A Novel Digital Twin Architecture with Similarity-Based Hybrid Modeling for Supporting Dependable Disaster Management...

Area of Science:

• Environmental Science
• Computer Science
• Disaster Management

Background:

• Accurate disaster monitoring and prediction are crucial for effective disaster management systems.
• Digital twins offer potential for disaster prediction but face challenges with high-fidelity physics-based models and data dependency in data-driven models.
• Inaccurate predictions lead to inefficient resource allocation, potentially increasing disaster impact.

Purpose of the Study:

• To propose a digital twin architecture for accurate disaster prediction services.
• To develop a similarity-based hybrid modeling scheme to enhance prediction accuracy.
• To mitigate errors in digital twin disaster predictions.

Main Methods:

• A hybrid modeling scheme combining physics-based and data-driven models was developed.
• A data-driven error correction model was used to compensate for physics-based prediction inaccuracies.
• A similarity-based approach was employed to construct training datasets by assessing similarities between target and historical disasters.

Main Results:

• The proposed digital twin architecture significantly improved disaster prediction accuracy.
• The hybrid modeling scheme reduced prediction errors by approximately 50% in wildfire scenario evaluations.
• The similarity-based training dataset construction mitigated data dependency errors.

Conclusions:

• The developed digital twin architecture with a similarity-based hybrid modeling scheme provides accurate disaster prediction services.
• This approach enhances the reliability of digital twins for natural disaster management.
• The findings suggest a more effective allocation of disaster response resources, reducing overall damages.