Stress acceptance criteria (peak stresses)

[JanEngineer]

Hi all,

At my company we want to establish acceptance criteria for FEM. This is mostly concerning peak stresses. Our safety factors are set by the company, we don't use any norm at the moment. The main issue we have is that we want to identify when a stress exceeding local stress it's a peak stress and when it isn't. This should not be engineering judgement but based on a norm. We will consider our loadcase to be static and therefore we don't take the fatigue into account.

I am not able to find a norm which we could use to determine if the stress can be stated as just being local (only a fatigue issue).

Preferably we would also go more into detail. For example how would you handle a compressive stress, shear stress or tensile stress. Once again we would like to pinpoint a norm.

We work in the dredging industry. It concerns on board cranes, suction tubes, bottom doors and more.

My question is if it is possible to use a norm for this (or the concept of determining the locality of a stress concentration from a norm)? I would greatly appreciate it if someone point me in the right direction.

Thanks

 

[EngineerMickeyMouse]

I have been through the similar survey for FEM quality assurance purposes... Below commentary is not a great advise you are looking for, but jest an input, hopefully helps.

Please note, various codes may distinct some definition differences between peak value and singularity value although it may not be given explicitly. In one word, it is frequent that standards' authors/creators do not discuss singularity subject, as it depends on experience and engineering judgement is a must. If you are creating some internal procedures for your company, in my opinion you have to use judgement (based on experience and FEA test runs) but once you will articulate it in the documented procedure you may call it a "norm". Nevertheless, I may refer you to ABS Guide for Building and Classing Drillships, which in section 5, table 2 gives allowable von Mises stress for various mesh size for small areas. What is a small area is left for the engineer interpretation. This is a codified indication that the subject is strongly based on one's judgement. Certainly what is given in a Ship design code cannot be applied to any other scale object analysis without special consideration/double check beforehand.

If I were to create a company internal procedure for FEA peaks I would start from the strict note for which object/structure procedure shall be applied to and for which shall not. My aim would be to provide tests models for various local mesh sizes and try to read out actual value in the singularity point. To achieve this it is recommended to follow e.g. Hot Spot Methodology linearisation, i.e. readout from 0.5t (t- thickness of the considered shell element) and 1.5t offset from the singularity point and linear extrapolation of the value and call it then an actual stress value (afterwards this value could be compared with your code allowable, false peak is excluded). But this will not be always true. You shall perform various linearisation tests and provide Excel-like graph plots against some hand calculated results to confirm yourself your interpretation is appropriate.

If you would go further to discover nonlienar static analysis you will quickly ask yourself what extent may have plastic area which experiences principal strains larger than elastic limit and is it still safe. Such a look into nonlinear could give you additional benefits of understanding that any peaks/singularities found in linear analysis are false, mathematic artefacts and would help you calibrating your peak allowable results. Refer to DNV-RP-C208 Determination of structural capacity by non-linear FEA.

Additionally please bear in mind that very frequently those peaks will be due to geometry which shall indicate need for a fatigue check, but this is another story.

 

[corus]

Use stress linearization to remove the peak stress. If you created the peak stress by application of a point load or restraint (which wouldn't exist in reality) then either ignore the stress there or model it better if that's a region of concern. There should be crane design codes available, but these tend to be written for those doing hand calculations. In general nuclear codes or pressure vessel codes go into design by analysis in more detail so I'd use those in conjunction with the appropriate crane design code.

 

[BILL_mce]

If I understand what your after correctly...which I may not..you want identify if the high stress at a location is critical or not? By critical I mean it can lead to failure or it just bleeds out on local yielding. If this is right then I would advise you run the problems with a perfect plasticity model and let them bleed out or cause the solution to fail due to loss of stiffness (ie yielding through the cross section). Of course this assumes you have access to a non linear solver. If this is not available then I fear you need to make a judgement call based on volume of material affected in the region of interest and how the resistance to the load is affected by local yielding.

Bill McEachern, P.Eng.
Redstone Six Consulting Ltd.

http://www.redstone6.com

 

[inline6]

if a stress value is a singularity (assume this is what you mean by peak stress) then another method is to check convergence. stress singularity will be heavily dependent on mesh density