Two-Phase Flow Dynamical Simulations And Modelling Of Bubble Column Reactors

Bubble column reactors are widely employed in petrochemical, energy, environmental and bio-engineering processes because of their features of being geometrically simple and superior gas-liquid interfacial mass and heat transfer. A fundamental investigation and understanding of fluid dynamics in bubble column reactors would be very beneficial to the optimization of operation, design of high efficiency reactor structure and extension of their applications. As the main flow inside bubble column is driven by rising bubbles, hydrodynamics involved exhibits complicated status due to the interaction and strong coupling between liquid and gas phases. In addition, the presence of a large amount of bubbles or bubble clusters causes bubble reactors to be opaque. These factors make it extremely difficult to experimentally study the details of flow patterns in bubble columns. Recent progress on CFD modeling and the knowledge of bubble dynamics has made this possible, i.e. the use of numerical simulation as an important tool to investigate multiphase flow behavior in bubble column reactors. This dissertation attempts to use numerical modeling approach to study the fluid dynamics in bubble columns, in particular seeking for scale-up modelling of PX oxidation reactors.This project has employed two-fluid model in all of the simulations due to its lower requirement for computational resource and potential for industrial applications. The model regards the dispersed bubble phase as a pseudo-fluid and postulates the gas and liquid phases to be permeable.In view of the existing difficulties and problems encountered in CFD modelling of flows in bubble columns, this dissertation has firstly conducted detailed numerical modelling experiments to quantitatively study the effects of different physical models currently employed in CFD modelling, in particular the interfacial force models and turbulence models in the simulations of gas-liquid flows. The examples to refelect these effects have been demonstrated in Chapter 3. The modelling results have clearly indicated that the turbulence model and lift force are the key factors influencing predictions of dynamic fluctuation behavior of bubble plume in bubble columns. The influence of added-mass force can be negligible. For a bubble column equiped with a single-pipe sparging gas distributor at the centre of the bottom of the bubble column, the simulation shows that it is necessary to include both the lift force and turbulent dispersion force in the modeling in order to obtain the reasonable gas phase distributions. While for a bubble column with an uniform gas sparger, reasonable simulation results can be still obtained when excluding the lift force and turbulent dispersion force in the two-fluid model.A single scalar transport equation which describes bubble size changes characterised by bubble interfacial areas was incorporated into the simulations through UDF subroutines. The effect of bubble coalescence and break-up was taken into consideration in the adopted models. The predicted time-averaged axial liquid velocities, gas hold-up and gas phase interfacial areas were compared with the available experimental results. It was revealed from the simulation that the predicted overall gas holdup, local gas holdup, local phase velocities and liquid phase turbulence kinetic energy are in good agreement with the experimental data while the Reynolds stresses for both gas and phases are poorly estimated, very likely due to the lost of the details of instantaneous flow in time-averaging process for deriving the k～εturbulent model.The bubble population balance model (BPBM or referred to as SBPBM in Chapter 6 of the dissertation) was systemically studied in Chapter 5 in order to properly predict the bubble size distribution in bubble columns. A numerical method to realise single-bubble-phase population balance model (SBPBM) was proposed and a corresponding code for execution of such method was programmed. Modification to "Prince bubble coalescence model" was introduced to account for the effect of the free-moving space among bubbles and effective bubble migration due to turbulence on the efficiency of bubble collision. In simulations, cases of both the use of Luo & Svendsen's model together with Prince's model and adoption of the modified Prince model were tested for prediction of hydrodynamics in a bubble column. The simulations show that the use of the original "Prince model" predicts that the mean bubble size increases initially but decreases with the increase of superficial velocity. However, adoption of the modified Prince's model forecasts that the mean bubble size increases with the increase of superficial velocity, consistent with the experimental observations reported in the literature.Based on the applications of the SBPBM model and perception of its limitations, an improved model to describe bubble size distribution-two-bubble-phase population balance model (TBPBM) was proposed in Chapter 6. The corresponding numerical method for exectution of such model was also developed and incorporated into the self-coding of UDF subroutines in a similar way to the work done based on the SBPBM model. The TBPBM model coupled with two-fluid model was used to simulate the flow inside a bubble column. The simulation has clearly demonstrated that the adoption of the TBPBM model significantly improves the predictions in comparison to the use of the mean bubble size model and SBPBM model. The mean bubble size predicted by using the TBPBM model is apparently larger than that by using the SBPBM model, consistent with the practical observations. Detailed examinations have been also performed in Chapter 6 to look into the influence of the boundary conditions, mesh size and numerical schemes adopted in the modelling on the simulations using the BPBM model. The obtained results indicate that the mesh size and numerical schemes employed in the simulations have a significant impact on gas holdup prediction since the volume fraction transport equation is susceptible to mesh size and the numerical schemes. As a result, a better meshing method was proposed.Based on the investigations in the preceding chapters, CFD modelling approach was employed to systemically assess the influence of adoption of a short size draft tube and configurations of gas distributors on fluid dynamic in bubble columns. Simulation results demonstrate that allocation of a short draft tube which is fixed in the lower part of the bubble column can remarkably increase the liquid circulation velocity and instantaneous velocity without reduction of the overall gas holdup, clearly being beneficial to suspension of solid particles in the bubble column. CFD simulation also shows that the configurations of gas distributor and the allocation of gas sparging pipes have a vital impact on flow patterns, mixing time and bubble sizes in the bubble column. An optimal allocation for four sparging pipe gas distributor in the bubble column is obtained by CFD modelling, which agrees well with the experimental observation.Finally, a modification to "Luo & Svendsen's bubble breakup model" was proposed to account for the constraint of eddy energy density limiting the the minmum fragmentation fration of a bubble, i.e. the energy density of an eddy (ρ1uγ2/2) should be larger than the capalliary force (σ/dmin) of the smaller fragmental daughter bubble. On this basis, three-dimensional numerical simulations of four different diameters of bubble columns were conducted to investigate the scale-up effect using the SBPBM model and the modified "Luo & Svendsen's bubble breakup model" together with "Prince coalescence model". The simulations show that the effect of the bubble column diameter on the overall gas holdup may be negligible, but the radial profiles of time-averaged local gas holdup become flatter when bubble column diameter increases. Furthermore, the predicted time-averaged axial liquid velocity in bubble column center was found to be proportional to the cubic root of the bubble column diameter. The simulations reveal that the diameter of a bubble column also affects the bubble size and bubble distributions but the effect is not notale as expected. However, significant differences of flow patterns in different bubble columns are predicted, indicating that the induced liquid flows up in the bubble column center with a meandering way for small diameter bubble column reactors, and a large circulation loop with induced liquid flowing up along the bubble column wall for larger bubble columns.