Polymer quality control in an industrial scale polyethylene fluidized bed reactor has been addressed. Since online measurements of polymer properties (melt index and density) are not available, they must be controlled indirectly via other available measurements. In the present paper, two algebraic equations correlating polyethylene melt index and density with the measureable concentrations of chemical components are obtained. Having the desired polyethylene properties and using these correlations, desired concentrations of chemical components are calculated and used via corresponding control loops. By using the infrequently available polyethylene property measurements, the correlation parameters are updated. In order to simulate the fluidized bed reactor behaviour, a comprehensive two phase model including bubble and emulsion phases is used. Six control loops with PI controllers are considered to regulate the reactor operating conditions. To improve the performances of the component concentration loops, ratio control strategy in a cascade framework is implemented. The effectiveness of proposed reactor control structure for several scenarios comprising the rejection of operational disturbances, set-point tracking for polyethylene production amount and grade changes and compensating the model uncertainties is demonstrated via computer simulation. The results indicate that the polyethylene properties are well controlled with the proposed inferential control strategy.