Design and operation issues using NLP superstructure modeling

Till present, models that determined batch plants configurations in the chemical process industry resorted to models with binary variables to represent the different admissible options. This approach allowed representing the problem in a simple way while considering a significant number of alternatives. Nevertheless, the non-convexity that arises when dealing with detailed models for representing the involved units operation prevents its correct resolution or has a low performance. This work presents a representation of the problem through a superstructure that takes explicitly into account all the alternatives without resorting to binary variables. By using extremely simple modeling, it is possible to manage an appropriate number of options for this type of problems by means of a non-linear programming (NLP) model. Moreover, it is possible to consider duplication in series of production stages, which is an alternative that has not been used till now. This approach is posed for the case of a fermentors network. The solution is reached with very low requirements as regards employed computer time and without the aforementioned difficulties.