ONSAS / solid-mechanics-ML

This repository contains code for a project that trains a neural network to solve solid mechanics problems faster than the traditional finite element method. It includes a pipeline for generating a database of FEM solutions and experiments comparing the neural network model to the FEM.

solid-mechanics-ML

First tests in automating a pipeline for generating automatically a database of FEM solutions to unidimensional compression/extension problem, and learning with a Neural Network to solve same mechanical problem.

The loop that generates the data is in the file download_data.sh and the script that trains the Neural Network is in surrogateMLP.py. The data is stored in the folder data.

The final documentation for the project is in the file documentacion.pdf.

Acknowledgments

This work was funded by the Comisión Sectorial de Investigación Científica agency of Universidad de la República, through the Project Definición de estrategias para la aplicación de métodos de identificación de material al diagnóstico no invasivo de Cáncer de mama.