An autoregressive model-based degradation trend prognosis considering health indicators with multiscale attention information

Prognostics and health management of rotating machinery plays an integral role in providing critical monitoring information for stable equipment operation. Thus, an autoregressive model-based prediction of bearing degradation trends considering health indicators with multiscale attention information is proposed. Specifically, a multiscale attention network is developed to construct a health indicator with the representation of degradation trends. Then, a comprehensive evaluation metric is designed to evaluate the metrics of the constructed health indicators. Also, an improved autoregressive integrated moving average model considering the elimination of heteroskedasticity attribute is proposed to predict the degradation trend of health indicators, and the order of the model are determined using the Bayesian information criterion. Finally, the degradation trend of the test set is predicted using the derived model. The results show that the prediction method constructed by a multiscale attention network considering health indicators can effectively utilize the vibration data to predict the degradation trend of bearings. Compared with other related methods, the proposed method has obvious advantages in the health monitoring of rotating machinery.