1) Traditional IC engine CFD simulations use simplified chemistry models that are insufficient for designing new high-efficiency engine concepts where kinetics effects dominate.
2) Detailed chemical kinetics models have been developed through the Model Fuels Consortium but cannot be directly incorporated into CFD due to high computational cost.
3) Reaction Design's CFD software uses novel parallel chemistry solvers that dramatically reduce simulation times, allowing the use of detailed chemical kinetics models for predictive 3D engine simulations.
Design & Development of Energy management strategies for the improvement of f...
CFD Design Flow for Predictive IC Engine Simulations
1. Efficient and Effective CFD
Design Flow for Internal
Combustion Engines
March 14, 2010
REACTION DESIGN
www.reactiondesign.com
+1 858-550-1920
2. Traditional IC engine combustion simulations involve CFD models that use a simplified chemistry
representation for fuel combustion. The chemistry in the models range from just a few molecular species
to ~50 species for Diesel fuel, for example. Alternative approaches use table-lookup strategies and
progress variables to avoid the cost of direct computation of the chemistry-flow interactions. For
conventional Diesel and Gasoline engines, these approaches have historically been good enough, because
the fluid-mixing effects dominated the kinetics effects in predicting engine performance.
New engine designs present new simulation challenges
New, high-efficiency, low-emissions designs present technical challenges that are dominated by kinetics
(e.g., dual-fuel engines, staged spray injections for improved efficiency, Premixed Charge Compression
Ignition (PCCI) combustion, low temperature conditions, etc.). What proved to be good enough for the
design of yesterday's engines is insufficient for today's new engine designs. A consistent complaint by the
industry is that they cannot rely on combustion CFD to predict values or even accurate trends in critical
combustion behaviors such as ignition, flame propagation and emissions. This problem is exacerbated by
the fact that the fuels landscape continues to evolve and become more complex. Where yesterday’s
engines were designed for a single fuel type, such as diesel or gasoline, today's engine specifications
demand fuel flexibility while achieving ultralow emissions.
The Model Fuels Consortium is an industry-led program, currently in its sixth year, which has developed
both the detailed chemical mechanisms and the tools required to simulate real fuel behavior. While the
MFC has been exceedingly successful in developing fuel mechanisms that accurately simulate real fuel
chemistry, it has proved the impracticality of reducing these mechanisms they can be incorporated into
contemporary CFD simulations without a substantial loss in accuracy. MFC researchers have recognized
that the focus should shift from trying to get reliable results with mechanisms so severely reduced that
they cannot capture real fuel behavior, to enhancing the ability of simulation tools to use mechanisms
with the necessary level of detail.
One of the Department of Energy’s premier scientific laboratories studying engine efficiency recently
acknowledged the critical link between the need to reduce greenhouse gas emissions and advanced
simulation in a white paper entitled: “Predictive Simulation of Combustion Engine Performance in an
Evolving Fuel Environment.”i The paper points out that engine manufacturers must move to “change
from a test-first culture to an Analysis-Led Design Process” and that “a predictive simulation toolkit
would accelerate the market transformation to high-efficiency, clean power sources for transportation.”
Kinetics is recognized as a critical area for advancement supporting the design of clean, fuel-flexible
engines that reduce greenhouse gas emissions.
Another key area of concern in engine simulation has been spray modeling. The choice of the spray
model can have a significant impact on both time-to-solution and the accuracy of results. Most of the
spray models used today are highly mesh dependent, which requires that valuable innovation time be
Reaction Design 1
3. spent adjusting or adding complexity to the mesh, to find the optimal combination of spray-model
parameters and grid. Problematically, this approach requires that the behavior of the spray in the
cylinder is known in order to tune the model to predict it. Even when a spray model can be calibrated to
a particular grid, it is unclear how effective the model will be on a different engine design, which may
require the whole process to be repeated. Understanding how to do this calibration requires specific
expertise and makes it difficult for widespread utilization of predictive CFD across the organization.
The lack of reliability in combustion simulations is likely caused by a lack of detail in the way the fuel-
spray and combustion kinetics are represented. Because the industry has been limited in the amount of
chemistry detail it could practically incorporate into a simulation, work has focused on turbulence-
mixing phenomena, use of approximate combustion models, and meshing. But, because of the
increasing challenges in today’s engine design environment, attention is once again turning to improved
modeling of the spray and kinetic phenomena.
How engine designers address the challenges today
The dominant way of dealing with time-to-solution issues over the last five years was simply to buy more
CPUs and use brute force to get a solution in a reasonable amount of time. Unfortunately, the inherent
limitations of conventional CFD approaches prevent the use of larger, more accurate mechanisms due to
solution complexity and numerical stability issues. Another common approach was to employ severely
reduced chemical mechanisms in CFD simulations, hoping that important combustion behavior might be
predicted even though most of the details had been removed. This approach worked for conventional
engine design by relying on vast amounts of empirical performance data, but these data do not exist for
today’s novel engine designs.
Some in the industry claim that predictive results are not achievable without engine calibration. This
means that in the end, the price of an inaccurate model is using extensive data to “tune” the simulation
model. The tuned CFD approach, however, often fails to translate to good results under different engine
operating conditions. This prevents in-cylinder combustion CFD from being a truly predictive design
tool. The impact of the lack of reliable results from existing CFD approaches is that production design
engineers cannot use them efficiently and this work must be done by expert R&D personnel or
outsourced to groups with specific expertise. Sometimes, combustion simulation is avoided completely
and non-reacting simulations are used to identify parameters such as local fuel/air ratio or spray
distribution and used to infer the effect on combustion performance.
Reaction Design 2
4. Figure 1: CFD design flow
The ideal CFD design flow
• Go directly from CAD drawings into running CFD cases
• Easy, graphical setup of the CFD case
• Incorporate experimental results as inputs to the CFD case
• Create parameter studies to conduct Design of Experiments on operating conditions
• Accurate fuel chemistry models to predict real fuel behavior and emissions formation
• Incorporate spray models that are truly predictive and independent of mesh size
• Spark ignition models must accurately and efficiently track the ignition, flame propagation and
onset of knock for today’s fuel and engine designs
• Powerful and smart chemistry solvers to tackle the daunting challenge of using accurate chemistry
• Seamlessly create, view and analyze the CFD results that an engine designer cares about without
the use of postprocessor at additional expense.
Treating each of these areas as point solutions builds inefficiencies into the CFD design flow that can have
dramatic impacts on its effectiveness. Improvements in one facet of the flow can slow down other facets
or affect accuracy. Weak or disjointed links in the flow can cause unnecessary delays or a loss of
information that also hinder CFD’s value as an effective design tool. Meshing can be handled
automatically or adaptively, but it can also generate a substantially larger number of cells or introduce
numerical errors that negatively impact run time and accuracy. Command-line software interfaces
require engineers to master a series of arcane user inputs and serve to inhibit wide use by developers.
Using progress variables and lookup tables as ways to manage computational complexity can also impair
the ability of CFD to be used as a predictive tool on cases where either high-EGR, low-temperature
Reaction Design 3
5. combustion, or alternative fuels are present. The overall success of a predictive CFD design flow depends
not only on the accuracy of the simulation results, but also on the timeliness and ease of generating those
results.
A new approach: achieving accuracy by modeling real fuel chemistry
For advanced-concept engines, chemical kinetics takes a front-seat role in controlling ignition behavior, as
well as emission and knock performance. Managing uncertainties in fuels and fuel composition requires
use of a high-fidelity fuel model in design calculations. Traditional CFD models are stymied by these
requirements, forcing designers to rely on expensive empirical methods for exploring and verifying new
ideas.
Powerful chemistry solutions
The barrier to good fuel representation in CFD simulations is not the lack of information about the
detailed chemical kinetics of fuel combustion. In fact, there has been huge growth in the understanding
of the combustion behavior of liquid transportation fuels over the last decade through work validated by
the Model Fuels Consortium. A surrogate-fuel approach was used in fuel-combustion studies, where a
small set of fuel-component molecules were selected to represent real fuels. In conjunction with this, the
MFC developed very detailed, molecular-based kinetics representations of the important surrogate fuel
components for conventional and alternative automotive fuels. Consortium researchers showed that
surrogate-fuel models that employ fundamental chemical kinetics information can capture details of fuel
ignition, flame propagation, pollutant emissions, particulate formation and engine knocking, as well as
the effects of fuel variability and multi-fuel strategies.
Results demonstrate both quantitative and qualitative prediction capability for combustion behavior, as
seen in Figure 2, where a reduced mechanism with ~100 species are compared to a more accurate
mechanism with 428 species. Experimental data are represented by the solid triangles. The larger
mechanism is shown to have sufficient accuracy required to provide excellent prediction of emissions
values and trends.
Figure 2: Dramatic improvement in the accuracy of CFD emissions results when using an accurate
mechanism with 428 species (solid line) compared to a reduced mechanism with ~100 species (dashed line).
Reaction Design 4
6. Figure 3: Dramatically reducing chemistry calculation time in CFD allow the use of more accurate
chemistry for good results without expert calibration.
Critical time time-to-solution advancement: Automatic Mesh Generation
Creating meshes for internal combustion engines is difficult. The typical engine CFD design project
begins with a lengthy process to construct an adequate representation of the cylinder and port geometry
using a mesh of computational cells. The construction must account for the fact that the mesh must
transform and shift dynamically with the motion of pistons and valves during the engine cycle. This
process can take weeks for a single-cylinder configuration, making a design- of-experiments that
considers geometry changes particularly challenging. Mesh generation has become the realm of a limited
number of experts who know all the tricks that are required to get an accurate and robust mesh.
Automatic mesh generation eliminates a key bottleneck from the design flow by importing CAD
drawings directly into the CFD environment. The key to success of this automation strategy is to ensure
that the implementation neither slows down other phases of the design flow nor introduces errors. From
an accuracy point of view, the ideal mesh created is one that is Cartesian, with perfectly orthogonal faces,
and one in which the boundary conditions are enforced exactly on the physical surfaces of the real
geometry. Automatic-mesh-generation methods that use a pure Cartesian based system avoid the
problems of highly skewed cells that can be introduced with other approaches.
Can you get accuracy in combustion CFD with reasonable solution times?
This is certainly the key question and time-to-solution has been a key barrier to incorporating sufficient
chemistry accuracy into CFD calculations. As most commercial CFD improvements directed toward
better accuracy have focused on enhancing meshing and turbulence modeling, there has been little effort
directed toward improving the fundamental chemistry calculations, to reflect the key engine behaviors
that are now beginning to dominate the design space. Given that chemistry calculation times in CFD can
Reaction Design 5
7. account for 90% of the total simulation time even when employing severely reduced mechanisms, there is
substantial opportunity for decreasing time-to-solution by accelerating these calculations.
Reaction Design’s CFD package, called FORTÉ, employs a novel solver approach that takes advantage of
the chemical similarity of groups of cells and implements a parallel processing algorithm to dramatically
reduce the chemistry calculation time. This technique can reduce simulation run times by almost two
orders of magnitude, as demonstrated in Figure 3. Chemistry models that previously were thought of as
only practical for 0-D simulations are now practical for full 3-D engine simulations complete with moving
pistons and valves. With innovative approaches to relieving the bottleneck in chemistry calculations,
predictive engine simulation is now a reality.
Reference
i
“Predicting Simulation of Combustion Engine Performance in an Evolving Fuel Environment,” US DOE Sandia
White Paper, submitted by Robert W. Carling, February 25, 2010.
Reaction Design 6