容错
计算机科学
基础(拓扑)
可靠性理论
可靠性(半导体)
控制理论(社会学)
可靠性工程
控制(管理)
自适应控制
断层(地质)
控制系统
故障检测与隔离
故障指示器
状态监测
错误检测和纠正
冗余(工程)
控制工程
作者
Qiangqiang He,Zhichao Feng,Changhua Hu,Zhijie Zhou,Yijie Sun
标识
DOI:10.1109/tr.2026.3750651
摘要
In this paper, a dynamic fault diagnosis and fault-tolerant control framework with adaptive memory length adjustment is developed for sensor failures based on the belief rule base (BRB) expert system. The framework addresses three key challenges in practical systems: limited high-value fault samples, high system complexity, and strict real-time requirements. While previous studies have explored BRB structure reduction, these works have mainly focused on static model simplification or constrained interpretability, lacking mechanisms for dynamic adaptation. In contrast, the proposed framework integrates BRB-based fault detection and sensor-output reconstruction with an online adaptive updating mechanism, where the active rule structure and memory length are dynamically adjusted according to diagnostic and reconstruction performance. This integration enables joint utilization of expert knowledge and observation data to handle nonfatal sensor failures efficiently, ensuring high fault diagnosis accuracy and online fault-tolerant control. The effectiveness and practical applicability of the framework are demonstrated through a case study on accelerometer failures in a rocket inertial navigation system. The proposed framework achieves an average fault diagnosis accuracy of 95.71%, with a false alarm rate of 2.12% and a missed detection rate of 2.17%. Moreover, the average runtime is reduced to 16.8 s per cycle, demonstrating its capability to achieve a favorable balance between diagnostic accuracy and real-time performance.
科研通智能强力驱动
Strongly Powered by AbleSci AI