First thing with all bellows is to make sure that the modelling is correct. Especially if these theoretical calcs are taken to extreme. Couple items to look at:
- for rubber bellows you have to enter all stiffnesses. If you do not do so CAESAR II calculates one or the other and it is not correct as the buil in formulas are for metal bellows.
- model tie rods using correct diameter, temperature and material. CAESAR II may calculate even temperature expansion etc. I have not done too many tests but better to be save than sorry. Include into the model pump stiffness. Nothing is rigid. Typical case of hair splitting can be seen on my site worksamples -
www.jat.co.za. There is one metal bellows with 2 tie rods side by side. Plant owner's designer figured out elongation of the rods was too much for the pump's nozzles. Only way was to add rods. Bellows were added in a first place because they could not build big enough concrete foundations to hold the pipe forces. My limited brain tells me that if the foundations cannot hold the pump then pump cannot be modelled as rigid point and therefore minor elongation of a rod is absorbed easily.
- double check what CAESAR II is doing with the pressure thrust. Nozzle should see as maximum the bellows pressure thrust less pressure thrust calculated using pipe ID.
In your case you have tie rods so the pump doesn't see bellows pressure thrust "directly". In your sketch you have horizontal section. If it expands you will see increasing axial force as the expansion is resisted by the vertical pipe (no compensators). This force is increasing until it reaches the full pressure thrust and then the rods are loose. Now you have untied bellows. See above for the nozzle load. If you have rod nuts on both sides of the flange/bracket situation is different.
If a pump manufacturer says that his pump is made out of paper and can take only small forces then ask them to give pump stiffnesses including foundations. Enter into CAESAR II and you will get different results.
If a calculation says that pump fails but there is an existing installation that works I scrap the calcs as I consider them wrong. Somebody calculated that a bee cannot fly. Last time I saw one it was happily flying.