Trending for aircraft fault detection

In this paper, we present advances of a research to apply multivariable trending as technique for fault detection in commercial aircrafts. The goal is to develop global vehicle health indicators that do not rely on mathematical models yet are capable of detecting process malfunctions. We base our work on the fact that observable system malfunctions leave a distinct trend in the sensors monitoring the process and these trends can. be suitably utilized in identifying the underlying abnormality. Instead of using mathematical models for comparing performances, we rely on extensive data collection of both normal and faulty performance data and develop classification schemes to detect and isolate faults. Here we consider the use of time measurements and test the hypothesis that the normal performance data can be enclosed in a convex set. Using the normal data collected we use a standard algorithm to define its convex hull. Measurements that fall outside the convex set are classified as indicating a fault. We collected data for the different scenarios using a linear model for a B747. The classification using convex hull shows a low rate of false and missed alarms.