Health Indicator Extraction for Electro-Mechanical Actuator with CHMM

Abstract

Electro-Mechanical Actuator (EMA) has played an important role as there are more EMAs incorporated in the flight control actuation of More Electric Aircraft. However, the difficulty of EMA Health Indicator (HI) extraction caused by limitation of sensor installation hinders the development of EMA Prognostic and Health Management (PHM). As a result, to address this issue, a new HI extraction method based on Continuous Hidden Markov Model (CHMM) is proposed for EMA. In the CHMM-based HI extraction method, the monitoring data in health condition of EMA are utilized to train a CHMM with log-likelihood function. Based on the CHMM and the monitoring data in the degradation condition of EMA, the output of log-likelihood function for EMA degradation condition can be obtained, which implies the similarity between the degradation condition and health condition. Furthermore, the normalized similarity is used as an EMA HI. Thus, the sensors with high correlation to the extracted HI (i.e. normalized similarity) no longer need to be installed or can be removed. This study provides a new way of EMA HI extraction with the limitation of sensor installation. To validate the effectiveness of CHMM-based HI extraction method for EMA, experiments are conducted, in which the data derived from NASAs Flyable Electro-Mechanical Actuator (FLEA) test stand are utilized. Experimental results show that the CHMM-based method has a good performance in EMA HI extraction.