The global demand of polyethylene has been growing vigorously in many industrial fields. With the increase of demand, the requirements of the customers have also been getting sophisticated and strict. In order to meet market demands for polyethylene products, most of polyethylene manufacturers are trying to enhance the flexibility of plant operation. That means they are forced to have frequent grade transitions that cause to produce large amounts of off-spec. products. To carry out grade transitions effectively, it is essential to develop an on-line inferential system - a virtual on-line analyzer (VOA) - for prediction of polymer properties.
A VOA system for prediction of the polymer properties - Melt index (MI) and density - in a unimodal gas phase process with the fluidized bed reactor has been developed. This VOA system is based on the combination of both $1^{st}$ principles model and empirical correlations. In the proposed VOA system, the polymer properties are predicted with 97% confidence level in comparison with manual analysis.
A VOA system for an industrial bimodal high density polyethylene process has been developed and applied successfully. The VOA system estimates the melt index and density of the reactor products, and the pellet properties such as melt index, density and melt flow rate ratio (MFRR), which can be used to characterize the molecular weight distribution (MWD).
Both SISO MPC and MIMO MPC system using the simplified VOA model as a state estimator have been developed for an industrial bimodal polyethylene process. The simulation results show that this MPC system is applicable to the target process.
In conclusion, proposed VOA and MPC systems will allow polyethylene processes to handle the increasing market demands for various products and achieve the reduction in the product cycle time in a timely and optimal way.
