Fault Diagnosis for China Space Station Circulating Pumps: Prototypical Network with Uncertainty Theory

• DOI: 10.3390/sym15040903
• Publication Date: 2023-04-13T08:33:26Z
• Authors: Wenbo Wu; Tianji Zou; Dong Guo; Lu Zhang; Ke Wang; Xuzhi Li
• Keywords: Similarity (geometry); Representation; Uncertainty Quantification

Methods for fault diagnosis based on metric learning, in which a query sample is classified by picking the closest prototype from support set their feature similarities, have been subject of many studies. In real-world applications in-orbit products, such as circulating pumps, computation similarity between different pairs prone to degrees inaccuracy, especially epistemic uncertainty. Knowing and considering uncertainty may improve detection accuracy. This article provides unique approach Prototypical Network (Pro-Net) Uncertainty Theory. particular, we use altering representation prototypes deterministic scalar an uncertain representation. To assess set, calculate distance using cross-entropy. Experiments with symmetrical structures reveal that our proposed method significantly enhances classification precision achieves state-of-the-art performance. It improves reliability reduces risk making erroneous judgments safety-critical systems, decreasing possibility adverse consequences.