深度学习辅助的等规聚丙烯硬弹性薄膜同步辐射SAXS图像高通量结构识别

高时空分辨的现代同步辐射光源结合原位在线研究装置是研究高分子薄膜在加工过程中多尺度结构演变的有力手段。然而，随之而来的一大挑战是实现对同步辐射科学大数据的高效处理和分析。针对同步辐射原位跟踪等规聚丙烯硬弹性薄膜在单向拉伸过程中所获取的海量小角X射线散射图（SAXS），本项目将利用深度学习算法开发高通量结构识别方法。其中，建立训练深度神经网络所需的数据集尤为关键。鉴于此，本项目计划开展等规聚丙烯硬弹性薄膜在温度和应变二维参数空间中的离线拉伸实验，制备具有不同应变的薄膜后拉伸样品，并对样品进行SAXS和扫描电子显微镜（SEM）结构表征，建立样品微观结构在倒易空间和实空间中的复杂映射关系，完成对SAXS图像的结构属性标记。将训练好的深度神经网络用于单向拉伸外场下等规聚丙烯硬弹性薄膜同步辐射SAXS图像所表征微观结构的高通量识别，揭示其在非平衡态条件下的多尺度结构演变规律及加工-结构关系。

The combination of modern synchrotron radiation light sources with high spatiotemporal resolution and in-situ instruments is a powerful tool to study the multi-scale structural evolutions of polymer films during processing. However, an enormous challenge is the high efficient treatment and analysis of synchrotron radiation scientific big data. In this project, by utilizing the deep learning algorithms, we will develop a method to realize the high-throughput identification of microstructures characterized by synchrotron radiation small-angle X-ray scattering (SAXS) images of isotactic polypropylene (iPP) hard-elastic films during in-situ uniaxial tensile deformations. Therein, the acquirement of the dataset for training the deep neural networks is very crucial. To this end, by performing ex-situ tensile experiments, we will prepare post-stretch iPP film specimens with different strains at the two-dimensional temperature-strain parameter space. Then, the microstructure of these specimens will be characterized by SAXS and scanning electron microscope. In this way, we can establish the complicated mappings between the microstructure of specimens in the reciprocal and real spaces, and accomplish the structural labeling of SAXS images. Finally, the well-trained deep neural networks can be used to achieve the high-throughput identification of microstructures characterized by large-scale synchrotron radiation SAXS images of iPP hard-elastic films under uniaxial tensile deformations. From this, we can unveil the multi-scale structural evolution of iPP hard-elastic film, and establish its process-structure relationship under the non-equilibrium conditions.