We present two alternative methods for fault detection and isolation (FDI) with redundant Microelectromechanical system (MEMS) inertial measurement units (IMUs) in inertial navigation systems (INS) based on nonlinear observers (NLOs). The first alternative is based on the parity space method, while the second is expanded with quaternion-based averaging and FDI. Both alternatives are implemented and validated using data gathered in a full-scale experiment on an offshore vessel. Data from three identical MEMS IMUs and the vessel's own industrial sensors are used to verify the methods' FDI capabilities. The results reveal that when it comes to FDI of the IMUs' angular rate sensors, there are differences between the two methods. The navigation algorithm based on quaternion weighting is essentially unaffected by the failure of an angular rate sensor, while the parity-space-method-based alternative experiences a perturbation.
Read full abstract