Search

Few rigorous models exist in literature for describing the temperature profile during start-up periods of distillation columns. In this work,a model is developed by applying several radial basis neural networks to data collected during the start-up period of a batch distillation column consisting of ethanol and water. Neural network efficiency training is introduced through rescale entry pre-processing series. To obtain the temperature profile,data points are obtained along different points of the column,and the results are applied to multiple networks. This allows construction of the temperature profile in the column consisting of a mean square error less than the maximum established values set during the efficiency pre-processing (mse = 0.001) of the networks. This model also allows observation of transition in the column from the empty cold state to the steady state,normally a challenge in conventional models. [ABSTRACT FROM AUTHOR]