AcuSolve 2017.1 Training Manual
AcuSolve 2017.1 Training Manual
AcuSolve 2017.1 Training Manual
1
AcuSolve Training Guide
Special Notice: Pre-release versions of Altair software are provided ‘as is’, without warranty of any
kind. Usage is strictly limited to non-production purposes.
The Altair web site is a valuable online companion to Altair software. Visit www.altairhyperworks.com
for tips and tricks, training course schedules, training/tutorial videos, and other useful information.
Altair training courses provide a hands-on introduction to our products, focusing on overall
functionality. Courses are conducted at our main and regional offices or at your facility. If you are
interested in training at your facility, please contact your account manager for more details. If you do
not know who your account manager is, please send an e-mail to [email protected] and your account
manager will contact you.
When contacting Altair support, please specify the product and version number you are using along
with a detailed description of the problem. Many times, it is very beneficial for the support engineer to
know what type of workstation, operating system, RAM, and graphics board you have, so please have
that information ready. If you send an e-mail, please specify the workstation type, operating system,
RAM, and graphics board information in the e-mail.
To contact an Altair support representative, reference the following table or the information available
on the HyperWorks website:
http://www.altairhyperworks.com/ClientCenterHWSupportProduct.aspx
For questions or comments about this help system, send an email to [email protected].
In addition, the following countries have resellers for Altair Engineering: Colombia, Czech Republic,
Ecuador, Israel, Russia, Netherlands, Turkey, Poland, Singapore, Vietnam, Indonesia
Official offices with resellers: Canada, China, France, Germany, India, Malaysia, Italy, Japan, Korea,
Spain, Taiwan, United Kingdom, USA
See www.altair.com for complete contact information.
Contents
3 Theoretical Background.............................................................................................20
1
3.5.4 Scalar Transport of Multiple Species................................................................ 143
3.5.5 Aeroacoustics................................................................................................ 143
3.5.6 Turbulence Modeling...................................................................................... 144
2
AcuSolve Training Manual
Introduction 1
1 AcuSolve Training Manual Introduction
The AcuSolve training class is to teach engineers or students who have no previous experience with
AcuSolve. It is designed to expose them to background knowledge regarding flow physics and CFD as
well as detailed information about the use of AcuSolve and what specific options do. The training class
is intended to be a modular format that allows instructors to choose what topics to cover. Each module
contains a detailed explanation of the content in the form of a reference (or training) manual, and
comes with a condensed version of the content in the form of a presentation. The presentation material
is intended to be used by the instructors for teaching the class, while the training manual is intended to
be a resource for those that want additional information on a given topic. The training manual contains
the following topics:
• Brief introduction about CFD, its history, applications and advantages
• Basics of flow phenomena, including flow models, governing equations
• Turbulence physics and its modeling
• Introduction to AcuSolve, its background and features
• Hands on training on the use of AcuSolve and its supporting features using tutorials.
Introduction to CFD and AcuSolve 2
2 Introduction to CFD and AcuSolve
This section details the various approaches available for analyzing a general fluid flow problem. It also
provides pros and cons for each of these approaches.
Analysis of any physical problem, in general, is governed by a set of basic governing equations. The
basic equations governing the motion of fluids are the Navier-Stokes equations, which were developed
more than 150 years ago.
Continuity Equation
Momentum Equation
Energy Equation
These equations are a set of coupled partial differential equations which describe the relation among
the flow properties like pressure, temperature and velocity. The solution to the above equations is in
general complex because of the non-linear terms and coupled nature of the equations. However, in
general the following three solution approaches are available.
Analytical Approach
An analytical solution can be obtained by integrating the above boundary value problem, resulting
in an algebraic equation for the dependent variables as a function of independent variables.
Because of the complexity of the Navier-Stokes equations a general solution for all physical
problems is an insurmountable task. However, analytical solution is possible in some simple cases
with certain approximations in geometry, dimensionality and compressibility of the flow. A few
examples of such flows include Couette flow, Hagen-Poiseuille flow and Parelle flows through
straight channels. An obvious limitation of this approach is the practical applications of these
flows.
Experimental Approach
This approach uses the physical experiments on the scale down models of the problem at hand
and extrapolates the results based on similarity laws. For example, to investigate the flow around
an actual aircraft a Reynolds number based miniature model is placed in the wind tunnel and
the results are extrapolated to that of the actual aircraft. Although a variety of experimental
techniques are available, the choice of the particular type to use depends on capturing the
parameter of interest. For example, Particle Image Velocimetry (PIV), Laser Doppler Velocimetry
(LDV) and Hot-wire anemometry are used to measure flow velocity. PIV is used to produce 2D
or 3D vector fields. Hot-wire anemometry and LDV are used when measuring the rapidly varying
velocities at a point with good spatial and time resolution is important. A few limitations of this
approach include the necessity of conducting a large number of experiments for the complete
description of flow, cost and the complexity of conducting them.
Numerical Approach (Computational Fluid Dynamics-CFD)
The third approach available for investigating fluid flow is CFD. In this approach, the fluid region
of interest is divided into finite regions and the complex governing equations are discretized to
obtain a set of algebraic equations that are advanced in time and space using computers. The end
result of the CFD analysis is a complete description of the fluid flow within the region of interest.
This section details the advancements in CFD research throughout the twentieth century.
Although the study of fluid mechanics is centuries old most of the evolution of CFD took place
throughout the twentieth century. By the start of the twentieth century all of the basic governing
equations were consolidated. Closed form analytical solution for simple problems were derived
putting aside the large set of problems of practical significance that are not suitable for analytical
solution. Before the 1950's the main focus was on developing a variety of approximated semi analytical
techniques such as perturbation methods for viscous boundary layer calculations, a method of
characteristics for compressible flow calculations alongside the numerical solution development. By the
mid-century the Finite difference based numerical methods were developed. However, the numerical
algorithms until the fifties provided a solution to the single partial differential equations limited by the
computation power. With the advent of digital computers the development of numerical simulation
was put on an accelerated path. During the fifties and sixties significant developments were made on
solution of inviscid compressible flows.
The seventies largely focused on inviscid flows over geometries like airfoils, wings and wing-body
combinations. During the eighties intense efforts were made on algorithms for solution of Euler
equations which include rotational flows but still inviscid. Research on grid generation was invested
in the late eighties which presented a challenge in generating suitable grids for the computational
domain of complex geometries. During the late eighties the transonic flow computations with viscous
compressible flow over airfoils was made possible. This resulted in designing the super critical airfoils
computationally with reduced wave drag. For the first time one could move away from the catalogued
airfoil shapes to custom-designed airfoils to suite design end.
In the nineties the power of parallel computing paved the way for turbulence modeling in the flow.
Turbulence was incorporated using turbulence modeling techniques such as Reynolds-averaged Navier
Stokes (RANS) and Large Eddy Simulation (LES) along with Direct Numerical Simulation (DNS) without
any modeling. This period marked the beginning of an era of commercial software capable of impressive
visualization of vast numerical data generated by the CFD simulation of flow fields.
This section details the few of the numerous industries in which CFD has become indispensable.
CFD has become an indispensable tool in diverse industries for a variety of applications.
Aerospace Industries
The development of CFD in its formative years was mostly driven by the Aerospace industry. From
the initial design and development phases of an aircraft the application of CFD is prominent in various
modules of aerodynamics design such as configuration design, estimation of aerodynamic loads,
estimation of stability derivatives and performance of engines. The design and development of today’s
state of the art missiles and launch vehicles would be unimaginable without CFD.
Figure 1:
Automotive Industries
The fuel efficient and stream lined automobiles in today’s market are results of extensive use of CFD
analysis. For example, CFD tools play a big part in external aerodynamics, the engine coolant jacket,
under hood thermal management and passenger comfort.
Figure 2:
Figure 3:
Civil Engineering
CFD is used for building aerodynamics, hydraulic systems, aeroacoustics, fire and smoke simulation
resulting in improved safety features at lower costs.
Figure 4:
Naval Engineering
CFD is a major tool in solving hydrodynamic problems associated with harbors, ships, submarines,
torpedoes, oil exploration platforms, and so on.
Figure 5:
HVACR Engineering
CFD is used for heating, ventilation, air conditioning and refrigeration applications. CFD analysis gives a
detailed description of fluid flow, heat transfer and chemical species transfer in HVAC systems.
Figure 6:
Figure 7:
Heavy Engineering
Heavy engineering is another industry that has a strong presence of CFD applications. CFD provides
greater insight for the design of products such as earth moving machinery, cranes, excavators, and so
on.
Figure 8:
Chemical Industries
CFD applications in the chemical industry involve combustion, mixing, multiphase flows and cavitation
for evaluating the accurate heat/mass transfer parameters.
Figure 9:
Electronic Industries
CFD analysis can provide great value in predicting temperature distribution over a chip and the
deposition rate therefore reducing the number of experiments.
Figure 10:
Medical Industries
The application of CFD in the medical industry includes analysis of blood circulation, impact of artery
blockage, air circulation in lungs, and so on.
Figure 11:
Aside from the conventional industries mentioned above CFD is also making strides into newer
platforms such as movies, computer graphics, food processing, and so on.
This section gives a brief background of Altair HyperWorks- AcuSolve and a brief introduction of
AcuSolve.
Altair HyperWorks, which was established in 1985, is a software giant in the Computer Aided
Engineering (CAE) Industry with varied product offerings like RADIOSS, OptiStruct, MotionSolve,
HyperXtrude, and so on. Altair acquired AcuSim Software Inc. in 2011 to cater to the growing needs of
the CFD industry.
Apart from CAE products, Altair operates business in product design consultancy and cloud computing
services.
AcuSolve is a leading general purpose CFD solver that is capable of solving the most demanding
industrial and scientific applications. Based on the Finite Element method, AcuSolve’s robust and
scalable solver technology empowers users by providing unparalleled accuracy on fully unstructured
meshes. Applications ranging from steady Reynolds-averaged Navier-Stokes (RANS) simulations
to complex, transient, multi-physics simulations are handled with ease and accuracy. AcuSolve is
used worldwide by scientists and engineers to explore applications in a variety of markets including
automotive, aerospace, energy, electronics cooling and chemical processing.
This section details the integration of AcuSolve with the industrially accepted pre and post-processing
tools of HyperWorks.
For example, traditional users of HyperMesh can prepare the model in HyperMesh and then can either
export the input files for AcuSolve or launch the AcuSolve solver directly by clicking File > Export >
Model.
After the solution is obtained from AcuSolve, HyperWorks can be used for various post-processing
applications. Although AcuSolve provides AcuProbe and AcuFieldView for post-processing, traditional
users of HyperWorks can make use of HyperWorks post-processing tools. For example, AcuSolve .log
files can be directly read from HyperView by clicking File > Load > Results and then choosing the
appropriate AcuSolve .log file.
HyperStudy is also integrated for AcuSolve. While performing a parametric study of the CFD simulation,
HyperStudy can be used to its fullest advantage. HyperGraph can also be used with AcuSolve to
visualize the graphical interpretation of the results. To do this you must convert the AcuSolve results
into a tabular format using AcuTrans (in the AcuSolve command prompt use the command acuTrans –
out ).
This section details the capabilities of AcuSolve in simulating the multi-physics problems by coupling
with structural solvers.
One of the rich features of AcuSolve lies in its ability to simulate FSI problems. AcuSolve simulates
these problems using two different technologies.
Practical FSI (P-FSI)
In P-FSI simulations, AcuSolve and the structural code are run separately. The Eigen solution
obtained from a structural solver is projected onto the CFD mesh and used as boundary conditions
to displace the fluid mesh. The mass, stiffness and damping arrays are also transferred from the
structural model to flexible body in the CFD model, completing the inputs necessary to compute
the structural deformation. P-FSI is preferred for simulations that have linear structural response
(linear material properties and small mesh deformations).
Direct Coupled FSI (DC-FSI)
For applications involving a nonlinear structural response, AcuSolve is coupled at run time to a
structural solver using the DC-FSI approach. The codes are run in tandem and exchange forces
and displacements at each time step. This technique captures all nonlinearities in the structural
model and can exploit the full capabilities of the structural solver. The DC-FSI technology within
AcuSolve performs all projection and interpolation between the structural and CFD meshes
without the use of any middleware. The supported structural solvers for AcuSolve include
RADIOSS, Abaqus and MD-Nastran.
AcuSolve can be directly coupled with MotionSolve, a HyperWorks multibody dynamics solver, to solve
multi-physics problems that involve both fluid and rigid body dynamics. The wetted surfaces are paired
with rigid bodies and loads. Displacements are exchanged during the run time using AcuSolve’s code
coupling interface. The coupling is currently supported for rigid body motion only and not for a flexible
body. However, AcuSolve supports both one way and two way coupling with MotionSolve. With one
way coupling, the displacements are read from a file generated from MotionSolve and projected onto
AcuSolve. With two way coupling, the codes exchange forces and displacement at each time step.
This section on mathematical background covers the various notations and operators used to formulate
and define the equations of fluid flow.
The conservation laws for a continuum medium involve vector and tensor quantities as well as several
operators such as gradient and divergence. In order to have a comprehensive understanding of these
equations it is essential to get a good grasp of the notations and operators used in this manual.
In this notation, the governing equation is independent of the choice of the coordinate system. For the
purposes of this manual the scalar quantities would be denoted with italicized letters. For example,
pressure field is denoted as . The vector quantities would be denoted with an arrow above the letters,
for example velocity field denoted as . Tensor quantities are donated with bold face letters, for
example stress tensor represented as .
In this notation, an index subscript is written after the quantity which corresponds to a component of
the quantity. Single subscript is used to denote a component of vector, for example for the vector
field and two subscripts are used to denote a component of tensor, for example for the stress
tensor .
The Kroneker Delta is a second order isotropic tensor which is defined as:
3.1.4 Operators
The most frequently used math operator related to flow equations is the referred to as nabla, grad
or del. The nabla operator operates on the quantity to the right of it and the rules of a derivative of a
product still hold. Otherwise the nabla operator behaves like any other vector in an algebraic operation.
It is a vector operator and for a Cartesian coordinate system, it is defined as
appears in several different ways when applied to scalar, vector and tensor quantities.
Gradient
When the nabla operator is applied on a scalar quantity , it is called Gradient and it gives a vector
whose components are the partial derivatives.
The gradient of points to the direction of greatest change of and has a magnitude equal to the rate
of change of with respect to distance in that direction.
When gradient is applied to a tensor quantity it produces a tensor one rank higher. When applied to a
vector quantity it gives a second order tensor given by
The gradient of a vector quantity represents the gradient of each component of the vector field
individually, each of which is a scalar.
Divergence
The divergence of a vector quantity is defined as a scalar quantity given by
The divergence of a vector field represents the flux generation per unit volume (flux density) at each
point on the field.
The divergence of a tensor gives a tensor which is one rank lower. When applied to a second order
stress tensor it yields a vector field whose ith component is expressed as
Div
The div operator is defined as and is expressed as
It maps a vector quantity to a scalar which can then be applied to a scalar, vector or higher order
tensors.
Laplacian
The Laplacian of a scalar quantity is a scalar defined as
Total Derivative
The total derivative, also known as material derivative or substantial derivative, describes the rate of
change of a physical quantity of a fluid element with time as it moves along a trajectory in a velocity
field. The total derivative of a scalar quantity is a scalar defined as
The total derivative of a quantity can be seen as a sum of local time derivative and convective
derivative.
This section on basics of fluid mechanics covers topics describing the fundamental concepts of fluid
mechanics such as the concept of continuum, the governing equations of a fluid flow, definition of
similitude and importance of non-dimensional numbers, different types of flow models and boundary
layer theory.
Any material, solid, liquid or gas is composed of billions of individual molecules, in a very small region,
separated by empty spaces in between them. The continuum assumption considers that matter is
continuously distributed and fills the entire region of the space it occupies.
A continuum material can be continually sub divided into infinitesimal elements with properties being
those of the bulk material. Properties such as density, pressure, temperature and velocity are taken to
be well-defined at infinitely small points.
For a continuum fluid, at each point of the region of the fluid it is possible to construct one volume
infinitesimally small enough compared to the region of the fluid and still big enough compared to the
molecular mean free path. This condition is mathematically expressed using a non dimensional number
known as a Knudsen number.
Describing a fluid flow quantitatively makes it necessary to assume that flow variables, such as pressure
and velocity, and fluid properties vary continuously from one point to another. Mathematical description
of flow on this basis have proved to be reliable and treatment of fluid medium as a continuum has firmly
become established.
This section describes the definition and formulation of the equations governing the conservations of
mass, momentum and energy in a fluid flow and obtaining a closed form solution from these equations.
The equations which govern the motion of a fluid are called the Navier-Stokes equations. These
equations refer to the basic governing continuity equations for a compressible, viscous, heat conducting
fluid.
They describe the transport of three conserved quantities for a local system: mass, momentum and
energy and can be expressed in an integral form when applied to a finite region known as the control
volume, or a differential form when applied at an infinitesimally small fluid element.
Continuity Equation
The instantaneous mass conservation equation commonly termed as the continuity equation, is derived
by applying mass conservation to a control volume for a general fluid.
where
The first term describes the rate of change of density with respect to time and the second term
describes the divergence of the vector field at a particular point fixed in space. The continuity
equation can also be expressed using the substantial derivative and is written as:
Momentum Equation
The momentum conservation equation is derived by applying Newton’s second law of motion to the fluid
control volume, which applies as the rate of change of momentum of the fluid particle equals the sum of
surface and body forces.
The surface forces comprise of viscous forces (Tensile and shear) and pressure (compressive) and the
body (volume) forces can be gravity, centrifugal force, Coriolis force, electromagnetic force, and so on.
where
is the pressure
is the source term (generally gravity)
is the viscous stress tensor
If the fluid is Newtonian in nature, the viscous stresses are proportional to the time rate at which strain
occurs. Stokes hypothesis for a Newtonian fluid simplifies the stress tensor to:
The general form of the momentum equation can be seen to be comprised of the following terms:
The momentum conservation equation for the flow velocity vector can be written as three separate
equations in an orthogonal system. These equations describe the conservation of momentum in x, y, z
directions with flow velocity components u, v and w.
Energy Equation
The energy conservation equation is derived by applying first law of thermodynamics to fluid control
volume, which applies as rate of change of energy of a fluid particle equals the sum of rate of heat
addition and rate of work done.
The energy of a fluid particle comprises of internal (thermal) energy and kinetic energy. Heat addition
generally takes place in the form of conduction due to a temperature difference through sources such
as chemical reactions, potential energy, and so on. Work is done on the fluid volume by the surface and
body forces.
where
There are different possible versions of the energy equation. The form written above is known as the
total energy form. Another form which describes the convection of internal energy and more commonly
used is written as:
where
is the temperature
is the thermal conductivity of the fluid
is the double dot product defined as describing the irreversible transfer of mechanical
where
is the energy transfer due to work done by compressive forces, that is, pressure work
The five equation sets described above, one continuity, three momentum and one energy equation
contain unknowns: , , , , , , . In order to obtain a closed form solution for this system of
differential equations, two more equations are needed. For a compressible flow these equations come in
the form of equations of state. They are stated as:
and
and
These relations give an additional equation for enthalpy given by . When substituted into the
energy equation it simplifies to:
This section describes the concepts of similitude and non-dimensional numbers and their importance in
fluid mechanics.
Similitude is a concept that relates the behavior of an object in a given flow field with its behavior in
a different flow field under different operating conditions. This concept allows you to compare any two
scenarios (simulations or experiments) and determine the similarities in the flow fields between them.
This concept becomes useful when there is a prototype of a vehicle, aircraft wing or any such
application to be tested under laboratory conditions. It is not always feasible to test the full scale
prototype owing to the large size or requirement of large scale flow conditions. Scaled models
allow testing of a prototype prior to its production which can greatly increase the efficiency of the
development process.
Geometric Similarity
Geometric similarity refers to similarity relations between the linear dimensions of the test scale model
and the prototype. This means the model has the same shape as the prototype. More precisely, the
model can be superimposed on the prototype by geometric operations like scaling, translation, rotation
and reflection.
Kinematic Similarity
Kinematic similarity refers to the similarity of motion between the model and prototype in a fluid. This
means that the flow streamlines are similar for the model and the prototype. More specifically, the
velocities at corresponding points are in the same direction and are related in magnitude by a constant
scale factor.
Dynamic Similarity
Dynamic similarity refers to the similarity between the force ratios at corresponding points and
boundaries of the model and the prototype. This means that force distributions of the same type on the
model and prototype have the same direction and are related in magnitude by a constant scale factor.
In order to analyse the above similarities certain non-dimensional numbers are used. The use of these
numbers drastically simplifies the task of inferring the experimental data from the model and applying it
to prototype design. Another advantage is that these numbers are independent of units of measurement
and hence can be interpreted to suit any measurement system.
The most widely used non-dimensional numbers in fluid flow analysis are the following.
Reynolds Number
Reynolds number is the ratio of inertial (momentum) forces to viscous forces. It is an important
parameter that can be used to determine the dynamic similarity between two cases of fluid flow or it
can be used to characterize the flow regime in a fluid, this is, the laminar or turbulent nature of the
flow. It is also used to study the transition between laminar and turbulent flows.
A laminar flow occurs at low Reynolds number when the viscous forces dominate resulting in a smooth
velocity and pressure field where as a turbulent flow occurs at high Reynolds number where the inertial
forces dominate the flow resulting in instabilities having a large range of time and length scales.
It is defined as
where
The Reynolds number can also be expressed as the ratio of total momentum transfer to molecular
momentum transfer.
The characteristic length is geometry dependent and defines the scale of physical system. For example,
the characteristic length for a flow in a pipe with circular, square or annular cross section is taken as the
hydraulic diameter of the pipe.
where
For a fluid moving between two plane parallel surfaces, where the width is much greater than the space
between the plates, the characteristic dimension is twice the distance between the plates.
There is no universal definition for a characteristic length. Any geometric length that significantly
simplifies the governing equations when converted to their non-dimensional form is considered the
characteristic length.
Mach Number
Mach number is the ratio of flow velocity to local speed of sound. It is a key parameter that
characterizes the compressibility effects in a fluid flow.
When the Mach number is small (less than 0.3) the inertial forces experienced by the flow can be
considered sufficiently small to not cause any changes to the fluid density. Under these conditions the
compressibility effects of the fluid can be ignored.
It is written as
The local speed of sound and hence the Mach number depend on the flow conditions, particularly on
temperature and pressure of the fluid.
The Mach number is used to classify the flow regime into subsonic, transonic, supersonic and
hypersonic flows and they are defined as:
Subsonic flow
Transonic flow
Supersonic flow
Hypersonic flow
Knudsen Number
Knudsen number is the ratio of the molecular mean free path length to the representative physical
length scale (characteristic length). It is a measure of rarefaction of the flow. Knudsen number
determines whether continuum mechanics or statistical mechanics formulation should be used to model
the flow.
It is expressed as
where
is the molecular mean free path understood as the mean distance travelled by a molecule
between two successive collisions with other molecules
is the characteristic length
Prandtl Number
Prandtl number is the ratio of molecular diffusivity to thermal diffusivity in a fluid. When the Prandtl
number is low it means that the heat diffuses quickly compared to momentum and vice versa.
It is written as
where
Nusselt Number
The average Nusselt number is the ratio of convective heat transfer across a chosen boundary or
surface to conductive heat transfer within a fluid. The convection heat transfer includes both advection
and diffusion heat transfer.
It is defined as
where
A Nusselt number close to one is characteristic of a laminar flow whereas a large value corresponds to
turbulent flow.
The local Nusslet number at a point at a distance x from the boundary is expressed as
Pressure Coefficient
The pressure coefficient is the ratio of pressure difference to the dynamic pressure. It describes the
relative pressure across a fluid field. It is expressed as
where
It is expressed as
The drag coefficient is the ratio of drag force to the dynamic force experienced by a body in a flow field.
It is expressed as
where
is the lift force experienced by the body
is the drag force experienced by the body
The reference area depends on the type of coefficient being measured. For automobiles and many bluff
bodies the reference area while calculating the drag coefficient is taken as the projected frontal area.
For airfoils the reference area is taken as the normal wing area.
The non-dimensional numbers used to determine similitude depend on the type of flow. For example,
in a case of incompressible flow (without free surface), the Reynolds number can satisfy the similarity
condition. If the flow is compressible Reynolds number, Mach number and specific heat ratio are
required in order to establish similarity.
It is not always possible to achieve absolute similitude between a model and prototype. It becomes
even harder as you diverge from the prototype’s operating conditions, therefore it becomes important
to focus on only the most important parameters.
The following table lists the most frequently encountered non-dimensional number, their description and
applications.
This section describes the simplification of the governing equations to various flow models by
assumptions on time dependence, density and viscosity of the fluid flow.
Depending on the flow conditions the Navier-Stokes equations, primarily the momentum equations, can
be simplified. The simplification of these equations depends on which effects in the flow are significant
or insignificant.
The simplified flow models that are most broadly used are the following:
• Steady flow
• Euler flow or Inviscid flow
• Stokes flow
• Incompressible flow
Steady Flow
The time dependence of flow field parameters is an important factor in the analysis of a fluid flow. A
majority of flows are not steady but transient in nature. In a steady state flow the flow properties at
a point, such as pressure and velocity, do not change with time. Steady state flows are of interest in
cases where the flow properties need to be studied after the flow field has stabilized.
This is achieved in simulations by taking a very large time step which causes the time derivative of
properties in the governing equations to reach to zero.
While performing simulations, a steady state flow result gives a preliminary insight into whether the
problem is set up correctly or not. If there are a large number of oscillations in the residuals it can be
inferred that the flow is transient and not steady. If the residuals show a smooth converge it implies
that the flow field becomes stable and steady state is achieved. The results can also be used to fine
tune the setup for further simulations or use them as initial conditions for a transient simulation.
Such flow situations can be identified in cases with a high value of Reynolds number, where the viscous
effects are concentrated to regions close to solid boundary and can be neglected for the regions far
away from the boundary.
These equations can form a closed form solution by assuming the equation of state.
Results obtained from these assumptions in the flow field are widely used in designing flying vehicles,
rockets and their engines, turbines and compressors. Studies of inviscid flows are carried out in gas
dynamics, acoustics, electro and magneto gas dynamics, the dynamics of rarefied gases, plasma
dynamics, and so on.
Strokes Flow
Stokes flow is a representation of fluid flow where the viscosity of the fluid is high. In such flows the
viscous effects dominate the advective inertial effects.
The Reynolds number in such flows is low, hence, it is also termed as creeping flow or Low Reynolds
number flow.
This results in the linearization of the governing equations and thus they can be solved by a number of
linear differential solvers available.
If the governing equations are non-dimensionalized and the Reynolds number is assumed to be very
low, the momentum equation reduces to
where
Incompressible Flow
All fluids (gas or liquid) exhibit some change in volume when subjected to compressive stresses. The
degree of compressibility for a fluid can be quantified using the bulk modulus of elasticity, E, defined as
where
A flow can be classified as incompressible if the density within fluid particle does not change during its
motion. It is also termed as isochoric flow and implies that under certain conditions a compressible fluid
can undergoincompressible flow.
Incompressibility is a property of flow and not of the fluid itself. Therefore the density field does not
need to be uniform for the flow to be incompressible. When the above equation is combined with the
continuity equation the following relation is obtained:
which states that in an incompressible flow the velocity field is solenoidal (having zero divergence).
When an incompressibility assumption is made it is important to know under what conditions this
assumption is valid. For a steady flow the condition is that the flow velocity must be much smaller
compared to the local speed of sound, that is,
In case of an unsteady flow there is an additional condition which needs to satisfied, stated as .
Physically this condition states the distance travelled by a sound wave in time must be much greater
than the distance travelled by the fluid particle. This implies that the propagation of pressure signals
(sound waves) is instantaneous compared to the interval over which the flow field changes significantly.
When the limit for maximum relative change in density is set to five percent as the criteria for an
incompressible flow, the maximum value of Mach number achieved is 0.3. This criteria states that any
flow with a Mach number less than 0.3 (without heat source) can be assumed to be incompressible.
One of the implications of assuming the flow to be incompressible is that there is no equation of state
as in a case of a compressible flow. In practice this means that the energy equation is decoupled from
the continuity and momentum equations, assuming fluid properties are not a function of temperature.
If the fluid properties change with temperature the equations again become coupled as in the case of a
compressible flow.
The pressure in such a flow is no longer a thermodynamic quantity and cannot be related to
temperature or density through an equation of state and must be obtained from the continuity and
momentum equation while satisfying zero divergence for the velocity field. In the continuity equation
there is no pressure term and in the momentum equation there are only the derivatives of pressure, but
not the pressure itself. This means that the actual value of pressure in an incompressible flow solution is
not important, only the changes of pressure in space are important.
This section covers concepts such as boundary layer type, flow, separation and transition.
When a fluid moves past an object or an object moves past a fluid, the fluid molecules right next to the
boundary surface stick to it. This causes the molecules just above the surface to slow down because of
collision with the molecules sticking to the surface. These molecules consequently slow down the flow
just above them. As we move farther away from the surface of the object, the collisions affected by the
object’s surface reduces. Thus a layer of fluid is created within which the velocity of the flow gradually
changes from zero at the object’s surface to the free stream velocity at a certain distance away from
the surface.
This distance at which the flow velocity inside the boundary essentially reaches the free stream velocity
is termed as the boundary layer thickness. Mathematically it is defined as the distance at which the flow
velocity is 99 percent of the free stream velocity.
As a result of the loss of velocity, a shear stress is imparted on the object surface defined as
The distance at which the difference between the flow temperatures inside the boundary layer and the
boundary surface essentially reaches the difference between free stream temperature and the boundary
surface temperature is termed as the thermal boundary layer thickness. It is mathematically expressed
as:
At y = , = 0.99
The local heat flux in a thermal boundary layer is analogous to the shear stress in a velocity boundary
layer. The heat flux at the surface is proportional to the temperature gradient. It is defined as
The flow inside the velocity boundary layer can be classified into two types
The shear stresses associated with this flow can be calculated using the molecular viscosity .
Figure 18 depicts the typical velocity profile for a laminar boundary layer.
In this case the scales of mixing cannot be handled by molecular viscosity alone, therefore turbulent or
eddy viscosity has to be modeled in order to proceed with the calculations.
Figure 19 depicts the velocity profile for the turbulent boundary layer.
The velocity profile for a turbulent boundary layer is quite different from a laminar boundary layer. It
comprises of three regions or layers:
• Outer layer: This layer is sensitive to the properties of the external flow.
• Inner layer: This layer has turbulent mixing as the dominant physics.
• Laminar or viscous sublayer: This layer is attached to the wall where the no slip condition is
applied. The shear stress in this layer is dominated by molecular viscosity.
These layers blend into each other and the blending region between the inner layer and laminar
sublayer is termed as the Buffer Zone.
As a consequence of intense mixing, a turbulent boundary layer has a much steeper gradient of velocity
resulting in a larger shear stress and higher skin friction drag at the wall compared to a laminar
boundary layer. In addition heat transfer rates are higher for a turbulent boundary layer. The presence
of eddies brings fluid particles at different temperatures into close contact with each other and the heat
transfer comes more effective.
Another consequence of mixing is that a turbulent boundary layer has higher momentum near the wall,
so it can withstand adverse pressure gradient longer and is less easily driven by the changes in the free
stream pressure compared to a laminar boundary layer. This property allows a turbulent boundary layer
to be attached to the object surface longer.
The pressure gradient is one of the parameters that greatly influences a flow. The shear stress
caused by viscosity, near an object surface, has a retarding effect on the flow. This retardation can
be overcome by a implementing a negative pressure gradient in the direction of the flow. A negative
pressure gradient is termed as a favourable pressure gradient because it enables the flow.
A positive pressure gradient, termed as adverse pressure gradient, has the opposite effect on the
flow. It further retards the flow leading to a reduction in the velocity near the wall and an increased
boundary layer thickness. This continuous retardation causes the shear stress at the wall to drop to
zero which then becomes negative causing a reversal in the flow direction near the surface. A region
of recirculating flow is developed with formation of large slowly rotating eddies resulting in the flow
detaching from the surface of the object. The point at which the velocity gradient and essentially the
shear stress reaches zero is termed as the point of separation.
The separation leads to the formation of a large wake of vortices behind the body, where the fluid
pressure is much lower compared to the regions where the fluid is attached to the surface since the
eddies cannot convert the rotational energy into pressure head. As described in the previous section
the flow velocity near the object surface is higher in case of a turbulent boundary layer. This higher
fluid momentum means that the fluid separation occurs further downstream than in case of a laminar
boundary layer when subject to the same adverse pressure gradient, resulting in a narrower wake and
consequently less pressure drag.
Figure 22: Boundary Layer Separation in a Diverging Channel (source: National committee for Fluid
Mechanics Films; http://web.mit.edu/hml/ncfmf.html)
Figure 22 shows the separation of a laminar (top) and turbulent boundary (bottom) layer in a diverging
channel.
Depending on the flow conditions, the recirculating flow may reattach to the surface of the body. A
number of factors could influence the reattachment. The pressure gradient may turn favourable due to
changing flow conditions or the geometry of the body. The other factor is that the boundary layer may
transition from laminar to turbulent resulting in a fuller velocity profile that could sustain the adverse
pressure gradient and causing the flow to reattach.
These factors along with many other sources of disturbance can cause a smooth laminar flow to break
down, enter a transition region and then develop into a turbulent flow. The mechanisms through which
a boundary layer transitions can be one of the following:
• Flow over an airfoil: In the aerodynamics industry, the boundary layer is particularly important
because it is responsible for a considerable amount of drag on the surface. The airfoils are
designed to be thin and streamlined in order to keep the boundary layer laminar, thus reducing
the skin friction drag. In situations where there are higher chances of flow separation, it is
always preferable that the boundary layer separates towards the leading edge of the airfoil and
reattaches somewhere downstream. It is dangerous if the separation occurs near the trailing
edge in which case the flow would not reattach, resulting in sudden loss of lift and stalling of the
aircraft. There are several approaches used to prevent this problem. These approaches include
turbulence generation by tripping the boundary layer using vortex generators, boundary layer
suction at the leading edge and reenergizing the boundary layer through blowing.
• Heat exchangers: A large number of production facilities in many industries use processes in
which heat transfer takes place within the fluid or between different fluids. In such cases it is
always preferable to increase the heat transfer and mixing without paying too high a penalty
in terms of increased pressure loss. Some of the approaches used for this purpose include use
of internal ribs or fins to increase the internal surface area of the domain, deformation in the
flow domain to decrease the hydraulic diameter and disruption of the boundary layer to induce
turbulence.
• CFD: The approximations obtained from the thin boundary layer theory greatly reduce the
complexity of the equations solved. Mathematically the application of boundary layer theory
changes the character of the governing Navier-Stokes equation from elliptic to parabolic. This
allows marching in the flow direction, as the solution at any location is independent of the
conditions farther downstream.
Aside from the topics discussed above, there are numerous cases where the boundary layer and the
physics associated with it have significant applications. A few examples are natural boundary layer
in the atmosphere, biometrics of marine organisms, automobile aerodynamics and design of sports
merchandise such as golf balls.
3.3 Turbulence
This section on turbulence covers the topics describing the physics of turbulence and turbulent flow. It
also covers the modelling of turbulence with brief descriptions of commonly used turbulence models.
This section on physics of turbulence introduces a brief history of turbulence and covers the theory
behind turbulence generation, turbulence transition and energy cascade in fluid flows.
On a daily basis you encounter turbulent flows in many places including rivers, ocean currents,
atmospheric turbulence, bush fires, flows over blunt bodies (for example, cars, aircraft, ships and
buildings), cross flows of chimney plumes, internal flows in pipelines, turbines and engines and many
other various forms. Despite the prevalence of turbulent flows in your daily life there is no consensus on
the definition of turbulence.
The following are several definitions quoted from various sources of literature.
• Bradshaw (1971): “Turbulence is a three dimensional time dependent motion in which vortex
stretching causes velocity fluctuations to spread to all wavelengths between a minimum
determined by viscous forces and a maximum determined by the boundary conditions. It is the
usual state of fluid motion except at low Reynolds numbers.”
• Hinze (1975): “Turbulent flow motion is an irregular condition of flow in which various quantities
show a random variation with time and space coordinates, so that statistically distinct average
values can be discerned.”
• Launder (1991): “At moderate Reynolds numbers the restraining effects of viscosity are too weak
to prevent small, random disturbances in a shear flow from amplifying. The disturbances grow,
become non linear and interact with neighboring disturbances. This mutual interaction leads
to a tangling of vorticity filaments. Eventually the flow reaches a chaotic, non repeating form
describable only in statistical terms. This is the turbulent flow.”
• Pope (2000): "An essential feature of turbulent flows is that the fluid velocity field varies
significantly and irregularly in both position and time."
Considering these definitions you can arrive at the following set of characteristic features of turbulent
flows:
• Irregularity. Turbulent flow is highly chaotic and irregular in both space and time (Tennekes and
Lumley, 1972; Hinze, 1975; Pope, 2000).
• Non-repetition. Turbulent flow is a non-repeating form (Launder, 1991).
• Three dimensional unsteady eddies with different scales. Turbulent flow is always three
dimensional and unsteady. Turbulent flow contains eddies with different scales, which extract
energy from shear stress of the mean flow and produce fluctuations in the velocity and pressure
fields. The eddies also interact with one another and exchange momentum and energy (Bradshaw,
1971; Tennekes and Lumley, 1972).
• Diffusivity. Turbulent flow is diffusive (that is, the spreading of fluctuations) (Tennekes and
Lumley, 1972; Pope, 2000). The chaotic eddies (vortices) within the turbulent flow increase the
momentum exchange in boundary layers delaying separation on the surface of curved bodies and
increasing the resistance (wall friction) and heat transfer.
• Dissipation. Turbulent flow is dissipative. This suggests that the kinetic energy within small eddies
is transformed into thermal energy. The small eddies receive their energy from larger eddies,
which receive their energy from even larger eddies, and so on. The largest eddies extract their
energy from the mean flow. This process of transferring energy from the larger eddies to the
smaller eddies is called the energy cascade process. Turbulence is not self sustainable without a
source of energy input.
• Amplification. The flow disturbances amplify when the Reynolds number exceeds critical Reynolds
numbers (~2300 for internal flows, ~500000 for external flows).
• Flow property. Turbulent flow is a property of flow, it is not a fluid property.
• Sensitivity. Turbulence is extremely sensitive to flow disturbances (Pope, 2000).
Leonardo da Vinci was the first one to report the visualization of turbulent flow in his famous sketches
from 1510. He described the existence of whirlpools of water in his notes. He wrote, “Observe the
motion of the surface of the water, which resembles that of hair, which has two motions, of which one
is caused by the weight of the hair, the other by the direction of the curls; thus the water has eddying
motions, one part of which is due to the principal current, the other to random and reverse motion.”
This notion is a precursor to the Reynolds flow decomposition of velocity into mean and fluctuating
parts, which Osborne Reynolds suggested 400 years later. He also noted, “The small eddies are almost
numberless, and large things are rotated only by large eddies and not by small ones and small things
are turned by both small eddies and large.” Nowadays, the “small eddies” and the “large things” refer
to turbulence and large-scale coherent eddies, respectively. Leonardo da Vinci utilized his sketches as a
flow visualization method, providing clear information of turbulence behaviors and their effects.
Figure 23: Leonardo da Vinci’s Sketches: Whirlpools of the Water (left), Vortices Behind Obstacles
(right): Arrows Indicate Flow Directions
In 1883, Osborne Reynolds conducted an experiment where dye was injected into a glass tube with
water flowing due to gravity. He studied the conditions of transition from laminar flow to turbulent
flow, indicated by the flow patterns in the tube. He analyzed that the flow patterns differed as the flow
velocity was changed. At low speed, the dye showed regular straight lines of a laminar flow pattern.
When the flow velocity increased, the dye became random (or chaotic) with eddies filling the entire
tube, and mixed with water. This random flow pattern is a turbulent flow. He further noticed that the
appearance of turbulent flow highly depended on a dimensionless parameter of velocity and length
scale, as well as viscosity. This parameter is now known as the Reynolds number.
The Reynolds number is not only used to characterize the flow patterns, such as laminar or turbulent
flow, but also to determine the dynamic similitude between two different flow cases. The latter is an
important concept to ensure valid experimental or computational results of numerous scaled models for
industrial applications.
The Reynolds number is the ratio of the inertial force to the viscous force, defined as
where is the fluid density (kg/m3), the mean flow velocity (m/s), the characteristic length (m), for
example, hydraulic diameter for internal flows, and the dynamic viscosity of the fluid (kg/(m-s).
The characteristic length in the Reynolds number can be anything convenient, as long as it is consistent,
especially when comparing different geometries. For example, the radius and the diameter are both
valid for spheres or circles in flows with the diameter primarily used by convention. When computing
the Reynolds number for airfoils and wings, the chord length is often chosen over the span because the
former is a representative function of the lift. For pipes, the diameter is the characteristic length. For
rectangular pipes a suitable choice is the hydraulic diameter defined as
where A is the cross-section area (m2) and P is the wetted perimeter (m). The wetted perimeter is the
total perimeter of subject walls in contact with the flow.
Fluid flows are laminar when their Reynolds number is below a certain critical value and they are
turbulent when they are larger than this critical value, termed the critical Reynolds number (Recr). This
critical Reynolds number is between 2,300 and 4,000 for pipe flows.
The critical Reynolds number is also referred to as the transition Reynolds number. It varies widely
depending on the conditions of surface roughness, flow disturbances, flow velocity and geometric
considerations. For example, the critical Reynolds number for the boundary layer flow over a flat
plate reaches about 500,000 based on the free stream velocity outside of the boundary layer and
characteristic length of the distance from the leading edge of the plate.
At low values of Reynolds number the viscous force is large compared to the inertial force. In this
range, viscous forcing dampens out disturbances in the flow field that are a result of surface roughness
or pressure gradients. As the Reynolds number increases, the viscous force becomes relatively smaller
and at some point it becomes possible for small perturbation to grow. The flows become unstable and
can transition to turbulence where large fluctuations in the velocity field continue to develop.
Figure 25 shows two different flow patterns in pipes for laminar flow and turbulent flow. It is evident
in the image that turbulent flow undergoes irregular flow patterns while laminar flow moves in smooth
layers while maintaining a constant flow direction. The turbulent flows happen at a high Reynolds
number where inertial forces are higher than viscous forces and perturbations can become amplified,
whereas the laminar flows occur at a low Reynolds number in which any induced perturbations are
damped out due to relatively strong viscous forces.
Figure 26 shows the time history of local velocity variations for laminar flows and turbulent flows,
respectively. These patterns can be obtained from hot wire anemometer measurements or CFD
simulations. For the pipe flow cases, laminar flows have nearly constant velocity, while turbulent flows
have random or chaotic velocity fluctuation patterns. For a cylinder in cross flow, laminar flows produce
a sine wave velocity pattern at a downstream location of the cylinder, while the turbulent flows have
similar wave patterns but with embedded fluctuations.
If you consider a flat plate immersed in a flow field with finite viscosity, a thin boundary layer will
begin to develop as a function of the distance traveled along the plate (x). The image below shows a
comparison of time-averaged velocity profiles for laminar flow and turbulent flow over a flat plate. The
time-averaged velocity profile in a turbulent flow appears more uniform than in a laminar flow because
the eddy motions in turbulent flow transport momentum more actively from one place to another. This
process results in a more uniform profile outside the boundary layer. The velocity gradient near the wall
is higher than the one seen in a laminar flow resulting in a larger skin friction coefficient ( ) than the
laminar flow. The skin friction coefficient can be defined as
where is the wall shear stress, is the dynamic viscosity, is the density and is the mean
velocity. The local Reynolds number ( ) is calculated by using the distance from the leading edge of
the flat plate as the length scale. measures the local ratio of inertial to viscous forcing and provides
an indication of the state of the flow regime as it moves from laminar to turbulent.
Figure 27: Time Averaged Velocity Profiles for Laminar and Turbulent Flows
Figure 28: Time Averaged Skin Friction Coefficient for Laminar Flow and Turbulent Flow
The information below includes equations of boundary layer thickness and skin friction coefficient for
turbulent flow and laminar flow. These equations where developed through empirical relationships
between local Reynolds number and boundary layer characteristics. The information below also provides
a summary of general characteristics that are present in turbulent flow and laminar flow.
Table 2: Laminar Flow vs Turbulent Flow: Boundary Layer Thickness and Skin Friction Coefficient
Friction drag (wall shear stress) Low (low velocity gradient near High (high velocity gradient near
walls) wall)
Separation Separation can occur at weak High momentum flow near wall
adverse pressure gradient delay flow separation
Transition Flow
In the real world, laminar flow and turbulent flow coexist when obstacles are located inside of a fluid
flow. Industrial examples of the transition from laminar to turbulent flow regimes can be found in many
different applications, as shown in Figure 29.
A transition from laminar flow to turbulent flow occurs due to various external factors such as
freestream turbulence level (vortical disturbances), sound waves (acoustic disturbances), temperature
fluctuations (entropy disturbances), streamwise pressure gradients, surface roughness, surface
curvature and vibration. Depending on the external factor or factors, the development of various
transition (instability) mechanisms cause the flow to trip from laminar to turbulent. Since transition
affects the flow development drag, heat transfer and many other factors it is important to understand
the transition phenomena properly. The following section overviews the three main transition processes:
natural transition, bypass transition and separation-induced transition.
Natural Transition
In the laminar flow regime viscous forces usually damp out the disturbances. However, when the free
stream turbulence is below one percent and the Reynolds number is higher than the critical Reynolds
number, viscous forces destabilize the shear layer such that it becomes unstable two dimensional
Tollmien-Schlichting (T-S) waves. These waves develop into three dimensional waves and hairpin
vortices, which eventually break down, elongate, and roll up. They then develop into triangular
turbulent spots past certain downstream locations. These turbulent spots grow by spreading sideways
and burst into a turbulent flow.
Although T-S waves are the main instability mechanisms of the natural transition process for two
dimensional boundary layers, such as flows over flat-plates, there are other kinds of instability
mechanisms for natural transition. These include Görtler vortices and the crossflow instability.
Görtler vortices generate on any surfaces with concave curvatures from the effects by centrifugal forces.
These are counter-rotating pairs of stationary streamwise vortices. Examples of industrial applications
with concave surfaces include supercritical airfoils designed for laminar flow control, high speed wind
tunnels, as well as forebody compression surfaces ahead of scramjet engine inlets. Details about these
vortices are discussed in Saric (1994).
The third instability mechanism is crossflow instability. Crossflow instability develops when the pressure
gradient over a swept wing combines with the cross flow from a wing root to a wing tip. It leads to the
formation of three dimensional streamwise, co-rotating vortices, inducing transition.
Having an understanding of these instability mechanisms and controlling them with flow controls have
received considerable attention over the past decade with the goal of delaying turbulent transition. This
is beneficial in reducing the drag of slim bodies, for example wings, by keeping laminar flows as long as
possible.
Bypass Transition
Another transition process is bypass transition. When the initial disturbances are high due to surface
roughness or freestream turbulence levels higher than one percent, turbulent spots are generated
without development of the three initial disturbances, T-S waves, spanwise vorticity and vortex
breakdown, observed in the natural transition. Although the bypass transition phenomena has been
investigated in experimental and computational studies it has not been very well understood. A common
industrial example of bypass transition is that of turbomachinery. The bypass transition occurs when
blades in a gas turbine engine encounter wakes from upstream vanes.
Separation-Induced Transition
The third kind of transition process is separation-induced transition. When a laminar flow experiences
adverse pressure gradients, such as airfoil suction surfaces and flow over a sphere, the fluid flow
detaches from the wall surface. If the disturbances are low, separation can cause the generation of
structures found in natural transition. Larger disturbances generate Kelvin-Helmholtz instabilities,
where vortices roll up before breaking down into turbulence. The process involved in separation-
induced transition depends on the size of the adverse pressure gradient and the presence of additional
disturbances, for example obstacles.
Many balls have some type of surface roughness, such as the seams on baseballs and the fuzz on tennis
balls. A rough surface can assist in reducing the drag by promoting an early transition.
A number of industrial applications take advantage of this observation for practical design. These
include vortex generators on various types of airfoils, such as wind turbine blades, wings and fences on
aircraft and automobiles, as well as wings enhanced with surface roughness.
Turbulent Wake
At the low Reynolds number, the downstream flow of the cylinder has a symmetric flow pattern and
produces near zero pressure drag. When flow reaches Re ~ 100,000 vortices begin to shed from
each side of the cylinder. This particular shedding pattern is called a von Karman vortex street. When
the Reynolds number is about 1,000,000 a turbulent wake begins to form behind the cylinder. It is
generated by the adverse pressure gradient at the peak position of the cylinder. This forms a large
separation region where vortical structures interact and decay downstream. As the Reynolds Number
increases the size of the turbulent wake decreases which in turn results in the reduction of the drag.
This mainly attributes to the turbulent boundary formation on the cylinder, delaying the flow separation
as it has a higher momentum near the cylinder than laminar flow.
As mentioned before, vortex shedding forms behind the cylinder when the Reynolds Number is between
47 and 100,000. This is referred to as a von Kaman vortex street and is caused by unsteady flow
separation around the cylinder (see Figure 30). Since vortices induce low-pressure zones, the cylinder
experiences lateral (sideways) forces. When the frequency of lateral forces is close to the natural
frequency of the cylinder, unwanted structural vibration will occur. In order to remedy this issue, spirals
are sometimes installed on objects subject to wind loading, such as a chimney.
Figure 30: Vortex Shedding Formation Downstream of a Cylinder (left) and Chimney with Spirals
(right)
Blunt bodies experience strong pressure drag compared to skin friction. This occurs due to the pressure
imbalance between the front cross-section of the blunt bodies and the downstream cross-section.
This pressure imbalance is attributed to the formation of downstream turbulent wakes, which reduces
the local pressure. The high-pressure drag is a determinant factor of the flying range of a ball. As an
example, it is common practice to make dimples on the surface of golf balls so that turbulent boundary
layers form on the surface of a ball. This results in delayed flow separation and consequently reduces
the size of the turbulent wake. Since the reduction of turbulent wake induces the smaller pressure drag,
the golf ball with dimples should fly for a longer range when compared to the smooth ball.
Compared to the sphere in an airfoil at a low angle-of-attack (AOA) produces a smaller wake size, which
results in the reduction of pressure drag. That is why the total drag is smaller for the airfoil than the
ball, although the friction (viscous) drag is higher on the airfoil. Consequently, designers focus on the
friction drag reduction by employing smooth surfaces. This is contrary to a ball with dimples for long
flying range. However, since the turbulent wake size increases for high AOA, designers will shift their
focus on reducing wake size by controlling flow on the airfoil suction side where it separates. Several
methods are available, including a leading-edge extension, a vortex generator and a flow injection
method.
Navier-Stokes Equation
The physics of turbulent flows have been discussed by presenting experimental observations and
comparing it to laminar flows. In this chapter, the focus will shift to the governing equations of these
flow fields.
The Navier-Stokes (NS) equations were named after Claude-Louis Navier and George Gabriel Stokes.
These equations govern the flow motion of a fluid and usually consist of the mass conservation equation
and the momentum conservation equation. Sometimes, the momentum conservation equations alone
represent the Navier-Stokes equations. The NS equations govern the behavior of a viscous fluid by
balancing the forces with Newton’s second law and assuming that the stress in the fluid is the sum of
a diffusing viscous term. In the case of incompressible Newtonian fluid the instantaneous continuity
equation is the mass conservation equation, which reads in vector notation as:
where is the fluid density, is the pressure and is the viscous stress tensor. In tensor notation, the
incompressible momentum equation is given by
The left hand side of the equation describes acceleration, which includes the unsteady term and
convective term. The right hand side of the equation represents the summation of pressure and the
shear-stress divergence terms. The convective term is non-linear due to an acceleration associated with
the change in velocity over position. This term can be disregarded in one-dimensional flow and Stokes
flow (or creeping flow). For the Newtonian fluid, the viscous shear stress is assumed to be proportional
to the shear strain rate. Thus, viscous stresses can be obtained by
In vector notation,
In tensor notation,
Since turbulent flows contain eddies that are rotational, a transport equation of local rotation (vorticity)
of the fluid is briefly discussed.
The vorticity vector, , is the curl of the velocity vector and is defined as follows:
or
The vorticity transport equation can be obtained by taking the curl of the momentum conservation
equation:
After substituting above terms, the vorticity transport equation can be obtained as:
The vortex stretching term, , on the right-hand side of the equation corresponds to the
enhancement of vorticity when a fluid element is stretched <0. In other words, when the cross
section of the fluid element is decreased, the vorticity is increased. The image below shows the concept
of vortex stretching. In the image, there are two cylindrical fluid elements within the streamwise fluid
flow. The concept shows that compared to a thick element a thin element has stronger vorticity due to
the angular momentum conservation.
In two dimensional flow, the vortex stretching term vanishes since , and .
Therefore, the vortex stretching term is essential to the energy cascade for three dimensional turbulent
flow.
Turbulence is composed of turbulent eddies of different sizes. At high Reynolds numbers, a scale
separation exists between the largest eddies and smallest eddies.
The largest eddies extract kinetic energy from the mean flow as the energy production. The length
scale is comparable to the flow dimensions. These processes are highly anisotropic and mostly are not
influenced by viscosity. Most transport and mixing happen in this range.
The smallest eddies have universal characters independent of the flow geometry and conditions. Those
eddies in this range usually receive energy from the larger eddies and dissipate their energy into
heat through the fluid’s molecular viscosity. These eddies are isotropic with length scales as described
by Kolmogorov scales. It is assumed that the small scale eddies are determined by viscosity and
dissipation.
The intermediate eddies are called Taylor length scale eddies with the length scales in the inertial
subrange. Compared to the dissipation range, the turbulence in this region is also isotropic but
the eddies in this region are independent of both the largest eddies and the smallest eddies in the
dissipation range. It is assumed that the eddies in this region can be characterized by the turbulent
dissipation and the wave number (or eddy size l). Through dimensional analysis you obtain
If the flow is fully turbulent flow at a high Reynolds number, the energy spectra should exhibit a -5/3
decay in the inertia region. This is called Kolmogorov spectrum law or the -5/3 law.
Richardson (1922) introduced the energy cascade concept, “Big whorls have little whorls that feed on
their velocity; And little whorls have lesser and so on to viscosity.”
The largest eddies with integral length scale are unstable and break up, transferring kinetic energy
to smaller eddies. A similar break-up process with these smaller eddies transfer their energy to
subsequent smaller eddies. These processes continue until the eddy size is reduced to the smallest size
and the eddy motion is stable. This process is known as the energy cascade, as shown in Figure 32.
The image above illustrates a typical energy spectrum (E) of turbulent eddies in wave number space
(κ). Since the turbulent flow has a wide range of eddies with different length scales, it is convenient to
utilize the energy spectrum of the velocity field for the classification of eddies into three representative
length scales, the integral length scale, the Taylor microscale and the Kolmogorov length scale.
The integral length scale is the largest eddy size of the energy spectrum. These are the most energetic
and highly anisotropic eddies that produce energy via the interaction with the mean flow. Eddies in the
integral range (A) are very sensitive to flow conditions and their sizes are close to the characteristic
length of the flow (for example hydraulic diameter). Its spectrum is defined as . The integral
The Taylor microscale is an intermediate eddy size of the energy spectrum between the largest eddy
and the smallest eddy where the flow is inviscid. In this inertial subrange (B), the kinetic energy is
transferred from large turbulent eddies to the small eddies in what is defined as the energy cascade.
The occurrence of the energy cascade is related to the process of the vortex stretching. The vortex
stretching causes the rotational rate of the turbulent eddies to increase and the radius of their cross
sections to decrease. The energy spectrum of the inertial subrange is obtained through dimensional
analysis and shows that the energy spectrum is proportional to the product of the wave number and the
dissipation rate .
The Kolmogorov scale is the smallest eddy size and is in the dissipative range (C) of the energy
spectrum. Turbulent eddies in the dissipative range are isotropic because the diffusive actions of strong
viscosity smear out anisotropic characteristics of the larger eddies. In this region the kinetic energy
received from larger eddies is dissipated into heat. The length scale of this region is assumed be a
function of the kinematic viscosity ( ) and the turbulent dissipation rate per unit mass ( ). Using
dimensional analysis you obtain
•
The Kolmogorov length scale
•
The Kolmogorov time scale
•
The Kolmogorov velocity scale
• The energy spectrum .
The ratio of the integral length scale to the Kolmogorov length scale
= .
Corresponding time scale and velocity scale ratios are, respectively, given as and
.
Considering these scale ratios, you can see larger scale separation between the integral length scale
and the Kolmogorov length scale with an increase of the turbulent Reynolds number. This suggests that
the Kolmogorov length scale is much smaller than the integral length scale for high turbulent Reynolds
number.
References
Bradshaw, P., 1971, “An Introduction to Turbulence and Its Measurement,” Pergamon Press, Oxford,
England.
Hinze, J. O., 1975, “Turbulence,” 2nd Edition, McGraw-Hill, New York, USA.
Launder, B.E., 1991, “An Introduction to the Modeling of Turbulence,’ VKI Lecture Series 1991-02.
Richardson, L. F., 1922, “Weather Prediction by Numerical Process,” University Press, Cambridge, USA.
Saric, W.S., 1994, “Görtler Vortices,” Annual Review of Fluid Mechanics, Vol. 26, pp. 379-409.
Tennekes, H. and Lumley, J.L., 1972, “A First Course in Turbulence,” The MIT Press, Cambridge, USA.
van Dyke, M., 1982, “An Album of Fluid Motion,” Stanford University, California, USA.
White, F.M., 1991, “Viscous Fluid Flow,” McGraw-Hill, New York, USA.
This section covers the numerical modeling of turbulence by various turbulence models, near wall
modeling and inlet turbulence parameters specified for turbulence models.
Although turbulence has been researched for centuries it continues to pose some of the most difficult
and fundamental problems in physics because it is a complex, nonlinear phenomenon.
Considering these turbulence characteristics, resolving the turbulent quantities directly by means
of Direct Numerical Simulation (DNS) is impractical due to the limitations of currently available
computational resources. For this reason, the preferred approach is not to resolve turbulence but to
model it with consideration of the following observations:
In this section representative turbulent flow cases are provided to explain why turbulent flow
simulations could be challenging. Resource requirements for DNS are examined in order to justify
the need of turbulence modeling. After a brief comparison of resource requirements for various
turbulence models the Reynolds-averaged Navier-Stokes equations (RANS) are introduced and
turbulence models are presented to close the RANS equations. The focus then shifts to Large Eddy
Simulation (LES).
Overall, this discussion is primarily aimed at motivating the need for turbulence models when
investigating high Reynolds number turbulent flow applications and to introduce commonly
available turbulence models to new users.
In order to set the context for the modeling of turbulent flow it is essential to understand why turbulent
flow simulations are challenging.
First, examine the turbulence characteristic non-repetition. Figure 34 shows two snapshots of smoke
plumes from a burning incense stick. Both instantaneous plumes show the chaotic, continuously
evolving movement of the smoke. Due to the sensitivity of the smoke plumes to surrounding conditions
(temperature, density and cross flow velocity), the pattern of their movement never repeats. Thus,
prediction on such instantaneous turbulent smoke evolutions is understood to be quite challenging.
The Navier-Stokes (NS) equations are the set of equations that govern the motion of a fluid. These
equations consist of the mass conservation equation and the momentum conservation equations for
incompressible flows. The NS equations govern the behavior of a viscous fluid by balancing the forces
with Newton’s second law and assuming that the stress in the fluid is the sum of a diffusing viscous
term and a pressure term. In the case of a Newtonian fluid, the instantaneous continuity equation is
defined as
where is the pressure, is the viscous stress tensor and is the Kronecker delta.
The left hand side of the equation describes advection, which includes two terms, the acceleration term
and convective term. The right hand side of the equation represents the summation of pressure and the
shear-stress divergence terms. The convective term is non linear due to an acceleration associated with
the change in velocity as a function of position. This term can be disregarded in one-dimensional flow
and Stokes flow (or creeping flow).
For Newtonian fluid the viscous shear stress is assumed to be proportional to the shear strain rate,
following Stokes Law. Thus, viscous stresses can be given by
Now, you have a full set of the Navier-Stokes equations to govern fluid flows. The next step is how
to solve these equations. There are two approaches available: analytical approaches and numerical
approaches. Since most flows involve convective accelerations, it is difficult to obtain the analytical
solutions of the Navier-Stokes equations due to the non-linearity associated with the convective term
and coupled nature of the equations. Although analytical solutions do exist for some simple flow cases
with certain approximations, they are mostly limited to laminar flows. These are Couette flow and
Hangen-Poiseuille flow. The former is a laminar flow between two parallel plates (one of which is
moving relative to the others), and the latter is a laminar flow in a long cylindrical pipe of constant
cross section. Therefore, a complete description of a turbulent flow can only be obtained by numerically
solving the Navier-Stokes equations (Moin and Mahesh, 1998). These are divided into a numerical
approach resolving turbulent flows by Direct Numerical Simulation (DNS) and numerical approaches not
resolving turbulence but modeling it.
Direct Numerical Simulation (DNS) solves the time dependent Navier-Stokes equations, resolving from
the largest length scale of a computational domain size to the smallest length scale of turbulence eddy
(Kolmogorov length scale). Considering a vast range of length scales within the computational domain,
you can rightly claim that huge computer resource would be required for DNS. Examining the computer
resource requirements for DNS will support this claim.
For a computational square box domain, the number of grid points ( ) in one direction depends on the
grid spacing .
In order to resolve the Kolmogorov length scale ( ), the grid spacing and the Kolmogorov length
scale should have the same order of magnitude.
Since the Kolmogorov length scale is a function of the fluid kinematic viscosity ( ) and the turbulent
dissipation rate ( ), the number of grid points in one direction can be estimated from
For three dimensional flows, the total grid size for DNS can be computed as
For a flat plate, turbulent flow occurs when the Re > 500,000 (Schlichting and Gersten, 2000). The
above relation shows the estimation of six trillion nodes, which easily exceeds the capacity of even the
most advanced high performance computers.
In addition to the demanding grid size requirements of DNS, the computing time step size ( )
should be small enough to resolve the Kolmogorov time scales. Since the Kolmogorov time scale
depends on the Kolmogorov length scale and the characteristic velocity ( ), the time step size can be
approximated as
Given this, the number of time steps ( ) can be estimated from the total duration of the simulation (
) and the time step size ( )
Utilizing the equation of the number of the time steps can be estimated as
Finally, the number of floating point operations needed to perform DNS can be computed from the
multiplication of total grid size ( ) and the number of the time steps
Therefore, the computational cost for DNS is very expensive, confirming that DNS is not feasible for
high Reynolds Number turbulent flows.
In addition, high order (third order or higher) numerical schemes are commonly used in order to reduce
the numerical dissipation and to keep the problem size tractable. These include spectral methods or
spectral element methods. Although these methods are very efficient in resolving the small scales of
turbulence, they require that the computational domain be relatively simple. The direct result is that
these high-order schemes have little flexibility in dealing with complex industrial geometries because of
the structured (blocked) mesh approach. Finally, DNS requires special treatments for realistic initial and
boundary conditions
Considering these observations it can be concluded that DNS attempting to resolve all turbulent length
and time scales is restricted to the low Reynolds number range and is impractical for industrial flows
due to huge computing resource requirements. Most DNS applications are served as benchmark
databases for tuning turbulence models and have been used for fundamental turbulent flow studies,
including homogeneous turbulent flows with mean strain, free shear layers, fully developed channel
flows, jets and so on.
Turbulence Modeling
Three dimensional industrial scale problems are concerned with the time averaged (mean) flow, not the
instantaneous motion. The preferred approach is to model turbulence using simplifying approximations,
and not resolve it. Turbulence modeling is a procedure to solve a modified set of the Navier-Stokes
equations by means of developing a mathematical model of the turbulent flow that represents the time-
averaged characteristics of the flow. Turbulence modeling is used to compute the impact of eddies
on the mean flow field. This approach is based on the assumption that the turbulent eddy motion is
“universal” and can be related to the large-scale average motion.
Over the course of the past few decades, turbulence models of various complexities have been
developed. Depending on the simplifications made to the Navier-Stokes equations, the turbulence
models can be classified as shown below.
Hybrid
Hybrid simulations are the bridge between LES and RANS by utilizing RANS for attached boundary
layers and LES for separated flow regions. In general, hybrid simulations need spatial filtering
processes to determine the local sub grid turbulent viscosity. Compared to LES and DNS, hybrid
simulations are much more tractable since the numerical requirement is less severe than the
other two approaches.
Common types include:
• Detached Eddy Simulation (DES)
• Delayed DES (DDES)
• Improved DDES (IDDES)
• Scale Adaptive Simulation (SAS)
• Wall Modeled LES (WMES)
• Zonal LES
Figure 35 shows turbulence models and their corresponding energy spectrum ranges for modeling. For
example, RANS models rely on their transport equations to model the entire wave number range, while
LES needs a subgrid model to model behaviors of eddies in the dissipative range, but explicitly resolves
the large eddies. The hybrid RANS/LES approach needs a model to cover the inertial range and the
dissipative range.
Figure 35: Turbulent Energy Spectrum Modeled by Various Turbulent Flow Simulation Approaches
As expected, computing requirements for these models differ since their models cover a different
turbulent spectrum for the range of wave numbers. Spalart (2000) summarized resource requirements
for each model, as shown below. Strictly speaking, DNS is not a turbulence model since it is resolving
all scales of motion, however, it is included for comparison purposes. Grid numbers and time steps were
estimated for a clean wing (Spalart, 2000). Grid numbers and time steps increase from RANS to DNS.
For incompressible turbulent flow the instantaneous velocity field can be decomposed into a
time averaged velocity and its corresponding fluctuation. This is commonly called the Reynolds’
decomposition, which is shown in Figure 36.
where
• : instantaneous velocity,
•
• velocity fluctuation.
Similarly, instantaneous pressure can be decomposed into the time averaged and fluctuation terms.
Once this concept is substituted into the instantaneous Navier-Stokes equations and then time
averaging is performed the RANS equations are obtained. The following equations are, respectively, the
Reynolds-averaged continuity and Reynolds-averaged momentum equations:
where
•
is the mean strain rate tensor,
It is noted that the momentum equation has an unsteady term (the first term in the left hand side of
the momentum equation). Wegner et al. (2004) argued that the RANS equations with the time term
(unsteady RANS) can be applied to simulate unsteady flows when a time scale (T2) for large scale
motions is larger than a time averaging period for RANS (T1) (see Figure 37). It is worth mentioning
that RANS time averaging period (T1) should be considerably larger than the largest turbulence
(integral) time scale, as URANS can only give a time averaged mean value for the velocity field, not
solving turbulence. Thus, a representative URANS solution is demonstrated by the smoothed red line in
the image below. Whereas the blue line represents the velocity profile with instantaneous fluctuations.
Steady RANS can be obtained by removing the time term when there is no large scale unsteadiness
observed in turbulent flow.
Figure 37: Time Scales for T2 for Large Unsteady Oscillation vs T1 for Turbulent Time Scale
Although Reynolds-averaging eliminates a need to compute the instantaneous flow field, it introduces a
new unknown term in comparison to the Navier-Stokes equations. This unknown term is referred to as
the Reynolds stress tensor, which represents an added stress due to the turbulent motions. This term
arises from the decomposition of the convective term using the Reynolds decomposition. Since your
goal is to eliminate any dependence on the instantaneous flow field in these simplified equations, this
term should be modeled in terms of the mean flow. This challenge is known as the closure problem of
turbulence.
One method that is available for computing the Reynolds Stresses that appear in the RANS equations
is to use the Boussinesq approximation. This approach assumes a linear relationship between the
turbulent Reynolds stresses and the mean rate of strain tensor:
where
•
: the turbulent kinetic energy per unit mass,
The only unknown in this equation is the eddy viscosity, which must be determined by the turbulence
model. There are hundreds of available turbulence models that have been developed to provide this
eddy viscosity. Some common Boussinesq models include
This approach is valid for many turbulent flows (for example, boundary layer, channel flow, round
jets, turbulent shear flow and mixing layer) but not for all (for example, impinging flow and swirling
flow). Under these circumstances, a model that is capable of predicting anisotropic Reynolds stresses
is more appropriate. This can include a full Reynolds stress model (RSM), which determines the
turbulent stresses by solving a transport equation for each stress component, or an eddy viscosity
model that uses a non linear stress strain relationship to compute each Reynolds stress component
separately. Although the improvement in accuracy achieved from this type of modeling is appealing, the
introduction of anisotropic stresses is often accompanied by increased compute cost and a decrease in
robustness.
The following section provides a review of some commonly used RANS models, ranging from algebraic
models (zero equation models) up to the seven equation Reynolds Stress Model.
where is a turbulence model variable, is the eddy viscosity, is the material viscosity and is a
constant.
The left hand side of the equation describes advection, which includes two terms, the acceleration
term and convective term. The right hand side of the equation represents the summation of the
diffusion, production and dissipation terms. The unsteady term represents the time dependence of
turbulence model variables, while the convective term accounts for the rate of change of variables
due to convection by the mean flow. The diffusion term on the right side of the equation describes the
transport of turbulent variables due to the summation of material viscosity and eddy viscosity. The
production term indicates the production rate of turbulent variables from the mean flow gradient, while
the dissipation term represents the dissipation rate of turbulent variables due to viscous stresses.
Mixing-Length Model
The Prandtl mixing-length model for Reynolds-averaged Navier-Stokes (RANS) is based on the
assumption that turbulent eddies can be expressed in terms of a product of a velocity scale and a
length scale. It has the advantages of fast computation and relatively accurate prediction for simple two
dimensional flows.
However, there exists a limitation in predicting turbulent flows where the turbulent length scale varies
(for example, separation or circulation). Currently, the applications of mixing-length models are hardly
adopted in commercial CFD codes, but are usually employed to describe near wall behaviors such as
wall boundary treatments for more complex turbulence models.
One equation Reynolds-averaged Navier-Stokes (RANS) models solve a single scalar transport equation
to compute the eddy viscosity. Common one equation models include the Spalart-Allmaras (SA) model
and the Nut-92 model. The SA model is discussed in this manual due to its application in general
purpose CFD codes and its popularity for the simulation of external flows and internal flows. The details
of the Nut-92 model can be found in Shur et al. (1995).
The SA model uses a transport equation to solve for a modified kinematic eddy viscosity, , as a
function of the kinematic eddy viscosity ( ) (Spalart and Allmaras, 1992). In this model, a length
scale (d) in a dissipation term of the modified kinematic eddy viscosity transport equation is specified to
determine the dissipation rate. This model has an advantage of having economic solutions for attached
flows and moderately separated flows, but it is not recommended for massively separated flows, free
shear flows and decaying turbulence.
Transport Equations
where and are constants and is the fluid dynamic viscosity. P and D are the production term and
destruction term of the modified turbulent viscosity, respectively.
Production of
where
•
,
•
is the rotation tensor,
•
,
Destruction of
where
• is a constant.
where
•
is the viscous damping function.
•
.
Model Coefficients
, , , , , = , .
The effects of system rotation and streamline curvature are present in turbomachinery components.
Some examples include axial turbines, radial turbines, axial fans, compressors and centrifugal impellors.
This is where most linear eddy viscosity models fail. In order to incorporate the rotational and curvature
effects for the SA model, Shur et al. (2000) introduced a version with the modified production term of
the transport equation by multiplying the rotation function .
Modified Production of
where
•
is the strain rate magnitude.
•
is the strain rate tensor.
•
is the modified vorticity magnitude.
•
is the rotation tensor.
Model Coefficients
=1.0, =12.0, =1.0
Two equation turbulence models are perhaps the most commonly used in industrial flow applications.
Although there are many two equation models available, some of the more popular ones include
Launder and Spalding (1974) proposed the standard k-ε turbulence model utilizing the relationships
described below. The model has been shown to be relatively accurate for high Reynolds number flows
in which the turbulence behavior is close to homogeneous, and the turbulence production is nearly
balanced by dissipation. Nevertheless, its performance deteriorates when predicting boundary layers
under adverse pressure gradients (Bradshaw, 1997). It also has difficulty in predicting the viscous
sublayer. To resolve this issue, it is suggested that a correction be made to reproduce the law of the
wall for incompressible flat-plate boundary layers (Wilcox, 2000). For low Reynolds number flows,
the difference between the turbulent kinetic energy production and the dissipation rates may depart
from their equilibrium value of zero, thus, ad-hoc adjustments of empirical parameters are inevitable.
Furthermore, the standard k-ε turbulence model does not perform well in shear layers and jets, where
the turbulent kinetic energy is not balanced with the dissipation rates (Versteeg and Malalasekera,
2007). Finally, this model is not recommended for high swirling/curvature flows, diverging passage
flows as well as flows with a body force under the influence of a rotating reference frame.
Transport Equations
Turbulent Kinetic Energy k
Production Modeling
Turbulent Kinetic Energy k
where
•
is the strain rate magnitude.
• is dynamic eddy viscosity.
Dissipation Modeling
Turbulent Kinetic Energy k
Model Coefficients
=1.44, =1.92, =0.09, =1.0, =1.3.
The RNG k-ε turbulence model (Yakhot and Orszag, 1986) deduces the behavior of large scale eddies
from that of the smaller ones by utilizing the scale similarity properties that are inherent in the energy
cascade (Bradshaw, 1997). This model employs a modified coefficient in the dissipation rate equation
to account for the interaction between the turbulent dissipation and mean shear. It results in better
prediction of flows containing high streamline curvature, flows over a backward facing step (Yakhot
et al, 1992), and flows in an expansion duct than the standard k-ε turbulence model. However, its
performance worsens when predicting flows in a contraction duct (Hanjalic, 2004).
Transport Equations
Turbulent Kinetic Energy k
Production Modeling
Turbulent Kinetic Energy k
Dissipation Modeling
Turbulent Kinetic Energy k
Model Coefficients
=1.44, =1.92, =0.09, =1.0, =1.3,
The standard k-ε turbulence model and RNG k-ε turbulence model do not satisfy mathematical
constraints on the Reynolds stresses for the consistency with physics of turbulence. These constraints
include the positivity of normal stresses ( ) and Schwarz's inequality (
). The realizable k-ε model proposed by Shih et al. (1995) employs a formulation of turbulent viscosity
with a variation of to satisfy these constraints. This is why this model is referred to as realizable.
The second improvement made with the realizable model is to have a source term in the dissipation
rate transport equation, which is derived from the transport equation of the mean square vorticity
fluctuation. Compared to the standard k-ε turbulence model, it performs better when simulating planar
jet flows, round jet flows and flows under adverse pressure gradients. Although the realizable model
has significant performance improvement over the standard model, it has a problem in predicting
impinging flow, swirling flow and secondary flows in a square duct since this model is based on the
Boussinesq assumption. In addition, this model needs additional functions to simulate the near wall
effects.
Transport Equations
Turbulent Kinetic Energy k
Production Modeling
Turbulent Kinetic Energy k
Dissipation Modeling
Turbulent Kinetic Energy k
Model Coefficients
= 1.44, = 1.9, = 1.0, = 1.22.
where
= , , , , ,
, ,
Since all three k-ε turbulence models cannot be integrated all the way to walls, wall damping wall
functions must be employed to provide correct near wall behavior. It is also known that the standard k-ε
turbulence model fails to predict the flow separation under adverse pressure gradients.
Wilcox proposed a turbulence model similar to the standard k-ε turbulence model but replaced the
dissipation rate (ε) equation with the eddy frequency (ω) equation (Wilcox, 2006; Wilcox, 2008). The
eddy frequency (ω) is often referred to the specific dissipation rate and is defined as . The
Wilcox k-ω turbulence model has an advantage over the k-ε turbulence model as the k-ω model does
not require any wall functions for the calculation of the velocity distribution near walls. As a result, the
k-ω turbulence model has better performance for flows with adverse pressure gradient when compared
to the k-ε turbulence models. However, the k-ω model exhibits a strong sensitivity to the freestream
boundary condition (Wilcox, 2006) for external flow applications.
Transport Equations
Turbulent Kinetic Energy k
Production Modeling
Turbulent Kinetic Energy k
Eddy Frequency ω
where
, , , , , ,
Dissipation Modeling
Turbulent Kinetic Energy (k)
where
, , ,
Model Coefficients
Menter (1994) suggested the SST model to overcome the freestream value sensitivity of the standard
k-ω turbulence model by transforming the k-ε model into the k-ω model in the near-wall region,
and by utilizing the k-ε model in the turbulent region far from the wall. The SST model employs a
modified source term in the eddy frequency equation. Among the different SST versions, the SST 2003
model (Menter, 2003) will be focused on in this section as it has been shown to be more accurate
when predicting flows near stagnation zones by introducing a production limiter to constrain the
kinetic energy production. This version also employs the strain rate magnitude in the definition of
eddy viscosity rather than the vorticity magnitude employed in the standard SST model (1994). When
compared to the k-ε model, the SST 2003 model achieves better accuracy for attached boundary
layers and flow separation. The SST 2003 model also overcomes the freestream sensitivity of the k-ω
turbulence model. However, this model shares a similar range of weakness with the k-ε equation models
for the predictions of flows with strong streamline curvature and/or rotation, as well as flows under
extra strains and body forces.
Transport Equations
Turbulent Kinetic Energy k
where
•
The blending function ,
•
,
Production Modeling
Turbulent Kinetic Energy k
where
• Production
•
Strain rate magnitude
•
Strain rate tensor
• Constant
where
Dissipation Modeling
Turbulent Kinetic Energy k
Eddy Frequency ω
where
•
The second blending function ,
Model Coefficients
= 0.09.
, ,
, ,
, .
Since the SST model relies on the Boussineq approximation, it also has poor performance for the
prediction of flows with streamline curvature and system rotation because the Reynolds stress tensor
is aligned to the mean strain rate tensor. In order to remedy this problem, the rotation curvature
correction for the SST 2003 model modifies the production terms in both the kinetic energy equation
and eddy frequency equation (Smirnov and Menter, 2009). The corrected model yields better
performance over the SST base model when predicting flows with strong streamline curvature and
rotating flows caused by strong swirl or geometric constraints.
Production Modeling
Turbulent Kinetic Energy k
where
•
=
•
=
Eddy Frequency ω
Model Coefficients
= 0.09, = 1.0, = 2.0, = 1.0.
v2-f Model
In order to account for the near wall turbulence anisotropy and non local pressure strain effects, Durbin
(1995) introduced a velocity scale v2 and the elliptic relaxation function f to the standard k-ε turbulence
model. The velocity scale v2 represents the velocity fluctuation normal to the streamline and represents
a proper scaling of the turbulence damping near the wall, while the elliptic relaxation function f is used
to model the anisotropic wall effects. Compared to the k-ε turbulence models, the v2-f model produces
more accurate predictions of wall-bounded flows dominated by separation but suffers from numerical
stability issues.
Transport Equations
Turbulent Kinetic Energy k
2
Velocity Scale v
where
•
: the length scale,
•
: the time scale.
Production Modeling
Turbulent Kinetic Energy k
2
Velocity Scale v
Dissipation Modeling
Turbulent Kinetic Energy k
2
Velocity Scale v
Model Coefficients
= 1.44, = 1.92, = 0.22, = 1.0, = 1.3.
Zeta-F Model
The base model of the zeta-f model is the v2f model described by Durbin (1995). However, by
introducing a normalizing velocity scale, the numerical stability issues found in the v2f model have been
improved (Hanjalic et al., 2004; Laurence et al., 2004; Popovac and Hanjalic, 2007).
Transport Equations
Turbulent Kinetic Energy k
where
Production Modeling
Turbulent Kinetic Energy k
Velocity Scale
Dissipation Modeling
Turbulent Kinetic Energy k
Velocity Scale
Model Coefficients
= 1.44, = 1.92, = 0.22, = 1.0, = 1.3.
The Reynolds stress model (RSM) determines the turbulent stresses by solving a transport equation
for each stress component. The RSM accounts for the effects of flow history and streamline curvature,
as well as system rotation and stratification (Wilcox, 2000). Reynolds stress models are known to
give superior results over one and two equation models when dealing with flows with streamline
curvature, flows with sudden change in strain rate, and flows with secondary motions, all at the cost
of an increased computing time (Bradshaw, 1997). There are many types of Reynolds stress models,
the two most common being the RSM based on the dissipation rate (ε) and the RSM based on the eddy
frequency (ω). In this section the dissipation rate (ε) model is discussed.
Transport Equations
Reynolds Stresses
Production Modeling
Reynolds Stresses
Dissipation Modeling
Reynolds Stresses
Pressure-Strain Modeling
Reynolds Stresses
where
where is the component of the unit normal to the wall and is the normal distance from the wall.
Model Coefficients
= 1.44, = 1.92, = 0.09, = 1.0, = 1.0,
All of the previously described models are incapable of predicting boundary layer transition. To include
the effects of transition additional equations are necessary. Transitional flows can be found in many
industrial application cases including gas turbine blades, airplane wings and wind turbines. It is known
that conventional turbulence models over predict the wall shear stress for transitional flows. Thus,
transition models can be used to improve the accuracy of CFD solutions when flows encounter turbulent
transition of the boundary layer.
γ-Reθ Model
Langtry and Menter (2009) proposed one of the most commonly used transition models in industrial
CFD applications. The model is a correlation based intermittency model that predicts natural,
bypass and separation induced transition mechanisms. The model is coupled with the Shear
Stress Transport (SST) turbulence model and requires two additional transport equations for the
turbulence intermittency ( ) and transition momentum thickness Reynolds number ( ). The
turbulence intermittency is a measure of the flow regime and is defined as
where represents the time that the flow is turbulent at a given location, while represents
the time that the flow is laminar.
For example, while the intermittency value is zero, the flow is considered to be laminar. A value of
one is for fully turbulent flow. The transition momentum thickness Reynolds number is responsible for
capturing the non local effects of turbulence intensity and is defined as
To couple with the SST model, Langtry and Menter (2009) modified the production term and dissipation
term of the turbulent kinetic energy to account for the changes in the flow intermittency.
Transport Equations
Turbulent Kinetic Energy k
Turbulent Intermittency
Production Modeling
Turbulent Kinetic Energy k
where
• Production
•
Strain rate magnitude
•
Strain rate tensor
• Constant
where
Turbulent Intermittency
where
where
, , , ,
Dissipation Modeling
Turbulent Kinetic Energy k
Turbulent Intermittency
where
where
•
The second blending function ,
Correlations
Model Coefficients
= 0.09.
= 0.5, = 0.856,
= 0.85, = 1.00,
, = 0.0828.
LES is often regarded as an impractical tool for industrial CFD applications as it requires large
computational resources. LES is associated with the severe requirements of small mesh size and time
step size so that the turbulent length scale and time scale in the inertial range is properly resolved.
These simulations also have an issue with resolving flows close to walls where energetic vortical
structures become very small with the increase of Reynolds number. However, there are cases where
LES provides significant benefits, including combustion chemical reaction problems and acoustic
problems.
While Reynolds-averaged Navier-Stokes (RANS) resolves the mean flow and requires a turbulence
model to account for the effect of turbulence on the mean flow, Large Eddy Simulation (LES) computes
both the mean flow and the large energy containing eddies. A subgrid model is used to capture the
effects of small scale turbulent structures. The spatial filtering process is used to filter out the turbulent
structures from the instantaneous flow field which are smaller than a given filter size. This filtering
process is based on the decomposition of instantaneous variables (velocity, pressure) into filtered
(resolved) and sub filter (unresolved or residual) variables. Here, velocity is used as an example.
where
• : instantaneous velocity,
• : filtered velocity,
• : sub filtered (unresolved) velocity.
The filtered velocity field is obtained from a low pass filtering operation due to a weighted filter G,
defined as
: a cut-off filter,
•
: a Gaussian filter,
•
: a top-hat filter,
where is a cutoff width, representing a spatial averaging over a grid element. Among
those, a top-hat filter (or similar one) is a popular choice for commercial CFD codes where unstructured
meshes are usually adopted.
Although the LES decomposition method resembles the Reynolds method, they have an important
difference due to the following.
Once this concept is substituted into the instantaneous Navier-Stokes equations, and then the spatial-
averaging (or filtering) is made, the filtered Navier-Stokes equations are obtained. The filtered Navier-
Stokes equations include the equations for the filtered continuity and filtered momentum equations,
which are given below.
where
•
is due to the decomposition of the nonlinear convective term
in the momentum equation.
•
is the filtered strain rate tensor.
where
•
is the cross stress tensor, representing the interactions between large and
small turbulence eddies.
•
is the Reynolds subgrid stress tensor.
Since in the convection term of the filtered momentum equation needs a secondary filtering
process, it is rewritten as shown below (Leonard, 1974),
where is the Leonard stress tensor, representing interactions among large turbulent
eddies.
Utilizing the decomposition process shown above, the filtered momentum equations can be rewritten as
where
•
is the filtered strain rate tensor,
Similar to RANS, the subgrid stress tensor is an unknown and must be computed by a subgrid model.
However, the subgrid stress tensor differs from the Reynolds stress tensor. The table below summarizes
the differences between time averaging for RANS and spatial averaging for LES. Common subgrid scale
(SGS) models available to solve the filtered Navier-Stokes equations include:
RANS LES
Double averaging
Turbulence averaging
Turbulent stress
The Smagorinsky-Lilly SGS model is based on the Prandtl’s mixing length model and assumes that a
kinematic SGS viscosity can be expressed in terms of the length scale and the strain rate magnitude of
the resolved flow.
where
•
is the subgrid turbulent viscosity,
•
is the filtered (resolved) strain rate tensor
A major drawback of this model is that the model constant ( ) does not vary in space and time.
Furthermore, this model has no correct wall behavior and it is too dissipative for laminar turbulent
transition cases. These limitations led to the development of a dynamic model for which the model
constant is allowed to vary depending on the grid resolution and flow regime.
Recognizing variations in space and time, Germano et al. (1991) proposed the dynamic model to
compute the value of rather than specifying it explicitly. It is implemented by utilizing two filters: a
cutoff filter and a test (coarse) cutoff filter
Figure 38 shows a resolved turbulence region utilizing Large Eddy Simulation (LES) and a modeled
region assuming the subgrid tensor
Figure 38: Energy Spectrum for LES Using the Cutoff Filter Width ( )
The test subgrid stress tensor with the coarse filter ( ) can be written as
where
•
is the coarse filtered strain rate magnitude.
•
is the filtered strain rate tensor, using the coarse cutoff filter.
Figure 39 shows a resolved LES region and a corresponding subgrid modelled region (T) when the
coarse filter is employed.
Figure 39: Energy Spectrum for LES Using the Test Cutoff Filter Width ( )
Because of the coarse filtering, the test (coarse) subgrid stress tensor should be a summation of the
coarse filtered subgrid stress tensor and the Leonard stress tensor .
where
• is the subgrid tensor for the cutoff filter (or grid filtered), then test filtered.
• is the Leonard subgrid stress tensor, representing the contribution to the subgrid stresses by
turbulence length scales smaller than the test filter but larger than the cutoff filter.
where .
where
Since the above equation is overdetermined a minimum least square error method is used to determine
the coefficient .
In order to avoid numerical instabilities associated with the above equation, as the numerator could
become negative, averaging of the error in the minimization is employed.
Hybrid Simulations
In recent years, hybrid methods have increasingly been employed for the simulation of unsteady
turbulent flows. These models form a bridge between Large Eddy Simulation (LES) and Reynolds-
averaged Navier-Stokes (RANS) by utilizing RANS for attached boundary layer calculations and LES
for the separated regions. In general, hybrid simulations need spatial filtering processes to determine
the local sub grid turbulent viscosity. Compared to LES and Direct Numerical Simulation (DNS), hybrid
simulations are extremely robust since the numerical requirement is less severe than the two other
approaches. Hybrid simulations include Detached Eddy Simulation (DES), Delayed Detached Eddy
Simulation (DDES) and Improved Delayed Detached Eddy Simulation (IDDES).
Detached Eddy Simulation (DES) uses one of the one or two equation Reynolds-averaged Navier-Stokes
(RANS) turbulence models to define the turbulence length scale or distance from the wall. The distance
is then used to determine the region for which the RANS or Large Eddy Simulation (LES) models will be
used.
DES using the Spalart-Allmaras (SA) model employs a modified distance function to replace the
distance (d) in the SA model’s dissipation term (Spalart et al., 1997).
where d is the distance to the closest wall, is a constant and is a metric of the local element
size, often chosen as the largest grid spacing in all three directions ( ).
The SA model is activated in a boundary layer region where d < ∆, while a Smagorinsky-like LES model
is used in regions where ∆ < d. Similarly, Strelets (2001) adopted the SST model as the baseline model
for DES, which computes a turbulence length scale and compares it with a grid length size for a switch
between LES and RANS. However, both approaches need careful determinations of local grid sizes
because these grid sizes act as the switch for the activation of LES and RANS.
Delayed Detached Eddy Simulation is an improved version of DES that avoids the undesired activation
of Large Eddy Simulation (LES) in boundary layer regions when the maximum grid size is less than the
distance to the closest wall (∆ < d).
In this model (Spalart et al., 2006), the length scale is redefined as:
where , , .
Shur et al. (2008) proposed an improved DDES to ensure that most of the turbulence is resolved when
the model is operating as a wall modeled Large Eddy Simulation (LES). This is achieved by introducing a
blending function of length scales defined as:
where
•
,
• ,
•
,
•
,
•
,
•
,
•
,
• .
Near-Wall Modeling
For internal wall bounded flows, proper mesh resolution is required in order to calculate the steep
gradients of the velocity components, turbulent kinetic energy, dissipation, as well as the temperature.
Furthermore, the first grid locations from the wall and the stretching ratio between subsequent points
are also determinant factors that affect solution accuracy. Since boundary layer thickness is reduced
as the Reynolds number increases, the computational cost increases for higher Reynolds number flows
due to the dense grid requirements near the walls. Additionally, it is hard to determine the first grid
locations for complex three dimensional industrial problems without adequate testing and simulation.
For high Reynolds number flows, it is numerically efficient if flows within boundary layers are modeled
rather than resolved down to the wall. This suggests that a coarse mesh is used at the expense of
numerical accuracy when compared to fully wall resolved approaches.
Wall Function
Figure 40 shows velocity profiles over a flat plate under the zero pressure gradient. Velocity profile and
wall distances are scaled by the friction velocity ( where is the wall shear stress and is
the fluid density). This is referred to as the log-law velocity profile. As shown in Figure 40, the velocity
profiles are divided into three distinguished regions, the viscous sublayer, the buffer layer and the
logarithmic layer.
The wall function utilizes the universality of the log-law velocity profile to obtain the wall shear stress.
This wall function approach allows the reduction of mesh requirement near the wall as the first mesh
location is placed outside the viscous sublayer.
Overall, the log-law based wall model is economical and reasonably accurate in most flow conditions,
especially for high Reynolds number flows. However, it tends to display poor performance in situations
with low Reynolds number flows, strong body forces (rotational effect, buoyancy effect), and massively
separated flows with adverse pressure gradients.
where is the velocity ( ) parallel to the wall, normalized by the friction velocity. The friction
velocity is defined as , is the wall shear stress and is the fluid density. is the
normalized wall distance (or wall unit). The distance from the wall is and is the fluid dynamic
viscosity.
The wall shear stress can be estimated from the above equation via the iterative solution procedure.
The two equation turbulence models based on eddy frequency (ω) and the Spalart-Allmaras (SA) model
do not require special wall treatments to solve the boundary layer as these models are valid through the
viscous sub layer. However, the two equation turbulence models based on turbulence dissipation rate (ε)
need additional functions to simulate the near wall effects. The two layer wall model is one of them. In
the two layer model, turbulent kinetic energy is determined from the turbulent kinetic energy transport
equation, while dissipation rate is resolved with a one equation turbulence model (Wolfstein, 1969) in
where
•
,
• ,
• is the von Karman constant,
• = 5.08 (Chen and Patel, 1988).
where
•
,
• = 70.
CFD simulations require specification of turbulence variables at inlet boundaries. When measured
turbulence data is available you can explicitly specify turbulence variables. For example, eddy viscosity
for the Spalart-Allmaras (SA) model, turbulent kinetic energy and eddy frequency/dissipation rate for k-
ε and k-ω based models. When measured data is not available, there are estimations for the turbulence
values that are based on the turbulence intensity and turbulence length scale or eddy viscosity ratio.
Turbulence Intensity
The turbulence intensity (I) is defined as the ratio of the root-mean-square of the turbulent velocity
fluctuations ( ) and the mean velocity ( ):
where and .
For fully developed internal flows the turbulence intensity can be estimated as
• 5 percent < < 20 percent for rotating machineries, for example, turbines and compressors
• 1 percent < < 5 percent for internal flows
• ~0.05 percent for external flows
where is the turbulent boundary layer thickness over a flat plate, is the
Reynolds number and x is the distance from the start of the boundary layer.
•
Option 3: Turbulence intensity and viscosity ratio . For the kinetic energy, . For
eddy frequency, .
•
Option 3: Turbulence intensity and viscosity ratio . For the kinetic energy, . For
Summary
During the past decades turbulence models of various complexities have been developed. Turbulence
models that employ the most assumptions are typically the least demanding from a CPU cost
standpoint. As the number of assumptions associated with the model decreases, the cost and accuracy
of the model typically increases. Since no universal turbulence model exists, CFD users need to
choose the best turbulence model specific to their application. The most cost effective approach is
typically steady Reynolds-averaged Navier-Stokes (RANS). When the flow is naturally unsteady and
exhibits oscillations in the quantities of interest, URANS may be employed to improve the accuracy of
solutions and gain a better understanding of how much the flow field fluctuates as a function time. For
applications when the flow transitions from laminar to turbulent as a function of space, a dedicated
transition model is required in order to accurately predict the near wall behavior. When the application
demands additional knowledge of the turbulent fluctuations for separated, reattaching or impinging
flow fields, hybrid simulations (Detached Eddy Simulation (DES), Delayed Detached Eddy Simulation
(DDES)) may be employed. Although the modeling assumption is still made within the boundary layer,
the tradeoff between accuracy and computational efficiency is often worthwhile compared to Large Eddy
Simulation (LES) or Direct Numerical Simulation (DNS). Finally, when the application requires detailed
information about nearfield fluctuations, or for combustion and acoustic applications, LES should be
considered.
Based on the information presented it is up to you to understand the complexity of your simulation
and make an appropriate decision on which turbulence model to use. Generally, it is recommended to
start with a simple turbulence model since a majority of industrial problems can be well defined with
a steady state assumption. Even if the application is inherently unsteady, this is still a good place to
start as a significant amount of information can be garnered from running RANS. If the initial RANS
fails to provide the level of accuracy required for the specific application you will need to consider the
use of a hybrid model or LES. Often times the flow regime exhibits sensitivities to the selection of the
turbulence model, specifically RANS vs. Hybrid/LES. In any case, the selection of the turbulence model
should be verified against previous simulation or validated with experimental/analytical data for the flow
in question.
This section on numerical approximation techniques covers topics, which describe the numerical
modelling of the fluid flow equations on a computational domain such as spatial discretization using
finite difference, finite element and finite volume techniques, temporal discretization and solution
methods.
3.4.1 Overview
The governing equations, that is, the Navier – Stokes equations in continuum mechanics are a set
of coupled non–linear partial differential equations derived from the conservation laws for mass,
momentum and energy.
For a numerical solution of a mathematical continuum model, the focus is to devise efficient, robust,
and reliable algorithms for the solution of the partial differential equations. To do this the partial
differential equations are converted to a discrete system of algebraic equation using a discretization
procedure. In case of a CFD simulation, the discretization of the governing equations, that is, the
derivation of equivalent algebraic relations should accurately represent the equations and the physics
modelled.
There are a number of discretization techniques that have been developed to solve different kinds of
problems in CFD. These methods can be broadly classified into two categories:
• Mesh based methods – These methods require division of the problem domain into a grid or
mesh over which the governing equations are discretized. The points are positioned according
to a topological connectivity which ensures the compatibility of the numerical technique used.
Examples of these methods include Finite Difference (FD), Finite Volume (FV) and Finite Element
(FE) methods.
• Mesh free methods – These methods use a collection of nodes in the domain which don’t
have any apparent connectivity. These methods are particularly important in simulations where
the nodes are created or destroyed, simulations where the deformations are so large that the
connectivity might introduce distortion and in cases where the domain possesses discontinuities
or singularities. A few examples of these methods are: Smoothed Particle Hydrodynamics (SPH),
Finite Pointset Method (FPM), Meshless Local Petrov Galerkin (MLPG) and Lattice Boltzman
methods.
Of the methods discussed above, mesh based methods are more popular and are among the most
widely used in CFD codes. The following sections describe the mesh based discretization methods
(FD, FV and FE) and the time discretization techniques used to solve the partial differential equations
governing a fluid flow.
This section describes the formulation and methodology of finite difference method to solve the
governing equations on a computational domain.
The finite difference method is the oldest method for the numerical solution of partial differential
equations. It is also the easiest to formulate and program for problems which have a simple geometry.
When calculating derivatives finite difference replaces the infinitesimal limiting process with a finite
quantity. The derivative of a function at point expressed as:
is replaced by:
The preceding expression is commonly referred to as the forward difference approximation. This
derivative can have more refined approximations using a number of approaches such as truncated
Taylor series expansions and polynomial fitting.
The term gives an indication of the magnitude of the error as a function of the mesh spacing and
is therefore termed as the order of magnitude of the finite difference method. In the above formulation,
the finite difference approximation employed is first order accurate. A second order approximation
would have the order of magnitude expressed as . Most of the finite difference methods used
in practice are second order accurate. An example of a second order method is the central difference
approximation of the first derivative. It is expressed as:
The basic methodology in the simulation process using finite difference approach consists of the
following steps.
Figure 41: Basic Methodology in the Simulation Process Using Finite Difference Approach
The finite difference formulation generally employs a structured grid. The most commonly used indices
are , and for the grid lines at , and , respectively. The function value at such a
Consider the one dimension advection-diffusion equation for a scalar quantity governed by
where is the specified velocity, is the density of the fluid and is the diffusivity.
If the domain is defined by the boundary and and the boundary conditions by and
, the domain can be discretised (non-uniformly) by a total of grid points for the finite
difference solution of the problem.
The diffusion term can be approximated using the central difference (both the inner and outer
derivative) as:
The convection term can also be approximated using the central difference as:
If the grid is uniform and the values of density, velocity and coefficient of diffusion are constant
throughout the domain the above equations reduce to:
The resulting set of linear algebraic equations can be solved to get the values of at the grid points.
Finite difference methods are advantageous for the numerical solution of partial differential equations
because of their simplicity, efficiency and low computational cost. Their major drawback is their
geometrical inflexibility such as application on an unstructured grid or moving boundaries. Their
formulation increases in complexity as the complexity of the domain increases.
The restrictions resulting from the above mentioned geometrical complexities can be alleviated by the
use of methods such as grid transformation and immersed boundary techniques.
This section describes the formulation and methodology of finite volume method to solve the governing
equations on a computational domain. It also describes the cell centered and face centered approaches
used for finite volume formulation.
The finite volume formulation is based on the approximate solution of the integral form of the
conservation equations. The problem domain is divided into a set of non overlapping control volumes
referred to as finite volumes, where the variable of interest is usually taken at the centroid of the finite
volume. The finite volume is also referred to as a cell or element.
The governing equations are integrated over each finite volume and interpolation profiles are assumed
in order to describe the variation of the concerned variable of interest between the cell centroids. The
resulting discretization equation expresses the conservation principle for the variable inside the finite
volume.
The governing differential equations discussed in this manual each have a dependent variable that
obeys a generalized conservation principle. If the dependent variable (scalar or vector) is denoted by ,
the generic differential equation is:
where
• The convection term accounts for the transport of due to the velocity field .
• The diffusion term accounts for the transport of due to its gradients.
• The source term accounts for any sources or sinks that affect the quantity .
When the above equation is integrated over a three dimensional control volume it yields
The divergence terms (convection and diffusion) can be converted into surface integrals using Gauss
divergence theorem and the above equation can be rewritten as
where
The integral conservation in the scalar transport equation applies to each control volume as well as
the complete solution domain thus satisfying the global conservation of quantities such as mass,
momentum and energy. These quantities can be evaluated as fluxes at the surfaces of each control
volume.
In order to obtain an algebraic (discretised) equation for each control volume the surface and volume
integrals are approximated using quadrature formulae in terms of function values at the storage
location. This may require values of variable at points other than the computational nodes of the control
volume. Values at these locations are obtained using interpolation schemes.
• Cell centered approach: The domain is defined by a suitable grid of finite volumes and the
computational nodes are assigned at the centroid of the control volumes.
• Face/Node centered approach: The domain is first defined by a set of nodes and control volumes
are constructed around these nodes as cell centers such that the faces of the control volume lie
between these nodes.
The cell centered approach and node centered approach have nearly the same accuracy and efficiency
for most of the cases which use a structured grid.
To illustrate how the conservation equations in CFD can be discretised using finite volume method
an example involving the steady transport of x-momentum in a uniform 2D rectangular grid can be
considered:
A cell centered approach is employed with the points P, W, E, N, S representing the cell centers and the
notations n, e, s, w representing the faces of cell P. The notation N (n), E(e), S(s), W(w) represent the
north, east, south and west directions, respectively.
In the example considered the pressure gradient at center of cell P can be evaluated by
interpolating the pressure values from surrounding nodes. Other source terms can be evaluated
similarly, interpolating where necessary to estimate the cell center values.
Evaluation of Diffusive Fluxes
The diffusive fluxes can be approximated as
The gradients of velocity at the cell faces (e, w, n, s) can be approximated using central difference
comprising of the cell center values. This results in diffusive fluxes being represented as
The discretised equation which is second order accurate takes the form
where , and so on are mass fluxes through the east, west, north and south faces.
The values of at cell faces need to be obtained using appropriate interpolation schemes between
the cell center values. A few examples of such schemes are linear interpolation (CDS), quadratic
upwind interpolation (QUICK), total variation diminishing (TVD).
The finite element approach is based on the discretization of the domain into a set of finite elements
(which are usually triangles or quadrilaterals in a two dimensions and tetrahedral, hexahedra, pyramids
or wedges in three dimensions) and using variational principles to solve the problem by minimizing an
associated error function or residual.
The finite element approach is based on the discretization of the domain into a set of finite elements
(which are usually triangles or quadrilaterals in a two dimensions and tetrahedral, hexahedra, pyramids
or wedges in three dimensions) and using variational principles to solve the problem by minimizing an
associated error function or residual.
The unknown variables are approximated over the domain using interpolation procedure in terms of
nodal values and set of known functions called shape functions. This approximation is substituted into
the governing (conservation) equations in their differential form and the resulting error (residual) is
minimized in an average sense using a weighted residual approach.
The variational or weighted residual formulations transforms the governing equations into an integral
form called the global weak form. The weak form when applied to each finite element results in a set of
discrete equations in terms of the nodal unknown which can then be solved by a number of methods.
In this formulation the governing differential equation for a quantity is expressed as:
where
•
is the diffusive flux vector
where
Finite element method requires transformation of the governing differential equation into an integral
equation over the domain. This is accomplished through approaches such as weighted residual
formulation and least squares formulation.
A few of the approaches used to convert the differential equations into their weak integral forms are
discussed below.
In order to determine the nodal values , the inner product of the residual with a prescribed weighted
function is set to zero
The above equation is called the strong form of the weighted residual method. In many cases it is
possible to perform integration by parts and obtain the weak form of the equation which contains lower
order derivatives than the ones occurring in and has reduced continuity requirement for the weighted
function. The weak form is expressed as:
where , and are the differential operators with lower order derivatives than and represents
the domain and the boundary of the domain.
Different solution methods are obtained based on the choice of the weight function. Some of these
approaches are:
• Point Collocation: where is the Dirac-Delta function. This is analogous to the finite
difference approach.
• Subdomain Collocation: for the subdomain (element) and zero elsewhere. This leads to
a finite volume formulation.
• Galerkin method: where is a shape function assumed over an element.
• Petrov Galerkin method: . This represents a generalization of all methods except the
Galerkin method.
Using variational principles leads to a least square weak form written as:
where
This section describes the various approaches used to discretize the governing equation in the temporal
domain such as two step, multistep and multistage methods.
In order to determine a numerical solution of the governing differential equations the temporal domain
must also be discretised apart from the spatial domain. The direction of influence of the time coordinate
is only in the future. Therefore, all the solution methods for time dependent problems advance in time
from a given initial data.
A vast majority of the methods used for temporal discretization are linear in nature. The time
dependent variable is updated using a linear combination of the variable and its time derivatives. Linear
approaches can be broadly categorized based on the number of steps, stages and derivatives used in
the discretization.
Some of the widely used time discretization approaches are described below.
Consider a first order ordinary differential equation for a dependent variable expressed as:
A generalized scheme using a weighted average value for the approximation of the variable value at the
th
n+1 time step is expressed as:
where
th
represents the function value at the n time step
The nature and stability of the temporal discretization scheme depends on the choice of the weight
and represent the forward Euler, Crank-Nicholson and the backward Euler schemes,
respectively.
Multistep Methods
Multistep methods involve function values at more than two instances of time. These methods are
generally derived by fitting a polynomial to the temporal derivative of the dependent variable, that is,
.
These methods include the Adams-Bashforth and Adams-Moutlton methods. The order of the method
depends on the number of points in time at which the polynomial fitting it used. A third order accurate
Adams-Moulton method is expressed as:
These methods require initial data at many steps, hence they are not self starting.
Multistage Methods
Multistage methods involve computation of the function values multiple times at the same time step.
th
They generally involve predictor and corrector steps to compute the values at the n+1 time step.
Numerical solution schemes are often referred to as being explicit or implicit. When a direct
computation of the dependent variables can be made in terms of known quantities the computation is
said to be explicit. When the dependent variables are defined by coupled set of equations, and either
a matrix or iterative technique is needed to obtain the solution, the numerical method is said to be
implicit.
Explicit methods are easy to program but are conditionally stable whereas implicit methods offer better
stability but are computationally expensive. Predictor-Corrector methods offer a compromise between
these choices. A variety of methods exist based on the choice of base method and the time instants
used in the predictor and corrector steps.
The most popular methods in this category are the Runge-Kutta methods. A fourth Runge-Kutta method
is constructed as follows:
Generalized- Method
The generalized method is an implicit method of time integration which achieves high frequency
numerical dissipation while at the same time minimizing unwanted low frequency dissipation and offers
unconditional stability for linear problems. It is a variant of the generalized two step theta scheme
discussed above where the first temporal derivatives are evaluated as variables.
The generalized method for integration from time step to is constructed as follows:
where
This section describes the direct and iterative solution methods used to solve the linear system of
equations obtained after spatial and temporal discretization of the governing equations.
Most of the methods of discretization discussed (FDM, FVM and FEM) yield a linear system of equations
that need to be solved.
that is
where
The methods used to solve the linear system of equations can be classified into two categories:
Direct Methods
Direct methods obtain an exact solution of the linear system of equations. These methods generally
consist of elementary matrix operations to simplify the equations that will allow an exact computation.
Some of these methods are:
• Cramer’s Rule: The solution to is computed directly by and using Cramer’s rule to
compute the inverse of the coefficient matrix by calculating the co factor and adjoint of matrix .
• LU Decomposition: This approach is based on the assumption that matrix is composed
of an upper triangular and a lower triangular matrix, that is . This results in
after which the systems and are solved separately.
• Gaussian Elimination: The given linear system of equations is converted to an augmented
after elementary operations. The solution to the original system of equations is found by back
substitution from the last row of the augmented matrix.
Direct methods work well for simple problems which generate lower order matrices but they are
computationally expensive and inefficient for large sparse matrices. Moreover the coefficient matrix
depends on the properties of the fluid, such as viscosity or density which evolves as the solution is
refined for the primary variables. Therefore it is not efficient to seek an exact solution to for
every global iteration because the variables used to build and are evolving.
Iterative Methods
Iterative methods obtain an approximate solution of the linear system of equations that is based on
iterations. An approximate solution of the primary variables is calculated which is then fed back into
the primary equations to refine the variables. For large scale CFD problems it is more efficient to
slightly improve over each variable at one turn and then cycle over to all the coupled variables in the
subsequent iteration to reduce the error.
Iterative methods generate an approximate solution to the system that tends to converge to an exact
solution. After iterations the approximation is described as:
where is the residual after iterations. The error in the approximation is defined as:
The purpose of the iterative methods is to drive this residual below the convergence criteria set.
Based on the way that the solution at every iteration is updated, iterative methods can be classified into
two types:
• Stationary Iterative Methods: The terms and are not dependent on the iteration count
. Some examples of these methods include Jacobi method, Gauss-Seidel method, Successive
Overrelaxation method (SOR) and Symmetric Successive Overrelaxation method (SSOR).
• Non-Stationary Iterative Methods: The terms and are updated at every iteration. Some
examples of these methods include Conjugate Gradient method (CG), Generalized Minimal
Residual method (GMRES) and Conjugate Gradient Squared Method (CGS).
This section on AcuSolve solver features covers the description of various solver features available in
AcuSolve such as heat transfer, fluid structure interaction and turbulence modeling.
Commercial solvers for fluid flow analysis come in many forms. They use a collection of numerical
methods and approximation approaches to solve the governing continuum equations discussed in this
manual.
AcuSolve is a general purpose CFD flow solver that is used in a wide variety of applications and
industries. The flow solver is architected for parallel execution on shared and distributed memory
computer systems and provides fast and efficient transient and steady state solutions for standard
unstructured element topologies.
AcuSolve is based on Galerkin Least Squares (GLS) finite element method. The GLS formulation
provides second order accuracy for spatial discretization of all variables and utilizes tightly controlled
numerical diffusion operators to obtain stability and maintain accuracy. In addition to satisfying
conservation laws globally, the formulation implemented in AcuSolve ensures local conservation
for individual elements. Equal order nodal interpolation is used for all working variables, including
pressure and turbulence equations. The semi discrete generalized alpha method is used to integrate the
equations in time. This approach has been verified as being second order accurate in time.
AcuSolve uses a time marching procedure to solve both steady state and transient simulations. In
the case of steady state simulations, the inertia (mass) terms of the conservation equations are only
included in the Galerkin part of the finite element weighted residual formulation. This inclusion adds
stability to the nonlinear iterations. Other parts of the finite element formulation (such as the least-
squares operator) do not include the inertia terms. This exclusion accelerates non linear convergence
to steady state at the expense of time accuracy. The initial time step size is set to a significantly large
value (1.0e+10).
For transient simulations the inertia terms are included in all the operators of the finite element
formulation in order to preserve time accuracy.
The resultant system of equations is solved as a fully coupled pressure/velocity matrix system using a
preconditioned iterative linear solver. The iterative solver yields robustness and rapid convergence on
large unstructured meshes even when high aspect ratio and badly distorted elements are present.
AcuSolve consists of multiple features designed to tackle various flow physics such as temperature flow,
solar radiation and rigid body dynamics.
AcuSolve supports various features for the modeling of the heat transfer phenomenon such as:
• Convective Heat Transfer: Heat transfer due to the motion of molecules within a fluid. This
includes both natural and forced convection.
• Conjugate Heat Transfer: Heat transfer between solids and fluids.
• Radiation Heat Transfer: Heat transfer due to radiation between surfaces having high
temperatures.
• Solar Radiation: Heat transfer due to incident solar radiation on a surface.
• Additional features: Thermal shells, viscous heating and compression heating.
This approach permits the conservation of energy with the Galerkin Least Squares formulation.The
temperature is derived from the state equation specified for the flow.
There are two common approaches used for the solution of thermal problems in AcuSolve.
• Segregated approach: The enthalpy equation is solved separately from the flow equations. This
approach can be further divided into two parts:
• Sequential solving: The enthalpy equation is solved in a sequential manner after solving the
coupled flow equations using feed- forward of the velocity field into the enthalpy equation. This
approach is appropriate for cases where the temperature field does not affect the flow field
significantly. This can offer benefits in speed since the iterations on flow field are not carried along
as part of the temperature equation. If the flow field has already converged sufficiently, additional
iterations are wasted. This approach also enables “frozen flow” thermal simulations where the flow
field is held constant, but different scenarios of the temperature field are investigated.
• Concurrent solving: The enthalpy equation is solved in a sequential manner after the flow
equations but there is feedback of the temperature dependent variable from the enthalpy equation
back to the flow equations during each iteration. This approach is appropriate for cases involving
temperature dependent material properties.
• Coupled approach: The enthalpy equation is coupled into the global system and solved in
conjunction with the flow equations. This approach is often times more stable for cases exhibiting
a large degree of coupling between the temperature and flow fields, that is, buoyancy driven
flows.
• View factor computation: In radiative heat transfer view factor is the proportion of radiation
incident on one surface due to another surface. The view factors for each facet defining the
radiation enclosure are computed using the hemicube algorithm and smoothed using least squares
method as a pre-processing step. The view factors are not recomputed during the simulation.
• Heat flux addition: The radiative heat flux (based on view factors) computed using the Stefan-
Boltzmann law and the total radiosity (based on Kirchoff's law) is added to the enthalpy transport
equation during the solver run.
The enclosure radiation model is supported only on fluid mediums, that is, the fluid side of the fluid/
solid surface.
View factor computation is an important point when dealing with moving mesh simulation. Since the
view factors are not recomputed during the solver run the boundary elements forming the enclosure
should not deform.
AcuSolve can model shell layers of varying thickness and material properties. There are two
mechanisms by which the heat transfer equation is solved in the thermal shell depending on the
number of layers:
• Single Layer: The elements in a single layer shell are considered as full volume elements. The
coordinates of the nodes on the opposite faces of the shell are offset by the thickness provided
and the heat transport equations are used to determine the heat transfer within the shell. For this
scenario the thermal shell includes all heat transfer effects that are present in a three dimensional
volume element.
• Multiple Layer: The nodes on the opposite faces have the same coordinates but have different
node numbers in order to support the temperature differences across the shell. A one dimensional
heat equation, through the shell thickness at the element nodes, is derived by neglecting thermal
inertia and conduction parallel to the surface. This means that the heat flux is the same in all
the layers. If each layer material model has a constant conductivity then this simplified heat
equation is solved exactly. If any of the conductivities is a function of temperature then a two-pass
procedure is used to approximately solve the resulting nonlinear system. Once the temperatures
are known at the corners of all the layers in each element they can be interpolated to the
quadrature points in the usual manner.
where
One of the rich features of AcuSolve lies in its ability to simulate FSI problems. AcuSolve simulates
these problems using two different technologies.
Practical FSI (P-FSI)
In P-FSI simulations, AcuSolve and the structural code are run separately. The Eigen solution
obtained from a structural solver is projected onto the CFD mesh and used as boundary conditions
to displace the fluid mesh. The mass, stiffness and damping arrays are also transferred from the
structural model to flexible body in the CFD model, completing the inputs necessary to compute
the structural deformation. P-FSI is preferred for simulations that have linear structural response
(linear material properties and small mesh deformations).
Direct Coupled FSI (DC-FSI)
For applications involving a nonlinear structural response, AcuSolve is coupled at run time to a
structural solver using the DC-FSI approach. The codes are run in tandem and exchange forces
and displacements at each time step. This technique captures all nonlinearities in the structural
model and can exploit the full capabilities of the structural solver. The DC-FSI technology within
AcuSolve performs all projection and interpolation between the structural and CFD meshes
without the use of any middleware. The supported structural solvers for AcuSolve include
RADIOSS, Abaqus and MD-Nastran.
AcuSolve provides an extensive set of features to model the different material properties of the fluid
and the surrounding media. The different material properties that can be modeled are discussed in the
following sections.
Density
The density models can be specified for all types of media supported by AcuSolve (solid, fluid and
shell). They are:
• Constant Density: A constant density for the media. This is generally used for incompressible flow
simulations.
• Boussinesq Model: Use of the Boussinesq approximation, where the variation in density is
assumed to be a linear function of temperature and the density variation is accounted only in the
body force terms.
• Isentropic Model: The constant entropy gas model, where the variation in density is taken into
account for all the terms in the momentum equations.
• Ideal Gas Model: The density is based on the ideal gas law. The variation in density is taken into
account for all the terms in the momentum equations.
• Customized Models: Custom variable property functions for density can be specified using curve
fit options (piecewise linear and cubic spline) as a function of single independent variable. User
functions can be used to model more complex density models. In case of customized models the
density variation is applicable to all the terms.
Viscosity
The molecular viscosity models used in the flow (momentum) equations, specified for the fluid medium
are:
• Constant Viscosity: A constant viscosity for the fluid medium. This is the simplest case of a
Newtonian model for calculation of the stress tensor.
• Ramped Viscosity: The viscosity is ramped down 1,000 times from the viscosity value specified at
time step one to the viscosity value at time step ten, after which the value from time step ten is
used.
• Non-Newtonian Models: Non-Newtonian viscosity models based on Power Law, Bingham model,
Carreau-Yasuda model can be specified.
• Customized Models: Custom variable property functions for viscosity can be specified using curve
fit options (piecewise linear and cubic spline) as a function of single independent variable. User
functions can be used to model more complex viscosity models.
Porosity
AcuSolve uses the Darcy-Forchheimer porosity model for the flow (momentum) equations, specified for
the fluid medium. The porosity model modifies the momentum equation as follows:
where
• is the field vector contribution due to porous media, also referred to as porous media forces.
• is the rotation tensor which rotates to the global coordinate system.
where
Viscoelasticity
AcuSolve supports simulations involving viscoelastic materials using the Upper Convected Maxwell
Model and the Upper Convected Maxwell Log Model. The viscoelastic models used in the flow
(momentum) equations, have various sub models to determine the rate of stress build up and rate of
stress decay. These include:
• Oldroy-B Model: Stress build up and decay rate obtained by defining the relaxation time and
polymer viscosity of the fluid.
• Giesekus Model: Stress build up and decay rate obtained by defining the mobility factor along with
the relaxation time and polymer viscosity of the fluid.
• Phan-Thien-Tanner Model (PTT): Stress build up and decay rate obtained by defining the PTT
extensibility factor and the diffential ratio along with the relaxation time and polymer viscosity of
the fluid.
• Customized Models: Customized models based on a user function.
Additional Properties
AcuSolve also supports the modeling of several additional properties such as Surface Tension,
Conductivity, Diffusivity, Contact Angle and Specific Heat of the fluid medium.
All of these material properties except Contact Angle can be specified using customized models based
on curve fit options (piecewise linear and cubic spline) as a function of single independent variable or by
using user functions.
The Conductivity can be additionally specified to be based on an Isotropic Prandtl number Model,
ramped to a value or set as anisotropic using the customized models.
Contact Angle for free surfaces with surface tension can be set to a constant value or automatically
computed from the boundary conditions.
AcuSolve has the capability to track multiple species in a fluid flow by using the scalar transport
equations for each of them individually. The species transport equation is similar to the generalized
governing equations for the fluid flow and it is expressed as:
where
th
• represents the i scalar species.
where is the diffusivity for species . The diffusivity can be modeled as constant, ramped against
time step or customized using variable property and user functions.
The material properties can be set to be functions of species concentration. This models a ‘miscible’
property relative to mixing of multiple fluids.
3.5.5 Aeroacoustics
AcuSolve performs acoustic computations in flow simulations as a post-processing step. There are three
types of output parameters used.
simulations where the physical domain is finite but the location of interest to the observer is often
beyond those boundaries.
• Vibration Acoustics (VAO): Specifies output parameters to support vibro-acoustic outputs.
AcuSolve supports a variety of turbulence models for fluid flow simulations. For steady state simulations
the Reynolds-averaged Navier-Stokes equations are solved to arrive directly at the time averaged
flow field. In case of transient simulations the governing equations are integrated in time to yield
an accurate description of the flow field. There are also many different turbulence closure methods
available for each type of simulation.
• Fully Resolved: This approach integrates equations directly to the wall and uses near wall damping
functions to produce appropriate behavior.
• Wall Function: This approach uses a wall model based on the standard Law of the Wall for
turbulent boundary layers.
• Running Average Wall Function: This approach enforces the Law of the Wall on the running
average flow field.
The wall functions used in AcuSolve are valid through the viscous sublayer and the buffer layer and
have the advantage of having no lower bound limit. The upper bound limit for is 300.
This section gives a brief description of a general workflow of a typical CFD process. It describes in short
the problem description, pre-processing, solving and post-processing aspects of a CFD workflow.
AcuSolve is a robust software that does all of the above steps of CFD and provides the accurate solution
to the problem.
General Workflow
The accurate simulation of a flow field using CFD requires a number of steps to be taken. In order to set
the problem up properly for simulation it is necessary to first understand the physics. Once the physics
have been understood a model can be constructed. The following steps describe the set of steps that
should be taken when performing a CFD simulation.
Problem Description
Though often neglected, this step is the foremost important aspect of a successful CFD simulation.
Before attempting a CFD simulation a clear description of the problem is necessary. This step
involves the following tasks.
• Identify the domain of interest: Though in principle domain of any size can be analysed
with CFD simulation, it is often wise to choose the critical regions in which flow field is of
importance and simulate the flow in that domain. Minimizing the size of the domain keeps
the simulation costs low, but comes at the expense of introducing non physical behavior in
the simulation by having boundaries too close to regions in the flow where the results are
of interest. It is important to select boundaries where the flow character is known so that
proper boundary conditions can be applied.
• Identify the nature of the flow: It is always helpful to make an educated guess on the nature
of the flow, such as whether it is steady/transient, laminar/turbulent, thermal/non thermal.
For example, while analysing a flow in a pipe that has low Reynolds number (say Re ~ 100),
it is useful to perform a laminar analysis rather than doing an expensive turbulent flow
analysis with a larger mesh count.
In case of turbulent flows, it is advisable to choose turbulence model based on the physics
present, while keeping in mind the areas where the flow will be critical.
Pre-Processing
This process involves geometry import, mesh generation and specifying the relevant physics
equations to be solved. Pre-processing also involves defining the fluid properties, boundary
conditions and body force parameters. AcuSolve uses AcuConsole for the pre-processing of the
geometry and AcuMeshSim (which is launched from AcuConsole itself) for the mesh generation.
Solving
Solving is the critical part of a CFD simulation. The governing flow equations are discretized. All
these algebraic equations are then solved for the unknown flow parameters. The final solution
is the quantitative representation of these parameters at the mesh generated nodes. As a first
step of solving, the AcuPrep module of AcuSolve checks for the compatibility of the information
provided in the pre-processing step. The source of errors in AcuPrep could be because of
insufficient inputs, syntax errors, and so on. After AcuPrep the actual solving of equations is
performed.
Post-Processing
Having solved for the unknowns, the next step is the analysis of this huge data to obtain relevant
insight. This data can be analysed either in the form of contour plots or tabular data to identify
the critical portions of the flow field. AcuSolve is integrated with the post-processing tools like
AcuTrans, AcuProbe and AcuFieldView which provide various options of post-processing.
4.2 Pre-Processing
This section details the several sub-categories of pre-processing in AcuSolve, including the use of user
interface, solver parameters, model parameters and mesh generation.
This section gives details on using the user interface of AcuSolve by describing several groupings of the
parameters available for pre-processing.
1. Click Start > All Programs > Altair HyperWorks > AcuSolve > AcuConsole.
The main window contains the following components: Data Tree, panel area, View Manager, Data Tree
Manager, Information window and graphics window.
Figure 46:
Data Tree
Data Tree is an area at the top left portion of the AcuConsole main window, which contains items in a
tree structure. These items represent the models or panels to set the problem parameters. A detailed
description of the parameters in Data Tree is provided in the following sections.
Panel Area
The panel area is located in bottom left portion of the AcuConsole main window. This area will open
the panels which set the problem parameters. These panels are a set of data entries such as booleans,
numbers, strings, tabs, lists, arrays and references to other panels.
View Manager
View Manager is a vertical narrow area in the middle of the AcuConsole main window. This area contains
command tools. These tools act on the model in the Visualization Area and are introduced in detail in
this manual.
Visualization Area
The Visualization Area is located in the right portion of the AcuConsole main window. This area is
responsible for displaying the model that is generated by AcuConsole as a result of rendering and
analyzing the visualization data. The area is interactive and gives the ability to rotate, scale, translate
and other common transformations by using mouse clicks and drag and drops.
Information Window
The Information window is located in the bottom right portion of the AcuConsole main window. All of the
information that you request will be printed in this window. You can save this information into a text file
and you can also clear it.
Global
The items in the Global Tree define the parameters that are applied to the entire problem domain. This
section includes various sub-sections and summarizes the sub-sections and their functions.
Problem Description Specifies analysis type and equations to be solved in the problem.
Solar Radiation Parameters Specifies global solar radiation algorithms and parameters.
External Code Parameters Specifies communication parameters for coupling with an external
solid/structural code.
Particle Trace Code Parameters Specifies communication parameters for coupling with AcuTrace
(particle tracing module of AcuSolve).
Auto Solution Strategy Automatically creates a solution strategy by issuing all the other
commands in this section.
Advanced Solution Strategy Specifies advanced features of solution strategy such as time
sequence, time integration and linear solver parameters.
Surface Tension Model Specifies surface tension model for free surfaces.
Contact Angle Model Specifies a contact model for free surfaces with surface tension.
Reference Frame Specifies a reference frame for element sets and boundary
conditions.
Output Specifies the nodal data output that is required for post-
processing.
Nodal Initial Conditions Specifies the initialization of nodal values at the start of
simulation.
Overall, parameters that are independent of the geometry and mesh of the problem are specified in this
section.
Model
Model is the section where parameters specific to the geometry and mesh of a problem are defined.
Model has the following sub-sections.
• Volumes: This section gives the list of all of the volumes that are available. All of the elements
belonging to a particular volume are grouped into a single element set. All of the elements in an
element set have the same material model and body force.
• Surfaces: The subsection Surfaces lists all of the surfaces available in which the boundary
conditions are specified.
Periodic boundary conditions are a set of boundary conditions that are often used to simulate a large
system by modeling a small part that is representative of the actual model. Periodicity can be set
between surfaces using AcuConsole.
1. To create a new periodic condition right-click Periodics in the Data Tree and select New.
Periodic 1 will be created.
In the Periodic BC Dialog, you need to specify the two sides to which the periodicity needs to be
applied, the type of periodicity and the distance or angle between those sides.
3. Select the two sides of the periodic condition in one of the following ways.
• Type in the name of the surfaces in the box next to Side 1 and Side 2 and press Enter.
• Select the surface from the drop down list next to Side1 and Side2.
• Click the eye icon. A Periodic pick model dialog opens. Select the surface and click OK.
Mesh Extrusions
Mesh extrusion is a powerful feature that allows the generation of structured mesh in the entire volume
or only on the surface. This technology extrudes the mesh on one surface to the other surface and can
also be used with other meshing features. Extrusions can be created in both 2D and 3D.
1. To create a new mesh extrusion, right-click Mesh Extrusions > New.
Mesh Extrusion 1 is created.
2. Right-click on Mesh Extrusion 1 to see the available options.
3. To define the extrusion options, click Define.
The Mesh Extrusion Dialog opens.
The mesh on Side 1 will be extruded onto Side 2. Therefore the order of surfaces selected is
important.
4. Select the extrusion surfaces using one of the following methods.
• Type the name of the surface in the box next to Side 1 and Side 2.
• Select the surface using the drop down lists.
• Click the eye icon and select the desired surface in the graphics window.
The Swap button next to Swap Sides is used for interchanging the Side 1 and Side 2 surfaces.
5. Select the swap option using one of the following methods.
• Number of Layers: Specify the number of layers and the type of elements.
• Layer height: Specify the Height of Layers and the type of elements.
The type of element can be selected from the Extrusion options drop down. Selecting mixed
elements provides mixed elements and All tets provide tetrahedral elements. The Extruded surface
option provides extrusion elements only on the surfaces. All the volume elements are generated
based on the global or volume mesh parameters.
Problem Description
The Problem Description branch under the Global Tree describes the general parameters of the
simulation such as title, subtitle, analysis type and flow equation.
Title
Provides a description of the problem. The title should not exceed 1,023 characters. Title is an
optional description for the problem and has no impact on the solution.
Sub title
The sub title of the problem is used to further describe the problem. This parameter is also limited
to 1,023 characters and is an optional description of the problem.
Analysis Type
This parameter specifies the type of analysis being performed, namely steady state or transient.
In either case, a time marching scheme is used to solve the problem. In the case of steady state
type, the inertia (mass) terms of the conversation equations are only included in the Galerkin part
of the finite element weighted residual formulation. This inclusion adds stability to the non linear
iterations. Other parts of the finite element formulation (such as the least-squares operator)
do not include the inertia terms. This exclusion accelerates non linear convergence to steady
state at the expense of time accuracy. For transient type, the inertia terms are included in all the
operators of the finite element formulation in order to preserve time accuracy.
Flow Equation
This parameter specifies the equations related to the flow physics to be solved in the problem. It
has two options of Navier- Stokes and Stokes equations. The Navier-Stokes equation describes
the motion of general viscous fluid. The Stokes equation is used to describe the low velocity, high
viscosity flows in which the convective inertial forces are negligible compared to viscous forces.
The only change for stokes flow is the removal of the (convective) terms from the Navier-Stokes
equation.
Abs. pressure offset
Usually in an incompressible flow regime the absolute pressures at the boundaries do not affect
the solution. Only the pressure difference plays the key role. However, in the compressible flow
regime the absolute pressures are important and this parameter Abs. pressure offset specifies the
offset to convert to absolute pressure units. For example, if Abs. pressure offset = 101325 Pa and
if you define the outlet Pressure as zero, then the absolute pressure at the outlet = 101325+0 =
101325 Pa.
Viscoelastic equation
This parameter specifies the viscoelastic material model equation to be used to obtain the
viscoelastic stresses in the simulation. AcuSolve supports two types of equations namely, Upper
Convected Maxwell and Upper Convected Maxwell Log Model. If the simulation does not involve
viscoelastic material effects you can use the option None for this parameter.
Temperature Equation
This parameter specifies the type of energy equation to be solved in the problem. AcuSolve
supports the Advective-diffusive equation for transport of temperature. If the problem does not
involve temperature as a variable then this equation can be set to None.
Abs. temperature offset
This parameter specifies the offset to convert to absolute temperature units. If specified, this
value is added to all temperatures specified in the simulation.
Radiation equation
This parameter specifies the addition of radiation type. AcuSolve supports Enclosure type radiation
equation. If the problem does not involve radiation then this equation can be set to None.
Species equation
This parameter specifies multi-species transport equation type. AcuSolve supports Advective
Diffusive equation for multiple species transport.
Num. species
This parameter specifies the number of species present in the problem and is used only with
Advective Diffusive species equation.
Turbulence Equation
Specifies type of Turbulence equation to be solved. AcuSolve provides a wide range of options in
turbulence modeling. The available options for turbulence modeling are listed below.
Mesh Type
This parameter specifies the type of mesh motion involved in the simulation. The following are the
available options with the mesh type.
• Fixed: The mesh motion is fixed which means there is no mesh movement of any kind.
• Fully Specified: This option is used when the mesh motion is completely specified using the
MESH MOTION command.
• Arbitrary Mesh Movement (ALE): Mesh motion with this option is based on the Arbitrary
Lagrangian-Eulerian (ALE) technique involving the benefits of both the Lagrangian based
and Eulerian based mesh motions.
External Code
This option of on/off specifies whether an external solid/structural code is coupled with AcuSolve
to solve a coupled simulation. Forces, moments and heat fluxes are communicated to the solid/
structural code while displacements are returned to AcuSolve. The code coupling interface (CCI)
library handles the required communication and interpolations. Once a solid/structural code is
linked to this library, that code can use CCI functions to run in conjunction with AcuSolve.
Particle Trace
This option of on/off specifies whether the AcuTrace code is coupled with AcuSolve to solve a
particle-flow problem. Flow quantities like velocity, pressure, temperature, species and so on are
communicated to AcuTrace and particle states and source terms are returned to AcuSolve.
Running average
This option of on/off specifies whether tocreate running average solution fields. If running average
is set on, then a running average field will be created for every equation field in the problem.
The creation of running average fields does not itself have any effect on the solution. A running
average field is defined by
The goal of the Auto Solution Strategy branch is to completely automate the specification of all the
solution strategy commands based on the minimum possible parameters defined in this section.
Analysis Type
This parameter specifies the type of analysis to be used. There are two types of analysis Steady
and Transient.
Final Time
This parameter is active only with Transient analysis type and specifies the final time of the run.
The run will terminate when the analysis time reaches the final time. The run may terminate due
to other criteria. If this parameter is zero, this option is ignored.
Convergence Tolerance
This parameter specifies the tolerance value for residuals and the solution is assumed to be
converged once all of the residuals of all parameters (such as velocity, pressure, eddy viscosity
and temperature) measured is less than this convergence tolerance. The smaller the convergence
tolerance the more accurate the solution is, but there is larger simulation time.
Relaxation Factor
This parameter is used to smooth the solution convergence for hard problems. Increasing the
relaxation factor smoothens the solution convergence. If zero, this parameter is ignored.
Flow
This parameter specifies whether or not to solve the flow equations. This parameter is set to on
if the flow equations are to be solved in the simulation and set to off if the solution from flow
equations is not desired.
Temperature
This parameter specifies whether or not to solve the temperature equations. This parameter is set
to on if the temperature equations are to be solved in the simulation and set to off if the solution
from temperature equations is not desired.
Temperature Flow
This parameter is active only when both Flow and Temperature are on and specifies whether
or not to solve thermal-flow problems fully coupled. To solve a thermal-flow problems with a
fully-coupled strategy Temperature flow is set to on, otherwise these problems are solved with
a staggered strategy. The choice generally depends on degree of coupling (that is buoyancy)
between the momentum and thermal equations. A forced convection problem can be solved
staggered, while a free convection problem usually benefits form the more expensive coupled
approach.
Enclosure Radiation
This parameter specifies whether or not to solve the radiation equations. This parameter is set to
on if the radiation equations are to be solved in the simulation and set to off if the solution from
radiation equations is not desired.
Species 1
This parameter specifies whether or not to solve the corresponding species equation. Depending
on the number of species specified under the Num. species parameter in Problem Description all
of the species can be individually provided whether or not to solve the species equation.
Turbulence
This parameter specifies whether or not to solve the turbulence equations. This parameter is set
to on if the turbulence equations are to be solved in the simulation and set to off if the solution
from turbulence equations is not desired.
Mesh
This parameter specifies whether or not to solve the mesh equations. This parameter is set to
on if the mesh equations are to be solved in the simulation and set to off if the solution is not
affected by mesh displacement.
Viscoelastic
This parameter specifies whether or not to solve the viscoelastic stress equations. This parameter
is set to on if the viscoelastic stresses are to be solved in the simulation and set to off if the
solution is not affected by viscoelastic stresses.
External Code
This parameter specifies whether or not to solve solid/structural equations with external code.
Particle Trace
This parameter specifies whether or not to solve the particle equations in a coupled particle-flow
interaction using AcuTrace.
Material Model
All of the properties of materials to be used in the simulation are defined in this branch.
Every volume (fluid or solid) must reference a material model. The same material model may be shared
by any number of volumes. Any material models that are not referenced are ignored. AcuConsole has
three pre-defined commonly used material models, Air, Aluminium and Water.
You can also import most of the standard material models from a repository. Apart from these, you can
define a completely new material by right-clicking Material Model > New.
You can then define the individual properties of each material model. In the material model definition,
the first parameter is Material Type which allows you to choose if the material defined is a fluid or solid.
A material model must contain all of the necessary properties associated with the equations being
solved in the problem. These equations are defined in the Auto Solution Strategy. A material model
which has Material Type as fluid must include the following parameters:
Density
Any equation
Specific Heat
Temperature equations
Viscosity
Flow and turbulence equations
Conductivity
Temperature equations
Viscoelastic
Viscoelasticity equations
Diffusivity*
Species equations
Porosity
Flow with porous material
Diffusivity properties can be defined individually for each species when solving the species equations, as
shown below.
In general, it is advised to always define the viscosity properties even if the flow equations are not
being solved. Some material properties (such as the conductivity model with a constant Prandtl
number) may require the viscosity defined.
Since only energy and mesh displacement equations are defined with Material Type-Solid, the
corresponding material models need definition of only Density, Specific Heat and Conductivity.
A fully defined material model can be exported so that it can be used for future simulations. To export a
material model, right-click on the model and select Export.
This section details the geometry parameters such as geometry import and graphics window controls.
It also involves discussion on boundary condition types, volume and surface parameters available in
AcuSolve.
Geometry Import
AcuConsole accepts different geometry file formats such as Parasolid, Pro-E part file, Pro-E assembly
file, CATIA part file and CATIA assembly file STL.
To import a geometry file, click File > Import. After you select a file to open the Import Geometry
dialog opens.
Geometry Size in X/Y/Z The first three rows in the dialog provide the size of the geometry
in X, Y and Z directions.
Geometry Units AcuConsole also provides the option of scaling the geometry
being loaded. You can specify the scale factor in the box next to
geometry units. By default a value of one is provided which means
the geometry is loaded without scaling and AcuConsole considers
all the dimensions to be SI units.
Geometry Resolution [Low- A slider next to Geometry resolution [Low-High] is used for setting
High] the resolution of geometry being imported. Geometry will be
imported with low resolution if the slider is to the extreme left
and high resolution if slider is to the extreme right. If a simple
geometry is being loaded then setting this option to Low speeds
up the importing process. If the geometry has a complicated
topology with minute sharp edges and so on, then it is best to
set to high resolution so as to preserve the exact topology of the
geometry.
Volume Group Option and You are provided with different options for grouping the volumes
Surface Group Option and surfaces.
Read Edge Data You can choose to import edge data with the geometry. On for this
option imports the edge data.
Edge Group Option This parameter is available only when Read edge data is On.
It provides two options of All in Default and By attributes with
functioning similar to that of Volume/Surface Group Option.
Full Geometry Check and By putting the full geometry check option to On the geometry will
Verbose Level be checked. Any errors found will be displayed in the information
window. Increasing the Verbose level provides more details of the
import process.
Full Voids in the Geometry If any part of the geometry has only surfaces and no material
filled in, then turning this option to On fills the enclosed volume
with material.
Suppress Small Features This option should be used if features smaller than a particular
value need to be suppressed in the geometry. Having Suppress
small features On provides two more options. One is a small
feature tolerance type and the other is the tolerance.
AcuConsole can perform many useful transformations on the model for better viewing. View Manager,
which is located adjacent to Data Tree, has user friendly commands for frequently used transformations.
Home icon
Changes all of the model properties such as size and view to initial state.
Fit icon
Translates the model to the center and scales it so that it fits the Visualization Area
Snap icon
Snaps the object to the closest xyz transformation.
Snap Z icon
Snaps out of screen vector to closest x, y, z axis.
Zoom icon
Switches the Visualization Area into a rubber-band state. The Visualization Area will zoom in the
area selected by mouse press and drag.
Center icon
Moves the clicked surface to the center of the Visualization Area. First select this icon and then
click on the surface to be moved to center of Visualization Area.
Annotation icon
Adds annotation to the Visualization Area.
Views
The Views menu shows the model from the corresponding axis view.
Figure 60:
The Rotate 45 degrees menu rotates the model 45 degrees over the corresponding axis.
Figure 61:
Mouse Controls
AcuConsole is equipped with the following mouse control features for easy transformations.
Scroll
Zoom-in and zoom-out
Left mouse press and drag
Rotate the model
Right mouse press and drag
Zoom-in and zoom-out
Middle mouse press and drag
Translate the model
All of the problems in fluid dynamics are governed by the same standard conservation equations.
What makes the solution of each problem unique is the set of boundary conditions. Hence boundary
conditions have great impact on the accuracy of the simulation. This section discusses the variety of
boundary conditions generally used in a CFD simulation.
Velocity Type
A user prescribed velocity is imposed on all of the nodes of the surface. Mass flux through the
surface depends on the variation of the prescribed velocity profile.
Pressure Type
A user prescribed pressure values are imposed on all the nodes of the surface. This type of
condition is useful when velocity or flow rate is not known, such as buoyancy driven flows.
Mass Flux
A user prescribed mass flux rate is imposed on the surface. A boundary layer profile for velocity
and turbulence fields is calculated based on the distance from no-slip walls and estimated
Reynolds number. In AcuSolve the profile is recomputed at each time step such that the
deforming meshes are properly accounted for in the calculation.
Flow Rate
A user prescribed flow rate is imposed on the surface. A boundary layer profile for velocity and
turbulence fields is calculated based on the distance from no-slip walls and estimated Reynolds
number. In AcuSolve the profile is recomputed at each time step such that the deforming meshes
are properly accounted for in the calculation.
Average Velocity
An average velocity is imposed at the surface. A velocity profile similar to the mass flux boundary
condition is used in this case.
and
Volume Parameters
The volumetric parameters such as Material Model, Body Force and Mesh Motion are defined in the
Global Tree. These parameters are assigned to the appropriate volumes using Volume Parameters in the
Model Tree. These parameters can be accessed by clicking Model > Volumes. The volume parameters
are categorized into three parts, namely Element Set, Element Output and Volume Mesh Attributes.
Element Set
This command defines a set of elements for the given volume in the problem. All elements in
the volume have same material model, body force, and so on. Element set defines the following
parameters.
Medium
This parameter specifies whether the volume is of type Fluid, Solid, Thermal Shell or None.
A fluid element set solves for all equations specified in the Auto Solution Strategy. A solid or
Thermal Shell type element set solves only the temperature and mesh displacement equations,
while all other equations such as flow, turbulence and species are ignored. A type of none
effectively deactivates the element set and all of its surfaces, the elements do not contribute to
any equation. The nodes may still be moved through the use of the Mesh Motion parameter.
Quadrature
This parameter defines the quadrature rule used to integrate element contributions.
The actual quadrature rule used depends on the element shape.
Table 6:
Four node tet 4-point 1-point 4-point nodal 4-point nodal 20-point
All of the points in quadrature rules except for nodal satisfy the following properties:
• Are Gauss quadrature
• Are inside the elements
• Have positive weights
• Are node order independent
• Use minimal number of points for a given integration order
Appropriate surface and edge quadrature rules corresponding to these interior rules are used
automatically for those surface and edges associated with a given element set.
Material Model
This parameter assigns the material properties using the Material models defined in Global
section. Only those models defined in the Material Model section in Global are applicable for this
parameter.
Reference Frame
This parameter specifies frame of reference for the volume. Only those reference frames defined
in Reference Frame section in Global are applicable for this parameter.
Residual Control
Residual control provides stabilization of the equations by adding additional diffusion in areas
where unresolved gradients are present. This option improves the robustness of the solver and
should be left on.
Oscillation control
Oscillation control triggers the use of an operator that is designed to prevent over/under shoot of
scalar equations. This option improves the accuracy and robustness of the solution and should be
left on.
Mesh Motion
This parameter specifies the type of mesh motion to be imposed on the element set. Only those
Mesh Motion models defined in Mesh Motion section in Global are applicable for this parameter.
Mesh Motion Precedence
This parameter specifies precedence of the mesh motion command above, in case of conflict from
other commands. The higher the precedence value the higher the priority. It is active only when
the Mesh Motion parameter is set to a mesh motion model.
Element Output
This command specifies parameters for the output of the average of quantities integrated through the
volume. This command then outputs the averaged quantities to the disk. By default, AcuSolve writes
out this output at the last time step of the simulation.
Surface Parameters
Surface parameters specify the boundary conditions pertaining to the surface.
Type
Specifies the type of boundary condition for the surface.
Active type
Determines which surfaces in this surface set will have boundary conditions imposed by this
command.
All: All surfaces in this set are active.
where
=-1 for inflow
=1 for outflow
Back flow Conditions
Specifies whether to specify nodal boundary conditions for scalar variables on nodes on outflow
surfaces where there is flow into the fluid domain. This parameter is active only for Outflow and
Far-Field boundary types.
Temperature back flow type
TSpecifies how the temperature condition is defined under back flow condition. Options for this
parameter are shown below.
• Value: Specifies a value for temperature.
• Area Average: Average of solution over entire outflow surface.
• Exiting Area Average: Area average of solution over subset of outflow surface that has flow
exiting the domain.
• Bulk: Bulk average of solution over entire outflow surface.
• Exiting Bulk: Bulk average of solution over subset of outflow surface that has flow exiting
the domain.
Flexible Body
Specifies the name of Flexible Body. It is active only with Flexible Body mesh displacement type.
Nodal x modes
Specifies the array of node number and x-component Eigen vector values. It is active only with
Flexible Body mesh displacement type.
Nodal y modes
Specifies the array of node number and y-component Eigen vector values. It is active only with
Flexible Body mesh displacement type.
Nodal z modes
Specifies the array of node number and z-component Eigen vector values. It is active only with
Flexible Body mesh displacement type.
Guide Surface
This parameter specifies name of the Guide Surface for the mesh to follow. It is used only when
mesh displacement boundary condition type is Guide Surface.
Mesh Motion
This parameter gives the name of the Mesh motion for determining mesh displacement boundary
conditions. If None this parameter has no effect. This feature is used with Fixed and‘Flexible Body
mesh displacement types.
Surface Output
This command specifies parameters for the output of the average of quantities integrated through the
surface. This command then outputs the averaged quantities to disk. By default, AcuSolve writes out
this output at the last time step of the simulation.
This section details a brief overview of meshing aspects in general and then provides various meshing
attributes available in AcuConsole. This section also provides details on meshing process in AcuConsole.
Mesh Overview
The basis for any numerical analysis is solving the governing equations over a set of discrete volumes
that enclose the physical domain. Mesh generation or grid generation is the process of generating such
discrete volumes (elements) of the geometry. The conservation equations of flow integrated over these
discrete volumes is the basis of CFD. The accuracy of the solution is affected by the quality of the grid.
There are different shapes of the elements that can be used for CFD analysis. However, the following
are some of the commonly used shapes.
2D Elements
3D Elements
Mesh Quality
Since the mesh holds the key for convergence and accuracy of the solution it is important to evaluate
the quality of mesh. The following are some of the quality metrics to determine the mesh quality.
Skewness is a measure of how much the element’s shape deviates from the optimal element shape.
Among the different ways of skewness measurement, AcuSolve uses equi-area or equi-volume
skewness which is defined as
Skewness ranges from zero (for equilateral triangle and tetrahedron) to one. The lower the skewness,
the better the convergence.
The aspect ratio of an element is defined as the ratio of the longest edge length of the element to the
shortest edge length.
The aspect ration can range from 1 to . The smaller the aspect ratio the better the mesh quality is.
Although different numerical analysis have different requirements on the minimum mesh quality,
AcuSolve is capable of dealing with the bad elements effectively. The only necessary requirement for
AcuSolve is that all of the elements have a positive jacobian, that is, elements with positive volumes.
In the regions of the flow with large gradients, the mesh has to be fine enough to capture those sharp
gradients. The boundary layer near the wall is one such region in the flow with large gradients. Hence
the proper resolution of the boundary layer is important for accurate solutions. Finer mesh at the wall
with tetrahedrons would require larger computation time. However, by exploiting the fact that at the
wall gradients along the stream are less compared to the transverse gradients, it is advantageous to
use high aspect ratio prism elements for boundary layer. To avoid inaccuracies due to abrupt change in
mesh sizes the mesh size has to transition smoothly from fine mesh at boundary to much coarser mesh
at the core. Hence, the boundary layer mesh size is gradually increased from boundary to the core by a
ratio (>1) called Stretch ratio, to approximately match mesh size in the inner core.
Since it is noted that mesh has to be fine at the boundary layers, the next question is how fine the
mesh should be. Taking into account the first cell height and flow velocity a non dimensional parameter
+
y is defined as
+
With a higher y the approximation of numerical solution cannot capture the boundary layer well.
+ +
However, with a smaller y the boundary layer is captured very well. Hence the y plays a significant
role in solution for the near wall region.
To accurately capture the near wall effects, a sufficient number of elements need to be present inside
+
the viscous sub layer (typically y < 10). However to reduce the computational efforts, the near wall
sub layers are modeled using wall functions, which uses semi-empirical approximations for the viscous
+
sub layer. The appropriate y values depends on the turbulence models used and the type of wall
functions.
Meshing in AcuConsole
AcuConsole uses AcuMeshSim as the mesher. AcuMeshSim operates directly on the underlying CAD
model. When it performs surface meshing, it actually queries the underlying CAD kernel to determine
where the node should fall in parametric space on CAD face. This ensures precise representation
of the CAD. The input CAD model must be a non-manifold model. A non-manifold model is created
automatically when you import a CAD model into AcuConsole. The process used to create the non-
manifold representation of the geometry resolves overlapping volumes, eliminates duplicate faces,
knits surfaces to form solids (if directed to), and performs a union between adjacent components. In all
cases, AcuMeshSim requires a clean CAD. This means that the geometry must not have unjoined edges,
unattached surfaces, gaps between common edges of surfaces, invalid spline representations of faces,
and so on.
Mesh attributes can be set on a global (whole model) and/or local (model entity) basis. An attribute
specification applies to the closure of an entity, so if a size is specified on a model face the model edges
and model vertices on the boundary of the face will inherit the same size specification. However, a
separate size specification may be given for the edges and vertices in each case. If conflicting mesh
sizes are specified, regardless of whether they are from the same type of mesh attributes or not,
AcuMeshSim will pick the smaller of the resulting sizes.
All of the parameters related to the meshing can be viewed by clicking MSH in the Data Tree Manager.
The Global Mesh attributes represent mesh settings that are applied to the entire model.
a radial array of elements is created around the sharp edge to provide better resolution of flow
features. The sweep angle is used to control how many degrees each radial division spans.
Figure 78: Meshing with Boundary Layer Blend On, Sweeping Angle = 15°
The Meshing Process attributes represent settings that are applied during the meshing process.
AcuConsole provides you with the ability to refine the mesh at particular regions in the domain
depending on the flow gradients. Mesh size can be specified as a function of location. Geometric shapes
such as spheres and box cylinders can be drawn irrespective of the CAD model and the mesh sizes are
specified for the volume enclosing these Zone Mesh Attributes.
A new zone mesh can be created by right-clicking Zone Mesh Attributes > New.
Every geometric model in AcuConsole contains one or more volume groups. The meshing of each of
these volumes can be customized using the volume mesh parameters.
If Volume Mesh attributes are not specified for any of the volumes then the global mesh attributes will
be applied for such volumes.
Similar to volumes, meshing of the surfaces can also be customized by using Surface Mesh Attributes.
If Surface Mesh Attributes are not specified for any surface group, mesh attributes for that surface will
be inherited from its parent volume group or the global mesh attributes. In the event that all three
mesh attributes are specified (global, parent volume and surface) AcuConsole uses the smallest of the
three values.
AcuConsole writes all of the mesh parameters to an .xml file with an .ams extension. Before the start
of the meshing process, the geometry data of the model is written to a folder named CAD.DIR in the
working directory. The .ams file and the CAD.DIR are required sets of data for the mesh process.
After specifying the mesh attributes the meshing can be launched by clicking Tools > Generate
Mesh. Alternatively, it can be launched from the toolbar by clicking , which opens the Launch
AcuMeshSim dialog.
After setting the above parameters clicking OK starts the meshing process. During the meshing process
certain files are written to the MESHSIM.DIR folder. After the meshing is done, an .arm (AcuSolve raw
mesh) file is created. Once the meshing is done MESHSIM.DIR can be deleted if required, which has
no effect on the mesh or the simulation. The .arm file is an .xml file pointing to different mesh files in
MESH.DIR. The complete mesh of the model can be transferred to a different machine using the .arm file
and MESH.DIR.
4.2.5 Solving
This section details the AcuSolve launch settings, input file structure, checking the compatibility of the
parameters using AcuPrep, parallel processing model, various solution files generated by AcuSolve and
the log file structure.
After meshing the model and setting the boundary conditions the model setup is ready for solving.
The solver can be launched by clicking Tools > AcuSolve or by clicking the icon.
Problem Name
Specifies the name of the problem. All of the files during the solutions are created based on this
name.
Problem Directory
Specifies the directory in which the input files for the problem are present.
Working directory
Specifies the directory in which the solution files are written.
Solver Modules
Specifies the type of solvers to be run. Solver modules include Prep, View, Solve, Prep-Solve and
all.
Prep
Performs only the pre-processing of the model using AcuPrep module.
View
Performs only pre-processing of radiation parameters using the AcuView module.
Solve
Performs only the solver module of AcuSolve.
Prep-Solve
Performs the pre-processing of the flow parameters and solves the problem.
all
Performs all the modules of pre-processing and solving modules.
User Library list
If the simulation uses a user-defined function (UDF) then the name of compiled file should be
placed for this parameter.
User argument list
Specifies the list of arguments to be passed on to the User Library list.
Number of Processors
Specifies the number of processors for running the problem.
In AcuSolve a simulation can be restarted from a previous solution by using the Restart tab in the
Launch AcuSolve dialog. A description of the parameters for restart analysis is provided below.
From problem
Name of the problem from which restart has to be initiated.
From directory
Directory in which the previous solution files are present.
From run
Specifies the number of the run from which restart has to be initiated. A value of zero uses the
last run available.
From time step
Specifies the time step from which the run has to be initiated. A value of zero uses the last
available time step.
Reset time step
Flag specifying whether the time step numbering should be followed from the previous solution.
Reset time increment
Flag specifying whether to use the time increment from the previous problem. Using on uses the
time increment mentioned in the present run.
AcuConsole is the graphical user interface (GUI) wrapper for the AcuSolve solver.
After launching AcuSolve a folder is written out named MESH.DIR containing mesh files and an input file
containing the solver commands. This input file can be edited directly for any appropriate changes in
the problem. The commands have the following general syntax.
COMMAND (qualifier){
Parameter1=value1
….
parameterN = valueN
}
Where COMMAND is the name of the command, such as ANALYSIS or MATERIAL_MODEL, qualifier is
a mandatory qualifier of the command, parameter1 to parameter are the optional parameters of
the command and value1 to valueN are the corresponding values of the parameters. Below is the
description of a simple pipe flow with the comparison between GUI and input file commands.
Figure 83:
Figure 84:
Figure 85:
Figure 86:
Figure 87:
As an alternative to the GUI based launch of AcuSolve, once you have the input file and mesh files
(MESH.DIR) the AcuSolve solver can be launched in batch process from the AcuSolve command prompt.
To open the command prompt in Windows click Start > All Programs > Altair HyperWorks >
AcuSolve > AcuSolve Cmd Prompt.
The command used for launching AcuSolve is acuRun. To know the various options of acuRun use the
command acuRun –h. By using the appropriate options you can launch the problem in batch mode.
For example, to launch a problem with the problem name pipe_laminar with the corresponding input
file pipe.inp with 12 processors the following command can be used
AcuPrep
After launching AcuSolve, the input file is first read by AcuPrep and all of the higher level commands
are translated into the machine language that AcuSolve can understand. During the AcuPrep process all
of the syntax errors or the incompatible commands are pointed out to you. A typical AcuPrep log looks
similar to the image shown below.
Domain decomposition is used to distribute the elements and nodes to different processors. Message
Passing Interface (MPI) is used to communicate between distributed memory computers and shared
memory data copy is used between subdomains of a single shared memory box. This ensures minimum
communication costs. All I/O during the computation is done in parallel. AcuSolve performs profiling of
each run to determine where the majority of the CPU/elapsed time is spent during the simulation. The
results of this profiling are printed to the .log file at the conclusion of each run and can be reviewed to
assess the performance of the code on your computing system. The results from a representative .log
file will be reviewed to explain the output in the following sections.
During the solving process AcuSolve writes the relevant solution files to a folder named as ACUSIM.DIR.
Depending on the nodal and element output parameters corresponding solution files are written to the
disk. The post-processing of the solution in AcuFieldView requires access to the files in ACUSIM.DIR.
During the solution process AcuSolve writes out the information to the .log file with the .log
extension.
This .log file contains the information on failure, if there is any, and solution statistics such as residual
ratios, solution ratios, CGP and GMRES iterations and Elapse time statistics. A typical sample of such
information is shown below.
At the end of the final time step overall statistics on time and memory consumption are listed.
4.3 Post-Processing
This section details the post-processing tools of AcuSolve, namely AcuProbe, which gives graphical
representation of parameters and AcuFieldView, which gives the detailed description of simulations,
such as plots and animations.
4.3.1 AcuProbe
During the process of solution you can track the solution behavior in terms of residual ratios. AcuProbe
can be launched from AcuConsole by clicking Tools > AcuProbe.
AcuProbe provides various statistics for tracking and plotting, as shown below.
You can plot a particular variable by right-clicking and selecting Plot or can you can plot all of the
variables in a category by right-clicking and selecting Plot All.
Data Types
Below is a description of the data types available in AcuProbe.
Run data
Contains the general simulation parameters such as Time, Time Increment, CPU Time and Elapsed
Time.
Residual ratio
Contains the residual ratios for each of the flow parameters such as pressure, velocity, eddy
viscosity and temperature. These ratios can be plotted with three different options Initial, Final
and All,
Solution ratio
Contains the solution ratios for each of the flow parameters such as pressure, velocity, eddy
viscosity and temperature. These ratios can be plotted with three different options Initial, Final
and All.
Surface output
Contains surface integrated values of variables such as pressure, momentum, mass flux and wall
shear stress.
Surface statistics
Contains the surface statistics of minimum and maximum of variables such as pressure,
momentum, mass flux and wall shear stress.
Residual Types
Below is a description of the residual types that are available.
Initial
Provides the residual ratios at the start of each time step.
Final
Provides the residual ratios at the end of each time step.
All
Provides the residual ratios for all of the staggers within a time step.
Apart from the variables provided by AcuProbe, you can define the specific variable not mentioned in
AcuProbe by using the user-defined function (UDF). A UDF can be defined by clicking Edit > User-
Function.
4. For getting x-traction on surface Body, right-click x_traction under Surface Output > Body and
select copy name and paste it in the text area.
5. Click Apply.
AcuProbe will plot the value of Cd.
6. The UDF can be saved by clicking Save so that it can be retrieved at a later time. Saved UDFs can
be imported in AcuProbe by clicking File > Import Curve/UDF.
4.3.2 AcuFieldView
AcuFieldView is a dedicated CFD post-processor for AcuSolve, an OEM version of FieldView, and
supports only AcuSolve results.
AcuFieldView can process both steady and transient data with flow animations, stream lines, boundary,
coordinate and iso surface contours. It can also be used for 2D plots of a parameter and point probes
for quantitative analysis of the data. You can integrate the parameters over a given surface and you can
also define new variables of particular interest.
Loading data
You can load AcuSolve data into AcuFieldView using AcuSolve .log files or using the .fv files created
from AcuTrans.
AcuFieldView can directly read the AcuSolve .log files. Log files can be read by clicking File >
Data Input > AcuSolve [Direct Reader]. AcuSolve .log files should be accompanied with the
corresponding solution files in ACUSIM.DIR.
The results can be read into AcuFieldView using the .fv files. To create .fv files use the following
command in the AcuSolve command prompt.
The above command creates .fv files, which does not require AcuSolve solution files of ACUSIM.DIR.
Once the .fv files are created they can be read into AcuFieldView by clicking File > Data Input >
AcuSolve [FV-UNS Export].
This section details the various tools and utilities available in AcuSolve that make the solver user-
friendly.
4.4.1 Tools
Aside from providing robust solver for CFD simulations, AcuSolve also provides easy to use tools for
smooth work flow.
Package Mesh
AcuConsole supports the import of a mesh from third party software. AcuConsole accepts .arm
(AcuSolve raw mesh) files for the import of mesh. Many of the commercially available meshing tools
generate the mesh in the format directly compatible with AcuConsole. If the mesh file is not in the .arm
format you can still package the separately available coordinate and connectivity files, surface files, and
so on, using the Package Mesh Tool.
For a given directory, you can selectively extract the required files by using the corresponding get
command. You can also extract nodal boundary condition (NBC) files and periodic boundary condition
(PBC) files. After extracting the required files, you can package them into an .arm file using the pack
command in the tool. This creates the .arm file that can be imported into AcuConsole.
Eigenmode Manager
Eigenmode Manager is used for P-FSI type of fluid structure interaction simulations in which the
eigenmodes extracted from a structural solver (such as RADIOSS) are projected on the flexible body in
AcuSolve. Eigenmode Manager also provides options to view the mode shapes before projection.
JMAG Manager
JMAG is a comprehensive software suite for electromechanical equipment design and development. The
temperature results from such an electromechanical simulation can be exported as .nas files and those
files can be used as heat sources for flow simulation of the same problem in AcuSolve using the JMAG
Manager tool.
From the JMAG Manager, you import .nas files. If the JMAG simulation has a coordinate system different
from that of AcuSolve, you can use the Transform Coordinates dialog to adjust the results for the
AcuSolve system before importing the results.
You can transform the coordinates with the options of scale, shift, center and rotate.
From the JMAG Manager dialog you can use the transfer to element set option to select the element set
on to which the JMAG results are to be transferred. The Show tab in the dialog can be used to visualize
the heat sources distribution.
Export
Clicking Tools > Export provides the option to create the setup files, which includes input file (*.inp)
and mesh files (MESH.DIR).
Generate Mesh
This tool launches AcuMeshSim, the meshing module of AcuSolve. The tool provides options to select
the mesh type (volume/surface mesh) and output file format.
AcuSolve
This tool launches the AcuSolve solver command. You can provide the parameters such as problem
name, problem directory and working directory. You can also choose the specific solver modules to be
launched and create input files without launching the solver by setting Launch AcuSolve = off .
PBS Queue
Portable Batch System (PBS) is the name of a computer software that performs job scheduling and
allocates computational tasks, that is, batch jobs, among the available computing resources. It is often
used in conjunction with Unix cluster environments. PBS is available in two forms, OpenPBS and PBS
Professional. OpenPBS is the unsupported original open source version of PBS, while PBS Professional is
a commercial product offered by Altair only. PBS job scheduler can be used for scheduling AcuSolve jobs
on cluster and the tool PBS Queue launches such tasks.
LSF Queue
Load Sharing Facility (LSF) is a commercial computer software job scheduler sold by Platform
Computing. It can be used to execute batch jobs on networked Unix and Windows systems on many
different architectures. AcuSolve jobs can be submitted using the LSF Queue but you need to have a
license for LSF, which is not provided by Altair.
AcuProbe
This tool launches AcuProbe, which is used as one of the post-processing tools in AcuSolve. AcuProbe
provides the options to plot the residuals, surface outputs, surface statistics, element output and time
history output. You can also plot user specific variables of interest using the user function, which is
basically a python script.
AcuOut
AcuOut is a post-processing tool in AcuSolve. This tool converts the AcuSolve data into different formats
such as Table, EnSight, H3D and CNGS. You can choose the specific data belonging to a time step and
specific variable to be converted to other formats.
Directory Manager
The directory manager module AcuDmg helps manage the working directory (ACUSIM.DIR by default).
This tool is helpful for reviewing the available data and deleting all of the unnecessary data for that
particular simulation. By frequently using this tool you can optimize the usage of disk space on your
machine. This feature is useful while running large transient simulations.
AcuFieldView
AcuFieldView is a light weight version of the post-processing tool of FieldView from Intelligent Light.
AcuFieldView works with AcuSolve only. There is no additional cost to get AcuFieldView as it comes with
the AcuSolve distribution.
EnSight
This tool launches installed versions of EnSight reader. EnSight is a product of Computational
Engineering International (CEI). It provides a set of tools to help with many types of engineering
analysis, visualization and communication. AcuConsole has a direct reader for EnSight. Please note that
EnSight reader is not shipped with the AcuSolve installation package.
AcuTail
AcuTail allows a file to be monitored. Instead of just displaying the last few lines AcuTail displays the
lines and then monitors the file. As new lines are added to the file by another process AcuTail updates
the display. This is particularly useful for monitoring log files.
AcuTrace
AcuTrace is a particle tracing post and co-processor to AcuSolve. AcuTrace uses the time discontinuous
Galerkin method to compute particle traces for flow fields generated by AcuSolve and optionally
provides source terms to AcuSolve for coupled fluid particle simulations.
AcuFwh
AcuFwh is an acoustic post-processor which is based on the Fowks-Williams-Hawkings method.
AcuSolve outputs the acoustic sources on the FWH surfaces that you set. AcuFwh then propagates these
acoustic sources to the far field observers.
Material Models
The Material Model under the Global tree displays three material models (Air, Aluminium, Water) by
default. You can create material models or you can import the standard material model using this tool.
The list of material models include various commonly used fluids and solid materials.
Plugin Manager
AcuSolve can be customised and streamlined for a particular type of CFD analysis. The data tree
structure can be customised to display new labels with only necessary parameters under them.
Such a module can be made with Python scripting and this tool is used to import such modules into
AcuConsole.
The plugin can be used either only for the current session or it can be loaded at the start of every
session using the option type. After browsing the file, using the Load Plugin option loads the
corresponding module. Some of the regularly used plugins such as Rotary Machinery and Virtual Wind
Tunnel are provided with the AcuSolve installations.
Project Solution
When running a simulation that is complex, say in terms of huge mesh, it will be wise to have
appropriate initial conditions for faster convergence. For this, one way is to run the simulation on
coarser mesh and use the nodal output of this simulation as initial condition for complex simulation.
Project solution helps in importing such data and projects that solution onto the complex problem,
thus helping in faster convergence. You can also select specific variables to be projected on the final
simulation.
Group Manager
Sometimes the naming of the parts is done in the CAD modeler itself. Any CAD modeler tracks the
changes made to a geometry by assigning an identity to each feature it created. Group manager helps
you in taking advantage of such CAD naming for volume and surface grouping in AcuConsole. Group
Manager also helps you in reorganizing the parts and its attributes.
You can group the volumes and surfaces based on CAD model attributes such as the part names,
assembly names, tag names or even based on the attributes of the adjacent volumes.
For a given geometry model with two parts whose names blower_impller_1 and blower_main_1 are
defined in the CAD model itself, these part names can be used as volume groups in AcuSolve. This can
be achieved by creating a rule for such grouping using the group manager by selecting
‘Rule_Type = Volume, Rule Active = on, Part Name = (.*) and Group = ‘\P1’ ‘
The above rule prompts AcuConsole to look into the geometry model for the part names and the
corresponding volumes are named the same as those of the part names. At the end of this operation
AcuConsole creates two volumes with names blower_impeller_1 and blower_main_1.
AcuConsole can also allow for the use of partial part names (viz. impeller, main) as the volume names.
This can be achieved by creating the following rule
‘Rule_Type = Volumes, Group = \P1, and Part Name = blower_ (.*) _1. ‘
The above rule prompts AcuConsole to look into all of the part names and pick up the part names that
matches the string blower_ some string_1 resulting in two such part names blower_impeller_1 and
blower_main_1, then the corresponding volumes are named as impeller and main.
The volume names can also be set to impeller_blower and main_blower. You can then use the following
rule parameters.
The above rule prompts AcuConsole to look into all of the part names with the pattern string 1_string
2 resulting in two such parts, for which the \P1 corresponds to first matching string, that is, string 1
and \P2 corresponds to the second matching string, that is, string 2. There will be two volumes with
names impeller_blower, main_blower corresponding to the two parts with names blower_impeller_1,
blower_main_1, respectively.
For the above case all of the surfaces belonging to a particular part as one surface group, then the
following rule can be used.
Rule_Type = surfaces, Active Type = on, Group = \P1 and Part Name = (.*)’
By using the Rule_Type = surfaces, AcuConsole groups only the boundary surfaces of a part. Grouping
of the internal faces can be done separately using Rule_Type = Internal Faces.
AcuConsole allows for grouping based on a variety of CAD model parameters such as assembly name,
tag name, tag value and volumes. For internal faces, in addition to the above parameters, additional
parameters such as neighboring part name, neighboring assembly name and neighboring tag name are
also available. These parameters can be activated using the columns option.
Similar to the syntax of \P1, \P2 for matching strings in part names, similar syntax can be used for
other variables, as shown below.
\P1, \P2
Part name
\A1, \A2
Assembly name
\T1, \T2
Tag name
\TV1, \TV2
Tag value
\L1, \L2
Levels
\V1, \V2
Volume
\NP1,\NP2
Neighbor part name
\NA1,\NA2
Neighbor assembly name
\NT1,\NT2
Neighbor tag name
\NTV1,\NTV2
Neighbor tag value
\NV1,\NV2
Neighbor volume
Another useful feature of the Group Manager is that apart from grouping the volumes and surfaces,
the meshing parameters and the boundary conditions of already existing volumes or surfaces can be
applied to the new grouping. You can select such volumes and surfaces under the Copy Settings From
command in the Group Manager.
All of the defined rules can be exported using the Export command which saves the file in AcuConsole
Group file (AGF) format and the Import command can be used to import the already defined rules set
(.agf file) in a different simulation.
AcuConsole has a great feature of defining all of the variables in a single panel and then using them in
the model setup. This feature is useful while performing studies such as grid refinement and time step
sensitivity.
For example, while performing a grid sensitivity study, you can set all of the mesh parameters in the
Variable dialog and use them in different panels. Using this feature makes the simulation setup process
simple and less error prone.
The Variable Manager can be opened by clicking Edit > Variable List. Variables can be defined using
this dialog and can be used in panel areas when entering floating point and integer values. The image
below shows the Variable Manager dialog with four variables defined.
The first step in the process is to define the variables in the Variable Manager dialog. Click Add for
defining a new variable. Rename the variable as desired by clicking in the dialog. Enter the expression
in the Expression column by using equal (=) or colon equal (:=) symbol before it. The expression will
be valid only if any of the two symbols mentioned are used. Equal to (=) calculates the value of the
expression when defined and uses it, whereas colon equal (:=) recalculates the value of the expression
if any relative variable is changed. The third column is the Value column in which the current value
of the variable is shown. Any information related to the variable can be included in the Description
column. For defining a new variable, again click Add as only one variable can be defined in each row.
The first three variables are assigned to floating point and integer values and the fourth variables have
an expression for the calculation of mass flux.
Move up, move down and sort buttons in the Variable Manager can be used for setting the order of
variables. To delete a variable, select the row and click the Del button. Instead of manually defining
variables one by one, they can be imported from a file using the Import button. Similarly the defined
variables can be exported by clicking Export. The variables exported are saved in AcuConsole Material
Library (ACL) file format and can be used for future simulations.
The variables defined can be used in the panels area replacing the integer and floating values. The
image below shows the variable m used for specifying the mass flux at the inlet.
You can customize the layout by setting certain default settings in AcuConsole. In the Preference dialog
you can change certain settings such as fonts and icon sizes, display of units, setting default surface
output, default mesh file type and the color scheme.
General
Specifies the number of recent files to show up in the file menu, icon size, font type, display of
units and width of unit buttons. Allows automatic loading of mesh and appending of the mesh.
Data base
Specifies the driver type, auto flush and database compression option.
Active
Specifies whether or not to default the surface output.
Export
Specifies the maximum size of the array that can be written to the input file and mesh file type.
Vis window
Specifies visualization parameters such as type of transparency and shading, phone crease angle
and feature angle, point size and line width.
Mesh display
Specifies the default display types for volumes, surfaces, edges, nodes and PBCs. These options
are effective if visible entity is set to mesh.
Geom display
Specifies the default display types for volumes, surfaces, edges, nodes and PBCs. These options
are effective if visible entity is set to mesh.
Surface highlight
Specifies the display types of a highlighted surface.
Volume highlight
Specifies the display types of a highlighted volume.
Color scheme
Specifies the default color scheme, outline color and wireframe color.
The field of computational field dynamics (CFD) has seen tremendous advancement in recent decades,
riding on the back of the phenomenal advances in computational science and power.
CFD evolved throughout the 20th century. In the pre-1950s, semi-analytical methods were developed
to solve the complex problems that were still unsolved using analytical solution methods. During
the same time period, numerical methods also advanced and became more mature with respect to
numerical stability, speed and accuracy. Post 1950, with the availability of low cost computing power
and advancement in computer science, researchers all over the world were able to develop complex
CFD models and test them on a variety of applications, enabling the advancement of CFD at an
accelerated scale. High-fidelity CFD solvers running on powerful computers today can accurately model
complex fluid flows, solving the governing equations over billions of grid points, performing trillions of
calculations at speeds unthinkable of a few years ago.
However, access to large amounts of computing power does not necessarily equate to high quality
results from a CFD analysis. Modern CFD solvers are specialized and complex computer programs. The
accuracy of the results provided by the solvers is subject to the accuracy of the inputs. It is necessary
that a complete and correct set of inputs be provided to the solvers so that they can perform their task
as correctly as possible. Some of the essential aspects of these inputs to be provided to the solvers
are discussed in this section. You will be able to apply the basic yet important CFD analysis guidelines
that are discussed to your own applications. Following these guidelines in most cases will ensure that
the results obtained from the CFD analyses are reasonably accurate with only a small margin of error.
However, it cannot be stressed enough that these guidelines should be treated as such, only guidelines,
and not an authoritative set of instructions that will ensure correct analyses every time and for every
application. You are both encouraged and solicited to apply sound engineering judgement to individual
cases, as well as explore validation of your modeling methodology for each application of interest.
A CFD simulation involves solving of the complex Navier-Stokes equations, often coupled with other
equations to determine distribution of flow, temperature and turbulence fields across a discretized
simulation domain. The Navier-Stokes equations are highly nonlinear and are impossible to solve
analytically with available methods. CFD solvers solve these equations numerically, often with a set
of assumptions to simplify the solution and a set of parameters specific to the numerical method
employed, to produce a stable solution. The numerical approach employed can vary from solver to
solver. Also critical to the solution method is the set of constraints that you provide to the solver. This
requires sound understanding of the problem definition and the objective of the simulation. At best,
a CFD solver is a tool to simplify the task of solving the complex flow-field equations and must be
applied carefully for the results to be a true and reliable representation of the actual physical processes.
However, since even the best solver technique is based on a numerical approximation method, every
CFD result has a degree of error and uncertainty introduced in it by these approximations.
Obtaining accurate results from a CFD simulation depends on many factors, such as how well you have
modeled the conditions of interest and how well the CFD software has solved the equations necessary to
model these conditions. In order to minimize the errors associated with a solution you must understand
what these errors are. Some of these common error sources are listed below.
• Modeling error: This type of error occurs because of the difference between the true physical
processes and the equations that are used to model the processes.
• Numerical or Discretization error: This type of error is induced because the model equations are
not solved continuously but on a finite number of grid points in the domain.
• Iteration or Convergence error: This type of error is induced because the solution is iterated to a
finite convergence level, which is not zero.
• Round-off error: This type of error is induced because the numbers are rounded off when stored in
computer memory because of storage limitations.
• Input errors: These are the errors induced due to uncertainties in inputs, such as the choice of
physical model pertinent to the application, boundary condition specification and inaccurate input.
• Code errors: These are the errors in the CFD solver code.
Ensuring accurate results requires extensive verification of the software and extensive validation of the
modeling approach for the application of interest. Verification and validation are important to minimize
the solver-side sources of error and uncertainty in the CFD analysis. Verification will ensure that the
solver solves the equations correctly and validation tests the extent to which a chosen model accurately
represents reality. Verification and validation are an important aspect of the CFD modeling process but
will not be the focus of this discussion. Instead, this section focuses on general CFD quality assurance
guidelines that you should follow to ensure that you can expect reasonable results from your analysis.
The discussion is divided into several topic areas that represents an important aspect of modeling that
addresses one of the error types discussed above. You need to be focused on these aspects, besides the
user errors, to ensure high quality results. The following modeling aspects are discussed.
For many CFD applications the true solution to the equations is not always known. In the absence
of an analytical solution, or high quality experimental data to compare against, it is difficult to judge
the accuracy of a CFD simulation. However, you can still investigate the suitability of your modeling
practices by performing sensitivity studies. This practice reveals aspects of your model that have a
significant impact on your solution. Once these aspects are identified further work can be done to
determine the most appropriate modeling practices for each of these aspects that will minimize the
error associated with the individual aspects, ensuring good quality of results overall. It should be noted
that performing a full suite of sensitivity studies to evaluate the solution is a costly and time consuming
process. Since there are a number of aspects to be evaluated and all of them can impact the solution
combining all of them in a single study is not advisable. It is best to study them one at a time. When
performing a sensitivity study one aspect of the model that is being studied should alone be changed
and everything else should be kept unchanged until a satisfactory solution is achieved. This process is
then repeated for all of the aspects that you are interested in studying. In practice, many investigations
are avoided and engineering judgment is used in its place. However, for applications where accuracy is
of high importance a full set of sensitivity studies needs to be performed to ensure that modeling errors
are minimized in the solution. It should be remembered that the sensitivity studies cannot take into
account user and code errors and do not warrant that the solution achieved will be the true physical
solution. The following discussion describes different sensitivity studies that are commonly performed
when investigating the modeling practices to produce a quality solution.
As an exercise, and also to demonstrate to you the effect of the choices made for each of these aspects
over the course of a sensitivity study, the discussion of the topics covered will be accompanied with the
discussion of a simple but interesting case, the backward facing step case. For a detailed description of
the case refer to the AcuSolve Validation manual in your AcuSolve installation directory. However, here
you will not analyze the case with the objective of validating the model but with emphasis on educating
you about guidelines for setting up the model correctly and identifying the appropriate modeling choices
to make. A brief discussion of the case is as follows.
The problem consists of a fluid flowing in a channel with a sharp expansion. The step in the channel
causes the flow to detach at the step and reattach some distance downstream of the step. The detached
flow also causes recirculation in the region between the step and the reattachment point. The fluid
properties are close to that of air. The Reynolds number at the inlet is 36,000, which is based on the
step height. The flow is turbulent and you will pursue a steady state solution. The objective of the
analysis is the accurate representation of the recirculation zone and determination of the reattachment
point. The image below shows the schematic geometry of the case setup.
This situation roughly mimics the behavior of flow exiting from a flow channel into a plenum of a
plate type heat exchanger. An engineer is interested in placing a pressure tap along the lower wall to
measure the total pressure drop through the system. In order to avoid spurious measurements from
the pressure tap it is necessary to place the tap downstream of the recirculation region. Therefore, the
location of the reattachment point must be known.
This section discusses and illustrates the effects of the geometrical representation of the problem, and
the discretization of the geometry on your CFD solution.
CFD simulation involves the creation of a discrete mathematical model of an object, assembly, region in
space, and so on. In order to obtain accurate results from the CFD analysis, analysts must ensure that
they are using an appropriate representation of the geometry that they are trying to simulate. When
simulating flow involving complex geometric features, it is necessary to evaluate the degree to which
the discretized representation of the geometry matches the actual component of interest. For example,
it is common practice to remove small features such as small bolt holes and fasteners when simulating
complex assemblies. These features may or may not have a significant impact on the results of the
simulation. Analysts need to determine the appropriate level of detail required to achieve an accurate
representation of the flow field. This often times requires a geometry sensitivity study to determine.
Another aspect of proper geometry modeling is how well the mesh matches the geometry. For curved
geometries, it is necessary to have sufficient element density to resolve the curvature smoothly and
provide an appropriate reproduction of the shapes. In addition to the mesh density, it is also necessary
to ensure that the underlying CAD model provided appropriate resolution of the geometry. When
constructing meshes from discrete CAD models you must pay particular attention to this to ensure that
the discrete model is an appropriate representation of the true geometry.
The third geometric modeling aspect is the extent of the domain used for analysis. When modeling
external flows, it is necessary to model a sufficiently large domain that minimizes the impact of the
numerical constraints applied to the model. For example, when trying to simulate an airfoil in an
infinite medium it is necessary to size the domain large enough so that the surrounding region is
not constraining the flow around the airfoil in an unrealistic manner. Similarly, for internal flows, the
boundaries of the domain should be located far away from the location of any phenomena of interest in
the domain to minimize any interference in the results due to the presence of the boundaries.
A final aspect of the geometric modeling applies to the problems that have a rotating component
within the domain. Examples of such flows are the flows in pumps and turbomachines where fluid
interacts with a rotating part, adding swirl to the flow. The common approach while modeling such
cases is splitting the domain into a rotating part which immediately surrounds the rotating component
(rotor), and the static part that makes up the rest of the domain. While modeling such cases, it is very
important to understand the sensitivity of the location of the interface which will separate the static and
the rotating part. An interface located either too close or too far from the rotor reduces the accuracy of
the solution. For an accurate solution it is advisable to test the sensitivity of the solution to the location
of this interface. Unfortunately, there are no set guidelines to determine this, as the optimum location is
a function of the dimensions of the domain and the rotor, the relative velocity between the fluid and the
rotor, and so on.
The extent to which the boundaries should be distanced from the region of interest should be evaluated
on a case by case basis. It can be argued that using a distance of 50 times the step height is a
very conservative estimate for the backward facing step (BFS) problem, which may be considered
a relatively simple problem. Using such a conservative value is not always feasible especially as
geometries and models become complex. The modeling aim should be to minimize the interference
of boundaries on the results. Unless there is a constraint on the model that prevents shifting the
boundaries sufficiently far from the region of interest, or you are interested in studying the effect of
boundaries on the model, this practice should be followed.
This section illustrates the various effects that mesh refinement, mesh quality and topology have on
your CFD solution.
The role of a CFD solver is to approximate the solution to a coupled set of partial differential equations.
This typically entails solving for the pressure, velocity and turbulence fields within a given domain.
These fields can be both spatially and temporally varying. To obtain a solution to the governing
equations, the simulation domain is discretized into small, finite regions (referred to as elements or
cells) that are made up of faces that are connected by nodes or vertices. The solver then solves the
governing equations over each of these finite regions to approximate the solution over the whole
domain. In the case of a finite element solver, the solution is obtained and stored at the nodal locations.
In other words, the solution is permitted to vary at the locations of the nodes in finite element solvers.
The solution between the nodal locations is assumed to vary linearly (in most cases). With this in
mind, it is clear that the mesh resolution thus must be sufficient so that all variations in the flow are
sufficiently captured by an assumed linear profile between the locations where the solution is stored.
Mesh resolution is a measure of the distance between two consecutive measurement points in the
mesh.
A consistent numerical analysis will provide a result which approaches the actual result as the mesh
resolution approaches zero. In this case the elements become infinitesimally small and the mesh
behaves almost like a continuous representation of the domain. Thus, the discretized equations will
approach the solution of the actual equations. One significant issue in numerical computations is
what level of mesh resolution is appropriate. This is a function of the flow conditions, type of analysis,
geometry and other variables. An insufficiently resolved mesh that has nodes far apart will not capture
the gradients well enough when assuming linear variations between nodes and will lead to a less
accurate numerical solution. An unnecessarily finely resolved mesh will lead to higher solution costs
without any gain in accuracy. You are often left to start with a coarse mesh resolution and then
conduct a series of mesh refinements to assess the effect of mesh resolution. This is known as a mesh
refinement study. Strategic mesh refinement will call for you running a solution on the coarse mesh,
studying the results and identifying the areas that require enhanced resolution to improve the solution.
The refinement process should be continued until the results do not change with further refinement.
Such a solution may be referred to as the grid converged solution.
CFD practitioners must recognize the distinction between a numerical result which approaches an
asymptotic numerical value and one which approaches the true solution. It is hoped that as the mesh
is refined and resolution improves the computed solution will not change much and approach an
asymptotic grid converged value, that is, the true numerical solution. However, there still may be error
between this asymptotic value and the true physical solution to the equations. This behavior is shown in
the image below.
Figure 106: Effect of Increasing Mesh Density on Accuracy of the CFD Solution
Another possible sensitivity associated with the mesh used for a CFD analysis applies to the quality
and type of elements that are used. Each CFD solver has its own guidelines for what types of elements
behave best from a stability, accuracy and convergence stand point. In addition to this, there are
generally guidelines about the quality of each type of element that is suitable for use within the solver.
Many finite volume solvers cite the aspect ratio of the elements, element skewness, orthogonality of
the element faces and the element growth factor as critical metrics for determining their suitability for
use. For this reason, it is sometimes recommended by code vendors that you use hexahedra elements
for finite volume codes. Finite element solvers typically do not have this same constraint and will
often recommend the use of tetrahedral elements. You should consider the needs of your solver when
constructing a mesh to determine the most suitable discretization for your application.
Figure 107: Streamlines Showing Separation Zone in Backward Facing Step Model
However, if there are insufficient points in the mesh to provide a good estimate of the gradients
between the solution points the solver will not be able to resolve the flow features properly. Using
this example, the shear layer that develops in the separation zone is under resolved and this leads to
a poor representation of the reattachment point. In order to improve the solution it is necessary to
refine the mesh in the locations of high velocity gradients. By doing this, the solver is better able to
represent the local variations in the solution leading to a better approximation of the true solution to the
governing equations. The image below shows five different meshes created for the backward facing step
problem. Each mesh is progressively refined with respect to the previous mesh to illustrate the areas of
refinement that must be paid careful attention to for this particular application.
(a) Mesh 1; (b) Mesh 2; (c) Mesh 3; (d) Mesh 4; (e) Mesh5
Figure 108: Five Meshes Used in Backward Facing Step Mesh Sensitivity Study
Figure 4 shows the velocity profiles obtained from the solutions of each of the meshes using a
consistent solver setup. Images (a) to (d) each correspond to the solution from mesh one through four,
respectively. It is easy to see how the solution improves with each refinement in the mesh. A detailed
discussion of these figures is covered in the mesh refinement process below.
Figure 109: Velocity Profiles Downstream of the Step Depicting Mesh Sensitivity for the Backward
Facing Step Case.
where gradients are stronger. The mesh displayed in image 4(b) is generated using the following
global mesh size specification.
This will give about five times more elements in vertical direction.
This solution (image 5b) shows much better resolution of the velocity field, but the shear layer
and the recirculation zone are not properly represented in the solution as irregularities can be
seen in the velocity profiles in these locations.
The reattachment point for this mesh is observed to be at a distance of 8.44 m downstream of the
step.
After refinement the shear layer and the recirculation zone are much better represented, as
shown in image (c).
The reattachment point for this mesh is observed to be at a distance of 7.42 m downstream of the
step.
Wall-Step does not need a boundary layer because it is not expected to have attached flow. The
step edge has been resolved with a fine mesh because it is the location of separation.
Results from this mesh can be seen to have a proper representation of all of the features of
interest for this case. The recirculation zone is properly represented and the location of the
reattachment point can be easily identified.
The reattachment point for this mesh is observed to be at a distance of 5.73 m downstream of the
step.
The reattachment point for this mesh is observed to be at a distance of 5.73 m downstream of the
step.
The table below gives a summary of the refinement process discussed above along with the mesh size
generated in each case.
1 688 1,737
The table below shows the results achieved using five different meshes. Listed in the table is the
variation of the location of the reattachment point for each mesh.
Table 8: Reattachment Point Location as Observed for Each of the Different Meshes
2 4,508 8.44
3 51,056 7.42
4 68,014 5.73
5 207,492 5.73
It is evident that refining the mesh in the shear layer, recirculation zone and the near wall area where
the velocity gradients are highest leads to a change in the solution between meshes one through four.
However, it is interesting to note that as the mesh is further refined between grids four and five there
is no change seen in the solution. This implies that the solution error associated with grid refinement
has been minimized and you have reached an asymptotically grid converged solution with the mesh
density and topology utilized in mesh four. All of the critical regions have been refined at this stage and
suitable mesh topology as well as sufficient mesh resolution has been achieved. Further refinement of
the mesh did not change the solution quantity of interest. A review of the table above indicates that
the separation location changed by almost 50 percent between mesh one and the final grid converged
solution of mesh four. If no grid sensitivity study was performed the engineer could have made very
poor conclusions about the location of the pressure tap.
Appropriate grid refinement is one of the most important aspects of obtaining accurate CFD results.
Unfortunately, there is currently no reliable method of automatically generating appropriate meshes
for all applications. Additionally, different numerical methods will require different levels of mesh
refinement, that is, each code will be different. Ensuring the suitability of a mesh for a given application
is your responsibility. The level of accuracy required from a CFD analysis depends on the desired use
of the results. A conceptual design effort may be content with general depiction of the flow features,
whereas a detailed design may require accurate determination of the flow field. Each quantity to be
determined generally has its own accuracy requirement. Levels of effort invested into the CFD model
may vary according to the information required. It is often recommended that a minimum of three
different meshes are constructed that cover a wide range of refinement and differ significantly in mesh
density and resolution. For applications that require high levels of accuracy this number may be more
depending on the number of mesh refinements it takes to achieve a grid converged solution. This
allows you to evaluate the sensitivity of the solution quantity of interest to mesh density. Some CFD
practitioners endorse the use of a Grid Convergence Index, or GCI, to estimate the level to which a
simulation has reached a grid independent solution. More information about the GCI metric is available
in the book by Patrick J. Roache, listed in reference [1].
This section discusses and illustrates the effects of the location of your model boundaries and the
boundary conditions you apply on your CFD solution.
Boundary conditions specify the set of constraints on the solution of a CFD problem at the bounds of the
modeled domain. There are many different types of boundary conditions that exist. The available set of
constraints and conditions under which they should be used is specific to each CFD solver. However, you
should consider the following points when applying boundary conditions.
Proximity of the boundary conditions with respect to the results of interest from the
simulation.
In the discussion of the geometric sensitivity of the model it is mentioned that the boundaries
must be modeled far away from the region of interest to prevent any interference on the results
due to proximity to the boundary. To understand the possible reasons behind this interference it
is important to understand that the boundary conditions impose a constraint on the solution. This
constraint is an assumption about the behavior of the flow. If the boundary is placed too close
to a location in which the results are of interest it is possible that the solution will be impacted
drastically by the constraint applied at the boundary.
• Proximity of the boundary conditions with respect to the results of interest from the simulation.
If the boundaries in the backward step facing case are placed too close to the step it will have
an adverse effect on the solution. If the outlet is too close to the step the presence of outlet
constraints may interfere with the solution, preventing accurate representation of the recirculation
zone and the reattachment point.
• Realism of the constraints. Two cases are presented to demonstrate the effect of inlet condition
sensitivity on the solution of the backward facing step problem. The first case uses a constant
value of velocity and turbulent eddy viscosity at the inlet. This case is derived from Mesh_4 from
Mesh Sensitivity uses the same setup as mentioned in that section. The second case provides the
inlet variables as a profile similar to the one used in an experimental study of the same problem.
Figure 110 shows the profiles with boundary layer definition for the second case.
Figure 111: Inlet Turbulent Eddy Viscosity Profile for Inlet Boundary Condition Definition
The inlet velocity profile shows the presence of boundary layers at the edges while the eddy viscosity
profile shows a concentrated presence of eddy viscosity in the boundary layer.
When comparing the results of the simulations run with the boundary conditions described above you
can see that the profiled inlet condition reflects the experimental data much more accurately. The
following plot compares the velocity profiles for both of the cases mentioned above with experimental
measurements at the location four times of step height upstream of the step.
Figure 112: Velocity Profiles at x/h=-4 for Profiled and Constant Inlet Boundary Conditions Compared
Against Experimental Data at the Same Location
You can see from the above comparison that using a boundary condition specification that resembles
the real physics is much more likely to provide a correct solution to the problem.
The table below shows the difference between the reattachment length observed in the backward facing
step problem when a profile similar to the one in the experiment is provided at the inlet, rather than
using a constant value for the variables.
Constant 5.73
Presence of reverse flow at outlet conditions. Figure 113 shows three candidate locations for an outlet
boundary in the backward facing step model. The location represented by the white line falls in a region
of recirculating flow and would cause inaccurate results as well as numerical difficulties due to the flow
re-entering the domain. The location represented by the blue line falls outside of the recirculation zone,
but there is still a significant amount of flow non-uniformity and falls fairly close to the recirculation
region. To minimize the impact of the outflow conditions on the solution it should be moved further
downstream, as shown by the green line.
Figure 113: Candidate Locations for an Outlet Boundary on the Backward Facing Step Model
This section discusses and illustrates the effects of the mathematical models you choose to represent
the problem physics, flow type, material properties, and so on, in your problem.
Setting up a CFD simulation requires the selection of a number of physical models. The purpose of the
physical model is to mathematically represent the physics of a process that occurs in the simulation.
One of the first steps in defining the simulation is the identification of the relevant physics that need to
be modeled. Once this is done, a suitable model needs to be selected for each aspect of the simulation.
Because various models exist for different characteristics of a simulation, it is important to investigate
the sensitivity of the results to each of them. The following list mentions some of the possible physical
models that need to be considered.
The smallest turbulence eddies occurring in flows can have extremely small length scales. Direct
simulation of these eddies requires mesh and temporal resolution which exceeds the computational
capabilities of most modern computers. The computationally effective way is to model the behavior of
the turbulent eddies in a time averaged flow field.
Turbulence models are used to predict the effects of turbulence in the fluid flow without resolving all
of the small scales of motion that are associated with a turbulent flow. This reduces the amount of
computing time necessary to obtain a solution to a fluid flow problem.
Most CFD solvers contain a broad selection of different turbulence models. The most commonly
employed models in the industry are based on the Reynolds-averaged Navier-Stokes (RANS) equations.
Among RANS models there are eddy viscosity and Reynolds Stress formulations that need to be
considered. Alternative approaches such as Detached Eddy Simulation (DES), Large Eddy Simulation
(LES) and Direct Numerical Simulation (DNS) are also available in many CFD solvers. DNS approach,
as the name suggests, is the approach that directly solves for eddies of all scales and does not involve
any kind of modeling. It is mostly employed only for research purposes. Accuracy and computational
requirements for these approaches is as follows.
When setting up a simulation it is your responsibility to select a suitable turbulence model for your
application. Many models provide a reasonable level of accuracy for a broad range of applications while
others are formulated specifically for certain flows. Among the most popular general purpose models
are the one equation Spalart-Allmaras (SA) model and the two equation Shear Stress Transport (SST),
and k-omega (k – ω) models.
Even while using general purpose turbulence models the choice of the model is important with respect
to the problem being analyzed. Some models are better suited to some flows than others. The accuracy
of a given turbulence model is highly case dependent and there are no established rules that can ensure
correctness of the chosen model for a specific application. You must always proceed with caution and
choose a turbulence model only after verifying its suitability to previously benchmarked problems of a
similar nature.
In addition to the selection of a turbulence model it is also important to consider whether the flow is
fully turbulent or transitional. If the flow is transitional it is important to include a transition model to
properly account for this effect.
Setting up a solid model requires specification of density, specific heat and conductivity.
It is not required for a CFD model to have a solid material in the domain. These should be added only as
required when an interaction between the fluid and the solid is present. Of the properties mentioned the
specific heat and the conductivity are required only when energy equation is being solved.
• Density
• Viscosity
• Specific Heat
• Conductivity
• Viscoelastic properties, if fluid exhibits viscoelastic behavior
• Diffusivity for mass diffusion problems
• Porosity for porous materials
Of the properties mentioned above specific heat and the conductivity are required only when energy
equation is being solved.
The material properties in a CFD simulation can be defined in a number of ways. The simplest definition
is specifying the properties as a constant value. Once specified these constant values do not change
during the simulation. For solids, the conductivity can also be specified as constant, but anisotropic.
That is a different but constant conductivity in each normal direction.
For some cases, however, specifying a constant value of properties is not exactly representative of the
actual material behavior. For example, for highly compressible flows specifying a constant density will
lead to inaccurate results. For such flows it is prudent to use the ideal gas model for the fluid density,
which uses the ideal gas law to calculate density at a given time step.
There are various options available to specify the different properties for the materials. You should
choose the option that resembles the actual conditions as closely as possible. Some options are as
follows:
The heat transfer modes available are conduction, convection (natural and forced) and radiation.
For highly viscous flows viscous heating of the fluid must be taken into account. An example of flows
where viscous heating can have a significant effect are the internal flows in microchannels.
For simulating processes where the fluid undergoes adiabatic compression compression heating of the
fluid must be taken into account as well.
Prandtl number (Pr) is an important parameter when heat transfer within the boundary layer is being
studied. Prandtl number can, in simple terms, be described as being proportional to the rate of growth
of the velocity and thermal boundary layers. Smaller values of Pr indicate that thermal diffusion is more
prominent than momentum diffusion. Thermal boundary layer grows faster than the velocity boundary
layer.
When temperature effects are being studied and the flow is turbulent it is also critical to take into
account the additional effect of turbulence on heat transfer properties. The turbulent Prandtl number
is an important parameter in this case. Experiments show the average value to be 0.85 for most
applications, but the actual value can vary between 0.7 and 0.9 depending on the specific problem.
If radiation effects are present in the problem you must choose the radiation parameters accordingly. In
case of enclosure radiation view factor computation between the enclosure surfaces is critical. In case of
external incident radiation, for example solar radiation, solar flux vector should be specified.
Chemistry Models
Selection of chemistry models can play a critical role in the results of simulations of combusting and
reacting flows. When selecting a chemistry model you need to consider the rate of the chemical reaction
in comparison to the times scales of other events in the flow. Some models are geared specifically
towards turbulent flows where the turbulent time scale is much slower than the chemical reaction.
Others are intended for the opposite scenario.
In the solver setups used in previous sections the backward facing step case was solved using the SA
turbulence model. To illustrate the sensitivity of the solution to the selection of turbulence model, the
results from two models, the one equation SA model and the two-equation SST turbulence model,
are shown in the table below. For this application, experimental data shows a reattachment length of
6.26 m. The results show that the SST model has a lower deviation from the experimental observation
indicating that it is more suitable for this flow than the SA model.
Table 10: Turbulence Model Sensitivity for Backward Facing Step Simulation
This section discusses and illustrates the effects of the level of convergence tolerance you provide to the
solver, and also stresses on the need to look for other solution parameters before accepting a solution.
The convergence of the CFD solver gives an indication of how well it has solved the discretized equation
set for the application of interest. Most CFD solvers measure convergence by monitoring the residuals,
which represent the imbalance of terms in the governing equations at each iteration. As a general
rule, the lower the residuals the lower the error due to imbalance in the terms in the equation. Some
solvers also monitor the degree to which the solution is changing between iterations. As the simulation
progresses, the residuals should fall and also the solutions to the equations should reach a stabilized
value. It is a good practice to measure a solution quantity which may also capture the behavior of other
solution variables to determine convergence. A converged solution means that the solver has done its
job and produced the best solution it can for the given setup. However, that does not guarantee that the
solution is correct in comparison to test or physics. The accuracy of the solution is a function of many
other characteristics, some of which are described above.
CFD solver packages typically set default convergence criteria that are suitable for many applications.
However, if high levels of accuracy are important it is necessary to monitor the impact of convergence
on the solution for the application of interest. For steady state simulations this means monitoring the
solution quantity of interest as the simulation is converged further. For transient applications it is also
necessary to evaluate the convergence at each time step to see how it impacts the overall transient.
A closer inspection of the reattachment length convergence history in the second plot shows that
the size of the separation zone does not stabilize until about time step 42. Therefore, the level of
convergence necessary to obtain a high level of accuracy for the separation location corresponds to the
levels corresponding to step 42 of the simulation. This value may or may not coincide with the solver's
default convergence tolerance settings.
The solution can also be visualized to further inspect the convergence of the flow. The image below
shows the solution at various time steps as the solution approaches convergence. Inspection of these
figures indicates very little change in the flow after time step 20 when observed visually. However, as
seen in the plot of reattachment length, the results change until stabilizing at time step 42. The decision
of desired convergence tolerance thus depends on accuracy required from the CFD solution.
Figure 115: Convergence Behavior of the Backward Facing Step Simulation using the Shear Stress
Transport (SST) Turbulence Model
Figure 116: Convergence Behavior of the Backward Facing Step Simulation using the SST Turbulence
Model
Each solver will be different and will perform differently on each application. It is up to you to
investigate the proper convergence levels for your application. It should also be noted that convergence
cannot be gauged by the residual metrics alone. Judging suitable convergence levels involves
monitoring the convergence metrics, the solution quantity of interest, as well as looking at momentum
and mass balances within the flow field when possible.
5.7 Conclusion
The modeling aspects described in this section are by no means complete for all of the flow simulations
that can be imagined. This discussion is intended only to convey some of the common sources of
variability among CFD model results. A rigorous CFD analysis involves investigating all of the items
described in this section. However, it is acknowledged that in practice it is necessary to rely on
engineering judgment to reach a balance between producing results in a reasonable amount of time and
performing a suitable level of quality assurance on the model. You should always consider the desired
level of accuracy of the CFD model when deciding how much effort needs to be invested in quality
assurance efforts.
An important observation to be made here is that this CFD solution may not be the true physical
solution to the backward facing step problem. Over the course of this study, an attempt has been made
to minimize the uncertainties by careful optimization of the problem setup with reference to each of the
aspects discussed in the beginning of the section. It is meaningful now to go back to that discussion
and recall that although these uncertainties can be minimized it is nearly impossible to eliminate them
completely. At this stage, through optimization of all the critical aspects of your model, the uncertainties
associated with these modeling aspects have been minimized. The solution at this stage can be deemed
as the best possible CFD solution with respect to the aspects of the model discussed in this section. The
difference observed between this numerical solution and the true physical solution can be attributed to
the following factors.
• Mathematical modeling: The mathematical models which are used to represent the physical
processes are not an exact representation of the physical phenomena. Even the best available
models are based on certain approximations and simplifications.
• Input errors: These are the errors due to uncertainties of certain inputs such as the boundary
condition specification. For example, even when the boundary conditions are specified as
equivalent to the available experimental values there might be some interpolation errors when
these values are projected on to the CFD domain.
• Uncertainties in experimental results and measurements: This discussion focused on a few
guidelines to setup the CFD simulation model for a given problem. However, one key aspect
that cannot be overlooked while comparing CFD results with experimental results of a problem
is the uncertainties present in the experiment. An experiment, just like CFD analysis, is a
controlled simulation of the problem. There are various sources of errors and uncertainties in
the experiments such as uncertainties in the setup and environmental conditions in which the
experiment is conducted, uncertanties in the measurement instruments used in the experiment
and uncertainties in the post-processing equipment.
Due diligence is required to carry out an experiment with application of refined experimental methods
to achieve a reliable set of results. A similar diligence, when applied to carrying out a CFD simulation,
will help achieve a reliable set of results from CFD. An analysis completed with the above mentioned
guidelines supporting the procedure will provide a result which for most purposes can be used as a
substitute of the true value of the solution.
5.8 References
[1] Roache, P.J., Verification and Validation in Computational Science and Engineering, Hermosa
Publishers, Albuquerque, New Mexico, 1998.