Real-time life prediction of equipment based on optimized ARMA model

Prediction with large error by traditional Autoregressive Moving Average (ARMA) theory has long been hampering accuracy in the life prediction. In this paper, methodology based on optimized ARMA model is proposed to provide real-time life prediction for equipment by utilizing information of degradation of similar equipment. Firstly, average relative change is used to optimize the orders of ARMA model, and the optimal model parameters are obtained. Afterwards, the ARMA model for similar equipment is established to get its degradation path sets. Then we get the degradation path sets of similar equipment that has the maximum similarity with specific equipment degradation path by K-means clustering. After that we get the specific equipment degradation path by weighting the equipment degradation path sets which are obtained by figuring out K-means clustering center with least similarity. By this algorithm, we can update the degradation path through real-time measured value, so as to predict the life of equipment timely. To test the model, operating current degenerating data of a laser is applied in this case study, and our study revealed that predicting accuracy using optimized model is evidently better than using model based on function degradation data of single equipment.