石油学报(石油加工) ›› 2018, Vol. 34 ›› Issue (1): 175-181.doi: 10.3969/j.issn.1001-8719.2018.01.024

• 研究报道 • 上一篇    下一篇

基于M-QSPR的乙醇-汽油参比燃料混合物辛烷值的理论预测

张彭非,潘勇,管进,蒋军成   

  1. 南京工业大学 安全科学与工程学院, 江苏 南京 210009
  • 收稿日期:2017-03-23 修回日期:2017-07-16 出版日期:2018-01-25 发布日期:2018-03-26
  • 通讯作者: 潘勇,男,教授,从事危险化学品安全研究;Tel: 025-83587305; E-mail: yongpannjut@163.com E-mail:yongpannjut@163.com
  • 作者简介:第一作者:张彭非,男,硕士研究生,从事化学物质危险特性及其分析鉴定技术研究;E-mail:zhang978398158@live.com
  • 基金资助:
    国家自然科学基金项目(21576136,21436006)和国家重点研发计划项目(2017YFC0804801)资助

Prediction of Octane Number for Ethanol-Primary Reference Fuel Mixtures Based on Quantitative Structure-Property Relationship Studies for Mixtures

ZHANG Pengfei, PAN Yong, GUAN Jin, JIANG Juncheng   

  1. College of Safety Science and Engineering, Nanjing Tech University, Nanjing 210009, China
  • Received:2017-03-23 Revised:2017-07-16 Online:2018-01-25 Published:2018-03-26

摘要: 从分子结构角度出发,针对乙醇汽油参比燃料混合物的辛烷值开展混合物的定量结构性质相关性(MQSPR)研究,建立相应的理论预测模型。采用原子碎片描述符对乙醇汽油参比燃料混合体系的结构特征进行表征,应用遗传多元线性回归算法从上述描述符中优化筛选出与其辛烷值最为密切相关的特征结构描述符,建立相应的两参数线性预测模型。模型复相关系数为0987,预测平均绝对误差为1478。模型内外部验证及稳定性分析结果表明,所建模型具有较强的稳健性及外部预测能力。对模型进行机理解释,揭示了影响乙醇汽油参比燃料混合物辛烷值的特征结构因素及其影响规律。该研究为工程上提供了一种根据分子结构预测乙醇汽油参比燃料混合物辛烷值的新方法。

关键词: 乙醇-汽油参比燃料混合物, 辛烷值, 预测, 混合物定量结构-性质相关性, 遗传-多元线性回归

Abstract: Ethanol gasoline has been widely used in engineering due to their higher octane ratings and wide range of sources. This study developed the Quantitative StructureProperty Relationship models for Mixtures (MQSPR) for ethanol primary reference fuel mixtures based on molecular structures. The Simplex Representation of Molecular Structure (SiRMS)descriptors of tetratomic fragments were employed to represent the molecular structures of the ethanol primary reference fuel mixtures. Genetic algorithm based multiple linear regression (GAMLR) method was employed to select the optimal and most relevant SiRMS descriptors related to the research octane number (RON), and develop the prediction model. The multiple correlation coefficients (R2) and the root mean squared error (RMSE) of the resulted model is 0987 and 1478, respectively. The developed model was then rigorously validated using multiple strategies. The results demonstrated the robustness, validity and satisfactory predictivity of the proposed model. The predominant structure characteristics responsible for RON of ethanol primary reference fuel mixtures were also identified through model interpretation. This study could be reasonably expected to provide a new method for reliable prediction the RON of ethanol primary reference fuel mixtures for engineering.

Key words: ethanol-primary reference fuel mixtures, octane number, prediction, M-QSPR, GA-MLR