Multistage Degradation Prediction Based on Degradation Pattern Recognition and Long Short-Term Memory

The degradation of equipment is characterized by randomness and complexity. For some equipment, there are many degradation stages, and the degradation law of each stage is quite different. Aiming at the problem of multi-stage degradation prediction, this paper proposes a method combined with degradation pattern recognition and life prediction method. Based on the degradation stage identification, life prediction for each stage is carried out separately, and an oxygen concentrator is taken as an example to verify the method. The results show that the prediction accuracy of oxygen concentrator life by this method is much higher than that by direct prediction without subsection.