Neural adaptive observer-based sensor and actuator fault detection in nonlinear systems: Application in UAV

Area of Science:

• Aerospace Engineering
• Control Systems
• Artificial Intelligence

Background:

• Unmanned aerial vehicle (UAV) systems rely heavily on sensors and actuators, making fault detection critical for operational safety and reliability.
• Existing fault detection methods may struggle with the complex dynamics and diverse fault types (abrupt, intermittent, incipient) inherent in UAV systems.

Purpose of the Study:

• To develop and evaluate a new online fault detection strategy for UAV sensors and actuators.
• To enhance the accuracy and robustness of fault detection by utilizing adaptive neural network parameters.

Main Methods:

• Implementation of a Neural Network (NN) with weighting parameters updated via an Extended Kalman Filter (EKF).
• Application of the proposed fault detection system to a nonlinear dynamic model of the WVU YF-22 unmanned aircraft.
• Comparative analysis against conventional recurrent neural network-based fault detection strategies.

Main Results:

• The proposed EKF-updated NN strategy demonstrated accurate detection of abrupt, intermittent, and incipient faults.
• Simulation results indicated superior performance of the new method compared to traditional recurrent NN approaches.
• The online adaptation capability of the NN weighting parameters was key to the enhanced detection performance.

Conclusions:

• The developed online detection strategy offers a significant improvement for UAV fault detection.
• The integration of EKF with NN provides a robust framework for identifying diverse sensor and actuator faults in UAVs.
• This approach enhances the safety and reliability of unmanned aerial vehicle operations.