ABC dataset

This is the repository for codes for ML, domain generation, graph generation of Asymmetric Buckling Columns (ABC) dataset in the paper "Learning Mechanically Driven Emergent Behavior with Message Passing Neural Networks".

Link to paper

The ABC data consists of spatially heterogenous beams under compression with fixed-fixed boundary conditions. The objective is to classify each column into its buckling direction. The dataset is seperated into 3 sub-datasets, each with varying geometry.

Link to dataset

If using the geometry files, then start from folders Domain - FEA (if you need labels) - Graph - ML. If using the json files, then start from folders FEA (if you need labels ) - Graph - ML.

The FEA folder can be skipped if labels are already obtained.

Domain

The Domain folder contains code to generate the domain geometry for ABC dataset. More details of the code are contained in the folder.

FEA

The FEA folder contains code to generate labels for the ABC dataset. More details of the code are contained in the folder.

Graph

The Graph folder contains code to generate graphs from domain geometry and json files. More deatils of the code are contained in the folder.

ML

The ML folder contains code for our protocol for train, validation and test split, PointNet++ implementation, ML architecture, and voting. More details of the code are contained in the folder.