Numerical methods for design, validation and optimisation of high-performance waterjets

This study introduces a novel waterjet propulsion concept, featuring a single unit housing the entire system, thus eliminating the need for internal components within the hull. Positioned significantly below the waterline, this design optimally operates at deep submergence, generating thrust perfectly aligned with the advancing direction, processing an axial, uniform capture stream tube, unaffected by the hull boundary layer. The main objective is to establish a comprehensive numerical framework for fluid dynamic behaviour, addressing system design, computational model validation, and optimization. The methodology progresses from a 1D nacelle approach inspired by aero-engines intakes principles to a meanline code for blade design employing Blade Element Method (BEM) integrated with empirical correlations. While stable, the approaches exhibit discrepancies, with accurate thrust coefficient predictions but overestimated efficiency and hydraulic design. To verify the study, Computational Fluid Dynamics (CFD) is employed, comparing numerical models with public domain experiments. The study considers various turbulence models and grid refinements for Reynolds-Averaged Navier Stokes (RANS) equations. A 2D-axisymmetric simplification reliably reproduces test measurements for revolved profile inlets, while a single-channel periodic reduction is suitable for bladed geometries. Discrepancies in nominal operations prompt systematic manipulation of Zwart cavitation model coefficients, enhancing accuracy in predicting thrust breakdown pressure rise. Numerical analysis of installed pump operations follows an integration strategy for previously isolated sub-systems. Hydraulic statistics show minimal influence from the upstream intake, but an off-design assessment reveals a deficiency in the matching strategy, resulting in peak performance at differing advancing speeds for the inlet and pump. Optimisation is deemed necessary, initiating with Latin Hypercube Sampling (LHS) as the first step for a Genetic Algorithm (GA). The investigation manipulates the 2D-axisymmetric intake geometry with a focus on axial and radial highlight locations and throat radius. Two-objective oriented analysis indicates readily reducible drag contribution but inherently limited pressure recovery maximisation. Concluding with a multi-fidelity strategy, the study employs the 3D installed model for three optimised solutions. While cruise operations may improve, a multipoint optimisation, including off-design performance, is deemed necessary to prevent unfeasible behaviour, emphasised by cavitation evolution.