Maintaining the integrity of mooring systems in floating offshore wind turbines (FOWTs) is essential, as any deterioration can impact the dynamics of the platform and its performance. A generalizable machine-learning-based monitoring system that continuously monitors various mooring systems for FOWTs requires dynamic response data from all the installed systems under diverse health, operational, and metocean conditions. To accomplish this, we propose a conditional hierarchical variational autoencoder (CHVAE) generative model designed for domain translation to generate the necessary data. We employ a pretraining-finetuning procedure to train the proposed model using the majority of healthy fairlead tension responses and the minority of damaged responses from the FOWT platform moored by the source mooring system. Subsequently, by using healthy responses from an unseen mooring system for the platform and CHVAE, the corresponding real-scale damage-associated tension responses are generated. These responses are then used to train a binary classifier to distinguish between healthy and damaged states. Given the limited availability of measurements, the training and evaluation datasets are generated using the open-source software OpenFast, with the OC4-DeepCWind semisubmersible platform as the reference benchmark. A thorough analysis is conducted to evaluate the similarity between the simulated records generated by OpenFast and those created by the CHVAE framework, employing visual, statistical, and behavioural methods. Additionally, the performance of the proposed generative model is initially evaluated on MNIST data to test its effectiveness for data augmentation. The generated records for unobserved random sea states closely resemble real-world dynamic behaviours during downstream binary classification, demonstrating the efficacy and adaptability of the CHVAE.

Unseen Mooring Monitoring by Conditional Hierarchical Variational Domain Translation / Fathnejat, Hamed; Nava, Vincenzo. - 674 LNCE:(2025), pp. 781-792. (Intervento presentato al convegno 11th International Conference on Experimental Vibration Analysis for Civil Engineering Structures, EVACES 2025 tenutosi a prt nel 2025) [10.1007/978-3-031-96110-6_77].

Unseen Mooring Monitoring by Conditional Hierarchical Variational Domain Translation

Vincenzo Nava
2025

Abstract

Maintaining the integrity of mooring systems in floating offshore wind turbines (FOWTs) is essential, as any deterioration can impact the dynamics of the platform and its performance. A generalizable machine-learning-based monitoring system that continuously monitors various mooring systems for FOWTs requires dynamic response data from all the installed systems under diverse health, operational, and metocean conditions. To accomplish this, we propose a conditional hierarchical variational autoencoder (CHVAE) generative model designed for domain translation to generate the necessary data. We employ a pretraining-finetuning procedure to train the proposed model using the majority of healthy fairlead tension responses and the minority of damaged responses from the FOWT platform moored by the source mooring system. Subsequently, by using healthy responses from an unseen mooring system for the platform and CHVAE, the corresponding real-scale damage-associated tension responses are generated. These responses are then used to train a binary classifier to distinguish between healthy and damaged states. Given the limited availability of measurements, the training and evaluation datasets are generated using the open-source software OpenFast, with the OC4-DeepCWind semisubmersible platform as the reference benchmark. A thorough analysis is conducted to evaluate the similarity between the simulated records generated by OpenFast and those created by the CHVAE framework, employing visual, statistical, and behavioural methods. Additionally, the performance of the proposed generative model is initially evaluated on MNIST data to test its effectiveness for data augmentation. The generated records for unobserved random sea states closely resemble real-world dynamic behaviours during downstream binary classification, demonstrating the efficacy and adaptability of the CHVAE.
2025
9783031961090
9783031961106
File in questo prodotto:
Non ci sono file associati a questo prodotto.
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11583/3004111
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo