I have a 20" 90deg elbow as a part of the pipeline that conveys raw gas from horizontal down to vertical line. Due to abrasiveness of the service the outer half of elbow needs to be made thicker then inner half. Due to thermal stresses they cannot afford to have the whole elbow thicker. Layout is fixed.
For the most accurate analysis in Caesar and using FEA I believe we have to do the following:
1. do calcs in FEA and get the SIFS and flexibilities.
2. Take this info and input in Caesar. The elbow should be modeled with average thickness for calculation of stress and loads.
What is the forums opinion about the above approach? All comments will be appreciated.
Ranka

why arn't you using sacrificial elbows?

I'd thought the cost of your special elbow+analysis will be astronomical??

(I've never done raw gas lines by the way.)

Someone else is doing this design. I was only asked for advise 'how to calculate / simulate' situation as is.
My biggest question is: If I do calcs as per method described above, are results of stress / loads calculated in Caesar based on average thickness are going to be realistic and conservative enough???

The loads you will get from Caesar will be good enough. You will then need to return them to your FEA model for full evaluation. From your description you have through wall thermal gradients and stresses that need to be taken account of. In other words the radial stress (which is one of the principal stresses) has a strong thermal component in addition to the pressure component. [From equilibrium principles it will also affect the hoop stress.]This should be taken into account. I surmise your pipe must be really thick for this to be pertinent.
I would be interested in how they fabricate a bend with thicker extrados than intrados.
By the way the bend will have differing in-plane and out-plane flexibilities because of this. Caesar can-not really model this but I do not think this issue as inportant as that radial thermal stress.

Dear 02061989, they will actually provide a weld overlay on inside of the top part of the split elbow. The base material is same and of the same thickness. The split was the solution for access to apply this overlay to improve abrasion resistance.
You said: "...radial stress (which is one of the principal stresses) has a strong thermal component in addition to the pressure component.[From equilibrium principles it will also affect the hoop stress.]..." and "...I do not think this issue as important as that radial thermal stress..."

Radial stress is normally caused only by difference between internal and external pressure. Please elaborate...

What is this jiu-jitsu?
It is you sir who says a priori "Due to thermal stresses they cannot afford to have the whole elbow thicker." This is a wonderful statement in itself. Have the "they" made a code submission for an unlisted component? And think about it you are the guy doing pipe stress analysis. So without a pipe stress analysis there is "thermal stress". It was my fair normative assumption that this "thermal stress" must be deriving from heat transfer temperature gradients through the pipe wall.

Michael

Dear Michael, I could not fully understand your previous reply perhaps due to my own limited knowledge of English language. The terminology you use is slightly different then what I am used to.
In any case I appreciate you taking time to make comments and discuss this topic.

Dear RS,
I have looked up on this. When you have thick pipe it is not necessarily isothermal and the temperature can vary from inside to outside to produce thermal stresses. You would need a Heat Transfer analysis to get the wall temperatures since the wall convection coefficients will be more dominant than the pipe wall conduction. Stresses are obtained from a thermal analogue to the thick pipe Lame formulas.
See Timoshenko and Goodier "Theory of Elasticity" (3rd Edn)Art 151 p449. Formulae 254, 255 and 256. With a Fourier analysis temperature profile for constant conductivity coefficient Formulae 257 give the three stresses and 258 with 259 the maximum values. Fig 228 shows stress distributions.