Sustainable ductile iron design: Leveraging machine learning and genetic algorithms for critical raw materials reduction

One of this century's key challenges is ensuring the supply of critical raw materials essential for economic stability. These materials face supply risks due to factors like scarcity, geopolitical instability in supplier countries, and lack of substitutes. The European Union maintains a list of...

Full description

Saved in:
Bibliographic Details
Published in:Engineering applications of artificial intelligence Vol. 163; p. 113117
Main Authors: Ferro, P., Spiller, S., Tang, K., Berto, F., Bonollo, F.
Format: Journal Article
Language:English
Published: Elsevier Ltd 01.01.2026
Subjects:
Alloy optimization
Cast iron
Critical raw materials
Genetic algorithm
Machine learning
Sustainable materials design
Critical raw materials
Alloy optimization
Genetic algorithm
Machine learning
Cast iron
Sustainable materials design
ISSN:0952-1976
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:One of this century's key challenges is ensuring the supply of critical raw materials essential for economic stability. These materials face supply risks due to factors like scarcity, geopolitical instability in supplier countries, and lack of substitutes. The European Union maintains a list of critical raw materials, updated every three years, and is strongly committed to developing strategies to mitigate these supply risks. Although ductile iron is a strategic material for mass production, offering low cost and high mechanical performance, its production requires critical raw materials, as well, such as Ti, Mg, Nb, and Sb, among others. To address this issue, researchers are exploring substitutes that could replace critical raw materials in ductile iron's composition without compromising or potentially even improving its mechanical properties. In this work, a deep neural networks-based model has been developed to assess the mechanical properties of ductile cast iron based on its chemical composition. The model has then been integrated with a genetic algorithm to optimize the chemical composition of the alloy in a critical raw material perspective, while maintaining equivalent mechanical performance. Three case studies were presented to illustrate the potentialities of the proposed approach. [Display omitted] •A hybrid DNN-GA framework was developed to design ductile iron alloys under CRM constraints.•The DNN predicts UTS, YS, and Elongation with high accuracy (R > 0.96).•The GA minimizes the Alloy Criticality Index while preserving mechanical properties.•The model supports sustainable alloy design aligned with circular economy goals.•Novel approach supports development of sustainable alloys with reduced CRM reliance.
ISSN:0952-1976
DOI:10.1016/j.engappai.2025.113117