questions about reference control input

This forum if for discussing controls. Questions about how to implement controls in FAST are more appropriate to the CAE Tools forum.

Moderator: Bonnie.Jonkman

Cheng.Zhang
Posts: 57
Joined: Tue Nov 14, 2017 6:50 am
Organization: ECN
Location: france

Re: questions about reference control input

Postby Cheng.Zhang » Thu Jul 12, 2018 3:24 am

Dear Jason,

I want to use mlife to calculate the Damage Equivalent Load (DEL) as a performance index for my controller, and I read the topic concerning the Mlife and the Mlife_user, but I still don't understand how to calculate it.

For example.
If I want to calculate the DEL of blade flap-wise bending (RootMxb1) and tower fore-aft bending (TwrBsMyt).
How can I choose the parameters of NSlopes, SNslopeLst, BinFlag, BinWidth/Number, TypeLMF, LUlt.

I read a paper that you are a co-author https://arc.aiaa.org/doi/abs/10.2514/6.2008-1307. In this paper, you also use DEL as a performance index (e.g. 'Flap DEQL', 'Tower FA DEQL', 'Tower SS DEQL', 'LSS DEQL'). Could you tell me how to calculate the DEL of those item?

Thank you!

Best regards,
Cheng

Jason.Jonkman
Posts: 4338
Joined: Thu Nov 03, 2005 4:38 pm
Location: Boulder, CO
Contact:

Re: questions about reference control input

Postby Jason.Jonkman » Fri Jul 13, 2018 12:11 am

Dear Cheng,

SNslopeLst is the slope of the SN curve. If not known for the specific components you are analyzing, it is often appropriate to assume the slope is 3, 4, or 5 for steel structures or 8, 10, or 12 for composite structures (you can choose one or use all three to assess the sensitivity). You can check the sensitivity of the binning based on your own results. TypeLMF and LUlt are discussed in our FAQ (https://nwtc.nrel.gov/FAQ#TypeLMF), but are unused if you are not using the Goodman correction to account for the influence of the load means on the DELs.

Best regards,
Jason Jonkman, Ph.D.
Senior Engineer | National Wind Technology Center (NWTC)

National Renewable Energy Laboratory (NREL)
15013 Denver West Parkway | Golden, CO 80401
+1 (303) 384 – 7026 | Fax: +1 (303) 384 – 6901
nwtc.nrel.gov

Kamineni.Akhil
Posts: 5
Joined: Sun Nov 04, 2018 4:08 pm
Organization: SELF
Location: Germany

Re: questions about reference control input

Postby Kamineni.Akhil » Thu Jan 31, 2019 4:16 pm

Jason.Jonkman wrote:Dear Cheng,

Yes, the structural DOFs can impact the performance of the controller.

The pitch angle can be negative, depending on definition of aerodynamic twist (the sum of the pitch + twist angles equals the angle between the chord and rotor plane). The twist is often defined such that "optimal" pitch is near zero degrees. It is common for the minimum blade-pitch angle to be saturated at the optimal pitch.
Jason.Jonkman wrote:Dear Kumara,

FAST uses the following definitions:
  • The structural twist is the angle (about minus z) from the blade coordinate system to the local principal elastic axes of bending at a given cross section for the undeflected blade.
  • The aerodynamic twist is the angle (about minus z) from the blade coordinate system to the chord at a given cross section for the undeflected blade.
  • The blade-pitch angle is the angle (about minus z) from the hub to the blade coordinate system and is constant along the blade.
So, the sum of pitch plus twist is the angle (about minus z) from the rotor plane to the appropriate local axis (principle or chord) for the undeflected blade.

I hope that helps.


Best regards,

Dear Jason
Here, "Aero twist = pitch angle + twist angle", I believe twist angle meant structural twist. But is the Pitch angle same as "The blade-pitch angle is the angle (about minus z) from the hub to the blade coordinate system and is constant along the blade."? (Defined by you in topic: Definition of parameters)
By the definition, Pitch angle must be the angle between local elastic axis and the chord. Am I correct?
In the document "Definition of a 5-MW Reference Wind Turbine for Offshore System Development" Aero twist and structural twist are assumed to be identical so that the pitch angle will be zero along the blade. So if I use this data can I ignore pitch angle in my calculations? or Can you please provide different values of Aero and Structural twist?

But Dear Jason in the book "http://www.innwind.eu/-/media/Sites/innwind/Publications/Deliverables/DeliverableD1-21ReferenceWindTurbinereport_INNWIND-EU.ashx?la=da&hash=14D9E77225A5F451D153082579B808E7C9E4C75C" in the section 4.1.3 third paragraph, the definition of structural twist angle is mentioned as the angle between principal elastic axis and chord. I am baffled. Please help me go through this.
Thanks in advance.

Jason.Jonkman
Posts: 4338
Joined: Thu Nov 03, 2005 4:38 pm
Location: Boulder, CO
Contact:

Re: questions about reference control input

Postby Jason.Jonkman » Thu Jan 31, 2019 4:35 pm

Dear Kamineni,

I've defined the structural twist, aerodynamic twist, and blade-pitch angle in the forum topic you quoted. I'm not sure how to define these any clearer.

I don't agree with your statement "Aero twist = pitch angle + twist angle" or your conclusion that pitch is zero. Instead, I would say that the angle (about minus z) from the rotor plane to the chord equals pitch + aerodynamic twist. Likewise, I would say that the angle (about minus z) from the rotor plane to the principle axis equals pitch + structural twist.

DTU uses the same definition I use, except that DTU defines the twist and pitch angles about positive z (so, you must flip the sign for use in FAST).

Best regards,
Jason Jonkman, Ph.D.
Senior Engineer | National Wind Technology Center (NWTC)

National Renewable Energy Laboratory (NREL)
15013 Denver West Parkway | Golden, CO 80401
+1 (303) 384 – 7026 | Fax: +1 (303) 384 – 6901
nwtc.nrel.gov

Behnam.Sanaei
Posts: 2
Joined: Wed Mar 06, 2019 11:30 pm
Organization: Sharif University of Technology
Location: Tehran

Re: questions about reference control input

Postby Behnam.Sanaei » Wed Jun 19, 2019 3:27 am

Dear Dr. Jonkman
I'm working on LPV modeling of NREL 5 MW wind Turbine. I started with a simple 2 DOF model(Generator,Drive Train). For evaluation of FAST v8 linear models I used a uniform wind speed with small wind speed step change but linear and nonlinear models don't comply. I have read all your hints on forum like NLin>=18, TLin=500, and considering one complete revolution.
I checked this problem with FAST v7 and something that made me more confused is that the operating point for this 2 DOF model in FAST v8 is:
wr=12.1 Betta=15.85052 wind=18
But, in FAST v7 for this 2 DOF model I have:
wr=12.1 Betta=14.75951 wind=18
Do you have any idea about the point where this difference stems from?
Thanks

Jason.Jonkman
Posts: 4338
Joined: Thu Nov 03, 2005 4:38 pm
Location: Boulder, CO
Contact:

Re: questions about reference control input

Postby Jason.Jonkman » Wed Jun 19, 2019 6:50 am

Dear Behnam,

Yes, I can't comment on your specific results, but I would expect some small differences in the steady-state pitch angle associated with a given mean wind speed and rotor speed between FAST v8 and FAST v7 due to differences in the BEM solution between AeroDyn v15 in FAST v8 and AeroDyn v13 in FAST v7.

FYI: We've been working with Envision Energy recently to introduce a steady-state (with controller trim) solution in OpenFAST. While we haven't yet released this capability, it will be very useful for calculating pertinent operating points (including controller trim) before linearization.

Best regards,
Jason Jonkman, Ph.D.
Senior Engineer | National Wind Technology Center (NWTC)

National Renewable Energy Laboratory (NREL)
15013 Denver West Parkway | Golden, CO 80401
+1 (303) 384 – 7026 | Fax: +1 (303) 384 – 6901
nwtc.nrel.gov

Behnam.Sanaei
Posts: 2
Joined: Wed Mar 06, 2019 11:30 pm
Organization: Sharif University of Technology
Location: Tehran

Re: questions about reference control input

Postby Behnam.Sanaei » Wed Jun 19, 2019 8:42 am

Dear Dr. Jonkman
Thanks for your response
I think 1 degree difference for blade pitch angle is remarkable.
In FAST V8, I calculated the operating points by means of baseline controller. I mean by choosing a steady wind, I ran the model with baseline and after 500 seconds all the states were in equilibrium(for 2 dof) model. Then I set the initial conditions for pitch angle equal to the calculated ones from previous run and started the linearization. But, linear models dc gains in comparison with nonlinear model for 0.5 m/s wind speed step change around a operating point is very smal. I mean generator speed change in FAST is four times the linear model output. I checked outputs during linearization and everything was ok at the time of linearization. Is there any required setting in input files before linearization?
I'm working on Spar Buoy Platform.

Jason.Jonkman
Posts: 4338
Joined: Thu Nov 03, 2005 4:38 pm
Location: Boulder, CO
Contact:

Re: questions about reference control input

Postby Jason.Jonkman » Wed Jun 19, 2019 9:17 am

Dear Behnam,

In general, I would expect the linear and nonlinear system should have similar behavior for small perturbations about the operating point for models that are set up the same. I don't see any problem with the approach from your overview, but the devil is in the details.

I would recommend upgrading to OpenFAST because there have been several issues with AeroDyn that have been fixed between FAST v8 and OpenFAST v2.

Best regards,
Jason Jonkman, Ph.D.
Senior Engineer | National Wind Technology Center (NWTC)

National Renewable Energy Laboratory (NREL)
15013 Denver West Parkway | Golden, CO 80401
+1 (303) 384 – 7026 | Fax: +1 (303) 384 – 6901
nwtc.nrel.gov

Cheng.Zhang
Posts: 57
Joined: Tue Nov 14, 2017 6:50 am
Organization: ECN
Location: france

Re: questions about reference control input

Postby Cheng.Zhang » Tue Jun 25, 2019 3:35 am

Dear Dr. Jason,

I made several kinds of sliding mode controllers of floating wind turbines with OC3 spar-buoy platform, and I calculated the DEL (TwrBsMxt, TwrBsMyt, TwrBsMzt, RootMxb1, RootMyb1) by using Mlife.
However, the DEL results of different controllers are similar.

For example, I compared the DEL of GSPI and ASTW (the controllers I designed), see the figure , you can find the normalized RMS and DEL of different performance index, as I said, the DEL of the two controllers are very close, it seems unreasonable.

I'm not sure whether the DEL results I got are correct, and I can't find anyone who can help me with that. So could you help check the .mlif file? Also, you can find the results I got in the Excel file, is it reasonable?

I very appreciate for your help!

Best regards,
Cheng

Code: Select all

-----  MLife version 1.0 Input File  -------------------------------------------
Test #02 Testing %12.4e formatting, NamesLIne=0, NumChans>0, WrDELsXLS=true
-----  Job Options  ------------------------------------------------------------
true              EchoInp           Echo input to <rootname>.echo as this file is being read.
false             StrNames          Use channel names following a "$" instead of numbers when specifying channels in this input file.
false             OutData           Output modified data array after scaling and calculated channels. (currently unavailable)
"%12.4e"          RealFmt           Format for outputting floating-point values.
"PI_SIM_18p15"           AggRoot           Root name for aggregate output files.
-----  Input-Data Layout  ------------------------------------------------------
5                 TitleLine         The row with the file title on it (zero if no title is available).
7                 NamesLine         The row with the channel names on it (zero if no names are available or are specified below).
8                 UnitsLine         The row with the channel units on it (zero if no units are available or are specified below).
9                 FirstDataLine     The first row of data.
0               NumChans          The number of channels in each input file.
ChanTitle    ChanUnits    Scale  Offset   PSF_Type  NumCols rows of data follow.  Title and units strings must be 10 characters or less.
-----  Calculated Channels  ----------------------------------------------------
0                 NumCChan          The number calculated channels to generate.
1234567890        Seed              The integer seed for the random number generator (-2,147,483,648 to 2,147,483,647)
Col_Title   Units    Equation       Put each field in quotes.  Titles and units are limited to 10 characters.  NumCChan rows of data follow.
-----  Load Roses  -------------------------------------------------------------
0                 NumRoses          The number of load roses to generate.
Rose Name     Units    Channel1       Channel2     nSectors
-----  Time and Wind Speed  ----------------------------------------------------
1                 TimeChan          The channel containing time.
2                 WSChan            The primary wind-speed channel (used for mean wind speed and turbulence intensity, 0 for none)
-----  Statistics and Extreme Events  ------------------------------------------
true              DoStats           Generate statistics of all the channels.
true              WrStatsTxt        Write the stats to a text file?
false             WrStatsXLS        Write the stats to an Excel file?
0                 NumSFChans        Number of channels that will have summary statistics generated for them.
1                 SFChans           List of channels that will have summary statistics generated for them.  Must number NumSFChans.
-----  Distributions -----------------------------------------------------------
false             UserDistrib         User defined distribution?  true = load user-specified distribution, false = only use Weibull wind distribution
2                 WeibullShape        Weibull shape factor. If WeibullShape=2, enter the mean wind speed for WeibullScale.
10                WeibullScale        Weibull scale factor. If WeibullShape<>2.  Otherwise, enter the mean wind speed.
3                 WSin                Cut-in wind speed for the turbine.
21                WSout               Cut-out wind speed for the turbine.
31                WSmax               Maximum  wind speed value for the wind-speed bins.
2                 WSMaxBinSize        Maximum width of a wind-speed bin.
0                 nDistribVars        Number of independent variables in the user-specified distribution, ignored if UserDistrib = false
""                DistribName         Filename of the user-supplied distribution table, ignored if UserDistrib = false
-----  Fatigue  ----------------------------------------------------------------
5                 nFatigueChannels  The number of rainflow channels.  Next six lines ignored if zero.
0.0               FiltRatio         The fraction of the maximum range of each channel used as a cutoff range for the racetrack filter.  Use zero for no filter.
630720000         DesignLife        Number of seconds in the design lifetime (20 years = 630720000 seconds).
1.0               Availability      Fraction of the design life the turbine is operating when winds are between Vin and Vout
true             BinCycles         Bin the rainflow cycles?
0.5               UCMult            Multiplier for binning unclosed cycles.  (0 discards, 1 counts as a full cycle)
true              DoShortTerm       Compute simple (unweighted) damage-equivalent loads and damage rates.
true              DoLife            Do lifetime-related calculations?
false             DoAggregate       Compute a DELs and a damage result based on an aggregate of all the input files (does not use the wind spped distribution)
false             WrShortTermTxt      Write short-term results to plain-text files?
true              WrShortTermXLS      Write short-term resultsto an Excel workbook?
false             WrLifeTxt         Write lifetime results to plain-text files?
true              WrLifeXLS         Write lifetime results to an Excel workbook?
1                 EquivalentFrequency          The frequency of the damage equivalent load (Hz)
true              DEL_AsRange       true = report DELs as a range value,  false = report as a one-sided amplitude
3                 DEL_Type          1 = fixed mean, 2 = zero mean, 3 = both
0                 GoodmanFlag       0 = no Goodman correction, 1 = use Goodman correction, 2 = compute results with and without Goodman correction
Channel#  NSlopes  SNslopeLst      BinFlag  BinWidth/Number  TypeLMF     LUlt    BinWidth not used when BinCycles is false. nFatigueChannels rows of data follow.  LUlt >> LMF
    57        1       10           BN        128.0      161       5700
    58        1       10           BN      128.0       33       5700
    59        1       10           BN      128.0      490      13680
    60        1       10           BN      128.0      490      13680
   61        1       10           BN      128.0      490      13680
1                 NumDELGroups
NChannels      ChannelList
"OUT" 5              1 2 3 4 5
-----  Input Files  ------------------------------------------------------------
1                 FileFormat         Format of input files.  1 = FAST ascii, 2 = FAST binary
1  1.1   1.3   1.5   1.7    (Weibull-Weighted Normal Operation: NumNormFiles, PSF1, PSF2, PSF3, PSF4)
"PI_SIM_18p15.out"
0  1.1   1.3   1.5   1.7    (Weibull-Weighted Idling: NumIdleFiles, PSF1, PSF2, PSF3, PSF4)
0  1.2   1.3   1.4   1.6    (Discrete Events: NumDiscFiles, PSF1, PSF2, PSF3, PSF4)2  1.1   1.3   1.5   1.7    (Weibull-Weighted Normal Operation: NumNormFiles, PSF1, PSF2, PSF3, PSF4)
==EOF==                             DO NOT REMOVE OR CHANGE.  MUST COME JUST AFTER LAST LINE OF VALID INPUT.

Attachments
PI_SIM_18p15_Lifetime.xlsx
(15.17 KiB) Downloaded 15 times
case4_del_rms.jpg
case4_del_rms.jpg (131.2 KiB) Viewed 1008 times

Jason.Jonkman
Posts: 4338
Joined: Thu Nov 03, 2005 4:38 pm
Location: Boulder, CO
Contact:

Re: questions about reference control input

Postby Jason.Jonkman » Tue Jun 25, 2019 7:32 am

Dear Cheng,

I'm not sure I know enough about your approach to offer a detailed response, but here are a few comments:
  • Your fatigue settings seem reasonable to me, except that I would normally expect different values of the Whöler exponent "m" for towers and blades. Normally, m is between 3 and 5 for steel towers and between 8 and 12 for composite blades; you are using m=10 for both.
  • Normally I'd recommend applying the Goodman correction so that differences in the mean loads show up in the DELs, but it is certainly simpler to not use the correction.
  • With only one time series, you could compare short-term DELs; lifetime DELs are useful when running many simulations at different mean wind speeds and different turbulence seeds.
  • If instead of comparing DELs between your controllers, you compare standard deviations (or RMS) of the loads, do you see the same trends? I would expect a similar trend in general, though the DELs tend to weight lower or higher amplitude cycles differently than a simple standard deviation would (depending on the value of "m").
Best regards,
Jason Jonkman, Ph.D.
Senior Engineer | National Wind Technology Center (NWTC)

National Renewable Energy Laboratory (NREL)
15013 Denver West Parkway | Golden, CO 80401
+1 (303) 384 – 7026 | Fax: +1 (303) 384 – 6901
nwtc.nrel.gov

Cheng.Zhang
Posts: 57
Joined: Tue Nov 14, 2017 6:50 am
Organization: ECN
Location: france

Re: questions about reference control input

Postby Cheng.Zhang » Wed Jun 26, 2019 1:42 am

Dear Dr. Jason,

Thank you for your reply!
I made some modification according to your comments,
    1. I used m = 4 for towers and m = 10 for the blade and made the same DEL calculation, the normalized DEL results hardly changed.
    2. Yes, I don't know how to use Goodman correction, so I didn't use it. If it is used, the results trend will change or not?
    3. Yes, I have only one time series, so the short-term DELs and lifetime DELs are the same.
    4. I calculated the STD of TwrBsMxt, TwrBsMyt, TwrBsMzt, RootMxb1 and RootMyb1. The normalized (GSPI = 1) results are [0.84, 0.91, 0.91, 1.00, 1.02 ], it is hard to say have similar trend with the DELs.

I still doubt the result, will Goodman correction help?
Which results are reasonable according to the RMS of platform and blade motion, DELs or STD?
Could I regard STD of the tower and blade moment as the reference of fatigue?

Thank you!

Best regards,
Cheng

Jason.Jonkman
Posts: 4338
Joined: Thu Nov 03, 2005 4:38 pm
Location: Boulder, CO
Contact:

Re: questions about reference control input

Postby Jason.Jonkman » Wed Jun 26, 2019 8:02 am

Dear Cheng,

The Goodman correction will not contribute much for a single time series. The Goodman correction includes the effects of the mean load on the fatigue calculation (the higher the mean, the higher the fatigue damage for a given cyclic amplitude). So, if you have many time series with different means, the fatigue damage will be more accurately calculated with the Goodman correction.

A single time series will likely not be sufficient for statistical convergence. Typically, many time series are needed across different operating conditions to properly calculate fatigue.

A simple comparison of standard deviation or RMS may be sufficient for your purposes. Fatigue damage or DELs are more accurate representations of fatigue because they consider the S-N curve in the calculation. But if you don't know the properties of the S-N curve (L_Ult and "m"), then there may be little use in calculating DELs.

Best regards,
Jason Jonkman, Ph.D.
Senior Engineer | National Wind Technology Center (NWTC)

National Renewable Energy Laboratory (NREL)
15013 Denver West Parkway | Golden, CO 80401
+1 (303) 384 – 7026 | Fax: +1 (303) 384 – 6901
nwtc.nrel.gov

Cheng.Zhang
Posts: 57
Joined: Tue Nov 14, 2017 6:50 am
Organization: ECN
Location: france

Re: questions about reference control input

Postby Cheng.Zhang » Wed Jun 26, 2019 8:13 am

Dear Dr. Jason,

Thank you very much for your detailed answer!
It's very helpful.

Best regards,
Cheng

KumaraRaja.Eedara
Posts: 26
Joined: Mon Sep 12, 2016 8:34 pm
Organization: IIT Bombay
Location: Mumbai

Re: questions about reference control input

Postby KumaraRaja.Eedara » Fri Jul 12, 2019 2:23 pm

Dear Jason,
I understand this part of reply from you completely, but, I have one more question:
Jason.Jonkman wrote:Dear Kamineni,

I would say that the angle (about minus z) from the rotor plane to the chord equals pitch + aerodynamic twist. Likewise, I would say that the angle (about minus z) from the rotor plane to the principle axis equals pitch + structural twist.

Best regards,

Can you please help me understand the sign convention followed in defining pitch in FAST. Whether the clockwise rotation (seen from root to tip of blade along pitch axis) of the blade is considered positive or negative? Thanks.

Regards,
Kumara

KumaraRaja.Eedara
Posts: 26
Joined: Mon Sep 12, 2016 8:34 pm
Organization: IIT Bombay
Location: Mumbai

Re: questions about reference control input

Postby KumaraRaja.Eedara » Fri Jul 12, 2019 3:00 pm

Dear Jason,

One more question, rather would like to confirm my understanding.

In the input file to FAST (AeroDynV14), for each blade, the 'AeroTwst' align with chord,structural twist etc,is specified at different blade sections. And I observed that 'AeroTwst' value decreases from root section to tip section (with almost zero for the tip section) of the blade.
Assuming pitch angle is zero, the chord of the tip section will lie in the rotor plane. Correct?
Chords of airfoil at other sections along blade are making the corresponding 'aerotwist' angle with the rotor plane. Correct? If yes,when we define 'aerotwst' of at a particular section of blade to be +10 degrees, is it the angle between 'rotor plane' and 'local chord' (about minus z) in the clockwise or counter-clockwise direction? Thanks.

Regards,
Kumara[/quote]


Return to “Controls”

Who is online

Users browsing this forum: No registered users and 1 guest