Reliable and interpretable predictive maintenance in digital twin: A hybrid modeling approach powered by XAI

Li, Shancang and Zhao, Shanshan 2025. Reliable and interpretable predictive maintenance in digital twin: A hybrid modeling approach powered by XAI. Journal of Management Analytics Item availability restricted.

	PDF - Accepted Post-Print Version Restricted to Repository staff only Download (1MB)
	PDF (Provisional file) - Accepted Post-Print Version Download (17kB)

Abstract

In this work, we propose a real-time predictive maintenance digital twin framework that integrates industrial IoT sensors, adaptive deep learning architectures, and physics-informed analytics to enable proactive equipment health management. Our solution introduces a hybrid tunable LSTM network with dynamic hyperparameter optimization, designed to process multivariate time-series sensor data while addressing temporal degradation patterns and operational noise. By coupling explainable SHAP-driven diagnostics with bidirectional digital twin synchronization, the framework achieves early fault detection (identifying anomalies 15–30\% earlier than threshold-based methods) and precise remaining useful life (RUL) prediction. Validated across cross-industry case studies—including automotive manufacturing and pump system systems—the model reduces unplanned downtime by up to 41\% and optimizes maintenance costs through condition-based interventions, demonstrating robust generalisability in noisy industrial environments.

Item Type:	Article
Status:	In Press
Schools:	Schools > Computer Science & Informatics
Subjects:	A General Works > AI Indexes (General)
ISSN:	2327-0012
Date of First Compliant Deposit:	3 July 2025
Last Modified:	03 Jul 2025 14:30
URI:	https://orca.cardiff.ac.uk/id/eprint/179031

Actions (repository staff only)

Edit Item