Sensor Fault Detection in nonlinear system using Adaptive Weighted Fusion Algorithm

In view of the problem that the state estimation contains fault values when detecting sensor faults containing noise based on the Kalman filter algorithm, a sensor fault detection method based on adaptive weighted fusion is proposed in this paper. The adaptive weighted fusion algorithm gives different weights to several groups of measurement output data to generate the measurement output without faults. Then, the composite embedded cubature Kalman filter algorithm is used to get the state estimation without faults, which will make the mean or variance of the generated residuals change, so as to realize fault detection. Finally, the single-link direct drive manipulator system driven by permanent magnet brush DC motor is simulated and verified. The results show that, compared with the CECKF algorithm alone, the proposed method can eliminate the faults contained in the Kalman filter state estimation, and detect all the output sensor faults more effectively.