Predictive Maintenance for Rotating Equipment in Process Manufacturing: A Reliability-Based Framework Using Vibration, Thermal, and Electrical Signatures

Abstract

This article presents a reliability-based predictive maintenance framework that integrates vibration condition indicators, infrared thermography, and motor electrical signatures into a unified decision pipeline that supports early detection, fault isolation, and maintenance prioritization under realistic noise, load variation, and data completeness constraints. The framework is engineering-oriented rather than tool-oriented, meaning that it defines measurable acceptance limits, uncertainty handling, and verification pathways so that maintenance decisions are defensible and repeatable, and it evaluates performance using detection reliability, lead time, false alarm governance, and economic consequence rather than using model accuracy in isolation. A generic case design is developed for typical process industry assets, and representative datasets are used to demonstrate how feature distributions shift across failure modes such as bearing wear, misalignment, imbalance, cavitation, and electrical asymmetry, while showing how multi-sensor fusion reduces decision uncertainty and improves time-to-decision compared with single-signal approaches. Results indicate that vibration broadband velocity and acceleration envelope metrics provide the earliest warning for mechanical degradation but are vulnerable to confounding under load and hydraulic variability, that thermal imaging strengthens fault localization and reduces nuisance alarms when integrated as a verification layer, and that motor current signature indicators improve detection of electrical defects and rotor bar issues that do not immediately appear in mechanical vibration. The study concludes with practical guidance for implementation, including recommended monitoring intervals, threshold governance strategies aligned with reliability criticality, and data-driven visualization templates suitable for Techne submissions and industrial reporting.