Materials Big Data: Technologies and Applications
Tong-Yi Zhang，Quan Qian，Wencong Lu，Yuexing Han, Yue Liu, Runhai Ouyang, Jiong Yang
Center of Materials informatics and Data Science, Shanghai University, Shanghai 200444,China
ABSTRACT: Materials data, especially experimental data on the mechanical behaviors of materials, are usually “small” in comparison with these internet “big data”. Most normal data are easily understood without any special knowledge, while it is a must and prerequisite that the background in materials science and engineering is required to understand, interpret, and analyze materials data. Materials data are often high-dimensional in feature space and small in size, viz., very sparse, and scatter greatly viz., big noisy. Great challenges are faced in how to effectively standardize, collect, and store materials data. The goal is to promote the integrated expert-knowledge and data-driven new materials discovery and the informatization and intelligentization of materials green manufacture and production with low cost and high quality, and to improve the performance and safety during materials service.
This presentation briefly introduces the progress of the Shanghai MGI Data platform in terms of data standardization, collection, storage, security, and protection of intellectual property. The Shanghai MGI Data platform includes materials databases, machine learning tools, high-throughput thermoelectric computation algorithm, and phase field simulation software. The presented examples focus on the strategy of data-driven novel materials discovery with the integration of domain knowledge in different stages of machine learning to improve the quality of sample data and model accuracy; the enhancement of model predictive performance by “formula deduction” via novel methods such as Evolutionary Symbolic Regression and the SISSO (Sure Independence Screening and Sparsifying Operator); the image de-noising, feature extraction, phase segmentation, and microstructure identification by digital-image-processing methods for revealing the underlying relationships of “composition-processing-structure-performance” of materials.
Keywords: Material Big Data; Machine learning; data-driven materials discovery
Tong-Yi Zhang earned Master degree in 1982 and PhD in 1985, majoring in materials physics, from University of Science and Technology
Beijing, China. From 1993 to 2015, he worked at Hong Kong University of Science and Technology, as Lecturer, Associate Professor, Professor, Chair Professor, and Fang Professor of Engineering. He is the Founding Dean of the Materials Genome Institute, Shanghai University, and the Founding Director of the MGI division in the Chinese Materials Research Society (CMRS), which
organizes the MGI symposium in the CMRS annual meeting every year. He is also a professor at Harbin Institute of Technology (Shenzhen) since 2020. He was a vice president of the International Congress on Fracture (ICF) 2013-2017 and now is a director of ICF executive committee. He was a recipient of the 2018 Prize for Scientific and Technological Progress from the HLHL Foundation, the Second Prizes of 2007 and 1987 State Natural Science Award, China, and the 1988 National Award for Young Scientists, China. He became ICF Fellow in 2013, Fellow of the Hong Kong Academy of Engineering Sciences in 2012, Member of Chinese Academy of Sciences in 2011, Senior Research Fellow of Croucher Foundation, Hong Kong, in 2003, Fellow of ASM International, USA, in 2001. He was Associate Editor-in-Chief of Science China Technological Sciences 2013 – 2017 and Editor-in-Chief since 2018. He is also Fracture and Continuum Mechanics Subject-Editor of the journal, Theoretical and Applied Fracture Mechanics (2013 – present). He is one of the active scholars and scientists who promote and foster the developments of Materials Genome Engineering and Materials/Mechanics Informatics, which integrate data science, artificial intelligence, and machine learning with materials science and engineering, and mechanics science and engineering, respectively.