Why does COMPRESS increase my shell thickness? 3

[TomBarsh]

Speaking of 'splitting hairs', some customers recently had similar issues in which they did not understand why COMPRESS was overriding the thicknesses they had entered for their shell components. This question comes up regularly and is worth discussing.

Any of a number of different Code criteria may govern the thickness of a shell component (formed head, cylinder, or cone), including internal pressure, external pressure, and longitudinal stress due to weight combined with pressure and external loads (wind, seismic, etc).

For components whose thickness is governed by internal pressure the required thickness is a function of the allowable membrane tensile stress from Section II Part D. The allowable stress is a function of the material and temperature. For any configuration of the component the allowable stress remains constant (assuming constant temperature).

But for components whose thickness is governed by external pressure the allowable stress is also a function of the component's nominal thickness. The allowable compressive stress is a function of the nominal thickness, the diameter, and the distance between lines of support, as well as the material and temperature. The allowable compressive stress is found as factor "B" in the material charts of Section II Part D; Factor B is in part a function of factor "A" found in Figure G, Section II Part D. Factor A is a function of the ratios L/Do and Do/t determined by the component's geometry. Thus the required thickness of the component is indirectly a function of its nominal thickness. Of course, if the required thickness calculated on the basis of the allowable stress for a given nominal thickness is greater than that nominal thickness, then the part does not "pass" and its nominal thickness must be increased.

The allowable longitudinal compressive stress is found using the same factors "A" and "B" found in Section II Part D (see Section VIII, Division 1, UG-23(b)). Even if there is no external pressure acting on the vessel in some cases the thickness of the component may be governed by longitudinal compressive stress; this might be the case, for example, for a component near the bottom of a tall vessel subjected to high wind loads.

Another issue exists when the compressive stresses due to external pressure are combined with the longitudinal compressive stress due to weight plus wind or seisimc. Again, if this condition governs the thickness of the component then the required thickness will be a function of its nominal thickness. COMPRESS applies the "Bergman" check for this load condition. This analysis is based on the paper "The Design of Vertical Pressure Vessels Subjected to Applied Forces" by E. O. Bergman (this paper is available for download from Codeware's website). The Bergman check determines the buckling mode of the shell under the combined action of circumferential compression due to external pressure and the longitudinal compression due to weight plus wind or seismic. Essentially, a factor (called the "ratio Pe&quot multiplies the design external pressure; the MAEP of the component must be greater than this product, otherwise the component is susceptible to buckling.

Thus we see four different load conditions or stress situations listed: design for internal pressure (tensile stress), external pressure (compressive stress), external loads such as wind or seismic (compressive stress), and external pressure combined with external loads (buckling). In the latter three cases the required thickness is an indirect function of the nominal thickness; only for the case of internal pressure is the required thickness independent of the nominal thickness.

COMPRESS provides two calculation/reporting modes: Design mode and Rating mode. In Design mode COMPRESS enforces that each component meets the design conditions; COMPRESS automatically increases the shell thickness to ensure that the component meets the design conditons. This may result in the thickness being increased over what the designer has entered. COMPRESS always forces a thickness increase to the next "commercially available" thickness. The designer can manually enter lesser thickness values to determine the absolute minimum thickness value that is possible. In Rating mode COMPRESS retains the thickness value entered. If the thickness entered is inadequate for the design conditions then some sort of Deficiency will be reported. This makes the Rating mode useful for investigating why a particular thickness is not accepted by COMPRESS in Design mode.

See

http://www.codeware.com/support/faq/1009.html

for more information.

We will see in a moment why I refer to "splitting hairs".


Tom Barsh
Codeware Technical Support

 

[TomBarsh]

Now that we've learned the "theory", let's take a look at some actual cases in which this issue has come up.

The three cases discussed also involve the issue of "precision" and how precise the curves of Figure G can be read. The precision issue affects only a small proportion of the total number of cases for which the shell required thickness is governed by compressive stress allowable or buckling but it is interesting to point out.


Case 1:

In this case there was no external pressure on the vessel. COMPRESS indicated that the required thickness of a cone was 0.3569" (the required thickness is the value shown in the gray background of the design dialog to the right of the input value) and had applied 7/16" (0.4375&quot material. Of course, 0.375" is greater than 0.3569", so the user was curious why COMPRESS changed the thickness value to 0.4375". The user wanted to use 3/8" (0.375&quot plate for this component; adding to the urgency was the fact that this vessel was already being fabricated on that basis.

It was apparent that longitudinal compressive stress was controlling the required thickness for this component. Using the technique of switching to "Rating" mode and changing the cone to 3/8" thickness revealed the overstress in the calculations. Switching back to "Design" mode allows the user to manually enter thicknesses for the component and to observe the corresponding required thickness. It's a simple matter to keep entering smaller and smaller thickness values until the "absolute minimum" thickness is reached.

This particular case was interersting. Entering 0.3756" resulted in an overstress and COMPRESS automatically increased the nominal thickness to the next standard increment (0.4375&quot. But 0.3757" was accepted without complaint. This value is barely greater than the 3/8" nominal plate actually being used for fabrication. However, a computer program does not see the gray areas but sees only in black and white: does it pass the "if/then" test? is it good/no good?

I do not think that the empirical curves of Figure G (Section II Part D) can be determined (or interpolated) to the level of precision implied by working to 4 decimal places. Even the method of interpolation between known points may influence the matter; eg: linear interpolation will produce a different result than logarithmic interpolation.

For this example the 3/8" is likely "good". However, in any case the design is right to the ragged edge.


Case 2:

This is a common issue: Design for external pressure governs the required thickness of the component. I can't find the specific example I was thinking of but recently someone had a case where they wanted to use 1/4" plate but COMPRESS was enforcing an increase to 5/16". Based on the spacing of their stiffener rings the L/Do was such that the required thickness was again a very small amount greater than 0.25" (as I remember it was like 0.2504&quot and the same comments apply as for Case 1 about being able to read and interpolate the curves of Figure G to an accuracy implied by the use of 4 to 6 decimal places.

A possible easy fix to this problem is to adjust the locations of vaccum rings slightly in order to decrease the unsupported length of the affected section. If the designer is lucky this can be done without adding an additional stiffener ring.


Case 3:

Another user recently wanted to use 1/2" minimum thickness on a cylindrical shell but COMPRESS was forcing the thickness to 9/16". The problem affected only one of a number of similar shells on a taller vessel.

The problem cylinder's required thickness was governed by the Bergman check for combined external pressure plus wind. Here the shell thickness could be manually "iterated" down to 0.5014", ie: this value was accepted but if 0.5013" was entered COMPRESS would change to 9/16". This was confirmed by running the calculations in Rating mode with either 1/2" or 0.5013" shell, in both cases the Bergman check failed. Again, this case was so close that using the 1/2" plate would possibly be justified but at the cost of there being no reserve strength left in the plate.

Only this particular cylinder was affected because of the higher weight acting near the bottom of the vessel. Slightly reducing the unsupported length of the component by decreasing the spacing of the vacuum rings at the problem cylinder so that its MAEP was increased slightly was enough to solve this problem.



In these 3 cases the shell thickness originally selected was found to be marginal at best. But all 3 cases illustrate the more common situations wherein COMPRESS may enforce a slightly greater shell thickness than what is expected and is simply doing its job to prevent an under-designed pressure vessel. The designer can review the calculations using the investigative techniques discussed.

Tom Barsh
Codeware Technical Support