Fatigue postprocessing with MLIFE: damage benchmarking

Wind & Water Computer-Aided Engineering Software Tools
1

Hi!

First of all, I’d like to congratulate all of you for providing a great value for the wind energy community. I’m following you for a while, but this is my first post in the forum.

I’m working in my PhD related to Robust Wind Turbine Control design, and I’m quite familiar with control design and aeroelastic simulation. My previous research showed me that most work done in control follows similar tactics:
1- select a reference wind turbine aeroelastic model, wind profiles and a baseline controller
2- apply whatever advanced control methodology to design a study controller
3- run a set of FAST simulations for both controllers in the same conditions
4- compute equivalent loads (MLIFE) and compare them.

In my case, I’m trying to deal with dynamic problems given by the wind turbine design that are difficult to tacke with classic control theory, by using robust control techniques. In order to do it, I’m modifying mechanical properties of the reference wind turbine. Once I obtain a realistic mechanical design where a baseline controller does not perform good enough, I design a robust controller and I compare them following the typical approach.

I’m having good diferential resutls comparing equivalent loads, nevertheless in my first seminar presenting the results, I had some critics on the methodology targeting the results comparison. For example,
1- using the onshore NREL 5MW 90HH I change mass distribution in the blades or tower, to emulate split blades, hybrid towers, etc.
2- I simulate fatigue cases with a baseline controller, obtaining worse equivalent load results compared to reference wind turbine with the same controller.
3- I simulate fatigue cases with a robust control design, obtaining better equivalent load resutls compared to 2- but never as good as 1-. Because of the mass increments, I think it is impossible to reach 1- values.

In order to provide a more satisfactory comparison, I thought on comparing absolute damage values for components. If the reference wind turbine component is lasting +20 years, and the modified component does not last 20 years with the baseline controller, but it will with the controller under study, then I have a feasible solution, even when it is not as good as the reference one.

My problem now is that I’m not capable of doing a post-process where the reference wind turbine last longer than 20 years in any of the components under study. I iterated the wind using non aggressive class IIIb profiles with low mean Weibull distributions, and used some baseline controllers and an improved one, obtaining smooth and realistic time-series. In order to compute damage, I followed some of the posts in this forum to obtain ultimate load values using cylinder models for blade root or tower base. Numbers seem to match with calculations from other users.

I searched previous publications looking for someone that provides a full fatigue study with damage calculation for the reference NREL 5MW 90HH onshore wind turbine, but I found nothing relevant.

Does anybody know any publication with a fatigue damage study using NREL 5MW, FAST and MLIFE? I’d like to follow it to find my errors. I think that something should be already published, but I’m not finding anything.

Do you think that there is another smarter or more practical way of comparing my resutls?

Kind regards,

 Jesús
2

Dear Jesús,

The fatigue loads for the land-based version of the NREL 5-MW turbine are calculated as damage-equivalent loads (DELs) for a range of material exponents (m) and ultimate load factors (ULF), as described in the calculation process documented in section 4.3 of this report: nrel.gov/docs/fy10osti/45891.pdf. Is this what you are looking for?

Best regards,

3

Dear Jason,

Thank you very much for your fast response. The content in the section 4.3 of the linked report is similar to the process that I’m following, and it is very well explained by the way. I read before this approach for computing the ultimate strength by obtaining the maximum load from ultimate load cases, and multiplying it times an increasing ultimate load factor, chosen when it asintotically converges to a constant value (Figure 34 in the report).

When we assume that the wind turbine components, wind profile, controller etc. are designed all together this approach may be valid, but I’m not sure if it is also valid for independent component design. Let me explain myself: in the linked report I assume that the design is done to last +20 years, then I obtain the ultimate strenght that matches it. But when I am modifying components, for example, I add a couple of tons at the middle of the blade to model a bolted joint, I guess that the ultimate load value from design load cases will be higher, and if I apply the same rule ultimate strength will be also higher, even when I did not change the blade root. This is just an assumption, I’ll calculate it, but in this example case ultimate strength for the blade root is the same with or without the bolted joint.

Following the same example, in case I change the geometry of the blade root I don’t know if my design is lasting 20 years, and if I compute ultimate strength via simulation I think that I’m asuming it as granted.

An alternative option that I found in this forum is to compute the ultimate strength with the properties of the material and geometry. Asuming a cylinder. This is what I tried but final damage value was very high, even when timeseries seemed smooth.

One thing I can do is to compare those ultimate strength values from the linked report and calculated from the material and geometry properties, to find if there is a big difference. Even when the numbers are not in the report, I can work with the variable presented as example in the Figure 34.

If you find some missintrepretation in my assumptions, please let me know. I’m really greatful for your support.

Kind regards,

 Jesús
4

Dear Jesús,

I’ll just clarify that the calculation of the ultimate strength would ideally be based on the geometry and material of the cross section. The calculation is quite simple for simplified cross sections such as cylinders made of isotropic material, but requires special cross-sectional analysis for complex cross sections and composites. See, e.g., the discussion here: Mlife - User Defined Distribution - #15 by Jason.Jonkman. In this case, if you’re analyzing the fatigue of the blade root but change the design of the blade mid-span (without changing the design of the blade root), the ultimate strength of the blade root would not be influenced by the design change.

In conceptual design, often the 1D beam properties are known, but the cross sectional geometry and material are not known (such is the case for the NREL 5-MW wind turbine). In this case, one can model the turbine in OpenFAST, but can’t compute the ultimate strength for the fatigue post-processing. In this situation, we have assumed that the ultimate strength can be derived by scaling up the ultimate load found from OpenFAST simulations (assuming that the cross section is designed sufficiently to not fail when the ultimate load from OpenFAST is reached), as is done in the report I linked above.

I hope that helps.

Best regards,

5

Dear Jason,

Your advice is really helping. Perhaps I’m doing it too complex, and as I’m closer to a conceptual design rather than a detailed design, this approach could be enough. I can use the report that you provided as a reference. I’m about to do some numbers with this approach to see if it fits.

Thank you very much,

 Jesús