Factory Analytics
Adaptive Fault Detection and Prognostics
Self-learning and transferable AI for detecting faults, predicting failures, and adapting across machines, products, and operating conditions.
Continual Domain AdaptationSource-Free Domain AdaptationOpen-Set GeneralizationTarget-Free GeneralizationDiffusion-Assisted Diagnosis
Domain shifts in manufacturing
Fault diagnosis models often fail when machine type, operating condition, product specification, sensor setting, or factory environment changes. Adaptive learning addresses this domain-shift problem.
Five research directions
We investigate continual domain adaptation, source-free domain adaptation, open-set domain generalization, domain generalization without target adaptation, and diffusion-assisted cross-domain fault diagnosis.
Toward reliable prognostics
The goal is to build fault detection and prognostics models that remain reliable under unseen faults, limited labels, privacy constraints, and changing production domains.