(197d) A Design and Computation Software for Separating Multicomponent Mixtures

Mathew, T. J., Purdue University
Mobed, P., Purdue University
Tawarmalani, M., Purdue University
Agrawal, R., Purdue University
Tumbalam Gooty, R., Purdue University
Separation is one of the most frequently used processes for separation of multicomponent mixtures in the chemical industries, including oil and gas, air separation, biochemical, pharmaceutical, etc. Multicomponent mixtures are separated in sequential steps in multiple distillation columns, referred to as distillation configuration. The number of configurations that are feasible for achieving the separation task increases rapidly as the number of components in the feed mixture increases. Also, performing rigorous calculations for validating the design and economics of a distillation configuration is tedious, and there is no systematic way for identifying alternative configuration options without performing exhaustive calculations. For these reasons, it is challenging for practitioners to retrofit the configurations currently in operation or find the most economic configuration(s) among the vast search space of configurations in the early stages of process design.

We propose a short-cut method by combining Underwood’s method for estimating vapor flows in pinched columns with mass balance equations to estimate the flows and vapor duties in a distillation configuration. This method reduces the computation load drastically while maintaining a good approximation compared to the rigorous calculations. For a feed mixture of n components, the separation task can be achieved in n-1 columns, referred to as basic configuration, or in less/greater than n-1 columns, referred to as non-basic configurations. The short-cut method gives us the ability to search within the basic and non-basic search space for an optimal configuration that has minimum vapor duty, exergy loss, and/or capital and operating cost. Therefore, a systematic tool that suggests attractive configurations based on recent advances in distillation technology would have a large impact on energy, environment and economic aspects of the U.S. manufacturing sector.

In this work, we are proposing a design and computation software for identifying the attractive basic configurations that can be used by both industrial practitioners and academic peers. The underlying problem is formulated as non-linear programing (NLP) and mixed-integer non-linear programing (MINLP). Our state-of-the-art software solves the underlying problem to global optima using GAMS/BARON, and displays the attractive configurations that address the user’s needs in a friendly user interface. The visual output of the software not only helps in absorbing the results, it allows the practicing engineers to explore various operable and dividing-wall column analogs. The software is based on our recent advances in developing efficient algorithms for identifying implementable energy-efficient low-cost multicomponent distillation trains.