Propeller hydrodynamic performance prediction based on PSO-BP neural network

In order to quickly calculate the hydrodynamic performance of propellers in propeller design, this paper proposes a PSOBP neural network-based propeller hydrodynamic performance prediction method. Taking the B-type 4-bladed propeller as an example, different propeller design parameters are divided into 317 groups of data for CFD simulation calculation; 85% of the calculated data sets are randomly selected as the training set of PSO-BP neural network, and the remaining 15% is used as the test set to train the propeller hydrodynamic performance prediction neural network model; the results show that: the error between predicted and real values of the trained neural network for the coefficients of thrust, KT, and torque, KQ, is very small. The results show that the trained neural network has a very small error between the predicted and real values of the thrust coefficient KT and torque coefficient KQ, and the R2 values of the validation set are above 0.95, which meets the requirement of accurate and fast prediction of hydrodynamic performance.