Advancing Fault Detection and Classification Through High-Fidelity Digital Twin Simulation and AI

This research study focuses on evaluating the efficacy of digital twin (DT) technology in real-time fault diagnosis within a DC buck converter. The primary role of the DT is to serve as a high-precision substitute for the actual system, providing a secure environment for users while preventing potential damage to the physical setup. The initial objective of this study is to establish a high-fidelity DT model as a genuine virtual representation of the physical system, striving for a Root Mean Square Error (RMSE) within 3% under three loading conditions. Subsequently, the second phase involves identifying faults and pinpointing their locations. The assessment is based on measurements derived from the confusion matrix using two supervised learning methods, including Support Vector Machines (SVM) and Ensemble. Results indicate that Ensemble outperforms SVM in the prognostic detection and classification process.