1 August 2025 in Physica Scripta
The nonlinear ultrasonic response to early plastic damage of metal materials is the result of multiple defects. The contributions of typical defects (vacancy atoms, interstitial atoms, edge dislocations, screw dislocations, grain boundaries, grains, twins and extrinsic stacking faults) to the ultrasonic nonlinear effects has not been reported. In this paper, the ultrasonic propagation model of a single defect in single crystal aluminum was established using the molecular dynamics method. The nonlinear ultrasonic response of each typical defect and the intrinsic relationship between the defect parameters and the acoustic nonlinearity were investigated. Meanwhile, the influence of these typical defects on linear acoustic parameters was also studied. Finally, a multi-defect coupling model incorporating vacancy atoms and other defects was established to analyze the interaction mechanisms of ultrasonic nonlinear parameters with positive/negative contribution coupling. This provides microscopic insights for determining critical defect configurations and assessing damage states associated with mechanical property degradation in metallic materials.
18 March 2026 in Measurement Science and Technology
Accurate fatigue life prediction is crucial to maintain the safety and structural integrity of mechanical structures. Magnetic Barkhausen Noise (MBN), a non-destructive testing (NDT) technique, is sensitive to microstructural changes induced by fatigue loading in ferromagnetic materials. It shows potential for processing vast and temporally dynamic data from MBN signals. However, accurate prediction of fatigue life of structures is still a challenge. This paper presents a multivariate time series analysis method using hybrid Transformers, named as MetaTran, for spatio-temporal feature fusion and improvement of fatigue life prediction using MBN signals. First, an adaptive window segmentation (AWS) method is designed to partition the preprocessed MBN signal to effectively capture intricate patterns and temporal dynamics. Then, a multi-domain feature selection using the Boruta algorithm identifies the most relevant features. Finally, based on multi-source features, a novel multi-perspective hybrid Transformer (MPHT) is constructed for fatigue life prediction. It consists of a feature tower and a time tower. The former employs weighted embeddings and self-attention mechanism to enhance feature extraction. The latter obtains temporal dependencies of multivariate data. Experimental data were collected from two specimens 20R and Q345 structural steels. Experimental results demonstrate that the proposed MetaTran has higher prediction accuracy in various tests. It is a robust and practical prediction model for both multivariate time series analysis and fatigue life assessment.