Early Fault Detection and Diagnosis in Industrial Machines: An Approach Using Digital Twins

The emergence of Industry 4.0 has catalyzed the integration of advanced technologies to enhance manufacturing efficiency, reliability, and competitiveness. Fault Detection and Diagnosis (FDD) systems are critical for minimizing downtime and ensuring operational continuity. This research investigates the integration of Digital Twin (DT) technology with Machine Learning (ML) models for real-time fault detection and diagnosis (RT-FDD) in discrete manufacturing machines. Two industrial systems—a Pick-and-Place machine and a Furnace—were modeled using linear and non-linear models to develop Digital Twins. The method is validated on a simulated environment replicating real industrial behavior, combining DT-generated features with conventional real-time process data. The proposed approach improved F1 scores by up to 11% and demonstrated enhanced robustness in both inter-cycle and intra-cycle fault detection tasks. Notably, for the Furnace machine, the method enabled fault detection 40% earlier in the cycle while maintaining the same F1 Score of 94%, and provided reliable diagnosis with an F1 Score of 80% at only 15% of cycle completion. A comprehensive evaluation of 16 ML algorithms highlighted the effectiveness of DT features in boosting predictive performance.