Cross-sensor domain adaptation multi-point monitoring network for mechanical fault diagnosis

Zhijie Zhou, Jiao Lu, Yu Ren, Songbin Li · 2026

Read Cross-sensor domain adaptation multi-point monitoring network for mechanical fault diagnosis by Zhijie Zhou, Jiao Lu, Yu Ren, Songbin Li (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/plos-pone-0358240.txt.

Description

Zhijie Zhou and colleagues published a 2026 methods paper in PLOS ONE on diagnosing machine faults when sensors sit at many different points on the equipment. A model trained at one monitoring point often fails at a new point, because the vibration data look different and labeled fault examples are scarce there. Their network, UCTL, learns from a labeled source point and adapts itself to an unlabeled target point without human tagging.

Inside, a reshaped transformer reads raw one-dimensional vibration signals directly and picks out patterns at several time scales. A new alignment method then pulls source and target data together, balancing average and spread differences so fault classes are less likely to mismatch. Across six transfer tasks on public bearing and gear datasets, the method averaged accuracy above ninety-nine percent. Starting from a pretrained model also cut training time, easing deployment across large monitoring systems.

Catalog record

CREATORZhijie Zhou, Jiao Lu, Yu Ren, Songbin Li
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE67375
LICENSECC_BY
RIGHTSCC BY
SOURCEPLOS

Source: https://doi.org/10.1371/journal.pone.0358240

Browse the full library · About the Tower of Babel