- 12:50 PM

MODELING of Unit Operations by Molecular Simulations

Danijel Babic, Artur Pereira Neto, and Andreas Pfennig. AVT - Thermal Process Engineering, RWTH Aachen University, Wuellnerstr. 5, Aachen, 52062, Germany

The properties and interactions of the molecules involved in chemical-engineering processes determine the behavior and thus the design of the used equipment. Nowadays it is still considered impossible to bridge the length and time scales between molecules and technical equipment in only one or a few steps of model hierarchy. Nevertheless it is generally accepted that the major effects in chemical engineering facilities can be modeled just with equilibrium and rate-based approaches. These effects have their origin on molecular level and corresponding data can be obtained by taking into account only the vicinity of the interface, i.e. the simulation of a few hundred or thousand particles is sufficient.

In this work a simulation tool has been developed that is able to predict the behavior of entire separation columns based on molecular simulations. In principle any countercurrent separation equipment can be described, until today we focus on distillation and solvent extraction. The benefit of this approach is that the simulations are performed at the operating conditions, i.e. composition, temperature, pressure and their gradients, that exactly correspond to the real apparatus. Thus it is not necessary to simulate data over the entire range of possible temperatures and compositions, where the data for the relevant conditions are then obtained afterwards by correlation or interpolation. Furthermore molecular simulations are able to predict multi-component effects of mass transfer as well as complex equilibrium behavior accurately and consistently.

In order to implement this concept in a first step the code of a molecular-simulation program has been modified as to allow different compartments and boundaries between them to be accounted for in a simulation. Then specific boundary properties can be applied to molecules that cross a border. By using different boundary properties all relevant aspects of a separation process can be depicted on molecular level: partly penetrable walls simulate the behavior of sieve trays, heating and cooling devices by influencing the molecular velocities and feed or product streams by inserting and removing particles from the simulation. It has already been shown that with this setup the principle behavior of a distillation column can be simulated with less than a thousand particles [1].

In the molecular approach the column is depicted by consecutive separation stages, i.e. the simulation of one separation stage is performed for two flowing phases in contact. Hereby the outflow of one stage is directly the inflow of the next stage with respect to flow rate, composition and temperature. In order to decrease the computation time, the simulation can be parallelized by simulating each stage on an individual PC. The connection of the individual stages (PCs) then depicts the separation column. The information that has to be exchanged between the stages, i.e. the PCs, doesn't have to be updated at every single simulation step. It is sufficient to update the time average of the feed and product fluxes only every few ten thousand steps. Thus the computational cost of the data exchange between the stages is negligible and the number of separation stages can be set to any desired value.

By performing a dimensional analysis the necessary residence time of the fluids in a molecular simulation that corresponds to a macroscopic separation stage can be determined. This means that the simulations don't just behave like adequately connected equilibrium calculations but rather they also include the non-ideal character of a stage, which can be either caused by multi-component effects of mass transfer or by the finite residence time of the fluids on one stage.

In the next steps we will focus on the experimental validation of the proposed simulation method. Therefore the simulated concentrations will be compared with the data obtained from distillation-column experiments. In this presentation we will highlight the modeling framework needed to predict the behavior of entire separation columns based on molecular simulations and present the simulation results for extraction and distillation processes. It will be demonstrated how multi-component and non-equilibrium effects influence the separation behavior.

1. A. Pfennig, “Distillation Simulated on Molecular Level”, Mol. Sim. 30(6), 361 366 (2004).

Web Page: www.avt.rwth-aachen.de/AVT/index.php?id=120&L=1