From Tony Paulin:
=========================================================================
There is a brief set of equations and notes on our web site that may be of help.
http://www.paulin.com/WEB_Markl_SIFs_ASME_VIII_2.aspx Refering to the problem with the bend described in the post: the failure in the pipe bend is generally along the side starting from the ID. In this case, the FEA model using brick or shell elements should match the B31 calculation iM/Z within about 10% providing d/t and R/d ratios are typical and the mesh is adequately refined to trap the sidewall bending stresses. Hopefully you're using an in-plane moment load and about 3-to-5 d of pipe on either end before the boundary condition. For the brick model the peak stress on the ID of the bend should be approximately equal to the linearized bending stress along the same line, and then in this case the stress (Pl+Pb+Q) should be about double iM/Z, where i = 0.9/h^2/3 and h= tR/r^2.
In our limited scope testing of numerous piping components with welds subject to cyclic loads in both the Markl machine and in a separate cyclic pressure test we have not found the sensitivity of the FSRFs published in the new 2007 Div 2 or in WRC 432 to inspection methods. For comparison with shell models at weldments we use the 2007 Div 2 recommendations for modelling, and an SCF=FSRF(pipe?)=1.35 to adequately envelope our compilation of tested Markl SIF results. Our testing has focused on low cycle fatigue, (less than 40,000 cycles), although in our spare time are building a high cycle machine to validate higher cycle failures in welded and non-welding piping components.
The FSRF and SCF have different meanings since they are intended to be used with different curves. We've recommended that the Markl curve be adjusted in the low cycle range (less than around 10,000 cycles) to bring it in line with the large set of other fatigue tests conducted around the world. We think the FSRF is a good concept that will ultimately be refined for use in the fossil and petrochemical industries, and further refined for the piping vs. pressure vessel industries. (You can typically inspect the ID of vessel girth weld, but more often can't easily inspect the ID of a pipe girth weld.)
There are significant issues with d/t as it gets smaller, and the effect it has on fatigue. The variation of the stress through the thickness is discussed in the referenced WRC documents, but it's effect on leakage prediction via the mechanism of fatigue is not always clear.
A more comprehensive way to use ASME VIII-II FSRF values to determine SIF’s is described below:
1. Determine the local M+B stress (or PL+Pb+Q).
2. Multiply by the FSRF for the weld joint, normally around 1.8 or 2.0 for most piping welds (girth welds and intersections).
3. Calculate the mean failure life using Langer’s original mean failure curve S = 8664/(N^0.50)+21.645 ksi.
4. Using the mean failure life from Step #3, calculate the SIF from Markl’s equation Sf*i = 490*Nf^-0.20.
a. Where Sf is the nominal bending stress in the pipe caused by M/Z.
Because the relationship between the Markl curve and Langer’s curve is a function of failure cycles, this is the only way to ensure that the FSRF is properly used and still get a valid SIF. The failure life is the only known parameter whereas stress is dependent on the particular definition to be applied.
Something that has caused confusion in the past is mixing of terms like “FSRF” and “SCF” and their application. An FSRF of 1.35 works for predicting experimental pipe failures because the Markl curve is lower than the ASME curve in the low cycle regime. However, if you were to use the ASME smooth bar curves for life predictions, those same tests would require higher FSRF’s. This is apparent simply on the basis that the FSRF is a scalar between the failure point and the mean curve. If the mean curves are the same, the FSRF would be the same. ASME VIII-II mean curve isn’t the same as the Markl curve and therefore we need different sets of FSRF’s to predict failures.
Nowadays, FSRF has a very special meaning in that it is tied to the ASME VIII-II fatigue curves. These FSRF’s can’t readily be used to predict Markl SIF’s since they have been developed from a different mean curve. This would be like attempting to use a SIF with ASME smooth bar curves; it just won’t work.