As industrial systems become increasingly digital, connected, and data-rich, predicting degradation and planning maintenance have become more challenging. In prognostics and health management (PHM), degradation modeling plays a central role in linking condition monitoring, health assessment, lifetime prediction, and maintenance decision-making.

This review focuses on the inverse Gaussian (IG) process, a stochastic process well suited to describing monotonic and cumulative degradation. Compared with other common degradation models, the IG process offers clear physical interpretability, analytical tractability, and the ability to capture heavy-tailed degradation behavior, making it useful for reliability analysis and remaining useful life prediction.

The paper systematically reviews the theoretical foundations of the IG process, major model extensions, and common inference methods. It also summarizes representative applications in accelerated degradation testing, burn-in testing, remaining useful life prediction, and maintenance optimization.

A key contribution of the review is its discussion of how AI can be integrated with IG-process-based modeling. Rather than replacing statistical models, AI can enhance their flexibility and data adaptability while preserving probabilistic consistency and interpretability. The review highlights promising directions such as AI-assisted parameter learning and physics-informed deep IG modeling, and points to future opportunities in multi-source data fusion, uncertainty-aware prediction, cross-domain transfer, and real-time PHM deployment.

The paper also points to several opportunities for future research, including modeling multi-source heterogeneous data, improving uncertainty-aware prediction, enabling cross-domain knowledge transfer, and developing scalable methods for real-time PHM deployment. Overall, the review presents the IG process as a unified and interpretable probabilistic foundation for intelligent maintenance, while suggesting that its integration with AI could further improve flexibility, scalability, and decision support in next-generation PHM systems.

The work titled “From statistical modeling to AI-integrated inverse Gaussian process: A comprehensive review for prognostics and health management” was published in ENGINEERING Management in Feb. 2026.
DOI: 10.1007/s42524-026-5388-8