Removal of Environmental Effects in Structural Health Monitoring Data
Translated title
Removal of Environmental Effects in Structural Health Monitoring Data: Structural Health Monitoring
Author
Vo, Tuan Viggo Luong
Term
4. semester
Education
Publication year
2018
Submitted on
2018-09-07
Pages
83
Abstract
Vindenergi er vokset kraftigt de seneste årtier. Turbinerne bliver større og placeres ofte til havs, hvilket gør det mere krævende at sikre deres strukturelle integritet. Visuel inspektion er ikke altid tilstrækkelig, så der er behov for fjernbaseret strukturel sundhedsovervågning (SHM), især af vindmøllevinger. En central udfordring er, at miljøforhold som temperatur og vind kan påvirke målinger på måder, der ligner skader, så ægte skadesindikatorer bliver skjult. Afhandlingens hovedfokus er at udvikle en databehandlingsmetode, der kan isolere et skadesfølsomt træk og samtidig fjerne indflydelsen fra miljøet. Til dette anvendes kointegration, en metode fra økonometrien. Når flere måleserier deler fælles, langsigtede tendenser, kan man finde en lineær kombination, hvor disse fælles tendenser ophæver hinanden. Det resterende signal er mere stabilt over tid og kan fungere som en indikator, der er følsom over for skader, men mindre påvirket af vejret. Metoden undersøges på en fuldskala, i drift, Vestas V27-vindmølle, der blev overvåget i cirka tre måneder i fem forskellige strukturelle tilstande, herunder kunstigt indførte skader. Samtidig blev der indsamlet meteorologiske data fra en nærliggende vejrmast. Robustheden af den udviklede skadesindikator vurderes med en outlieranalyse baseret på Mahalanobis-afstand, som kan afdække usædvanlige mønstre, der kan indikere skade.
Wind energy has expanded rapidly in recent decades. Turbines are getting larger and are often installed offshore, making it more demanding to ensure their structural integrity. Visual inspection alone is not always sufficient, so there is a need for remote structural health monitoring (SHM), with a particular focus on wind turbine blades. A key challenge is that environmental conditions, such as temperature and wind, can change sensor readings in ways that resemble damage, masking true damage indicators. The main focus of this thesis is to develop a data processing approach that isolates a damage-sensitive feature while removing the influence of the environment. To achieve this, the study uses cointegration, a method from econometrics. When several measurement series share long-term trends, it is possible to find a linear combination in which those common trends cancel out. The remaining signal is more stable over time and can serve as a feature that is sensitive to damage but less affected by weather. The approach is examined on a full-scale, operational Vestas V27 wind turbine monitored for about three months in five different structural states, including artificially introduced damages, while meteorological data were collected from a nearby weather mast. The robustness of the developed damage indicator is assessed using outlier analysis based on the Mahalanobis distance, which highlights unusual patterns that may indicate damage.
[This abstract was generated with the help of AI]
Documents