Tray efficiency effects in batch distillation
Computer simulation has long been recognised as a useful tool in improved process operation and design studies. Commercial simulation packages now available for batch distillation studies typically assume constant tray efficiency. Here, on the basis of both practical work and computer simulation, the effects of tray efficiency variation with tray liquid composition on model accuracy and column performance are investigated. Detailed modelling studies were carried out on a pilot batch distillation unit and tray efficiency was found to be an important factor affecting the model fidelity. Distillation of different methanol/water mixtures revealed that tray efficiency varies with the mixture composition on the tray, the form of the variation being for the efficiency to pass through a minimum at intermediate compositions. This variation of tray efficiency with tray composition is a known phenomenon, which has not been included in batch distillation simulations even though tray compositions change significantly during a batch run. The model developed in this work (Variable Efficiency Model) includes the tray efficiency variation with mixture composition and results in an evident improvement in model accuracy for methanol/water distillation. The potential effects of strong tray efficiency dependence on mixture composition, at a more general level, are investigated using two case studies, based on hypothetical extensions of the tray efficiency concentration dependence observed for methanol/water mixtures. In extreme cases, the efficiency-composition dependence could introduce a significant additional non-linearity to the process behaviour, resulting in unexpected composition and temperature movements. To quantify the potential significance of these effects, the economic performance of a column based on simulation using the Variable Efficiency Model was compared with its performance, using an overall column efficiency (which is the common practice). Using fixed column efficiency was found to under-predict column performance for low purity products and over-predict performance for high purity products.