I-GLIDE: Input Groups for Latent Health Indicators in Degradation Estimation
Lucas Thil, Jesse Read, Rim Kaddah, Guillaume Doquet
2025-11-27
Summary
This paper introduces a new way to predict how much longer a machine or system will function reliably, known as its remaining useful life (RUL). It focuses on improving the 'health indicators' used to make these predictions, especially when dealing with complex systems that have many sensors.
What's the problem?
Currently, predicting RUL is difficult because it's hard to figure out *how* things are breaking down within a complex system with lots of sensors. Existing methods struggle to separate different failure causes and don't always give a good sense of how trustworthy the health indicators actually are. Basically, we need a better way to understand what's going wrong and how confident we can be in our predictions.
What's the solution?
The researchers developed a framework called I-GLIDE. It uses a technique called RaPP to create health indicators that are better at spotting degradation than older methods. They also added ways to measure the uncertainty in these indicators, using methods like Monte Carlo dropout and probabilistic latent spaces. The key innovation is grouping sensors together to focus on specific parts of the system, allowing them to pinpoint exactly *where* the problems are occurring and understand the specific failure mechanisms. This makes the predictions more accurate and easier to understand.
Why it matters?
This work is important because it improves our ability to predict when machines will fail, which is crucial for things like aerospace and manufacturing. By not only predicting *when* something will break, but also *why*, it allows for more targeted maintenance and prevents unexpected downtime. It also bridges the gap between simply detecting problems and actually predicting how long something will last, offering a more complete and reliable approach to system health monitoring.
Abstract
Accurate remaining useful life (RUL) prediction hinges on the quality of health indicators (HIs), yet existing methods often fail to disentangle complex degradation mechanisms in multi-sensor systems or quantify uncertainty in HI reliability. This paper introduces a novel framework for HI construction, advancing three key contributions. First, we adapt Reconstruction along Projected Pathways (RaPP) as a health indicator (HI) for RUL prediction for the first time, showing that it outperforms traditional reconstruction error metrics. Second, we show that augmenting RaPP-derived HIs with aleatoric and epistemic uncertainty quantification (UQ) via Monte Carlo dropout and probabilistic latent spaces- significantly improves RUL-prediction robustness. Third, and most critically, we propose indicator groups, a paradigm that isolates sensor subsets to model system-specific degradations, giving rise to our novel method, I-GLIDE which enables interpretable, mechanism-specific diagnostics. Evaluated on data sourced from aerospace and manufacturing systems, our approach achieves marked improvements in accuracy and generalizability compared to state-of-the-art HI methods while providing actionable insights into system failure pathways. This work bridges the gap between anomaly detection and prognostics, offering a principled framework for uncertainty-aware degradation modeling in complex systems.