# Material Fracturing and Failure Simulation Datasets

## TL;DR

This paper introduces a large dataset of simulated material fractures and failures, useful for developing machine learning models to predict material behavior under stress.

## Contribution

The novelty lies in providing a diverse and comprehensive dataset of simulated fractures using two distinct numerical methods across multiple materials.

## Key findings

- The dataset includes simulations for five materials using two numerical solvers.
- Phase-field simulations cover 400,000 cases under different loading conditions.
- The dataset supports the development of machine learning models for material failure prediction.

## Abstract

Fracturing is a fundamental physics phenomena with broad relevance across multiple domains, ranging from infrastructure integrity, aerospace durability, reservoir production, and seismic events. We present a diverse dataset of simulated fracture evolution and material failure generated from two numerical solvers: the phase-field method and the combined finite-discrete element method (FDEM). These solvers differ in formulation, physical fidelity, and computational efficiency. The dataset includes five materials: PBX, anisotropic shale, tungsten, aluminum, and steel. For each, phase-field simulations span 400,000 cases: 200,000 under uniaxial tension and 200,000 under biaxial tension. The computationally expensive FDEM simulations include 90,000 split evenly among PBX, shale, and tungsten under uniaxial loading. All simulations begin with randomized initial fracture patterns. Each entry includes temporal data capturing fracture propagation dynamics. This comprehensive dataset is designed to support the development of foundational or surrogate machine learning approaches for predicting material failure. While no such models are introduced here, the dataset lays a robust foundation for advancing future research and innovation in these areas.