Fault detection of aircraft generator rotating rectifier based on SAE and SVDD method

Abstract

This paper presents a method of fault detection based on the stacked auto-encoder (SAE) and support vector data description (SVDD) for rotating rectifier of brushless ac synchronous generator. The rotating rectifier (RR) is a key component of synchronous generator, and its health state needs to be monitored to ensure the safety of the generator. The method is composed of three steps. In the first step, the health information of RR needs to be collected and preprocessed. In this study, the exciter generator field current is selected as the information source. Fast Fourier Transform (FFT) is firstly used to extract frequency components in this step, and the data information can be compressed this way. Second, the frequency information is input to a SAE, which is trained with some iterations to extract health features. Third, a pre-designed one-class classifier is employed to perform fault detection with the features. The SVDD is selected as the one-class classifier, and in this classifier, the Euclidean distance is chosen as the classification standard in this classifier because this method is direct and simple. The experiment is conducted with a 7KW three-stage synchronous generator test rig, and three load conditions (zero load, 1.5KW load and 3KW load) are considered. The results of experiment demonstrate that the presented method is valid and the fault occurrence can be detected.