Machine Learning Classification of Vortex Wakes from Oscillating Foils

A machine learning classification model is developed for wake patterns behind oscillating foils whose kinematics are tuned for energy harvesting from the oncoming freestream flow. The role of wake structure is particularly important for arrays of oscillating foils, since the coherent structures can cause constructive and/or destructive interference with downstream foils. This work explores 35 different oscillating foil kinematics within energy harvesting mode, with the goal of grouping and parameterizing the wake based on the input kinematic variables. An approach combining a convolutional neural network (CNN) with long short-term memory (LSTM) units is utilized to automatically classify the wakes into three groups based on foil's relative angle of attack. After revising group boundaries based on model performance, an average accuracy of 90% is achieved, demonstrating that foil relative angle of attack can be used to discern distinct wake patterns and hence provide insight for optimizing foil array configurations for energy harvesting.