Self-Monitoring Activities for Autonomous Fight Data Analysis

Abstract

With the growing adoption of comprehensive health and usage monitoring and evaluation systems tracking military aircraft operations on a flight-by-flight basis, the need for improved reliability of automated procedures that process data has become a priority. One approach to fulfill this need is for the automated data analysis procedures themselves to be capable of ascertaining when their processing may be in error. This can be achieved through the implementation of self-monitoring activities. In order to capture the level of automation truly necessary to process large quantities of data autonomously, a complimentary capability that allows the self-monitoring activity to learn from its processing experiences by readily incorporating that knowledge into its future processing is also necessary. In this study, self-monitoring activities capable of modifying their functionality autonomously in response to data content are presented. In particular, self-monitoring activities are developed for ensuring the proper operation of monitored signals and algorithms performing signal reconstruction, as well as event detection. The entire adaptive self-monitoring analysis is demonstrated and evaluated using flight data from a set of 47 flights obtained from several naval rotorcraft equipped with an Integrated Mechanical Diagnostic System. Additional results are obtained and evaluated for an unbounded data set represented by over 1,000 flights.