Uncertainty-Aware Real-Time Decision-Making in Digital Twins

Digital Twins are more than virtual replicas of physical systems. They create a dynamic connection between physical assets and their digital counterparts, enabling bi-directional interaction, continuous learning, and real-time decision support. By integrating sensor data, IoT systems, physics-based models, artificial intelligence, and advanced analytics, Digital Twins provide a powerful framework for understanding, predicting, and optimizing complex engineering systems.

This research thrust focuses on developing AI-enabled Digital Twin methods for system design, performance optimization, control, and lifecycle management. A central goal is to transform Digital Twins from passive monitoring tools into intelligent learning systems that can adapt to new data, reason under uncertainty, and support robust decision-making in changing operating environments.

Our work combines machine learning, uncertainty quantification, real-time data assimilation, model predictive control, reinforcement learning, etc, to bridge the gap between physical and digital domains. These methods enable Digital Twins to predict future system behavior, evaluate alternative decisions, optimize process conditions, and provide actionable insights with quantified confidence.