Session: CT-19-01 AI, Data Engineering and Data Analysis-1

Paper Number: 151723

151723 - Machine Learning to Predict Mechanical Properties of a Material, the Cucrzr Case Study 

Abstract: 

Material qualification for critical components, including mechanical properties quantification, often requires destructive testing, leading to costly scrapping. These tests, however, need to be performed to quantify mechanical properties, which need to comply with imposed requirements. This is particularly challenging for novel materials where minor variations in manufacturing or composition can significantly impact mechanical properties. Our study focuses on Copper Chromium Zirconium (CuCrZr), a crucial material for ITER's in-vessel blanket first wall panels.

We employed various machine learning models to predict CuCrZr's mechanical properties, identify key parameters affecting these properties and verify compliance with established requirements.

Our models achieved accuracies ranging from 91.1% to 98.1%, successfully correlating chemical composition and manufacturing processes with mechanical properties. Inverse modeling validated property limits against input parameters and quantified the impact of requirement variations on mechanical properties. Furthermore, the impact from variations in requirements onto mechanical properties can be quantified with a reliable accuracy with these models.

This approach shows promise for complementing traditional sample testing in material qualification, expediting the introduction of new materials into codes and standards and reducing the need for destructive testing and associated waste

By leveraging machine learning, we can enhance material development efficiency, reduce costs, and accelerate innovation in critical engineering applications.

Presenting Author: Maria Ortiz de Zuniga Fusion for Energy

Presenting Author Biography: María holds a Master's Degree in Computer Engineering from ICAI, having also spent a year at University of Michigan, as well as a Master's in Advanced Manufacturing Technologies from UNED, both in Spain. She is currently pursuing developing a Philosopher's Degree in Industrial Technologies on "Improvement methods for manufacturing of ITER components, through automation and artificial intelligence".

María was appointed in November 2022 as Senior Technical Officer, developing an Artificial Intelligence Program, as well as representative of the agency on Nuclear codes subcommittee. Prior to that, Maria held positions as the Deputy Program Manager for the ITER Antennas or the Vacuum Vessel Project Manager at the European Domestic Agency for the ITER Project, Fusion for Energy. The aim of the antennas is to heat up and control instabilities in the plasma, whereas the Vacuum Vessel is a large nuclear stainless double-wall torus-shaped container, made up of nine sectors where each one weighs 450 tons and characterised by a complex nuclear manufacturing, with a budget of 200M€. Prior to that, Maria worked as technical coordinator for production control and in project management.

María has worked for 17 years on the ITER Project. She enjoys interdisciplinary challenges, working in international environments and contributing to global energy solutions.

Authors:

Maria Ortiz de Zuniga Fusion for Energy
Samuli Heikkinen Fusion for Energy
Jose Miguel Pacheco Fusion for Energy
Margherita Sardo Fusion for Energy
Ana María Camacho Universidad Nacional de Educacion a Distancia
Álvaro Rodríguez-Prieto Universidad Nacional de Educacion a Distancia