Underwater Integrated Navigation System Fault Tolerance Method

Precision and reliability are the most important standards to decide the capability of Navigation System. This thesis studies the integrated navigation system of Autonomous Underwater Vehicles (AUV) and the particularity of underwater navigation is considered. Equipped with Strapdown Inertial Navigation System (SINS) as common reference system, Global Positioning System (GPS), Doppler Velocity Sonar (DVS), Magnetic Heading System and Depth Gauge as assistant equipment, the fault-tolerant method of integrated navigation system of AUV is studied. The main work and conclusions are as follows:1. Design of integrated navigation system of AUV based on federated filtering. Underwater integrated navigation system is designed based on federated filtering according to the particularity of underwater navigation. The performance of federated filtering is analyzed and compared with centralized filtering by mathematical simulation. When AUV is on the water, GPS signal is brought in to the system and SINS/GPS sub-filter is added to federated filtering. It shows this can restrain the divergence of SINS well and improve the precision of navigation greatly.2. Study on system fault detection and fault diagnosis. Fault detection module is designed for each sub-filter of federated filter and X~2 test method is used to detect in real time, the performance of two kind of test methods for different fault is analyzed. Fixed period fault diagnosis strategy is designed by using method to weed out outliers when outliers exist in measurement, simulation result shows this will restrain wrong alarm well.3. As the situation that system noise and measurement noise are uncertain, SINS/GPS integrated navigation system is designed based on BP Neural Networks which can approximate well to non-linear system feature and has powerful learning capacity. Simulation result shows the design is feasible when there are high-accuracy training samples and enough learning time. Meanwhile, structure of integrated navigation data fusion base on Neural Networks is put forward.