Critical hydraulic components remaining useful life prediction based on long-life test and Bayesian joint model with data augmentation.

Abstract

Remaining useful life (RUL) predictions of hydraulic components are critical to the operational reliability of hydraulic systems. Currently, research on highly reliable hydraulic components is primarily limited to simulation models, few faulty components or accelerated life tests. Moreover, RUL predictions are mainly limited to multi-source condition monitoring data, potentially leading to difficulties in ensuring long-term RUL prediction accuracy. To address these issues, this paper proposes a RUL prediction method based on long-life test and Bayesian joint model with data augmentation. First, seven solenoid valves were subjected to a long-life test lasting over 2.2 million times for 20 months. Second, a data augmentation method was utilized to increase the size of the RUL prediction training set. Finally, a Bayesian joint model was designed to identify random association relationships among condition monitoring, inspection and event data. The accuracy and confidence of the proposed method has been validated by long-life test datasets.