基于M-QSPR的含甲苯汽油辛烷值理论预测研究

Mixture quantitative structure,property relationship study for the octane number of toluene gasoline

从分子结构角度出发,针对含甲苯汽油辛烷值展开了混合物构效关系(M-QSPR)研究。应用SiRMS(Simplex Representation of Molecular Structure)描述符对含甲苯汽油混合体系的结构信息进行表征,运用遗传算法-偏最小二乘(GA-PLS)组合算法对SiRMS描述符进行筛选和建模,得到了相应的研究法辛烷值(RON)预测模型和马达法辛烷值(MON)预测模型型。随后对模型进行验证、应用域分析和机理解释。结果表明,两个预测模型均具有较优的稳定性和预测能力。两个预测模型对其测试集的预测复相关系数分别为0.999和0.987,均方根误差分别为0.714和1.248。此外本研究揭示了影响含甲苯汽油辛烷值的主要结构因素及其影响规律。本文研究方法的提出为工程上提供了一种根据分子结构预测含甲苯汽油辛烷值的新方法。

A mixture quantitative structure,property relationship （M-QSPR study was made for the toluene gasoline mixture in term of molecular structure by using the SiRMS （Simplex Representation of Molecular Structure descriptors to represent the molecular structure of the toluene gasoline mixtures. Then the Genetic Algorithm-Partial Least Squares （GA-PLS method was employed to select the optimal and most relevant SiRMS descriptors and the prediction model was developed. Subsequently, the validation and interpretation were performed and the Williams plot was drawn to analyze the applicability domain. As the result, both two models have good stability and predictivity. The values of multiple correlation coefficient （R2 for research octane number （RON model and motor octane number （MON models are 0.999 and 0.987, and the values of Root Mean Squared Error （RMSE for them are 0.714 and 1.248. Furthermore, the main structure factors influencing the octane number of toluene gasoline were discovered. This study provides a new method for predicting the octane numbers of toluene gasolines in engineering.