A Structured Probabilistic–AI Framework for Predictive Reliability Assessment and RUL Estimation of Asynchronous Rotating Machines

Toukap Nono Fernand Joseph; Tokoue Ngatcha Dianorré; Offole Florence; Nyatte Steyve; Mouzong Pemi Marcelin

doi:10.5935/jetia.v12i60.3969

Autores

Toukap Nono Fernand Joseph Laboratory of Mechatronics, Energiatronics and Sustainable Mobility (LaMEMD), National Higher polytechnic School of Douala, University of Douala.
Tokoue Ngatcha Dianorré Laboratory of Mechatronics, Energiatronics and Sustainable Mobility (LaMEMD), National Higher polytechnic School of Douala, University of Douala.
Offole Florence Laboratory of Mechanics and Materials (LaMM), National Higher polytechnic School of Douala, University of Douala
Nyatte Steyve Laboratory of technology and applied Sciences, University of Douala, Cameroun
Mouzong Pemi Marcelin Higher Technical Teacher’s Training College, University of Buea, Cameroun.

DOI:

https://doi.org/10.5935/jetia.v12i60.3969

Resumo

The industrial reliability of asynchronous rotating machines is a major strategic issue, where unplanned downtime causes considerable economic losses. This article proposes a structured and comprehensive methodological approach for predictive reliability assessment, integrating probabilistic reliability concepts with advanced condition monitoring and artificial intelligence tools. Our methodology, divided into six sequential phases (Acquisition, Feature Extraction, Diagnosis, RUL Modeling, Probabilistic Assessment, Decision Optimization), aims to transform multisensory data (vibrations, current, temperature) into optimal maintenance decisions. The theoretical application to a simulated asynchronous motor demonstrates the feasibility and expected benefits: early fault detection (up to 60 days before critical failure), reliable estimation of Remaining Useful Life (RUL), and economic optimization resulting in a 35% reduction in annual maintenance costs and a 0.5% increase in availability for a fleet of 10 motors. This approach provides a reproducible protocol for predictive maintenance, adaptable to different failure modes (mechanical, electrical, thermal).