An air turbine starter (ATS) is used to start the aero-engine before an aircraft takes off, which plays a significant role in the reliable operation of the aero-engine and is critical to the flight security, so it is vital to monitor the health and predict the remaining useful life (RUL) for the ATS. This paper proposes a fusion framework based on the combination of empirical mode decomposition (EMD) and relevance vector machine (RVM). EMD is used to smooth out the non-stationary data by pattern decomposition, and the multiple intrinsic mode functions (IMF) which can effectively reflect the fault characteristics, are carefully selected from all IMFs by kurtosis index technique. RVM is used to train the selected smooth IMFs samples and establish a regression model for remaining useful life prediction. In addition, the subtraction clustering technique is introduced to reduce the samples scale and speed up the RVM’s training efficiency. The effectiveness of the proposed fusion framework is illustrated via an experiment of ATS, and the results show that the proposed method has satisfactory prediction performance.
A Fusion Prognostic Method for Remaining Useful Life Prediction Based on an Extended Belief Rule Base and Particle Filters
