Bilinear identification of a binary distillation column
Thesis DisciplineChemical Engineering
Degree GrantorUniversity of Canterbury
Degree NameDoctor of Philosophy
Low order bilinear models of chemical processes, suitable for control applications over a wide operating region, were identified. A case study on a simulated heated tank system showed that there was considerable potential for the use of bilinear models for chemical processes in which mass or energy balance equations have a bilinear structure. Bilinear models were more accurate than linear models at fitting the simulated system's behaviour over a wide range of operating points. The U-D identification algorithm used, proved to be both robust and reliable. Multi-input multi-output bilinear models of both a simulated and an experimental binary distillation column were identified using the U-D algorithm. Two factors affected the trend in the column parameters; the turn-down of the column and the shifting of the composition profile along the column. The bilinear models were good at following changes due to the turn-down, but not the large fluctuations due to changes in the composition profile. This limitation was due to the simplified liquid/vapour equilibrium relation inherent in the bilinear model, which was invalid if there were large fluctuations in composition on any of the plates.