BPNN用于含氮、硫杂环润滑油添加剂抗磨性能的定量构效关系研究

A Quantitative Structure Tribo-ability Relationship Model for the Antiwear Properties of N/S-containing Heterocyclic Lubricant Additives using Back Propagation Neural Network

摘要: 采用反向传播神经网络法(Back Propagation Neural Network，简称：BPNN)对31种含氮、硫的2-烷基黄原酸酯类润滑油添加剂的抗磨性能进行了摩擦学定量构效关系(Quantitative Structure Tribo-ability Relationship，简称：QSTR)的研究，得到了具有良好的稳定性和预测能力的BPNN-QSTR模型(R²=0.998 4，R²(LOO)=0.695 9，q²=0.879 1). 参考输入层的12种2D和3D结构描述符的敏感度，对影响抗磨性能的分子结构进行了相应的探讨. 结果表明：分子中的N和S杂原子对其抗磨损性能有显著的影响；同时，分子长度、所含双键S原子和芳香环数量以及分子支化程度等都是影响抗磨性能的主要因素.

Abstract: Quantitative structure tribo-ability relationship (QSTR) was studied by back propagation neural network (BPNN) method for 31 kinds of N/S-containing heterocyclic lubricant additives. The BPNN-QSTR model exhibits good accuracy and predictability (R²=0.998 4, R²(LOO)=0.695 9, q²=0.879 1). Considering the sensitivities of 13 kinds of 2D and 3D structural descriptors included in BPNN inputs, the effects of each descriptor on antiwear performance were discussed. The results show that N and S heteroatoms had significant impacts on the antiwear performance. In addition, the antiwear performance was also affected on the length of molecule, the number of S with one double bond and aromatic rings, and branching degree of molecule.