I am using FAST to model the loads on the tower (Monopile turbine) and then access the short term 10 minutes fatigue damage in the tower to be used on a probabilistic analysis.
So far I have implemented a code that allows me to do this by running FAST, doing the rainflow counting and then, using the S-N curve, calculating the 10min_damage generated during the 10 minutes. I worked with stresses (using the section D=6m and t=0.027) and the rainflow methodology was the same used in m-life package. Used Goodman correction for 0 mean and using a commercial steel properties.
Then, I tried to validate the results for damage rate with m-life and some questions started to appear. If anyone can help me with them I will be really grateful.
a) When I was modeling short-term damage the highest values of generated damage occurred around the rated wind speed (see figure attached), but when using m-life I realized that the highest damage rate occurs at 20 m/s (see file attached. From the wind speeds I ran, I was expecting it to occur at 13 m/s).
Is this something I should be expecting? At the beginning I was quite happy that peaks of damage were located around rated power as it was where I was expecting more extreme loading.
b) When assessing the Lifetime_Damage with m-life I realised that the values of cumulated lifetime damage that I get are quite high (see file attached for the corresponding m-life simulations mentioned before).
Using the peak short-term damage rate that I obtained with m-life, and assuming that it is going to happen during 20 years (624365*20 x 10 minutes) I get a value that is smaller than the Lifetime_Damage given by m-life’s output. I have been checking the theory manual but wasn’t able to answer. Will it be because I don’t have wind speeds in all the Weibull bins so my results are not valid ?
I know that the results are highly dependent on the S-N curve, but I wanted to ask if there is any reference value for the damage rate of the NREL’s OWT tower damage rate (e.g. considering a simple tubular steel tower or other considerations)?
Best regards and many thanks in advance,
Test_reference_Lifetime_Damage.txt (871 Bytes)
Test_reference_Short-term_Damage_Rate.txt (7.07 KB)
I would normally expect short-term fatigue damage to increase with frequency because the standard deviation of wind speed increases with frequency.
Are you saying that you developed your own tool that performs rainflow cycle counting, applies the Goodman correction, and calculates the short-term damage, and the results do not match those calculated by MLife for the same time series? I would debug by comparing the output of each step e.g. the rainflow counting, followed by the Goodman correction, followed by the short-term damage.
The short term damage from MLife is reported as damage/second. I don’t see that you’ve converted applied the second/minute conversion in your calculation, so, perhaps you’re off by a factor of 60?
We have published fatigue damage equivalent loads for the NREL 5-MW turbine, but I don’t recall publishing damage rates.
Thank you for your reply, it was very helpful.
As I was interested in the rates I created my own code just to “glue” the different tools I was using, nothing really new. The rainflow I am using is one available online and is the same one used by m-life.
As you suggested I ran a step by step analysis and I was in fact missing the */s rate, so I was off by 600. Also, the fact that I was using a double slope S-N was introducing some deviations in the results. Using the same S-N formulation as m-life I get the exact same results.
When you refer to the variation of the wind standard deviation with the frequency, you mean inside the wind spectra used to generate the wind 10 min time series? So a wind spectra for a certain mean wind speed with high density at high frequency is expected to contribute the most for the short-term damage rate at that wind speed?
So its hard to tell which mean wind speed is expected to create more damage in average? Sorry, if I am being confusing. I am not sure if I got the whole picture.
Yes, I would expect to see more energy in the wind spectra (i.e. higher wind speed standard deviations) at higher wind speeds, so, I would expect the short-term damage rates to increase with wind speed. Of course, the importance of this increase to the total damage/lifetime depends on when the turbine is cut-out and the probability of occurrence of the higher wind speeds.
Thank you for your analysis of this particular topic and multiple others on this forum. They are very informative in my studies for Wind Technology.
I checked the results for the damage-rate function of the 3 classes of turbulence after validating with m-life and again I got a peak of damage rate around 12-13 m/s, then a decrease and a new increase to the same levels of damage-rate calculated for 13/ms around 20+ m/s. I suspect that this can be related to S-N curve that I am considering which has m=3 for higher ranges and m=5 for smaller ranges.
Initially I assumed also that I would have high damage at 12-13 m/s as I saw in other works that these winds were where the biggest loads in the turbine happen. Is this correct?
And then the further increase around 20+ m/s I was suspecting to be related to the cut-out speed which could be influencing the model and its behavior.
FAST control system does not consider the turbine shut-down, correct? In practice how valid is to model mean wind speeds around 22, 23, 24 m/s (if cut out assumed to be 25 m/s)? I unfortunately don’t have any sensation of how the turbine shuts down in practice…
Well, I will investigate a bit further and check the results with a 1 slope S-N curve.
Many thanks again as always!
Yes, I would expect high short-term damage rates loads near cut-out wind speed and perhaps at rated wind speed if the loads are quite high there.
I wouldn’t expect the high wind-speed controller shutdown to itself bring about high short-term damage. But in fact, the possibility for high damage is the reason for the cut-out wind speed in the first place.
I’m not an expert on controller shutdown behavior, but expect many turbines would time-average the wind speed from an anemometer and shut down the turbine when the time-average exceeds the cut-out wind speed. So, it is quite normal to simulate normal operational loads at 20, 22, and 24 m/s mean wind speed etc., without bothering to trigger the high wind-speed controller shutdown.
Most control functionality in FAST is user defined, but your right that the simple controllers NREL has provided as examples don’t include logic for high wind-speed controller shutdowns.