Fault Detection and Diagnosis of Sensors and Actuators for Unmanned Surface Vehicles

Abstract

This paper describes an algorithm for fault detection and diagnosis (FDD) of unmanned surface vehicles (USV). The algorithm estimates the faults in sensor measurements and actuation force on which autonomous navigation of a USV highly relies: faults in measurements of acceleration and angular rate of an inertial measurement unit (IMU) and thrust force from thrusters. The models of fault dynamics and measurements for FDD are used for the algorithm which is based on Kalman filter, specifically iterated optimal two stage extended Kalman filter. The FDD procedure consists of 4 steps: fault-free filter, fault filter, coupling matrix calculation, and fusion. The fault model for sensor measurements does not depends on damping, so time delay effect dose not affects seriously. On the contrary, the hydrodynamic model for the thrust FDD includes effects by added mass, damping, Coriolis and centrifugal forces. The hydrodynamic effect causes delay in FDD. The algorithm is tested using sea-trial data and simulated data. The implementation is verified for three fault types: bias, oscillation, and drift. The FDD for IMU measurements estimates the fault immediately when fault takes place, while the FDD for thrust force exhibits long delay and high dependency on the model accuracy.