Assessment of a 40-year-old induction motor using hybrid diagnostic and AI-based predictive techniques

Abstract Electric motors represent critical assets in industrial systems, where reliability and longevity directly influence operational continuity. Although the nominal service life of induction motors is typically 20–25 years, many units remain in operation beyond this threshold under effective maintenance. This study evaluates the continued performance and insulation health of a 40-year-old, 150 kW low-tension induction motor deployed in a water transfer pump system. A comprehensive diagnostic protocol was applied, including insulation resistance, polarization index, dielectric absorption ratio, leakage current, and DC winding resistance measurements. Results indicated insulation resistance values between 2.39 GΩ and 10.3 GΩ, an R-phase polarization index of 1.87, and marginal performance in Y and B phases. Infrared thermography identified localized temperature gradients associated with incipient faults. AI-assisted analytics using a Random Forest classifier achieved an overall accuracy of 86.7% and ROC-AUC of 0.81, demonstrating moderate predictive capability. The framework illustrates the potential for integrating conventional diagnostics with data-driven decision support in condition-based maintenance applications. The motor maintained an availability of 99.94%, confirming its extended viability under structured monitoring. The combined framework merging conventional electrical diagnostics, thermal imaging, and machine-learning inference provides a scalable approach for condition-based maintenance and life-extension assessment of aged assets.