Using openfast with MLIfe for fatigue and ccyles to failure

[lt72884]

I have now completed the tasks of creating a turbsim.bts file that works. I have tested the .bts file in my 5MW_LAND_DLL_WTURB.fst and it works great. Thank you so much for all the help. I am very grateful.

Now that i have the one outb file and can view it in matlab and produce the plots. I would like to use MLife next for fatigue

I have installed Mlife and i have the documentation. What do i need to do specifically in Mlife to have it read the .outb file and calculate the fatigue and cycles to failure.

1. Do i need to have any of the mlife files in the 5MW_LAND_DLL_WTRUB folder?
   2.Do i need to edit any files and if so, what are the file extension endings fo those files?
2. What code command do i need to run in matlab to use mlife and calculate values?
3. Once values are calculated, how do i view them?
4. Is there by chance a video or teams meeting or demonstration on Mlife that could help me?

thanks for all the help thus far. It is really appreciated.

[jjonkman]

Dear @lt72884,

I'm not aware of a video or workshop tutorial stepping through the details of setting up and executing MLife post-processing. The MLife user and theory documentation and sample models from the MLife archive (https://www.nrel.gov/wind/nwtc/mlife.html) are you best source of information.

Best regards,

[lt72884]

ok, some progress has been made.

I created the following:
C:/fatigue/data
C:/fatigue/outputs

I opened the mlife.m file located in the MLife/Source directory. I copied the test01.mlif file from the MLife cert test folder and placed it in the MLife source directory. I changed its name to settingsFile.mlif. I then copied the DLC2.0_1.out to the C:/fatgue/data directory

i then ran the following lines of code

settingsFile = 'settingsFile.mlif'
dataDirectory = 'C:/fatigue/data,
outputsDirectory = 'C:?fatigue/outputs'

[Fatigue, Statistics]= mlife( settingsFile, dataDirectory, outputDirectory );

it then ran and created some fles that i dont understand the output. Matlab also threw the following errors:
'wblcdf' requires Statistics and Machine Learning Toolbox.

Error in compute_windspeed_bins/compute_bins (line 74)
bins(iBin) = wblcdf(WShi, c, k) - wblcdf(WSlo, c, k) ;

Error in compute_windspeed_bins (line 32)
[nBins1, binWidth1, binProbabilites1] = compute_bins(0, Fatigue.WSin, Fatigue.WSMaxBinSize);

Error in mlife (line 565)
Fatigue = compute_windspeed_bins(Fatigue, windChannelMeans,FileInfo.DLCs(1).NumFiles, FileInfo.DLCs(2).NumFiles );

so we are one step closer. What is next to fix this?
How do i use my .outb file generated from open fast IN Mlife to calculate fatigue and etc?

thanks

[jjonkman]

Dear @lt72884,

According to the MLife User's Guide, you need the Statistics Toolbox of MATLAB to run MLife within MATLAB. If you don't have that, you'll need to purchase it or replace the functions supplied by that Toolbox with your own. Alternatively, you can use the provided precompiled version of MLife.

Best regards,

[lt72884]

Ok, I'll look into that. Thanks

Does MLife use the .outb file produced by openfast to calculate the stats and fatigue?

[jjonkman]

Yes. MLife can process one or many OpenFAST time-series output files.

[ebranlard]

For simple fatigue analysis (equivalent load), you can also adapt the following python script:
https://github.com/OpenFAST/python-toolbox/blob/dev/pyFAST/postpro/examples/Example_EquivalentLoad.py

[lt72884]

ok, great. How do i do that? :)

thanks

[lt72884]

ok, something i have noticed.the first link you gave me supplied a version of fatigue.py that does not have find range count. as your second link shows. The first link you gave me for "equivalent load" i downloaded all the files and the fatigue.py in that is not the same as the second. can i just replace it?

thanks

[ebranlard]

yes you can replace it.

[lt72884]

For simple fatigue analysis (equivalent load), you can also adapt the following python script: https://github.com/OpenFAST/python-toolbox/blob/dev/pyFAST/postpro/examples/Example_EquivalentLoad.py

How do i run this? i do not program at all and have ZERO education in programming.

i did a little googling and i used powershell to run the script you provided but i get an error about numpy in line 7

File "C:\fatigue\fatigue.py", line 7, in
import numpy as np
ModuleNotFoundError: No module named 'numpy'

and this error as well

from pyFAST.input_output import FASTOutputFile
ModuleNotFoundError: No module named 'pyFAST'

for kicks and giggles i ran this from the power shell which completed sucessfully but it still throws the same error

pip install pyFAST

so i have no idea what to do now

[lt72884]

thank you:) the Leq value i provided above, i do not know wha the units are. the output is just a number so i am not sure what the units are.

I have been reading the theory manual and it has helped alot. I have been doing some calculations by hand as well. My professor showed me the print() function that i can use to see the output of certain things in the python.

Ill see if i can back out the cycles to failure somehow from the information i have.

[ebranlard]

As Jason mention, the equivalent load is returned in the same unit as your input variable. If your input is in [N], your output will be in [N].

[lt72884]

ok, thank you. I understand that now. I get what he means by saying "the time series load you provide"

as for the python question, i will replace the file

Next, in order to run find_range_count (or even the function eq_load_cycles), do i need to change line 10 from the Example_EquivalentLoad.py from

"from pyFAST.postpro import equivalent_load" to "from pyFAST.postpro import find_range_count" and make sure init.py in pyFAST>postpro has

from pyFAST.tools.fatigue import find_range_count

listed as well?

If this is not the correct procedure to run find_range_count or eq_load_cycles, would you be willing to share how to run those? thanks

as of right now, the ONLY output im getting is Leq:1244.76543

when i do get find_range_count to work, and i get the cycles for each bin, is that cycles to failure? or how many cycles it went under? IE 10^6, 10^7, or 10^8? and will it explicitly tell me?

thank you very very much

PS, i may have found an error in your code. The comment at the top of fatigue.py says m is a default value of 3 for the wohler exponent. BUT on line 87

DELi = S**m * N / neq
print(m)
print(DELi)

when i do a print(m) to see the output, it gives me that m=1 and not 3. I also verified it by doing the math by hand and m=1.

[(241.96)(13)]/60 = 52.426 which is the first DELi value listed in the table that is out put to vs code

here is the output of the print(m) and first few rows of print(DELi)

1

52.42614038 0. 96.7867207 0. 120.98340088
0. 96.7867207 0. 80.65560059 0.
120.98340088 0. 0. 0. 32.26224023
0. 0. 0. 0. 0.
44.36058032 0. 96.7867207 0. 0.

is it possible to change the value of m to 3?

[lt72884]

so DEL is NOT the same as Leq then? If i am reading the code correctly (Leq = DELi.sum() ** (1/m)), then Leq = DEL^(1/m) where m = 3 (by default) so in my case my leq is 1244 newtons so if i do the math then DEL = (1244)^3 which is a large number. This DEL value of (1244)^3 is the load produced under constant amplitude cycles that has the same degree of damage as real irregular loads?

suppose my Neq is 60 meaning "Number of equivalent cycles to failure." So failing at 60 cycles is that failing under the DEL value or the Leq value?

IE, 60 cycles to fialure at 1244 newtons or 60 cycles to fialure with (1244)^3 newtons?

thank you:)

[ebranlard]

The script uses DELi to mention the comtribution of cycle "i" to the DEL (or Leq both mean the same).

If you want to look at "damage", you need to apply the theory presented in MLife theory manual, and combine the cycles (S) and counts (N) yourself as currently we don't have a python-script for that.

[lt72884]

I THINK i got it. Here are some screenshots:

https://photos.app.goo.gl/Jz8ATyiSQq5qsXQR9
https://photos.app.goo.gl/UfCC2tUf4uxsB3xt7

i then used equation 26 and 30 to validate the math. However, the m value in the script is NOT defaulted to 3 like the comment says, its defaulted to 1. I used excel to test out and verify that it is indeed m=1 and it is. But i can change that value in excel now and get new numbers.

Can you confirm that the output i show in the images is indeed cycles and load range according to the script comments?

thanks a million guys

[ebranlard]

Yes, "S" and "N" in the code corresponds to the cycle ranges and number of cycles.

[lt72884]

now i have a theory question of the script. I noticed it is making 100 bins and some of those bins are empty as shown in the provided screen shots. Why are some of them empty and how is it generating these bins? And lastly, what is a bin?

thanks

[ebranlard]

Here's a link that explains data binning: https://en.wikipedia.org/wiki/Data_binning .
In the script, the amplitude ranges of the signals are split into multiple bins and counted according to these bins.

[lt72884]

question. line 26 of the script you made

Leq = equivalent_load(df['Time_[s]'], df['RootMyc1_[kN-m]'], m=m)

The RootMyc1 i know is a parameter that i added to the outlist section of my elastodyn.dat file. However, can i change that RootMyc1 in the script to some other parameter?

second, since the RootMyc1 is Kn-m that means leq is in Kn-m correct?

thanks

[ebranlard]

Yes you can use anyother channel present in the pandas dataframe. You can see them using print(df.keys()).
The units are indeed kN-m for that channel, and therefore the same for the damage equivalent load.

[lt72884]

Ok, thank you. I will try and find one that is Newton's only. Thank you again for all the help. I'll try the print function to see what the pandas tell me haha

Does the print function you list go in the fatigue.py script?

[lt72884]

ok, i have been working on the script and have some good results. I switched the parmeter from RootMyc1 to RootFxb1 so my DEL is in KN and not a moment.

Reading the theory manual, i am trying to find Neq cycles until failure. I have neq which is a differnt value, according to the manual neq is equivalent counts for time-series j. neq =feqT so 160. neq=60.

ok, now i know that Dj^st=neq/Neq (eq 25 in theory manual)
is Dj^st the same as DEL? or is Dj^st different? If it is differnet, does your script calculate Dj^st at all? if so, i would like to know where so i can find Neq which is cycles until failure.

If i knew the fixed mean load, and the ultimate load, i could find Neq as well. Does the script calculate the fixed mean load and ultimate load?

thanks you very much

[lt72884]

oh, i forgot to ask, is there any way to change Teq from the default of 1hrz to something else?

Teq : The equivalent period (Default 1, for 1Hz)

the reason i ask, when i look at the graphical output of RootFxb1 in either pydatview or matlab, in a 15 second span of time, there are only 3 periods. But the python script is set to a default of 1. The manual says "You must enter the equivalent frequency (Teq) of the DEL" im trying to find out what that Teq value is.

thanks