Blade mode shapes with BModes

Hallo,

Currently I try to get an idea on how to get the mode shapes needed in FAST. Therefore I am trying to recalculate the mode shapes of the NREL 5MW baseline windturbine which are given in the CertTest input files of the FAST distribution. I read through some topics in this forum, so I know that these modes shapes were calculated by using ADAMS. Nevertheless I thought that it should be possible to calculate at least an approximation of these mode shapes also with BModes, but unfortunately I wasn’t successful with that. My results show quite big differences from that given in the CertTest input files. See this picture:

As I was not able to recalculate the CertTest mode shapes I am now quite unshure whether I took the correct input parameters and what input parameters I should use in future when I have to calculate the mode shapes of other blades. Therefore it would be nice to get some further advices on these parameters from an expert.

What I found so far are two older posts by Jason Jonkman which gave the following suggestions on the input parameters when deriving blade mode shapes:
#1 (from Tower Eigenfrequencies of NREL 5MW Turbine)

and #2 (from NREL 5MW Blade Mode Shape)

Unfortunately they are different concerning the suggestions for str_tw and tw_iner. As far as I understand FAST expects uncoupled modes so the suggestion of #2 should be more appropriate, isn’t it?
As a “very high number” I took 1.00E+10 and as “very small number” I took 1.00E-10. Do you think they are high/small enough?

I know, calculating the mode shapes is not the easiest task but correct mode shapes are quite important to get realistic simulation results. So I hope you can help me a bit in this issue.

Kind regards,
Paul Schünemann

Dear Paul,

Good question.

In reality, both methods have some problems. The blade mode shapes would be more accurately predicted by BModes using method #1 than method #2, but as you point out, the structural model of FAST v7 or the ElastoDyn module of FAST v8 expects uncoupled blade modes, which FAST then couples through the structural pretwist, so, the use of method #1 is not completely consistent with what FAST expects. Using method #2 is more consistent with what FAST expects, but the mode shapes predicted by BModes will not be as accurately predicted because the lack of the coupling will have an influence on the effect of centrifugal stiffening.

The ADAMS-derived mode shapes for FAST for the NREL 5-MW blade were based on method #1. Which method are you using in BModes?

From my experience, both methods have yielded adequate results as the structural response predicted by FAST is not too sensitive to the specified blade mode shapes, as long as the natural frequencies are predicted accurately. I’d be curious if you arrive at a different conclusion.

Best regards,

Dear Jason,

first of all thank you very much for your helpful and as always amazingly fast answer.

In fact, I tried both methods (#1 and #2) and as you stated the resulting mode shapes are indeed very similar, which can be seen on the following figures. There you see the mode shapes calculated with ModeShapePolyFitting.xls from the results of BModes. For the #1 cases I input only the flap disp generated by BModes and ignored the lag disp.
However as I mentioned in my last post all cases I tried differs from the mode shapes I extracted from the CertTest input files. That’s because I’m so unshure about this thing.

The natural frequencies derived from BModes (see next picture) are also not really sensitive to the used method but quite sensitive to the used rotor speed (romg) in the BModes input file. However as far as I understood that’s not suprising.
NaturalFrequencies.JPG
In your post you mentioned that for FAST it is important to accurately predict the natural frequencies. Unfortunatly I don’t know the corresponding natural frequencies to the mode shapes of the CertTest input file. The only values I found concerning the NREL 5MW turbine are given in Table 9-1 of http://www.nrel.gov/docs/fy09osti/38060.pdf, but I’m not shure with which value I should compare my results. Maybe you can help me with that and give me a hint whether you think my results are accurate enough.

Last but not least: Until now I did not do any FAST simulations to investigate the sensitivity of the use of different mode shapes. What would be a good output channel to see the influence of the different mode shapes? Maybe the blade deflection (OoPDefl1 and IPDefl1)?

Sorry for that long post and thanks for any help in advance.

Kind regards,
Paul Schünemann

Dear Paul,

I’m surprised by the difference you are seeing between the BModes-generated mode shapes and the mode shapes from the CertTest. Are you sure you’re plotting the mode shapes from the CertTest correctly? I plotted those mode shapes myself, and they are much closer to the results you are showing from BModes than the results you are showing from the CertTest. See attached.


Table 9-1 from the NREL 5-MW specifications report lists the full-system natural frequencies of the stationary wind turbine. So, these results should be most consistent with the 0 rpm results from BModes, but I would expect some differences as Table 9-1 involves full-system modes i.e. the blade modes may have some coupling with the support structure. If you are using the latest version of FAST v8, you can use the new linearization functionality of FAST to derive the full-system natural frequencies of a stationary or spinning wind turbine.

From my first comment, I suspect the BModes and CertTest mode shapes are actually quite consistent, so, there is probably no need to make a sensitivity comparison between the mode shapes. However, if you still wanted to do this, I would suggest comparing blade-tip deflections and blade-root loads for a few operational cases between cut-in and cut-out wind speeds.

Best regards,

Dear Jason,

you are completly right. Encouraged by your post I checked my caluclation of the mode shapes in MS Excel and indeed there was a mistake (i.e. I took the wrong cell (x^2 instead of x) for calculating the values of the polynomial). After correcting this mistake the mode shapes now match quite well. See the two attached figures if you are interested. Thank you very much for guiding me to this error.
All in all that helped me a lot and I really feel more confident with the calculation of blade mode shapes now.

So next will be the calculation of the tower mode shapes. I already had a short look into that but stucked on how to deal with the tower top mass. However I already found some other topics in the forum concerning this issue and I will read them again and post any questions there if I can’t solve the problem for myself.

Concerning the full-system natural frequencies I will follow your advice and use the linearization functionality to have a look on which frequencies I derive. I’m curious whether I get similar results to that in Table 9-1.

Again thank you very much for all your ideas and advices.

Kind regards,
Paul Schünemann


Dear Jason,

I’m a bit confused about the mode definitions in FAST and BModes. In the BModes user’s guide the flap displacement is described as a displacement normal to the reference plane BB. So the flap displacement is an out-of plane displacement. But for FAST I have to use the flap displacement from BModes to calculate the poynomial coefficients for the flapwise mode shape. According to the FAST v6 user’s guide the flapwise mode shape is “defined with respect to the local structural twist”. So it is normal to the local chord of the blade and “the tip will deflect in both the in-plane and out-of-plane directions due to a pure flapwise deflection”.

So why can I use the BModes outputs for FAST? Why can I use an out-of plane mode from BModes as a flapwise mode for FAST?

Thank you again for your answer.

Best regards,
René

Dear René,

Normally the blade structural twist is set to zero in BModes when obtaining mode shapes for FAST. When the resulting mode shapes are specified in FAST, along with the proper structural twist, the flapwise and edgewise mode shapes get coupled. This is discussed more in the following forum topic: http://forums.nrel.gov/t/coupled-blade-modes-in-fast/314/1. Obviously there is some approximation here, but from my experience, the result is quite accurate.

Best regards,

Dear Jason,

Thank you for clarification of my understanding of the approximation.
I have already read the other topics and the discussion of using structural twist in BModes or not.

Best regards,
René