Answers are grounded in this paper and cite their sources.
This study presents a machine learning–assisted optimisation framework for the design of contra-rotating propellers (CRPs) to improve hydrodynamic efficiency while accounting for cavitation and noise constraints. A Multiple-Input Multiple-Output Artificial Neural Network (MIMO-ANN) is developed as a surrogate model to predict open-water performance using the thrust loading coefficient (KT/J2), expanded area ratio (EAR) and pitch-to-diameter ratio (P/D) as inputs. The model achieves high predictive accuracy across multiple propeller series, with coefficients of determination (R2) exceeding 0.9996 and root-mean-square errors (RMSE) below 0.005. The ANN is coupled with constrained nonlinear optimisation to maximise open-water efficiency subject to the Keller cavitation criterion and a blade-tip-speed limit. The best-performing CRP configurations achieve efficiencies exceeding 68%, with a maximum of 70.30%, compared with 65.0% for the reference KP505 and 66.71% for the best optimised fixed-pitch propeller (FPP). Including propeller diameter as a decision variable affects thrust loading, rotational speed, and cavitation constraint. Some high-efficiency solutions marginally violate the cavitation constraint because of high operating load and geometric limits of the systematic propeller series; these cases are treated as near-feasible performance boundaries. The framework is intended for preliminary CRP design and design-space screening before higher-fidelity computational fluid dynamics (CFD) or experimental verification.Abbreviations: ANN: Artificial Neural Network; b(ℓ): Bias vector; B: Breadth; CAD: Computer-aided design; CFD: Computational fluid dynamics; CRP: Contra-rotating propeller; D: Diameter; DES: Detached Eddy Simulation; Δ: Displacement; T: Draft; EAR: Expanded area ratio; EARmin: Minimum EAR required to avoid cavitation; EEDI: Energy Efficiency Design Index; ESD: Energy-saving devices; f: Function; FPP: Fixed-pitch propeller; FSI: Fluid–structure interaction; FW-H: Ffowcs Williams–Hawkings; GA: Genetic algorithms; GHG: Greenhouse gas; h: Propeller centerline immersion; IMO: International Maritime Organization; JA: Advance coefficient; k: Constant; KCS: KRISO Container Ship; KQ: Torque coefficient; KRISO: Korea Research Institute of Ships and Ocean Engineering; KT: Thrust coefficient; KT/J2: Thrust loading coefficient; L: Loss function; ℓ: Hidden layer; LFC: Lifetime fuel consumption; LNG: Liquefied natural gas; LWL: Length water line; MFO: Moth-flame optimisation; MIMO: Multiple-Input Multiple-Output; ML: Machine learning; N: Samples; n: propeller rotational speed; NSGA: Non-dominated Sorting Genetic Algorithm; NURBS: Non-Uniform Rational B-Splines; P/D: Pitch-to-diameter ratio; Patm: Atmospheric pressure; PSO: Particle swarm optimisation; Pv: Vapor pressure; R: Correlation coefficients; R2: Coefficients of determination; RANS: Reynolds-Averaged Navier–Stokes; RBFN: Radial basis function networks; RMSE: Root mean square error; S: Wetted surface area; T: Draft; T: Thrust; URN: Underwater radiated noise; VA: Propeller speed of advance; VS: Ship speed; Vtip: Blade tip speed; Vtip-max: Maximum permissible tip speed; w: Wake fraction; W(ℓ): Weight matrix; WCT: Wide-chord-tip; x: Input vector; y: Output vector; Z: Blade numbers; γ: Specific weight; Δ: Displacement; ηo: Open water efficiency; Θ: Biases; σ: Non-linear activation function.
No takes yet. Share an insight, caveat, or question.
Tadros et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: