Nonlinear Model-Based Fault Detection for a Hydraulic Actuator

This paper presents a model-based fault detection algorithm for a specific fault scenario of the ADDSAFE project. The fault considered is the disconnection of a control surface from its hydraulic actuator. Detecting this type of fault as fast as possible helps to operate an aircraft more cost effective and can help to avoid an undetected increase in fuel consumption. The method proposed here uses an Adaptive Extended Kalman Filter (A-EKF) to detect the disconnection using only local measurements (control signal to the actuator and actuator rod position). For this purpose, an accurate physical model of the hydraulic actuator is needed and the fault is detected by parameter estimation. It is shown that the A-EKF performs better than the regular Extended Kalman Filter (EKF) for this application.