92 Example 1 Suspension Polymerization Reactor

Suspension polymerization is an old and relatively simple process to produce polymers and copolymers for various applications. It is mainly used to produce specialty copolymers which have high value but low volume demand. Because of the operational problems associated with the transport of highly viscous droplets in suspension, continuous operation of suspension polymerization process is difficult: polymerization is carried out in a stirred reactor operated in batch mode. Controlling the particle size distribution (PSD) is one of the major reactor engineering objectives apart from high conversion and selectivity. Although the process has been widely studied, the understanding of factors that affect the PSD and several other physical phenomena that occur inside the vessel is still limited. Most suspension polymerization reactors are designed and operated based on wisdom accumulated from prior experience. In this example, the potential of using computational flow modeling to enhance our understanding and, thereby, enhance control of the performance of a suspension polymerization reactor is discussed.

Let us consider a typical suspension polymerization process for manufacturing polymer beads (for example, polystyrene beads for ion-exchange resins). For such a process, control of particle size distribution is crucial as it determines the usable yield from the process. Different applications demand different ranges of particle sizes. Typical ion-exchange applications may require polymer bead diameters within the range 250 to 1000 ^m. Any polymer particles falling outside this range are a waste of raw materials. For specialty applications, demands on particle size distribution are even more stringent. A typical suspension polymerization reactor is shown schematically in Fig. 9.1. While many operating protocols are used, it is common to disperse catalyzed monomer into aqueous phase containing a suspending or stabilizing agent (Leng and Quarderer, 1982). A certain time period is allowed for drops to attain a stable size distribution, after which the batch is heated to polymerization temperature. It is important that coalescence or agglomeration is prevented during polymerization. Failure to achieve adequate stabilization may lead to mass polymerization and reactor shutdown. Of course, avoiding such a possibility is no longer a problem, but enhancing the yield of polymer beads within the usable range is still a challenge for a reactor engineer.

Despite several studies spanning five decades, the understanding of factors which control PSD and the several phenomena that occur inside the stirred reactor is still limited. Several (design and operating) parameters affect the PSD. Some of the more important parameters include:

• type of impeller (shape and number of blades),

• number of impellers and their locations,

• impeller diameter,

• reactor geometry (shape, height to diameter ratio);

• water to monomer ratio;

• chemical recipe (type and concentration of initiator, stabilizer, surfactant, catalyst and so on).

It is not possible (and also not necessary) to review all the available information on suspension polymerization here. Some recent reviews may be referred to for more information (Vivaldo-Lima et al., 1997, 1998). Most of the answers to the questions related to chemical recipe may have to be obtained by conducting specific experiments. Other questions, such as selection of operating temperature, batch time etc., can be answered by developing a conventional reaction engineering model (a single-drop polymerization reactor model, by assuming complete mixing, will give information about the progress of polymerization with time). The need for a detailed understanding of fluid dynamics is, however, essential to an understanding of the liquid-liquid dispersion process occurring in the reactor. The breakage of monomer liquid phase into individual droplets and their dispersion within the reactor ultimately controls the particle size distribution. This is where computational flow modeling can make significant contributions.

The literature on drop breakage is extensive (see the recent review by Zhou and Kresta, 1998). Grossly approximating, one may state that the main drop breakage and coalescence events occur at the impeller at intervals defined by the mean circulation time. Different drop size distributions arise due to equilibrium between drop breakup and coalescence. For the case of stabilized dispersion (such as used in suspension polymerization reactors), the extent of coalescence is significantly reduced. For such systems, it is necessary to understand the relationship between fluid dynamics and drop breakage as well as between reactor hardware and the resulting fluid dynamics to evolve suitable reactor design. Several models have been developed to describe drop breakage in turbulent flows (see, for example, Kumar et al., 1998). A suitable drop breakage model (and a coalescence model if necessary) can be combined with a population balance model to simulate drop size distribution and thereby bead size distribution of a batch suspension polymerization reactor.

These models require information about mean velocity and the turbulence field within the stirred vessels. Computational flow models can be developed to provide such fluid dynamic information required by the reactor models. Although in principle, it is possible to solve the population balance model equations within the CFM framework, a simplified compartment-mixing model may be adequate to simulate an industrial reactor. In this approach, a CFD model is developed to establish the relationship between reactor hardware and the resulting fluid dynamics. This information is used by a relatively simple, compartment-mixing model coupled with a population balance model (Vivaldo-Lima et al., 1998). The approach is shown schematically in Fig. 9.2. Detailed polymerization kinetics can be included. Vivaldo-Lima et al. (1998) have successfully used such an approach to predict particle size distribution (PSD) of the product polymer. Their two-compartment model was able to capture the bi-modal behavior observed in the experimental PSD data. After adequate validation, such a computational model can be used to optimize reactor configuration and operation to enhance reactor performance.

It is also possible to use computational flow models without explicitly linking them with models predicting the particle size distribution. In such an approach, one can use insight gained from the results of the CFD model by implicitly combining it with an understanding of drop breakage phenomena and polymerization reaction, to evolve the desired reactor engineering solution. Recently, Ranade (1999a) demonstrated such an application to enhance the performance of an industrial suspension polymerization reactor. Ranade (1999a) considered the case of an existing industrial polymerization reactor agitated with two pitched blade impellers. The reactor was designed based on laboratory-scale experiments carried out in a 5 liter reactor and a few experiments on a pilot scale reactor (700 liter). The plant scale reactor was operated in batch mode. The yield of usable polymer beads (particles) was about 65%, which clearly indicated scope for enhancing the reactor performance. The approach of Ranade (1999a) is discussed below.

As mentioned earlier, both chemical (catalyst, surfactants, stabilizers) and physical (fluid dynamics, energy dissipation rates, circulation time and so on) factors control the performance of the suspension polymerization reactor. It is first necessary to examine the available experimental data to clearly understand the role of these chemical and physical factors. The available data indicates that the yield of usable polymer beads in laboratory scale reactor is more than 85%. Laboratory experiments were then planned to examine the sensitivity of the yield to various parameters of the polymerization recipe under the same hydrodynamic conditions. These experiments showed that the yield is relatively insensitive to small deviations in the chemical recipe. Analysis of the available data on pilot and plant scale indicated a progressive decrease in the yield of usable polymer beads from laboratory to pilot to plant scale. This analysis and some indirect evidence suggested that it may be possible to re-design the plant-scale reactor hardware to generate better fluid dynamics and mixing to increase the yield of particles in the desired size range.

One may use an approach discussed earlier in which a comprehensive mathematical model is developed, which relates the reactor hardware to fluid dynamics and polymerization reactions in one framework. However, it is extremely difficult

Population balance models /

coalescence breakup models: drop size distributions

CFD models: flow characteristics of impeller and bulk regions, exchange between these two regions

Population balance models /

coalescence breakup models: drop size distributions

FIGURE 9.2 Modeling a suspension polymerization reactor (approach of Vivaldo-Lima et al., 1998).

FIGURE 9.2 Modeling a suspension polymerization reactor (approach of Vivaldo-Lima et al., 1998).

and time consuming to develop such an approach for an industrial polymerization reactor. The uncertainties in kinetics and physicochemical properties coupled with the complexities of coalescence-break-up models may raise doubts about the direct correspondence between model predictions and plant performance. Ranade (1999a) therefore used an approximate approach, in which the first step is to understand the role of fluid dynamics in controlling the PSD without mathematically relating them. It is first necessary to identify the controlling drop breakage mechanism (shear, elongation, turbulence and so on). Without going into specific details of any particular application, it can be said that:

Wide distribution of shear strain rates within the reactor will result in a wider PSD. This means impeller rotational speed should be kept small enough to ensure narrower strain rate distribution. It should, of course, be able to provide the necessary bulk flow and should be able to keep the monomer droplets well dispersed within the reactor. Adequate dispersion of monomer droplets ensures that they all experience the same environment and leads to narrower PSD.

• Impeller size, shape and speed should ensure that the turbulence energy dissipation rates within the impeller zone are not excessive so as to avoid unwanted finer particles.

Several other qualitative requirements may be added. These may, however, be sufficient to illustrate the possible application of CFM. These fluid dynamical requirements form the 'wish list' of the reactor engineer, who has to evolve a suitable reactor configuration (height to diameter ratio, impeller type, size, location, number and so on) to satisfy this wish list. The conventional way is to modify some standard configuration, test it in pilot scale and then scale it up for the full-scale plant. Unfortunately, because of the costs and time involved in testing new configurations at pilot scale, usually new concepts/designs are sidelined in favor of known configurations. In such cases, CFD-based models can make substantial contributions. CFD models can be used to select an appropriate impeller, number of impellers, location of feed pipes, to satisfy the fluid dynamical 'wish list'. If the fluid dynamical 'wish list' is evolved carefully, the CFD-based model will be useful in optimizing the polymerization reactor without explicitly developing the detailed reaction and particle size distribution model. It may be necessary to modify the 'wish list' based on the understanding gained via CFD models. In most cases, however, the whole process converges to an appropriate solution in a couple of iterations.

After establishing the fluid dynamical requirement, the first step is to analyze the fluid dynamics of the existing industrial reactor. Details of the computational modeling of flow generated in a stirred reactor are discussed in Chapter 10. Without discussing those details, results reported by Ranade (1999a) are discussed here. A typical computational grid and predicted results for the flow generated by two pitched blade turbines are shown in Fig. 9.3. The simulated flow field was used to predict

confidentiality constraints). (a) Grid; (b) vector plot; (c) contours of turbulent KE (white: highest value; black: lowest value). Reproduced in colour plate section between pages 210 and 211.

FIGURE 9.4 Two alternatives to enhance reactor performance. (a) Four pitched blade turbine, (b) two pitched blade turbine with cage.

circulation time distributions, volume averaged velocity and other relevant flow characteristics. The average as well as maximum and minimum shear rates and energy dissipation rates were examined. Evaluation of these simulation results by combining the available information/understanding of drop breakage and the performance of a working reactor in the plant, indicated that it may be beneficial to reduce the impeller rotational speed without reducing the volume-averaged velocity to ensure the required bulk flow. One way to achieve this is to use more impellers and/or use larger diameter impellers. Another alternative is to combine the two pitched blade impellers with a cage to facilitate dispersion at lower speeds. To narrow the circulation time distribution, a draft tube could be used, however, in view of the possibility of polymer deposition on a draft tube and the subsequent cleaning problems, the option of a draft tube was not considered further. Various alternative reactor configurations

FIGURE 9.5 Simulated flow field for two alternative reactor configurations (white: highest value; black: lowest value; legend not shown due to confidentiality constraints). (a) Four pitched blade turbine, (b) two pitched blade turbine with cage (left: vector plots; right: contours of turbulent kinetic energy.) Reproduced in colour plate section between pages 210 and 211.

FIGURE 9.5 Simulated flow field for two alternative reactor configurations (white: highest value; black: lowest value; legend not shown due to confidentiality constraints). (a) Four pitched blade turbine, (b) two pitched blade turbine with cage (left: vector plots; right: contours of turbulent kinetic energy.) Reproduced in colour plate section between pages 210 and 211.

were evolved and screened heuristically to check whether they satisfied the 'wish list'. Shortlisted configurations were then studied using the computational model. The two alternative reactor configurations are shown in Fig. 9.4. Flows generated by these configurations were then simulated. The simulated flow field indicated that the volume-averaged velocity within the reactor is adequate (see Fig. 9.5 for sample results). The four-impeller configuration was found to be more effective in reducing the heterogeneity within the vessel. The pitched blade impellers, however, generate strong velocity gradients just below the impellers, which may widen the resulting PSD. The shape of the blade was, therefore, modified to generate as uniform a velocity as possible across the blade length. To achieve this, a simple phenomenological model which approximately relates the axial flow generated by the blade with blade angle and blade width was used:

This proportionality was used to determine suitable blade width (BW ) and blade angle (0 ) profiles along its length. The flow generated by these hydrofoil impellers was then simulated. Critical evaluation and comparison of the flow generated by hydrofoil impellers with that generated by PTD indicated that with four hydrofoil impellers, the polymerization reactor may be operated at an impeller speed which gives minimum variation in circulation time and much less heterogeneous distribution of turbulent energy dissipation rates, without affecting the bulk flow within the vessel. Plant trials with the new hydrofoil impellers resulted in 15-20% increase in the yield of usable polymer beads (Ranade, 1999a). Thus, in this case, even without combining the PSD model and CFD framework, it was possible to enhance reactor performance by judicious application of computational flow modeling. In many practical industrial reactor engineering problems, such an approach may have to be adopted.

0 0

Post a comment