Neural network and structural coding method to predict the chromatographic retention index of straight-run gasoline

neural networks;  structural coding;  Gas chromatography retention index

The molecular structure of monomeric hydrocarbons in straight-run gasoline is digitally encoded, and the nonlinear correlation model between the gas chromatographic retention index and its molecular structure of monomeric hydrocarbons in straight-run gasoline is constructed by error backpropagation neural network algorithm, the neural network structure is 3 layers, the hidden layer nodes are 7, there are 15 inputs, corresponding to the 15-bit digital code of the monomer hydrocarbons, and 1 output, corresponding to the gas chromatographic retention index. The prediction results show that the correlation coefficient and standard deviation obtained by the error back-transmission algorithm are better than those obtained by the multiple linear regression method.