Neuro-based formulation to predict fouling threshold in crude preheaters

For preheat exchangers of a crude distillation unit (CDU), operating in conditions such that fouling minimized is crucial. A number of semi-empirical models called “threshold fouling models” were developed by various researchers to predict crude oil fouling behavior in crude preheaters of CDUs. In this study, an artificial neural network (ANN) model has been employed to develop a set of mathematical formulations to identify regions where there is less/no fouling. The comparisons between results of the developed neuro-based formulation and three threshold fouling models showed the use of neuro-based model resulted in significant improvements in terms of predicting crude oil fouling behavior of various laboratory and plant data. The approach of developing neuro-based models to predict fouling behavior can be readily applied in CDUs to identify more accurate fouling/no-fouling operating zones leading to an enhancement in the operation of crude preheaters.