A prediction on structural stress and deformation of fish cage in waves using machine-learning method

Abstract

Generally, the hydrodynamics of a fish cage are investigated using numerical simulation, physical model experiments, and field measurements. However, these traditional research methods are time consuming and low in efficiency. In this study, an artificial neural network (ANN) model is built such that the hydrodynamic characteristics of a fish cage in waves can be predicted rapidly. The training data of the ANN model are generated by a well-developed numerical model from our previous studies. The parameters in the hidden layer of the ANN model are determined considering the prediction accuracy of the hydrodynamic results of the fish cage. The ANN model is validated against a well-developed numerical model with satisfactory accuracy. Using the proposed ANN model, the hydrodynamic results of the fish cage including the maximum tension in mooring lines, minimum effective-volume ratio, and maximum stress of the floating collar are predicted for various waves. Overall, the predicted results indicate a trend consistent with that of the previous studies. The present model can potentially forecast disasters for an oncoming typhoon, which is important for the hazard prevention of a fish farm.