FESeR: A data-driven framework to enhance sensor reliability for the system condition monitoring

Abstract

The collected system information is critical for condition monitoring (CM) which is mainly implemented by utilizing various types of sensors. Hence, the reliability of sensors directly influences the evaluation result of CM. One type of data-driven framework to enhance sensor reliability is realized in this article. To be specific, the combination of sensor selection method and data anomaly detection is achieved by information theory and Kernel Principle Component Analysis (KPCA). The sensors which can provide more valuable information for system CM are selected. The correlations among sensors are analysed by mutual information. Finally, the data anomaly detection is conducted by utilizing the correlations among sensor data sets and KPCA. The effectiveness is proved by employing sensor data sets from National Aeronautics and Space Administration Ames Research Center.