Virtual sensing and strain estimation on an offshore wind turbine using supervised learning

Marius Tarpø; Sandro Amador; Evangelos Katsanos; Mattias Skog; Johan Gjødvad; Rune Brincker

Virtual sensing and strain estimation on an offshore wind turbine using supervised learning

Marius Tarpø, Sandro Amador, Evangelos Katsanos, Mattias Skog, Johan Gjødvad, Rune Brincker

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

140 Downloads (Pure)

Abstract

Virtual sensing enables expansion and transformation of measured quantities from physical sensors into new quantities at unmeasured locations. This allows for estimating the strain and stress in unmeasured locations of a system by transforming the physical sensors (input) into the desired strain and stress response (output). This transformation model can be based on either knowledge of the systems, data from the system, or any combination of these. In this paper, supervised learning and data-driven models are applied to strain estimation of an offshore wind turbine through Principal Component Analysis (PCA). Training data are used to establish the data-driven model that enables a versatile strain estimation that functions well under different wind scenarios than the training data set.

Original language	English
Title of host publication	COMPDYN 2021 : 8^th ECCOMAS Thematic Conference on Computational Methods in Structural Dynamics and Earthquake Engineering
Volume	2021
Publication date	2021
Pages	2112-2120
Publication status	Published - 2021
Event	8^th ECCOMAS Thematic Conference on Computational Methods in Structural Dynamics and Earthquake Engineering - Online, Athens, Greece Duration: 28 Jun 2021 → 30 Jun 2021 Conference number: 8

Conference

Conference	8^th ECCOMAS Thematic Conference on Computational Methods in Structural Dynamics and Earthquake Engineering
Number	8
Location	Online
Country/Territory	Greece
City	Athens
Period	28/06/2021 → 30/06/2021

Series	COMPDYN Proceedings
ISSN	2623-3347

Keywords

Data-driven model
Principal component analysis
Stress estimation
Structural health monitoring
Virtual sensing

Access to Document

fulltextFinal published version, 595 KB

OpenUrl availability

Full text

Cite this

@inproceedings{c76e2489b536473098c52a1f4c05984c,

title = "Virtual sensing and strain estimation on an offshore wind turbine using supervised learning",

abstract = "Virtual sensing enables expansion and transformation of measured quantities from physical sensors into new quantities at unmeasured locations. This allows for estimating the strain and stress in unmeasured locations of a system by transforming the physical sensors (input) into the desired strain and stress response (output). This transformation model can be based on either knowledge of the systems, data from the system, or any combination of these. In this paper, supervised learning and data-driven models are applied to strain estimation of an offshore wind turbine through Principal Component Analysis (PCA). Training data are used to establish the data-driven model that enables a versatile strain estimation that functions well under different wind scenarios than the training data set.",

keywords = "Data-driven model, Principal component analysis, Stress estimation, Structural health monitoring, Virtual sensing",

author = "Marius Tarp{\o} and Sandro Amador and Evangelos Katsanos and Mattias Skog and Johan Gj{\o}dvad and Rune Brincker",

year = "2021",

language = "English",

volume = "2021",

series = "COMPDYN Proceedings",

publisher = "National Technical University of Athens",

pages = "2112--2120",

booktitle = "COMPDYN 2021",

note = "8<sup>th</sup> ECCOMAS Thematic Conference on Computational Methods in Structural Dynamics and Earthquake Engineering, COMPDYN 2021 ; Conference date: 28-06-2021 Through 30-06-2021",

}

Tarpø, M, Amador, S , Katsanos, E, Skog, M, Gjødvad, J & Brincker, R 2021, Virtual sensing and strain estimation on an offshore wind turbine using supervised learning. in COMPDYN 2021 : 8^th ECCOMAS Thematic Conference on Computational Methods in Structural Dynamics and Earthquake Engineering. vol. 2021, COMPDYN Proceedings, pp. 2112-2120, 8^th ECCOMAS Thematic Conference on Computational Methods in Structural Dynamics and Earthquake Engineering, Athens, Greece, 28/06/2021.

Virtual sensing and strain estimation on an offshore wind turbine using supervised learning. / Tarpø, Marius; Amador, Sandro ; Katsanos, Evangelos et al.
COMPDYN 2021 : 8^th ECCOMAS Thematic Conference on Computational Methods in Structural Dynamics and Earthquake Engineering. Vol. 2021 2021. p. 2112-2120 (COMPDYN Proceedings).

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

TY - GEN

T1 - Virtual sensing and strain estimation on an offshore wind turbine using supervised learning

AU - Tarpø, Marius

AU - Amador, Sandro

AU - Katsanos, Evangelos

AU - Skog, Mattias

AU - Gjødvad, Johan

AU - Brincker, Rune

N1 - Conference code: 8

PY - 2021

Y1 - 2021

N2 - Virtual sensing enables expansion and transformation of measured quantities from physical sensors into new quantities at unmeasured locations. This allows for estimating the strain and stress in unmeasured locations of a system by transforming the physical sensors (input) into the desired strain and stress response (output). This transformation model can be based on either knowledge of the systems, data from the system, or any combination of these. In this paper, supervised learning and data-driven models are applied to strain estimation of an offshore wind turbine through Principal Component Analysis (PCA). Training data are used to establish the data-driven model that enables a versatile strain estimation that functions well under different wind scenarios than the training data set.

AB - Virtual sensing enables expansion and transformation of measured quantities from physical sensors into new quantities at unmeasured locations. This allows for estimating the strain and stress in unmeasured locations of a system by transforming the physical sensors (input) into the desired strain and stress response (output). This transformation model can be based on either knowledge of the systems, data from the system, or any combination of these. In this paper, supervised learning and data-driven models are applied to strain estimation of an offshore wind turbine through Principal Component Analysis (PCA). Training data are used to establish the data-driven model that enables a versatile strain estimation that functions well under different wind scenarios than the training data set.

KW - Data-driven model

KW - Principal component analysis

KW - Stress estimation

KW - Structural health monitoring

KW - Virtual sensing

M3 - Article in proceedings

VL - 2021

T3 - COMPDYN Proceedings

SP - 2112

EP - 2120

BT - COMPDYN 2021

T2 - 8<sup>th</sup> ECCOMAS Thematic Conference on Computational Methods in Structural Dynamics and Earthquake Engineering

Y2 - 28 June 2021 through 30 June 2021

ER -

Virtual sensing and strain estimation on an offshore wind turbine using supervised learning

Abstract

Conference

Keywords

Access to Document

OpenUrl availability

Fingerprint

Cite this