Fault diagnosis of underwater vehicle based on improved continuous-discrete unscented Kalman filter

To diagnose the faults of an underwater vehicle's continuous nonlinear system, we propose an improved continuous-discrete unscented Kalman filtering (CDUKF) algorithm. Firstly, according to the estimation mean values and covariance matrix of the state of the continuous nonlinear system and the parameters of its faults, we construct its unscented state matrices and forecast its unscented state array through integrating the unscented state differential functions. Then we obtain the estimation mean values and covariance matrix of its state and parameters through calculating the mean values and updating the estimation. Secondly, we fuse the faults of the actuator of the underwater vehicle into the continuous nonlinear system in the form of proportional coefficient or affixation parameter and estimate the parameters of the faults of the actuator through constructing respectively the state and the parameters of the CDUKF, thus diagnosing the faults of the actuator. Finally, to diagnose the faults of the actuator in simulating the docking of the underwater vehicle on a horizontal plane, we use the CDUKF algorithm to effectively estimate the faults of the underwater vehicle in their parameter form, thus verifying its feasibility and effectiveness.