TY - JOUR
T1 - Quantitative structure-property relationship studies for predicting flash points of alkanes using group bond contribution method with back-propagation neural network
AU - Pan, Yong
AU - Jiang, Juncheng
AU - Wang, Zhirong
PY - 2007/8/17
Y1 - 2007/8/17
N2 - Models of relationships between structure and flash point of 92 alkanes were constructed by means of artificial neural network (ANN) using group bond contribution method. Group bonds were used as molecular structure descriptors which contained information of both group property and group connectivity in molecules, and the back-propagation (BP) neural network was employed for fitting the possible nonlinear relationship existed between the structure and property. The dataset of 92 alkanes was randomly divided into a training set (62), a validation set (15) and a testing set (15). The optimal condition of the neural network was obtained by adjusting various parameters by trial-and-error. Simulated with the final optimum BP neural network [9-5-1], the results showed that the predicted flash points were in good agreement with the experimental data, with the average absolute deviation being 4.8 K, and the root mean square error (RMS) being 6.86, which were shown to be more accurate than those of the multilinear regression method. The model proposed can be used not only to reveal the quantitative relation between flash points and molecular structures of alkanes, but also to predict the flash points of alkanes for chemical engineering.
AB - Models of relationships between structure and flash point of 92 alkanes were constructed by means of artificial neural network (ANN) using group bond contribution method. Group bonds were used as molecular structure descriptors which contained information of both group property and group connectivity in molecules, and the back-propagation (BP) neural network was employed for fitting the possible nonlinear relationship existed between the structure and property. The dataset of 92 alkanes was randomly divided into a training set (62), a validation set (15) and a testing set (15). The optimal condition of the neural network was obtained by adjusting various parameters by trial-and-error. Simulated with the final optimum BP neural network [9-5-1], the results showed that the predicted flash points were in good agreement with the experimental data, with the average absolute deviation being 4.8 K, and the root mean square error (RMS) being 6.86, which were shown to be more accurate than those of the multilinear regression method. The model proposed can be used not only to reveal the quantitative relation between flash points and molecular structures of alkanes, but also to predict the flash points of alkanes for chemical engineering.
KW - Alkane
KW - Back-propagation (BP) neural network
KW - Flash point
KW - Group bond contribution method
KW - Quantitative structure-property relationship (QSPR)
UR - http://www.scopus.com/inward/record.url?scp=34447637403&partnerID=8YFLogxK
U2 - 10.1016/j.jhazmat.2007.01.025
DO - 10.1016/j.jhazmat.2007.01.025
M3 - 文章
C2 - 17292543
AN - SCOPUS:34447637403
SN - 0304-3894
VL - 147
SP - 424
EP - 430
JO - Journal of Hazardous Materials
JF - Journal of Hazardous Materials
IS - 1-2
ER -