Digital Twin for Industrial Supervision: Review of Techniques, Contributions, Limitations, and Emerging Trends in Automation

Abstract

This paper investigates digital twins as key components of cyber-physical supervision architectures for industrial systems through a systematic analysis of 65 publications (2019-2026). A tripartite taxonomy physical first, data-driven, and hybrid approaches is established and evaluated using operational criteria including data requirements, robustness to process drift, interpretability, and decision latency. The study highlights the evolution of digital twins transitioning from virtual replicas toward intelligent systems through the synergy and the integration of physical models, measured data, and artificial intelligence. Despite these advances, a gap remains between academic developments and industrial deployment due to sensitivity to data quality, limited explainability, and interoperability challenges. Six cross-cutting challenges are identified architectural standardization, physics informed robustness, self-supervised learning, explainable AI, cybersecurity, and scientific reproducibility to support the development of autonomous and prescriptive digital twins for resilient Industry 4.0/5.0 supervision systems. This work contributes through a supervision-oriented perspective combining a unified taxonomy,and quantitative benchmarking framework of digital twins.