Developing an Airfoil Simulator with Machine Learning-based Performance Prediction

Voiceover

Presenter(s)

Ayan Vaishnav

Abstract or Description

Background: Simulations are capable tools in the aerospace industry, allowing engineers to test their designs virtually before making physical forms. However, simulations are not the final step in determining the design of an aerospace system - physical testing is still required to verify the performance calculations. With the recent growth in machine learning technologies, machine learning proves to be very useful. It can be incorporated into simulators to provide highly realistic approximations of the performance metrics. This is what I aim to do in my project, by creating an airfoil simulator that utilizes machine learning to predict its performance.

Abstract: Aerospace Engineering is a complex yet captivating field which advances the frontiers of humanity further into the universe. To be able to reach the next frontier, aerospace systems have to be designed with immaculate technical details, addressing multiple disciplines such as fluid dynamics, structures, propulsion, and more. In an aircraft, one of the most important disciplines to be addressed is the aerodynamics of the wing. The cross-sectional shape of the wing (known as an airfoil) plays a key role in this discipline. The performance of an airfoil is determined using aerodynamic metrics such as Lift, Drag, and the Lift-to-Drag (L/D) Ratio. However, testing for such metrics is a physical, labor-intensive process. Simulations are a viable alternative to physical testing; however, current simulation methodologies are only an approximation of the performance, meaning that physical testing is still required to gauge the accuracy of simulated performance calculations. Additionally, available simulation software is often outdated and does not leverage modern computational resources. In this project, I develop a modern airfoil simulator that leverages modern computational power through machine learning-based performance prediction. I also model and test my airfoil innovations to validate my simulation’s accuracy through the performance metrics derived in my previous projects.

Previous Works:

"Effects of Different Kinds of Airfoils on Lift, Weight, Thrust, and Drag"

• A basic introduction to the world of aerodynamics, testing different airfoil designs and working to determine their effects on the four forces of flight.

"What’s Up Th-air? Utilizing a Simulation to Design the Impeccable Airfoil for a Commercial Airliner"

• More in-depth exploration on airfoil designs. Incorporated the use of the FoilSim III simulation provided by NASA's Glenn Research Center to verify my experimental data, and extrapolated metrics for higher wind-speed data.

"Designing and Evaluating the Performance of Novel Airfoil Innovations using SOLIDWORKS"

• Modeled a life-sized Boeing 777 in SOLIDWORKS. Incorporated numerous different airfoil designs into the wings, including four of my novel airfoil design innovations. Compared their performance to an approximation of Boeing's proprietary supercritical airfoil design. Found that one of my innovations (the Modified High Camber) provided a significantly higher Lift-to-Drag ratio than any other airfoil tested, including the approximation of the supercritical airfoil.