DNV RP-C203 Fatigue Analysis

[Crannog]

Hello,
I've a small offshore structure, with fatigue design to DNV RP-203 and analysed in FE (solid body). The design was sufficient to meet requirements of 2.11 such that all cyclic ranges well below Fatigue Limit [FL]. However the loads have changed and a single case is showing a design ratio in excess of FL.

2.11 requires that local stresses fall below the FL value for that detail's S-N curve.

The model shows relatively complex geometry and loading which might make the standard S-N curves less valid anyway. I am fairly sure that if I model in the welds I can reduce the peak stresses to an acceptable level, particularly using the Notch methodology covered in D.11 Comm. 4.1.

However I can see no guidance to show that it is appropriate to replace local stress with notch stress when justifying the omission of detailed fatigue analysis.

Does anyone have any experience with this regard? Thanks.

-D

NB: I believe that the geometry and loading would exclude use of the hotspot method, though the same question could be asked, can a interpolated hot-spot stress be used compared against FL.

 

[Reitzel]

Hi Cannog

It is always acceptable to increase the modelling detail by including the welds. You can include the welds and compare the new stress to the fatigue limit. But remember to compare the hotspot stress to the fatigue limit for the appropriate SN curve.. For a detailed FE analysis you already include the geometric part of the stress, i.e. the portion of the hotspot stress due to the geometri. For this reason it is typically appropriate to apply the D-curve in a hotspot analysis. The SCF embedded in the D-curve is 1.0 because the curve only needs to cover local notch effects.. But be careful! If you are dealing with anything else than double bevel butt welds you might need to consider a worse SN classification. Also, I don't think you should probe the stress value directly in the geometric centre of the hotspot. Are you considering extrapolation, as described in CHapter 4 of C203?

Another thing: Why apply solid elements? Are you dealing with very thick plates, for which plate theory doesn't apply (small transverse stiffness in comparison to membrane and bending stiffness)? 9 out of 10 times I apply shell elements in my models. If you apply shells you can use the hot spot stress as extrapolated according to chapter 4 in C203. This stress can be compared to the fatigue limit directly.

 

[Crannog]

Reitzel,
Thanks for the thoughts.

Yes, I've got a thick plate in the joint - I think a solid model is needed to accurately capture the localised bending within the joint and welds. For similar designs with thinner plates I have shelled the models.

I don't believe the DNV-hotspot method (covered in C203 section 4) is wholly applicable to the geometry and loading that I have in this case. I cannot think of a case where probing a peak however would give a result that wouldn't be considered to be conservative?

Yes, I agree with your point that the SCF for a detailed analysis is captured by the geometry in the D.11 Comm 4.1 'notch' geometry analysis. However is it still permissible to design to the fatigue limit, if stress range is applicable, for the detailed analysis D-curve? I would presume so.

The same is not directly stated in the C203 code that FL design is appropriate for derived hotspot stress ranges - though from fatigue theory would would say that it should?

Cheers

 

[Reitzel]

Hi Crannog

Regarding: I cannot think of a case where probing a peak however would give a result that wouldn't be considered to be conservative?

My thoughts: Probing the hotspot stress directly in the hotspot is more conservative than the extrapolation indeed! The C203 from last year, made by DNV and GL, states, that the hotspot method is applicable to both shell and solid elements. Why is it, that you don't think it's applicable to your problem?

Regarding: However is it still permissible to design to the fatigue limit, if stress range is applicable, for the detailed analysis D-curve?

My thoughts: The D-curve is more conservative than the Notch-Stress curve, i.e. for the same stress range in the two curves you get more damage out of the D-curve.. If you choose to follow the notch stress methodology and apply the D-curve you will be on the safe side by a large margin. According to Section 2.11 in C203 you can always compare the largest stress range a detail is subjected to to the fatigue limit. If all stress ranges throughout the lifetime of the detail are below the fatigue limit (stress range at 10^7 cycles) then a detailed fatigue analysis can be omitted.

 

[Crannog]

Hello Reitzel,
With regard the applicability of the hospot method? If I remember correctly the 1.5t (about 90mm in this case) distance for the extrapolation took the point around a 'corner' of the model into a different region so the loadings couldn't be considered to be representative of an extrapolation.

I feel there was another too reason but don't recall it.

 

[Reitzel]

Hello Crannog

If you can't make a straight line of 1.5t in the normal direction from the hotspot then you can apply the so-called method 1 (as i recall it), which only employs the 0.5t value, but multiplies this value with 1,11 as i recall.. You can look up the expressions for the effective stress range, it's written there.