PSI - Issue 68

Available online at www.sciencedirect.com Structural Integrity Procedia 00 (2025) 000–000 Available online at www.sciencedirect.com ScienceDirect Structural Integrity Procedia 00 (2025) 000–000 Available online at www.sciencedirect.com ScienceDirect

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

ScienceDirect

Procedia Structural Integrity 68 (2025) 458–464

European Conference on Fracture 2024 Temperature and strain rate sensitivity characterization of a Ti65 alloy by machinal learning method Wenqi Liu a , Zibiao Wang a, *, Tao Shi a , Jianrong Liu b , Guian Qian a European Conference on Fracture 2024 Temperature and strain rate sensitivity characterization of a Ti65 alloy by machinal learning method Wenqi Liu a , Zibiao Wang a, *, Tao Shi a , Jianrong Liu b , Guian Qian a Abstract This study proposes a straightforward and efficient investigation method to reveal the complex temperature and strain rate sensitivity of Ti alloys. A concise experimental program was designed for a Ti65 alloy to cover the temperature range of 25~650 °C and the strain rates from 10 -5 to 10 -2 s -1 with the smooth round bar samples. Tensile properties including the elastic modulus, yield strength, tensile strength, and fracture elongation were analyzed. It is indicated that the Ti65 alloy performed the dynamic strain aging effect in a certain temperature–strain rate interval and creep behavior at high temperatures, resulting in the non-monotonous and non-linear temperature and strain rate effects on the tensile properties of Ti65. The classical theory, i.e. the Johnson–Cook model, and the machine learning technique, i.e. support vector regression (SVR) algorithm, were adopted to predict the tensile properties of the investigated Ti65 at extensive temperature and strain rate ranges. It is demonstrated that the SVR algorithm is a suitable machine learning solution for the small amount of sample data with complicated non-linear dependence. With only 11 groups of experimental input data, the prediction performance of the SVR algorithm on strength is 7 times better than the Johnson– Cook model and the deviation between the predicted and measured properties is less than 3% for both strength and elongation prediction. © 2025 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 ECF24 organizers © 2025 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 ECF24 organizers a Institute of Mechanics, Chinese Academy of Science, Beijing 100190, China b Institute of Metal Research, Chinese Academy of Science, Shenyang 110016, China a Institute of Mechanics, Chinese Academy of Science, Beijing 100190, China b Institute of Metal Research, Chinese Academy of Science, Shenyang 110016, China Abstract This study proposes a straightforward and efficient investigation method to reveal the complex temperature and strain rate sensitivity of Ti alloys. A concise experimental program was designed for a Ti65 alloy to cover the temperature range of 25~650 °C and the strain rates from 10 -5 to 10 -2 s -1 with the smooth round bar samples. Tensile properties including the elastic modulus, yield strength, tensile strength, and fracture elongation were analyzed. It is indicated that the Ti65 alloy performed the dynamic strain aging effect in a certain temperature–strain rate interval and creep behavior at high temperatures, resulting in the non-monotonous and non-linear temperature and strain rate effects on the tensile properties of Ti65. The classical theory, i.e. the Johnson–Cook model, and the machine learning technique, i.e. support vector regression (SVR) algorithm, were adopted to predict the tensile properties of the investigated Ti65 at extensive temperature and strain rate ranges. It is demonstrated that the SVR algorithm is a suitable machine learning solution for the small amount of sample data with complicated non-linear dependence. With only 11 groups of experimental input data, the prediction performance of the SVR algorithm on strength is 7 times better than the Johnson– Cook model and the deviation between the predicted and measured properties is less than 3% for both strength and elongation prediction. © 2025 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 ECF24 organizers Keywords: Ti65 alloy; Support vector regression; Johnson–Cook model; Temperature and strain rate sensitivity; Dynamic strain aging Keywords: Ti65 alloy; Support vector regression; Johnson–Cook model; Temperature and strain rate sensitivity; Dynamic strain aging

* Corresponding author. Tel.: +86 13021113511; fax: +010-82543893. E-mail address: scientistbiao@163.com * Corresponding author. Tel.: +86 13021113511; fax: +010-82543893. E-mail address: scientistbiao@163.com

2452-3216 © 2025 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 ECF24 organizers 2452-3216 © 2025 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 ECF24 organizers

2452-3216 © 2025 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 ECF24 organizers 10.1016/j.prostr.2025.06.082