Hybrid Surrogate–Physics Framework for Rapid Initialization of Granular Packings in Discrete Element Simulations

This repository contains the datasets, source code, and benchmark outputs used in the development and validation of the surrogate modelling approach presented in the associated manuscript. The dataset includes the training data generated using the OxDEM simulation framework and the PyTorch implementation of the multilayer perceptron (MLP) surrogate model. Animated results are also provided to visually compare the OxDEM solution with the surrogate model predictions. Files included:

training_data.txt – Dataset generated using OxDEM simulations and used to train the surrogate model.
mlp_surrogate.py – PyTorch implementation of the MLP-based surrogate model used in the study.
OxDEM_solution.avi – Animation of the reference OxDEM simulation results used as the benchmark solution.
mlp_surrogate_solution.avi – Animation of the surrogate model predictions for comparison with the OxDEM solution.

These materials enable reproducibility of the training procedure and provide visual validation of the surrogate model performance.