Hello,

The best place to look for the technical background information regarding structural design of SRV installations is ANSI/ASME B31.1, Appendix II. This (non-mandatory) appendix even provides examples of the calculations commonly used to check the structural design.

I think it is important to mention that the analyst should be careful about WHERE the Stress Intensification Factors (SIF's)are applied. Analytical models represent the piping system as centerlines with associated section properties. It is implicit in these models that the SIF's will be applied at (for example) the intersection of the header and the branch. If the Header (run pipe) is a large diameter, it may be that the point of attachment of the branch pipe (or Weld-O-Let (WOL)) is in fact a significant distance from the intersection of centerlines. It is equally important to feed the correct section modulus into the SIF equation - the WOL's section modulus is appropriate to use for the "branch".

When modeling large diameter headers with relief valves (e.g., some power plant high energy piping system headers - Main Steam, Hot Reheat, Cold Reheat....) I will use a rigid link from the header centerline to the header OD where the outlet to the SRV begins. Providing a "node" at this point will cause CAESAR II to calculate data (e.g., moments) there. This would be the attachment point for the WOL. My point here is that the correct "moment arm" would be calculated at the header OD and this is where the SIF should be applied. I would model the WOL as a piece of rather robust pipe (do some hand calculations to get it about right) from the header OD to the Weld Neck Flange (WNF). The WNF and the SRV body would be modeled as a sequence of rigid links (representing the valve body geometry) with appropriate weights (adding up to the WNF + SRV's total weight). The exit elbow could be modeled in the normal way (as pipe) and CAESAR II would calculate "Code Stresses" there.

I suppose the distance from the header centerline to the WOL attachment point at the OD would not matter much if the header was a relatively small OD. What do you "lurkers" think?

Best regards, John.