Pre-trained artificial neural network for prediction of long-rod penetration depth in a semi-infinite target

Published: 12 December 2022| Version 1 | DOI: 10.17632/3rgkbwnzdb.1
Robbert Rietkerk,


A pre-trained artificial neural network is provided for predicting the scaled penetration depth P/L of a rod penetrating into a semi-infinite target, based on the rod length-over-diameter ratio L/D, the impact velocity and the density and hardness of the target and projectile materials. The tensorflow keras sequential network is stored as neural_network.h5 in hierarchical data format (HDF5). The script demonstrates an application of the neural network. Alternatively, the notebook main.ipynb or its static version main.html may be consulted.


Steps to reproduce

1. Install required python packages with: conda env create -f environment.yaml conda activate LRP 2. Run the main script to create a figure called finetuned_model_prediction.pdf Alternatively, run the jupyter notebook main.ipynb


Fraunhofer-Institut fur Kurzzeitdynamik Ernst-Mach-Institut


Machine Learning, Terminal Ballistics