Position parameters optimization of surface piercing propeller by artificial neural network

Improving the performance of surface-piercing propellers is achieved by investigating the influential factors. In this study, Artificial Neural Network is used to identify nonlinear models for estimating various phenomena. Non-Dominated Sorting Genetic Algorithm II is considered as an optimization tool. In this study, in order to optimize the position parameters, including the immersion ratio, angle of attack, and yaw angle, data from experimental tests at the HYDROTECH center of IUST were collected as the initial data field for the generation of training data by the artificial neural network, then experimental tests were implemented in the position of the Non-Dominated Sorting Genetic Algorithm II proposed as the output, and the results were compared. The Artificial Neural Network results showed that the mean error of the trained verified and test data is 7.5e-5, 1e-4, and 1e-4, respectively. Comparing the experimental and optimization results, the thrust coefficient showed a relative error of 9.7%, while the torque coefficient showed a relative error of 7.5%, this algorithm can be used as a cost-effective, time-saving method for a similar problem.