Numerical Optimization of a propeller in a given wake and behind a ship

Licentiatavhandling, 2005

The work presented in this thesis includes two stages in a research project at the Rolls-Royce University Technology Center (UTC) at Chalmers Technology University. The objective of the project is to develop an optimization technique for ship/propeller interaction. In the first stage, the optimization of a cavitating propeller blade in a given wake is made to maximize the efficiency and minimize the propeller induced pressure fluctuations by tuning the propeller blade geometry. Both single objective and multi objective optimization models are set up. Cavitation constraints are switched on separately or simultaneously to investigate their influence on cavitation and efficiency. A gradient-based optimization method DHC (Dynamic Hill Climbing) is adopted. Optimizations starting from a near optimum and an off-design propeller are made and optimums obtained show better performance of efficiency and cavitation compared with the original propeller. In the second stage, the optimization of a propeller behind a fixed ship (Hamburg test case) at full scale is carried out to minimize the delivered power by adjusting the propeller geometry. To reduce the large computation effort required, the minimum iteration number and grid density are carefully selected. A global optimization method GA (Genetic Algorithm) is applied for the optimizations of a near optimum and an off-design propeller. A procedure of verification and correction of the objective function and design variables because of the unsatisfied constraint in the finer grids is introduced and applied for the near-optimum optimization. The aim of this work is to automatically optimize the geometry of a propeller with different objective functions and constraints. Especially once RANS solver is involved, the reduction of the computation effort required and the verification of the optimal design in the finer grids are vital to 3D optimization of the propeller/ship interaction at full scale. Although such procedures have only been applied for few cases, it is believed that they will produce clues and guidance for future more 3D optimizations with RANS solvers.