Epilogue

Almost all the processes relevant to the manufacturing industry (chemical, petrochemical, fertilizer, metallurgical, power, cement and so on) involve flow of fluids in some way or the other. Innovative and competitive edge in any manufacturing industry rests on how well these flow processes are designed and operated. In view of the central role of reactors in chemical process industries, there is tremendous potential for applying new flow modeling tools to chemical reactor engineering. Reactor engineering requires expertise from different fields ranging from chemistry and catalysis to fluid mixing and transport phenomena. Reactor engineering has to marry chemistry and catalysis with reactor hardware to evolve the best possible way to carry out the process under consideration. It is obvious that reactor engineers need to use several modeling tools to achieve their objectives. Computational flow modeling or CFD is being increasingly used for reactor engineering practice and research. In recent symposiums on reactor engineering (ISCRE and GLS conferences), more than 50% of the papers mentioned CFD. I hope that this book conveys the potential of computational flow modeling for reactor engineering applications and facilitates realization of this potential.

I have made an attempt to provide sufficient information to understand and to define the specific role of computational flow modeling in reactor engineering applications. Discussions on the main features of reactor engineering, computational flow modeling and their interrelationship will help to select appropriate models, and to apply these computational models to link reactor hardware to reactor performance. Mathematical modeling of flow processes (including turbulent flows, multiphase flows and reactive flows) and corresponding numerical methods to solve these model equations are discussed. Implementation of these mathematical models and corresponding numerical methods on computer, and key issues for evaluating available computational tools are also discussed. The overall methodology of achieving the objectives of reactor engineering via computational flow modeling is discussed with the help of practical examples. Aspects of the application of computational flow modeling to four major reactor types: stirred tank reactor, bubble column reactor, fluidized bed reactor and fixed (or trickle) bed reactor, and some other types of reactor are discussed. The selection of examples used in this book may appear somewhat biased since many of these are drawn from our own research and consulting experience. An attempt is, however, made to evolve general guidelines, which may be useful for solving practical reactor engineering problems.

At this juncture, it would be useful to re-examine the lessons learnt from our experience of the application of computational flow modeling to reactor engineering. From our experience, it is extremely important to correctly:

• identify and pose the problem;

• analyze key issues relevant to achieving the defined objectives; and

• select an appropriate modeling approach/tools which are consistent with the set objectives.

For any engineering discipline, the so-called Occam's razor always provides guidelines for selecting appropriate methods/tools. Occam's razor can be stated as, 'it is futile to do with more, what can be done with less'. There are many instances where simple, conventional models may provide elegant and adequate solutions. Even if complete solutions are not possible with simple, conventional models, conventional analysis and modeling is essential to understand the problem correctly and for appropriate formulation of the flow-modeling problem. If the time and space scale analysis indicates that a complete mixing assumption is more or less valid for the stirred reactor under consideration, it may not be necessary to use a computational flow model to simulate conversion obtainable in such a reactor. Conventional flow modeling and accumulated empirical knowledge about the equipment under consideration must be used to get whatever useful information that can be obtained, before undertaking rigorous CFD modeling.

It is, however, important to emphasize here the maxim that says 'one should always try to make things as simple as possible (following the Occam's razor) but not simpler'. It may be necessary to match the complexity of the problem with complexity of the analyzing tool. One may try to find simple solutions to complex problems, which may not be right all the time! Distinguishing the 'simple' (keeping the essential aspects intact and ignoring non-essential aspects) and 'simpler' (ignoring some of the crucial issues along with the non-essential issues; akin to throwing the baby away with the bath water) formulations is a very important step towards finding useful solutions to practical problems. One should have the expertise and skill to select an appropriate level of complexity of the analyzing tools to suit the set objectives. This is one of the most important prerequisites for successful execution of reactor-engineering projects. In many reactor-engineering applications, detailed fluid dynamic models may become necessary and may substantially contribute towards performance enhancement. Thus, playing with problem definition (evaluating symptoms/set objectives, identifying and separating essential and non-essential issues, reframing problem objectives in the light of such analysis) and selecting an appropriate modeling approach, is one of the most important tasks of a reactor engineer. More often than not computational flow modeling projects are likely to overrun the budget (of time and other resources) due to inadequate attention being paid to this initial step of the overall project. Inadequate attention to this step may even lead to failure in achieving the set objectives. Some of the examples discussed in Chapters 9 to 13 are useful to understand the importance of selecting an appropriate modeling approach using different sets of models.

Another important lesson is that it is beneficial and more efficient to develop mathematical and computational models in several stages, rather than directly working with and developing a one-stage comprehensive model. For example, even if the objective is to simulate non-isothermal reactive flows, it is always useful to undertake a stage-wise development and validation of computational models. Such stages could be: (1) simulate laminar flow; (2) examine these results and select an appropriate turbulence model; carry out simulation of turbulent flow; (3) evaluate isothermal turbulent simulations, verify existence of key flow features (go back to step 2 if results are not satisfactory), try to validate quantitatively wherever possible; (4) include non-isothermal effects (without reactions); (5) include reactive mixing models in the non-isothermal turbulent models; and (6) validate and apply. Development efforts and simulated results from each stage enhance understanding of the flow phenomena. For each stage of model development, quantitative evaluation of limiting solutions (may be with drastic simplifications) is often useful to enhance confidence in the developed computational model. The simulated results also provide information about the relative importance of different processes, which helps to make a judicious choice between 'simple' and 'simpler' representations. Such a multistage development process also greatly reduces numerical problems, as the results from each stage serve as a convenient starting point for the next stage.

Apart from appropriate model formulation, it is also essential to understand the influence of numerical issues (grid spacing, time step, degree of convergence and so on) on simulation results before one can use the results obtained from a computational flow model for engineering applications. One must resist the temptation to use physically realistic simulated results without quantitatively assessing grid dependence. This is true even when the objective is just to understand key flow features qualitatively. It is possible, in some cases, that a different grid spacing may show different key flow features. Sometimes it is observed that computational results obtained with a specific grid show good agreement with the available data. This acceptable agreement often encourages immediate application of the computational model to the problem under consideration. It must be remembered that no matter how good the agreement one finds between available data and results simulated on a specific grid, if the solution is not grid independent, the agreement is probably an artefact of the specific grid size. It is, therefore, necessary to make an attempt to obtain grid independent results before they are used for reactor engineering applications.

In many situations, however, it may not be possible to obtain grid independent solutions for flow in complex industrial equipment (due to the constraints on available time and computational resources). In such cases, the reactor engineer may still use these simulations for practical applications, provided some of the following precautionary steps are carried out:

• quantitative evaluation of special cases/limiting solutions;

• qualitative verification of key flow features;

• assessing dependence on grid spacing by extrapolating key results to zero grid spacing (results may not be grid independent even for the finest grid used in these simulations).

Another related issue is about finding some flow features in the simulation results, which are not really expected from the underlying mathematical model. For example, the Eulerian-Eulerian simulations of a bubble column containing finite vapor space above the liquid pool often show a gas-liquid interface when contours of gas volume fraction are plotted. Usually, in such simulations, the two-fluid model and the interphase drag coefficients suitable for regions in which gas is dispersed in a liquid are used. These models are not suitable to simulate the region in which the gas phase is continuous. Because of this and since by its very nature, Eulerian-Eulerian simulations are not suitable to simulate a gas-liquid interface, results near the interface are highly inaccurate. It is necessary to consider these aspects before one proceeds to use the simulated information about the pseudo-gas-liquid interface for engineering decision making. It is very important to understand the capabilities and limitations of the underlying mathematical model. Considerations of limiting solutions are often useful for this purpose. The necessity of verification and validation of computational flow models is repeatedly emphasized throughout this book. In many industrial applications, data required for adequate validation is not available and a reactor engineer has to rely only on 'indirect validation' of some gross quantities. Reactor engineers must, therefore, develop their skills in assessing the quality of simulations in the absence of direct validation. Such skills may be acquired by studying known case studies and through hands-on experience of applying computational flow modeling to reactor engineering. If some of these issues are properly taken care of, computational flow modeling (CFM) may be used to provide invaluable information for reactor engineering applications. CFM may be the only way to realize the 'wish list' of a reactor engineer in practice. CFM may also be used to study aspects of flow which are not amenable to experiments (due to high temperature/pressure or corrosive conditions). Detailed flow modeling of industrial processes offers new possibilities for performance enhancement and innovation in the design of industrial reactors. Because of these unique capabilities of computational flow modeling, CFM will have tremendous impact on current as well as future reactor engineering practices.

With the emergence of cheap, high speed computing platforms and the availability of commercial CFD codes and support, flow modeling needs to be harnessed to devise the best possible reactor hardware. Some comments on future trends and needs may be appropriate at this juncture. Each advance in the CFD community's capability to perform a particular class of computations, has led to a corresponding increase in the engineer's expectations. These expectations can be translated to define research and developmental requirements. These requirements may be classified into two categories: computational and physical. The most important areas of computational character, in which further work is needed, are:

• cheaper ways of conducting fine-grid computations;

• minimizing numerical diffusion without jeopardizing robustness;

• preserving the order and flexibility in CFD codes as the complexity of their physical content increases.

Further research on problems of physical character, needs to focus primarily on the development of better turbulence models, better multiphase flow models and better reactive flow models. Recent advances in applying renormalization group (RNG) theory to formulate turbulence models appear to be promising. However, much work is needed to understand the intricacies of turbulent, multiphase flows. Recent advances in direct numerical simulations and database of DNS results may provide useful guidelines for further development of physical models. Detailed and comprehensive experimental programs are needed to verify the applicability of the existing and new models. Quality experimental data collected through such programs and close communication between experimentalists and those developing numerical models and methods are essential to advance the applicability of multiphase CFD in practice. Experience of applying computational flow modeling to any practical reactor engineering problem may suggest several areas in which further studies are needed for development of better physical models. In view of the wide range of reactor engineering applications, it is practically impossible to make a comprehensive list of all such suggestions. Ranade (1995) listed some areas which need further research for better simulations of dispersed gas-liquid flows (in stirred and bubble column reactors). His list of important issues on which further work is needed may be generalized for any dispersed multiphase flow:

• interphase momentum exchange terms/influence of dispersed phase volume fraction;

• motion of dispersed phase particles near the wall, wall boundary conditions;

• role of particle wakes on dispersion/inter-phase momentum exchange;

• turbulent transport of dispersed phase particles/dependence on particle size;

• turbulence modification by dispersed phase particles;

• particle-particle interactions (collisions, coalescence/agglomeration, breakup);

• interphase heat and mass transfer/phase change models.

The list is merely suggestive. Complexity of reactive flows may greatly expand the list of issues on which further research is required. Another area which deserves mention here is modeling of inherently unsteady flows. Most flows in engineering equipment are unsteady (gas-liquid flow in a bubble column reactor, gas-solid flow in a riser reactor and so on). However, for most engineering purposes, all the details of these unsteady flows are not required to be known. Further work is necessary to evolve adequate representation of such flows within the CFD framework without resorting to full, unsteady simulations. This development is especially necessary to simulate inherently unsteady flows in large industrial reactors where full, unsteady simulations may require unaffordable resources (and therefore, may not be cost effective). Different reactor types and different classes of multiphase flows will have different research requirements based on current and future applications under consideration.

Apart from such research requirements to enhance the capabilities of CFD tools, more and more studies on the application of available tools to simulate engineering equipment are necessary. Accepting the limitations of knowledge of underlying physics and invoking model calibration whenever necessary, is mandatory to expand the application horizons of computational flow modeling. Such experience will provide invaluable information and may guide future developments. Besides this, such applications will significantly enhance the current reactor engineering practice.

Chemical and process engineers today routinely use process simulation tools to design and to optimize overall plant operations. Computational flow modeling tools are also expected to be used as widely as process simulators in the near future. Currently efforts are under way to integrate CFD and process simulation tools (for example, CFX and Hyprotech or FLUENT and Aspen). Several attempts are also being made to couple CFD tools with reactor simulation tools (with automated information flow between CFD and other simulation tools). Advances in software technology and enhanced computing resources allow efficient coupling of CFD codes with physical and chemical property databases (or predictive tools) on one hand and process or reactor simulation tools on the other hand. Such seamlessly integrated tools will allow evaluation of changes in the reactor hardware on overall process performance in the near future. Such capabilities will significantly influence the reactor-engineering practice of tomorrow.

Adequate attention to the key issues mentioned in this book and creative use of computational flow modeling will make significant contributions to enhancing chemical reactor engineering. The field of computational flow modeling for reactor engineering is evolving and being continuously updated. New advances may be assimilated using the framework discussed in this book. I hope that this book will stimulate applications of computational flow modeling to chemical reactor engineering.

0 0

Post a comment