The oxygen concentrator is an important airborne part of the aircraft life support system, and its performance degradation data has the characteristics of multivariate and strong noise. In order to solve the problem of lack of univariate information and low prediction accuracy in the prediction of oxygen concentrator life, the correlation between multi-dimensional degradation variables was considered to select suitable variables. By introducing the influencing factor of oxygen concentration into the oxygen partial pressure variable, the long-term trend in the sequence was extracted with the Hodrick Prescott (HP) filter, and the fuzzy C-means (FCM) method was used to stage the oxygen partial pressure to establish a multi-stage dynamic regression model. Modeling the degradation of an oxygen concentrator. The findings indicate that the multivariate autoregressive integrated moving average (ARIMA) model with HP filtering improves prediction accuracy by 98.20%, 81.21%, and 77.87%, respectively, when compared to the univariate ARIMA model, single-stage multivariate ARIMA model, and multi-stage dynamic regression model.