有机化合物粘度的定量构性关系研究

A quantitative structure-property model for prediction of the liquid viscosity of organic compounds

本文对177个有机化合物,用HyperChem6.0软件中分子力学方法MM+进行几何结构优化并经构象搜寻得到了最低能量构象的几何结构,并通过分子描述符软件计算了表征分子的电子、拓扑、几何结构的共102分子个描述符,用偏最小二乘法(PLS)建立其粘度定量构性关系(QSPR).本文采用Monte Carlo模拟退火方法进行变量筛选,最终筛选出14个描述符.模型验证采用了留一法.所建立的最后的最优模型的结果为:Q2=0.85,R2=0.87,表明该模型具有较好的预测能力.

A quantitative structure-property relationship （QSPR） technique is developed for prediction of liquid viscosities of various organic compounds. The geometries of the 177 organic compounds are optimized by molecular mechanics MM＋ approach, and 102 descriptors including electronic, topological, geometric and molecular shape indices for each molecule are calculated with descriptor software. 14 descriptors are selected by using Monte Carlo annealing approach. Then the relationship between liquid viscosity and molecular parameters is established by Partial Least Square method （PLS）. The result shows that the model has good prediction ability.