End-to-end lifecycle machine learning framework for predictive maintenance of critical equipment
Jérémie Marchand, Jannik Laval, Aïcha Sekhari, Vincent Cheutet, Jean-Baptiste Danielou · Enterprise Information Systems · 2025
The industrial adoption of data-driven predictive maintenance (PdM) is increasing, with machine learning (ML) methods playing a key role in preventing equipment failures. However, ML models assume stationary data, a condition rarely met in non-stationary industrial environments. This paper proposes a comprehensive framework for managing ML systems in PdM to address concept drift and maintain performance throughout their lifecycle, particularly during usage and maintenance. The framework includes dual-level drift detection, drift severity quantification, integration of human expertise, and end-to-end lifecycle management, offering a robust solution for long-term reliability and adaptability in dynamic industrial settings.