9.3 Aircraft Maneuver Prediction with Machine Learning Applications
- Event
- ettc2022 - European Test and Telemetry Conference
2022-05-10 - 2022-05-12
Nuremberg - Chapter
- 9. Machine Learning & Artificial Intelligence
- Author(s)
- E. Cakici, O. Yörük - Turkish Aerospace, Kahramankazan Ankara (Turkey)
- Pages
- 214 - 219
- DOI
- 10.5162/ettc2022/9.3
- ISBN
- 978-3-9819376-6-4
- Price
- free
Abstract
The flight tests are an important phase of aircraft development programs. Currently, parameters of manned or unmanned test flights are analyzed by the test engineers. Maybe the process is still best choice for the reliability concerns, but on the other hand it is time consuming. So we want to propose flight maneuver predictor with using Machine Learning techniques. For this purpose, a collected dataset of a fixed-wing propeller aircraft is used. A machine learning model was created that can predict seven different maneuvers using the gathered data. During flight test every test maneuver's start and stop time tagged and labeled as test points by the flight test engineers. These labels are Takeoff, Landing, LSS, Phugoid, Loop, Wind Up Turn, and Aileron Roll. After gathering data, preprocessing is performed such as fixing row size of all attributes by timestamps. Also some other attributes which had less frequent data than the others reproduced. For the creation of prediction model, support vector machine (SVM) applied. Overall prediction score of the model is 0.90.