Fault diagnosis of gravitational wave detection system operation with limited computing resources using semi-qualitative symbolic directed graph model

Recently, there has been significant progress in the field of gravitational wave detection. China is planning to launch Taiji II in 2024, aiming to establish a muti-probe system in a predetermined orbit. Similarly, the United States' LISA project aims to complete the scientific research and development of a three-spacecraft system by 2035. However, due to the special detection environment, including serious electromagnetic interference and other uncertain factors, various system failures occur from time to time, and missions will also be affected or even interrupted. Therefore, it is crucial for the system to promptly diagnose faults under conditions of long communication time with ground control and limited on-board resources in order to ensure safe operation. In this paper, a semi-qualitative model-based fault diagnosis method is proposed. Firstly, a qualitative model of a single spacecraft for gravitational wave detection is built based on the improved symbolic directed graph method. By defining nodes of different stages in the fault model, part of quantitative information is given to the directed graph of the system and the quantitative information of the system is transformed into different stages of the qualitative model, thus solving the shortcoming of too much information missing in the process of qualitative model modeling. Subsequently, leveraging this qualitative model information from each spacecraft involved, we establish a collaborative fault-diagnosis model considering limited computing resources. This comprehensive model incorporates constraints related to single spacecraft computations as well as inter-spacecraft interactions and different fault modes while fulfilling prerequisite requirements for diagnosing faults within gravitational wave detection systems. Finally, based on the qualitative constraints among the nodes of the system model, the incompatible branches are identified through a combination of forward propagation and backward tracking to determine the set of fault sources, thereby completing the ranking of fault possibilities for potential fault sources. The integration of forward propagation and backward tracking in diagnosing results addresses the limitation problem associated with unidirectional search for minimum diagnosis sets in traditional diagnostic methods. Moreover, it eliminates redundant and incorrect incompatible branches from a qualitative logic perspective, effectively enhancing the accuracy of the final diagnostic solution. Simulation results also demonstrate that by injecting various fault modes, this proposed method can promptly respond to faults in gravitational wave detection systems and identify their sources even under limited computing resources conditions. Consequently, it improves diagnosis efficiency compared to traditional model-based approaches.