Determination of quantitative structure-octane rating relationships of hydrocarbons by genetic algorithms

The influence of the molecular structure of organic compounds on their knocking behavior was determined using a nonbinary genetic algorithm (GA). The molecular structures of 240 potential gasoline components were described by use of 16 different structural groups. Partial octane numbers were calculated for these structural groups dependent on the substance classes paraffins, naphthenes, olefins, aromatics and oxygenates. The sum of the calculated partial octane numbers supplies the octane number of the compound. A multiple linear regression (MLR), a neural network and a GA were used for the computations of the connections between the structural groups and the knock ratings. Results obtained by GA were significantly better than these obtained by MLR. The correlation coefficients between the calculated and the test engine determined blended research octane numbers were for paraffins R-GA = 0.988 (R-MLR = 0.954), naphthenes R = 0.975 (0.877), olefins R = 0.984 (0.959), aromatics R = 0.945 (0.877) and for the oxygenates R = 0.964 (0.919). The calculated partial octane numbers allows the quantitative determination of influences of structure modifications on the knocking characteristics of gasoline components. (C) 1999 Elsevier Science B.V. All rights reserved.