摘要
传统配方产品设计通常是基于经验和大量的实验。计算机辅助分子设计(CAMD)将配方设计问题转化为确定与目标产品功能相匹配的化学物质及其分子组成的问题,显著提高设计效率和准确性。然而,对于结构性产品(如洗涤剂、牙膏等),其性能主要由微观结构决定,且一些难以描述的性质(如泡沫性能)缺乏可量化的数据,导致定量构效关系(QSPR)模型的建立存在困难。针对这一问题,本文提出了一种结合分子动力学(MD)模拟与机器学习的计算机辅助配方设计方法。首先,利用分子动力学模拟研究配方产品中不同成分的微观行为及其形成的微观结构,揭示了成分间的相互作用机理。然后,从模拟结果中提取可量化的分子动力学描述符,结合贝叶斯神经网络构建QSPR模型,预测配方成分与产品性能之间的关系。最后,通过数学优化算法求解最优配方组成。本文以牙膏产品为例,设计了满足泡沫性能需求的配方。结果表明,该方法能够显著减少配方设计的搜索空间,提高设计效率和科学性,为配方设计指明了方向。未来,随着技术进步和市场需求的变化,计算机辅助设计方法将在化工领域发挥更加重要的作用。
Traditional formulated product design relies heavily on experience and extensive experimentation.Computer-aided molecular design(CAMD)transforms this process by identifying chemical substances and their molecular compositions which match the target product’s functionality,greatly enhancing design efficiency and accuracy.However,for structural products(e.g.,detergents,toothpaste),performance is mainly determined by microstructure,and quantifying certain properties(e.g.,foaming)is challenging due to insufficient data,making it difficult to establish quantitative structure-property relationship(QSPR)models.To address this,this paper proposes a computer-aided formulation design method that combines molecular dynamics(MD)simulation with machine learning.MD simulation is first used to study the microscopic behavior of different ingredients and their formed microstructure in the formulated product,revealing the interaction mechanisms among components.Then,quantifiable MD descriptors are extracted from the simulation results to construct a QSPR model in combination with Bayesian neural network,predicting the relationship between formulation composition and product performance.Finally,a mathematical optimization algorithm is employed to solve for the optimal formulation composition.A toothpaste formulation case study is provided,showing that this method can significantly reduce the formulation design search space and enhance design efficiency and scientificity.It also indicates the direction for formulation design.As technology advances and market demands evolve,computer-aided design methods are expected to play a more important role in the chemical engineering field in the future.
作者
齐妍
常昊
张磊
QI Yan;CHANG Hao;ZHANG Lei(Institute of Chemical Process Systems Engineering,School of Chemical Engineering,Dalian University of Technology,Dalian 116024,Liaoning,China)
出处
《化工进展》
北大核心
2025年第8期4341-4351,共11页
Chemical Industry and Engineering Progress
基金
国家自然科学基金(22278053,22078041)
国家自然科学基金优秀青年科学基金(22422801)。
关键词
分子动力学模拟
计算机辅助分子设计
配方设计
表面活性剂
定量构效关系
molecular dynamics simulation
computer-aided molecular design
formulation design
surfactants
quantitati ve structure-property relationship
