Learning Forum

[Jeone]

Fatigue analysis (Time-Series SN) is done as linear superposition in DesignLife, however my FEA is nonlinear. Is this situation a problem? What can I do for this difference? 

In addition, I do 1 Newton analysis in FEA and the result does not reach to yield strength of material. Then, this result is used time-series fatigue analysis. When, the stress is multiplied by time series datas, it appears to have exceeded its yield strength. What should be my method for this?  Should I make a separate analysis for inputs that pass yield strength?

[danielshaw]

Yes, the time series load provider uses linear superposition.  If that assumption is not valid, then you may need to use the time step load provider.

Technically the stress-life method (SN) is only valid for elastic stresses, although its application is sometimes stretched to be used when the elastically calculated stress is slightly above yield.  SN is not valid if the stress is significantly above yield.

With time series loading, the fatigue stress is the FE stress multiplied by the time series scalars.  For SN to be strictly valid, that stress (FE*time series) must be less than yield.  If elastically calculated stress is significantly above the yield stress, you should use the strain-life method (EN).

[Jeone]

I did time-series fatigue analysis and I saw some nodes are more than yield, then I did EN analysis. I obtained a good result. 

The only thing that confuses me is that when I do structural analysis of the same loads in the time-series, it comes out close to the yield strength. In time-series, it exceeds the yield strength. In fact, the difference in tension between the two is about 6%.

If I use time-step load provider, I have to change just the scale factor. However, there are many loads in datas. Actually, I did today, but the life was more then 1E8, and this is not true. Is there any method? 

 

[danielshaw]

I do not know why your time step loading model is not producing valid results. 

You can use the time series load provider in an EN analysis even if the elastically calculated stresses exceed yield.  The time series load provider will use linear superpositon which is valid if the stresses are elastically calculated.  The EN engine will use an elastic-plastic correction (Hoffman-Seeger or Neuber) on the final stresses to calculate the plastic strains to be used in the fatigue evaluation.

You cannot use the time series load provider if plastic strains were calculated in the FE.  Linear superposition is not valid for nonlinear stress results.

[Jeone]

The stress does not pass the yield, because I use unit load(1N,1g etc.), and I hava to do nonlinear analysis because of contacts. When I use time-series load provider, I can see that some nodes' stress more than material's yield, so I do EN fatigue analysis. Consequently, I think, I have to use time-series. Is it possible, isn't it?

[danielshaw]

probably not.  Time series uses linear superposition.  If you have non-linear contacts in the model, linear superposition may not be valid.

[Jeone]

Even if I use a time-step load provider, I have too many load conditions. For example, I have a 2g load and I have a 1.8g load. These are approximately 3 hours of data. How many percent wrong will I be if I use time-series?

[danielshaw]

I do not know how much the non-linearity will affect your results.  It depends on how much the non-linearity affects the calculated stresses.  Constant Amplitude and Time Series load providers use linear superposition to combine stresses.  They do not check if the analysis was linear.  They essentially just assume that the stresses are from linear analysis.  If the non-linearity has a small effect on the calculated stresses, then CA and TS might produce reasonable results even with non-linear analysis.  However, if the non-linearity has a large effect on the calculated stresses, then CA and TS might produce incorrect results.

[Jeone]

You are right, it is depend on the stress. Thank you for taking the time to answer.