Aircraft Health Monitoring System Using Interacting Multiple Model Estimation

Abstract

An effective approach for modeling and simulating failure detection and identification (FDI) of aircraft components is presented. A telescoping main landing gear and pressure control system is modeled and simulated subject to various fault conditions. Systems subject to faults cannot be modeled well by a single set of equations. A more appropriate model would be a model whose state not only varies continuously, but may also jump from one state to another, which is the so-called stochastic hybrid system. The use of multiple models for FDI where each model represents a fault or the nominal mode fits well into such a system. The interacting multiple-model (IMM) estimation algorithm is one of the most effective approaches for FDI. It is able to detect and identify multiple faults more quickly and reliably then many existing approaches. It runs a bank of Kalman filters in parallel and switches from one model to another in a probabilistic manner. In this thesis, the dynamics of the landing gear and pressure control system are modeled as a stochastic hybrid system where FDI based on the IMM estimation algorithm is simulated and performance is evaluated. Simulation results show that FDI based on IMM can detect and identify sensor failures and failures in system components.