Two papers accepted for presentation in 2025: I-GLIDE at ECML PKDD in Porto, and a complementary uncertainty-quantification paper at MIMAR in Nancy. Both explore health indicators derived from latent spaces in complex degradation processes.
I-GLIDE: Input Groups for Latent Health Indicators in Degradation Estimation
Lucas Thil, Jesse Read, Rim Kaddah, Guillaume Doquet
Accepted at ECML PKDD 2025, taking place in Porto this September. The paper explores how structured input grouping can improve latent health indicator learning in complex degradation processes.
Uncertainty Quantification as a Complementary Latent Health Indicator
Lucas Thil, Jesse Read, Rim Kaddah, Guillaume Doquet
Accepted at the 13th IMA International Conference on Modeling in Industrial and Reliability (MIMAR 2025), held in Nancy, 7 to 10 July. The work highlights how uncertainty measures can complement latent health indicators for more robust remaining-useful-life prediction on turbofan engines.
Across both papers, we investigated health indicators derived from latent spaces, enhanced with uncertainty quantification. Our findings suggest that it is the quality of health indicators, not necessarily the complexity of the models, that drives prognostics performance.
We had the opportunity to present these contributions at the JNI Days alongside industrial partners at CEA and IRT SystemX, and we are grateful for the discussions that followed.