A Bayesian Framework With Omni-Dimensional Normalizing Flows for Infrared Thermography-Based Machine Prognostics
IEEE Transactions on Instrumentation and Measurement · 2026 · 75 · 3514912-3514912
DOI: 10.1109/tim.2026.3697048 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Instrumentation and Measurement, 2026. Verfasst von Yimin Jiang und 5 weiteren Autor:innen. Die Arbeit nennt 30 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Natural Science Foundation of Shanghai, Ministry of Education-China Mobile Research Foundation.
Abstract (Originalsprache)
Infrared thermography (IRT) has attracted increasing attention in noncontact condition monitoring and prognostics. However, the high dimensionality and spatiotemporal degradation process in infrared images present a significant challenge for uncertainty quantification of IRT-based prognostic models. In this article, a variational Bayesian framework with omni-dimensional normalizing flow (VBONF) is presented. First, an ONF mechanism is proposed to construct flexible approximate posterior distributions. It improves the quality of variational inference (VI) by leveraging multidimensional convolutional kernel space to learn auxiliary random variables along all dimensions of the kernel space. These variables are exploited to construct a family of compound approximate posteriors. Furthermore, a Sigmoid inverse autoregressive transformation (SIAT) method is designed to enable successive invertible mappings with trivial determinants of the Jacobian. SIAT stably controls the range of auxiliary variables to make optimizing Bayesian models smoother. Experiments using an infrared image dataset demonstrate that the proposed VBONF-based prognostic framework is effective and superior. The experimental and simulated infrared degradation image datasets are used to evaluate the proposed approach, which implements considerably lower prediction errors and higher effectiveness of uncertainty quantification in comparison with advanced Bayesian-based prognostic models.
Förderung
- National Natural Science Foundation of China
- Natural Science Foundation of Shanghai
- Ministry of Education-China Mobile Research Foundation
- National Key Research and Development Program of China
Zitieren
APA
Jiang, Y., Xia, T., Xu, Y., Li, R., Wang, D., & Xi, L. (2026). A Bayesian Framework With Omni-Dimensional Normalizing Flows for Infrared Thermography-Based Machine Prognostics. IEEE Transactions on Instrumentation and Measurement, 75, 3514912-3514912. https://doi.org/10.1109/tim.2026.3697048
Vancouver
Jiang Y, Xia T, Xu Y, Li R, Wang D, Xi L. A Bayesian Framework With Omni-Dimensional Normalizing Flows for Infrared Thermography-Based Machine Prognostics. IEEE Transactions on Instrumentation and Measurement. 2026;75:3514912-3514912. doi:10.1109/tim.2026.3697048
BibTeX
@article{Jiang2026,
author = {Jiang, Yimin and Xia, Tangbin and Xu, Yuhui and Li, Rourou and Wang, Dong and Xi, Lifeng},
title = {A Bayesian Framework With Omni-Dimensional Normalizing Flows for Infrared Thermography-Based Machine Prognostics},
journal = {IEEE Transactions on Instrumentation and Measurement},
year = {2026},
volume = {75},
pages = {3514912--3514912},
doi = {10.1109/tim.2026.3697048}
}
EUVIMED