Magnesium (Mg) alloys are prime candidates for lightweight structures owing to their low density and high specific strength, but the pronounced basal texture that develops during wrought processing produces marked tension–compression asymmetry. Under low‑cycle fatigue (LCF) regime, this asymmetry is driven chiefly by repeated twinning–detwinning, making reliable life assessment exceptionally difficult. Therefore, cycle‑informed fatigue‑life estimation (CIFLE) is presented as a unified, physics‑aware machine learning framework for predicting LCF lives of wrought Mg alloys that display pronounced asymmetry. CIFLE links three data‑driven modules: a neural network that synthesizes representative hysteresis loops from basic loading inputs, an automated routine that interprets each loop into a concise set of mechanical and microstructural damage parameters, and a Bayesian network that maps those parameters to the life fraction while quantifying predictive uncertainty. The framework is trained and validated with cyclic deformation data from extruded AZ91 and SEN9 alloys, covering multiple strain amplitudes and extrusion conditions. Compared with conventional ε–N and energy‑based models, CIFLE achieves higher accuracy and well-calibrated uncertainty, delivering reliable life estimates at untested strain amplitudes by bracketing and refining energy bounds from neighboring tests. It also augments sparse datasets through loop synthesis and preserves accuracy even when most tests are withheld. In a case at an untested strain amplitude, the framework narrows the initial empirical life window and yields estimates that closely follow measured lives, whereas traditional models require additional experiments or extensive parameter tuning. By combining physics-based energy concepts with data-driven cycle synthesis, the framework provides an accurate, interpretable and data-efficient route for fatigue design of wrought Mg alloys.

Mg alloys are extremely valuable in the automotive and aerospace industries because of their lightweight properties and excellent machinability. The applications in these industries necessitate the accurate prediction of fatigue life under cyclic loading. However, this is challenging for many wrought Mg alloys owing to their pronounced plastic anisotropy. Conventional predictive methods such as the Coffin-Manson equation require manual parameter adjustment for different conditions, thus limiting their applicability. Accordingly, a novel predictive model for low-cycle fatigue (LCF) life that combines machine learning (ML) with an energy-based physical model, referred to as the hybrid ML/E model, is proposed herein. The hybrid ML/E model leverages a substantial hysteresis-loop dataset generated from LCF tests on a rolled AZ31 Mg alloy to effectively predict fatigue life. The proposed approach addresses the inherent challenges of small fatigue datasets, hysteresis-loop perception, and algorithm selection. The hybrid ML/E model demonstrates superior predictive accuracy and robustness in various loading directions, based on validation against conventional methods. The integration of ML and physical principles offers a unified framework for the LCF life prediction of anisotropic materials and represents a significant advancement for industrial applications.

Mg alloys possess an inherent plastic anisotropy owing to the selective activation of deformation mechanisms depending on the loading condition. This characteristic results in a diverse range of flow curves that vary with a deformation condition. This study proposes a novel approach for accurately predicting an anisotropic deformation behavior of wrought Mg alloys using machine learning (ML) with data augmentation. The developed model combines four key strategies from data science: learning the entire flow curves, generative adversarial networks (GAN), algorithm-driven hyperparameter tuning, and gated recurrent unit (GRU) architecture. The proposed model, namely GAN-aided GRU, was extensively evaluated for various predictive scenarios, such as interpolation, extrapolation, and a limited dataset size. The model exhibited significant predictability and improved generalizability for estimating the anisotropic compressive behavior of ZK60 Mg alloys under 11 annealing conditions and for three loading directions. The GAN-aided GRU results were superior to those of previous ML models and constitutive equations. The superior performance was attributed to hyperparameter optimization, GAN-based data augmentation, and the inherent predictivity of the GRU for extrapolation. As a first attempt to employ ML techniques other than artificial neural networks, this study proposes a novel perspective on predicting the anisotropic deformation behaviors of wrought Mg alloys.