This paper presents a new methodology for synthesizing a static inferential-feedforward control scheme for petroleum fractionator fed with undefined blended mixture of hydrocarbons. Traditional steady-state analysis are used to evaluate two control schemes, direct and indirect products quality control schemes. Results indicate that despite the simplicity of implementation of the indirect quality control scheme which infers unmeasurable products quality from secondary measurements, it is very sensitive to modeling errors, and therefore the direct quality control scheme is more favorable. In order to compensate for undesirable time delays caused by the on-line composition analysers in the direct quality feedback control scheme, methods are presented to design an inferential-feedforward control structure that infers unmeasurable feed composition disturbances from secondary measurements and adjusts the nominal feedforward ratios of the manipulated variables to the feed flow rate, to maintain the desired products quality. First, the feed composition is represented by three groups, namely, paraffins, naphthenes, and aromatics (PNA). Then, the sensitivity analysis via a steady-state simulation are used in determining how the nominal feedforward ratios can be adapted. Singular value decomposition (SVD) technique is performed to select two tray temperatures which can detect changes in the PNA composition of the feed. This proposed structure consists of correlations obtained by running a simulation of the column, and has a great potential for enhancing the performance of the direct quality feedback control scheme.