GOALI: Combining Discontinuous Molecular Dynamics and Chemical Process Simulation

摘要：0075883J。R Elliott在这项为期两年的探索性赠款中，将评估将分子建模工具与化学过程模拟包相结合的前景，以提供完整、严格和准确的物理性质预测和关联框架。化工过程模拟器正在成为工业化学知识的主要接口。最近人们发现微扰理论比以前认为的要准确得多，特别是对于多原子分子。不连续分子动力学(DMD)模拟与微扰理论和维里展开相结合，为分子模拟的各个方面的高杠杆计算工作提供了基础。现有的DMD程序将在低成本微处理器上以合理的速度运行必要的最少数量的模拟。与传统的工程基团贡献模型类似，所得到的链节势能模型将作为分子尺度的基团贡献。虽然这些链节势基本上是作为第一基团的贡献，但从分子几何出发的高阶效应将通过用户指定的分子结构中一阶基团的连接性得到明确的解决。最初的目标是测试纯流体和混合物平衡性质的不连续势模型的可转换性。将开发球体、二聚体、三聚体、6聚体、8聚体和苯的波动人口数据库，并用于开发最佳阶跃电位，以表示甲烷、乙烷、正己烷、正辛烷、苯、水、甲醇和乙醇的蒸汽压、密度和内能的实验值。将根据对正丁烷、正戊烷、正庚烷和正十烷的预测准确性来评估所产生的势的可转移性。对于混合物，将根据对三元混合物和正丙醇甲醇等混合物预测的准确性来评估可转移性，其中正丙醇中的羟基被具体回归。第二个目标是除了过去计算的平衡和共存性质外，还将实施计算输送性质的方法，如粘度、导热系数和扩散系数。由此产生的模拟将探索完全基于平衡性质开发的潜在模型在多大程度上可以应用于估计运输性质。人们试图“解剖”分子间势函数的模型，以确定哪些势与特定的物理性质最相关。通过类似于平衡性质的微扰观点，还将寻求给出参考流体的DMD模拟性质的吸引效应对传输性质的广泛适用的映射。例如，将检验从正辛烷扩散系数的关联结果和纯排斥性正庚烷的DMD模拟可以预测正庚烷扩散系数的程度。这项工作将在阿克伦大学和ChemStations，Inc.之间的合作下进行。如果这项探索性工作显示出希望，最终目标将是一个客户可以访问的互联网网站，客户可以预先零成本访问，并且根据所要求的服务器端计算的强度而相对较低的每小时费用。提供的服务将包括传输和平衡性质的分子建模，如蒸汽压、活度、水溶解度、辛醇分配系数、粘度，以及通过通用分子模型从其他性质的测量中推断有关一种性质的知识的能力。在可供选择的范围内，将全面收集半经验方法，并估计每项财产的准确性。该项目的范围包括根据大约1300种化合物的数据库对所有半经验模型和基于分子的模型的准确性进行彻底的评估。这些评价将构成探索阶段以外工作的重要部分。网络版还将包括基于快捷装置操作模型的流程图和流程模拟。过程资本和生产成本对估计的物理属性的敏感度将是一个菜单选项。快捷方式模型将作为直接在基于网络的环境中进行严格过程模拟的先驱。

ABSTRACTCTS-0075883J. R ElliottIn this two-year exploratory grant prospects will be evaluated for integrating molecular modeling tools with a chemical process simulation package to provide a complete, rigorous, and accurate framework for physical property prediction and correlation. Chemical process simulators are becoming the primary interface for industry chemical knowledge. It has been recently discovered that perturbation theory is much more accurate than previously appreciated, especially for polyatomic molecules. Discontinuous molecular dynamics (DMD) simulation combines with perturbation theory and virial expansion to provide a basis for highly leveraged computational effort in all aspects of molecular modeling. An existing DMD program will run the minimal number of simulations necessary with reasonable speed on low-cost microprocessors. The resulting segmental potential models will act as molecular scale group contributions, analogous to conventional engineering group contribution models. Although these segmental potentials act essentially as first group contributions, higher order effects derived from molecular geometry will be explicitly addressed through the connectivity of the first order groups in the user-designated molecular structure.The initial goal will be to test the transferability of discontinuous potential models for the equilibrium properties of pure fluids and mixtures. A database of fluctuation populations for spheres, dimers, trimers, 6mers, 8mers, and benzene will be developed and be used to develop optimal step potentials for representing experimental values for vapor pressures, density, and internal energy for methane, ethane, n-hexane, n-octane, benzene, water, methanol and ethanol. The transferability of the resulting potentials will be evaluated in terms of the accuracy of predictions for n-butane, n-pentane, n-heptane, n-nonane, and n-decane. For mixtures, the transferability will be evaluated in terms of the accuracy of predictions for ternary mixtures and mixtures like n-propanol+methanol, for which the hydroxyl group in n-propanol was specifically regressed.The second goal will be to implement methods of calculating transport properties like viscosity, thermal conductivity, and diffusivity in addition to the equilibrium and coexistence properties that have been computed in the past. Resulting simulations will probe the extent to which potential models developed solely on the basis of equilibrium properties can be applied to estimating transport properties. It is sought to "dissect" model intermolecular potential functions in the sense of identifying which pieces of the potential correlate most strongly with specific physical properties. Also sought will be broadly applicable mappings of attractive effects on transport properties given DMD simulated properties for reference fluids, by analogy to the perturbation perspective for equilibrium properties. For example, the extent to which the diffusivity of n-decane can be predicted from correlated results for the diffusivity of n-octane and DMD simulations for purely repulsive n-decane will be examined. The work will be performed with collaboration between The University of Akron and ChemStations, Inc. If this exploratory work shows promise, the ultimate goal will be an internet site which clients can access for zero cost up front and relatively low hourly fees varying according to the intensity of the server side computation requested. Services provided will include molecular modeling of transport and equilibrium properties like vapor pressure, activity, water solubility, octanol partition coefficients, viscosity, and the ability to infer knowledge about one property from measurements of other properties through a common molecular model. Within the range of options will be a comprehensive collection of semi-empirical methods with estimates of the accuracy of each property. The scope of this project includes a thorough evaluation of the accuracy all the semi-empirical models and the molecular based models against a database of approximately 1300 compounds. Those evaluations will comprise a significant portion of the work beyond the exploratory phase. Also included in the web-accessible version will be flowsheeting and process simulation based on shortcut unit operation models. The sensitivity of the process capital and production costs to the estimated physical properties will be a menu option. The shortcut model will serve as a precursor for rigorous process simulations directly within a web-based environment.