Ester rate soft-sensor in PET process

In this paper, through the analysis of esterification reaction, a soft-sensor model of ester rate which is the quality index in PET process is established. A modeling method is presented, which uses subtractive clustering to generate an initial T-S fuzzy model, and then the optimal fuzzy model is selected by rude-tuning, combined with fine-tuning, the radius of a cluster center. Before modeling secondary variable are selected and computed, errors of data samples are eliminated and normalized. Simulation results show that the model can be built fastly and has perfect generalization capability, and it can estimate ester rate efficiently.