Design of pump-jet propulsor based on data-driven optimization method

Abstract

This study optimizes the hydrodynamic performance of a pump-jet propulsor using a novel surrogate-assisted evolutionary algorithm. Incorporating Piecewise Cubic Hermite Interpolating Polynomials (PCHIP) for blade parameterization, a comprehensive parameterization methodology is developed. This approach establishes the propulsor's geometry while significantly expanding the design space, facilitating the generation of smooth and diverse geometric models. Besides, integrating the duct, rotor, stator, and hub, it efficiently establishes the entire models with 23 variables. Key design variables of rotor and stator include the distribution of pitch, camber, thickness, skew and mounting angle, besides attack angle of duct and hub as well as the distance between rotor disc and stator disc are also involved. The optimization process seeks to maximize efficiency while considering thrust, torque, and minimum pressure requirements through one objective and three constraint functions. This approach not only enhances optimization efficiency but also prevents unrealistic design outcomes. As a result, the efficiency improves by 7.67% compared to the initial design. The optimized model changes the distribution of the rotor pitch ratio at higher radius (especially over 0.75 r/R), drops the duct's angle of attack by about 1°, and forms a gentler pressure distribution that decreases the risk of flow separation. These adjustments lead to improved thrust stability and better coordination among component groups, making the system more efficient and well-integrated. Further unsteady analysis revealed enhanced hydrodynamic performance. With minimal designer intervention, this data-driven optimization method effectively fine-tunes each component of the pump-jet, enhancing overall system performance.