PSI - Issue 28

Available online at www.sciencedirect.com Available online at www.sciencedirect.com ScienceDir ct Structural Integrity Procedia 00 (2019) 000–000 Available online at www.sciencedirect.com ScienceDirect Structural Integrity Procedia 00 (2019) 000–000

www.elsevier.com/locate/procedia www.elsevier.com/locate/procedia

ScienceDirect

Procedia Structural Integrity 28 (2020) 1392–1398

© 2020 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of the European Structural Integrity Society (ESIS) ExCo Abstract There was studied a jump-like deformation of AMg6 aluminum alloy at static tensile test in the soft mode of loading. The experimental methods of jump-like deformation study of this alloy are often complicated, expensive and time-consuming. Therefore, it is more effective to model the stress-strain diagram numerically. One of the promising ways to predict the stress-strain diagram of AMg6 material is based on the application of machine learning methods, in particular by neural networks, boosted trees, support-vector machines, and k - nearest neighbors. It was discovered that the method of neural networks gives the least prediction error equal to 6.9%. The errors of the methods of boosted trees, support-vector machines and k - nearest neighbours are 9.1, 7.7, and 9.4% in test sets, respectively. © 2020 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of the European Structural Integrity Society (ESIS) ExCo Keywords: AMg6 aluminum alloy; jump-like deformation; machine learning 1. Introduction Improving the reliability of the important structural elements and avoiding industrial accidents is an actual task. In particular, the failure of details is related to changes in their structure and mechanical properties during operation. Therefore, the prediction of material stress-strain diagrams is of high importance. Abstract There was studied a jump-like deformation of AMg6 aluminum alloy at static tensile test in the soft mode of loading. The experimental m thods of jump-like deformation study of this alloy are often complicated, exp nsive an time-consuming. Therefore, it is more effective to model the stress-strain diagram numeric lly. One of the promising way to pre ic th stre s-strain diagram of AMg6 material is based on the application of machine learn ng methods, in particular by neural networks, boosted trees, support-vector machin s, and k - nearest neighbors. It was discovered that th method of neural networks gives the least pr iction error equal 6.9%. The errors of the methods of boosted trees, support-vector machines and k - nearest neighbours are 9.1, 7.7, and 9.4% in test sets, resp ctively. © 2020 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review u der re ponsibility of European Structural Integri y Society (ESIS) ExCo Keywords: AMg6 aluminum alloy; jump-like deformation; machine learning 1. Introduction Improving the reliability of the important structural elements and avoiding industrial accidents is an actual task. In particular, the failure of details is related to changes in their tructure and mechanical prop r ies during operation. Therefore predicti n of material str ss-strain diagrams is of high importance. 1st Virtual European Conference on Fracture Modeling of AMg6 aluminum alloy jump-like deformation properties by machine learning methods Oleh Yasniy a, *, Iryna Didych a , Sergiy Fedak a ,Yuri Lapusta b a Ternopil Ivan Pul’uj National Technical University, Ruska str. 56, Ternopil, 46001, Ukraine b University Clermont Auvergne, CNRS, SIGMA Clermont, Institut Pascal, F-63000 Clermont-Ferrand, France 1st Virtual European Conference on Fracture Modeling of AMg6 aluminum alloy jump-like deformation properties by machine learning methods Oleh Yasniy a, *, Iryna Didych a , Sergiy Fedak a ,Yuri Lapusta b a Ternopil Ivan Pul’uj National Technical University, Ruska str. 56, Ternopil, 46001, Ukraine b University Clermont Auvergne, CNRS, SIGMA Clermont, Institut Pascal, F-63000 Clermont-Ferrand, France

* Corresponding author E-mail address: oleh.yasniy@gmail.com * Corresponding author E-mail address: oleh.yasniy@gmail.com

2452-3216 © 2020 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of the European Structural Integrity Society (ESIS) ExCo 2452-3216 © 2020 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review u der responsibility of t European Structural Integrity So i ty (ESIS) ExCo

2452-3216 © 2020 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of the European Structural Integrity Society (ESIS) ExCo 10.1016/j.prostr.2020.10.110