Simulação operacional de uma torre de destilação atmosférica via Aspen Plus e avaliação de modelos de analisadores virtuais

The atmospheric tower is one of the most important equipments of an unit of distillation of raw oil in a refinery. The optimization of the operation of this equipment is a constant goal in this industry type, because it makes possible plenty financial returns attractive. To take the operation of the close column of atmospheric distillation to its optimal operational point it is necessary to measure some key properties, in real time, for the performance of the control system. However some properties are not measured easily in real time, or still its measurement depends on the purchase of on-line analyzers that they can present expressive costs. An alternative for the obtaining of those properties is the use of inferences (virtual analyzers), that predict the variable wanted through mathematical models using secondary variables as entrance parameters. Two inferences, granted by REFAP S/A, were appraised with the purpose of determining that with larger predictive power of the temperature that distils 85% of the current of wanted hydrocarbons (T85%). The source of data, for the tests, it was originated in the stationary state by the software Aspen Plus, after the validation of a model for medium operational data collected in the own refinery. Two operational manners were proposed and sensibility analyses were generated so that their results were used as source of data for the inferential models. T85% coming of the simulator were considered pattern for the comparison with the results obtained by the inferences. The virtual analyzers were shown sensitive to the operation way the one that is submitted. This means that the adjustments done in the parameters of the models should be reviewed in case there is a significant operational change in the distillation system. For the operation in that the models were developed the acting of the two inferences was similar, could conclude that in the stationary state the predictions of both are satisfactory.