(248n) Computationally Efficient Evaluation of Energetically Improved Distillation Processes for the Separation of Non-Ideal Mixtures

The majority of fluid separations in the chemical industry are still performed by means of distillation columns, accounting for about half of the total energy consumption in this industry and 10-15% of the US industrial energy use [1]. Besides the pure quantity of these processes, with approximately 40,000 distillation columns in the US [2], another reason for this high energy consumption is to be found in the low thermodynamic efficiency of 10-15% for conventional distillation processes [3]. Consequently, energetically improving these processes provides a huge potential for energy savings [4]. While a variety of different options for heat integration, thermal coupling and heat pump configurations have been proposed in order to improve the energy efficiency of distillation processes, even academic investigations are almost always restricted to specific options, such as dividing wall columns (DWC) [5], or the implementation of heat-pump designs up to a fully heat-integrated distillation column (HIDiC) [6]. The same holds true for the investigation of hybrid separation processes, such as membrane-assisted distillation or the combination of crystallization and distillation [7], for which the achieved improvements are almost always related to conventional distillation processes as benchmark.

One of the reasons for this non-compliance of energetically improved distillation processes is the limited availability of efficient means for their evaluation. For the separation of zeotropic mixtures, powerful synthesis methods to derive all simple and thermally coupled configurations have been developed (see e.g. [8]) and superstructure-based optimization approaches have been proposed to determine the optimal design of these configurations [9]. However, in the way these approaches have been described the application of the methods was limited to the separation of ideal mixtures.

Similar statements can be given for the design of heat-integrated distillation processes, for which elegant superstructure optimization approaches have been presented [10-11], but the applicability was limited to the separation of ideal mixtures as well. Despite this limitation, these methods are also not available in commercial process modeling and simulation software, such that complex configurations like DWCs have to be assembled from the connection of a thermodynamically equivalent combination of conventional distillation column models [12]. Current approaches for screening thermally coupled configurations rely on shortcut methods like the Fenske-Underwood-Gilliland method (FUGM) for the determination of the most suitable options. These are further verified by means of rigorous simulation [13]. However, the FUGM is limited in applicability and accuracy by the underlying assumptions of constant molar overflow and constant relative volatilities. The rigorous simulation of the thermally coupled configurations is complicated by the absence of readily available models, while the complex configurations, which are required in order to depict the thermally coupled configurations by means of conventional column models, have to be initialized manually [14].

In order to efficiently address the problem of evaluating a large set of process configurations, including heat-integration, thermal coupling, heat pumps and multi-effect distillation, a systematic screening procedure is proposed, which operates in two steps. In a first step a full enumeration of all process variants based on reliable shortcut methods is proposed. In order to account for non-ideal thermodynamic models, the rectification body method (RBM) [15] is integrated in an automated evaluation procedure which automatically investigates various ways of heat-integration between single columns by means of variating the operating pressures in the different columns. Thermally coupled configurations and the thermodynamically equivalent DWC configurations are evaluated based on the decomposition method proposed by Carlberg and Westerberg [16], while the distribution of the intermediate boiling component is optimized by means of a simple bisection algorithm. The application of vapor recompression is performed based on isentropic compression computation. For the separation of a mixture into three products more than 30 different configurations are evaluated, including the non-improved direct, indirect and intermediate split. Due to the computationally efficiency of the shortcut-based approach even uncertainty associated to the feed composition, utility costs and the thermodynamic model can be addressed, based on a screening of these configurations for several hundred different scenarios. Based on the full enumeration of all the configurations and potentially all resulting scenarios the most efficient configurations as well as the most relevant scenarios can be identified, and can further be investigated by means of a rigorous optimization approach based on equilibrium-tray models.

In order to evaluate the cost optimal process design from the various configurations, an efficient optimization-based method is proposed. The method relies on a superstructure approach including rigorous thermodynamic models for equilibrium and enthalpy computations [17], making it applicable to ideal as well as non-ideal mixtures. Optimal designs are determined on the basis of the total annualized costs (TAC), considering investment and operating costs. All configurations associated to one sequence of splits are initialized by means of the common column sequence, while subsequently the optimization of each energetically improved configuration is performed by means of an automated adaption of the superstructure. The different process options are optimized for minimum energy demand and TAC. The application of the overall method is demonstrated for the separation of ternary non-ideal mixtures.

REFERENCES

[1] D.S. Scholl and R.P. Lively, Nature, Vol. 532 (2016), 435â??437

[2] P.C. Wankat, â??Separation Process Engineeringâ?, Prentice Hall, 3^rdedn. (2012)

[3] G.M. Koeijer and S. Kjelstrup, IJOT, 3 (2010), 105-110

[4] A.A. Kiss, J. Chem. Technol. Biot. 89 (2014), 479â??498

[7] A. Harwardt and W. Marquardt, AIChE J., 58 (2012), 3740-3750

[6] I.J. Halvorsen and S. Skogestad, J. Nat. Gas Sci. Eng., 3 (2011), 571-580

[7] M. Skiborowski, A. Harwardt and W. Marquardt, Annu. Rev. Chem. Biomol. Eng., 4 (2013), 45-68

[8] V.H. Shah and R. Agrawal, AIChE J., 56 (2010), 1759-1775

[9] G. Dünnebier and C.C. Pantelides, Ind. Eng. Chem. Res., 38 (1999), 162-176

[10] H.Y. Yeomans and I.E. Grossmann, Comp. Chem. Eng., 23 (1999), 1135-1151

[11] M. Barttfield and P.A. Aguirre, Ind. Eng. Chem. Res., 42 (2003), 3441-3457

[12] A.C. Dimian, C.S. Bildea and A.A. Kiss, Computer-Aided Chem. Eng., 35, Elsevir, (2014)

[13] M. Errico, B.-G. Rong, G. Tola and I. Turunen, Chem. Eng. Proc.: Proc. Int. 48 (2009), 907-920.

[14] I. DejanoviÄ?, L. MatijaÅ¡eviÄ? and Ž. OlujiÄ?, Chem. Eng. Proc.: Proc. Int. 49 (2010), 559-580.

[15] J. Bausa, R. von Watzdorf and W. Marquardt, AIChE J. 44 (1998) 2181â??2198.

[16] N.A. Carlberg and A.W. Westerberg, Ind. Eng. Chem. Res., 28 (1989), 1379

[17] M. Skiborowski, A. Harwardt and W. Marquardt, 2015, Comput. Chem. Eng. 72, 34â??51.