A generalized degradation tendency tracking strategy for gearbox remaining useful life prediction. (January 2023)

• Record Type: Journal Article
• Authors: Chen, Xieyi; Wang, Yi; Sun, Haoran; Ruan, Hulin; Qin, Yi; Tang, Baoping
• Is Part Of: Measurement. Volume 206(2023)
• Publication Date: 2023-01
• Subjects: Health indicator -- Degradation tendency tracking -- Tradeoff loss function -- Remaining useful life; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2022.112313
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: A generalized degradation tendency tracking strategy for gearbox RUL prediction. A tradeoff loss function is used for iterative training of the deep model, which provided a robust feature learning strategy. An improved health indicator fusion method to obtain HI with good tendency. The experimental results indicate the proposed strategy can apparently improve the prediction performance of the RUL prediction models. Abstract: Gear is an important component of mechanical equipment. Its health state will affect the operation of the whole equipment. Therefore, the remaining useful life (RUL) prediction of gearbox is very important. However, most of the current deep learning-based RUL prediction methods inevitably focus too much on the fluctuation characteristics of degradation data, while not capturing trend characteristics well. In order to obtain accurate and reliable prediction results, a generalized degradation tendency tracking strategy (GDTTS) for gearbox RUL prediction is proposed. Firstly, a health indicator (HI) of gearbox degradation process is constructed based on improved HI fusion method. Then, a tradeoff loss function (TLF) is proposed to guide the feature learning of the model. The proposed TLF enables the model to grasp the tendency characteristics of the data itself and robust to singularities which can cause the entire prediction model to deteriorate. Finally, a new end-of-life determination criterion is established. The prediction results on the actual gearbox datasets show that the proposed strategy is a generalized strategy for gearbox RUL prediction which can apparently improve the prediction performance of the life prediction models.