A data-based fault detection scheme for the stratospheric airship control system

Abstract

This brief proposed an innovative fault detection method based on analytical data for the stratospheric airship control system. The control system considered is subject to both space disturbance and nonlinear characteristics; the faults of sensors and actuators are all taken into account. The proposed method is developed in two phases. In the first phase, the moving window kernel principal component analysis is employed to construct the fault detection model with the training data under normal operating conditions and update the fault detection model online until abnormal data are detected. Second, a fault detection model updating mechanism is designed to reduce computational complexity and cost with a clustering algorithm, which compounds the mean shift clustering with weighted Euclidean distance to reflect the data density distribution feature to make the updating to be adaptive. Finally, the proposed method is applied to detect fault for an illustrative simulation stratospheric airship control model. The fault detection results validate the effectiveness of proposed fault detection method for different sensor and actuator fault cases. Comparing to some extended moving window kernel principal component analysis method, the proposed method reduces the computational cost significantly.