The high explosives & affected targets (HEAT) dataset.

Area of Science:

• Computational physics
• Materials science
• Artificial intelligence

Background:

• Simulating shock propagation through multiple materials is computationally intensive, requiring complex physics models.
• Existing datasets are insufficient for training and validating AI/ML models for high-explosive driven shocks.
• Publicly available data for multi-material shock dynamics is lacking for the AI/ML community.

Purpose of the Study:

• To introduce the High-Explosives and Affected Targets (HEAT) Dataset, a novel resource for AI/ML research.
• To provide a physics-rich dataset for training, testing, and validation of AI surrogate models for shock propagation.
• To enable the development of efficient AI/ML emulations of complex shock phenomena.

Main Methods:

• Generated a collection of 2D, cylindrically symmetric simulations using an Eulerian, multi-material shock propagation code.
• Created two partitions: expanding shock-cylinder (CYL) and Perturbed Layered Interface (PLI) simulations.
• Included time-series data of thermodynamic and kinematic fields for various materials, including high explosives.

Main Results:

• The HEAT dataset comprises detailed simulation data capturing phenomena like momentum transfer, shock propagation, plastic deformation, and thermal effects.
• The CYL partition includes diverse materials (solids, liquids, gases, high explosive), while PLI focuses on specific material combinations.
• The dataset serves as a valuable benchmark for AI/ML model development in multi-material shock dynamics.

Conclusions:

• The HEAT dataset addresses a critical gap in publicly available data for multi-material shock propagation.
• It facilitates the advancement of AI/ML surrogate models for computationally challenging physics problems.
• HEAT enables more efficient and accurate emulation of high-explosive driven shock phenomena.