(508c) Multiple Hydrodynamic States in Trickle Flow Reactors: Towards Optimising Reactor Performance by Manipulation of the Hydrodynamic State

The hydrodynamic description of trickle flow is a complex field of study. One of the main contributors to the complexity is the existence of multiple hydrodynamic states (MHS). As early as 1978, Kan & Greenfield reported different pressure drop and holdup readings for the same liquid and gas feed rates. The study revealed that the flow history experienced by the bed determined the specific hydrodynamic state. Once this was established numerous studies confirmed the effect by using different flow histories (Christensen et al., 1986, Levec et al., 1986, Lazzaroni et al., 1988).  In these studies the MHS were reported as pressure drop and liquid holdup hysteresis loops.

The initial focus on MHS (or hysteresis) was to understand the mechanism that caused the variation in pressure drop and holdup measurements. The following have been suggested as explanations:

·        The difference between film and rivulet flow (Christensen et al., 1986)

• The change in tortuosity of the gas flow channels (Kan & Greenfield, 1978 &1979)
• The difference between advancing and receding contact angles at the gas-liquid-solid interface (Levec et al., 1986 & 1988)
• The non-uniformity of the flow (Wang et al., 1995).

These approaches were helpful in clarifying some of the MHS results but no theory could encompass all the observations.

In the past three years trickle flow MHS have been extensively studied at the
University of
Pretoria in
South Africa. The project is performed by The Reaction Engineering Group of the Department of Chemical Engineering and is sponsored by SASOL. The main aim of the project is to utilise the MHS principles on an industrial scale, whether by design or by the optimisation of existing trickle bed reactors. Although the work done up to date is fundamental by nature, the intention is to apply the insight on a commercial scale. In this abstract the highlights of the research done up to date are given in short and projections are made on the possible implications of the research.

As mentioned, the existence as well as possible reasons for the existence of MHS is well established in open literature. Since our goal is one of improving trickle bed reactor performance, the extent of possible variation of the hydrodynamic state is of crucial importance. This has not been addressed in literature where the hysteresis loops commonly exhibits small variation in the measured variable (approximately 10-40% variation). Furthermore it is also preferable to extend hydrodynamic measurements beyond pressure drop and liquid holdup. The extremes of MHS are expected to be dependent on the extremes of the flow history and in our work we propose that the extremes can be attained by a set of different prewetting procedures. The following prewetting procedures were used as limiting cases for MHS. These are:

·        Non-prewetted

·        Levec prewetted: the bed is flooded and drained and after residual holdup stabilisation the gas and liquid flows are introduced

·        Kan_L prewetted: the bed is operated in the pulse flow regime (by increasing liquid velocity) after which liquid flow rate is reduced to the desired set point (all at the desired gas flow rate)

·        Kan_G prewetted: the bed is operated in the pulse flow regime (by increasing gas velocity) after which gas flow rate is reduced to the desired set point (all at the desired liquid flow rate)

·        Super prewetted: the bed is flooded and gas and liquid flows are introduced once draining commences

Using the conventional water-nitrogen-glass beads (3mm) system, holdup, pressure drop as well as volumetric gas-liquid mass transfer (oxygen stripping from the water) was measured for the different prewetting procedures. Some of the results are shown in figure 1.

One will firstly note the significant amount of variation for the pressure drop and mass transfer coefficient for the different prewetted beds (up to 500% variation in pressure drop and 400% variation in mass transfer). One will further note that the lower extreme for prewetted beds is Levec prewetting, while no specific prewetting procedure could be identified as the upper extreme for all the measured variables. The Kan_G  prewetting exhibits high liquid holdup and relatively lower pressure drops, a phenomenon which can not be explained by any of the mentioned theories on MHS.

The extensive variation indicates that different prewetting procedures entail severe differences in the liquid morphology. A similar experiment to the one reported in figure 1 was performed, using g-alumina as packing and X-ray computed tomography to visualise the differences. The following shows the tomographic reconstruction of a cross sectional cut through the catalyst bed. Figure 2 (f-i) represents a visual enhancement of the image where the solid spheres are subtracted from the image.

The non-prewetted bed exhibits severe maldistribution where only certain areas of the bed are contacted by liquid. All the prewetted beds show a bed scale even distribution over the cross section, but the Levec prewetted bed exhibits larger (and less) localised liquid ?spots', indicating poorer spreading of the liquid. No evidence of liquid films are present and closer investigation suggest that smaller liquid ?spots' (compared to
films around a particle) might get lost in the analysis of the tomograph.

The liquid-contacting on a particle scale level can be better investigated by employing a colorimetric technique. By using the colorant Chrome-Azurol S in water it is possible to stain g-alumina particles only where liquid contacting occurs. The fractionally stained particles can be removed from the bed, photographed and analysed for wetting fraction. For Levec and Kan_L prewetting the particle wetting distributions (population of 3000 particles) are given in figure 3.

One will note severe differences between the prewetting modes, not only in terms of average wetting efficiency, but also in the particle wetting distribution. It is evident that Levec wetting results in a significant fraction of non-wetted or poorly wetted particles, while this is not the case for Kan_L prewetting.

The results give a clear picture of the distinct differences between the different MHS. Although only the limiting cases of the MHS were shown in the results, it should be noted that a continuum of MHS exists and that any point on the continuum can be accessed by combining the desired prewetting procedure with a gas/liquid hysteresis loop. At a first glance one might generalise by opting for the upper boundary of the MHS (optimum liquid distribution via
Kan or Super prewetting) as most lab scale trickle bed studies do. One should however keep in mind that most industrial trickle bed reactors operate with no prewetting or Levec prewetting at best, since the higher flow rates required for pulsing are beyond the capability of the equipment. Furthermore one might have a gas limited or volatile liquid phase reactions where dry catalyst surface actually enhances the overall reaction rate via direct gas-solid mass transfer. We are accordingly convinced that numerous industrial examples exist where the optimum operating point will lie somewhere within the continuum of MHS. For the cases where maximum liquid distribution is required, the super prewetting method (as opposed to the
Kan prewetting procedures) poses a practical solution for industrial reactors.

We have recently performed a study on  a commercial  trickle bed reactor and have confirmed the existence of MHS in this unit, regardless of its high pressure (40 bar) and low liquid surface tension (10 mN/m). Consequently we are hopeful that our current research can be extended to enhance trickle bed reactor performance. In view of this we are currently in the process of constructing a 2? high pressure bench scale trickle bed reactor. Within time we will be able to suggest pre-treatment procedures to establish the ideal hydrodynamic state for a specific reaction.     

References

Christensen, G., McGovern, S.J. and Sundaresan, S. (1986) A.I.Ch.E. J., 32, (10), 1677 - 1689.

Kan, K.M. and Greenfield, P.F. (1978) Ind.
Eng. Chem. Process Des. Dev., 17, 482 - 485.

Lazzaroni, C.L., Keselman, H.R. and Figoli, N.S. (1988) Ind.
Eng. Chem. Res., 27, 1132 - 1135.

Levec, J., Grosser, K. and Carbonell, R.G. (1988) A.I.Ch.E. J., 34, 1027 - 1030.

Levec, J., Saez, A.E. and Carbonell, R.G. (1986) A.I.Ch.E. J., 32, 369 - 380.

Wang, R., Mao, Z. and Chen, J. (1995) Chem.
Eng. Sci., 50, (14), 2321 - 2328.