Machine learning assisted propeller design

Machine learning assisted propeller design

연구 분야: Artificial Intelligence

학회: ICCPS '21: Proceedings of the ACM/IEEE 12th International Conference on Cyber-Physical Systems

Propellers are one of the most widely used propulsive devices for generating thrust from rotational engine motion both in marine vehicles and subsonic air-crafts. Due to their simplicity, robustness and high efficiency, propellers remained the mainstream design choice over the last hundred years. On the other hand, finding the optimal application-specific geometry is still challenging. This work in progress report describes application of modern and rapidly developing Machine Learning (ML) techniques to gain novel designs. We rely on a rich set of preexisting parametric design patterns and accumulated engineering knowledge supplemented by high-fidelity simulation models to formulate the design process as a supervised learning problem. The aim of our work is to develop and evaluate machine learning models for the parametric design of propellers based on application-specific constraints. While the application of ML techniques in optimal propeller design is at a very nascent level, we believe that our early results are promising with a potentially significant impact on the overall design process. The ML-assisted design flow allows for a more automated design space exploration process with less dependency on human intuition and engineering guidance.