A computer-implemented control system and method for dynamically managing airflow in an air lubrication system of a vessel are disclosed. Real-time sensor data including vessel motion, acceleration, and speed-through-water are processed by a controller executing a machine learning model to predict near-future sea-state conditions. The controller accesses and interpolates computational fluid dynamics (CFD) efficiency profiles to determine target airflow values and generate valve and compressor commands that modulate airflow to multiple outlets beneath the hull. Feedback from hull and pressure sensors is used to stabilize the air layer, while a learning-based prediction engine continuously refines CFD profiles to improve operational efficiency and reduce parasitic load. The system enhances vessel hydrodynamic performance and fuel economy under varying environmental conditions.