Carbide Volume Fraction Estimation in as-cast HCCI Alloys using Machine Learning

This repo contains a literature collection of compositional data and experimentally determined carbide volume fractions (CVF) for as-cast high chromium cast iron (HCCI) alloys as well as a machine learning (ML) model to predict CVF based on the chemical composition.

The zip file "Dataset_HCCI_CVF.zip" contains the raw data compiled from literature, as well as the train and test splits that were used for training the ML model. The raw data compilation ("20240213_HCCI CVF Composition Database_zenodo.xlsx") lists the chemical compositions and experimentally determined CVF with corresponding references. Carbon-to-Chromium ratio has been added as an additional column. Moreover, CVF has been calculated according to existing literatures formulas (six in total). The deviation (in %) from experimental CVF for each calculation is also given.

A separate list of all references that have been included in the dataset is also provided as .bib and .ris files ("References for Excel Database.zip").

The zip file "ML_model_HCCI_CVF.zip" contains the final trained ML model (MATLAB file) and the corresponding MATLAB script that can be run in order to predict the CVF based on the chemical composition ("model_inference_CVF_HCCI.m"). The script accesses the trained ML model "GPR_final_all_data.mat" that must be stored in the same location as the MATLAB script. Input of the chemical composition can be done either directly in the MATLAB script or by loading an excel or csv spreadsheet. Further details about usage of the code are also mentioned in the MATLAB script.