Research on life prediction based on wavelet transform and ARMA model for space relay

The electrical life is an important index to evaluate the reliability of relay, and it is closely related with many characteristic parameters of relay such as contact resistance, pick-up time, over-travel time, etc. By using the time series analysis and by taking some characteristic parameters as predicted variables, the life of relay can be obtained by the life prediction which is a nondestructive and short-cycling life test method. However, the predicted accuracy is greatly influenced by the complex variations of characteristic parameters, and as a result it sometimes becomes too low to be accepted. For the purpose of improving the predicted accuracy, a life prediction method for space relays based on wavelet transform and ARMA (auto-regression moving average) time series is proposed in this paper. Taking the pick-up time of a relay (which can be regarded as a nonstationary time series) as an example, we divide it into three parts- trend term, cyclic term and stochastic term by adopting wavelet decomposition and reconstruction. Afterwards, models of each term are built with different methods and then they are employed to predict the future value, respectively. Finally, the predicted results of the original time series are obtained by superimposing the respective prediction of each term. The comparison between the method used in this paper and the conventional regression analysis method indicates that the former has higher predicted accuracy.