Outline

As the propeller sweeps through the wake field of the hull, the pressure as well as the amount of cavitation varies with time and space, and these variations enhance the noise level which can adversely affect the migration and reproduction of underwater life, so a low-noise marine propeller design is important to reduce underwater noise.

The use of ANN can predict noise faster, but the prerequisite is the need of sufficient data for training. The accuracy of the data source, the structure of the network and the selection of the network parameters all have significant impact on the prediction results.

Targets:

The input to the ANN will be the geometric parameters and the operating conditions. The geometric parameters include the distribution of thickness and camber, and the nose radius of the airfoil. The operating conditions will be the inflow velocity (Reynolds number), geometric angle of attack, and cavitation number. The cavitation volume can be tentatively defined as the output of the ANN. Since the combination of different operating conditions and different geometries generates a large number of calculations, thus, all numerical calculations should be performed based on a two-dimensional airfoil. The application of a symmetry condition can be used to reduce the time consumed by the computations.

Tasks:

• Literature review and collection of the experimental data of the airfoil.
• Selection of turbulence model and mesh parameters by comparing with the existing experiments.
• Automatic generation of 2D volume mesh based on given geometric parameters.
• Development of a forced motion model to determine the relationship between the angle of attack and the cavitation volume. The position of the cavitation on the airfoil will also be a subject of great interest.
• Collection of input and output data for training ANN.
• Implementation of ANN structure and tuning hyperparameters to complete your own network.

Type & Start

Master Thesis

start immediately.

Contact

Keqi Wang