I know, plenty was said on SIFs in this forum, newsletters, manuals and Codes, but I would like to hear your opinions about the modeling technique of the long radius elbows (or tees) of significantly higher actual thickness then the matching pipe (tb>>tp).

According to ASME B31 codes, stress should be calculated based on the matching pipe thickness and SIFs based on the actual bend thickness.
Just as an example, Imagine simple L pipe (10” 9.3mm thick pipe and 18mm thick bend), anchored on one end and loaded on the other end with a force attempting to close the bend.

(A Model) To comply with the Code, the thickness in the input sheet for the pipe and bend remains the same (tp), but the fitting thickness (tb) is added in the bend input sheet for SIF calcs. The stress results in MPa are:
1. inlet weld – pipe side, SIF(1 / 1), Code Stress 80 (80)
2. inlet weld – bend side, SIF(1.666 / 1.388), Code Stress 100 (147)
3. inlet to the centre of the bend curve, SIF(1.666 / 1.388), Code Stress 95 (142)
4. outlet from the centre of the bend curve, SIF(1.666 / 1.388), Code Stress 95 (142)
5. outlet weld – bend side, SIF(1.666 / 1.388), Code Stress 88 (130)
6. outlet weld – pipe side, SIF(1 / 1), Code Stress 72 (72)
(For comparison, the values in the brackets are for the case if the bend was of the same thickness as pipe(tb=tp), SIF (in / out) = 2.564 / 2.12).

It seems to me that the calculated stress at nodes 2 & 5 is too conservative.

(B Model) Altough the Code has its requirements as per above, I believe that we should use the engineering judgement in this situation and model the bend as a separate element, with the actual thickness in the input sheet, and the SIFs user specified (same as per A). Here the calculated stress is significantly reduced!
1. inlet weld – pipe side, SIF(1 / 1), Code Stress 80
2. inlet weld – bend side, SIF(1.666 / 1.388), Code Stress 52
3. inlet to the centre of the bend curve, SIF(1.666 / 1.388), Code Stress 50
4. outlet from the centre of the bend curve, SIF(1.666 / 1.388), Code Stress 50
5. outlet weld – bend side, SIF(1.666 / 1.388), Code Stress 46
6. outlet weld – pipe side, SIF(1 / 1), Code Stress 72

In reality, considering the surface imperfections at the butt weld and its consequent stress risers, I would say that the pipe would either fail on the thin section of the pipe side of the weld, nodes 1 & 6, where the ovalization is not so prominent (???), or at the centre of the bend curve, nodes 3 & 4, where the effect of ovalization is at its maximum (if fitting is flexible enough). Actually, in case of very thick elbows (up to 3xtp for FRP), the pipe might have more tendency to ovalize (more flexible) then the bend, so the bend would make the 1,2 and 5,6 section more rigid then the straight pipe further away from the joint.
The basic two questions are:
- In reality, the extent of ovalization is not the same along the bend curve. What would be the actual stress causing the failure at the butt weld (thin side)? On which SIF value is it based; the actual butt weld's SIFs (unpredictable), or bend's SIFs, or SIF=1? If the first is the answer then we would have to account for it at every butt weld in the system to be safe.
- Are the SIFs calculated per ASME 31codes average values along a bend (tee) or max. values (at the centre of a fitting)? If the first is the correct answer, is the stress (B) calculated at nodes 3 & 4 insufficiently conservative?

With exotic piping materials or FRP piping, design to A or B would make major cost difference. We all know A is the safe and costly option. But I would like to hear from the more experianced engineers what they think about the B option, if it is safe enough?