2:1 Elliptical Vs 2:1 Torispherical Heads 3

[darrenyee88]

Hi all,

I understand that 2:1 Ellipsoidal head has crown radius (L) = 0.9D and Knuckle radius (r) = 0.17D. So is it true that 2:1 Torispherical head has the same crown radius and knuckle radius as 2:1 Ellipsoidal head ? If yes, Can someone guide me where to find this information.

Or someone can guide me where to find information regarding above to let me have a good understanding.

Thanks and cheers.

 

[shamino]

As per UG-32(e), the crown radius = outside diameter of skirt for torispherical heads
For ellipsoidal the crown radius L is obtained from Table UG-37.

 

[r6155]

Ellipse has infinite radius
Torispherical (spherical dished) head has L= inside spherical or crown radius and r = inside knuckle radius
Ellipsoidal head 2:1 means that inside diameter is 2 x inside depth of head

UG-32 (C): An acceptable approximation of a 2:1 ellipsoidal head is one with a knuckle radius of 0.17D and a spherical radius
of 0.90D

Regards
r6155

 

[TGS4]

r6155 has it correct that ASME Section VIII, Division 1, paragraph UG-32(c) does indeed allow that a permissible approximation of a 2:1 semi-elliptical head is a torispherical head with a crown radius of 0.9D and a knuckle radius of 0.17D.

However, in my opinion, that is a geometric approximation only. The two heads have very different stress results, particularly at the edge of the so-called knuckle. In fact, if you calculate the required thickness of this "equivalent" 90-17" torispherical head using the rules in Appendix 1-4, you will end up with a thicker head than a pure 2:1 semi-elliptical. This is recognized as a problem/inconsistency in the Code, and one that I have the privilege of acting as Technical Project Manager (in SG-Design of VIII) to correct. I (along with a few colleagues) are also writing a "whitepaper" on the subject that will be published at the 2017 PVP Conference.

So, I know this topic inside and out. Go ahead and ask any more questions...

 

[r6155]

To avoid problems I propose to delete UG-32 (c) entirely.

Regards
r6155

 

[r6155]

Sorry my mistake

To avoid problems I propose to delete the last sentence of UG-32 (c): "An acceptable approximation of a 2:1 ellipsoidal head is one with a knuckle radius of 0.17D and a spherical radius of 0.90D"

Regards
r6155

 

[JStephen]

I just assumed that the 90%/17% approximation was based on the assumption that heads weren't formed much more accurately than that, anyway, so you were just fooling yourself to think that one head was truly elliptical and one truly torospherical in the comparison of the two profiles.
Come to think of it, is the inside surface supposed to be an ellipse, or the outside, or the middle surface, or the middle surface of the corroded profile or what? I'm thinking that when you offset an ellipse by a given thickness, the offset curve is no longer an ellipse. Normally, theoretical analyses assume the middle surface is the exact profile.

 

[TGS4]

r6155Thank you for your opinion. That is one of the options that I am considering. However, the rules for heads are very different between VIII-1 and VIII-2, with the latest technology skipping VIII-1 and going straight into VIII-2. UG-32(d) and/or 1-4 doesn't align with 4.3.6 (even using an equivalent allowable stress basis), so there's a consistency problem, too.

JStephenIt's the inner surface. Because when the head manufacturer forges the head, his die is for the ID. he can make many different thickness heads with the same die just by changing the blank thickness.

Here's where things get a little bit more complicated - the information that I have been provided by head manufacturers (and I am very interested to hear from other head manufacturers) is that sizes greater than ~4' dia will not be forged with an elliptical die, but rather bumped and spun, meaning that they will be torispherical heads. So, with UG-32(c), even if you order a 2:1 semi-elliptical head, over 4' dia you will likely get the equivalent torispherical head.

What makes this important for me is that, as an FEA guy who often gets called when manufacturers need to do things like put nozzles into head knuckles, even if the drawing says a 2:1 semi-elliptical (and the calculations show the thickness for a 2:1 semi-elliptical), it may in fact be a 90-17 torispherical. And this would make a huge difference in the stress state for a nozzle in said knuckle.

 

[SnTMan]

(Sighhhhh) I miss the old days

The problem with sloppy work is that the supply FAR EXCEEDS the demand

 

[JStephen]

I may misunderstand the forming process- but I thought on a spun head, the crown/knuckle radius came from the movement of the rollers, not from the curvature of the rollers, so it wasn't necessarily a torispherical head even if spun.

 

[r6155]

TGS4 – Another inconsitency is in UG-81
(b) Hemispherical heads or any spherical portion of a
torispherical or ellipsoidal head designed for external
pressure shall, in addition to satisfying (a) above, meet
the tolerances specified for spheres in UG-80(b) using a
value of 0.5 for L/Do .

We can’t do thickness calculation as ellipsoidal head and then inspect as torispherical. I always inspect ellipsoidal head with template made with dimensions according to ellipsoidal.

Regards
r6155

 

[TomBarsh]

"Come to think of it, is the inside surface supposed to be an ellipse, or the outside, or the middle surface, or the middle surface of the corroded profile or what? I'm thinking that when you offset an ellipse by a given thickness, the offset curve is no longer an ellipse. Normally, theoretical analyses assume the middle surface is the exact profile."

1- In my discussions with head manufacturers they will generally form the heads so that the inner surface is ellipsoidal. But this needn't always be the case, one could pay the piper (or for head fabricator) to form the head to any desired configuration.

2- Yes, it's mathematically true (and practically so, as well) that a curved line "parallel" to an ellipse is not itself an ellipse.

 

[RonJeremy]

T.E heads?
I've been working in pressure vessels since '86 and never built, bought or specified a T.E head. Ever.
Back in the 1950s when vessels were designed on a slide-rule, T.E heads made some sense.
Why?
Well, stresses in spheres and toroids were very well understood at the time [Timoshenko, etc.], and a T.E head is simply a part of a sphere attached to a part of a torus, attached to a cylinder: a fairly simple calculation. S.E heads were a bit of a black art back then, being designed on empirical bases.
In the modern (computer) age I cannot see any justification for a T.E over an S.E and I do not think I will use them ever again.
PS: If a sub-supplier has difficulty rolling an S.E head to your requirements, maybe he is not the one for you .....

 

[TGS4]

Have you ever checked any of your 2:1 semi-elliptical heads with a template to see whether or not they are true semi-elliptical heads or if they are equivalent 90-17 torispherical heads? And can you even tell with the tolerances in UG-81? With the statement in UG-36(c) (note that this used to be UG-36(d)), it is likely that you weren't even told of the switcheroo, because it was/is all perfectly legal.

I will also refer you VIII-2 4.3.6 and 4.3.7, whereby the thickness of 2:1 semi-elliptical heads are explicitly calculated using the 90-17 torispherical head approximation.

p.s. your abbreviations T.E and S.E are not understood on this side of the pond. I find that abbreviation generally don't translate well - which is why I have taken the pains to spell out as much as I can...

p.p.s. Elliptical heads are generally pressed, not rolled. Torispherical heads are generally rolled/spun.

 

[RonJeremy]

Have you ever checked any of your 2:1 semi-elliptical heads with a template to see whether or not they are true semi-elliptical heads or if they are equivalent 90-17 torispherical heads?

YES, and Code tolerances apply.

I will also refer you VIII-2 4.3.6 and 4.3.7, whereby the thickness of 2:1 semi-elliptical heads are explicitly calculated using the 90-17 torispherical head approximation.

Yes, and what is wrong with that? I never said that design of S.E heads is anything other that empirical (see above). Code design rules.

p.s. your abbreviations T.E and S.E are not understood on this side of the pond...

Do not care. You say tomato and I say tomato. We are not all Americans, please accept.

p.p.s. Elliptical heads are generally pressed, not rolled. Torispherical heads are generally rolled/spun.

Yep, my fault. Thanks!

 

[TGS4]

Have you ever checked any of your 2:1 semi-elliptical heads with a template to see whether or not they are true semi-elliptical heads or if they are equivalent 90-17 torispherical heads?

YES, and Code tolerances apply.

Very interesting. Do you have any anecdotal data to say whether or not you were getting 2:1 semi-elliptical or 90-17 torispherical? At what sizes? Most head manufacturers that I have talked with say that below 4' diameter you will usually get a 2:1 semi-elliptical, but above that size will be the 90-17 equivalent torispherical.

I never said that design of S.E heads is anything other that empirical (see above). Code design rules.

The design rules in 4.3.6 are most certainly not empirical, but quote rigorously scientific, evaluating multiple failure modes.

We are not all Americans, please accept.

Indeed we are not - I am not American either. Many on here don't even speak English as a first language. If you want yourself to be understood, you may want to keep the lingo/abbreviations/shorthand to a minimum.

 

[RonJeremy]

You first said that:

(I) "the thickness of 2:1 semi-elliptical heads are explicitly calculated using the 90-17 torispherical head approximation" [i.e., empirical]

and yet later you say that:

(II) "the design rules are most certainly not empirical, but quote rigorously scientific [ ... ?], evaluating multiple failure modes"

Could you explain whether the design rules for S.E heads are based on rigorous science (as you say in II), or empirical bases (as you say in I)?

 

[TGS4]

I said that the rules in 4.3.6 are rigorous. The rules in 4.3.6 are for torispherical heads.

I'm not so sure how happy I am with the approximation in 4.3.7 (approximating a 2:1 semi-elliptical head with a 90-17 torispherical head), but if that is, in fact, what head manufacturers are providing for semi-elliptical heads above certain sizes anyway, then it really doesn't matter, does it?

Nevertheless, the biggest concern is that although the 2:1 semi-elliptical head and the 90-17 torispherical head, although from a geometric perspective they appear to be close enough approximations, the stress results from them are anything but. Apparently, shells of revolution are extremely sensitive to minor variations in shape

 

[Don56]

I made tank heads for 28 years. Back in the olden days we made SE 2:1 heads using an 85-15 approximation. Later we switched to the 90-17 method because it is more economical. However we always checked our SE heads against templates that were cut to actual 2:1 elliptical radii. Using either the 85-15 or 90-17 method produced parts well within the tolerances of UG-81. We're not making Swiss watches here guys/gals. FYI the depth of a 2:1 elliptical head is not 1/2 the diameter it is 1/2 the radius or 1/4 of the diameter.

 

[SnTMan]

We're not making Swiss watches here guys/gals.

Thank you for that

Regards,

Mike

The problem with sloppy work is that the supply FAR EXCEEDS the demand