An Explainable Regression Framework for Predicting Remaining Useful Life of Machines

Explainability builds trust beyond accuracy: The paper demonstrates that for critical applications like machine maintenance, stakeholders need more than just performance metrics (like error rates). Understanding why a machine is predicted to fail soon—not just that it will—is essential for decision-makers to trust and act on AI predictions.

Black-box AI poses risks in high-stakes settings: The authors highlight that many machine learning systems used for predicting equipment failures operate as unexplainable "black boxes." This lack of transparency is problematic when maintenance decisions could affect safety, costs, or operations.

Explainability techniques can be applied across different AI models: The framework uses two model-agnostic explanation methods (LIME and SHAP) that work with various machine learning algorithms, suggesting that explainability doesn't require choosing between performance and interpretability—both can be achieved together.

Responsible AI requires identifying decision drivers: The paper emphasizes that understanding which machine conditions or sensor readings most influence failure predictions is crucial for validating whether the AI system is making decisions based on sensible, real-world factors rather than spurious patterns.

Explainable AI is foundational for future development: By establishing this baseline framework combining prediction accuracy with explainability, the authors suggest that responsible AI practices—not just technical performance—should be built into machine learning systems from the start.