Design by Rule VS Design by Analysis

[Paulettea]

Dear All

I am performing calculations for a pressure vessel according to ASME BPVC Sec. VIII Div. 2. First, I modeled the vessel in PV Elite which performs Design by Rule as per Part 4. Of the VIII-2. All the nozzles and shells passed the strength calculations based on the design conditions. Then I started to design the vessel based on the guidelines in Part 5 of the VIII-2. First, I used the Ansys workbench software to obtain the stresses in operating pressure and temperature conditions. I continued the procedure and the fatigue calculations for components passed and there will be no failure during the service life of the vessel. However, I noticed that despite the fact that these nozzles have passed the Design by Rule conditions they fail the conditions in Design by Analysis. I mean when the calculations are done if I define some SCL’s and obtain stresses P_m & P_L+P_b they fail the protection against plastic failure conditions in par 5.2.
The question is:
Can I ignore these results and use design by rule as a tools for passing static loads? After all, I am using Design by Analysis for protection against cyclic loading not plastic failure.

Warm Regards

 

[TGS4]

Unfortunately, you're probably misapplying the DBA rules for plastic collapse. Nevertheless, the rules are that if you decide to design the thickness using DBA, then you have to demonstrate all of the failure modes, and not cherry-pick.

If something passes DBR and fails DBA, I'd like to know the details, as this is a potentially serious problem.

 

[Paulettea]

Thank you TGS4 for your reply

After your reply I read the Code and found something:

(b) Protection Against Local Failure – these requirements apply to all components where the thickness and configuration of the component is established using design-by-analysis rules. It is not necessary to evaluate the protection against local failure, 5.3, if the component design is in accordance with Part 4 (e.g., component wall thickness and weld detail per 4.2).

(d) Protection Against Failure From Cyclic Loading – these requirements apply to all components where the thickness and configuration of the component is established using design-by-analysis rules and the applied loads are cyclic. In addition, these requirements can also be used to qualify a component for cyclic loading where the thickness and size of the component are established using the design-by-rule requirements of Part 4.

I think these sentences allow using just some parts of DBA while other aspects of design are based on DBR or maybe I am misunderstanding them.

Warm Regards

 

[TGS4]

Not really a misunderstanding, as much as bad guidance in the words. I believe that we fixed some of that in the 2017 Edition - please check the 2017 Edition.

Nominally, thought, my expectation would be that anything that passes DBR, would pass DBA (standard flanges notwithstanding).

 

[Paulettea]

the rules are that if you decide to design the thickness using DBA, then you have to demonstrate all of the failure modes, and not cherry-pick.

TGS4, I do not have access to 2017 edition but I think I got the root problem. I have not decided to design the thickness by DBA. Rather, I want to use DBR in order to design the thickness. Now, if I am supposed to follow all requirements of design by rule then I have to design the component against fatigue failure. And there is no rule in Part.4 regarding fatigue design of a component. Instead, in Part.4 it is always necessary to check the thickness against fatigue using screening criteria in Part.5.

4.1.1.4 A screening criterion shall be applied to all vessel parts designed in accordance with this Division to determine if a fatigue analysis is required. The fatigue screening criterion shall be performed in accordance with 5.5.2. If the results of this screening indicate that a fatigue analysis is required, then the analysis shall be performed in accordance with 5.5.2. If the allowable stress at the design temperature is governed by time-dependent properties, then a fatigue screening analysis based on experience with comparable equipment shall be satisfied (see 5.5.2.2).

So I think what I did was not picking some criteria in Design By Analysis and ignoring the others. It actually was doing the whole design in accordance with requirements of Design By Rule.

 

[TGS4]

Ok. I think I understand what you're doing. I stand by my original comment that you are applying the plastic collapse rules incorrectly. Nevertheless, carry on with your fatigue analysis.

 

[Paulettea]

TGS4, I am in trouble interpreting the code for elastic stress fatigue analysis.

First of all, do I need to linearize the stresses for fatigue analysis? I do not think it is necessary to linearize the results of stress analysis since for fatigue the total stress is required to be evaluated including stresses resulting from stress concentrations and other local effects.

Then I need to know the validity of this method with respect to the mesh size of finite elements analysis. Since a linear elastic procedure is to be followed, there will be no limit on the stress of the material in material properties and no yielding and subsequent strain hardening is considered for the material. In such cases if there is a sharp edge in the geometry the stresses go to infinity based on linear elasticity theory. When it comes to finite elements analysis this phenomenon shows itself with change in the stress results when the elements size change. Here, the finer the meshes are generated for the analysis, the higher the resulting stresses will be. Even if there is no sharp edge in the model, the areas around a discontinuity are very sensitive to mesh size when it comes to calculating maximum equivalent stress.

If the elastic stress analysis is applied in protection against plastic collapse this problem does not exist since linearization is done and the effect of local stress concentrations are neglected for being a part of peak stresses.

I personally modeled a nozzle in a shell (using solid elements) and ran the finite element model in Ansys workbench and checked the effects of element size on the stresses. The results show that when I used finer mesh the stresses resulting from linearization (membrane and bending stresses) are not much sensitive to mesh size (maximum 1 or 2 MPa in the 150 MPa). However, the maximum stress continues to increase by decreasing the mesh size. It is very important since the expected life of a component can change considerably with the change in the mesh size.

Please help me come out of this doubt.

Warm Regards

 

[Paulettea]

And one more question:

Suppose that I am designing all the thicknesses based on DBA. Is it possible to do the protection against plastic collapse using elastic stress analysis and protection against cyclic failure using elastic-plastic approach? Or, design against fatigue with elastic stress analysis and design against ratchet using elastic plastic method?

Warm Regards

 

[TGS4]

You point out a good and well-known point regarding stress-at-a-point. It's important to not have such discontinuities in your model.

Regarding mixing and matching the methods, it is generally acceptable.

 

[Paulettea]

Thank you TGS4.

At least now I can understand that I am not the only one who has faced this issue. Nevertheless, I think it is not a good engineering practice to not model discontinuities in order to prevent stresses to go high. You see these type of sharp edges are actually present in the real vessel. But using linear elastic model for these points leads to results that are sensitive to mesh size for fatigue. For plastic collapse it is OK because of linearization but in fatigue design the validity of linear elastic stress method is quite questionable.

Warm Regards

 

[marty007]

Paulettea,

One method that I've seen to deal with the issue of the discontinuity is described in a document by the International Institute of Welding (IIW) titled: "Recommendations for Fatigue Design of Welded Joints and Components".

In this document it talks about reading the stresses at a couple of points away from the discontinuity, then extrapolating these stresses back to the discontinuity. Using this method, you should be able to refine the mesh enough to converge the stress at the points away from the discontinuity, using these to calculate a peak stress for use in the ASME VIII-2 formulas for fatigue life.

Take a read through the article though, as it discusses a few different options for extrapolation.

 

[BJI]

Why not use the mesh insensitive structural stress method? It would be the preferred method for welded fatigue analysis. It is only fully mesh insensitive if you are extracting line forces, however with typical expected levels of mesh refinement through the thickness, linearizing stress should give you the same results.