TY - GEN

T1 - Assessment of sensitivity analysis methods of different complexity for offshore wind turbines

AU - Hübler, Clemens

AU - Gebhardt, Cristian G.

AU - Rolfes, Raimund

N1 - Funding information: This work was supported by the compute cluster, which is funded by Leibniz Universität Hannover, the Lower Saxony Ministry of Science and Culture (MWK) and the German Research Association (DFG).

PY - 2019

Y1 - 2019

N2 - Sensitivity analyses (SA), being used to identify the most influential inputs of simulation models, are still rarely conducted for offshore wind turbines (OWTs). If SA are performed, in most cases, no reasons for the choice of the method are given and the selection is based on expert knowledge only. Objective criteria are not applied. Frequently, simplified SA are utilized, e.g. one-at-a-time analyses, although it is known that these approaches are not suitable for non-linear or stochastic systems. Sophisticated SA, like variance-based methods, are time-consuming and limited suitable for OWT models. As a consequence, here, the performance of various SA methods of different complexity in the context of structural analyses of OWTs is investigated. The analyzed methods range from simple testing of parameter constellations, to elementary effects and meta-model-based approaches, all the way to variance-based approaches. Objective recommendations with regard to the most suitable SA methods for OWT analyses based on the two criteria, accuracy and computing time, are given. Results show that neither plain linear SA (not selecting the right probabilistic subset) nor variance-based approaches (being computational inefficient) are the best choice, but meta-model-based approaches represent the recommended compromise. Furthermore, challenges of SA for stochastic systems are highlighted.

AB - Sensitivity analyses (SA), being used to identify the most influential inputs of simulation models, are still rarely conducted for offshore wind turbines (OWTs). If SA are performed, in most cases, no reasons for the choice of the method are given and the selection is based on expert knowledge only. Objective criteria are not applied. Frequently, simplified SA are utilized, e.g. one-at-a-time analyses, although it is known that these approaches are not suitable for non-linear or stochastic systems. Sophisticated SA, like variance-based methods, are time-consuming and limited suitable for OWT models. As a consequence, here, the performance of various SA methods of different complexity in the context of structural analyses of OWTs is investigated. The analyzed methods range from simple testing of parameter constellations, to elementary effects and meta-model-based approaches, all the way to variance-based approaches. Objective recommendations with regard to the most suitable SA methods for OWT analyses based on the two criteria, accuracy and computing time, are given. Results show that neither plain linear SA (not selecting the right probabilistic subset) nor variance-based approaches (being computational inefficient) are the best choice, but meta-model-based approaches represent the recommended compromise. Furthermore, challenges of SA for stochastic systems are highlighted.

KW - Global sensitivity

KW - Offshore substructure

KW - Sensitivity analysis

KW - Stochastic models

KW - Structural dynamics

KW - Uncertainty

KW - Wind energy

UR - http://www.scopus.com/inward/record.url?scp=85075002027&partnerID=8YFLogxK

U2 - 10.3850/978-981-11-2724-3_0140-cd

DO - 10.3850/978-981-11-2724-3_0140-cd

M3 - Conference contribution

AN - SCOPUS:85075002027

T3 - Proceedings of the 29th European Safety and Reliability Conference, ESREL 2019

SP - 3144

EP - 3151

BT - Proceedings of the 29th European Safety and Reliability Conference (ESREL)

A2 - Beer, Michael

A2 - Zio, Enrico

CY - Europe

T2 - 29th European Safety and Reliability Conference, ESREL 2019

Y2 - 22 September 2019 through 26 September 2019

ER -