Hydrodynamic Responses and Machine Learning-Based Shape Classification of Harbor Seal Whiskers in the Wake of Bluff Bodies

Harbor seals, equipped with their uniquely structured whiskers, demonstrate remarkable proficiency in tracking the trajectories of prey within dark and turbid marine environments. This study experimentally investigates the wake-induced vibrations of an elastically supported whisker model placed in the wakes of circular, square, and equilateral triangular cylinders of varying dimensions. Thereafter, a machine learning model is trained to identify and classify these intrinsic responses. The findings reveal a positive correlation between the amplitude of vibration and the total circulation shed by the bluff bodies. In the wake flow fields of triangular and circular cylinders, the mean drag is quite similar. Meanwhile, the whisker's vibration amplitude and drag fluctuation show that the triangular cylinder is comparable to the square cylinder, and both are higher than the circular cylinder. To classify the wake-generating body shapes based on the hydrodynamic characteristics, hydrodynamic features encompassing vibration amplitudes, fluid forces, and frequency-related information were extracted to train an LSTM-based model, and it was found that the mean drag significantly enhances the model's flow velocity generalization performance.