1-24. High-Throughput Design and Screening for Apatite and Porous Materials
Dingguo Xu1,2, Mingli Yang2 and Xindong Zhang2,3

1. College of Chemistry, Sichuan University
2. Research Center for Material Genome Engineering, Sichuan University 3. National Engineering Research Center for Biomaterials, Sichuan University

Abstract:As one of three key techniques in material genome engineering research, high-throughput (HT) computing could significantly reduce the research and deployment time and cost for new materials. Assisted by machine learning, it can further predict or design new material structure, and further screen the candidate materials with good properties. In this presentation, we constructed machine learning and HT computation models to bone repairing materials (e.g., apatite) and porous materials (e.g., MOF), respectively. For apatite, metal ion doping is widely accepted as one of ways to regulate the mechanic properties of implanted biomaterials. On the basis of artificial neural network and HT models, we successfully realized fast prediction of the stable ion-doped apatite structures. In industry, the production of H2 can be fulfilled using the steam methane reforming technique. Metal of organic framework (MOF) materials have considered to be one of efficient mechanisms to isolate the produced mixed gases. Several machine learning approaches were employed here, preliminary experiments can confirm our results. Finally, we have implemented all these machine learning models into our previously designed material HT computation software.

       磷灰石及多孔材料的高通量设计和筛选

徐定国 1, 2，杨明理 2，张兴栋 2, 3 

1.四川大学化学学院， 2.四川大学材料基因工程研究中心,

3.国家生物医学材料工程技术研究中心，成都，四川 

摘要:高通量计算技术作为“材料基因工程”三大支撑平台和关键技术之一，与材料数据库及高通量制备表征技术的交互融合有助于大幅度缩短材料研发时间和减少成本。结合机器学习的高 通量计算模拟技术能够预测和设计材料结构，筛选具有优异性能的候选材料。我们以骨修复材 料(磷灰石体系)和多孔材料(有机金属框架材料，MOF)为研究对象分别建立了相应的高通 量和机器学习模型。针对磷灰石体系，以金属离子掺杂后的稳定结构预测为研究对象，结合人 工神经网络和第一性原理高通量计算实现了快速预测稳定的掺杂结构。针对 MOF 材料，以筛选 能够选择性吸附分离蒸汽甲烷混合气体的材料为研究对象，结合 KNN，人工神经网络，随机森 林，SVM 和逻辑回归等机器学习方法建立的模型用于材料性能筛选，初步的实验也证明我们筛 选出的 MOF 材料能够有效分离蒸汽甲烷重整化混合气体。最后我们将这些机器学习方法融合进 了我们之前开发的材料高通量计算平台。