Utilizing machine learning to predict static pressure over a wingtip

Abstract

Our goal was to find expected pressure around different wingtip shapes to predict vortice behavior. This project focused on aerodynamics, specifically the location where high and low-pressure air mixes over a lifting surface. High-pressure air mixes with the low-pressure air at the wingtip of a plane creating vortices that cause drag, which wastes fuel and slows down the aircraft. Not only is this bad for the environment, but it increases the cost of flight and affects the distance that larger planes can fly ahead of smaller planes due to wake turbulence. As planes have gotten lighter, faster, and safer, the issue of wingtip vortices and drag has continued to be a problem. The approach we used to answer this problem was to select an applicable data set using continuous machine learning models and later, discrete models to predict a pressure coefficient above the wing. We combined multiple datasets from the same research paper created by NASA to have numerous factors for the machine learning model to predict. As a result, we produced accurate static pressure predictions with 80% to 90% accuracy. Even more accurate were our model recall scores which were within 99%. As a result of the work done on this project, accurate predictions of expected pressure over an airfoil are achievable. With only a few input variables about speed and dimensions, an accurate static pressure can be found.