In this paper a general model is presented for the prediction of octane blending values based on gasoline compositions containing Methyl Tertiary Butyl Ether. The models were fitted to data coming from design experiments for five Venezuelan refineries using data transformations and regression techniques such as stepwise model adjustments.
The models eliminate to some extent the dependency of prediction on components (alkylates, naphthas, reformates, etc.). The regression variables are generated from detailed gas chromatography data for each component, depending on the response variable.
The regression adjustments found were up to 94 % for research and 86 % for motor premium gasoline octane numbers. Regular gasolines had 92 % regression adjustments for research and 48 % for motor octane number. The residual standard deviations are 0.327, 0.529, 0.430 and 0.788 respectively.
The models for unleaded gasoline require only compositional data of the components and apply over all the hydrocarbon range of the experiments. The models found for the prediction of octane blending values are independent of the refineries and they can be used to maintain operational flexibility of the Venezuelan state-owned petroleum industry.