结合力场开发，高通量分子模拟计算和机器学习方法、预测分子热力学及输运性质的方法和软件，为软物资材料研发提供基础数据，解决实验数据不足的问题。研究内容主要包括：开发普适的具有可迁移性的全原子力场；优化用分子模拟计算热力学和输运性质的方法；结合力场和优选的计算方法形成完整的计算包，实现高效自动运行的方案；通过大规模计算对比实验数据优化和验证力场和算法参数形成标准的分子模拟计算程序；用高通量分子模拟计算数据弥补实验数据的不足，开发深度机器学习方法，形成一个数值预测模型。在线发布计算方法力场参数和计算预测数据。

This project intends to develop methods and software for predicting molecular thermodynamic and transport properties by combining high-throughput molecular simulation and machine learning. The purpose is to provide data and to solve the problem of insufficient data for developments of soft materials. The main content of the research includes: developing a general all-atom force field with transferability; optimizing calculation methods for predicting thermodynamic and transport properties based on molecular simulation data; forming a calculation package by combining the general force field and optimized methods; carrying out high-throughput molecular simulations to supplement the experimental data and provide sufficient data for machine learning; and to build a numerical prediction model based on deep machine learning. The computational methods, force field parameters and the prediction data will be published online.