Volume 42, No 1, 2020, Pages 1-9

Download full text in PDF

Machine Learning Approaches to Predict the Hardness of Cast Iron


C. Fragassa , M. Babic , E. Domingues dos Santos

DOI: 10.24874/ti.2020.42.01.01

Received: 21 March 2019
Revised: 27 June 2019
Accepted: 19 September 2019
Published: 15 March 2020


The accurate prediction of the mechanical properties of foundry alloys is a rather complex task given the substantial variability of metallurgical conditions that can be created during casting even in the presence of minimal variations in the constituents and in the process parameters. In this study an application of different intelligent methods of classification, based on the machine learning, to the estimation of the hardness of a traditional spheroidal cast iron and of a less common compact graphite cast iron is proposed. Microstructures are used as inputs to train the neural networks, while hardness is obtained as outputs. As general result, it is possible to admit that ‘light’ open source self-learning algorithms, combined with databases consisting of about 20-30 measures are already able to predict hardness properties with errors below 15 %.


Hardness prediction, Spheroidal cast iron (SGI), Compact graphite cast iron (CGI), Machine Learning (RF), Random Forest (RF), Artificial Neural Network (NN), k-nearest neighbors (kNN)

Last Edition


Volume 42
Number 2
June 2020

Crossref logo


TiI news RSS 2.0

Table of contents