Active fault-tolerant control with prescribed performance and reachability judgement for the altitude ground test facility

• Published in: Aerospace science and technology Vol. 155; p. 109653
• Main Authors: Lun, Yuebin, Wang, Honglun, Hu, Jintao, Yan, Guocheng, Wu, Tiancai, Ren, Bin, Qian, Qiumeng
• Format: Journal Article
• Language: English
• Published: Elsevier Masson SAS 01.12.2024
• Subjects: Altitude ground test facility; Fault tolerant control; Neural network; Prescribed performance control; Reachable set approximation; Reachable set approximation; Neural network; Altitude ground test facility; Fault tolerant control; Prescribed performance control
• ISSN: 1270-9638
• DOI: 10.1016/j.ast.2024.109653

•A fault diagnosis method based on the long short-term memory (LSTM)neural network is designed.•A reachability judgment network is proposed to ensure the safety of system in real time.•An asymmetric appointed-time prescribed performance control is proposed to improve the control performance. This research proposes an active fault-tolerant control (AFTC) method based on reachability judgement with asymmetric appointed-time prescribed performance to ensure the safety of altitude ground test facility (AGTF) under potential valve actuator and sensor faults. Firstly, a long short-term memory (LSTM) neural network is employed for fault diagnosis. Given the unique characteristics of AGTF, the equivalent multiplicative fault is introduced to ensure the successful fitting of the fault diagnosis network. Then, a reachability judgement network based on the distance field on grids (DFOG) method is designed to assess the impact of faults on subsequent tasks of the system, thereby ensuring the safety of the AGTF. Subsequently, this study proposes an error transformation function of prescribed performance control (PPC) that ensures error convergence even in cases of asymmetric performance function. This enables the performance boundaries to be further reduced to improve the control performance of the system. Finally, experiments validate the effectiveness of the methods proposed in this manuscript.