Dynamic multi-fault diagnosis model based on support vector domain description

To improve the adaptability of multi-fault diagnosis, a novel dynamic diagnosis model is proposed, which includes two main techniques; independent training and online adjusting for multi-fault modes. The data of every fault class are trained by support vector domain description (SVDD) to obtain the optimal enclosing feature spaces. Due to the independence of these distribution spaces, new classes can be generated facilely. The relative distances between fault data and the distribution spaces decide which class they belong to. Moreover, an online SVDD algorithm is developed to update the feature space, and the distribution information of new data is appended to the existing feature space timely. And the simulated and practical data are analyzed to verify the effectiveness of the model for dynamic multi-fault diagnosis.