April 2018

Special Focus: Petrochemical Technology

CFD simulation in chemical reaction engineering

Reaction engineering is a branch of chemical engineering that deals with the design and optimization of chemical reactors.

Aglave, R., Eppinger, T., Siemens PML Software

Reaction engineering is a branch of chemical engineering that deals with the design and optimization of chemical reactors. The goal is to optimize the transport processes (heat transfer, mass transfer and mixing) to improve the yield/conversion of desired products and to ensure the safe operation of the reactor. This means maximizing yield while minimizing costs. These costs could be related to the cost of feedstock, energy input, heat removal or cooling, stirring or agitation, pumping to increase pressure, frictional pressure loss, etc. 

FIG. 1. Essential aspects of chemical reactor design.
FIG. 1. Essential aspects of chemical reactor design.

From the standpoint of computational fluid dynamics (CFD), reaction engineering is the application of transport phenomenon and chemical kinetics knowledge to industrial systems. Chemical kinetics—the study of rates of chemical processes—is founded on the experimental study of how different conditions influence the speed of a chemical reaction, its mechanism and transition states, and the development of mathematical models of the reaction’s characteristics.

Reactor design and CFD

Reactor design includes several key facets: Phase, or state of the reactants and products (solid, gas, liquid or aqueous/dissolved in water); reaction type (single, multiple, parallel series or polymerization); whether a catalyst is involved; flow distribution and mixing; transport of the species; and mode of operation (i.e., batch, semi-batch or continuous). Most crucial are the underlying transport processes—fluid flow, heat transfer, mass transfer and reactions—which is where CFD simulation can add substantial value.

In reactor design, the process of taking the discovery of a new chemical with novel properties from lab to commercial production involves many steps (Fig. 1):

  • Conceptualization: Analysis of new chemistry and the business case
  • Lab scale: Analysis of kinetics, catalysis, thermodynamics, material properties and toxicity
  • Reactor selection: Analysis of flow regime, heat release, residence time distribution (RTD), liquid hourly space velocity (LHSV), as well as gas hourly space velocity (GHSV)
  • Engineering with idealized models: Analysis of plug flow or continuous stirred tank reactors (CSTR), volume sizing, overall heat transfer
  • Preliminary configuration: Analysis of vessel configuration, internals, baffles and coils
  • Scale-up simulations: Pilot-scale design, scale-up parameters and design space exploration
  • Final design: Extrapolation from scale-up rules, flow modeling, safety, risk and runaway analysis, dynamic modeling
  • Commercializing: Startup, troubleshooting and operating training.

Initial engineering begins with an idealized reactor model sized to lab/bench scale and is used to predict key variables in reactor behavior, including the reaction parameters, material properties, toxicity, ideal operating conditions and catalysts best suited for the job, as well as preliminary reactor dimensions. Next comes the preliminary reactor configuration, followed by scale-up and simulation of a pilot-scale design. Finding the “sweet spot” for the pilot-scale design is typically the stiffest challenge during the design process, as the affected parameters are linked non-linearly to each other, which means that each of the sub-processes scales differently. 

FIG. 2. Phases of reactor design and scaleup where CFD can add visibility, fidelity and confidence beyond experimental data alone.
FIG. 2. Phases of reactor design and scaleup where CFD can add visibility, fidelity and confidence beyond experimental data alone.

If a larger capacity is desired, an increase in the geometric size may be sufficient; however, this does not apply to the reaction, heat transfer or mixing. If the model is scaled with turbulence scales, it may result in extremely high revolutions per minute, or a geometrically unfeasible design. As a result, many different designs can be created, depending on the scale-up rule used. Achieving the perfect scale-up model requires design space exploration to find the optimum result. 

Final design is accomplished through extrapolation from scale-up rules, detailed flow modeling, assessment of safety risks and runaway reactions, and dynamic modeling of the entire system to see how the predicted reactions will work at plant scale. At each stage of the design process (Fig. 2), CFD can provide predictive capability, design exploration and optimization at better fidelity, speed and cost than can experimental data alone.

Reaction models

Depending on the physics of the reaction, a wide range of chemical reaction models are available in a proprietary modeling system,a regardless of whether the reaction is carried out in the gas phase or liquid phase, and whether a catalyst is present. Models that are supported include:

  • Gas phase reactions may be simple combustion reactions, or reactions that take place inside tubes, such as those in cracker furnaces where heat is supplied externally.
  • Liquid phase reactions may involve polymerization schemes or series-parallel liquid reactions with meso-/micro-mixing.
  • Custom reactions involve enzymatic reactions and fermentation, as well as user coding.
  • In cases where detailed chemical kinetics of a reaction need to be developed, modified or modeled in ideal flow conditions, a dedicated, proprietary tool is available.
FIG. 3. List of available reaction models for gas phase reaction flows.
FIG. 3. List of available reaction models for gas phase reaction flows.

Gas phase reacting flow models

The most basic types of flow models are for gas phase reacting flows where the different species (feedstocks) enter the reactor non-premixed, completely mixed or partly premixed. For each category, a host of proprietary models are available to simulate the reaction (Fig. 3). 

Some of these models are regarded as tabulated chemistry models to reduce computation times, while others make use of detailed chemistry using the previously mentioned proprietary tool. An example is a simple glass furnace, where air and fuel enter the domain non-premixed. One simple proprietary model, shown in Fig. 4, highlights the flame zone, the approximate region where NOX is formed, and the remainder of the combustion chamber. In this case, critical considerations in design and operation of the reactor include not only heat transfer, but also the production of pollutant species, such as NOx.

Validation studies conducted for this reactor design are summarized in Fig. 4. The top right plot shows the temperature profile at a location of 0.9 m downstream from the fuel inlet. The green dots show experimental measurements, while the blue line shows simulation results from a proprietary modeling system. Good agreement is visible between the two, providing the necessary confidence and predictive capability for designing and operating this type of reactor safely.

FIG. 4. Eddy breakup model of glass furnace showing flame zone and region of NO<sub>X</sub> formation within overall combustion chamber.
FIG. 4. Eddy breakup model of glass furnace showing flame zone and region of NOX formation within overall combustion chamber.

Also important is species concentration, such as the O2 mole fraction at the same location, which is shown 0.9 m from the fuel inlet. In Fig. 4, the bottom plot demonstrates good agreement between experimental measurements and simulation results. In situations such as this, where high-temperature processes make experimental work difficult and expensive, simulation provides an easier, more cost-effective way to explore quantities of interest.

Modeling process for heaters and crackers

In process heaters and naphtha crackers, reactants pass through a tube while heat is supplied by combustion outside the tube. The proprietary modeling system provides a simplified way to simulate these reactions by modeling the tubes as 1D plug flow reactors—an idealized model used to describe chemical reactions in continuous, flowing systems of cylindrical geometry—while the external combustion is modeled in 3D. This approach is much less computationally expensive than simulating the entire system in 3D.

FIG. 5. Proprietary modeling results for a coupled 1D/3D simulation of a steam methane reformer.
FIG. 5. Proprietary modeling results for a coupled 1D/3D simulation of a steam methane reformer.

On the outer tube wall, conduction, convection and radiation are modeled in 3D, while inside the tube, heat transfer via convection is modeled in 1D. The two simulations are then coupled at the junction of the tube wall. This allows the designer to observe how mass fractions of species behave inside the tubes, and at the same time view the heat transfer and temperature profiles outside the tubes in the furnace. Fig. 5 shows the model results for a coupled 1D/3D simulation of a steam methane reformer providing the axial distribution of temperature, heat flux and species concentration.

Chemical kinetics modeling tool

A standalone, gas phase, detailed chemical kinetics modeling toolb can be used for simulating ideal (simplified) reactor models. These models help develop and import reaction mechanisms to perform sensitivity analysis, validate experimental data and simplify the reaction mechanisms for use in CFD simulations (Fig. 6).

Surface chemistry formulation

An important class of detailed chemistry simulations is surface chemistry. In these actions, the reactants are initially adsorbed onto a surface medium that acts as a catalyst for the reaction. After the reaction, the products are desorbed and the surface is left unchanged. 

FIG. 6. Digital analysis providing flamelet libraries and lookup tables for gas and surface chemistry to the modeling system.
FIG. 6. Digital analysis providing flamelet libraries and lookup tables for gas and surface chemistry to the modeling system.

Surface chemistry can be modeled either with detailed chemistry formulations using the stiff differential equation solver in the proprietary model, or using global reaction mechanisms. One application where surface chemistry is important is designing packed-bed reactors, which consist of tubes filled with a packing material impregnated with catalysts to improve the contact between the two phases in the reaction. 

Design challenges include the accurate prediction of heat transfer, which is critical for the safe operation of such reactors. The modeling of such reactors provides critical insights into heat transfer if representative random packing is generated, and if contact resolution in meshing can be carried out efficiently and accurately (Fig. 7).

FIG. 7. The proprietary model provides an automated process to model and simulate packed-bed reactors.
FIG. 7. The proprietary model provides an automated process to model and simulate packed-bed reactors.

The proprietary modeling system provides an automated way to model and simulate packed-bed reactors. Through a graphical user interface, the designer can specify geometry conditions, particle properties, wall properties and particle-to-particle interactions, as well as other fluid properties and heat transfer simulations. Once these specifications are defined, the catalyst bed is generated using a built-in element modeling capability. The modeling system creates a mesh including boundary layers, and then progresses to carry out the CFD simulation, followed by post-processing, to show radial and axial porosity, velocity profiles, heat transfer and reactions, if necessary.

Multiphase systems

Performing lab-scale tests and validating them in a CFD simulation is an essential part of the scale-up operation. Such validation provides the necessary confidence in the robustness and fidelity of the model to make predictions at plant scales, where measurements may not be possible. One such example of determining the power calculation for a varying solids concentration (10 wt%, 20 wt%, 30 wt% and 40 wt%) is given. 

FIG. 8. Schematic of the lab-scale mixing vessel used in solids suspension experiments.
FIG. 8. Schematic of the lab-scale mixing vessel used in solids suspension experiments.

In this example, a four-bladed, pitched-bladed turbine (Fig. 8) was used to suspend sand (particle size 190 µm) in water at a speed of 600 rpm. The liquid and solid densities were 1,000 kg/m3 and 2,483 kg/m3, respectively. The tank had a height and diameter of 0.34 m, whereas the impeller diameter was 0.19 m.

Results (Fig. 9) show that simulation predictions not only accurately validate power consumption data, but also predict that the increase in power is sub-parity. Common correlations that are used to predict power require the lookup of a power number from available charts in literature. The conditions of the given system must be matched to those available in literature for geometric ratios, material properties and solids concentrations.

Most of the time, it is impossible to find a good match from the geometries and material properties for cases available in literature for factor S to those for a given design requirement; therefore, the confidence to choose the right power number is very low. This uncertainty results in inaccurate power prediction from correlation-based methods, as shown in Fig. 10.

FIG. 9. Power consumption comparison of experiment, simulation and correlation. The contours plot on the right shows the solids volume fraction for various solids loadings.
FIG. 9. Power consumption comparison of experiment, simulation and correlation. The contours plot on the right shows the solids volume fraction for various solids loadings.

Liquid phase reactions: Micro-mixing

Liquid phase reactions differ significantly from gas phase reactions because the diffusivity of liquids is much lower than their viscosity. Therefore, reactions can be strongly influenced by scalar gradients. The most important difference present in many liquid phase reactions is the phenomenon of micro-mixing, or mixing at the molecular scale. 

In the reaction shown in Fig. 10, whether the reaction will preferentially form product S or product R depends on the physical configuration of the reactor. In this schematic, for example, the location at which species B is added—at the top of the vessel, or instead close to the impeller—can make a significant difference in results. 

FIG. 10. Physical configuration of reactors is critical to predicting the product of many liquid phase micro-mixing reactions.
FIG. 10. Physical configuration of reactors is critical to predicting the product of many liquid phase micro-mixing reactions.

The proprietary model allows accurate modeling and prediction of these micro-mixing effects by providing an eddy contact micro-mixing model, which gives higher accuracy compared to an eddy breakup model (used as standard for gas phase reactions). The eddy contact model calculates a reaction time scale based on the scalar dissipation rate, and uses this scale to calculate the reaction rate (Fig. 11).

FIG. 11. Comparison of predicted yield values obtained by eddy contact micro-mixing, and eddy breakup models with experimental measurements for liquid phase reaction.
FIG. 11. Comparison of predicted yield values obtained by eddy contact micro-mixing, and eddy breakup models with experimental measurements for liquid phase reaction.

Customized reaction models: Fermentation and biochemical reactions

Another kind of flow reaction consists of fermentation and biochemical reactions. In these reactions, sugar is converted to acid, and either gases or alcohol are produced through a complex reaction chemistry. Reaction rates differ, starting with a log phase followed by an exponential phase, then a deceleration phase and, finally, a stationary phase. Each phase has a defined reaction rate. All of these phases can be defined in the modeling system using the custom reaction definitions available through the user-defined functions or field functions (Fig. 12).

FIG. 12. The proprietary modeling system provides numerous ways and capabilities to model reactors.
FIG. 12. The proprietary modeling system provides numerous ways and capabilities to model reactors.

Takeaway

The capabilities available in the modeling system allow reaction engineers to look at the transport processes in various reactor types, including packed-bed reactors, fluidization, stirred reactors, bubble columns and membrane reactors, as well as the various high-temperature processes in gas phase reactions. 

Even in established processes, the underlying transport processes are crucial for reactor design, and they open up possibilities for improvement. For each of those transport processes, the modeling system provides ways and capabilities to model and discover better reactor designs, faster. HP

Notes

     a The STAR-CCM+ modeling system

     b DARS (Digital Analysis of Reacting Systems)

The Authors

Related Articles

From the Archive

Comments