(145c) Improving Design and Scale-up of Columns with Structured Packings Using Artificial Intelligence | AIChE

(145c) Improving Design and Scale-up of Columns with Structured Packings Using Artificial Intelligence


Rehfeldt, S., Technical University of Munich
Klein, H., Technical University of Munich
Emissions of the process industry contribute significantly to global warming. A substantial reduction of these emissions may be accomplished by establishing more efficient processes with properly designed structured packings in distillation columns. Potential benefits range from lower pressure drop to the feasibility of a more sophisticated process integration [1]. Retrofitting existing tray columns with structured packings can yield better separation efficiency in addition to a capacity increase. Therefore, the development of an improved design and scale-up method for columns with structured packings is the primary goal of the joint project ReProvAP.

One major disadvantage of using structured packings in distillation columns are the insecurities in their design process. Either multiple correlations to predict the mass transfer in the packing or small-scale experiments for scale-up are necessary to empirically determine the required packing height for a given separation task [1]. Especially with non-ideal mixtures, this leads to unreliable results and high safety margins.

The present work gives an overview of currently used strategies and problems in scale-up of packed columns in the process industry. Furthermore, existing mass transfer models are compared with insights from industrial scale-up approaches for packed column design and examined for improvement potential. The goal is to reduce the reliance on experimental investigations. To this end, an approach to predict HETP (Height Equivalent to a Theoretical Plate) values based on artificial intelligence (AI) is presented.

A similar study can be found in literature [2], where the influence of thermophysical properties and other relevant parameters on HETP values is investigated and compared. The present work will expand on this by including additional influencing factors in the model that were previously not considered, as well as incorporating aqueous and wide-boiling systems, too. These results can be used to improve correlations and scale-up for non-ideal mixtures.


[1] J. Stichlmair, H. Klein, S. Rehfeldt: Distillation: principles and practice, 2nd edition. Wiley-VCH, 2021.

[2] G. S. Pollock, R. B. Eldridge: Neural Network Modeling of Structured Packing Height Equivalent to a Theoretical Plate. Ind. Eng. Chem. Res. 39 (5) (2000), 1520–1525. DOI: 10.1021/ie9908128.