Fault Detection and Diagnosis for Aerostat Sensors Based on PCA and Contribution Graph

In order to solve the sensor fault diagnosis problem of multi-propeller aerostats, considering the factors of strong coupling, nonlinearity and model uncertainty, this paper adopts principal component analysis (PCA) and traditional contribution graph as the fault detection and diagnostic methods. Firstly, each sensor data under normal operating conditions is collected as training data to establish a principal component model. The normal data is then projected into the principal component space and the residual space, and the statistical control limits under normal conditions are calculated. Finally, the data collected by the sensor in real time is projected into the principal component space, and the real-time statistics and control limits are judged to determine the presence or absence of the fault and the type of fault. When the fault occurs, the specific position of the fault is further determined by calculating the contribution value of each sensor variable to the statistic. The simulation results show that the fault diagnosis method based on principal component analysis and contribution graph can detect and diagnose the fault of the aerostat sensor in real time.