Transfer learning-based prediction of high-temperature fatigue life in Fe-based structural alloys with limited data

Abstract

High-temperature (HT) fatigue life serves as a crucial performance metric for assessing the structural integrity and service safety of materials under elevated-temperature conditions. However, conventional assessment approaches rely on HT fatigue testing, which is typically time-consuming, costly, and experimentally challenging. Consequently, the scarcity of reliable HT fatigue data poses a major obstacle to accurate fatigue life prediction and limits the effectiveness of conventional machine learning models. In this study, a transfer learning strategy is proposed to address this data limitation by leveraging data-rich room-temperature fatigue datasets. An optimal feature subset is used with a gradient boosting decision tree model, and limited HT samples are progressively incorporated through incremental retraining to enable effective knowledge transfer across temperature domains. The results demonstrate significantly improved predictive accuracy and confirm the consistency of the dominant features across room and high temperatures. Overall, the proposed method offers a practical strategy for materials design and HT fatigue life assessment in engineering applications.

Keywords

High-temperature fatigue / fatigue life prediction / transfer learning / gradient boosting decision tree / feature selection

References

[1] Li Y,Huang W.Microstructure related analysis of tensile and fatigue properties for sand casting aluminum alloy cylinder head.Eng Fail Anal2022;136:106210

[2] Schütz W.A history of fatigue.Eng Fract Mech1996;54:263-300

[3] Peng W,Ge R.The influential factors on very high cycle fatigue testing results.MATEC Web Conf2018;165:20002

[4] Yang G,Li Q.Methodology to evaluate fatigue damage of high-speed train welded bogie frames based on on-track dynamic stress test data.Chin J Mech Eng2019;32:365

[5] Jimenez-Martinez M.Manufacturing effects on fatigue strength.Eng Fail Anal2020;108:104339

[6] Dixon WJ.A method for obtaining and analyzing sensitivity data.J Am Stat Assoc1948;43:109-26

[7] Veile G,Grözinger N,Grimm M.Improving fatigue testing of AISI 304L stainless steel in high temperature water regarding their complex hardening and softening material behaviour.Int J Press Vessel Pip2025;218:105612

[8] Hirose T,Nakajima M,Miyazawa T.Effects of test environment on high temperature fatigue properties of reduced activation ferritic/martensitic steel, F82H.Fusion Eng Des2018;136:1073-6

[9] Kamal M.Advances in fatigue life modeling: a review.Renew Sustain Energy Rev2018;82:940-9

[10] Zhu S,Correia J,Berto F.Evaluation and comparison of critical plane criteria for multiaxial fatigue analysis of ductile and brittle materials.Int J Fatigue2018;112:279-88

[11] Fatemi A.A critical plane approach to multiaxial fatigue damage including out‐of‐phase loading.Fatigue Fract Eng Mat Struct1988;11:149-65

[12] Wang H,Chen T.Prediction and evaluation of fatigue life via modified energy method considering surface processing.Int J Damage Mech2022;31:426-43

[13] Cui W.A state-of-the-art review on fatigue life prediction methods for metal structures.J Mar Sci Technol2002;7:43-56

[14] Shang C,Wu H.Efficient design of hydrogen-resistant ultra-high-strength steels via active learning and multiscale characterization.Corros Commun2026;21:60-72

[15] Zhang M,Zhang X.High cycle fatigue life prediction of laser additive manufactured stainless steel: a machine learning approach.Int J Fatigue2019;128:105194

[16] Yang J,Liu Y.A novel method of multiaxial fatigue life prediction based on deep learning.Int J Fatigue2021;151:106356

[17] Arvanitis K,Pavlou D.Machine learning-based fatigue lifetime prediction of structural steels.Alexandria Eng J2025;125:55-66

[18] Freed Y.Machine learning-based predictions of crack growth rates in an aeronautical aluminum alloy.Theor Appl Fract Mech2024;130:104278

[19] Rathore S,Kumar A,Singh A.Prediction of sub-critical fatigue crack growth rate in a high-carbon tempered martensitic steel at varying R ratio: experimental investigation and machine learning based modelling.Int J Fatigue2025;193:108804

[20] Shiraiwa T,Enoki M.Prediction of fatigue strength in steels by linear regression and neural network.Mater Trans2018;60:189-98

[21] Choi D.Data-driven materials modeling with XGBoost algorithm and statistical inference analysis for prediction of fatigue strength of steels.Int J Precis Eng Manuf2019;20:129-38

[22] Yan F,Liu Y,Chen J.Predictions and mechanism analyses of the fatigue strength of steel based on machine learning.J Mater Sci2020;55:15334-49

[23] Pan SJ.A survey on transfer learning.IEEE Trans Knowl Data Eng2010;22:1345-59

[24] Tang S,Yan Z,Khoo BC.Deep transfer learning strategy in intelligent fault diagnosis of rotating machinery.Eng Appl Artif Intell2024;134:108678

[25] Hu J,Tang H.An adversarial transfer learning method based on domain distribution prediction for aero-engine fault diagnosis.Eng Appl Artif Intell2024;133:108287

[26] Li Y,Xiang G,Liu H.Gear contact fatigue life prediction based on transfer learning.Int J Fatigue2023;173:107686

[27] Wei X,Han S,Wang C.High cycle fatigue S-N curve prediction of steels based on transfer learning guided long short term memory network.Int J Fatigue2022;163:107050

[28] Zhai Q,Zhu P.A transfer learning enhanced physics-informed neural network for predicting fatigue life of megacasting alloy with limited data sizes.Int J Fatigue2025;200:109129

[29] Furuya Y,Hirukawa H,Takeuchi E.Catalogue of NIMS fatigue data sheets.Sci Technol Adv Mater2019;20:1055-72 PMCID:PMC6853246

[30] Patro SGK. Normalization: a preprocessing stage. arXiv 2015, arXiv:1503.06462. https://doi.org/10.48550/arXiv.1503.06462. (accessed 2026-06-03)

[31] Miao B,Zhang Y,Yang H.Machine learning applications in metallic materials: recent advances and future perspectives.J Alloys Compd2026;1062:187486

[32] Shmuel A,Lazebnik T.A comprehensive benchmark of machine and deep learning models on structured data for regression and classification.Neurocomputing2025;655:131337

[33] Colin Cameron, A.; Windmeijer, F. A. An R-squared measure of goodness of fit for some common nonlinear regression models.J Econom1997;77:329-42

[34] Herlocker JL,Terveen LG.Evaluating collaborative filtering recommender systems.ACM Trans Inf Syst2004;22:5-53

[35] Cleger-Tamayo S,Huete J. On the use of weighted mean absolute error in recommender systems. 2012. https://api.semanticscholar.org/CorpusID:16692759. (accessed 2026-06-03)

[36] Sheugh L.A note on Pearson correlation coefficient as a metric of similarity in recommender system. In 2015 Ai & Robotics (IRANOPEN), Qazvin, Iran. Apr 12, 2015. IEEE; 2015. p. 1-6.

[37] Chai T.Root mean square error (RMSE) or mean absolute error (MAE)? - Arguments against avoiding RMSE in the literature.Geosci Model Dev2014;7:1247-50

[38] Shang C,Wu H.A quantitative relation for the ductile-brittle transition temperature in pipeline steel.Scr Mater2024;244:116023

[39] Efron B.Estimating the error rate of a prediction rule: improvement on cross-validation.J Am Stat Assoc1983;78:316-31

[40] Lundberg S. A unified approach to interpreting model predictions. arXiv 2017, arXiv:1705.07874. https://doi.org/10.48550/arXiv.1705.07874. (accessed 2026-06-03)

[41] Kohavi R.Wrappers for feature subset selection.Artif Intell1997;97:273-324

[42] Park JS,Lee CS.Effect of W addition on the low cycle fatigue behavior of high Cr ferritic steels.Mater Sci Eng A2001;298:127-36

[43] Lu X,Ran H.The role of layer strength ratio in enhancing strain hardening and achieving strength-ductility synergy in heterostructured materials.Acta Mater2025;289:120928