Chinewalking, and Porpoising, and Machine Learning, Oh My!

Small high-speed craft are prone to dynamic instabilities such as porpoising, chine walking, and spin-out which are not typically observed in larger vessels. To investigate these behaviors, experiments were conducted on a double-stepped hull at the U.S. Naval Academy, where porpoising and chine walking were observed. This study explores data-driven modeling approaches to understand and predict these instabilities, with an emphasis on real-time applicability. Preliminary analysis using Long Short-Term Memory (LSTM) networks demonstrates the ability to forecast key motion states with moderate accuracy. For pitch and heave, the models achieve mean absolute errors (MAE) of 0.9° and 0.3 in, respectively, and root mean square errors (RMSE) of 1.1° and 0.4 in. Acceleration predictions are less accurate (MAE = 0.9 g, RMSE = 1.2 g), likely due to sensor noise and high-frequency content. Complementary analysis using Dynamic Mode Decomposition (DMD) applied to pitch and heave signals suggests that porpoising may manifest as low-growth-rate oscillatory modes persisting across multiple speed regimes consistent with nonlinear, sustained limit-cycle behavior. These findings highlight the potential for data-driven methods to characterize and predict dynamic instabilities in planing craft. Ongoing work aims to extend this analysis to chine walking, which exhibits coupled roll-yaw dynamics, with the broader goal of enhancing the operational safety and design of high-performance small vessels.