A Semantic Data-Based Distributed Computing Framework to Accelerate Digital Twin Services for Large-Scale Disasters.

Area of Science:

• Environmental Science and Engineering
• Computer Science
• Disaster Management

Background:

• Increasingly extensive natural disasters, exacerbated by global warming, challenge traditional disaster management systems.
• Existing digital twin frameworks for disaster prediction struggle with complex natural phenomena and fail to account for environmental data correlations.
• This leads to inaccurate computing load predictions, unbalanced load partitioning, and increased prediction service times.

Purpose of the Study:

• To propose a novel distributed computing framework to accelerate disaster prediction services.
• To address the limitations of previous schemes by incorporating semantic analyses of environmental data correlations.

Main Methods:

• Developed a framework that combines environmental data into disaster semantic data, representing initial disaster states (e.g., wildfire burn scars, fuel models).
• Utilized a convolutional neural network-based algorithm for accurate computing load prediction based on semantic data.
• Implemented balanced partitioning of simulation models and allocation of sub-models to distributed computing nodes.

Main Results:

• The proposed framework demonstrated a significant decrease in disaster prediction times.
• Achieved up to a 38.5% reduction in prediction time compared to previous disaster management schemes.
• Successfully improved load partitioning accuracy through semantic analysis of environmental data correlations.

Conclusions:

• The novel distributed computing framework effectively accelerates disaster prediction services.
• Semantic analysis of environmental data correlations is crucial for accurate computing load prediction and balanced resource allocation.
• This approach enhances the efficiency and responsiveness of disaster management agencies in the face of complex natural disasters.