Propeller Design and Optimization Using a Robust Genetic Algorithm and a Computationally Efficient Solver

Abstract

A computationally efficient and reliable propeller design tool has been constructed using an advanced real coded Genetic Algorithm (GA) and a mid-fidelity potential flow solver. The GA constructs a population of propeller geometries using a series of Bernstein Polynomials (BP) which have a total of 63 coefficients. This population of propeller geometries is then tested using a reliable and efficient solver. The best four members from the population are then obtained by means of a tournament style selection followed by a round-robin style tournament to determine the true maximums. A following population is then built using the 63 characteristics from the four most optimal members. The process of build, test, select, and build is carried out for several demes or subpopulations which provide the initial population for the main generational loop. After all the deme and main generations have been executed, the GA will provide a propeller that matches the desired thrust input for the specified operating conditions and diameter while maintaining high propulsive efficiencies due to the nature of the fitness function. This thesis describes the technical details of the optimizer, solver, and associated tooling, validation cases for the solver, and sample optimization results.