PSI - Issue 72

Available online at www.sciencedirect.com

ScienceDirect

Procedia Structural Integrity 72 (2025) 181–187

12th Annual Conference of Society for Structural Integrity and Life (DIVK12) Prediction of structural elements lifetime of titanium alloy using neural network Oleh Yasniy a , Iryna Didych a, *, Dmytro Tymoshchuk a , Pavlo Maruschak a , Vladyslav Demchyk a a Ternopil Ivan Puluj National Technical University, Ruska str. 56, Ternopil, 46001, Ukraine Abstract In modern mechanics of materials, fatigue failure prediction problems are important for ensuring the reliability of structures. Machine learning methods, particularly neural networks, allow for effective modeling of complex relationships between loading parameters and failure characteristics. In this paper, a model was created using neural networks to accurately predict how the crack length increases with the number of loading cycles at different stress ratios. The neural network was trained on experimental data, which allowed the model to recognize patterns and predict crack length. In particular, the training data were the number of loading cycles N and the stress ratio R . Therefore, integrating traditional models with machine learning methods ensures high prediction accuracy and the reliability of structural elements. The prediction results are in good agreement with the experimental ones. In particular, the error of 0.2% at the stress ratio R = 0.03, 0.1, 0.3 and 0.4% at the stress ratio R = 0.03 were obtained by the neural network method in the test sample. By predicting these curves accurately, we can assess the remaining life of a material and make informed decisions regarding maintenance and operation. The neural network models provided accurate predictions and showed flexibility in adapting to different stress ratios and loading conditions. © 2026 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 Aleksandar Sedmak, Branislav Djordjevic, Simon Sedmak Dr. Simon Sedmak, ssedmak@mas.bg.ac.rs, Innovation Center of Faculty of Mechanical Engineering, Belgrade, Serbia

Keywords: fatigue crack growth rate; artificial intelligence; machine learning; neural networks

* Corresponding author. Tel.: +380972272074 E-mail address: iryna.didych1101@gmail.com

2452-3216 © 2026 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 Aleksandar Sedmak, Branislav Djordjevic, Simon Sedmak Dr. Simon Sedmak, ssedmak@mas.bg.ac.rs, Innovation Center of Faculty of Mechanical Engineering, Belgrade, Serbia 10.1016/j.prostr.2025.08.090