TY - CHAP

T1 - Investigations Towards Physics-Informed Gaussian Process Regression for the Estimation of Modal Parameters of a Lattice Tower Under Environmental Conditions

AU - Möller, Sören

AU - Jonscher, Clemens

AU - Grießmann, Tanja

AU - Rolfes, Raimund

N1 - The authors greatly acknowledge the financial support provided by Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) - CRC 1463, subproject C02.

PY - 2023

Y1 - 2023

N2 - In the context of structural health monitoring (SHM), it is stated that changing environmental conditions (ECs) affect the structure of interest. This fact makes it difficult to distinguish between structural changes caused by damage and those caused by changing ECs. This paper presents a simple physics-informed Gaussian process (GP) to predict the natural frequencies of a lattice tower structure for damage detection. It explores the idea of modelling the effects of different ECs rather than, for example, classifying them. For this purpose, ECs in terms of wind speed, humidity and temperature are used as inputs to a GP to estimate the first two bending modes in the x- and y-directions of the structure. Observed dependencies between inputs and outputs are incorporated by using basis functions to obtain a physically informed GP and hence a grey-box model. To use the estimations and the related confidence intervals as damage-sensitive features, the difference to the measured data is calculated and a threshold for subsequent damage detection is defined. The results are validated using the Leibniz University Test Structure for Monitoring (LUMO), an outdoor lattice tower. It is found that only a small amount of training data is required to achieve acceptable accuracy. Furthermore, it is shown that the presented approach can be used for the detection of artificially induced damage.

AB - In the context of structural health monitoring (SHM), it is stated that changing environmental conditions (ECs) affect the structure of interest. This fact makes it difficult to distinguish between structural changes caused by damage and those caused by changing ECs. This paper presents a simple physics-informed Gaussian process (GP) to predict the natural frequencies of a lattice tower structure for damage detection. It explores the idea of modelling the effects of different ECs rather than, for example, classifying them. For this purpose, ECs in terms of wind speed, humidity and temperature are used as inputs to a GP to estimate the first two bending modes in the x- and y-directions of the structure. Observed dependencies between inputs and outputs are incorporated by using basis functions to obtain a physically informed GP and hence a grey-box model. To use the estimations and the related confidence intervals as damage-sensitive features, the difference to the measured data is calculated and a threshold for subsequent damage detection is defined. The results are validated using the Leibniz University Test Structure for Monitoring (LUMO), an outdoor lattice tower. It is found that only a small amount of training data is required to achieve acceptable accuracy. Furthermore, it is shown that the presented approach can be used for the detection of artificially induced damage.

KW - Damage detection

KW - Data normalisation

KW - Gaussian process regression

KW - Grey-box modelling

UR - http://www.scopus.com/inward/record.url?scp=85174808056&partnerID=8YFLogxK

U2 - 10.1007/978-3-031-39117-0_41

DO - 10.1007/978-3-031-39117-0_41

M3 - Contribution to book/anthology

SN - 978-3-031-39116-3

T3 - Lecture Notes in Civil Engineering

SP - 401

EP - 410

BT - Experimental Vibration Analysis for Civil Engineering Structures

A2 - Limongelli, Maria Pina

A2 - Giordano, Pier Francesco

A2 - Gentile, Carmelo

A2 - Quqa, Said

A2 - Cigada, Alfredo

ER -