Digital Twin-Enabled Anomaly Detection for Industrial IoT Using Explainable AI

Abstract

A hybrid approach is then introduced in this paper to combine the DT technology with XAI to detect the anomaly in IIoT environment in real time. The system also integrates high-fidelity simulation models with sensor data in order to increase the accuracy of detection and decrease the number of false positives. It leverages SHAP-based explanations, counterfactual deliberation, and natural language normalization to render the system interpretable for the engineers or operators in charge of decision making. Experimental results on real industrial datasets achieve a detection accuracy of 95.3% and 78% of reduction in false positives with respect to the state of the art. The promising performance of XAI-DT integration with a decision-supported mechanism demonstrates its application value for reliable and transparent predictive maintenance in industrial domain.