New forecasting method of closing time for aerospace relay in storage accelerated degradation testing

Abstract

Space relays are affected by many nonlinear elements during storage, and the reason for predicting time series is to achieve nonlinear mapping. Combining artificial neural networks and grey system theory, we built a grey artificial neural network (GANN) model. The model effectively combined the characteristics of artificial-neural-network nonlinear adaptability and the characteristics of grey theory weakening data sequence volatility integration. We predicted the degradation value of the closing time of measured data in a relay accelerated storage test by using a variety of grey models and GANN models. By comparing several forecasting methods, the results showed the proposed grey neural network model has higher precision and is more accurate than a single grey model. The method also provides new ideas and methods for the life prediction of relay storage acceleration tests.