Hydrodynamic characteristics of coral nursery buoyancy system and prediction using machine learning methods under wave action

The buoyancy system is a critical structure that ensures the safety and stability of the floating coral nursery and reduces the rate of coral shedding. Firstly, experiments in a wave tank were conducted on the motion response characteristics of the buoyancy system with four different structural types and submerged states, using the buoyancy system as the experiment object. The experimental results showed that the submerged state of the floating sphere was an essential element in affecting the hydrodynamic characteristics of the buoyancy system. The variation trends of the maximum oscillating tension of mooring cable with increase in period, and of the floating sphere motion amplitude with increase in Kc number under different submerged states were significantly different. Subject to experiment conditions, it was hard to judge the change tendency of the floating sphere motion amplitude under the large Kc number. Machine learning models trained through experiment data samples can quickly and accurately predict physical phenomena over a broader parameter range. Secondly, the prediction results of various machine learning models were compared and analyzed. Then the traditional machine learning model was optimized based on the particle swarm algorithm, which can improve the accuracy by about 20%. Finally, the optimal PSO-RBF model was employed to predict the floating sphere motion amplitude under the large Kc number. The predicted results showed that Kc = 6 was the demarcation point where the motion amplitude of the floating sphere varies with the Kc number. In addition, the fully submerged single floating sphere of buoyancy system is the best choice for floating coral nurseries under the four experiment cases in this paper, which can effectively reduce the damage of extreme waves with large wave heights and periods to floating coral nurseries.