Abstract
Ensuring the structural integrity of aquaculture cages is crucial for the industry's economic viability, environmental impact, and aquatic life safety. Effective real-time monitoring of fish net cages is essential for damage detection and preventing system failures. Traditional monitoring methods, including numerical simulations and sensor deployment, face high computational demands, significant costs, and data loss risks. Addressing these issues, we develop machine learning-based reduced order models utilizing Gaussian process regression for the efficient real-time monitoring of net cage structural behavior. These models predict fast and accurately the cage deformations and mooring line loads under varying environmental conditions, validated on a full-scale net cage operated in SINTEF ACE industrial fish farm and numerical data from the Fh-Sim software. This advancement supports decision-making and complex operations in fish farms, securing industry's resilience and increased performance, while opens new avenues for machine learning models specifically tailored for fish cage structural dynamics.
