Remaining useful life prediction for CT X-ray tubes based on multi-dimensional and multi-domain feature fusion

Abstract

Remaining useful life (RUL) prediction of X-ray tubes is crucial for ensuring the reliable operation of computed tomography (CT) equipment and improving the quality of medical services. However, existing RUL prediction methods for X-ray tubes face challenges in extracting complex degradation information. To address these challenges, This paper proposes a novel RUL prediction method for CT X-ray tubes based on multi-dimensional and multi-domain (MDMD) feature fusion network. First, a parameter construction technique is developed to uncover hidden degradation information between different parameter combinations. Next, a MDMD feature extraction network is constructed, which extracts features from time, frequency, and spatial domains to comprehensively capture multi-dimensional data characteristics. In this regard, a feature fusion module is introduced to enhance the focus on key degradation features. Additionally, a segmented weighted loss function is designed to prioritize data from the degradation phase during model training. Experimental results demonstrate that the proposed method significantly outperforms several state-of-the-art prediction methods in terms of root mean square error, mean absolute error, and other evaluation metrics. The proposed method can assist the equipment maintenance team of hospitals in predictive maintenance of medical imaging equipment.