Abstract
Health assessment of complex systems is critical for Prognostics and Health Management (PHM), enhancing operational reliability and preventing catastrophic failures. However, in engineering systems, different testing intervals for subsystems complicate the evaluation of the system’s overall health state. Excluding data points with brief testing intervals could lead to a significant loss of critical information and compromise the accuracy of the health assessment. Towards this end, we develop a novel information fusion mechanism for non-equidistant data, introducing a federal evidential reasoning (FER) rule-based health assessment method. Firstly, we divide non-equidistant indicators into corresponding subsystems and define the calculation period and fusion period. On this basis, a corresponding non-equidistant information processing method is proposed. Secondly, to reduce the effect of non-equidistant data on the assessment results, the non-equidistant information discount coefficient is introduced to quantify the impact on indicator weight and reliability. Moreover, based on the calculation period and fusion period, the predicted health state and original health state can be obtained by the ER rule-based information fusion mechanism, respectively. Finally, a health assessment experiment of the Laser Inertial Measurement Unit (LIMU) is conducted. Comparative experiments demonstrate that the proposed FER rule-based model significantly outperforms the traditional ER rule-based model and other similar health assessment methods.
Funding Information
  • National Natural Science Foundation of China (62403474)
  • National Natural Science Foundation of China (62203365)
  • National Natural Science Foundation of China (62227814)
  • Shaanxi Province Science and Technology Innovation Team (2020TD-24)

This publication has 33 references indexed in Scilit: