Chemical Industry

of the Future

Technology Roadmap
for

Computational
Fluid Dynamics

January 1999
 Table of Contents
Page

1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Challenges on the Horizon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Industry’s Response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
The Role of Computational Fluid Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 What is CFD? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

3 Current Situation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Computing Power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Computational Fluid Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Chemical Process Design and Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

4 Trends and Drivers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

5 Performance Targets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

6 Technology Barriers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

7 Research Needs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
CFD Goals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
The Road to Follow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

Table of Contents i
ii Computational Fluid Dynamics Technology Roadmap
 1 Overview
Challenges on the Horizon

As the 21st century approaches, the chemical industry faces considerable economic, environmental and
societal challenges [VISION 1996]. Major forces for change include: increased globalization of
markets; societal demand for improved environmental performance; the need for increased profitability
and capital productivity; higher customer expectations; and changing work force requirements.
Technology research, development and deployment will be vital to meeting these challenges and seizing
future opportunities for market growth.

Globalization will promote the free movement of people, technology, capital, information and products
across international boundaries. While this will create many new market opportunities, it will also
require the development of advanced manufacturing technologies that allow the industry to be globally
competitive. Improving and sustaining environmental performance will provide opportunities to serve
the global population with higher performance and higher quality products. It will also require
manufacturing processes that are resource efficient, cost effective, and environmentally sound.

With an ever-increasing emphasis on the creation of products that generate revenues and satisfy
stockholder expectations, the industry will need to find ways to increase profitability and capital
productivity. Strategically-driven investments in R&D and new technologies can drive the industry
towards a higher level of financial performance. An expedient response to customer expectations will
be increasingly important in capturing new market opportunities and ensuring financial success for
chemical producers as well as customers. To meet this challenge, technology innovation will be needed
throughout all phases of R&D, production, and distribution. Finally, to keep pace with rapid
technological change and the increased complexity of chemical production technology, a more highly
skilled work force will be essential.

Industry’s Response

In response to these challenges and to a request from the White House Office of Science and
Technology, the chemical industry initiated a study in 1994 to examine the factors affecting the
competitiveness of the industry and the R&D essential to meeting their future needs. The results of this
study are contained in Technology Vision 2020: The U.S. Chemical Industry, a collaborative effort
representing inputs by more than 200 technical and business leaders from the

Overview 1
 Exhibit 1-1. Selected R&D Areas Identified By Technology Vision 2020

chemical community [VISION 1996]. This visionary document identifies major goals for the industry
that include, among others, improving operations and the efficient use of resources (e.g., raw materials,
energy).

To meet future industry goals, Technology Vision 2020 advocates the support of R&D in a number of
areas, including new chemical science and engineering technologies that will promote more cost-efficient
and higher performance products and processes (see Exhibit 1-1). An important component of this
R&D is the development of enabling technologies that improve the application of fundamental chemical
sciences throughout the industry’s process environment.

The Role of Computational Fluid Dynamics

Technologies to assist industrial computations may be considered “enabling” in that they are used in
nearly every aspect of chemical research, development, design and manufacture. The focus of this
report is on computational fluid dynamics (CFD), a highly sophisticated integration of applied
computer science, physics, chemistry, and engineering science. It is a subset of computational
technologies that is critical to meeting the industry’s future challenges. A focus on CFD R&D was
identified in Technology Vision 2020 as a high-priority for meeting the industry’s future goals.

2 Computational Fluid Dynamics Technology Roadmap

In 1995, the chemical industry began to define its needs in CFD through a workshop on the computer
simulation of reactive multiphase flow [LANL 1995]. Attendees included participants from twelve
companies representing the petroleum, chemical, environmental and consumer products industries,
along with representatives from the U.S. Department of Energy and Los Alamos National Laboratory.
The dialog at this meeting suggested that reactive multiphase flow simulation, a component of CFD, was
critical to optimizing a number of chemical processes, and also an excellent candidate for partnership
activities involving industry, government, and academia. A white paper was also presented at this
workshop to foster discussion on the formation of a consortia to improve the state-of-the-art in
computational technologies [DOW 1996]. This white paper outlined the problems with existing
computational packages as well as some of the challenges for the chemical process industry in modeling
specific chemical systems. The results of this workshop and the white paper later provided input for the
discussion of computational technologies in Technology Vision 2020.

Following the publication of Technology Vision 2020 in 1996, another workshop was held to build an
industry-wide consensus on the essential research and technology development pathways needed to
develop advanced CFD capabilities. The National Workshop on Computational Fluid Dynamics and
Multiphase Flow Modeling brought together members of the chemical processing, petroleum refining,
aluminum, and computer hardware industries [NATL 1997]. Participants also included representatives
from government agencies (U.S. Department of Energy, National Science Foundation), national
laboratories, and academia. The first half of the workshop concentrated on the exchange of information
between members of the chemical and computer hardware industries regarding needs, future industry
direction, and potential developments. The second half included brainstorming sessions to identify
specific research needed over the next ten years and to explore the foundation for a cohesive CFD
R&D roadmap for the chemical process industries.

A final workshop held in June 1997 completed the roadmap process by further linking technology
research needs with practical fluid flow problems, performance targets, and time frames where research
could be expected to yield tangible benefits. Together the results of these workshops provide the
foundation for this CFD roadmap. It is one of many such efforts to provide a link between the broad-
based goals defined in Technology Vision 2020 and the research portfolio that will be pursued through
cooperative R&D partnerships. The research priorities outlined in this roadmap will be used as the
basis for making new research investments by government and industry. It is a dynamic document, and
will be reevaluated periodically to incorporate new market and technical information and to ensure that
the research priorities remain relevant to the needs of both the chemical industry and its customers.

Overview 3
4 Computational Fluid Dynamics Technology Roadmap
 2 What is CFD?
Computational fluid dynamics, or CFD, provides
a quantitative description of flowing fluids in Characteristics of Fluid Flows
relation to the fluid’s surroundings. Fluid flow situations comprise liquids, gases, solids or a
Mathematical equations describe the physical combination of these phases. Fluid flows can be reactive,
and chemical phenomena that occur as fluids turbulent or laminar; steady or unsteady; Newtonian or non-
Newtonian.
flow. In computational applications the
mathematics are translated into computer Reactive Flow - fluid flow where chemical reactions
provide important contributions to heat and mass transfer.
language so that the thousands of calculations Combustion situations and many chemical manufacturing
required to describe the fluid flow can be processes fall in this category.
performed. Multiphase Flow - fluid flow of more than one phase (e.g.
gas-solid, gas-solid-liquid) the properties of which cannot be
Fluid flows found in industrial processes range described by the flow of a single phase alone.

from simple to highly complex. A simple flow Laminar Flow - flow in smooth, distinct streamlines, such
situation is liquid water flowing through a pipe. as water flow in a pipe at low flow rates. Steady laminar
flow of a single phase is one of the few cases with a
For this example it is straightforward to predict relatively straightforward mathematical solution.
the properties of the flow (such as the volumetric
Turbulent Flow - Characteristic of higher flow rates (or
flow rate) from characteristics of the flow higher Reynolds numbers) at which, instead of moving in
situation (length and material of the pipe, pump smooth, predictable paths, the fluid motion is chaotic.
pressures, flow temperature). For more Turbulent flow is characterized statistically. The time-
averaged flow properties can be approximated adequately
complex situations, such as gases, liquids, and for some single-phase cases. Expensive experimental
solids contacting together in a chemical reactor, methods are currently relied on to analyze and predict
turbulent flow, particularly for multiphase flow.
it is not possible to determine the properties of
the flow (such as concentrations of components, Reynolds Number - a dimensionless number, commonly
flow rates) by hand calculation. The properties used in characterizing fluid flow, that is proportional to the
inertial force divided by viscous force. Typically it is
must be calculated from “point to point” in the calculated as (length) x (mass flow rate)/(viscosity). At low
reactor, meaning that the mathematical relations Reynolds number the flow is laminar. At sufficiently high
Reynolds number it becomes turbulent.
describing the fluid flow must be solved
numerically, with the aid of a computer. The Steady Flow - The spatially averaged properties of which
are constant in time.
various aspects involved in these calculations are
inherent to “CFD.” Newtonian Flow - A flow for which stress is directly
proportional to strain rate. Flows of polymer materials
(molten plastics, for example) are usually non-Newtonian.
In practice non-Newtonian flows require additional energy

What is CFD? 5
More precisely, CFD technical areas include constitutive relations, numerical methods, and
experimental validation.

Constitutive relations describe the limitations of the fluids in relation to their surroundings, and the
interrelationships of the flow variables. Many constituitive relations are empirical, meaning that they are
derived from observations; other constitutive relations are based on theory or fundamental physical
principles. It is most common for fluid flow variables (pressure, velocity, temperature) to be related to
each other from point to point, infinitesimally close (i.e., in differential form). The corresponding
relations are called “differential equations.” CFD is applied to find numbers that represent these
variables in specified fluid flow situations.

Numerical methods are the tools used to solve the equations described above. The use of numerical
methods depends on the particular form of the equations, the desired accuracy of the solution, the type
of computer that is used in the calculations, and in many cases, the knowledge and skill of the
investigator doing the calculations. Advances in computer technologies are usually accompanied by
advances in numerical methods, with the goal of enhanced speed and amount of data that can be
processed.

Modeling and simulations mean nothing, of course, independent of the reality they are supposed to
represent. The accuracy of the simulation must be checked, or validated, against data obtained from
real operating systems. This experimental validation is critical to ensuring that the computational
tools are meaningful representations of the applications for which they are intended. The role of
experiment goes beyond merely confirming the computed results. Experiments must be designed to
challenge the model and push it to the limits at which it fails.

By predicting a system's performance in various

areas, CFD can potentially be used to improve Applications for
Computational Fluid Dynamics
the efficiency of existing operating systems as well in Chemical Processing
as the design of new systems. It can help to
shorten product and process development cycles, C Mixers
C Chemical Reactors
optimize processes to improve energy efficiency C Packed Beds
and environmental performance, and solve C Crystallizing/Dissolving Equipment
problems as they arise in plant operations. There C Pneumatic Conveyors & Classifiers
C Flows in Pipes
are many potential applications of CFD in C Sprays
chemical processes where predicting the C Biological Systems
characteristics of fluid flow are important (see
inset). CFD has also been applied to determine
optimum control strategies, optimize equipment in the automotive and electric power industries, and has
been especially successful in the aerospace industry. In some cases, the design to production time for
airplanes has been reduced by half or more over the past decade with the development and application
of CFD. A concerted effort by industry, in partnership with government and academia is needed to
make similar advances in CFD possible for the chemical and other low-temperature process industries.

6 Computational Fluid Dynamics Technology Roadmap

 3 Current Situation
Computing Power

Computational tools available today Computing Hardware

are greatly improved from those
developed just a few years ago, and Supercomputer - A system that meets the computer industry’s
evolving standards at any particular time. A supercomputer
are easier to use. Development of
must have not only very fast processors but also
parallel machines and parallel commensurate memory, internal communications, and data
algorithms designed for these machines storage technology. For example, in 1998 the U.S. Department
of Energy acquired a supercomputer that can perform 3.9
has allowed the solution of problems
trillion (million million) calculations per second and is
that were previously impossible to equipped with over 2.6 trillion bytes of memory. Such a
solve. Advances have also been made machine exceeds typical desktop computers by ten thousand
fold in both speed and memory.
that facilitate the handling of the
immense amount of data generated by Parallel Architecture - At least four (and usually more)
a complex simulation. Distributed and processors linked together, and controlled by hardware for
efficient use of the maximum number of processors at any
shared-memory computer
time. Processors may cooperate or work independently to
architectures have both emerged as solve a problem.
the way to increase computational
Distributed Architecture - A technique to use the capabilities
power. Visualization of complex
of more than one computer (processors and memory)
three-dimensional flows is now a simultaneously. The technique is usually applied with
practical tool for obtaining useful software controls.
information from simulations.
Vector Processing - Calculations are performed in linear,
assembly line fashion. Supercomputers based on Vector
Architecture include special processor hardware to stage and
rapidly execute the repetitive calculations that are typical of
Computational Fluid matrix arithmetic, as found in most scientific computer
Dynamics programs.

Considerable effort has gone into the

development of single-phase fluid dynamic modeling tools for applications in the aerospace, automotive,
and electric power industries. CFD for single-phase non-reactive fluid flows is relatively well-
developed, and can be used to model relatively complex flow fields. Reactive single-phase flows can
potentially be modeled with existing tools (such as simple combustion systems), provided the
computations required are not too large or complex.

Current Situation 7
 Currently available numerical methods for single-
Single Phase Numerical Methods
phase flow include Direct Numerical Simulation
Direct Numerical Simulation (DNS) - provides a (DNS), Large Eddy Simulation (LES), and
numerical solution (description) for all the scales Reynolds Averaged (RA) models. These
(size) of flow without simplifying or approximating
the important physical properties of the flow. models, as well as new approaches coupling full
reaction kinetics with turbulence, suggest that
Large Eddy Simulation (LES) - provides a modeling turbulent reactive multiphase flows is
numerical solution for large eddies (whirlpools) in
the flow while modeling (simplifying) the effects of possible on today’s computers. However,
very small eddies. development of hardware (parallel architecture),
software (advanced numerics for parallel
Reynolds Averaged Navier Stokes (RANS) Model
-simulates turbulent fluid flow more rapidly than machines, object-oriented modular
DNS or LES but with limited accuracy. programming), and theory (submodel
development) are required.

Commercial CFD vendors have not given emphasis to

Current Applications of
problems involving chemical reactions and turbulent Multi Phase CFD in Chemical
reacting flows in particular, primarily because their Processing
products have evolved for use in the aerospace,
automotive, and power industries. However, a C Gas Scrubbing
C Thermal Oxidation/Waste Treatment
number of existing single-phase CFD packages have C Packed Bed Reactions
been adapted for use in the chemical industry to assist C Crystallization
in solving practical problems that arise with chemical C Polymer Extrusion
processing equipment. C Mixing and Blending
C Rotary Kiln Incineration
C Drying
Limitations of Current CFD Packages C Membrane Separation
Commercially available CFD tools cannot or do not: C Dust Separation
• Pyrolysis (Cracking)
C use available computational horsepower (e.g.,
parallel computing), as well as they should,
C use leading edge numerical methods (e.g., adaptive gridding, solution-driven grid refinement) or
theoretical methods,
C include available submodels of key subprocesses (e.g.,crystal nucleation and growth), or submodels
are not coupled properly (i.e., turbulence and chemistry),
C have a common chemical engineering infrastructure to allow linking to programs common in the
chemical processing industries (e.g., physical and chemical properties data base).

Chemical Process Design and Operation

The efficient design and operation of multiphase flow systems is currently limited by a number of
factors, as shown in Exhibit 3-1. Some of these are due at least in part to the lack of accurate modeling
tools for multiphase flow regimes; others result from problems inherent to specific chemical processes.

8 Computational Fluid Dynamics Technology Roadmap

 Exhibit 3-1. Problems Limiting the Efficient
Design and Operation of Multiphase Flow Systems
Design Operation and Control Process-Related Issues

Current designs are artificially Lack of means for controlling Inefficient pneumatic handling of
constrained product attributes solids (feeds and products)
resulting from poor design and
Current designs are based on Poor utilization of existing reliability
precedence and empirical methods process vessels
Problems associated with chemical
Data at the macroscopic rather than Excessive down-times due to containment and safety
microscopic level is used in current corrosion and erosion
designs General process inefficiencies
Limited ability to optimize leading to unnecessary energy
Not all design alternatives are existing reactors and consumption, production of waste,
explored—limited possibility thinking is separation units, leading to and emission of pollutants
the rule low yields and poor
performance Limited availability of designs and
Steady state is often used to explain computational tools that target
transient and segregated flows High cost of experimenting in specific production processes or
full-scale production facilities plants
Limited ability to do real reactor design
Poor visualization of process Mass transfer-controlled operations
Current design simulations are based on phenomenon
idealized conditions (often quite Multiphase flow in channels
different from actual)
Viscous and non-Newtonian mass
Current codes provide inaccurate transfer operations
predictions when extended to other flow
conditions Safety pressure relief multiphase
discharge designs

Design
Many current designs are based on precedence and guided by entirely empirical methods (those based
on experiment or experience). This is the direct result of the lack of, or inaccuracies associated with,
computational methods available for use during the design cycle. Current codes designed for very
specific flow conditions often provide inaccurate predictions when applied to other flow conditions. In
many cases existing design simulations are based on idealized conditions that are quite different from
actual operating conditions. For example, available codes often use steady state simulations (steady
over time) to describe flows that are transient (changing with time). In general, the lack of accurate
computational packages for multiphase systems results in artificially constrained designs and a limited
ability to explore alternative designs.

Operation and Control

While CFD packages for single-phase systems have been used for optimizing operation and control of
existing processes in the chemical industry, their use is limited for systems containing reactive and/or
multiphase flows. Without the use of simulation tools, the plant engineer’s ability to visualize process
phenomena is limited and optimizing operations is made more difficult and time-consuming. Often the
only way to identify and solve the problem is to make experimental measurements—a costly
proposition in full-scale chemical production. Without an effective way to optimize operations, the

Current Situation 9
result is often poor control of processes and product attributes, poor utilization of equipment, excessive
equipment down-time, low yields, and generally low performance.

Process-Specific Issues
Several areas of chemical processing are particularly difficult from a design and operational aspect.
These include pneumatic handling of solids, chemical containment and safety, and general process
inefficiencies. Most of these could be addressed in part by effective CFD packages that model
turbulent reacting multiphase flows.

10 Computational Fluid Dynamics Technology Roadmap

 4 Trends and Drivers
Computer Technology
New computing technology is entering the marketplace almost daily. The speed of high performance
computing platforms has increased exponentially, and as of 1998 had reached 3.9 TFLOPS in some
models. Dramatic advances have been made in highly parallel processing and parallel numerical
algorithms and other high performance processors. The rapid advances in computing technology are
not expected to slow, and will lend further support to the development and effective use of
computational tools like CFD.

Energy Price and Supply

The availability and cost of fossil energy for use as fuel and feedstock is of vital importance to the
chemical processing industries. For some products, energy for heat, power, and feedstocks can
account for up to 85 percent of total production costs (overall, chemical industry energy costs are
about 8 percent of the value of shipments) [DOC 1994, CMA 1996]. Feedstock availability is also a
primary concern for many chemical producers—nearly 50 percent of energy consumed is in the form of
petroleum-based feedstocks. The industry is highly susceptible to volatility in energy feedstock price
and supply, a fact made evident during the oil embargo of 1973. Although energy prices are currently
low, history shows they can be subject to rapid change with devastating impacts. Advances in
technologies like CFD, which have the potential to improve process energy efficiency, will continue to
be of importance to the industry.

Government Regulations and Public Policy

Chemical production, use, storage, transportation, and disposal are heavily regulated by Federal and
state laws and regulations. Compliance with these statutes (e.g., Clean Air Act and its Amendments,
Toxic Substances Control Act, Resource Conservation and Recovery Act, and others) represent
serious challenges to the chemical industry in terms of capital and operating expenses. In 1994 the
industry spent about $4.6 billion on pollution abatement and control, nearly double the costs incurred in
1984 [CMA 1996]. New environmental programs are continually being proposed at the Federal, state
and other levels of government, and will likely further increase the difficulty and cost of compliance.
Consequently, cost-effective technologies that improve the environmental performance of processes will
be increasingly important.

Trends and Drivers 11

12 Computational Fluid Dynamics Technology Roadmap
 5 Performance Targets
Improvements in computational fluid dynamics can contribute directly to the goals stated in the chemical
industry vision. Specifically, the application of computational technologies could promote:

C Shortened product-process development cycles;

C Optimization and control of existing processes to improve yield and energy efficiency;
C Efficient design of new products and processes; and
C Improvements in health, safety, and environment.

With these broad goals as a base, specific

performance targets (both quantitative and
qualitative) have been identified for CFD, as
Exhibit 5-1. Performance Targets for
shown in Exhibit 5-1. These targets
Computational Fluid Dynamics
illustrate how improvements in
computational fluid dynamics can have a Quantitative Targets
wide-reaching impact throughout the entire
C Shorten lead times (from research to final plant
chemical industry. The basic tools of design) to 3-5 years.
numerical modeling and simulation of fluid C Reduce plant down times to 1%.
flows are broadly applicable and can lead C Reduce separation energy and improve separation
efficiency by 20%.
to improvements in nearly all unit operations
and processes. Qualitative Targets

The primary advantage of CFD is the C Increase reliability of design (reduce risk).
C Reduce/eliminate design errors.
potential to shorten lead times for process C Promote innovation.
development (and plant designs) as well as C Reduce fuel consumption per unit of product.
new product development. Using C Improve heat transfer (waste-heat recovery).
C Optimize processes to increase yield and aid
conventional development methods without incremental expansion.
CFD, lead times can be as long as 7 to 10 C Create a library of computational tools with both
years. The impact of CFD on the process single-phase and multiphase flow capability.
and product development cycle is greatest
in the earliest phases (i.e., experimental
optimization and scale-up). CFD facilitates the design process by increasing the reliability of the design,
reducing or eliminating design errors, and allowing developers to visualize the results of a process

Performance Targets 13
design or innovation. It promotes innovation by making the design process shorter, less risky, and
easier to accomplish.

Using CFD to resolve problems with existing processes can reduce equipment failures and minimize
poor operational performance, decreasing process shut-down time. When used to optimize plant and
equipment operation, CFD can increase yields, providing a mechanism for incremental expansion.

Applying CFD to the optimization of energy-intensive processes (e.g., separations) can improve
separation efficiency, reduce energy requirements, and improve environmental performance. CFD can
also provide a way to improve heat transfer (e.g., recovery of waste heat) and better integrate heat
transfer needs within the chemical plant.

The creation of a library of computational tools that are relatively easy for a non-specialist research
plant engineer to use is a key target. Such a library would greatly promote the use of CFD tools in the
chemical plant to solve real problems and optimize process and product designs. These tools would be
designed with a “plug-and-play” code to allow easy insertion of new submodels and new solution
algorithms on various computing platforms, including high performance desktop workstations; large, fast
vector-processor machines; and highly parallel processors. Proposed codes would be built
independently of current computer architecture, since architecture will change and code must change
along with it to retain usability. A CFD library with the above features will be more widely acceptable
and usable throughout the chemical community, and will help to promote environmentally safe, efficient
chemical production processes.

14 Computational Fluid Dynamics Technology Roadmap

 6 Technology Barriers
There are a number of barriers that inhibit the development and use of CFD. These are shown in
Exhibit 6-1, organized by topic.

Practical Usability
The practical use of current CFD packages by engineers in chemical plants is limited primarily by the
excessive time currently required for the engineer to set up the calculation. Furthermore, current CFD
tools do not always provide results that are easily used or interpreted by the plant engineer. In many
cases the plant engineer initially wants to use CFD to solve a problem arising with existing equipment or
processes. He/she then finds that the solutions provided by CFD are not conveniently accessible.
Another limit is the lack of CFD tools that can model simple versus complex situations. For example,
some problems or design issues could be resolved by easily accessible submodels of a larger CFD
package rather than setting up an entire simulation. These types of modular packages are currently not
available. In general, many personnel involved in plant operations have a poor perception of modeling
tools as a means of solving practical problems, primarily because of a lack of good information
dissemination concerning these tools, their usability, and the validity of results.

Data/Basic Sciences
Inadequate understanding of the basic chemistry and physics related to fluid flows is a limiting factor to
the advancement of CFD, particularly for reactive and/or turbulent multiphase flows. Lack of
knowledge of residence times, interfacial phenomenon, distribution of materials in the flow, interphase
coupling, turbulence, and general flow characteristics inhibits the development of relations needed to
model fluid flows in more complex situations. Experimental data and physical properties data that could
provide empirical relations for multiphase flows are also lacking.

Computational Limits
While computing power has increased dramatically, many computational tools have not been upgraded
to take advantage of new capabilities. Current computational methods place limits on the size of the
problem that can be simulated—limiting the development of CFD for many important problems (e.g.,
coupling of turbulence and interfacial chemistry, modeling complex polymer rheology). A more generic
limitation applicable to many higher level simulations is the inability to integrate multi-disciplinary
problems in a single computational tool. For multiphase flow simulations, the lack of general relations
for the coupling of turbulence, chemistry, phase

Technology Barriers 15
 Exhibit 6-1. Technology Barriers to the
Development and Use of Computational Fluid Dynamics

Commercial Computer
Data/Basic Computational Software Hardware/
Practical Usability Sciences Limits Development Architecture

Excessive time Lack of Limits on the size of Chemical Inadequate speed

required for engineer data/knowledge of the problem that can manufacturers lack of computation
to set up the residence times, currently be the resources to
simulation interfacial area, simulated, such as effectively Lack of solution-
distribution of commercialize dependent grid
Inadequate usability materials, scaling - turbulence software refinement
of computational - interfacial (relates to
tools/results by plant Insufficient chemistry Chemical computing
support engineers knowledge of fluid - segregation manufacturers lack efficiency)
flow characteristics - complex polymer the resources to
Solutions are not rheology support software Inadequate data
conveniently Lack of appropriate - complex processes documentation, help management
accessible coupling of - time dependence files, on-line help, devices for the
chemistry to multi- and training enormous data
Lack of different phase physics Inability to integrate sets generated
levels of modeling multi-disciplinary In-house codes
and simulation Poor problems in a single developed by
(simple vs. complex) understanding of computational tool industry are
underlying physics typically usable
Poor transfer of of multi-phase Lack of closures for only by specialists,
information about conditions (e.g., turbulence, not general
computational tools turbulence, chemistry, phase
and their usefulness interfacial interfaces, and Few chemical
phenomenon) boundary conditions companies are
Poor perception and motivated to refine
acceptance of Lack of Inadequate models for general
modeling tools and experimental data mathematical use
results by plant for multi-phase methods
personnel flows Developing
Fine-scale physical commercial software
Lack of support to Lack of physical features of multi- for small markets is
link current CFD properties data for phase flow are not cost-prohibitive for
tools to chemical multi-phase flows captured in any a single vendor
engineering canonical set of
infrastructure equations

interfaces and boundary conditions is a limiting factor in CFD development. In general, mathematical
methods are still inadequate for complex CFD.

Commercial Software Development

The manufacturing industries, particularly the chemical industry, are avid users of sophisticated scientific
and engineering software, but seldom develop it for commercial use. Manufacturing companies do not
have the resources needed to commercialize software, which includes support for auxiliary functions
(e.g., documentation, help files, on-line help support, training, debugging, and upgrading) to meet
customer needs. Commercial software companies find it difficult and expensive to develop, support,

16 Computational Fluid Dynamics Technology Roadmap

and sell specialized programs for relatively small markets. While the market for CFD in the chemical
processing industries is potentially very large, current CFD packages are not “usable” enough to
capture that market. A number of manufacturers have developed CFD codes for in-house use, but
many are only usable by the researchers who developed them (specialists), rather than by the broader
chemical engineering community (generalists). These in-house codes often become orphans and fall
into disuse because they are not updated to keep up with newer science, better concepts in software
architecture, or improved paradigms for user interfaces.

Computer Hardware/Architecture
In spite of the many advances in computer hardware and architecture, computer hardware still places
limits on highly complex computational tools like CFD. Current computer systems are still not quite fast
enough to effectively perform the many thousands of calculations required for a multiphase flow
simulation. The promise of parallel computing has been limited in general by software development and
code portability issues related to the lack of a parallel computer architecture standard. Computing
efficiency is also currently limited by the lack of solution-dependent grid refinement in commercial CFD
programs. In addition, while new storage mechanisms are emerging continually, none are entirely
adequate for managing the enormous sets of data generated by reactive, multiphase CFD solutions.

Technology Barriers 17
18 Computational Fluid Dynamics Technology Roadmap
 7 Research Needs
The primary goal of research in CFD is to extend current CFD capabilities to accurately model
multiphase flows, including reactive and/or turbulent flows. Some of the physical models and advanced
numerics needed to address these flows exist, but commercial CFD vendors have been slow to
incorporate them into readily available and fully supported CFD tools that are compatible with the
needs of the chemical industry. The following areas are of particular importance to chemical
processing:

C Turbulent/Reacting Flow - simulation of Three-Phase Flow Problems

turbulent flows with chemical reactions, including
gas phase, liquid phase, and multiphase reacting C Mixing
C Packed Beds
flows. For example, prediction of turbulence C Crystallizing/Dissolving
and its effect on fast reactions in a liquid phase C Sprays
stirred tank reactor remains an unsolved C Three-phase Flows in Pipes
C Biological Systems
problem. Applications include turbulent • Bubble Column Catalytic Reactors
combustion, chemical reactors with mixing,
gasification of solids, and hydrocarbon Two-Phase Flow Problems
production. C Riser Reactors
C Fluidized Beds
C Multiphase Flow - simulation of flows C Spray Dryers
C Pneumatic Conveyors/Classifiers
containing solids in gas, solids in liquid, gas C Gas-Solid Jets
bubbles in liquid, liquid emulsions, and liquid
sprays in gas. Applications include design of
safety relief devices for reactors and tanks,
optimizing burners and waste incinerators, eliminating wear failure in pneumatic conveyors,
controlling crystallization of fine chemicals, and improving reactor performance.

C Viscoelastic Polymer Flow - simulating flow of polymers with non-Newtonian rheology into a
mold, through a die, extruder barrel, or gear pump. Applications include coating and fiber spinning.

Research needed to overcome barriers to the development of CFD tools that can be effectively applied
to the simulation of chemical processes is shown in Exhibit 7-1, with different priority levels identified.
Exhibit 7-2 illustrates the time frame (near, mid, and long term) in which various R&D activities are
expected to be accomplished.

Research Needs 19
 Exhibit 7-1. Research Needs for Computational Fluid Dynamics

Phenomenology and
Numerical Methods Constitutive Relations Experimental Validation

Characterize/model dilute to dense Characterize/model dilute to dense Design/develop multiphase flow

phases k regimes (e.g., laminar/turbulent test beds k
flows) k
Develop relevant data sets for Perform experimental validation at
code verification and scaling k Characterize/model interactions small scale k
between phases k
Incorporate complex geometry Ž Conduct small and large scale
Develop reliable turbulence separate effects tests k
Develop adaptive computational closures for multi-phase flows k
grids Ž Develop new diagnostics and
Characterize boundary conditions sensors for experimental
Design modules to customize and interactions (e.g., inlets and measurement of multi-phase flows
complexity for different user needs wall and interior surface k
interactions) k - Non-invasive
Ž
- Full-field, rather than
Develop chemistry models for local or averaged
Improve parallelization techniques
Ž volume and surface phenomenon Ž - Increase spatial and
temporal resolution
Characterize/model polydisperse
Develop more efficient, accurate
systems • Enhance capability for analysis of
algorithms and solvers Ž
results Ž

Characterize/model chemistry and Incorporate population balance •

Develop new experimental methods
chemical coupling phenomenon for
Characterize/model multiphase heat applicable to large scale flows Ž
multi-phases •
transfer •
Develop algorithms to treat the
changing position of a free surface
(e.g., molten polymers) •

Develop wrap-around optimization

using large-scale CFD simulation
and small parameter models •

Key: k = Top Priority Ž = High Priority • = Medium Priority

Numerical simulation capabilities for multiphase CFD should be developed at three levels which parallel
those used in single phase CFD. These include direct numerical simulation (DNS), large eddy
simulations (LES), and Reynolds Averaged (RA) simulations (see Section 3 - Current Situation).
Extension of these single phase methods to multiphase systems will require an integrated effort that
involves development of numerical methods, constitutive theory and concurrent experimental validation.

Numerical Methods
Ultimately, the goal is to develop numerical capabilities that are efficient, accurate and allow simulation
of the complicated patterns involved in chemical processes (e.g., mixing), with the ability to model
conditions as they change over time. Research needs in numerical methods will help to overcome some

20 Computational Fluid Dynamics Technology Roadmap

of the more critical barriers associated with current computational limits and advanced computer
architectures. For example, the development of adaptive computational grids and improved
parallelization techniques will enable better integration of CFD with high performance computing
platforms. The development of more efficient, accurate algorithms and

Research Needs 21
22
Near Term Mid Term Long Term All Time
(0-3 years) (3-5 years) (5-10 years) Frames

Perform Conduct small

Experimental experimental and large
Develop new
Validation validation at scale
Develop experimental
Enhance small scale separate
multiphase methods effects tests
capability for k
flow test beds applicable to k
large scale analysis of
flows M results M Develop new diagnostics
and sensors for
experimental measurement
of multi-phase flows k

Characterize/ Characterize/
model dense model
Phenomenology phases (e.g., interactions
laminar / turbulent between phases Develop
and Constitutive flows k k chemistry
Relations Incorporate models for
Develop reliable population volume and
turbulence closures for balance F surface

Computational Fluid Dynamics Technology Roadmap

multi-phase flows k phenomenon
M

Characterize boundary conditions

and interactions (e.g. inlets and
wall and interior surface
interactions) k

Incorporate Develop Develop models to

complex adaptive customize complexity for

Exhibit 7-2. Research Needs Time Frame

Numerical geometry M computational user needs
Characterize/ Develop more grids M
Methods model dense efficient,
phase k accurate Develop relevant data sets
algorithms Improve Develop for code verification and
and solvers parallelization wrap-around scaling k
M techniques optimization
M using
large-scale Characterize/ model
Model CFD chemistry and
changing simulation and chemical coupling
positions of a small phenomenon for multi
free surface parameter phase F
F models F
mathematical solvers will enhance computational efficiency and speed. Research will also strive to
overcome barriers created by the lack of understanding of basic chemistry and physics of multiphase
flows (e.g., characterization of chemistry and chemical coupling phenomenon).

In terms of time frames for R&D (see Exhibit 7-2), early progress in the development of dense phase
models will be essential to the overall research program in numerical methods. Development of relevant
data sets for validation and scaling is needed throughout the program. The same is true for chemistry
and chemical coupling models—development should begin with less-detailed models that are
continually improved in terms of model complexity and flow coupling. Parallelization is an important part
of the research plan, but is not the primary focus of early work. In the area of optimization, both direct
optimization using large-scale CFD simulations and “small parameter” models that can be run separately
from the simulation will be explored. Modeling the changing position of a free surface (for polymer
flows), developing adaptive computational grids and the ability to model flows around complex
geometries will be a long-term effort.

Phenomenology and Constitutive Relations

Constitutive relations are needed to accurately describe the chemical and physical phenomenon
occurring in specific fluid flows. The development of accurate constitutive relations is a continuing
process of interactions between derivations based on theory, simulations, and experiments. The
accuracy of constitutive relations must be demonstrated through a verification process in which
simulation results are carefully compared with detailed experimental data. Verification requires
simplified, idealized experiments which allow one feature of the process to be isolated and examined.
In general, the desire is to obtain constitutive relations that are accurate, verified, simple, consistent,
hierarchical, general, and theoretically based.

Ideally research in this area will progress along with an increased understanding of the chemistry and
physics of multiphase flows (e.g., modeling of dense phases, characterization of boundary conditions,
development of chemistry models). The development of reliable turbulence closures for multiphase
flows and the incorporation of population balance will enhance computational efficiency and improve
the accuracy and reliability of results.

Early progress in dense phase models, along with development of reliable turbulence closures,
characterization of boundary conditions (inlets and wall and exterior surface interactions), and modeling
of interactions between phases will be essential to this research activity. Chemistry models will be
important throughout this research, including development of models of increasing complexity from
reduced models to full chemistry with surface chemical effects.

Experimental Validation
A well-planned experimental program will provide validation of model results for complex multiphase
flows and increase the usefulness of results to plant engineers, enhancing overall CFD usability.
Experimental validation goes hand in hand with research in numerical methods and constitutive relations.
Experimental tests at a small scale are needed for model development as well as closures for
turbulence, chemistry, phase interfaces, and boundary conditions. Large scale experimental
measurements will be equally important for validating model performance. A strong experimental

Research Needs 23
program is essential to the overall research plan, as the experimental measurements of multiphase flows
are extremely difficult to obtain and require careful analysis when comparing with model results.

The optimal multiphase flow diagnostic for validation of multiphase flow models provides non-invasive,
full-field, spatially (in space) and temporally (over time) resolved measurements of velocity, pressure,
temperature, material phase, and chemical composition. A portion of the research program is designed
to overcome the lack of diagnostics currently available for providing information at this level. A wide
variety of physical techniques (e.g., acoustic, electrical, optical, radiation) should be examined as
possible diagnostic tools. Development of advanced multiphase-flow test beds will be essential for
evaluating a wide range and scale of flows (e.g., size, velocity) and controlling flow regimes. They will
serve as platforms for evaluating diagnostics as well as acquiring data for model development and
validation.

24 Computational Fluid Dynamics Technology Roadmap

Research Needs 25
 8 Summary
CFD Goals

CFD can assist the design and optimization of new and existing processes and products. CFD can also
be used for reducing energy costs, improving environmental performance, and increasing productivity
and profit margins. Although single-phase CFD has been applied to great advantage in the aerospace,
automotive, and power industries, the current packages available are limited for use in the chemical
industry, particularly for reacting and/or turbulent multiphase flows. The chemical industry needs
advanced CFD software that can be used by a broad segment of the chemical community to solve
more complex flow problems. CFD tools for the chemical industry need to be:

C Versatile — portable to various machines, usable by generalists rather than just specialists, scalable,
reliable, and compatible with current chemical industry data bases.
C Fundamentally Based — containing more and better physics coupled with chemical reactions,
rather than non-mechanistic correlations that are not scalable.
C Experimentally Verified — simulations validated with experimental results.
C Computationally Efficient — adaptable to multiprocessor work stations and clusters of processors,
able to expediently provide useful results.
C User Friendly — accompanied by professional technical support that is appropriate for industrial
engineers as well as researchers.
C Information Transfer — include mechanisms for information dissemination and educational support.

The Road to Follow

Workshop results have demonstrated that precompetitive research and development is needed to
advance the state of the art and to build capabilities for computing multiphase flows. Research should
include activities in numerical methods and constitutive theory to advance computational capability for
reactive and turbulent multiphase flows. The research agenda should include a well-integrated
experimental program that will allow validation of models developed for practical chemical processes.

For effective resource leveraging, risk minimization, and providing a stable baseline for funding,
precompetitive research should be cooperatively supported through the chemical industry, commercial
software vendors, and the Federal government (Department of Energy and Department of Defense).

26 Computational Fluid Dynamics Technology Roadmap

For maximum use by industry, the development process will be directed by the technology users - the
U.S. petrochemical industry.

Industry has identified the Federal R&D Laboratory system as a candidate for leading CFD technology
development. As such, Federal support of R&D laboratories for CFD development, on behalf of
industry, could maximize the returns on the R&D investment. Project selection, control, and evaluation
should be carried out by a broad cross-section of potential industrial users. Federal laboratories can
partner with individual industries for technology development, and universities can address fundamental
issues in the development of this technology.

The technology cannot be widely applied without the customer service and support base of the
software business, and these businesses must be brought into the technology program as soon as
possible.

Summary 27
 References
CMA 1996. U.S. Chemical Industry Statistical Handbook 1996, Chemical Manufacturers
Association, Washington, D.C.

DOC 1994. 1994 Annual Survey of Manufactures, U.S. Department of Commerce.

DOW 1996. Tyler B. Thompson, “Computational Fluid Dynamics for the Chemical Processing
Industry,” June 1995, DOW Chemical Company, Midland, Michigan.

LANL 1995. W. Brian VanderHeyden, Reactive Multiphase Flow Simulation Workshop Summary,
Los Alamos National Laboratory, June 1995. [LA-UR-95-2194].

NATL 1996. Technology Roadmap Workshop for the Chemical Industries of the Future -
Computational Fluid Dynamics and Multiphase Flow, October 1996, multiple authors.

VISION 1996. Technology Vision 2020: The U.S. Chemical Industry, 1996, a collaborative effort of
the American Chemical Society, the Chemical Manufacturers Association, the Council
for Chemical Research, the American Institute of Chemical Engineers, and the Synthetic
Organic Chemical Manufacturers Association. Available from the American Chemical
Society, Washington, D.C.

28 Computational Fluid Dynamics Technology Roadmap

 Acknowledgments
The chemical industry and the U.S. Department of Energy gratefully acknowledge the efforts and
support of numerous individuals who helped to create this document. The following is a list of
contributors from companies, government agencies, national laboratories, and academic institutions.

Clyde O. Beasley, Jr. Don Hardesty

Oak Ridge National Laboratory Sandia National Laboratories

Dan Butler Jay Hargreaves

Los Alamos National Laboratory Amoco Research Center

Philip Colella Bill Heard

Lawrence Berkeley National Lab. Exxon Research & Engineering Co.

Bruce H. Cranford, Jr. Mike Henesey

U.S. Department of Energy IBM

Clayton Crowe Mark Holmes

Washington State University Silicon Graphics, Inc.

Jack Dorning John W. Hopson

University of Virginia Los Alamos National Laboratory

James A. Fort Phil Hsieh

Pacific Northwest Laboratory Alcoa Tech Center

Ehr Ping Huang Fu Edward L. Joyce, Jr.

U.S Department of Energy Los Alamos National Laboratory

Walter Fuchs Kailas Kailasanath

U.S. Department of Energy Naval Research Laboratory

Sue Gelderbloom Sang-Wook Kang

Dow Corning University of California

Paul Gillis Thomas A. Kitchens, Jr.

Dow Chemical U.S. Department of Energy

Mark Goldman Kostas Kontomaris

Silicon Graphics, Inc. DuPont Central Research

Acknowledgments 29
Dick LaRoche Massimo Sangalli
Silicon Graphics, Inc. Amoco Co.

Clarence Law Paul Sechrist

DuPont Central Research Universal Oil Products

Lucia Liljegren Dennis N. Smith

Pacific Northwest Laboratory U.S. Department of Energy

Robert Lyczkowski Philip J. Smith

Argonne National Laboratory University of Utah

Thomas J. O’Brien Professor Sankaran Sundaresan

U.S. Department of Energy Princeton University

John Oleson Tyler B. Thompson

Dow Corning Dow Chemical Company

James Park Jim Tippett

Oak Ridge National Laboratory DuPont Central Research

Arthur C. Ratzel, III Richard A. Valentin

Sandia National Laboratories Argonne National Laboratory

Mike Roco Brian Volintine

National Science Foundation U.S. Department of Energy

Upenda S. Rohatgi Ed Wolfrum

Brookhaven National Laboratory National Renewable Energy Lab

Charles Russomanno
U.S. Department of Energy

30 Computational Fluid Dynamics Technology Roadmap