(176b) Distillation Column Flooding Predictor | AIChE

(176b) Distillation Column Flooding Predictor


Mphahlele, M. G. - Presenter, University of Texas at Austin
Eldridge, R. B. - Presenter, University of Texas at Austin

The Dzyacky distillation column flooding predictor is an advanced process control strategy that utilizes a patented (US 5,748,538) pattern recognition system to identify the onset of pre-flood conditions in distillation columns. By avoiding flooding, the column operates in a stable mode with high throughput and the process becomes more energy efficient.

To facilitate implementation of the predictor methodology, a general method for predicting the occurrence of flooding in liquid-gas mass transfer devices, specifically trayed distillation columns, is required. This research will firstly develop a generalized dynamic model to describe the transient behavior of a sieve-tray distillation column operation. This model will be based on predicting the dynamic response of liquid hold-up, liquid level, pressure drop and mass transfer efficiency. Existing state-of-the-art correlations will be used in the model.

The ultimate goal of the research is to predict a priori the pre-flood alarm values for column operation based on first-principles models (equilibrium and rate-based models) using column dimensions and other column specifications (e.g. feed rates, feed composition, reflux ratio etc). These alarm values are key to predicting a pre-flood condition. Currently, the plant operator determines these values by using rules of thumb or guesswork.

The dynamic model results will be validated using pilot plant data from the Separations Research Program at the University of Texas, Austin. This validation will also involve using the model to identify pre-flood patterns. The dynamic model results will also be compared to data obtained from a commercial ethane/ethylene distillation column. Wavelet-based methods will be applied to the data to remove the background noise prior to model validation

Keywords: Flooding prediction, Distillation, Equilibrium stage model, Model validation, Dynamic modeling and Simulation, Pattern recognition system, Wavelet method.