A BiGRU method for remaining useful life prediction of machinery. (1st January 2021)

• Record Type: Journal Article
• Authors: She, Daoming; Jia, Minping
• Is Part Of: Measurement. Volume 167(2021)
• Publication Date: 2021-01-01
• Subjects: Prediction uncertainty -- RUL -- Bootstrap -- Deep learning; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2020.108277
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: BiGRU model can obtain the past and future degradation state between layers. BiGRU model prediction method can improve the accuracy of RUL prediction. Bootstrap method can express the uncertainty of the deep learning effectively. Abstract: Remaining useful life (RUL) prediction, allowing for mechanical prediction maintenance, reduces the unplanned expensive maintenance greatly. Deep learning methods have provided better point estimation for RUL prediction due to their powerful feature extraction capability. Because of the measurement noise and model parameters, the prediction results usually vary greatly. In order to express the uncertainty of prediction, it is necessary to calculate not only the determined RUL prediction value, but also the confidence interval (CI) of RUL. In this paper, a bidirectional gated recurrent unit (BiGRU) RUL prediction method based on bootstrap method is proposed. The confidence interval (CI) of RUL can be obtained through bootstrap method. The validity of the proposed method is demonstrated by ABLT-1A bearing data. Obtaining the uncertainty in the RUL prediction has great significance for the actual production and manufacturing.