Optimized Design of Wind Turbine Jacket Foundations: Improved efficiency and accuracy in the sequential integrated approach
Author
Augustyn, Dawid Jakub
Term
4. term
Education
Publication year
2016
Submitted on
2016-06-09
Pages
90
Abstract
Fundamenter til moderne havvindmøller designes i dag med en såkaldt Sequential Integrated-tilgang, hvor to parter samarbejder: vindmølleproducenten og fundamentsdesigneren. De udveksler data via forenklede (reducerede) og lineariserede computermodeller. Kvaliteten af det endelige design afhænger af, hvor godt disse modeller beskriver den virkelige konstruktion – herunder konstruktionens egen bevægelse (intern dynamik), bølgelaster og jordens opførsel. Afhandlingen vurderer den nuværende praksis og introducerer metoder, der kan øge både nøjagtighed og effektivitet. Først gennemgås teorien bag modelreduktion, herunder de udbredte Guyan- og Craig-Bampton-metoder samt den nyere Augmented Craig-Bampton-metode. Derudover beskrives jordmodelleringsmetoder, også ikke-lineære tilgange som fx Winkler-fjeder-metoden, og deres begrænsninger fremhæves. Derefter undersøges tre reducerede fundamentmodeller, tårnets modalrepræsentation og ikke-lineær jordmodellering i forhold til en detaljeret (ikke-reduceret) referencestruktur. Standard lastberegninger udføres med en højdetaljeret jacket-struktur i Rambølls in-house program ROSAP kombineret med reference-turbinen NREL 5 MW i Rambølls aero-elastiske kode. Resultaterne viser, at Guyan-metoden ikke beskriver de interne afstivningers dynamik tilstrækkeligt, og at Craig-Bampton i nogle tilfælde ikke effektivt fanger effekten af interne bølgelaster. Den mere robuste Augmented Craig-Bampton-metode leverer derimod samme nøjagtighed som referencesystemet. Kombineres den med den effektive Direct Expansion recovery-run-procedure, forbedres både beregningseffektiviteten og nøjagtigheden markant. Tårnets repræsentation i den aero-elastiske kode FLEX5 undersøges også. Koden anvender kun to bøjningsmodi for tårnet, hvilket sammenlignet med en ikke-reduceret tårnmodel giver en betydelig fejl. Et følsomhedsstudie viser, at ca. ti interne modi bør medtages for at beskrive tårnets stivhed korrekt. Endelig sammenlignes linearisering af jordens opførsel med en ikke-lineær jordmodel. Forskellene er store, især ved store forskydninger i jorden. To løsninger foreslås: Den mest præcise og robuste er at udelade pælene fra den lineære, reducerede jacket-model og implementere den ikke-lineære jordbeskrivelse eksternt. Dette giver samme resultater som den ikke-reducerede jacket med ikke-lineær jord, men kræver ekstra implementering i den aero-elastiske kode. For at omgå dette foreslås en brugerspecificeret lineariseret jordtilgang, som kan estimere den ultimative pæleforskydning, men som forudsætter geoteknisk ekspertise.
Foundations for modern offshore wind turbines are typically designed using a so-called Sequential Integrated approach, in which a wind turbine manufacturer and a foundation designer work together. They exchange information via simplified (reduced) and linearized computer models. The final design’s accuracy depends on how well these models capture the real structure, including its internal dynamics, wave loading, and the seabed’s behavior. This thesis evaluates the current practice and introduces methods to improve both accuracy and efficiency. It first explains the theory behind model reduction, covering common methods such as Guyan and Craig-Bampton, as well as the newer Augmented Craig-Bampton method. It also reviews soil modeling approaches, including nonlinear methods like the Winkler spring method, and highlights their limitations. Next, three reduced foundation models, the tower’s modal representation, and nonlinear soil modeling are assessed against a detailed (non-reduced) reference structure. Standard load calculations are carried out using a high-fidelity jacket structure in Rambøll’s in-house ROSAP software and the reference NREL 5 MW turbine in an aero-elastic code. The results show that the Guyan method does not adequately capture internal brace dynamics, and that Craig-Bampton can, in some cases, miss the effects of internal wave loading. In contrast, the more robust Augmented Craig-Bampton method achieves the same accuracy as the reference solution. When combined with an efficient Direct Expansion recovery-run procedure, it significantly improves both computational efficiency and accuracy. The tower model in the aero-elastic code FLEX5 is also examined. The code currently uses only two tower bending modes; comparison with a non-reduced tower model reveals a significant error. A sensitivity study indicates that about ten internal modes are needed to represent the tower’s stiffness adequately. Finally, a linearized soil model is compared with a nonlinear soil representation. The results differ substantially, especially at large soil displacements. Two solutions are proposed: The most precise and robust is to exclude the piles from the linear, reduced jacket model and implement the nonlinear soil model externally. This matches the non-reduced jacket with nonlinear soil but requires additional implementation in the aero-elastic code. To avoid that change, a user-defined linearized soil approach is proposed, which can estimate ultimate pile displacement but requires geotechnical expertise.
[This abstract was generated with the help of AI]
Keywords
Documents