A fault diagnosis approach based on PCA and dag-SVM for hydrostatic fluid gyro platform system

Hydrostatic fluid gyro platform systems (called as inertial platform system) are widely applied in missiles and other weapons. Due to the complex structure composition of the inertial platform system, the traditional fault diagnosis methods are difficult to achieve rapid and accurate fault location. In this paper, a new fault diagnosis method based on Directed Acyclic Graph and Support Vector Machine (DAG-SVM) are presented for fault diagnosis. The DAG-SVM fault diagnosis model is established based on the stabilization loop of inertial platform system, and Principal Component Analysis (PCA) method is used to reduce the dimensionality of measured datasets. The multi-class classification model of SVM is trained using the measured datasets of stabilization loop, and these classifiers trained by each two types of the training samples are used as the root node of the DAG to construct a complete fault diagnostic model. Simulation results show that the DAG-SVM fault diagnosis model has higher accuracy in diagnosis, and faster training and testing in velocity. It can achieve better diagnosis results with limited training datasets, which overcomes the problem of model inaccuracy while the traditional neural network trains with the same testing datasets.