Neural network model for assessing structural performance of netting in net-enclosure aquaculture facilities

Maintaining the integrity of netting is crucial for the protection of net-enclosure aquaculture facilities (NEAFs) against disastrous weather. To assess the expected structural performance of NEAF netting, a rapid prediction model based on a backpropagation neural network (BPNN) optimized by a genetic algorithm (GA) was devel-oped. This GA-BPNN model is a preferable alternative to time-consuming traditional methods such as numerical simulations and laboratory experiments. The maximum loads on the rope and twine of the NEAF were predicted using the proposed GA-BPNN model with ocean conditions and structural properties (including current velocity, wave height, pile spacing, and rope/twine diameters) as input parameters. The prediction error (3.81%) and number of training-convergence steps (18) of the GA-BPNN model were lower than those of the BPNN model (9.28% and 33), respectively. The netting structural performance of NEAF under various conditions was also assessed by the GA-BPNN model. The results showed that the model predicted NEAF hydrodynamic response quickly and accurately.