HEALTH MONITORING FOR COMMERCIAL AIRCRAFT SYSTEMS

摘要

This paper presents a methodology for the development of health monitoring solutions for complex systems. The goal of monitoring is early detection of system failures based on present and historic values of signals from built-in sensors and detectors. The methodology is illustrated on an example of health monitoring for an aircraft bleed air control system. The monitoring solution was developed using more than 100 signals collected from a real commercial aircraft over more than 50 flights. The system failures are detected using reasoning on a graphical probabilistic model; more specifically a Bayesian network. The model was created from both knowledge of the bleed air control system and learned from the data. The operation of the monitoring software was verified on test data and resulted in more than a 98% success rate of failure identification.