Wind energy is one of the most promising renewable energies available today. The continuous demand of wind energy production led the interest of the wind industry towards bigger turbines. This upscaling trend has imposed bigger (not quasi-static) loads that are significantly influencing the fatigue life of the wind turbine components and the tonalities generated by the drive train. To tackle noise and vibration problems and validate complex models it is of high interest to continuously track the modal parameters of the machines during operating conditions. To this purpose, operational modal analysis (OMA) represents a powerful tool. The work investigates and implements a completely automated operational modal analysis technique for continuously tracking the modes of a wind turbine drive train. This goal is achieved by implementing the state-of-the-art PolyMAX algorithm in a programming language capable of being coupled with machine learning algorithms and big data storage systems. In order to make it suitable for continuous analysis, the algorithm is improved by eliminating all the required computer-human interactions and adding a method that automatically tracks the modal parameters along different data sets. The work deals also with the problem of the application of OMA on rotating machines: the presence of harmonics. To this purpose, two different procedures are analyzed and implemented: the use of a cepstrum lifter and the use of order based modal analysis. The implemented procedure is used to process real data coming from a long-term monitoring campaign on an offshore wind turbine in the Belgian North Sea. Data sets coming from the turbine in different operating conditions are processed in order to take into account all the topics involved in the thesis. Interesting results are obtained. First of all it has been shown that when used on stand still data (i.e data acquired while the turbine is not producing energy) the implemented automatic procedure is able to detect modes close to narrow harmonic components by itself, without the need of pre-processing steps. Unfortunately it still fails when the harmonics influence a broader frequency band. Secondly the results obtained from the data coming from the rotating turbine and processed with the two methodologies proposed for dealing with the harmonics are compared with the results obtained from the stand still turbine. It has been shown that the order based modal analysis gives results more compatible with the results coming from multibody simulations available in literature. The latter result shows the power of the work implemented. This automatic modal parameter estimation that can be performed on long-term data can be implemented in a new concept of design: the simulated model can be integrated with the information learnt during the life time of the previous prototypes in order to have more robust design of the components.
L'energia eolica é una delle energie rinnovabili piú promettenti disponibili al giorno d'oggi. La continua domanda di produzione di energia ha condotto l'interesse dell'industria eolica verso turbine sempre piú grandi. Questo fenomeno, chiamato up-scaling trend, ha imposto maggiori (non quasi-statici) carichi che influenzano significativamente la vita a fatica dei componenti delle turbine e le tonalitá generate dagli organi di trasmissione. Per affrontare problemi di rumore e vibrazioni e validare i modelli, é di grande interesse tracciare continuamente i parametri modali delle turbine durante condizioni di normale funzionamento. A questo proposito l'analisi operazionale modale rappresenta un potente strumento. Il lavoro svolto implementa e investiga una versione completamente automatica dell'analisi operazionale modale al fine di tracciare continuamente i parametri modali dell'albero a trasmissione di una turbina eolica. Questo obiettivo é raggiunto implementando l'algoritmo PolyMax, cosí come formulato in letteratura, in un linguaggio che lo renda abbinabile ad algoritmi di machine learning e processazione di grandi quantitá di dati. Per rendere questo strumento adatto ad una continua processazione di dati, l'algoritmo é stato modificato, eliminando tutte le possibili interazioni richieste con l'utente e aggiungendo un algoritmo che permetta di tracciare i parametri modali ottenuti in diversi set di dati. Il lavoro tratta anche un secondo problema: l'uso dell'analisi operazionale modale su macchine rotative. Il problema in questione é la presenza di armoniche nel segnale di input. A questo proposito, due diverse procedure sono analizzate e implementate: l'uso dell'analisi cepstrale e l'uso di order based modal analysis. La procedura implementata é stata usata per processare dati reali provenienti da una turbina eolica off-shore installata nel mare del nord, a largo delle coste Belga. Dati provenienti da diverse condizioni di funzionamento della turbina sono stati utilizzati in modo da considerare tutte le problematiche analizzate durante la tesi. Prima di tutto é stato mostrato come la procedura automatica di analisi modale usata su dati provenienti dalla turbina in condizioni stazionarie (ovvero quando la turbina non produce energia) é in grado di rilevare modi molto vicini ad armoniche, senza il bisogno di pre-processare i dati, quando le ultime influenzano una banda di frequenza molto ristretta. Sfortunamente la procedura fallisce quando le armoniche influenzano una banda di frequenza piú ampia. In secondo luogo, i risultati ottenuti dai dati provenienti dalla turbina rotante e processati con le due metodologie implementate per risolvere il problema delle armoniche sono stati analizzati e attraverso una comparazione tra gli ultimi e i risultati provenienti dalla turbina in condizioni stazionarie, é stato mostrato che l'uso di Order Based Modal Analysis fornisce risultati piú compatibili con quelli ottenuti da simulazioni trovati in letteratura. L'automatica stima di parametri modali analizzata puó essere sfruttata per implementare un nuovo concetto di design: i modelli derivanti da simulazioni possono essere integrati con le informazioni imparate durante la vita produttiva dei prototipi precedenti per avere una progettazione basata su informazioni piú realistiche.
Continuous modal behaviour estimation of an offshore wind turbine drive train in the presence of harmonics
GIOIA, NICOLETTA
2016/2017
Abstract
Wind energy is one of the most promising renewable energies available today. The continuous demand of wind energy production led the interest of the wind industry towards bigger turbines. This upscaling trend has imposed bigger (not quasi-static) loads that are significantly influencing the fatigue life of the wind turbine components and the tonalities generated by the drive train. To tackle noise and vibration problems and validate complex models it is of high interest to continuously track the modal parameters of the machines during operating conditions. To this purpose, operational modal analysis (OMA) represents a powerful tool. The work investigates and implements a completely automated operational modal analysis technique for continuously tracking the modes of a wind turbine drive train. This goal is achieved by implementing the state-of-the-art PolyMAX algorithm in a programming language capable of being coupled with machine learning algorithms and big data storage systems. In order to make it suitable for continuous analysis, the algorithm is improved by eliminating all the required computer-human interactions and adding a method that automatically tracks the modal parameters along different data sets. The work deals also with the problem of the application of OMA on rotating machines: the presence of harmonics. To this purpose, two different procedures are analyzed and implemented: the use of a cepstrum lifter and the use of order based modal analysis. The implemented procedure is used to process real data coming from a long-term monitoring campaign on an offshore wind turbine in the Belgian North Sea. Data sets coming from the turbine in different operating conditions are processed in order to take into account all the topics involved in the thesis. Interesting results are obtained. First of all it has been shown that when used on stand still data (i.e data acquired while the turbine is not producing energy) the implemented automatic procedure is able to detect modes close to narrow harmonic components by itself, without the need of pre-processing steps. Unfortunately it still fails when the harmonics influence a broader frequency band. Secondly the results obtained from the data coming from the rotating turbine and processed with the two methodologies proposed for dealing with the harmonics are compared with the results obtained from the stand still turbine. It has been shown that the order based modal analysis gives results more compatible with the results coming from multibody simulations available in literature. The latter result shows the power of the work implemented. This automatic modal parameter estimation that can be performed on long-term data can be implemented in a new concept of design: the simulated model can be integrated with the information learnt during the life time of the previous prototypes in order to have more robust design of the components.| File | Dimensione | Formato | |
|---|---|---|---|
|
Tesi_NicolettaGioi_837006.pdf
non accessibile
Descrizione: Thesis text
Dimensione
6.76 MB
Formato
Adobe PDF
|
6.76 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/135432