Abstract: With the rapid increase in the number of spacecraft in orbit, the growing complexity and diversity of mission requirements and space environment, fault diagnosis and recovery are critical for ensuring spacecraft safety and sustainable operations. However, existing fault diagnosis and recovery methods are typically designed for specific fault types and are stored as fixed code on board, often requiring substantial manual intervention and software maintenance, especially in the event of unforeseen in-orbit failures. These methods are still far from achieving full autonomy. To address this issue and enhance the spacecraft's resilience to unknown faults, this paper proposes an Event-triggered Adaptive Fault Diagnosis and Recovery (EAFDR) framework. EAFDR is based on fault event trees and employs a hierarchical fault event analysis technique, prioritizing fault event trees by severity for diagnosis. It then generates fault recovery strategies through Event-Condition-Action (ECA) rules, enabling real-time responses and dynamic monitoring during recovery to ensure system safety. Furthermore, EAFDR decouples the fault diagnosis and recovery processes from the control cycle, allowing for dynamic modification and maintenance of the fault event tree and ECA rules. Ground-based simulation results from a real-world system demonstrate that EAFDR can diagnose and recover from spacecraft faults in orbit with acceptable computational overhead, providing effective support for the safe and stable operation of spacecraft.

Key words: control software, fault diagnosis and recovery, on-orbit adaptation, event-triggered, fault event tree, event-condition-action rule