In 2011 American and German standards codes introduced the seismic load combination for wind turbine design. It accounts for the likelihood of an earthquake occurrence during the operational state.
The operational loads and the seismic loads are dealt with separately and then are combined by means of combination coefficients.
Referring to the ASCE/AWEA RP2011 pag.31, the suggested “best practise” combinations including earthquake are for an operational load equal to the greater of:
- loads during normal power production at the rated wind speed
- characteristic loads calculated for an emergency stop at the rated wind speed
Now, referring to “Wind Energy Handbook 2011” (page 202), I have a clear definition for the “characteristic loads calculated for an emergency stop at the rated wind speed”, which is:
Load case 5.1, Emergency shut down in combination with normal turbulence model for rated wind speed U_r ± 2m/s, 12 10-minute simulations with the characteristic load as the MEAN OF THE SIX LARGEST TEN-MINUTE EXTREMES.
But what about the “loads during normal power production at the rated wind speed”? Referring again to “Wind Energy Handbook 2011” (page 199), if I look at the Load case 1.1, I have to carry out a minimum of 15 10-minute simulations for a range of wind speeds (cut-in:2m/s:cut-out) and then extrapolate the extreme value, for the characteristic extreme loads. But for the seismic load combination, I just have to compute the simulations only for the rated wind speed and I am not looking for extreme loads. But I still have 15 simulations at the rated wind speed and I need to evaluate the characteristic loads? Do I take also in this case the MEAN OF THE SIX LARGEST TEN-MINUTE EXTREMES?
Or, would you suggest me to apply simultaneously aerodynamic and seismic loads with FAST using the seismic module and then follow the rules of EUROCODE 8? Here a minimum of 3 seismic simulation with independent accelerograms should be used and the maximum of the three responses is taken. Alternatively, for 7 accelerograms (7 simulations), the average values of response quantities is chosen.
Thanks for the help.
After a discussion with other experts, I would like to add few explanatory notes.
According to IEC 61400-1- Paragraph 11.6, the earthquake loading shall be superposed with operational loading that shall be equal to the higher of
a) loads during normal power production by averaging over the lifetime;
b) loads during emergency shutdown for a wind speed selected so that the loads prior to the shutdown are equal to those obtained with a).
These prescriptions are slightly different than the one from ASCE/AWEA RP2011 (see previous post).
However, few points of my previous post are now clear.
The Load Case 1.1 (IEC 61400-1 -7.4.1 page 36) refers to extreme loads and prescribes the statistical extrapolation with 50-years return period for the characteristic load. One must be careful not to confuse this characteristic load with the operational load for the seismic combination a).
A definition for the operational load is given in IEC 61400-1- Paragraph 11.6 :
"The seismic load evaluation may be carried out through time-domain methods, in which case, sufficient simulations shall be undertaken to ensure that the operational load is representative
of the time averaged values referred to above."
Here’s my own interpretation:
One may run a sufficient number of transient aerodynamic simulations under NTW conditions and take the MEAN OF THE EXTREMES. One may stick to the principle in Load case 1.1 and run 15 simulations. No extrapolation is required. The representative load will be then combined with the earthquake loads as in IEC61400-1 ANNEX C or with other reasonable methods.
For the combination b) one may follow the:
ASCE/AWEA RP2011 --> characteristic loads calculated for an emergency stop at the rated wind speed, 12 10-minute simulations with the characteristic load as the MEAN OF THE SIX LARGEST TEN-MINUTE EXTREMES
or IEC 61400-1 --> loads during emergency shutdown for a wind speed selected so that the loads prior to the shutdown are equal to those obtained with a); again running a sufficient number of transient aerodynamic simulations under NTW conditions and take the MEAN OF THE EXTREMES.
Please correct me, if I am wrong.
I hope it helps.
I am performing a seismic analysis of wind turbine considering soil-structure interaction (SSI) with a commercial finite element software. The implemented model is a beam with a distributed mass and a lumped mass on the top, representing the tower head. The soil-structure interaction plays a central role in my research and it is modeled with the boundary element method. My model is equivalent to the NREL 5MW Reference Model. I checked it by means of modal analysis and the dynamics of the two systems seems be in perfect accordance. I also carried out a seismic analysis with FAST-seismic and with my model, with the same input motion, and I obtained the same dynamic response.
According to the seismic load combination (see previous posts) I would like to carry out transient seismic analysis including the aerodynamic loads. For the aerodynamic loads I use FAST. The Idea is to extract the aerodynamic loads at the tower head and apply them to my own model, where also seismic base shaking + SSI effects are included.
If I run a normal turbulence analysis with FAST and I read the aerodynamic loadings collective at the tower head:
“YawBrFxp,YawBrFyp,YawBrFzn” - Tower-top / yaw bearing force
“YawBrMxp,YawBrMyp,YawBrMzn” - Tower-top / yaw bearing moments
and I apply them to my model at the tower head with the whole mass, I have the feeling that the presence of the mass will be counted twice. Or that the presence of the lumped mass influences the applied loadings, transforming them into large displacements. I tested it (with my model without SSI effects) and I get wrong results at the tower base, that is, in my model the internal force at the base are much higher. If I apply the FAST loadings to the tower top without mass, it works well and I get the right internal force at the base of the tower.
However, in order to preserve the modal response of my model for the seismic excitation I cannot remove the tower head mass from my own model. This would change its natural frequencies.
So how can I extract the load form fast, in order to be able to apply them to an equivalent system with a mass at the top of the tower?
Any suggestions? Or maybe any reference where I can read more about substructuring/condensation methods for dynamic analysis?
Thanks in advance
Yes, the yaw-bearing loads output by FAST include the applied aerodynamic loads, as well as the gravity and inertia loads of the nacelle and rotor. So, you would definitely be double-booking the gravity and inertia terms if your finite-element model includes the tower head mass.
Normally, if one is interested in the pure aerodynamic applied load I recommend that they time-average the FAST output (time averaging eliminates the inertial effect and the gravity load can be easily subtracted out). However, my guess is you want to preserve the time history of the aerodynamic load. Adding the applied aerodynamic loads as outputs is on our FAST development to-do list, but is not available in a current version. You could modify the source code to add them yourself if you need them now.
Do you need to preserve the “modal response” of your finite element model if the applied loads at the yaw bearing already include the acceleration-dependent inertia terms?
thank you very much for your answer. Yes, I am interested in keeping the time history of the aerodynamic loads, which is important for detecting the influence of the soil-structure interaction in time domain. I think I have to keep the tower top mass in order to get the right natural frequencies of the structure, which will govern the seismic response. For example the natural frequency of the 5MW wind turbine should be about 0.32 Hz, including the top mass. It is about 0.88 Hz without top mass. This would give me completely different seismic responses. Of course the mass and its effects are included inside the time history of the aerodynamic loads, but I am not that this would be enough. I did a small check and I attached here the results. Please have a look and tell me your opinion. To me it seems that, taking off the mass and simply applying the aerodyn. loads, the tower base forces are ok. However the tower top forces are not as expexted.
I hope I am not talking nonsense
I hope you find the example relevant.
Thank you very much.
Example.pdf (133 KB)
I’m confused by your attachement. A couple questions:
*Does System 2 have a seismic displacement applied? The figure on page 2 does not indicate so, but the results seem to imply that it does.
*How is it possible that the applied top loads (force/moment) do not equal the output top loads in system 2? That is, why doesn’t F_ouput_0 = F_output_2? Is F_output_2 not actually at the top?
If you wish to remove the tower head mass from your FE model, I would think you would need to derive the tower-top loads from a model that includes the seismic displacement. That is, if you take F_output_1 and apply it as the load to system 2 with seismic diplacement, shouldn’t the response agree between system 1 and system 2 because F_output_1 should equal F_output_2 and the seismic displacement is the same?
*Yes, the system 2 has seismic displacement applied, and this is the reason why F_ouput_0 ≠ F_output_2. I forgot to draw the seismic displ. on the figure.
- F_output_2 is at the top. In the system 2 I applied both the F_output_0 and the seismic displacements and this is why the response of the System 0 at the tower top is different than the one in System 2. In the system 0 there is only the “aerodynamic load” applied.
My idea was to calculate the tower top loads (F_output_0) with FAST without seismic loads (similar to System 0) and apply it to my ANSYS model (Similar to System 2), where I apply also the seismic loads (and later on the soil-structure interaction). Why I do so? Because I cannot compute the aerodynamic loads directly with ANSYS. Why do I not use directly FAST? Because I am using boundary elements for the soil, which are the focus of our research and they are implemented inside ANSYS. I may use FAST_seismic for computing the tower-top loads from a model that includes the seismic displacement, as you suggested. I investigated this case and it is attached here as a continuation of the example (attempt 2).
Please find attached the corrected and extended version of my example, also with the case of keeping the mass at the top and applying the aerodynamic loads(attempt 3).
It seems that attempt 2 gives a better result than attempt 1 with respect to the top forces. However, the two systems still response in a different way because of the presence of the mass.
With this example I wanted to check the equivalence of the FAST ans ANSYS models, before moving to the simultaneous analysis with aerodynamic loads, SSI and seismic loads. For normal operational conditions and emergency stop, I was wandering from which model should I take the aerodynamic loads:
*FAST model with wind and seismic loads
*FAST model with wind
*a modified FAST model with wind without gravity
For now, after discussing and checking various issues, my guess would be the third possibility.
Sorry for stealing your time. I hope it is an interesting issue.
Thanks for your help.
Example.pdf (302 KB)
I guess I’m still confused a bit. In “attempt 2”, you take the top loads from a model with a head mass (F_output_1) and apply it to an identical model without the head mass, correct?. The top loads include the effect of the head mass, so why would system 2 not match system 1 exactly and why does the load applied to system 2 (F_output_1) not equal the top load measured from system 2 (F_output_2)?
Eliminating gravity from the FAST model won’t eliminate the inertia load from the moving head mass.
Ideally, you’d want the pure aerodynamic applied load derived from FAST to apply to your ANSYS model. As I mentioned above, you should be able to modify the FAST source code to output these. The aerodynamic loads are calculated within FAST, they are just not yet output directly to a file.
You’d should also realize that the aerodynamic loads depend on the motion of the structure, so, will be different for a case with and without seismic motion.
It’s just a thought, but maybe you could run FAST twice. Use a completely rigid turbine and run once with aerodynamics and once without. Subtracting the two time histories will give you a (maybe) reasonable approximation of the aerodynamic load. Obviously, it will not include the aeroelastic response (there will be no effect on the aerodynamics due to turbine vibrations).
BTW, I am currently rewriting the AeroDyn I/O processing and you will be able to get aerodynamic loads directly once I am done. Probably in a month or two if I can stay on task.
It is correct. In attempt 2 I took the top loads from a model with a head mass (F_output_1) and apply it to an identical model without the head mass. In both cases I had seismic load as well. Why it is not different? I think because in the model there is not only the head mass but also a distributed mass along the tower. If there was the only head mass the output was definitely the same.
Maybe this example is not really helping that much. I am drifting from the problem.
You are right, it is not about gravity but inertia of the whole tower head.
Regarding the influence of motion of the structure on the aerodynamic loads, in order to make my analysis possible, I make a main assumption, that the aerodynamic loads are computed for a fixed base tower without seismic loads. I know, this is not true, the loads depends on the vibration of the tower (which in turns depends on the SSI). It is an approximation. Do you believe, that taking the aerodynamic loads without considering the tower vibrations and the SSI effects, will lead me to intolerable miscalculations? The focus of the whole research is on the influence of layered soil on the dynamic response of wind turbine during seismic analysis.
About modifying the code I probably wait for your version, avoiding mistakes.
Thank you very much for the help
thanks for your idea. I will try this option as well. I know, it is going to be an approximation. It is going to be anyway an approximation, as the vibrations are also affected by the layered soil, and this aspect would miss. In FAST I could only model the homogeneous half space through spring,mass and dashpot, but I cannot consider several layers over half space.
Otherwise, I will just consider seismic loads and soil-structure interaction effects, avoiding to extend the analysis to the aerodynamic loads, in order to avoid senseless conclusions.
I have to check how big will be the error of such an approximation.
Looking forward to the new AeroDyn processing.
Thanks again for the suggestion.
Deriving the aerodynamic loads independent of the seismic motion may be OK, but it probably depends on what you are after and the magnitude of the seismic load.
I would guess for large earthquakes, the seismic load would be much larger than the aerodynamic load. For small earthquakes, the reverse is probably true.
However, one contribution you are probably trying to get from the aerodynamic load is the aerodynamic damping from motion of the support structure, which may mitigate (at least somewhat) the seismic-induced loads. However, if the aerodynamic load is derived in the absence of the seismic load, you won’t get this aerodynamic damping.
I’m beginer of FAST software.
When I run the provided Seismic examples executing FAST_iwin32.exe, it aborts after input phase. The message told me “GBRatio must be set to 1.0 when using Kirk Pierce’s UserVSCont() routine”.
Could you give me some ideas to solve this problem. It is really difficult for me.
I looking forward to hearing from you.
It looks like the executable you are using doesn’t have the Bladed-style DLL interface, which the NREL 5-MW turbine uses. Where did you get the FAST executable FAST_iwin32.exe? The version of FAST provided in the Seismic archive (nwtc.nrel.gov/Seismic) is named FAST.exe, which is a version of FAST v7.02.00d-bjj compiled using Intel® Visual Fortran v11.0.074 with the seismic source code replacing the generic UserPtfmLd routine and with the Bladed-style DLL interface replacing the generic variable speed and blade-pitch controller.
I am performing a seismic analysis of wind turbine considering soil-structure interaction (SSI) with a recompile FAST version.
The version of FAST provided in the Seismic archive (nwtc.nrel.gov/Seismic), which is a version of FAST v7.02.00d-bjj compiled using Intel® Visual Fortran v11.0.074 with the seismic source code replacing the generic UserPtfmLd routine.
If I want to execute time history input about seismic acceleration on the seabed, which flexible-foundation type (coupled springs or distributed springs model) is more reasonable compared with actual situation?
I’m not really sure I understand your question, but the Seismic version of FAST v7.02 effectively uses the coupled springs model of the foundation, with the properties defined by an actuator frequency and damping. Modeling a different type of foundation (e.g. distributed springs via p-y curves) would recompile a change to the source code.
How do I define the actuator frequency and damping if the Seismic version of FAST v7.02 can use the coupled-spring model of the foundation with the properties defined by an actuator frequency and damping effectively?
I try to recompile the source code include CS-spring (OC3 phase II soil property ) and seismic module but I think I get a wrong result (the soil effect is larger than seismic effect).
Can you give me some suggestions about my attachments?
Thanks for your help.
attachment.docx (90.6 KB)
UserSubs_forBladedDLL - Seismic - CS.txt (8.39 KB)
Unchanged, the Seismic module of FAST v7.02 only applies to platform-translation DOFs. Without changing the source code, you could derive ActFreq to give the same stiffness as the horizontal translation coupled springs stiffness matrix, kTrans = TotalMass*ActFreq^2 = Stff(1,1), or ActFreq = SQRT( Stff(1,1)/TotalMass ). I would use the documentation’s recommendation of ActDamp = 60 to 70%. But you’ll have to modify the code if you want to use the complete 6x6 coupled springs stiffness matrix and if you want the earthquake to directly excite more than just the translational modes.
How do I model a different type of foundation (e.g. distributed springs via p-y curves) with earthquake if I use the Seismic version of FAST v7.02 ?
I can change the source code (e.g. distributed springs via p-y curves) and recompile it successfully.
However, I don’t know how to define the numerical modeling to simulate the seismic effect with p-y curve.
Do you give me some suggestions?
Thanks for your help.
I’m not sure I fully understand your question, but without getting into the details, my guess is you’d want to do the following. First, from the existing Seismic code, use the section of the code that determines the specified ground X-Y displacement (motion), but neglect the section that calculates the platform force required to achieve that motion. In your p-y implementation, define the displacement (y) as the relative difference between the actual tower X-Y positions and the displaced motion of the ground in the presence of the earthquake (the later would be zero in the absence of the earthquake).