Knowledge–Data Hybrid-Driven Discovery of Interpretable Design Rules for High-Efficiency Propellers

Discovering design rules is essential for advancing interpretable and intelligent propeller design. This paper proposes a knowledge–data hybrid-driven method for discovering design rules for high-efficiency propellers, with the aim of expanding the pool of comparable design data while reducing the search space for candidate rule parameters. Based on the propeller ideal efficiency theory, an efficiency factor is introduced to address the lack of comparability among performance data caused by differences in propeller loading, thereby enabling all designs in the database to be used for data mining. Dimensionless composite features are then constructed using the principle of dimensional homogeneity and screened through correlation analysis within a knowledge–data hybrid-driven framework. Six characteristic load ratios are identified as rule parameters, yielding the following design rule: a propeller has a high efficiency factor when all six ratios lie within their respective optimal ranges. Finally, the rule is validated on 12 independent designs, including three designs of a different propeller. The proposed knowledge–data hybrid-driven feature-engineering framework provides a viable approach to design-rule discovery and may support inverse design and intelligent propeller design.