Void characteristics are fundamentally correlated with the macroscopic deformation responses of materials, yet traditional modeling methods exhibit inherent limitations in data mining. In this study, a machine learning (ML) framework is proposed to predict the full-field strain evolution of Cu/Ni clad foils, and the impact of intrinsic voids is quantitatively assessed using interpretative analysis methods. The local strain and void data are extracted and integrated through digital image correlation and computed tomography. To accommodate the nature of the constructed dataset, a ML model is established with reference to the concept of time series forecasting. Subsequently, the influence of microstructural features such as volume fraction (VVF), area, and size of voids are investigated, alongside their role in driving local strain evolution. This approach successfully predicts strain localization, and accurately pinpoints the onset of plastic instability and the location of crack initiation. The VVF is identified as the most predominant factor, followed by void size along the tensile direction and grain size. The strongest association is observed between the VVF and grain size, which intensifies over extended time scales. Moreover, as void coalescence is almost completed, the promoting effect of the concentrated void distribution on macroscopic strain concentration will become increasingly pronounced. These findings provide novel perspectives for exploring the intricate relationship between deformation and damage.

中文翻译：

---

通过可解释机器学习研究 Cu/Ni 包层箔的全场应变演化行为


空隙特性与材料的宏观变形响应从根本上相关，但传统的建模方法在数据挖掘中表现出固有的局限性。在这项研究中，提出了一个机器学习 （ML） 框架来预测 Cu/Ni 包层箔的全场应变演变，并使用解释分析方法定量评估了本征空隙的影响。通过数字图像相关性和计算机断层扫描提取和整合局部应变和空隙数据。为了适应构建的数据集的性质，参考时间序列预测的概念建立了 ML 模型。随后，研究了体积分数 （VVF） 、面积和空隙大小等微观结构特征的影响，以及它们在驱动局部应变进化中的作用。这种方法成功地预测了应变定位，并准确地确定了塑性不稳定的开始和裂纹开始的位置。VVF 被确定为最主要的因素，其次是沿拉伸方向的空隙尺寸和晶粒尺寸。观察到 VVF 和晶粒尺寸之间的相关性最强，这种相关性随着时间尺度的增加而增强。此外，随着空隙聚结几乎完成，集中空隙分布对宏观应变浓度的促进作用将越来越明显。这些发现为探索变形和损伤之间的复杂关系提供了新的视角。