bug in Leq computation

wetb.prepost.Simulations.Cases.fatigue_lifetime contains a bug resulting Leq calculations that seem to be off with a factor x.

The intended implementation is outlined in the attached pdf, but there is an ambiguity whether or not S refers to range or amplitude.Leq_and_1Hz_eq_derivation.pdf

@cgal @rink

But isn't it the last equation you are using.

Yes, it's the last equation from Leq_and_1Hz_eq_derivation.pdf

Anything before that is what pdap and wetb are doing the same, and is what results in the 1Hz equivalent load.

So the last equation from Leq_and_1Hz_eq_derivation.pdf is implemented from L4801 to L4810 in wetb.prepost.Simulations.Cases.fatigue_lifetime

Here I attach my hand written notes

Calculation of the lifetime fatigue for a harmonic signal with amplitude=0.5 and freq=1Hz

I have add such a signal in a htc file and I did the analysis with the WE toolbox as well. I assumed only DLC1.2 with wdir=0 and 1seed per wsp.

The calculations are located under mimer here: X:\BLATIQUE\OlsenW14p3m_V27\A00006

I discussed with Jenni before a while, and the issue is in the calculation related with nT*neq, see my notes for the lifetime calculation. As Jenni said the 1Hz eq. loads are correctly calculated.

Here is the problem (I think)

weight = hours_pr_20year * 3600 / sim_time (https://gitlab.windenergy.dtu.dk/pdap/source/blob/master/appfuncs/dlc/sensor_statistics.py#L26)

while https://gitlab.windenergy.dtu.dk/toolbox/WindEnergyToolbox/blob/master/wetb/prepost/Simulations.py#L4808 uses the hours directly

Yes, I agree, that is exactly the difference. wetb.prepost.Simulations.Cases.fatigue_lifetime does not scale the 1Hz equivalent load to one hour based on the simulation length in seconds. I am not totally convinced that this scaling is actually correct though....I am preparing an example to show this, otherwise I will be plain wrong :-)

I was obviously plain wrong. fixed with e4bea73e. And as a bonus, there are also some unit tests now! See 7e700c88.

closed

@mmpe, do you think eq_load test from 7e700c88 should be moved to test_fatigue.py?

Yes, absolutely