A Physics‐Guided Neural Network for Probabilistic Fatigue Life Prediction Under Multiple Overload Effects
Fatigue & Fracture of Engineering Materials & Structures · 2025-04 · 48 (4) · 1612-1629
DOI: 10.1111/ffe.14557 ↗Kurz gefasst
Fachartikel in der Zeitschrift Fatigue & Fracture of Engineering Materials & Structures, online erschienen am 16.01.2025. Verfasst von Shan Jiang und 2 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 37 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Fundamental Research Funds for the Central Universities.
Abstract (Originalsprache)
A Physics‐guided Neural Network (PgNN) is proposed to provide a robust probability distribution of fatigue life under arbitrary multiple overloads, which integrates the physical mechanism model (PMM) and neural network (NN). Notably, the proposed PgNNs are trained solely using data under constant amplitude loading scenarios. Firstly, a PMM is developed to predict fatigue life based on linear elastic fracture mechanics, considering crack closure. A data preprocessing approach informing PMM is presented, transforming arbitrary overload conditions into equivalent constant amplitude loading with stress ratio . Moreover, a back‐propagation NN is constructed, where a loss function integrating the PMM and mean square error is designed. The PgNN framework encompasses the uncertainties associated with stress levels, material coefficients and equivalent initial flaw size. The fatigue data of aluminum alloy 7075‐T6 and Al‐Li alloy 2060 are used for model validation. The results affirm that the PgNN exhibits superior accuracy and robustness.
Förderung
- National Natural Science Foundation of China
- Fundamental Research Funds for the Central Universities
Zitieren
APA
Jiang, S., Zhang, Y., & Zhang, W. (2025). A Physics‐Guided Neural Network for Probabilistic Fatigue Life Prediction Under Multiple Overload Effects. Fatigue & Fracture of Engineering Materials & Structures, 48(4), 1612-1629. https://doi.org/10.1111/ffe.14557
Vancouver
Jiang S, Zhang Y, Zhang W. A Physics‐Guided Neural Network for Probabilistic Fatigue Life Prediction Under Multiple Overload Effects. Fatigue & Fracture of Engineering Materials & Structures. 2025;48(4):1612-1629. doi:10.1111/ffe.14557
BibTeX
@article{Jiang2025,
author = {Jiang, Shan and Zhang, Yingchun and Zhang, Wei},
title = {A Physics‐Guided Neural Network for Probabilistic Fatigue Life Prediction Under Multiple Overload Effects},
journal = {Fatigue & Fracture of Engineering Materials & Structures},
year = {2025},
volume = {48},
number = {4},
pages = {1612--1629},
doi = {10.1111/ffe.14557}
}
EUVIMED