Abstract:

To solve the fault prognostic problem caused by small samples, a model based on grey relevance vector machine (RVM) is presented. At the training stage, the discrete grey model (DGM) is established according to the characteristic data sequence, and the model based on RVM regression is trained by using the forecasting values of DGM as input and using the original data sequence as output; At the forecasting stage, a grey RVM model is established by combining DGM and model based on RVM regression, and the information contained in the data are updated through metabolism. The experiment results show that the model has a better performance than conventional grey models.