[James Pennock]

Modeling & Stress Analysis of Socket Weld Fittings
(From: The Discussion Forum for CAESAR II)
The question:
From: Dave McD.

What is the best way to model a piping system with socket weld elbows and tees?
To model an elbow, I was going to use three nodes; 1 at each end (use intersection type 8 at the sockets) and one node at the elbow intersection. The bend radius will be 1. However, does Caesar II consider the added stiffness due to the fitting thickness?
To model a tee, I was going to use 4 nodes; 1 at each end of the tee (type 8 at the socket), and 1 at the intersection. What type of intersection do I use here? (Type 2?). As with the elbow, will the additional stiffness be considered by the program?

The first response:
From: John Breen, Member of the ASME B31 Committee

Regarding socket weld fittings. That is an interesting question. Presumably we are discussing Standard ANSI B16.11 socket weld fittings.
First of all, remember that the Caesar II algorithm is written to address the Piping Codes – warts and all. Never lose sight of the fact that B31 has provided you with a SIMPLIFIED set of rules and expects you to use good engineering judgment to guide your analyses. And good engineering judgment includes an understanding of what is practical.
Regarding the “warts”: when you look at Appendix “D” you will see that we have flexibility factors (FF) and stress intensification factors (SIF) for (among other components) bends and TEE’s. Appendix “D” says that TEE’s have a flexibility factor of 1.0. In fact, they are much more flexible than that and there are several ASME papers that report the flexibility for TEE’s that is predicted by finite element analysis (FEA). Also, Appendix “D” would have you “turn on” the SIF’s and the FF’s at the weld line where the bend begins and “turn it off” at the weld line where the next section of straight pipe begins. Again, “it ain’t so”. Fact is, the ovalization carries out past the weld lines and it partially ovalizes the adjacent straight sections (or at least it can if there is enough bending – and there are no flanges). But we do not normally try to model these “end effects”; the B31 Codes don’t mention it and CII, following the Code, does not provide an easy accommodation for it. The net effect is to have a model that is not perfectly accurate as far as flexibility is concerned.
By the way, while you are looking at Appendix “D”, notice that there is no mention of socket weld fittings. There is no fatigue data available to the Committee.
The subject of ovalization deserves a closer look. Remember that we are using beam theory. Usually the pipe, as a beam, is equally as stiff (strong) on any pair of perpendicular axes on the cross section. When a pipe of (nearly) round cross section is made oval it will have a “strong” axis and a “weak” axis. This is just a result of the extreme fiber (relative to the long or short axis) being moved closer or further away from the neutral axis (centerline). If we bend a 90-degree elbow (about a centerline perpendicular to the plane of the elbow) such that the angle is reduced, the (nearly) round cross section will become oval with the long axis of the oval being perpendicular to the plane (parallel to the axis of rotation). If we pull the elbow open so that the 90-degree angle is increased, we will make the cross section oval with the long axis of the oval being parallel to the plane. So far, so good. CII can use the flexibility factor to adjust the calculated stiffness for the elbow to accommodate the relocated extreme fiber. However, if we bend the elbow as above AND introduce some tensional loading (all 3 dimensional piping systems have torsion) the ovalization varies around the bend. If we would look at the cross section at the weld line we would see the long axis of the oval is orientated at an angle from the plane. If we could view the cross section at, say, 10-degree increments around the bend we would see that the angle of the long axis continues to shift relative to the plane. It resembles the rifling in the barrel of a gun – going from one weld line to the other the long axis “twists” around the pipe centerline. A computer program cannot easily accommodate this “twist” (and incrementally apply a FF on the properly angled vector) because the torsion cannot be predicted prior to the problem solution.
Those comments don’t even address the fact that most bend manufacturers will start with thicker wall pipe so that after bending, the least wall thickness (thinned by bending on the extrados) will not be less than the required (or specified) wall thickness. It is also common practice for the manufacturers to make B16.9 elbows thicker than the schedule of the matching pipe. The thicker elbows are “end-bored” to make the wall match the straight pipe at the weld line – that (matching thickness at the weld line) is all that the B16.9 Standard requires for an elbow. The net effect is that you can have a model that is more flexible than the actual piping system. That is why CII facilitates a quick fix, allowing the analyst to make all the elbows a different thickness than the other components (i.e., runs of pipe) in the system. As an aside, I will mention that some analysts will be surprised to know HOW LITTLE in the way of geometry is actually standardized by B16.9 (or by B16.11).
As Rich Ay was mentioning in another discussion below, the Code (Appendix “D” again) gives us factors to use to “take away” some of the flexibility if we have a flange on one end of the bend or elbow. There is another factor to employ if there are flanges on both “ends” of an elbow or bend (it is estimated that a flange takes away the “extra” flexibility of about 30 degrees of elbow angle – so, if we have a 45 degree elbow with 2 flanges, how much flexibility do we have?). We even have a factor to use for big D/t ratios where the bend might be stiffened by internal pressure (bourdon effect). Frankly, these factors amount to “Band-Aids” on our simple beam element model. The bogus Code FF’s and the questionable (in hindsight) “other adjustments” that CAESAR II MUST employ (to satisfy the Code) assure that the final model will not be exactly accurate.
The real question is “how accurate do we really need to be” (?). Quite honestly, we will never know within 10 to 15 percent what the real loads are across a piping system (consider the vagaries of materials and the plastic response and “shakedown” at elbows). Also, from the above, we really do not have a perfect representation of the system in our model. The model will never perfectly reflect the flexibility of the “real world” piping system. This is why we have “factors of safety” (or indices of ignorance if you will), to allow for the vagaries of a non-perfect world. But (and this is a big BUT) the rules result in a safe and reliable piping system – those old guys on the B31 Code Committee in 1952 must have really known what they were doing (well, almost, we will have to fix the SIF rules soon).
So, let us consider the socket weld fittings. First of all by definition, they are NPS 4 (NPS 4 matching pipe) or smaller (they are relatively small). When we look at the dimensions as compared to a B16.9 butt-welding fitting, we see that they are pretty compact (more compact than a short radius elbow). They are very “close coupled” components – through the body, they are much thicker than pipe. Next, consider the fact that these components are not specified by “schedules”, they are pressure/temperature rated (much like flanges) and are specified by “classes”. Another important factor is that socket weld fittings are forged steel – relative to the matching pipe, they are very rigid. Now, consider the fact that we don’t really have allowable stresses for these things – what will you do with the calculated stresses (?). The socket ends will act somewhat like the flanges mentioned above. They will “take away” the tendency of the adjacent pipe runs to ovalize by holding the adjacent straight sections of pipe to a round cross. All this makes for an “elbow” with “much diminished” flexibility (same goes for the TEE).
What is the bottom line? It is a judgment to be made by the practical piping engineer (that’s why we get paid the “big bucks”). In my experience, 8 out of 10 practical piping engineers will model the socket weld fittings as rigid links (including appropriate weight). The flexibility will probably be no less accurate than about any other approximation that you might develop. Certainly, (at least for static loadings) it will be a conservative analysis. CII will not calculate a “stress” at the socket weld component, and there will be no stress comparisons for them in your B31 compliance report. The alternative is to develop FF’s and SIF’s experimentally or by FEA (Yuck!). Just make sure that you include the larger SIF that the Code recommends for the fillet welds where the pipe meets the B16.11 components (reference B31.1, Table D-1, note 11). Remember, by far the most common mode of failure that we see in the field with socket welds is fatigue at the toe of the fillet welds. If this is a system that would be expected to see vibration or if it is B31.3 “cyclic service” (also see severe cyclic conditions), don’t do it. Specify a weld like the illustration in B31.1 Figure 127.4.4(A) (d). Specify the examination of all finished fillet welds with repair for any undercut. Make sure that the welder provides the 1/16-inch gap shown in B31.1 Figure 127.4.4(B) (c).
Ha! You thought you asked a simple question.

The 2nd response
From: John Luff, Member of the ASME B31 Committee

I see you are involved in the steel industry. If you are at a high enough pressure I would also add…. that B31.3 “Chapter IX High Pressure Piping” specifically forbids socket welds K306.1.2 (a).
This prohibition is quite simple in that if one considers the loading of the socket fillet weld versus a circumferential butt weld it becomes readily apparent that the butt weld under pure tension rather than a combined shear/moment is far superior. I have had a first hand example of this demonstrated by two identical systems installed in separate plants operating at a rather high pressure. The butt welded system has been in service failure free for 12 years while the socket weld system has been the maintenance gang’s problem child for 9 years.
So you now know more about socket versus butt welds than you ever wanted to know!
One final thought… according to a rather widely known expert in fatigue research socket welds are the most fatigue researched fitting there are!
I am willing to bet that most piping engineers don’t even detail a weld profile or perhaps don’t even refer to a piping code as the minimum acceptable weld.
This coupled with the fact that usually the less skilled welders are put to work on the fillets can lead to an undercut, undersized fillet! So at the very least refer to a piping code for minimum acceptable weldment!
All welded out and headed to the gate……

The 3rd response
From: Shahid Rafiq
1) So as far as modeling is concerned, the socket weld fittings can be modeled as rigid elements. What do other forum members do?
2) It seems that the failure for socket weld fittings is mainly in the fillet weld. What about the socket weld fitting itself? Any failures experienced there by any forum member?

[Author]

Posts