Eventos Anais de eventos
COBEM 2023
27th International Congress of Mechanical Engineering
Quadrotor Fault Detection and Diagnosis using Multi-class Support Vector Machine
Submission Author:
yohan diaz , MG , Brazil
Co-Authors:
Petrus Fiore, yohan diaz, Guilherme Ferreira Gomes
Presenter: yohan diaz
doi://10.26678/ABCM.COBEM2023.COB2023-0179
Abstract
Quadrotors are rotary-wing UAVs capable of vertical take-off and landing. Due to several advantages and numerous applications in all areas of knowledge, their use becomes widespread worldwide. Despite the development of new control algorithms ensuring robustness in relation to disturbances and modeling errors, flight safety in emergency situations as in case of sensor or actuator failure is still a matter of concern to drone manufacturers and is the focus of current research. Active fault tolerant control (AFTC) techniques need a fault detection and diagnosis (FDD) system in such a way that in a timely manner, the controller can be reconfigured and stability could be recovered. The main objective of this work is to use the Multi-class Support Vector Machine (SVM) classification method in order to identify the occurrence of actuator hard-over fault in a quadrotor. A numerical model is used to simulate and collect data. The Principal Component Analysis (PCA) feature extraction technique was in conjuntion with SVM allowing to properly classify the type of fault injected on the system with a high accuracy. Results also demonstrated a high influence of the data collection time on the precision of the method.
Keywords
Quadrotor, Fault Detection and Diagnosis, Support vector machine, Principal component analysis (PCA)
