113 Application To Reactor Engineering

An overall procedure for reactor engineering is discussed in Chapter 1. Additional comments on the engineering of bubble column reactors are made in Section 11.1. Some of these are repeated here to emphasize their importance. As for any other reactor, conventional reaction engineering models are first used to evaluate the influence of various fluid dynamic characteristics (mixing, volume fraction, interfacial area, heat transfer coefficient and so on) on overall performance of a reactor. These studies lead to the formulation of specific duties for the reactor, which may be related to specific demands of the underlying fluid dynamics. After finalizing these demands on reactor fluid dynamics, the reactor engineer has to evolve a suitable hardware configuration to fulfill these demands. CFD-based models can make significant contributions at this stage.

After completing reaction-engineering work, it is first necessary to evolve a reactor configuration before one can start evaluating whether such hardware can perform the expected duties. In the case of bubble columns, evolving reactor hardware involves at least the following (also see Fig. 11.2):

• Bubble column configuration/dimensions: simple versus loop configuration, diameter, height to diameter ratio, internals (draft tube, radial baffles, cooling/heating coils, packings), feed inlet/outlet nozzles (for gas as well as liquid phase components), gas-liquid separator, foam breakers/entrainment reducers, necessary process monitoring sensors and so on. Each of these will further involve selection from various alternatives. Even for the very simple bubble column reactor with no internals, it is necessary to select a suitable sparger in addition to the basic sizing.

• Design of sparger: type, sparger resistance, sparger holes, gas velocity and bubble size generated at sparger, operability of the sparger at varying gas loads, design of plenum chamber (if necessary) and so on.

More often than not, the reactor engineer evolves more than one configuration to meet the expected duties. It is necessary to examine these alternatives and to select a few short-listed configurations for further studies. Normally, this short-listing procedure involves various heuristic arguments based on prior experience and available information. CFD models can be used to quickly evaluate the various configurations to assist this short-listing procedure. To estimate commonly required reactor duties like liquid phase mixing and heat transfer coefficient, it may be sufficient to predict time-averaged liquid velocity profiles and corresponding time-averaged gas volume fraction profiles. For such cases, it may be adequate to use two-dimensional models. One example of such a model developed by Ranade (1997) was discussed in an earlier section. The work of Krishna et al. (2000a) also confirms that two-dimensional models may give adequately accurate estimations of overall gas volume fraction and liquid circulation velocities. Such two-dimensional models may also be used to qualitatively evaluate the influence of different reactor internals, such as draft tubes and radial baffles, on liquid phase mixing in the reactor. Ranade (1993b) demonstrated such an application to evaluate the influence of radial baffles on mixing in bubble column reactors.

The two-dimensional models are, however, unable to capture details of flow structures. If it is essential to capture such flow structures in the simulated results, it is necessary to use three-dimensional models. For example, to evaluate different spargers, it will be necessary to examine the role of unsteady structures on mixing. Ranade and Tayalia (2001) evaluated liquid phase mixing caused by single and double ring spargers using a computational model. They considered an axis-symmetric, two-dimensional domain as well as the full 3D domain (which does not require imposition of symmetry at the column axis). Though estimated volume-averaged quantities such as gas volume fraction, liquid velocity are within 10% for the 2D and 3D models, the details of flow structures are quite different. Typical results obtained for a double-ring sparger are shown in Fig. 11.17. Comparison of the predicted flow field for single and double ring spargers using a 3D model are shown in Fig. 11.18. The complete 3D computational model was able to differentiate between single and double ring spargers and can, therefore, be used to evaluate different spargers. Recently Padial et al. (2000) used a three-dimensional model to evaluate the influence of size and location of draft tube on the fluid dynamics of bubble column reactors. Such models can then be extended to simulate the influence of draft tube on mixing in bubble column reactors.

Once a small number of reactor configurations have been short-listed based on the CFD models discussed above, more rigorous simulations and rigorous experimental verification (and calibration, if necessary) of the computational models can be undertaken. The behavior of gas-liquid dispersions is known to be very sensitive to impurities and therefore it is essential to undertake a systematic experimental program at this stage. Scale-down methodologies should be used to arrive at a suitable experimental program. These small-scale experiments are invariably carried out in

FIGURE 11.17 Typical results obtained for double ring sparger (From Ranade and Tayalia, 2001). (a) Vector plot, (b) particle streak lines.
FIGURE 11.18 Comparison of predicted flow field for double (a) and single (b) ring spargers (from Ranade and Tayalia, 2001).

simple geometries and different conditions than actual operating conditions. Available information on the influence of pressure and temperature should be used to select appropriate model fluids for these experiments. Detailed CFD models should then be developed to simulate the fluid dynamics of a small-scale experimental set-up under representative conditions. The computational model is then enhanced further until it leads to adequately accurate simulations of the observed fluid dynamics. The validated CFD model can then be used to extrapolate the experimental data and to simulate fluid dynamics under actual operating conditions. An example of the application of such a methodology to a loop reactor is discussed in Chapter 9. Here we briefly discuss two recent examples from the published literature.

The first is concerned with optimization of an industrial ozonation reactor (Cockx et al., 1999). Ozonation reactors are used to remove microorganisms or micropollu-tants from drinking water. The efficiency of these reactors depends on liquid phase mixing (since disinfection kinetics is approximately first order) and gas-liquid mass transfer. Cockx et al. (1999) developed a computational model using a Eulerian-Eulerian framework. The model was evaluated first by comparing predicted results with a pilot-scale airlift reactor. Different sub-models, such as drag coefficient, effective bubble diameter and so on, were calibrated to obtain adequate agreement between predicted and experimental results. The computational model was then used to simulate the fluid dynamics and performance of an industrial-scale (350 m3) ozonation reactor. Although local measurements of flow variables were not available for the industrial-scale reactor, some local measurements of ozone concentrations and residence time distribution data were available. These data were used to validate predictions of the computational model. These comparisons are shown in Fig. 11.19. It can be seen that agreement is adequate for most reactor engineering applications. The validated CFD model was then used to optimize a larger ozonation reactor by suitably modifying internals. The model was used to evaluate alternative reactor configurations and to evolve a final configuration. Initial and modified configurations of

0 0

Post a comment