Algorithms for Increasing the Fault Tolerance of Strapdown Redundant Inertial Navigation Systems

The fault tolerance of integrated inertial navigation systems is an important problem, which is critical for navigation safety in the air, on ground, and at sea. The paper proposes an approach providing the fault tolerance of redundant strapdown inertial navigation systems (SINSs) in the generation of navigation parameters and focuses on one subproblem, namely, fault detection and isolation (FDI). We compare the performance of 3 FDI algorithms for an integrated system comprising redundant SINSs. Unlike existing algorithms, where the FDI problem is solved at the integrated system level, i.e., using the output navigation and dynamic parameters of the redundant SINSs, in the proposed approach, the faults are diagnosed in the output measurements of the inertial sensors, which reduces the time from fault occurrence to its detection and isolation. The algorithms are based on the pairwise difference method, various modifications of which are proposed by the authors to enhance algorithm performance. Particularly, it is proposed to compute the weighted sum of pairwise differences, the adaptive threshold, and transformation to the principal components. The algorithms are compared using seminatural simulation, where simulated faults are introduced into raw output data. As an example, we consider additive single and multiple faults with random amplitudes. To compare the performance of the proposed algorithms, the confusion matrix is computed and further used to calculate the accuracy and recall rate metrics.