This study explores integrating models with different degrees of detail for optimizing azeotropic batch distillation systems. A detailed dynamic model is used from outside the optimization program both to verify feasibility of the design and to update the parameters needed by the optimization model. The updated parameters are the constant relative volatilities between pseudo-components, used in a binary Fenske-Underwood-Gilliland-type model. The approach was used to optimize the design of a batch process for the recovery of spent isopropyl alcohol, which works cyclically to separate the excess water, satisfying an environmentally acceptable specification, and using cyclohexane as entrainer, which in turn is recovered in the same process and recycled. The approach permitted optimizing the batch sizes, number of separation stages, the reflux ratios of a piecewise constant multilevel reflux policy, the extent of each separation, and the size of the intermediate cuts to be recycled, with affordable computation and problem setup times. (c) 2005 American Institute of Chemical Engineers.