Digital Twin and Industrial Internet of Things–Enabled Predictive Maintenance for Reliability Enhancement in Smart Manufacturing Systems

Abstract

This study quantitatively evaluated the effectiveness of Digital Twin and Industrial Internet of Things (IIoT)–enabled predictive maintenance for enhancing equipment reliability, maintenance efficiency, and production performance in smart manufacturing systems. A longitudinal quasi-experimental design was applied to 220 manufacturing assets representing 18,742 operating cycles and approximately 1.285 million cumulative machine-hours. The analytical dataset incorporated 18.64 million validated sensor observations, 1,086 maintenance events, and 428 baseline failure events across motors, bearings, gearboxes, CNC machines, cutting tools, pumps, compressors, robotic systems, conveyors, machining centers, and production-line assets. Baseline IIoT sensor coverage was 91.4%, data availability was 96.8%, and digital-twin fidelity was 92.6%. Following predictive-maintenance implementation, failure frequency decreased from 1.95 to 1.18 failures per asset, representing a 39.5% reduction, while mean time between failures increased from 1,486.3 to 2,176.8 hours, equivalent to a 46.5% improvement. Mean time to repair declined by 35.6%, availability increased from 91.8% to 96.1%, and unplanned downtime decreased by 45.9%. Overall equipment effectiveness increased from 76.9% to 85.6%, production interruptions declined by 44.7%, throughput loss decreased by 46.1%, and maintenance cost fell by 24.6%. Early fault detection increased from 54.3% to 87.2%, while anomaly-detection and fault-classification accuracy reached 94.1% and 92.8%, respectively. The highest-performing ensemble model achieved 96.2% classification accuracy, a 96.0% F1 score, ROC-AUC of 0.982, and remaining-useful-life R² of 0.94. Digital-twin fidelity increased to 96.4%, and synchronization consistency reached 98.1%. All principal baseline-to-post comparisons were statistically significant at p < 0.001, with large effects observed for MTBF, unplanned downtime, OEE, availability, and failure frequency. The findings demonstrated that reliable IIoT connectivity, high digital-twin fidelity, accurate predictive modeling, and earlier maintenance intervention were collectively associated with substantial improvements in manufacturing reliability, operational continuity, and maintenance performance.