In the pursuit of sustainable and resilient offshore infrastructure, the integration of renewable energy sources and aquaculture systems on floating multi-use platforms presents a pioneering approach to harnessing the ocean’s potential. Between March 2021 and January 2022, a 1:15 scale prototype was tested in Reggio Calabria, Italy. This experiment provided invaluable insights into the dynamic responses of such structures under varying wave climates. Our study introduces the use of Artificial Neural Networks (ANNs) to predict the variability in structural load measurements, focusing on specific structural points. By analysing metocean data, various ANN models and optimization techniques were evaluated to find the most accurate predictive model. Our findings, showing a Normalized Root Mean Square Error (NRMSE) of 1.7-4.7%, demonstrate ANNs’ potential in predicting offshore platform behaviours. This research underscores the value of machine learning in designing and operating sustainable ocean-based systems, paving the way for future data-driven marine resource management endeavours.
Advancements in predicting maritime structure dynamics using neural networks / Martzikos, N.; Ruzzo, C.; Malara, G.; Arena, F.. - (2025), pp. 483-490. (Intervento presentato al convegno 6th International Conference on Renewable Energies Offshore tenutosi a Lisbon - Portugal nel 19 - 21 November 2024) [10.1201/9781003558859-53].
Advancements in predicting maritime structure dynamics using neural networks
Malara, G.;Arena, F.
2025-01-01
Abstract
In the pursuit of sustainable and resilient offshore infrastructure, the integration of renewable energy sources and aquaculture systems on floating multi-use platforms presents a pioneering approach to harnessing the ocean’s potential. Between March 2021 and January 2022, a 1:15 scale prototype was tested in Reggio Calabria, Italy. This experiment provided invaluable insights into the dynamic responses of such structures under varying wave climates. Our study introduces the use of Artificial Neural Networks (ANNs) to predict the variability in structural load measurements, focusing on specific structural points. By analysing metocean data, various ANN models and optimization techniques were evaluated to find the most accurate predictive model. Our findings, showing a Normalized Root Mean Square Error (NRMSE) of 1.7-4.7%, demonstrate ANNs’ potential in predicting offshore platform behaviours. This research underscores the value of machine learning in designing and operating sustainable ocean-based systems, paving the way for future data-driven marine resource management endeavours.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.