Optimal section design of surface piercing propeller using artificial neural networks algorithm

Abstract

ABSTRACT Section optimisation of a surface piercing propeller (SPP) with the combination of artificial neural networks and NSGA II algorithm is investigated in this research. The blade section have a sharp leading edge and thick trailing edge. The geometry of this propeller is mostly designed experimentally, and comprehensive studies have not been performed to examine its parameters. In this regard, the propeller's parametric geometry is designed based on the cubic B-spline method. The SPP was first fabricated and tested, and the simulated data of the tested propeller were verified. Then, limited data were extracted according to a verified CFD model based on which the ANN was trained. The ANN results indicated the mean square error of the trained network is 3e-9, 1e-7, and 1e-7 for training, validation, and test data, respectively. comparing the optimisation and experimental results showed a relative error amounting to 3.5% in thrust and 4.24% in torque.