Establishing quantitative structure tribo-ability relationship model using Bayesian regularization neural network

• DOI: 10.1007/s40544-016-0104-z
• Publication Date: 2016-03-28T07:13:02Z
• Authors: Xinlei Gao; Kang Dai; Zhan Wang; Tingting Wang; Junbo He
• Keywords: Regularization; Heteroatom

Abstract Quantitative structure-activity relationship methods are used to study the quantitative structure triboability (QSTR), which refers tribology capability of a compound from calculation descriptors. Here, we Bayesian regularization neural network (BRNN) establish QSTR prediction model. Two-dimensional (2D) BRNN–QSTR models can flexibly and easily estimate lubricant-additive antiwear properties. Our results show that electron transfer heteroatoms (such as S, P, O, N) in molecule improve ability. We also found molecular connectivity indices good descriptors 2D models.