Data-driven framework to build process-porosity linkages in metal additive manufacturing

This repository contains Porosity datasets derived from (i) experimental light optical microscopy images, (ii) melt pool simulations, (iii) pyrometer sensor data. The tabular dataframe is used to build machine-learning based process-porosity linkages. These linkages and especially feature contributions to the model output are further investigated by a SHAP analysis. The images for pre-processing, feature extraction (expert features and CNN) can be found here.

schema