

The problem with sloppy work is that the supply FAR EXCEEDS the demand