Robust Dynamic State Estimation of a Binary Distillation Column

摘要

A nonlinear geometric observer is developed, which infers the distillate and residue compositions in a highly-nonlinear binary distillation column from temperature measurements. The estimator performance is evaluated by imposing severe step changes to the input variables. In particular, the capability of the observer reconstruction is assessed with regard to ill-conditioning of the observability matrix, initialisation of the composition estimates, and noise in the temperature measurements.