A Synergistic Fault-Tolerant Flight Control System based on Adaptive Neural Networks and an Enhanced Adaptive Super-Twisting Algorithm

Abstract

In This study introduces a sophisticated fault-tolerant flight control system designed to address the challenges of disturbance rejection and sensor fault compensation in aircraft systems, particularly under conditions of high angle of attack and severe noise. The proposed framework integrates an Adaptive Super-Twisting Observer (ASTO) with an Adaptive Neural Observer enhanced by an Extended Kalman Filter (ANO-EKF), complemented by a robust Backstepping control strategy that incorporates a smooth switching mechanism. A central innovation of this study lies in the development of the adaptive super-twisting observer, which improves estimation accuracy by dynamically adjusting observer gains according to system behavior. This approach significantly enhances disturbance rejection. The proposed method increases system robustness without requiring prior knowledge of disturbance bounds, ensuring superior adaptability in sensor noise and faults. To further enhance fault tolerance, the ANO-EKF is introduced, leveraging a neural network for adaptive state refinement, while the EKF ensures precise fault detection and compensation. This integrated approach guarantees reliable performance even under severe sensor faults. Additionally, a Backstepping-based control strategy is employed, incorporating a sigmoid-based smooth switching function to ensure seamless transitions between actual and estimated states, effectively reducing transient oscillations and minimizing control disruptions. Extensive nonlinear dynamic simulations using an F-18A fighter jet model highlight the enhanced effectiveness of the proposed approach in comparison with conventional methods. The outcomes indicate that incorporating ASTO into the Fault-Tolerant Control (FTC) framework enhances tracking accuracy by 48.3% and reduces chattering by 5.7%, underscoring its suitability for practical aerospace applications.