BENCH

MARK

Aerospace issue . . .
COMPOSITES RESEARCH ON THE RISE
COMPOSITE PROCESS SIMULATION
HOW TO GET THE PART DIMENSIONS RIGHT
RESIDUAL STRESS CALCULATION FOLLOWING A REPAIR PROCESS
DYNAMIC SIMULATION OF FLIGHT TEST MANOEUVRES
IMPROVING THE SIMULATION OF BIRD STRIKE ON PLASTIC WINDSHIELDS
HIGH LIFT SYSTEM VIRTUAL TEST
IMPROVING STRUCTURAL MODELLING
ROCKET SCIENCE
FATIGUE IN ALUMINIUM HONEYCOMB-CORE PLATES
COUPLING 1D AND 3D CFD

THE INTERNATIONAL MAGAZINE FOR ENGINEERING DESIGNERS & ANALYSTS FROM NAFEMS

Special Edition:
Engineering Analysis
& Simulation in the
Aerospace Industry

from your

editor
David Quinn
david.quinn@nafems.org
@benchtweet

Welcome to this special edition of Benchmark, which has brought

together a series of past articles that are strongly relevant to
simulation in the aerospace industry. Aerospace manufacturers and
suppliers are facing an increasingly challenging and competitive
marketplace. The current industry demands that engineers design
safe and reliable aircraft, meet increasingly stringent fuel-economy
standards, and invent cost-effective approaches to the use of
cutting-edge materials. Utilizing the latest simulation tools with
accuracy and efficiency has never been more critical, as aerospace
engineering continues to move into highly-advanced technological
space.
As part of our annual industry series, we are hosting an aerospace
event which will look specifically at the challenges that the industry
currently faces, and will also explore how simulation and analysis
can help meet the industrys goals in a cost-effective and efficient
manner. Within this special edition of Benchmark, you will find
articles on many topics related to the aerospace industry, all of
which give a best-in-class perspective on a range of the issues that
are prevalent to all involved.
NAFEMS is the only independent, international association dedicated
to engineering analysis and simulation. Our range of best-practise
guides, benchmarks, how to publications, as well as seminars,
courses, e-learning and conferences, allow us to bring industries
together to share and exchange experience and knowledge in order
to drive the technology forward. Our members come from every
industry around the world, giving a truly global perspective to our
activities and allowing our community to benefit from the wealth of
its own experience. You can find out more about NAFEMS and our
activities, as well as details on our industry and technology specific
events, by visiting nafems.org

Composites
Research on
the Rise

Dr. Robert N. Yancey

VP Aerospace and Composites, Altair Engineering

n the 70s, 80s, and into the 90s, there was a lot of research money spent on composite materials. It
started with polymeric composites but in the 90s, much of this went to metal-matrix and ceramicmatrix composites. Much of this research was funded by the US, European, and Japanese governments
and was directed towards aerospace applications of composites. By the end of the 90s, much of the
advanced materials government research funding went towards nanomaterials with the private sector
taking on the composite materials research effort. This led to the Boeing 787 and Airbus A350 which have
over 50% of their structure made from advanced composite materials. In the last few years, a resurgence
of interest in traditional composite materials has led to several government supported research initiatives
and consortiums in this area. Much of this interest is in the fields of non-aerospace composites which
dominated the research funding in the 70s, 80s, and 90s. A review of some of the higher profile initiatives
and consortiums is outlined here.

In the US, there are several initiatives launched recently

from the Department of Energy, NASA, and the White
House. In the Department of Energy, Oak Ridge National
Laboratory is home to the Department of Energys (DOE)
new Carbon Fiber Technology Facility (CFTF)a 42,000
square foot innovative technology facility. The CFTF offers
a highly flexible, highly instrumented carbon fiber line for
demonstrating advanced technology scalability and
producing market-development volumes of prototypical
carbon fibers. The CFTF serves as a national testbed for
government and commercial partners to scale-up
emerging carbon fiber technology. A major goal of the
carbon fiber line is to bring down the cost of producing
carbon fiber and part of this effort is looking at
alternative precursors that could be less expensive to
produce.

energy, and compressed gas storage industries. Part of

IACMI will focus on virtual simulation tools for
composites and this effort will be led by Purdue
University and Prof. Byron Pipes (see article on
cdmHUB).

Early in 2015, the White House and Department of Energy

announced the formation of an Institute for Advanced
Composites Manufacturing Innovation (IACMI) led by the
University of Tennessee with a large team of companies,
research institutes, universities and U.S. State
governments. The focus will be on manufacturing
innovation for applications in the automotive, wind

The SAE Aircraft Seat Committee is also developing a

standard for composite aircraft seats which will address
the unique aspects of composite seats. The work in
progress standard (ARP6337) will define and develop test
parameters, test methods, measurements, and
acceptable performance criteria for composite aircraft
seat structures. The rationale behind the standard is the

NASA has launched an Advanced Composite Research

Partnership to advance certification of composite
structures for aerospace applications. The team includes
Bell Helicopter, GE Aviation, Lockheed Martin, Northrop
Grumman, Boeing, and United Technologies. Certifying
composite structures is much more complex and costly
than certifying metallic structures and a major goal of
the effort is to utilize simulation methods to bring down
the cost of certifying new materials and structures made
from new composite materials.

Digitally Reinforcing High-Rate

Composite Manufacture

Sderlund Harald, Shailesh Chillal,Asha Koshy & Sushovan Roychowdhury

GKN Aerospace

Composites
Research
on the Rise

Miroslav Stojkovic MSc CEng MRaes

Engineering Capability Manager, Design Stress and Simulation, National Composites Centre UK

ibre reinforced polymer composites have been a

mainstay of research and development in aerospace
and motorsport industries for decades. With mass
reduction and increased structural efficiency becoming a
key priority in the transport, renewable energy, marine and
construction sectors, the composites industry faces new
challenges to meet the growing demand and cut
component costs.
Developing highly repeatable automated production lines
is critical for composite manufacturing industries to make
the transition from low volume business models to the
high volume cost-efficient manufacturing these new
markets demands.
The National Composites Centre UK (www.nccuk.com) is
the UK hub for composites manufacturing industry and
provides a focal point for research and development into
automated composite processing. The Bristol-based
centre houses a multidisciplinary team of over 140 staff
who are successfully supporting industrial partners in
understanding and solving a range of issues facing
composite manufacturers.
As part of the NCC technology development,
manufacturing process simulation plays a key role in
developing insight and guiding the development of
composites manufacturing processes to accelerate
innovation and reduce the cost and risk associated with
process development. Among the wide ranging research
into automated manufacturing, automated fibre
placement, resin infusion and induction welding
of thermoplastics stand out as examples where
challenging simulations have provided real
benefit to NCC members.

Automated fibre placement (AFP)

Automated fibre placement (AFP) is used in the
manufacture of high-value composite components,
where precision and repeatability of fibre placement are
key to the performance of safety-critical components.
Tapes of composite material, between 6.35mm and
25.4mm wide, are compacted onto a tool using a
compliant polymer roller mounted to a robotic
positioner. Most AFP systems also heat the incoming
material via laser, infra-red light or Xenon Flash-lamp
(a system manufactured by Heraeus NobleLight and
developed in collaboration with the NCC) to achieve the
ideal processing conditions.
AFP is a key technology at the NCC, with live
development programs for thermoset composites (such
as carbon/epoxy), dry fibre materials as well as
thermoplastic matrix composites already yielding
industrial benefits.
Simulating the AFP process presents many challenges,
with large gradients in both pressure and thermal fields
around the rapidly moving roller. However, by
addressing specific manufacturing issues, the NCC
simulation engineers are providing real benefits on the
shop floor.
Recently, the NCC has developed efficient methods for
predicting the maximum achievable course width (the
number of tapes deposited in a single pass)and also
how changes in applied heating power influence the asdeposited material state.

14

Dynamic Simulation of
Flight Test Manoeuvres
on the Diamond D-Jet

Residual Stress
Calculation Following
A Repair Process

14

The numerical simulation of the complex fluid-structure interaction

taking place when manoeuvring an aircraft remains a challenge. A
realistic analysis of the airplane manoeuvrability often involves the
presence of moving parts, such as the deflection of the elevators,
the ailerons, or the elevons. For conventional Computational Fluid
Dynamics (CFD) codes, dealing with such moving geometries is a
challenging task. The following work uses a software based on the
lattice-Boltzmann method (LBM) to overcome these issues.

35

Mr Jaymeen Amin, Dr Tobias Ulmer (Airbus Operations GmbH, Bremen),

Mr Philip Neuhaus (FTI Engineering Network GmbH,
on behalf of Airbus Operations GmbH, Bremen)

Winner of the NWC13 Best Paper Award for

Greatest Business Impact of Simulation

High Lift System

Virtual Test

Improving Structural
Modelling of High
Strain Rate Behaviour
of Composite Materials
Using High Speed
Imaging
Duncan A. Crump, Janice M. Dulieu-Barton, and Stephen W. Boyd
University of Southampton

There is a drive towards producing lighter vehicles that are faster, more manoeuvrable
and more fuel efficient to improve the sustainability of transport systems. The excellent
specific stiffness/strength properties of fibre reinforced polymer composite, e.g. carbon
and glass fibres, make them an increasingly attractive option for structures in high-end
and military applications.
These complex materials are being used in applications where there is a real risk of
impact or high velocity loading, whether this is bird strike on passenger aircraft (Figure
1), slamming loads on marine vessels or explosions in the proximity of military vehicles.
For efficient structural design it is vital that accurate and pertinent material properties
are available for input into finite element (FE) models. While the quasi-static behaviour
of composite materials is generally well understood [1], there is a need to for further
analysis at high velocity loading [2, 3].

honeycomb

35

40

47

Fatigue in
Aluminium
Honeycomb-core
Plates

Fatigue in Aluminium
Honeycomb-core
Plates

Laurent Wahl, Arno Zrbes, Stefan Maas and Danile Waldmann, from the
University of Luxembourg, investigate the fatigue properties of the honeycomb
core of aluminium sandwich panels, as used throughout the aerospace and
automotive industries.

47

The University of British Columbia

Convergent Manufacturing Technologies

Composite Process
Simulation
Digitally
Reinforcing HighRate Composite
Manufacture

omposite parts will not have the same dimensions

as the tool on which they were processed, because
of mechanisms such as tool dimensional change
during heat up and residual stress build-up within the
part during cure/solidification and cool-down. This is true
for all composite materials and processes, and only the
mechanisms differ slightly between different material
systems and processes. Dimensional change becomes a
problem if the magnitude of change is greater than the
dimensional tolerance requirements of the part.
Aerospace structure tolerances can be as tight as +/0.010 inches from nominal engineering dimensions, and
this can be difficult to achieve without a good
dimensional management strategy. Many composite
fabricators are familiar with spring-in or spring-back
which is the closing of angles due to strain anisotropy
[Nelson & Cairns]. However, dimensional management is
a bigger systems level problem and many other
parameters also affect final cured part dimensions. The
systems parameters that affect dimensional change, and
any other outcome in a composites process, can be
divided into three broad groups related to part, tooling or
process [Johnston et al.]. Some of the main drivers are:

Part: Geometry, Material behavior, Lay-up

Tooling: Geometry, Material behaviour

Process: Temperature, Pressure, Time, Heat transfer

Anyone who has baked a cake in their home kitchen

knows that how the cake turns out does not only depend
on the dough but also the type of pan used, type of oven,

location in the oven, temperature and time in the oven,

cool down, and removal from the pan. The same is true
for composites processing it is a systems problem
where the part, tooling and process all interact to
determine the outcome.

11

Coupling 1D and 3D CFD

The Challenges and
Rewards of Co-Simulation
Vincent Soumoy of EURO/CFD and David Kelsall of Flowmaster Ltd, both members
of the NAFEMS CFD Working Group, provide an overview of the recent NAFEMS
UK seminar on coupling 1D and 3D.

he benefits of coupling 1D and

3D CFD codes have long since
been r0ecognised.
Automotive and aerospace
companies have used 1D codes to
gain a better understanding of
system performance (such as fuels
systems), whilst 3D codes are used
to analyse detailed behaviour within
and around key components. With
that in mind, the NAFEMS CFD
Working Group recently arranged a
seminar at the Heritage Motor
Centre in Gaydon UK to understand
the benefits of such links and
assess the current state of the art.
Approximately 40 interested parties

from across the NAFEMS

membership attended to hear a
number of interesting and thoughtprovoking presentations from
various speakers.
Darren Morrison started the
technical presentations by sharing
an interesting view on the subject
from the perspective of a large
aerospace company (AIRBUS).
Validation is seen as desperately
important, so that much of their
work is to prove that any couplings
are producing realistic and
reasonably accurate predictions. In
designing fuel systems, much of

the analysis is done with 1D codes

for reasons of computational
economy but sometimes the
passages and fluid interactions are
so complex that only a 3D
treatment is felt appropriate.
Hitherto results have been passed
manually from 1D to 3D analyses.
There is a desire for such couplings
to be automatic but without
compromising the integrity of the
analysis.
Representing a vendors
perspective, Domonik Sholz from
ANSYS Germany called for
participating codes to develop a

How To Get the Part

Dimensions Right in
Composites Processing

Predicting Dimensional Change

If a tool is machined to the nominal engineering
dimensions of the composite part, dimensional
measurements on multiple parts made off that tool will
generally show a mean deviation from nominal and some
variability around the mean. If the total deviation from
nominal is less than the dimensional tolerances,
dimensional conformance is achieved and no further
action is required. If not, the part, tooling and/or process
have to be modified to achieve dimensional conformance.
This can be a costly and time consuming iterative
process, as it is difficult to anticipate the effect of system
parameter changes if there is no predictive model.
It is increasingly unacceptable to depend on trial-anderror to achieve dimensional conformance, especially
once the tool is made and full scale parts are produced.
Other typical options are experience, expert opinion,
tests, or simulation. Experience and expert opinion often
fall short if the part and process is complex or deviates
from the previous experience base. Test data are often of
limited use as final cured dimensions depend on part,
tooling and process, which make scaling of results from
small test coupons to the full-size part and process often
misleading and thus risky. The most effective option,
particularly for large and complex structures, is

11

20

29

Improving the
Simulation of Bird
Strike on Plastic
Windshields
Parts: Mecaplex Ltd, Grenchen, Switzerland
Simulation: Aerofem GmbH, Ennetburgen, Switzerland
Project: University of Applied Sciences and Arts Northwestern Switzerland FHNW / Institute
of Product and Production Engineering, Windisch, Switzerland

Dynamic Simulation
of Flight Test
Manoeuvres on the
Diamond D-Jet

This article, which won the Best Presented Paper award at the
2013 NAFEMS World Congress, presents a numerical study on the
dynamic simulation of flight test manoeuvres on the Diamond DJET, using the XFlow virtual wind tunnel. The pitch capture
manoeuvre is first simulated, studying the pitch oscillation response
of the aircraft. Dutch roll flight mode is then numerically
reproduced. Finally, the D-JET angle of attack is evaluated in the
post-stall regime under controlled movements of the elevator.

20

High Lift System

Virtual Test

Gran Fernlund1,2, Anoush Poursartip1,2, Abdul Arafath2, Corey Lynam2

2

Dr. Peter Giddings CEng MIMechE

Residual Stress
Calculation Following
A Repair Process

How To Get the Part

Dimensions Right in
Composites Processing

Research Engineer, Manufacturing Process Simulation, National Composites Centre UK

"...a resurgence of interest in traditional composite

materials has led to several government supported
research initiatives and consortiums..."

Composite Process
Simulation

Improving the
Simulation of Bird
Strike on Plastic
Windshields
29

40

Rocket Science
At the recent Siemens NX CAE Symposium, held in Charlotte, NC, USA,
benchmark took some time to speak to Nathan Christensen of ATK
Launch Systems about their analysis processes, and use of simulation.

Nathan joined ATK as a design engineer in composite structures, designing and analyzing missiles and rockets. He
spent a significant portion of his 28-year career working with PLM/CAD/CAE and computational tools for design and
analysis. Christensen is one of the technical founders of ATKs PLM system, which now manages hundreds of thousands
of pieces of product and engineering information used at ATK facilities across the US. He has published numerous
technical articles and papers on rocket motor design and analysis, CAE tools and computational methods. He also holds
a patent for hybrid pressure vessels.

44

Christensen was first appointed manager of the CAE group in 1992, with responsibilities for engineering computational
tools and methods. In his current position as manager of Engineering Tools and Analysis group, his responsibilities
include PLM/CAD/CAE tools, trend analysis, rocket motor performance databases, analytical methods and software
development, reliability engineering and high-performance computing.

Improving Structural
Modelling of High
Strain Rate
Behaviour of
Composite Materials
Using High Speed
Imaging

Rocket Science

44

Background image courtesy of ATK Launch Systems. Nathan

Christensen photographed by Branco Liu, Siemens

54
Coupling 1D & 3D CFD
The Challenges and Rewards of
Co-Simulation

54

Composites
Research on
the Rise

Dr. Robert N. Yancey

VP Aerospace and Composites, Altair Engineering

n the 70s, 80s, and into the 90s, there was a lot of research money spent on composite materials. It
started with polymeric composites but in the 90s, much of this went to metal-matrix and ceramicmatrix composites. Much of this research was funded by the US, European, and Japanese governments
and was directed towards aerospace applications of composites. By the end of the 90s, much of the
advanced materials government research funding went towards nanomaterials with the private sector
taking on the composite materials research effort. This led to the Boeing 787 and Airbus A350 which have
over 50% of their structure made from advanced composite materials. In the last few years, a resurgence
of interest in traditional composite materials has led to several government supported research initiatives
and consortiums in this area. Much of this interest is in the fields of non-aerospace composites which
dominated the research funding in the 70s, 80s, and 90s. A review of some of the higher profile initiatives
and consortiums is outlined here.

In the US, there are several initiatives launched recently

from the Department of Energy, NASA, and the White
House. In the Department of Energy, Oak Ridge National
Laboratory is home to the Department of Energys (DOE)
new Carbon Fiber Technology Facility (CFTF)a 42,000
square foot innovative technology facility. The CFTF offers
a highly flexible, highly instrumented carbon fiber line for
demonstrating advanced technology scalability and
producing market-development volumes of prototypical
carbon fibers. The CFTF serves as a national testbed for
government and commercial partners to scale-up
emerging carbon fiber technology. A major goal of the
carbon fiber line is to bring down the cost of producing
carbon fiber and part of this effort is looking at
alternative precursors that could be less expensive to
produce.

energy, and compressed gas storage industries. Part of

IACMI will focus on virtual simulation tools for
composites and this effort will be led by Purdue
University and Prof. Byron Pipes (see article on
cdmHUB).

Early in 2015, the White House and Department of Energy

announced the formation of an Institute for Advanced
Composites Manufacturing Innovation (IACMI) led by the
University of Tennessee with a large team of companies,
research institutes, universities and U.S. State
governments. The focus will be on manufacturing
innovation for applications in the automotive, wind

The SAE Aircraft Seat Committee is also developing a

standard for composite aircraft seats which will address
the unique aspects of composite seats. The work in
progress standard (ARP6337) will define and develop test
parameters, test methods, measurements, and
acceptable performance criteria for composite aircraft
seat structures. The rationale behind the standard is the

NASA has launched an Advanced Composite Research

Partnership to advance certification of composite
structures for aerospace applications. The team includes
Bell Helicopter, GE Aviation, Lockheed Martin, Northrop
Grumman, Boeing, and United Technologies. Certifying
composite structures is much more complex and costly
than certifying metallic structures and a major goal of
the effort is to utilize simulation methods to bring down
the cost of certifying new materials and structures made
from new composite materials.

"...a resurgence of interest in traditional composite

materials has led to several government supported
research initiatives and consortiums..."

recent interest in the use of composite structures for 9-g

static and 16-g dynamic aircraft seat applications.
Historically the design, fabrication, inspection and
maintenance of aircraft seats has centered on the use of
metallic structures. The use of composites materials
requires additional guidance and information to maintain
the current level of aircraft seat safety and performance.
This effort is tightly coupled with efforts on the
Committee to move to a certification by analysis standard
for aircraft seats.
For Aerospace and Marine applications, the US Office of
Naval Research (ONR) established the Composites
Manufacturing Technology Center (CMTC) of Excellence
as one of nine Centers of Excellence supporting Navy
Manufacturing Technology. The CMTC develops improved
manufacturing processes for composites and advanced
materials and facilitates technology transfer for the
resolution of manufacturing and repair issues identified
and prioritized by the Navy's Program Executive Offices
(PEO's), other Department of Defense (DoD) services and
industry.

"...many companies have

launched significant research
efforts to evaluate increased
use of carbon composites..."
In Europe, there are also several composite initiatives
newly established. In the UK, the National Composites
Centre (NCC) is one of several Catapult Centers focused
on Manufacturing Technologies. The Centre, led by the
University of Bristol, includes several companies
including Airbus, Rolls Royce, Agusta Westland, GE
Aviation, GKN, and Cytec. The NCC brings together
companies and academics to develop new technologies
for the design and rapid manufacture of high-quality
composite products. The combination of academic and
business strengths will speed progress from laboratory
to design to factory and into products.
In Germany, the Technical University of Munich (TUM)
established the Institute for Carbon Composites in 2009.
The Institute is sponsored by the SGL Group and is
composed of an interdisciplinary team that can go from
raw materials through implementation of manufacturing
technologies to complete composite components. A key
area of expertise of the Institute is simulation methods

that have been developed to virtually model the complete

composite manufacturing process. TUM has also
partnered with Singapore Polytechnic and ST Kinetics to
extend their research and efforts to Southeast Asia.
Recently, MAI Carbon Cluster Management GmbH was
established to research methods to dramatically
decrease the cost of carbon fiber. The $100 million
research project is backed by Germanys federal
government and more than 70 businesses and research
institutes with major involvement from BMW and Audi for
automotive applications. BMW is aggressively looking
beyond the carbon fiber rich i3 and i8 models to include
carbon fiber in other BMW models.
In Japan, Mitsui & Co. of Tokyo will work with the
Innovative Composite Materials Research and
Development Center of the Kanazawa Institute of
Technology (KIT) on experimental research for new
production methods for the fabrication of automotive
parts and other industrial products using carbon fiber
composite materials. This center is supported by the
Japanese Ministry of Economy as well as the Japanese
Automotive Industry.
In addition to these federal government initiatives, many
companies have launched significant research efforts to
evaluate increased use of carbon composites and develop
the simulation and manufacturing technologies to ease
the transition from metals to composites. This includes
partnerships between Ford and Dow, BMW and SGL,
General Motors and Teijin, Toyota, Toray, and FHI, and
others. These partnerships between carbon fiber
suppliers and automotive OEMs benefit both industries
by increasing the applications of carbon fiber and hence
increasing the demand while reducing vehicle weight to
reduce the fuel emissions for the automotive industry.
Also, BMW and Boeing have partnered to work
collaboratively on design and analysis methods for
composite structures taking the best of the aerospace
and automotive industries to advance the state of the art.
Overall, it is an exciting time to be involved in the
composites industry. With now a firm footing in the
commercial aviation, sporting goods, and marine
industries and growing applications in the automotive,
energy, and building industries, the composites industry
is poised to accelerate its growth. A key to this growth
will be modeling and simulation methods for composites
that are robust, proven, and accessible to the design and
analysis community. NAFEMS will play a critical role in
establishing the modeling methods, practices, and
protocols for composite materials and structures.

Composite Process
Simulation
Digitally Reinforcing High-Rate
Composite Manufacture
Dr. Peter Giddings CEng MIMechE
Research Engineer, Manufacturing Process Simulation, National Composites Centre UK

Miroslav Stojkovic MSc CEng MRaes

Engineering Capability Manager, Design Stress and Simulation, National Composites Centre UK

ibre reinforced polymer composites have been a

mainstay of research and development in aerospace
and motorsport industries for decades. With mass
reduction and increased structural efficiency becoming a
key priority in the transport, renewable energy, marine and
construction sectors, the composites industry faces new
challenges to meet the growing demand and cut
component costs.
Developing highly repeatable automated production lines
is critical for composite manufacturing industries to make
the transition from low volume business models to the
high volume cost-efficient manufacturing these new
markets demands.
The National Composites Centre UK (www.nccuk.com) is
the UK hub for composites manufacturing industry and
provides a focal point for research and development into
automated composite processing. The Bristol-based
centre houses a multidisciplinary team of over 140 staff
who are successfully supporting industrial partners in
understanding and solving a range of issues facing
composite manufacturers.
As part of the NCC technology development,
manufacturing process simulation plays a key role in
developing insight and guiding the development of
composites manufacturing processes to accelerate
innovation and reduce the cost and risk associated with
process development. Among the wide ranging research
into automated manufacturing, automated fibre
placement, resin infusion and induction welding
of thermoplastics stand out as examples where
challenging simulations have provided real
benefit to NCC members.

Automated fibre placement (AFP)

Automated fibre placement (AFP) is used in the
manufacture of high-value composite components,
where precision and repeatability of fibre placement are
key to the performance of safety-critical components.
Tapes of composite material, between 6.35mm and
25.4mm wide, are compacted onto a tool using a
compliant polymer roller mounted to a robotic
positioner. Most AFP systems also heat the incoming
material via laser, infra-red light or Xenon Flash-lamp
(a system manufactured by Heraeus NobleLight and
developed in collaboration with the NCC) to achieve the
ideal processing conditions.
AFP is a key technology at the NCC, with live
development programs for thermoset composites (such
as carbon/epoxy), dry fibre materials as well as
thermoplastic matrix composites already yielding
industrial benefits.
Simulating the AFP process presents many challenges,
with large gradients in both pressure and thermal fields
around the rapidly moving roller. However, by
addressing specific manufacturing issues, the NCC
simulation engineers are providing real benefits on the
shop floor.
Recently, the NCC has developed efficient methods for
predicting the maximum achievable course width (the
number of tapes deposited in a single pass)and also
how changes in applied heating power influence the asdeposited material state.

Figure 1: The NCC has two Coriolis Composites AFP machines (pictured with GKN composite winglet) and an
additional Accudyne machine with choice of laser, infrared and patented Xenon FlashLamp heat sources.

The method developed to predict maximum course width

provides clear programming rules for manufacturing
engineers that ensure material will receive sufficient
compaction pressure. This critical information is
extracted from local quasi-static finite element
simulation describing the compaction of a roller onto a
tool surface at critical locations (simulated boundary of
positive contact pressure shown as green ellipse in Fig. 2.
Built-in modelling options describing materials
behaviour, geometric non-linearity and sliding contact
within Abaqus Standard (provided by NCC member
Dassault Systmes) efficiently capture the complex
physical behaviour. By taking component geometry, fibre
orientation and experimentally measured load-deflection
response of the roller [1] as inputs, these models return
maximum course width within around 90 minutes for
each desired fibre orientation and feature on the tool.
For a complex component, just 2.5 days of simulation
effort is required to generate design rules to guide
process specification and ensure good manufacturability.
These rules reduce operator uncertainty and variability in
programming while saving days or weeks of costly onmachine trials.
The prediction of as-deposited material state has begun
by tackling AFP manufacturing using thermoplastic
composites as part of the Core Research Program.

Dr. Peter Giddings, simulation engineer responsible for

that effort explains Our objective was to quantify how we
could manipulate heater power to maximise quality of the
deposited material by predicting material state
parameters like degree of bonding between layers or
percentage of voids.
The underpinning simulation method is an in-house finite
difference code, written in MatLab, that predicts heat
diffusion within the deposited material as the thermal
and pressure boundary conditions imposed by the roller
move across a component. To enhance predictions for
temperature and material state distributions, the code
updates key material properties that influence thermal
diffusion [2,3], for example density, during each solution
increment.
Today these coupled thermo-chemical simulations are
helping to define process windows for high quality carbon
fibre/PEEK composites for aerospace structures. As the
capability is extended to cover the full range of materials
and heat sources used in AFP it will help more
customers, these simulations can offer guidance on
effective machine settings and usable design rules for
AFP manufacturing to help broaden the viability of
automated fibre placement and minimise
commissioning risk says Dr. Giddings.

Figure 2: Simulation of critical features in AFP layup to determine maximum course width showing manufacturing
challenge and an example FE contact patch output with extraction of maximum course width

Resin flow
Many composite components begin as preforms of dry
reinforcing fibres before being impregnated with an
uncured liquid resin and heated to cure the resin. For
components requiring excellent surface finish and
increased mechanical performance, that impregnation
occurs in a closed metallic tool in a process called resin
transfer moulding (RTM). Prediction of how the resin
flows through the preform to fill the mould, whether any
areas will fail to be completely impregnated and the
optimisation of injection location and pressure are all
challenges that the NCC is working toward resolving.
Over the past two years NCC core research has
developed effective RTM simulation approaches as Dr.
Christian Lira explains: today, if a customer comes to us
with a problem in their infusion, even if it is thick or
highly curved, we can help. Tooling design, where you
inject the resin and how you adjust the pressure can all
be included to guide them to a solution.
These successes have been achieved using ESIs
software PAM-RTM which Dr. Lira says provides a finite
element solution to Darcys flow equation (flow through
porous media) and allows us to make useful simulations
within industrial timescales. With infusion, the process
has inherent variability, small but unavoidable changes in
material permeability cause big changes in flow rate so
any simulation is indicative, not perfectly predictive[4].
However, the simulations are still extremely valuable for
comparing the effects of various parameter changes on
process outcomes.
The understanding of material and process variation
built up at the NCC has made it clear that flow
simulations cannot predict the exact dimensions of a
defect but do indicate whether defects may occur
and their likely locations . Within these limits, Dr.
Liras infusion simulations are already guiding
engineers through more efficient test plans and
have made simulation-led process design for
traditional RTM a reality at the NCC.

Figure 3: Prediction of resin infusion through

a carbon fibre preform of a vehicle wheel
using ESI's PAM-RTM software

To meet the tight timescales demanded by high volume

automotive customers and produce cured composite
parts in less than 5 minutes, resin infusion technology is
moving to higher injection pressures and faster curing
resins. High pressure RTM (HP-RTM) injects resin into a
preform at pressures of up to 140 bar to fill moulds in
seconds before the fast curing resin systems begin to
harden. The speed and violence of the HP-RTM process
means that the understanding of infusion simulation,
built up in traditional RTM development, is no longer
enough to effectively guide manufacture.
The challenges posed require new approaches as Dr. Lira
describes we have to update material permeability and
fluid viscosity during the simulations as resin pressure
deforms the fibres and fast reacting resins begin to cure
during injection. Were working with software providers to
help extend RTM simulation techniques to deal with
these effects, but the effects of small quantities of
polymeric compounds applied to the dry fibres to hold
performs together, known as binders, is not so simple.
The methods for capturing the influence of binders within
HP-RTM are not well understood even within the
scientific community and so arriving at a predictive

Figure 4: Europes only open-access 3.4m 2.6m press installed and making parts at the NCC

simulation of HP-RTM requires a longer-term

commitment to develop and refine the simulation
capability. That commitment has already begun as the
NCC works with academics, software vendors and
manufacturers around the world to bring insightful
simulation of HP-RTM towards industrial
implementation.

These efforts are made possible by the installation of a

Schuler 36-Kilotonne press in December 2014 at the
Bristol site to explore the infusion of large composite
components via HP-RTM. The unique combination of
open-access industrial scale equipment together with
on-site laboratory mean that the novel simulations
necessary to support rapid development of HP-RTM
processes can be grounded in high quality
experimentation. That industrial scale validation is
critical to understand how machinery and processes will
respond under the extreme conditions that HP-RTM
imposes.
Multiphysics simulation of induction welding
Together with enhanced environmental resistance,
recyclability and novel processing routes, one of the
benefits for thermoplastic composites is that
components may be joined structurally by welding two
components together. The resulting joints can replace
mechanical fasteners to help components retain more of
the strength of the pristine laminate by eliminating
drilled holes as well as reducing part count in large
composite assemblies.

A particular interest at the NCC is induction welding of

carbon fibre composites. In this emerging composites
process, a magnetic field is used to heat the carbon
fibres within composite materials through electrical eddy
currents generated by electromagnetic induction within
the conductive fibres. Through controlled application of
an oscillating magnetic field the heating effect can be
managed so that the polymer matrix melts in the desired
locations to permit welding to take place.
For simulation engineers at the NCC, capturing the
induction heating effect in layered anisotropic materials
has proven to be a hugely satisfying project. The process
simulation team chose MSC MARC Nonlinear FEA
software (supported by MSC Software Ltd, Frimley) to
build simulations for induction heating of thermoplastic
composite joints.
The task of improving simulation results to be of use in
process specification for composite welding required
coupling of thermal and electromagnetic models, and
careful specification of material parameters. The
research posed challenges to the materials test and
tooling manufacture supply chains as well.
Materials tests were identified or developed to provide
unusual but necessary simulation input data over
describing properties such as dielectric permeability,
among others, for anisotropic composite materials. Once
preliminary models were validated against literature
data, more detailed analyses were developed with MARC
to design test fixtures and induction coils suitable to
experimentally characterize the induction heating
process [5,6].

Figure 5: Induction welding simulation of lap shear test specimen conducted

at the NCC showing specimen dimensions and resulting heated area

The simulation of induction heating is a strong first

step towards developing the predictive tools for
induction welding of composites, and is already
helping tooling suppliers to refine fixture designs that
retain components without impacting the delivered
magnetic field. The effort expended in building these
capabilities has allowed the NCC to validate methods
for predicting heat input within induction heating of
composite plates as the centre moves towards
simulation of real industrial welding processes.

Conclusion
The outlook for process simulation in composite
manufacture is incredibly bright. There is vibrant
academic research activity extending our fundamental
understanding and coupled with strong growth in
industrial demand for composites. Automation
technology is becoming established in a broader
range of industrial applications and simulation tools
from ESI, Dassault Systmes, MSC Software and
others offer suitable platforms in which to build
useable and powerful process simulations.
The simulation successes at the NCC are just the tip
of the iceberg for composite process simulation; there
are some fascinating challenges and tangible
commercial opportunities for simulation engineers
within composites. However, real progress is needed
in bringing these complex simulations into the supply
chain to aid in industrialisation of automated
composite manufacture.
The NCC aims to pave the way for the simulation
supply chain to effectively support the composites
sector and help demonstrate that the fascinating
multiphysics problems bring real returns on the shop
floor and also in the finished product.

10

References
[1] Helenon, F. D. H.-J. A. Lukaszewicz, Ivanov,D and Potter,
K. Modelling slit tape deposition during automated fibre
placement. 19th International Conference on
Composite Materials (ICCM19), Montreal, Canada, 2013
[2] Cogswell, F. N. Thermoplastic aromatic polymer
composites. 1st Edition, Elsevier Science and
Technology. 1992
[3] Stokes-Griffin, C.M. Compston, P. A combined opticalthermal model for near-infrared laser heating of
thermoplastic composites in an automated fibre
placement process. Composites Part A (In Press) .
[4] Arbter, R. Experimental determination of the
permeability of textiles: A benchmark exercise.
Composites: Part A 42: 1157-68, (2011)
[5] Moser, L. Experimental Analysis and Modelling of
Susceptorless Induction Welding of High Performance
Thermoplastic Polymer Composites, PhD Thesis,
Institut fur Verbundwerkstoffe (2012 )
[6] Rudolf, R. Mitschang, P. & Neitzel, M. Induction heating
of continuous carbon-fibre-reinforced thermoplastics,
Composites: Part A 31: 1191-1202 (2000)

About the NCC: www.nccuk.com

The NCC is a 25m investment supported
by: the Department for Business, Innovation
and Skills (12m); the South West RDA
(Regional Development Agency) (4m); and
9m from the European Regional
Development Fund (ERDF). It is owned and
hosted by the University of Bristol. The
Government announced a further 28m in
the 2012 Autumn Statement for the
expansion of the NCC. The NCC is a partner
of the High Value Manufacturing Catapult.

How To Get the Part

Dimensions Right in
Composites Processing
Gran Fernlund1,2, Anoush Poursartip1,2, Abdul Arafath2, Corey Lynam2
1

The University of British Columbia

Convergent Manufacturing Technologies

omposite parts will not have the same dimensions

as the tool on which they were processed, because
of mechanisms such as tool dimensional change
during heat up and residual stress build-up within the
part during cure/solidification and cool-down. This is true
for all composite materials and processes, and only the
mechanisms differ slightly between different material
systems and processes. Dimensional change becomes a
problem if the magnitude of change is greater than the
dimensional tolerance requirements of the part.
Aerospace structure tolerances can be as tight as +/0.010 inches from nominal engineering dimensions, and
this can be difficult to achieve without a good
dimensional management strategy. Many composite
fabricators are familiar with spring-in or spring-back
which is the closing of angles due to strain anisotropy
[Nelson & Cairns]. However, dimensional management is
a bigger systems level problem and many other
parameters also affect final cured part dimensions. The
systems parameters that affect dimensional change, and
any other outcome in a composites process, can be
divided into three broad groups related to part, tooling or
process [Johnston et al.]. Some of the main drivers are:

Part: Geometry, Material behavior, Lay-up

Tooling: Geometry, Material behaviour

Process: Temperature, Pressure, Time, Heat transfer

Anyone who has baked a cake in their home kitchen

knows that how the cake turns out does not only depend
on the dough but also the type of pan used, type of oven,

location in the oven, temperature and time in the oven,

cool down, and removal from the pan. The same is true
for composites processing it is a systems problem
where the part, tooling and process all interact to
determine the outcome.

Predicting Dimensional Change

If a tool is machined to the nominal engineering
dimensions of the composite part, dimensional
measurements on multiple parts made off that tool will
generally show a mean deviation from nominal and some
variability around the mean. If the total deviation from
nominal is less than the dimensional tolerances,
dimensional conformance is achieved and no further
action is required. If not, the part, tooling and/or process
have to be modified to achieve dimensional conformance.
This can be a costly and time consuming iterative
process, as it is difficult to anticipate the effect of system
parameter changes if there is no predictive model.
It is increasingly unacceptable to depend on trial-anderror to achieve dimensional conformance, especially
once the tool is made and full scale parts are produced.
Other typical options are experience, expert opinion,
tests, or simulation. Experience and expert opinion often
fall short if the part and process is complex or deviates
from the previous experience base. Test data are often of
limited use as final cured dimensions depend on part,
tooling and process, which make scaling of results from
small test coupons to the full-size part and process often
misleading and thus risky. The most effective option,
particularly for large and complex structures, is

11

simulation where a physics-based model is generated.

This model links system parameters such as part, tooling
and process to the relevant process outcomes in this
instance dimensional change.

Physics-Based Process Models

To accurately predict dimensional change, a physicsbased process model must include a high fidelity
description of the part, including geometry, lay-up and a
detailed description of the curing material behavior as
the composite material properties evolve during the
cure/consolidation cycle. Also needed is a good
description of the tooling, which includes geometry and
thermo-physical material properties. Finally, the model
needs to capture the process: temperature and pressure
application over time, and heat transfer to the part and
tool [Fernlund et al.]. This type of multi-physics process
model is internally fairly complex but nowadays can be
relatively easy to set up and run if the right solution
package is selected. Without suggesting that this is an
exhaustive list, and looking beyond general capabilities
available in general purpose codes with user defined
capabilities, there is some embedded capability within
MSC.Marc and ANSYS, as well varying levels of more
focused capability within ESI PAM-DISTORTION, LUSAS
HPM, and Convergents RAVEN and COMPRO.
In terms of modelling details, the first step is to solve the
thermochemical problem: this consists of a transient

12

thermal analysis of the part on the tool with typically

convective heat transfer boundary conditions; a critical
feature is to have an accurate representation of the heat
generation due to the cure reaction of the matrix in the
composite part, as well as property evolution as a
function of both temperature and degree of cure. More
sophisticated analyses may include a flow and
compaction modelling stage, but the next necessary step
is to model the development of residual stress due to the
cumulative mismatch in free strains throughout the part
and tool, where the part is viscoelastic in nature; note
that prior to gelation (the beginnings of a 3-D network in
the polymer matrix allowing for residual stress to
develop) the matrix has no memory and no ability to
develop residual stress. As the matrix vitrifies, it
becomes increasingly able to develop residual stress, and
thus the effects of cure shrinkage towards the end of the
processing cycle and thermal cool-down effects become
critical. The constitutive representation of the material,
the complexity of the solution, the ability to characterize
and calibrate the model, and the efficiency of the solution
become key issues in getting meaningful answers. The
current state of the art is that it is possible to accurately
and efficiently solve for a wide range of important
industrial processes, and improvements in capability are
accelerating as demand grows rapidly.
Figure 1A shows an example of a finite element model of
a part on a tool subject to a cure cycle that was quickly
developed from existing CAD information using Dassault
Systemes CATIA and ABAQUS design and simulation

Figure 1. A) Finite element mesh of part and tool; B) Calculated temperature profile during heat-up and cure;
C) Calculated dimensional change.

software together with Convergents COMPRO process

simulation software. Figure 1B shows the predicted
temperature gradient during heat-up and Figure 1C, the
calculated dimensional change.

Managing Dimensional Change

Once a process model is created for the part, tool and
process of interest, it can be used both for predicting the
expected mean dimensional change as shown in Figure
1C but also to identify the systems parameters that drive
variability in the process. The most effective way to
address the mean dimensional change of the part is
often by geometric compensation of the tool surface as it
can be done without changing laminate or process
parameters. Using the COMPRO CATIA ABAQUS
solution set, geometric compensation of the tool can be
done automatically by transferring the calculated
dimensional change back to the CATIA design
environment and morphing the tool surface so that
dimensional conformance is achieved. Altering the lay-up
sequence and/or modifying the cure cycle are alternative
options that can be developed and evaluated in the same
simulation environment.
Once the system parameters have been adjusted to give a
mean dimensional outcome that matches the nominal
engineering dimensions, the model can be used to
identify and set bounds on the allowed variability of
systems parameters such that the dimensional variability
of the part is within dimensional tolerances.

Closing
The composites process simulation technology and
methodology presented here is currently increasingly and
routinely used by the large aerospace OEMs. It is clear we
have reached the tipping point of convergence with
powerful and effective process simulation tools and
cheap and fast computational power. Increasingly,
composites processing should no longer be treated as an
art and should leave the domain of empiricism.
Processing can be approached with the same analytical
mindset and design and simulation tool sets as any other
aspect of engineering. This is critical for us to succeed in
designing and building large complex composite
structures that can compete with metal structures. The
era of simulation supported, knowledge-based
composites manufacturing is here, and there is no
turning back if we want to remain competitive.

References
Nelson, R. H., & Cairns, D. S. (1989). Prediction of dimensional changes in
composite laminates during cure. Tomorrow's Materials: Today., 34, 23972410.
Johnston, A., Vaziri, R., & Poursartip, A. (2001). A plane strain model for
process-induced deformation of laminated composite structures. Journal
of composite materials, 35(16), 1435-1469.
Fernlund, G., Floyd, A., Shewfelt, M., & Hudek, M. (2007, September).
Process analysis and tool compensation for a complex composite panel.
In Proceedings of the 22nd American society for composites technical
conference (ASC), Seattle, Washington, USA.

13

Residual Stress
Calculation Following
A Repair Process
Sderlund Harald, Shailesh Chillal,Asha Koshy & Sushovan Roychowdhury
GKN Aerospace

14

Influence of Residual Stress

on The Life of a Bolted
Flange Subjected to Repair
from a Manufacturing Defect

15

ero engine structural components such as

casings are usually made as cast or fabricated
pieces and welded together before the final
assembly. Deviations during the machining
operations or mishandling during its
manufacturing stages could cause damage to the
structure resulting in permanent deformations such as
bends or dents. Such deviations on any part, in an
aerospace industry caused during the manufacturing
process, are commonly termed as Non-conformances
(NC).
Due to cost implications, considerable effort is spent to
repair such deviations, instead of completely rejecting the
part. However to accept the part, it is quintessential to
analyze and understand the impact of the repair process
on the components structural integrity both in terms of
strength and life requirement fulfillments. A carefully
designed simulation process can quantify this impact
more accurately. An understanding of how the repair
process is carried out is essential to estimate the
deformation and residual stresses that could arise due to
the repair.

A Case Study The problem

This paper features a case study on residual stress
computation and its impact on life evaluation of an aero
engine casing flange subjected to repair after the
manufacturing process. The approach is based on FE
analysis using ANSYS as software to simulate the repair
process in order to determine the residual stresses in the
structure after the repair.
An incident during handling of the component caused a
tool to impact the flange, causing bending of certain
parts of the flange as shown in Figure 1. The flange is
assembled with the adjacent components through bolted
joints. The bending of the flange needs to be straightened
out following standard repair process for proper
assembly with the adjacent structure and also to prevent
leakage during operation. The repair is carried out by
heating the component locally and cooling it down to
room temperature before assembly. The process of
straightening during repair induces plastic deformation
and residual stresses on the flange. The challenge lies in
simulating these processes in a manner that allows a
realistic computation of the induced stresses. The
residual stresses induced after repair is then combined
with other operational loads to compute the life of the
component.

FE Modeling Approach
The typical FE model considered in the analysis is shown
in Figure 2. The model includes part of two flanges, bolts

Figure 1: Bent Flange configuration

and nuts all modeled using 3D solid elements

(SOLID186) in ANSYS. Standard frictional contact is used
at the interfaces of bolt-to-flange, flange-to-flange and
nut-to-flange regions. Bolt preload is simulated through
pretension elements. Regions around the bolt holes and
the flange fillets are considered the most critical life
limiting locations in this assembly. This assembly model
is used to simulate both the repair process and the flight
mission. By doing so, it is easier to superpose the
addition of residual stresses to the stresses generated
from operational loads for life computation at each node.

Simulation of Repair
The FE model shown in Figure 2 corresponds to the final
design configuration. In order to simulate the repair
process, the affected flange has to be first deformed to
the NC configuration. This is achieved first by separating
out the flange component from the assembly by
numerically reducing the stiffness of the adjacent
components. The stiffness of components other than the
affected flange is made near zero using ANSYS EKILL
command. This helps in maintaining the element and
node numbering sequence in the model same throughout
the analysis and enables superposition of stresses at
later stages.
The overall process to numerically compute the residual
stress involves six major steps as shown schematically in
Figure 3. The steps 1 and 2 are carried out to obtain the
NC configuration before the repair process and steps 3
through 6 simulate the repair process. During these six
steps, only the affected flange is
considered from the whole assembly.
Nonlinear material model using
kinematic hardening (option KINH in
ANSYS) is used during all the steps.
This accounts for material behavior
under the reversed loads. Geometric
nonlinearity is included in the

The challenge lies in simulating these

processes in a manner that allows a realistic
computation of the induced stresses

16

Figure 3: Schematic representation of the FE simulation of repair process

analysis by using the large deformation option. The

different processes involved in each steps are briefly
described next.
In Step 1, the critical NC region is identified on the flange.
This is a loading condition where the nodal forces are
applied in the identified NC region and the analysis is
carried out to initiate flange deformation. The
deformation of the actual bent hardware is measured and
a scale factor based on the displaced configuration is
used as a basis for initial load application in Step 1.
Step 2 is an unloading part, wherein the applied forces
from Step 1 are removed. The deformed configuration at
the end of Step 2 should represent the NC configuration
both in shape and magnitude of deformation. Since all
the analyses are non-linear in nature, Steps 1 and 2
requires an iterative process to set the magnitude of
initial force field, in order to match the flange bending at
the end of second step with that of the actual hardware.
At the end of second step, the deformation at all the
nodes of the flange surfaces are measured and
stored in an array. Figure 4 shows
the match
between the
analysis and
measured data
from actual
hardware.

Once the NC configuration is achieved in Step 2, the

analysis is continued in Step 3 by locally subjecting the
component to an elevated temperature. This is because
such repair processes are normally associated with local
heating.
In Step 4, the repair process is initiated by applying a
scaled value of unit force to the NC region of the flange.
The force scaling is done using the flange deformation
pattern obtained at Step 2. The stored flange nodal
deformation is used to scale this unit force and the force
is applied in the opposite direction to simulate the repair
process.
In Step 5 the external force is removed. Note that in steps
3-5, the component is exposed to local heating.
In Step 6 the flange is brought to room temperature. The
deformations obtained from Step 6 are compared with
the actual repaired hardware to fine tune the scaling
factor to be used in Step 4. Here again an iterative
process is used to get an appropriate scale factor in Step
4 that would result in a fairly good match of the surface
profile at the end of Step 6 with that of the repaired
flange. Figure 5 compares the final flange surface profile
obtained from simulation with that of the actual hardware
after the repair process.

Figure 2: Bolted Joint Assembly

17

Figure 4: Flange deformation comparison between actual hardware and simulation

Figure 5: Flange surface profile comparison between actual hardware and simulation

Simulation of Results
At the end of Step 6, the stresses resulting in the flange
configuration is considered to be the residual stress
generated due to the repair process. It is observed that
the residual stress is not uniform in the flange after
repairing the flange nonconformance. In the conventional
process used in the current industry, a uniform or
constant residual stress is added all across the
component for life computation. This value is normally
obtained from past residual stress measurement data or
by experience. From the present analysis it was observed
that the flange regions are subjected to varying residual
stress as shown in Figure 6. The normalized stress
distribution for one of the critical mission loads is shown
in Figure 7.
A sensitivity study was further performed to evaluate the
variation in residual stress with final surface profile

18

achieved. It was found to be less sensitive in variation of

final surface profile. A variation of 2 mm in the final
surface profile resulted in stress varying up to 2.5%.
Additionally a variation of residual stress up to 30% had
very little impact on LCF life as shown in Figure 8.
In order to compute life, a linear superposition of residual
stress from repair simulation over the elastic stresses
obtained from mission loads is considered. The linear
superposition is valid only when the residual stresses are
below yield with very little or no plastic strain. These
stresses should also be much less compared with
mission stresses. The six component stresses from
repair simulation were added to those from each load
case of the elastic mission stresses. The life is computed
using the neuber corrected stresses for the complete
mission loads.

A sensitivity study was performed to

evaluate the variation in residual stress

Figure 6: Residual stress distribution on the flange after the repair simulation (section view)

Figure 7: Flange stresses from the most critical mission load (section view)

an FE simulation of the repair

process is recommended to
compute the distributed residual
stress more accurately
Conclusions
Figure 8: Variation of residual stress and LCF life
Computation of Life

Figure 9 shows the variation of life at bolt hole and flange fillet
using the conventional approach and the current approach. The
current approach is a more realistic way to compute the residual
stresses as compared to using a constant stress value. Using a
constant value across all the region could result in over or under
estimating the residual stresses and hence impacting the
computed life numbers.
Based on the observations of the current study, it is recommended
to perform FE simulation of repair process to compute the
distributed residual stress more accurately and use it for
estimation of component fatigue life. This would make the NC
evaluation process more robust and helps in making a realistic
justification on whether the part can be accepted or rejected after
repair.

Figure 9: LCF life comparison using conventional &

current approach

19

Dynamic Simulation of
Flight Test Manoeuvres
on the Diamond D-Jet

The numerical simulation of the complex fluid-structure interaction

taking place when manoeuvring an aircraft remains a challenge. A
realistic analysis of the airplane manoeuvrability often involves the
presence of moving parts, such as the deflection of the elevators,
the ailerons, or the elevons. For conventional Computational Fluid
Dynamics (CFD) codes, dealing with such moving geometries is a
challenging task. The following work uses a software based on the
lattice-Boltzmann method (LBM) to overcome these issues.

20

This article, which won the Best Presented Paper award at the
2013 NAFEMS World Congress, presents a numerical study on the
dynamic simulation of flight test manoeuvres on the Diamond DJET, using the XFlow virtual wind tunnel. The pitch capture
manoeuvre is first simulated, studying the pitch oscillation response
of the aircraft. Dutch roll flight mode is then numerically
reproduced. Finally, the D-JET angle of attack is evaluated in the
post-stall regime under controlled movements of the elevator.

Luc Van Bavel (Diamond Aircraft Industries, Canada);

David M. Holman, Ruddy Brionnaud,
Maria Garcia-Camprubi (Next Limit Technologies, Spain)

21

In literature, some CFD works on flight simulation

consist of generating a tabular database of fundamental
aerodynamic parameters, which are later used either to
calculate static and dynamic stability derivatives, or as
lookup tables by Six-Degree-of-Freedom simulations
(e.g. Ghoreyshi et al. 2010, Lemon, K.A., 2011). The
application of this classical two-step approach is limited
since the aerodynamic forces and moments of an aircraft
with high angle of attack and large amplitude
manoeuvres, responding to sudden changes of the flow,
depend on the time history of the motion. For instance,
this approach fails particularly when post-stall motions
or propeller slipstreams are considered.
More comprehensive CFD works on the simulation of
dynamic manoeuvres consider the flow equations on
dynamic meshes, e.g. Farhat et al. 2001. For conventional
CFD codes, i.e. Eulerian approach, the handling of
dynamic meshes requires a time-consuming remeshing
process at each time step that often leads to numerical
errors and convergence issues; thus being a challenge
even for simplified geometries (e.g. Shishkin & Wagner,
2010; Johnson, 2006).
A relatively new method which has been investigated the
last decades seems to offer new capabilities to overcome
these limitations: the lattice Boltzmann method (LBM).
The LBM is a mesoscopic particle-based approach to
CFD and circumvents those moving-mesh issues, while
its refinement algorithms allow the spatial discretization
to be dynamically adjusted during the simulation,
according to the wake structure.
The CFD software XFlow has been employed for this
study, since it is based on the LBM and allows moving
geometries. The ability of XFlow to conduct rigid body
simulations concurrently with CFD analysis including
fully turbulent airflow cases has been investigated as
part of the ongoing research and development studies for
the design of future aircraft at Diamond Aircraft
Industries.
The Diamond D-JET, shown in Figure 1, is a five-seat
single engine jet currently undergoing flight testing in
Canada. Its cruise speed is 315 knots (580 km/hr) and it
is powered by the Williams FJ33-4A-19 turbofan engine. A
sophisticated data acquisition system records hundreds
of air data and systems parameters at high frequency. In
addition to flight testing, the D-JET has also undergone

wind tunnel testing at the University of Washington

Aeronautical Laboratory (UWAL) in the US and at the
Large Amplitude Multi-Purpose (LAMP) wind tunnel in
Germany.

Numerical Approach
In the literature there are several particle-based
numerical approaches to solve the computational fluid
dynamics. They can be classified in three main
categories: algorithms modelling the behaviour of the
fluid at microscopic scale (e.g. Direct Simulation
Montecarlo); algorithms which solve the equations at a
macroscopic level, such as Smoothed Particle
Hydrodynamics (SPH) or Vortex Particle Method (VPM);
and finally, methods based on a mesoscopic framework,
such as the Lattice Gas Automata (LGA) and Lattice
Boltzmann Method (LBM).
The algorithms that work at molecular level have a
limited application, and they are used mainly in
theoretical analysis. The methods that solve macroscopic
continuum equations are employed most frequently, but
they also present several problems. SPH-like schemes
are computationally expensive and in their less
sophisticated implementations show lack of consistency
and have problems imposing accurate boundary
conditions. VPM schemes have also a high computational
cost and besides, they require additional solvers (e.g.
schemes based on boundary element method) to solve
the pressure field, since they only model the rotational
part of the flow.
Finally, LGA (Hardy et al. 1973) and LBM schemes have
been intensively studied in the last years being their
affinity to the computational calculation their main
advantage. Their main disadvantage is the complexity to
analyse theoretically the emergent behaviour of the
system from the laws imposed at mesoscopic scale.

Lattice Boltzmann method

While the LGA schemes use Boolean logic to represent
the occupation stage, the LBM method makes use of
statistical distribution functions fi with real variables,
preserving by construction the conservation of mass and
linear momentum.

Figure 1: Diamond D-JET

22

Figure 2: Unified Laws of the Wall

The Boltzmann transport equation is defined as follows:

(4)

where fi is the particle distribution function in the

direction i, ei the corresponding discrete velocity and i
the collision operator.
The stream-and-collide scheme of the LBM can be
interpreted as a discrete approximation of the
continuous Boltzmann equation. The streaming or
propagation step models the advection of the particle
distribution functions along discrete directions, while
most of the physical phenomena are modelled by the
collision operator which also has a strong impact on the
numerical stability of the scheme.
Two common formulation of collision operator exist: the
single-relaxation time (SRT) and the multiple-relaxation
time (MRT). The single-relaxation time approach, .e.g.
the Bhatnagar-Gross-Krook (BGK) approximation (Qian
et al. 1992), is commonly used because of its simplicity.
Some of the SRT limitations are addressed with
multiple-relaxation-time (MRT) collision operators
where the collision process is carried out in moment
space instead of the usual velocity space

(8)

where the collision matrix Sij is diagonal, meqi is the

equilibrium value of the moment mi and Mij is the
transformation matrix (Shan & Chen, 2007; d'Humieres,
2002).
The collision operator in XFlow is based on a multiplerelaxation time scheme. However, as opposed to

standard MRT, the scattering operator is implemented in

central moment space. The relaxation process is
performed in a moving reference frame by shifting the
discrete particle velocities with the local macroscopic
velocity, naturally improving the Galilean invariance and
the numerical stability for a given velocity set (Premnath
& Banerjee, 2011).
Raw moments can be defined as

(9)

and the central moments as

(10)
By means of the Chapman-Enskog expansion the
resulting scheme can be shown to reproduce the
hydrodynamic regime for low Mach numbers (Ran & Xu,
2008; Qian et al. 1992; Higuera & Jimenez, 1989).

Turbulence Modelling
The approach used for turbulence modelling is the Large
Eddy Simulation (LES). This scheme introduces an
additional viscosity, called turbulent eddy viscosity t, in
order to model the sub-grid turbulence. The LES
scheme used is the Wall-Adapting Local Eddy viscosity
model, which provides a consistent local eddy-viscosity
and near wall behaviour (Ducros et al. 1998).
A generalized law of the wall that takes into account for
the effect of adverse and favorable pressure gradients is
used to model the boundary layer (Shih et al. 1999). The
interpolating functions f1 and f2 given by Shih et al. are
depicted in Figure 2.

The algorithms that work at molecular level have a

limited application, and they are used mainly in
theoretical analysis.

23

Treatment of Moving Geometries

The treatment of moving boundary conditions is
straightforward and similar to the handling of
fixed boundaries. In basic LBM implementations
the wall boundary conditions for straight
boundaries are typically implemented following a
simple bounce-back rule for the no-slip boundary
condition and a bounce-forward rule for the freeslip. In XFlow the statistical distribution functions
fi coming from the boundaries are reconstructed
taking into account the wall distance, the velocity
and the surface properties. The set of statistical
distribution functions to be reconstructed is
recomputed each time-step based on the updated
position of the moving boundaries. A reference
distance to the wall, velocity, surface orientation
and curvatures are taken into account in order to
solve the wall boundary condition.

Figure 3: Pitch Capture Simulation

Simulations Setup
The simulation of tests points by XFlow has been
conducted in the virtual wind tunnel featured by
the software, designed for external aerodynamics
simulations. The size of the wind tunnel is set to
40x30x20 m and periodic boundary conditions are
applied at the top and bottom boundaries, as well
as at the lateral boundaries.
The required inputs to run the simulation are:
D-JET model geometry (actual loft) with
flow through inlet
D-JET mass, centre of gravity and full
inertia tensor at the test point time
Test point airspeed, air density,
temperature and dynamic viscosity
Flight controls deflections corresponding
to the test point, slightly reduced by a
factor determined from static wind tunnel
data validation where applicable.

Figure 4: Dutch roll simulation

Flight Test Manoeuvres

The model is placed at the initial angular positions

corresponding to the test point being evaluated,
and its behaviour set to rigid body dynamics with
the relevant Degrees Of Freedom (DOF). Once the
simulation starts, no further input from flight test
data is used by XFlow. The average setup time for
these simulations in XFlow is approximately 15
min.

This section presents the XFlow numerical results for the Diamond DJET performing three types of flight test manoeuvres, namely: (i) pitch
capture; (ii) Dutch roll; and (iii) stall. The performance of the CFD tool
is evaluated by comparing its results with flight test data for the
corresponding manoeuvres. Additionally, the ability of XFlow to
simulate other kind of manoeuvres is illustrated with the D-JET
spinning.

The rigid body dynamics simulation settings were

usually as follow: 0.5m resolved scale, 0.125m
wake resolution and 0.0625m target resolved
scale.

Pitch Capture
This maneuver involves flight at a predetermined speed in trimmed
conditions, aggressively pitching up five degrees for one or two seconds
without re- trimming, then return to the trimmed condition with flight

Dutch roll frequency and damping must meet

specific requirements for acceptable flight
handling characteristics.

24

Figure 3: Pitch Capture Simulation

controls fixed. The pitch oscillation frequency and

damping are resulting parameters used to qualify flight
handling qualities.
Pitch capture is simulated with one degree of freedom in
pitch, starting at the flight test out of trim pitch angle at
0.7 seconds. The elevator deflection is fixed to the
trimmed condition as in flight test.
Figure 3 shows the pitch evolution of the D-JET for the
given test conditions, where XFlow results are
represented in orange and flight test data in black. As it
is shown in the figure, numerical results yield a similar
pitch response curve, although at higher frequency and
lower damping than the experimental one.

Dutch Roll
Dutch roll is initiated in level flight with a rudder input to
excite the Dutch roll motion, after which the flight
controls are held fixed. The resulting yaw causes the
aircraft to roll due to the dihedral effect, and subsequent
oscillations in roll and pitch are analysed for frequency
and damping. As with pitch capture, Dutch roll frequency
and damping must meet specific requirements for
acceptable flight handling characteristics.
Dutch roll is simulated by XFlow with three degrees of
freedom: pitch, roll, and yaw. The elevator is set for
trimmed conditions at 100 KIAS and 20500 ft. The
simulation starts when the rudder is centred (7.6
seconds).
Figure 4 shows both the experimental and numerical
results of this test. The agreement between simulation
and flight test data is good, with a Dutch roll frequency
only 9% above flight test. Damping is a match for the
first oscillations. Similar results are obtained at higher
speeds (up to 200 KIAS) with a slightly higher
overestimate of the frequency, but still within 15%.
Simulations at coarser resolution have shown lower
damping. In this simulation, the resolution of XFlow
would need to be increased to improve the damping
match with flight test data for oscillations below 2
degrees.
Spiral stability causes the bank angle to slowly diverge
during the Dutch Roll manoeuvre. To facilitate
comparison of the curves, this long period parameter
has been removed from flight test and XFlow bank
angles shown in Figure 4.
This 13 seconds simulation was computed in 32 hours
on a Dell Precision 7400 with dual quad-core E5440 Xeon
processors.

The Dutch roll manoeuvre is illustrated in Figure 5,

where the position of the D-JET is captured in three
different moments of the test. The images highlight the
roll motion of the aircraft.

Stall
The test point simulated here involves stall and poststall behaviour at angle of attack approaching 30
degrees. When the angle of attack goes beyond 25
degrees, the pilot pushes the nose down as this
represents a flight test limit. The aircraft is in a clean
configuration (flaps and gear are retracted).
This simulation focuses on the evolution of the angle of
attack in the post-stall regime, and the effectiveness of
the elevator in bringing the nose of the aircraft down.
Elevator deflection and airspeed are simulation inputs,
the values of which are shown in Figure 6. The Angle of
Attack (AOA) is the simulation output and it is shown in
Figure 7.
From Figure 7 it can be stated that XFlow reasonably
predicts the elevator effectiveness while the aircraft is
fully stalled, though it underestimates the maximum
angle of attack by 4 degrees. The simulation may be
improved when feedback controls will be included in
XFlow, and allow the elevator to be scheduled to
maintain altitude up to the stall. This way, the Z axis can
be added as an additional degree of freedom for
additional realism.
Figures 8 and 9 show some images of the numerical
stall test. The one shown in Figure 9 corresponds to the
moment at which the D-JET reaches the maximum
angle of attack; it can be observed how the horizontal
tail is fully submerged in the turbulent wing wake.

Spin
Flight test data for the spin test of the D-JET is not
available. Nonetheless, spin simulations have been
conducted with D-Six, a Bihrle Applied Research 6- DOF
simulation software. The D-Six simulation uses dynamic
stability data obtained on a D-JET model at the Bihrle
Large-Amplitude-Multi-Purpose Wind Tunnel.
When setting up XFlow with mass properties and prospin flight controls deflections identical to the D-Six
simulation, it was found that XFlow reached the same
stabilized angle of attack of 47 degrees but the

25

Figure 6: Stall Simulation Inputs: Elevator Deflection and Airspeed

Figure 7: Stall Simulation Output: Angle Of Attack

Figure 8: Stall Manoeuvre

26

Figure 9: Stall Manoeuvre At Maximum Angle of Attack

yaw rate was nearly twice as high. In order to

investigate this discrepancy, dynamic derivatives were
subsequently determined by XFlow by measuring forces
and moments during pitch and yaw sweeps. A
comparison of several dynamic stability derivatives is
shown below:

Wind Tunnel XFlow

Cmq Pitch damping

-30.4

-30.9

Cnr Yaw damping

-0.271

-0.212

Clr Roll due to yaw rate

0.153

0.145

Cyr Side force due to yaw rate

1.42

1.00

Yaw damping calculated by XFlow is 22% lower than

determined by wind tunnel. Additionally, XFlow
overestimates rudder control power by one third in
static conditions at the coarse resolution settings used
in this simulation. Higher computing power not
available for this study may improve the level
correlation between XFlow and D-Six.

Conclusions
The lattice Boltzmann method offers the potential of
evaluating the flight handling characteristics of any
aircraft configuration at the conceptual design stage,
and can complement wind tunnel data with dynamic
stability data including power or propeller slipstream
effects.
Indeed, a total of four flight manoeuvre simulations
have been conducted with the LBM-based software
XFlow on the Diamond D-JET developed by Diamond
Aircraft Industries in Canada: the pitch capture, the
Dutch roll, the stall and spin simulation. Except for spin
rate, overall accuracy is showing good potential: the
pitch capture has the correct frequency but too high
amplitude, the Dutch roll had a perfect match on initial
amplitudes but shorter frequency, and the stall shows
similar patterns to experiment but with lower
amplitudes in the aircraft incidence angle
demonstrating elevator control effectiveness.
Further validation studies will determine its domain of
validity and possibly allow applications beyond aircraft
design. For example, XFlow may eventually be
considered as a flight test risk mitigation tool by
simulating a range of flight test manoeuvres such as
deep stall and spins prior to actual testing.

The lattice Boltzmann method offers the

potential of evaluating the flight handling
characteristics of any aircraft configuration
at the conceptual design stage

27

REFERENCES
Chen, H., Chen, S., & Matthaeus, W., 1992, Recovery of the
Navier-Stokes equations using a lattice-gas Boltzmann
method, Physical Review A, vol. 45, pp. 5339.
Ducros, F., Nicoud, F., & Poinsot, T., 1998, Wall-adapting
local eddy-viscosity models for simulations in complex
geometries, Proceedings of 6th ICFD Conference on
Numerical Methods for Fluid Dynamics, pp. 293-299.
Farhat, C., Pierson, K. & Degand, C., 2001, Multidisciplinary
Simulation of the Maneuvering of an Aircraft. Engineering
with Computers 17: 16-27.
Ghoreyshi, M., Vallespin, D., Da Ronch, A.,Badcockx, K. J.,
Vos, J. & Hitze, S., 2010, Simulation of Aircraft Manoeuvres
Based on Computational Fluid Dynamics. American
Institute of Aeronautics and Astronautics.
Hardy, J., Pomeau, Y., & de Pazzis, O., 1973, Time evolution
of a twodimensional model system. I. Invariant states and
time correlation functions. J. Math. Phys., 14(12):1746-1759.
Higuera, F.J., & Jimenez, J., 1989, Boltzmann approach to
lattice gas simulations, Europhysics Letters, vol. 9, pp. 663668.
Holman, D.M., Brionnaud, R.M., Martinez, F.J., & MierTorrecilla, M., 2012,
Advanced Aerodynamic Analysis of the NASA High-Lift Trap
Wing with a Moving Flap Configuration. 30th AIAA Applied
Aerodynamics Conference, New Orleans, Louisiana, 25 - 28
June.
d'Humieres, D., 2002, Multiple-relaxation-time lattice
Boltzmann models in three dimensions, Philosophical
Transactions of the Royal Society of London. Series A:
Mathematical, Physical and Engineering Sciences, Vol. 360,
No. 1792, 2002, pp. 437-451.
Johnson, A.A., 2006, Dynamic-mesh CFD and its application
to flapping-wing micro-air vehicles, 25th Army Science
Conference, Orlando.
Lemon, K.A., 2011, Application of a six degrees of freedom
adaptive controller to a general aviation aircraft. MSc
Thesis, Wichita State University.
Premnath, K., & Banerjee, S., 2011, On the ThreeDimensional Central Moment Lattice Boltzmann Method,
Journal of Statistical Physics, 2011, pp. 1- 48.
Qian, Y.H., DHumieres, D., & Lallemand, P., 1992, Lattice
BGK models for Navier-Stokes equation. EPL (Europhysics
Letters), 17:479.
Ran, Z., & Xu, Y., 2008, Entropy and weak solutions in the
thermal model for the compressible Euler equations,
axXiv:0810.3477.
Shan, X., & Chen, H., 2007, A general multiple-relaxationtime Boltzmann models in three dimensions, International
Journal of Modern Physics C, Vol. 18, No. 4, 2007, pp. 635643.
Shih, T., Povinelli, L., Liu, N., Potapczuk, M., & Lumley,
1999, J., A generalized wall function, NASA Technical
Report.
Shishkin, A. & Wagner, C., 2010, Numerical modeling of
flow dynamics induced by fruit flies during free-flight, V
European Conference on Computational Fluid Dynamics,
ECCOMAS CFD 2010, Lisbon (Portugal), 14- 17 June.

Figure 10: Tunnel Model (Top) - Spin Simulation on XFlow (Bottom)

28

Improving the
Simulation of Bird
Strike on Plastic
Windshields
Parts: Mecaplex Ltd, Grenchen, Switzerland
Simulation: Aerofem GmbH, Ennetburgen, Switzerland
Project: University of Applied Sciences and Arts Northwestern Switzerland FHNW / Institute
of Product and Production Engineering, Windisch, Switzerland

29

ircraft and other vehicle windshields are exposed to a wide variety of loads, one of the most dangerous being the
impact of solid objects at high velocities. For aircraft of all kinds, a relevant load case is bird strike, happening
mostly during the landing and take off phase of flight. It is vital that such events do not completely destroy the
windshield, which can be made of different materials, for example glass, PMMA, polycarbonate and combinations of
these materials joined by rubbery interlayer materials (see Figure 1).
To ensure safety, such glazing needs to fulfil a wide variety of tests, one of which is a bird strike test where an impactor
(e.g. dead chicken or gelatine block) is shot onto the windshield using a large pneumatic cannon. These tests are time
consuming and expensive; therefore FEM is used to optimize parameters like material type and layer thickness. This
should result in only a final homologation test and no other costly test series.
In a government supported project, the Institute for product and production engineering (IPPE) at the University of
Applied Sciences and Arts Northwestern Switzerland (FHNW), Mecaplex Ltd as manufacturer of such aircraft structures
and Aerofem GmbH as design and calculation company, have joined forces to increase the prediction accuracy of such
simulations. To achieve this goal it was necessary to conduct extensive material testing as well as examine a variety of
FE material models.

Important issues

Material model calibration: An example

Finite element programs, such as LS Dyna which was

used in this project, offer a wide range of material
models, from simple linear elastic to complex
nonlinear with damage models and so on. But the
answer to the important question, which one is best
suited to the task at hand, needs to be found by the
users themselves. To find a suitable model, knowledge
about the material and understanding its loading
conditions are necessary in order to know the
capabilities that need to be included in the material
model.

As already mentioned, layers of glassy polymers can be

combined using an interlayer material, which can for
example be a thermoplastic polyurethane (TPU). These
thin films of thermoplastic elastomers show a highly
nonlinear elastic behaviour, completed by strain rate
dependency and being nonsymmetric regarding tension
and compression loads.

The first task therefore was to give some thought to the

loading of the parts and the consequences of this, in
order to ensure a reliable FEM simulation.
For example these are:
High impact velocites lead to high strain rates in the
plastic materials used. Their behaviour is strongly
strain rate dependent, meaning their stiffness
behaviour at high rates of deformation differs from
ones measured at low rates.
The impact leads to a bending deformation of the
windshield, resulting in tensile and compressive
stresses. The materials behave differently in each
mode, not only regarding the stiffening but also the
failure behaviour.
As aircraft fly in a wide variety of weather
conditions, materials need to work from low to high
temperatures. Plastics properties also change with
altering temperatures.
There are many more considerations like this. To begin
with, we concentrated on analysing the strain rate and
load condition dependency of the most commonly used
materials and on how to represent these
characteristics correctly in FEM simulations of the
birdstrike.

To find a suitable modelling method, we first conducted

a thorough study of the literature to get a better
impression of the necessary behaviours in FE
simulation of these materials. While it would also be
feasible to model such interlayers by using special
contacts (*Tiebreak in LS DYNA) or cohesive
elements, we chose to use continuum elements for the
interlayer. Only with the complex material models
available there was it possible to include all relevant
material behaviours. Two candidates were chosen as
material model:
*Mat_ Plasticity_Compression _Tension (Mat_124),
an elasto plastic material model offering the
possibility to define different base curves and strain
rate dependencies for tension/compression.
Through a Maxwell type viscoelasticity included
with a Prony Series also the elastic part can be
influenced. One drawback is that only base curves
of true stress versus plastic strain can be inputted
which are then scaled for different strain rates, but
this does not allow for a change of the curve shape
e.g. with increasing strain rate.
*Mat_Simplified_Rubber (Mat_181), described as
being a quasi hyperelastic rubber model. The
quasi term is necessary, because there is no
real strain energy function used to determine the
stresses (as with "true" hyper elasticity), but only
the tangent stiffness is derived as if an energy
function was present. Using the model, one needs
to keep in mind that, with the chosen approaches,
effects like creep and stress relaxation cannot be
modelled in a correct way, as it lacks the capability
to represent hysteresis in a visco elastic sense.
Having chosen the FE modelling method, the next step
is to get the right input data.

30

Figure 1: Bird strike

damaged windshield,
example of material
layup in aircraft
windshield (below)

Glass
Adh. Interlayer
PMMA
Adh. Interlayer
PMMA

High speed camera

Cold light lamps
Upper clamp and catch
Specimen
Load cell
Support

31

To investigate the strain rate dependency, a

purpose built measurement system was developed at
the FHNW (Figure 2). Using a fast hydropulser,
waterjet cut specimens with a special shape are torn
at speeds up to 4m/s allowing for strain rates up to
200/s. Forces are measured using a piezoelectric load
cell and strains are recorded by digital image
correlation with pictures from a high speed camera.
Compressive tests were conducted using round
samples of stacked foil, compressed on a standard
universal test machine at low strain rates. The results
were then extrapolated to higher strain rates using the
tensile results and comparable literature data from e.g.
split Hopkinson pressure bar testing.
Further testing then included cyclic loading, shear,
confined compression and dynamic mechanical
analysis (DMA), as well as comparisons between in
plane and transverse data. After this comprehensive
test program, the material can be considered well
enough known under bird strike conditions.

Material Model Verification

The tensile and compression stress strain curves
(Figure 2) are the basis for both FE material models.
Simulations of these basic tests showed that the
material models represent the tests which they are
based on well. Now, as there are many more modes in
a real impact than just tension and compression, it was
interesting to see how the material models behave
under other loading conditions. For this Arcan type
shear tests (Figure 3) as well as instrumented
pendulum impact tests (IPIT, Figure 4) were used. To
gain useful results, for some of the tests it was
necessary to add covering layers to the rather soft
interlayer material. Using PC or PMMA would have
been ideal, but as these materials also pose simulation
problems themselves, in order to ease comparison
between simulation and test of the interlayer only, thin
aluminium sheets were used as top and bottom layer
(see e.g. Figure 4).
In Figures 3 and 4 it can be seen that MAT_181 is well
suited to also represent these loads which were not
directly used as input to the material model. Note that,
to achieve these results, no further modifications of the

Figure 3: Arcan bulk shear test setup, test and simulation with stress comparison, force vs displacement curves for
different material models

32

material cards were made; only the data from

tension/compression testing were used.
MAT_124 is generally acceptable, but fails when it
comes to unloading: This is only represented by linear
elasticity, whereas in MAT_181 an unloading load curve
is defined.
For the other materials used in windshields a similar
research and verification process has been conducted,
to achieve correct material models for all of them.
Where necessary, parameter identification also made
use of optimization methods to achieve the best accord
between test and simulation.

Impact Simulation Validation

Having identified the material parameters of all
individual materials, the next step is the validation of
combinations of materials in one layered plate. Here an
impact machine developed at the FHNW comes into
operation, allowing for tests of square sheets of
material under impact load with a steel impactor,
providing measurement of impact force versus time.
The measurement is performed by metering the
acceleration of a large, air suspended mass serving
as base plate for the impacted plate. Results of such a
simulation (in this example shown with layered shell
elements) the test setup and comparison of measured
and simulated forces are shown in Figure 5.
As for the material models, LS DYNA offers a wide
range of possibilities for contacts, furthermore the
refinement of geometrical modelling as well as mesh
densities and other parameters besides the material
model itself need to be correct. Only when all these
components match can a good correlation between test
and simulation be achieved, even in the simple impact
setup shown.
Finally, the most promising combination of material
models has been put to the test by simulation of a full
canopy FE model. This simulation (Figure 6) was
compared to results produced with models prior to this
project and with bird strike tests. It could be shown that
several parameters like deformation during impact,
failure and plastic deformation after test can now be
simulated with much higher accuracy.
Conclusions and Outlook
By using newly developed or improved measurement
methods an enhanced understanding of the behaviour
of the different plastics used for aircraft windshields

Figure 4: IPIT test setup, tested sample and

simulation, force vs. displacement curves for different
material models

...the most promising combination of material models has

been put to the test by simulation of a full canopy FE model

33

Figure 5: Impact test (left), results of test and

simulation (middle), test and simulation
curves of force versus time (right)

Figure 6: FE Simulation of bird strike on canopy

was achieved. This data allowed to select the best material models and to identify their parameters to simulate bird impact
on these parts. It could be shown that the new models provide significantly enhanced prediction quality. This now allows a
further optimization of the glazings regarding thickness and material types to minimize weight. Furthermore the risk of
failure during the homologation tests is minimized, thereby giving a competitive advantage to the participating companies.
Further work on the topic will focus on the materials in a temperature range between 40 and 80C under high strain rate
loading. A facility under construction is an instrumented test rig allowing for small scale birdstrike tests by shooting a
birdlike material like ballistic gelatine at high velocities.

References:
Fritzsche P. et al: A procedure for the simulation of failure in thermoplastic composites; Composite Structures 2007
Wyss I.: Bird Impact on Multilayer Aircraft Transparencies: Investigating the Interlayer Simulation; Master Thesis 2012
Ramakrishnan K.: Low Velocity Impact Behaviour of Unreinforced Bi layer Plastic Laminates; Master Thesis 2009
Rinaldi R. et al: Modeling of the mechanical behaviour of amorphous glassy polymer based on the quasi point defect
theory Part II: 3D formulation and finite element modeling ofpolycarbonate; Int. J. of Non Linear Mechanics 2011

34

High Lift System

Virtual Test
Mr Jaymeen Amin, Dr Tobias Ulmer (Airbus Operations GmbH, Bremen),
Mr Philip Neuhaus (FTI Engineering Network GmbH,
on behalf of Airbus Operations GmbH, Bremen)

Winner of the NWC13 Best Paper Award for

Greatest Business Impact of Simulation

35

irbus High System Lift Test

Department has been
continuously working for
more than couple of years to
introduce Virtual Testing as an
additional, and equivalent test
means. This is alongside existing
established physical means of
compliance, such as Test Rigs (TR)
and Real Time Simulators (also
known as Functional Integration
Benches (FIB).
The use of computer simulated
physical models for system test has
a huge potential to improve the
verification and certification
process for the high lift system.
To fully exploit this potential it is
mandatory to guarantee the quality
of the applied modelling and
simulation methods and also it is
crucial to ensure a completely
traceable simulation lifecycle
including simulation data and
process management (SDPM)
throughout the aircraft
development. This implies the
inclusion of creating and
maintaining the relationship
between all relevant data, such as;
key results extracted from
simulation results, the models and
the procedures that were used to
produce the results, as well as, the
inputs and the processes that were
used to generate the models.
Besides this mainly data
management related aspect a
working process management is
essential for an efficient use of
Virtual Testing, e.g. by automated or
semi-automated execution of
process steps during preprocessing, solving and postprocessing.
This article gives an overview of the
high lift systems test department,
its major tasks and major
responsibilities. It introduces how
multi-body simulation models are
used today and how they were used
in the past to support the system
certification process.
It demonstrates the evolution of the
corresponding simulation data and
process management which has
lead to the system that is used
today, i.e. the SimManager based
High Lift System Virtual Testing
Portal (HLSVT Portal). Apart from

36

Figure 1: Typical High Lift System Testrig.

Figure 2: Virtual Test History and Strategy Development

its main purpose of simulation data

management, the HLSVT Portal has
also been connected to the
departmental test management
system. This provides a full
embedment of Virtual Testing into
the global test process, thus
fulfilling the global requirement of a

complete track and traceability

between system requirements and
key simulation results that are used
for verification of these
requirements.
In addition to the overview of the
global testing process, the

Figure 3: High Lift System Multi Body Simulation Model

embedded HLSVT Portal and the

Virtual Testing process, the article
focuses on the current realization
of efficient post-processing. The
evaluation of simulation results,
including flexible, definable key
value extraction and plot generation
is realized by integration of Matlab
scripts into the HLSVT Portal.
Introduction
For aircraft system certification a
huge amount of testing is required
to guarantee safe, robust and error
free behaviour under all operating
and environmental conditions.
Typically these tests on system level
are performed on physical test
benches (example see Figure 1)
where all the relevant components
including actuators, sensors and
control computer are integrated.
Due to the conflicting trends of
increasing complexity of systems
and drastically reduced
development times, virtual testing
has become one of the solutions to
overcome this challenge.
For the high lift system multi body
simulation is the preferred
approach for virtual testing. Since
2002 the high lift test department
has developed experience on this
topic within R&T projects in
collaboration with the technical

university Hamburg-Harburg,
(Glzau and Carl 2007, 1), and with
partners within the EU project
VIVACE, (VIVACE Consortium
Members, 2007, 3).
As shown in Figure 2, virtual testing
activities has been expanded and
become more and more establishe,
e.g. by using it for risk mitigation
purposes or in contributing to
system certification.
The approach is based on the
strong coupling of physical and
virtual test to obtain highest
possible confidence in the
simulation results. Starting with a
model variant which represents the
physical test bench in all relevant
details (e.g. external load
application) as shown in Figure 3
the model will be validated using
results from the test bench and
then finally extended to a close
aircraft like variant. Main
differences between test bench like
models and aircraft like models are
the application of airloads (discrete
load cylinders vs. distributed
pressure loads), the interface
conditions (attaching the high lift
system to a rigid test bench vs. a
flexible wing) and the consideration
of load dependent wing deformation
(test bench without deformation vs.

Figure 4: Physical and Virtual Test Scenario

application of wing bending and

twist according to load case under
consideration).
Significant benefit can be achieved
using this integrated approach
(Figure 4) in terms of time and cost
reduction together with increasing
quality of results.

Modelling Process for

Virtual Testing
Modeling is performed according to
a detailed modeling process as
schematically shown in (Figure 5),
starting with creation of a blockwise conceptual model (example
shown in Figure 6). It is used to
identify the systems bodies and
their interfaces, i.e. the types of the
connecting joints. Decision is made,
which bodies can be modeled as
rigid and which have to be modeled
as flexible bodies.
After validation of the conceptual
model, a so called computerized
model is set up within
MSC/SimXpert respectively ADAMS.
Therefore information from
different domains and departments
has to be merged:
Kinematics information from
CAD (e.g. kinematic points for
joint location and orientation,
gear ratios, pitches)
Masses and CoGs (Bodies
inertias are calculated in
ADAMS using simplified
geometries and adjusting
densities to get the specified
masses. In some cases
replacement of simplified
geometries by CAD geometry
only for graphical purposes.)
Flexible bodies represented by
modal neutral files (.mnf files)
based on FEM Models.
Parameters describing frictions
and damping (e.g. for joints and
flexible bodies)

37

translated to a solver input deck

(.adm file) and then executed by a
corresponding .acf file.
In general model validation is
performed by comparing time
series data or scalar values of
computer simulations to
corresponding reference data
coming from physical test results
respectively other (already
validated) simulations, e.g. CAD
results (kinematics), FEM
calculations (statics) and
measurement data (statics and
dynamics). Based on the
reference data and admissible
deviations a tolerance band for
acceptable simulation results is
generated. Measurement
uncertainties are taken into
account (if applicable) by reducing
the tolerance band to a
guaranteed validation band, see
example in Figure 7.

Figure 5: Modelling Process

Test accuracy also increases the

band for a possible validation,
therefore simulation results
leaving the guaranteed validation
band dont automatically lead to a
rejected validation but to further
investigations.

Vt Process and Simulation

Data Management
The use simulation for aircraft
system certification is not just
related to build simulation models
with sufficient accuracy and
quality. Furthermore regulations
from airworthiness authorities
also request a well defined and
robust process for the complete
data chain involved in the
certification.

Figure 6: Example of Conceptual model

For performance and archiving reasons,

simulations are conducted by using
ADAMS Solver in interactive mode,
therefore the model to be simulated and
the applied loads and motions are

Currently used requirements

based engineering (RBE) is the
formal way of developing new
aircraft and their systems. Within
the test departments this has lead
to a requirements based testing
(RBT) process.
All required functions and
properties of the system in terms
of performance, safety, etc. are
specified verbally within single
requirements managed by a
database system based on Doors
(Telelogic).
Using a Test Management System
(TMS) the formal verification of
each of the requirements is

38

assigned to one or more of the

existing test tools. Each test tool
has a local process and data
management environment. After
test execution the TMS collects all
the test results from the local
platforms and generates
automatically the required test
analysis reports and finally the
certification documentation
(coverage report).
For successful implementation of
virtual testing in the existing test
process (Figure 8), a solution was
developed based on MSC
SimManager (MSC 2008, 2). This
so-called Virtual Test Portal
(Figure 9) fulfils following general
requirements:
Interface with global TMS
Simulation Data Management
Workflow Management &
Process Control & Process
automation
Configuration Management &
Lifecycle Management
Traceability & Reproducibility
Interface with other S/W (MSC
Adams)
Diverse model library,
supporting multidisciplinary
system simulation
User-friendly Graphical
Environment
One of the most important and
critical aspects was the correct
capturing of the virtual test
process itself and its interface
with the TMS.
A detailed specification capturing
all objects, process steps and
related attributes was established
and refined during setup of the
portal.
Besides the virtual test process
today the modeling process is also
driven by the virtual test portal.
The process described in Section
0Modelling Process for Virtual
Testing is captured and
implemented by execution of
scripts (e.g. SimXpert templates)
or interactive pre-processor
sessions triggered by the portal.
Besides configuration control in
the sense of answering the
question which model was used to
generate a certain simulation
result this enables also to capture
which input data and processes
were used to generate the used
model.

To increase simulation performance, i.e. to

enable sensitivity analysis, variations of
tests or design of experiment studies,
simulation multi-runs and corresponding
script based post processing launched by
the portal are possible.

Conclusions
The article shows, how the combination of
physical test and virtual test helps to
overcome the challenge of increasing
complexity of systems on one hand and
drastically reduced development times on
the other.

Figure 7: Model validation bands

It illustrates how the verification and

certification process of the high lift system
is improved and the quality of the applied
modelling methods and the complete
traceability of the simulation lifecycle along
the aircraft development is assured.
In addition to introducing the modelling
process the presentation also shows how
model data, simulation results and
corresponding evaluation data is integrated
in existing processes for the management
of all test related data by the High Lift
Virtual Test Portal and its connection to
the existing test management system
(TMS).

References

Figure 8: Global Test process for system certification

[1] Glzau, H and Carl, U.B., 2007.

Parametric Modelling and Experimental
Validation of multi body simulation of
elastic flap systems in transport
aircraft. In: O. Estorff, ed. 2007, 1st
International Workshop on Aircraft
System Technologies. Hamburg: TUHH,
pp. 39-48.
[2] MSC, 2008. MSC SimManager R3.1
Users Guide. MSC.Software.
[3] VIVACE Consortium Members, 2007.
Final Technical Achievements. System
Simulation, pp. 12ff.

Definitions / Abbreviations
VT
TMS
mnf
CoG
adm
acf
RBE
RBT
S/W
H/W

SDM
MBS
FEM

Virtual Test
Test Management System
Modal Neutral File
Center of Gravity
Adams Solver Dataset File
Adams Solver Command File
Requirements Based Engineering
Requirements Based Testing
Software
Hardware
Simulation Data Management
Multibody Simulation
Finite Element Method

Figure 9: Virtual Test Portal

39

Improving Structural
Modelling of High
Strain Rate Behaviour
of Composite Materials
Using High Speed
Imaging
Duncan A. Crump, Janice M. Dulieu-Barton, and Stephen W. Boyd
University of Southampton

There is a drive towards producing lighter vehicles that are faster, more manoeuvrable
and more fuel efficient to improve the sustainability of transport systems. The excellent
specific stiffness/strength properties of fibre reinforced polymer composite, e.g. carbon
and glass fibres, make them an increasingly attractive option for structures in high-end
and military applications.
These complex materials are being used in applications where there is a real risk of
impact or high velocity loading, whether this is bird strike on passenger aircraft (Figure
1), slamming loads on marine vessels or explosions in the proximity of military vehicles.
For efficient structural design it is vital that accurate and pertinent material properties
are available for input into finite element (FE) models. While the quasi-static behaviour
of composite materials is generally well understood [1], there is a need to for further
analysis at high velocity loading [2, 3].

40

n 2009 a research project was

initiated at the University of
Southampton with funding from
EPSRC and DSTL to use high speed
imaging to capture strain and
temperature evolutions from high
velocity tests to provide materials
characterisations for input into FE
models. During the project a
synchronised approach was
developed using, the now wellknown, digital image correlation
(DIC) technique [4] on images
captured using a high speed camera
and infra-red thermography (IRT) [5]
using a commercial detector from
FLIR Infrared systems, which
required considerable modification to
enable accurate temperature
measurements at high framing rates
(Figure 2). A key part of the set-up
was the use of a specialised servo
hydraulic test machine (VHS), from
Instron, capable of actuator speeds
up to 20 m/s and loads of up to 80
kN. The VHS machine is equipped
with a Kistler piezoelectric load cell
with a measuring range up to 400
kN, which is coupled to a charge
amplifier to output a full scale of 100
kN across 10 V.
A methodology has been devised that
enables the IR and white light
cameras to be triggered by the test
machine and capture images from
both systems simultaneously. It has
been shown that the methodology
can provide consistent material
characterisations and failure mode
information at a range of strain rates
up to 100 s-1.

Capturing Full-Field, DataRich Strain/Temperature at

High Speeds
A major consideration in using
optical techniques for high speeds
tests is the image framing rates, e.g.
for a test velocity of 10 m/s the test
on a composite specimen has
duration of approximately 1 ms and
hence image framing rates above 10
kHz are necessary. The following
describes the capture of strain and
temperature at such rates using fullfield optical techniques. DIC [4]
tracks a random speckle pattern
(usually applied by spray paint) on
the surface of a specimen by
comparing images in the deformed
and undeformed states. The
algorithm divides the image into

reproduced with permission from

Mick Bajcar and Airliners.net

Figure 1: Composites are Being used in Applications with an Increased

Risk of Impact such as Bird Strike on Passenger Aircraft.

Figure 2: To Fully Characterise Material Behaviour at Elevated Strain Rates a

Synchronised Approach has been Developed using DIC on Images Captured
with High Speed Cameras and IRT to Provide Strain and Temperature.

correlation cells of a fixed number of

pixels, e.g. 32 x 32 pixels, and then
tracks the pattern within each to find
the deformation vectors (Figure 3),
and hence by differentiation the
strain. The strain resolution is
dependent on the number of pixels in
these correlation cells, and therefore
it would be ideal to use as large a
cell as possible but this will result in
a decrease in spatial resolution. To
achieve an appropriate image
capture rate it is necessary to use
high speed cameras such as the
Photron SA series. To achieve frame
rates into the kHz range the sensor
element must be sub-windowed to
allow appropriate data processing

rates, which affects both the strain

and special resolution. Therefore
application of the DIC technique is a
compromise between temporal,
spatial and strain resolution. The
high frame rate also requires a fast
shutter speed to avoid image
blurring. Therefore high intensity
illumination is required to provide the
sensor with sufficient excitation over
the short capture period.
In contrast to white light imaging
infra-red thermography (IRT) [5]
relies on the detection of a
temperature evolution resulting from
a physical change in the material
under load. Therefore it is not

41

in the kHz range it was necessary to

sub window and perform a detector
by detector calibration using a frame
of 64 x 12 detector elements (Figure
4). For accurate temperature
measurement IRT requires a matt
surface with high emissivity free from
reflections, which had to be
considered alongside the lighting
requirements of the white light
imaging and the painted speckle
pattern required for the DIC.
Figure 3: Digital Image Correlation Process to Track the Movement of
Random Speckle Pattern to Measure Deformation and Strain.

Figure 4: Uncalibrated and Aalibrated IRT Data Captured at 15

kHz from a Commercial Detector Demonstrating the need for
an in-house Non-Uniformity and Calibration Routine.

Figure 5: Example of the Full-Field Strain and Temperature Data Captured

from a test on a Woven Glass Fibre Specimen at 1 m/s

possible to increase the intensity of

the emission by increasing
illumination. However all the other
image capture considerations apply,
so accurately measuring
temperatures at high speed using IR
imaging is extremely challenging as

42

the maximum framing rate is limited

by the sensitivity of the IR detectors.
A commercially available IRT system
was used that has a detector array of
256 x 320 sensors and can capture at
frame rates of 383 Hz at full frame.
To make the system work accurately

Notwithstanding the above, the

greatest challenge in developing the
methodology was combining the
white light and IRT to provide
synchronous image capture and
accounting for the conflicting
requirements of the two approaches.
In a nutshell, the image capture is
triggered by the test machine which
provides a consistent common
datum. Painting the test specimens
with speckle pattern is an art rather
than a science, so a technique has
been developed that provides the
required speckle in the white light
spectrum with the matt finish in the
IR spectrum. The effects of the
lighting in the IR measurements
have been resolved by using banks of
cold LED lights.
Example Data from Gfrp Specimens
To highlight the capability of the
methodology to obtain meaningful
inputs for structural models, some
example data from woven E-glass
epoxy specimens is provided here.
Strip specimens were manufactured
with dimensions 20 mm wide, 150
mm long and nominally 1.5 mm
thick. The specimens incorporated a
wasted section at the centre of the
gauge length to localise failure
ensuring the IR data was captured in
the region of failure. Specimens
were tested at a range of strain rates
from near quasi-static (0.001 m/s) to
immediate strain rate (10 m/s). The
high speed images were captured
using a Photron SA 5 high speed
camera at a range of framing rates
depending on the test speed, i.e. 50
Hz for 0.001 m/s and 30 kHz for 10
m/s test. The IRT detector was set to
capture at 50 Hz for the slowest test,
but at 15 kHz for the other speeds.
DIC was applied to high speed white
light images using the DaVis 8.0
image correlation software from
LaVision. Figure 5 shows an example

There is a real influence of strain

rate on failure stress and
temperature evolution, with an
almost 60% increase in failure stress
and 100% increase in heat
generation. The link between the
two is also of importance to produce
accurate material models to properly
simulate the viscoelastic behaviour
in these complex materials.

Figure 6: Example of Stress/Strain Curves Captured at 0.001, 1

and 10 m/s from Woven Glass Fibre Specimens

The outcome of the project is a

facility at the University of
Southampton that can offer high
quality research and characterisation
of high speed testing on composites
(or other materials) using high speed
imaging to provide a rich vein of
information to assist in structural
modelling and failure investigations.
The University of Southampton is
forming a high speed imaging
consortium to enable industry to
access the unique facility where
strain and temperature evolutions
can be captured at high speed.

Acknowledgements
The research was funded by the
Engineering and Physical Sciences
Research Council (EPSRC) and the
Defence and Science Technology
Laboratories (DSTL). We would like
to thank the continuing support of
the numerous academic and
industrial collaborators who formed
the industrial advisory board during
the project as follows; Aalborg
University, Airbus, Arts et Mtiers
ParisTech, FLIR ATS, GE Aviation,
LaVision, Lloyds Register.
Figure 7: The Effect of Strain Rate on Failure Stress and Temperature Change

of the full-field data captured at 1

m/s presenting the evolution of the
strain and temperature maps across
the specimen, and plots of the
stress/strain curve (taken from test
machine load cell and DIC) and
temperature change averaged across
the entire specimen. This rich vein of
information is able to provide
constitutive properties at a range of
strain rates, and can also give
detailed analysis of the failure modes
and heat generation across the
specimen. The benefit of using such
data rich full-field techniques
becomes apparent when applied to
tests at elevated strain rate where
the strain rate can be measured

directly across the entire specimen

and it is clear this is not constant
during the load application or across
the specimen.
Figure 6 plots the stress/strain curve
from a single result at each of the
test speeds used that is indicative of
the behaviour of all specimens. As
the test speed is increased there is a
significant increase in both the
failure stress and failure strain of the
specimen. There is also some
evidence of nonlinear behaviour.
Finally, Figure 7 plots the effect of
strain rate on both the failure stress
and the overall temperature change
of the woven specimens tested.

References
1.

2.

3.

4.

5.

I. M. Daniel and O. Ishai. Engineering

Mechanics of Composite Materials. Oxford
University Press, 1994.
I. M. Daniel, et al. Strain-rate-dependent
failure criteria for composites. Composites
Science and Technology, vol. 71, p 357-364,
2011.
A. M. S. Hamouda, and M. S. J. Hashmi.
Testing of composite materials at high
rates of strain: advances and challenges.
Journal of Materials Processing
Technology, vol. 77, p 327-336, 1998.
M. A. Sutton, J. Orteu and H. W. Schreier.
Image correlation for shape, motion and
deformation measurements. Springer,
2009.
X. P. V. Maldague. Theory and practice of
infrared technology for non-destructive
testing. John Wiley and Sons, ed. K.
Chang, 2001.

43

benchmark january13

Rocket Science
At the recent Siemens NX CAE Symposium, held in Charlotte, NC, USA,
benchmark took some time to speak to Nathan Christensen of ATK
Launch Systems about their analysis processes, and use of simulation.

Nathan joined ATK as a design engineer in composite structures, designing and analyzing missiles and rockets. He
spent a significant portion of his 28-year career working with PLM/CAD/CAE and computational tools for design and
analysis. Christensen is one of the technical founders of ATKs PLM system, which now manages hundreds of thousands
of pieces of product and engineering information used at ATK facilities across the US. He has published numerous
technical articles and papers on rocket motor design and analysis, CAE tools and computational methods. He also holds
a patent for hybrid pressure vessels.
Christensen was first appointed manager of the CAE group in 1992, with responsibilities for engineering computational
tools and methods. In his current position as manager of Engineering Tools and Analysis group, his responsibilities
include PLM/CAD/CAE tools, trend analysis, rocket motor performance databases, analytical methods and software
development, reliability engineering and high-performance computing.

44

Background image courtesy of ATK Launch Systems. Nathan

Christensen photographed by Branco Liu, Siemens

How do you benchmark your

use of simulation compared
to other organizations?
Traditionally we review our work with
our customers, prime contractors and
joint venture partners comparing
tools, methods and approaches with
their technical people. More recently,
weve been working with several
different partners, like GM and
Northrop Grumman to benchmark
inside and outside of our business
sector. Through organizations like
Siemens PLM World, we have made
cross-industry contacts that are

So this cross-industry
cooperation and discussion
must really help you drive
your processes forward?

interested in sharing best practices and

in benchmarking. By benchmarking
ourselves, we have adopted a strategy
to understand and apply cross-industry
best practices to improve what we do in
our own business sector.
We are under a lot of cost pressure
internally to do more with less. Both
the US economy and government
spending is down substantially in
certain areas of aerospace. The
challenge for ATK is to produce the
same great products for less money or
produce improved products for the
same price.

Weve done benchmarking with GM,

Ford, Lockheed, and Northrop
Grumman. Its been great working with
GM since were not in the same
industry. Thats one reason GM is
participating too. Since we dont
compete in the same market space, we
can share more openly than we would
with a competitor in our own market
sector. GM was looking to non
competing industries to share and
benchmark themselves. We thought
that was a great idea, and weve started
doing quite a bit of that too.

your own industry sector. Looking

across industry certainly has made us
think more about the processes we use.
I think that historically aerospace has
focused primarily on performance and
safety. Cost and efficiency has taken a
back seat. These days, we are using
tools like value stream mapping and
Toyota production system to improve
our cost and efficiency.

Its interesting to see that race shops

like Joe Gibbs Racing are making part
and system changes in a week that
would take aerospace a year or more to
implement. I think that there is a lot to
be learned when we share ideas and
methods across industries.

It seems like aerospace has

always been at the forefront
of analysis and simulation.
Things start with aerospace
and then falls down to
automotive etc etc?

Youre right, we in aerospace have been

using numerical simulation since the
beginning. Automotive has really
leveraged that development and
expertise from aerospace. What has
changed from the early days is that
analysis and simulation has become
much more trustworthy. People are

using it for things we never even

dreamed of before. Things like golf
clubs, recreation equipment, consumer
packaging, and high speed machines
used to make consumer goods like
Pringles potato chips, diapers or
shampoo.

So how has simulation

changed what you do over
the time youve been using
it?

dollar test, they dont want any

surprises. If our simulations dont
predict exactly what is going to happen
in the physical tests -- its a bad day for
our Engineering team.

I think the biggest change is the

acceptance of simulation. A major
simulation project I was involved in
early in my career was the Space
Shuttle Challenger failure in 1986.
During the failure investigation and
subsequent redesign, we relied heavily
on simulation to understand the O-ring
seal and joint failure mechanisms that
are attributed to causing the
Challenger disaster. At that time,
simulation was only accepted when
accompanied with significant validating
physical tests. Today, analysis leads
physical testing. Our customers
demand that we run extensive
simulations and predict exactly what
the physical testing will show prior to
testing. When we run a multi-million

When I first started as an engineer,

analysis was nice to have but it wasnt
really a trusted source or even viewed
as necessary for design sign-off.
Physical tests were the final word and
were required for design validation.
Today, we wouldnt dream of coming to
a critical design review with our
customers, without an array of
simulation models and results.
Nowadays, we have to predict what the
test is going to do with a high degree of
integrity well in advance of the test. We
have to understand and share with our
customers, exactly what the test results
will be. If we dont match up with test
results, all kinds of issues, problems
and questions will ensue. For example,

we conducted a large solid rocket

motor booster ground test about a year
ago where we had some problems with
a test controller. It took weeks to
simulate the issue, resolve the
controller issues and reschedule the
test.

I think that it does; I think it gives us a

different perspective. Sometimes its
easy to get too comfortable in your own
processes and in accepted practices in

Our business has become extremely

risk averse and simulation is a key
technology. In a lot of our designs, we
have to predict, within a very narrow
margin, what testing is needed and
what it is going happen. If we dont, its
back on us: whats wrong with the
analysis, what went wrong with the test,
and what didnt you understand?
So whats changed? Historically testing
was the final word; now testing is just
the final step in the part certification
process. ATK customers are very
demanding. NASA is extremely
demanding in that area, and if we dont

45

hit our predictions right on, then there

is a lot to explain. Were very careful
with our analyses, and we do a lot of
subscale testing to make sure we
correlate our models. Well conduct a
static motor ground test, sometimes
recording 200-300 channels of data.
Were actually looking at
accelerometers, strain gauges and
pressure transducers trying to match
everything weve already predicted in
simulations then correlate our models
to make sure our simulation is right.

We recently completed and test fired

the Orion Multi-Purpose Crew Vehicle
launch abort motor. The reverse thrust
motor with nozzles at the head end of
the motor was a concept first
demonstrated by the Russians, but this
was the first time its been
demonstrated on a US space program.
It is an interesting design aimed at
accomplishing two things: 1) re-locate
the pressure center and center of mass
in a manner that increases
aerodynamic stability and 2) move the
rocket motor exhaust and heat from the

nozzle as far away from the crew

capsule as possible. This design is a
substantial variation from the Apollo
launch abort systems, which used aft
thrust nozzles and produces
substantially less thermal load on the
crew capsule.

So now youre involved in

using multi-physics
simulations?

minimal intervention). We use

commercial off the shelf and internally
developed solvers with internally
developed solver coupling software to
perform multi-physics analyses.

of the motor. When the motor ignites,

the entire surface is burning from the
center radially outward until all of the
fuel is consumed. The expanding gases
travel along the center bore and exit the
nozzle to create the thrust that propels
the rocket. If not designed properly,
exiting gas creates a structural load
closing down the center bore resulting
in an overpressure that can cause the
rocket to explode catastrophically.
Thats just one of the typical multiphysics simulations we routinely run
today in our design and analysis
process.

Yes, we routinely do a lot of multiphysics analyses that we didnt do even

10 years ago. We regularly couple fluid
flow, thermal and structural analyses.
This is done either as loosely coupled (a
manual method where results from one
discipline fed as input to the next) or
tightly coupled (an automated method
where solvers iterate and converge on a
multi-physics based solution with

So what software do you use

for this?
We typically use a combination of
commercial off the shelf software
coupled with internal software. In the
aforementioned bore choking example,
SIMULIAs Abaqus and Ansyss Fluent
commercial codes are coupled with an
ATK internally developed FTSI (FluidThermal-Structural Interaction) code
called FEMBuilder. We term this
analysis method tightly coupled even
though the physics is not integrated
into a single solver, because the
computational interactions are largely
handled automatically by the FTSI code.
Were also using other commercial

What about data

management?
ATK is using Siemens Teamcenter
Enterprise and Teamcenter Unified
suites for PLM. Weve been using PLM
for nearly 30 years and have a very
mature implementation. We are long
time Tc Enterprise customers
migrating to Tc Unified. Tc Unified is
our 6th generation PLM system. ATK is
an agglomeration of
smaller aerospace
companies which have

46

For example, today we regularly

analyze fluid-structure interactions
(FSI) in solid propellant grain (shape)
design. A typical design concern we
regularly analyze with FSI is bore
choking. In a solid rocket motor, the
solid fuel typically has a center
perforation (hole) down the full length

codes like NX NASTRAN, NX Thermal,

NX Simulation, Hypermesh, Optstruct,
Ansys, iSight, ADAMS and MATLAB in
multi-physics simulations. We
generally try to use commercial
software wherever we can and only
develop internal code when we cant
find adequate or advanced capabilities
commercially. Since much of our
design and analysis work is export
controlled or restricted by ITAR
(International Traffic in Arms)
regulations, we use numerous
government and internally developed
software for multi-physics work.
Other ATK internally developed
software includes HERO

been united under the ATK brand. This

legacy brings challenges in managing
and uniting business processes under a
single PLM system. ATK PLM manages
5 CAD standards with the Teamcenter
suite (NX, Catia, Pro/E, AutoCAD and
Solidworks). PLM at ATK manages the
full gambit of product and process data
including procurement, design,
simulation, manufacture, inspection,
test and refurbishment.
PLM is implemented at more than half
of ATKs 60 sites in the US. PLM is

This first use on a US space program

was correct on the first test firing. In
this very complex rocket firing, we hit
all of our simulation predictions within
a few percent. Our customers including
NASA were very impressed.

(thermal/ablation solver) CaseBuilder

(composite design/analysis), SHARP
(fluid flow) and RECESS (propellant
ballistics).
We have benchmarked a few
commercial integrated solver multiphysics packages in the past, but they
didnt work out as well for us. Perhaps
future versions will do better. I think
these types of packages are getting
there, put I think its a little bit
premature for the level of sophistication
and complexity that we need. Generally
speaking, I think an integrated solver is
will be easiest to use, they just lack the
sophistication needed for complex
interactions.

administered through a corporate

Center of Excellence (COE). The COE
model minimizes development time
and the necessary investment. ATK
sites act as individual profit centers and
are sometimes hesitant about making
investments in PLM. We find that when
sites are hesitant to implement PLM its
because they dont understand the
business benefits. PLM is very
complex. It takes time, but people get
on-board once they understand it and
see the benefits.

David Quinn

honeycomb

Fatigue in
Aluminium
Honeycomb-core
Plates
Laurent Wahl, Arno Zrbes, Stefan Maas and Danile Waldmann, from the
University of Luxembourg, investigate the fatigue properties of the honeycomb
core of aluminium sandwich panels, as used throughout the aerospace and
automotive industries.

47

oneycomb composite
lightweight structures made
of aluminium or aramid
fibres are used in airplanes, railway
carriages and automobiles. These
structures are subjected to dynamic
loading but hardly any fatigue
properties of the honeycomb core
exist in current literature (A
summary of the state of the art:
[1]).

Table 1: Material and dimensions of examined Panels

The lightweight panels which were

investigated are made of a
honeycomb core of aluminium,
which is connected by an adhesive
layer with two outer sheets of
aluminium (Figure 1).
During this project, fatigue tests
with failures of the core structure
were conducted in parallel with
Finite Element calculations. An
analytical model was created,
which explains the experimental
results.

Figure 1: Sandwich structure with honeycomb core [2]

Since the behaviour of the panels is

orthotropic, the panels react
differently depending on the
direction of the loading. For this
reason, it is necessary to
distinguish between the three
directions of symmetry, which are
called L, W and T direction (Figure
2).
The walls of the honeycomb cells
have different wall thicknesses.
This is due to the manufacturing
process, where the foils are partly
glued together. The glued walls
with double thickness are called
ribbons (Figure 2). The dimensions
of the examined panels are shown
in Table 1.
Failure Modes of Honeycomb core
Sandwich Panels
In a 3-point bending test, sandwich
structures are mainly subjected to
three types of stress:
Tension / Compression in the
cover sheets due to bending
Shear stress in the core
Compressive stress in the core
in proximity of the load
application
Each stress type must be examined
in order to figure out which is the
critical one.

48

Figure 2: L, W and T directions [2]

(Ribbon has double thickness, due to the manufacturing process)

The bending stress leads to cracks

in the face sheets, which was
examined in a former project [3].
The core of the sandwich panels
usually fails due to shear or
compressive stress (Figure 3). The
type of stress which prevails,
depending of the geometry and the
load application, is responsible for
the core failure.
The distribution of the stresses in
Figure 3 was simulated in ANSYS by
moving the load horizontally. The
shear stress is maximal
somewhere between the two points
of the force application. The
compression stress in the core has
a maximum just below the middle

load. Core indentation is occurring,

when the compression stress
surpasses the buckling strength of
the honeycomb core. In this case,
the structure fails locally due to
buckling of the core (Figure 3).

Materials
The sandwich structure consists of
three different materials:
Glue
Aluminium alloy AlMg3 H44 (AW
5754) for the face sheets
Aluminium alloy AlMn1Cu H19
(AW3003) for the honeycomb
structure

Test Methodology
Dynamic 3-point bending tests were

Figure 3: Stress distribution and failure modes of the honeycomb core

performed in order to provoke core

failure. The test setup is powered by
a hydraulic cylinder from Instron
Structural Testing Systems (IST). The
hydraulic cylinder can be excited
displacement or force controlled.
In order to provoke the two failure
types of Figure 3, the load was
applied in two different ways:
Steel roll with a small diameter
(25mm), which implies a high
contact pressure and the
component fails by core
indentation (Figure 5a)
Elastomeric roll (Vulkollan 80
Shore A) with a big diameter
(76mm), which implies a low
contact pressure so that the core
fails due to the shear forces
(Figure 5b)

Fatigue Test Results

Dynamic tests were carried out to
study the fatigue properties of the
structure. The samples were loaded
in a three-point bending test with a
sinusoidal load with constant
amplitude at a power ratio of R=10.
The excitation was force controlled.
The soft load application (Figure 5b)
leads to shear failure in the

honeycomb core. Cracks are initiated

in the interior of the honeycomb core,
which grow predominantly in the
diagonal direction of the cells (Figure
6). These cracks are not exactly
under the load, but some cells away
from it. Here the shear stress is
maximal, as shown in Figure 3.
If a hard load application is chosen
(Figure 5a), the specimens fail due to
the pressure load induced by the load
(core indentation). In the damage
pattern of Figure 7, it can be seen,
that the cracks are exactly under the
load application. The W-specimen
shows horizontal and diagonal
cracks in the cell walls. The L
specimen shows only horizontal
cracks.
The tests showed that first cracks
occurred after less than 10% of the
total life period of the specimen. The
buckling process creates locally high
stresses and cracks, which are not
imperatively leading to the total
direct failure of the structure.
In Figure 8, the fatigue diagrams of L
and W-samples with identical
dimensions are shown. In the
ordinate of the fatigue diagram the

force amplitude is displayed and not

the stress amplitude at the location
of the crack initiation. These two
values are related, but the
relationship is not necessarily linear.
The number of cycles on the
abscissa corresponds to the number
of cycles to complete failure of the
part and not until the first crack.
These boundary conditions imply that
the diagrams are not conventional
SN-diagrams. The experimental
results, however, lie well along a
straight line.
The curves of the specimens which
fail due to buckling (core
indentation), are flat, compared to
the shear failure curves. This flat
curve is due to the high nonlinear
stress increase during buckling.

Simulations
A model of the sandwich structure
was created using ANSYS. The
structure is modelled with shell281
elements, which have 8 nodes with 6
degrees of freedom each. Shell281
elements are also suitable for large
deformations and plastic behaviour.
The roll for the load application is

Table 2: Mechanical properties of the materials used in the Sandwich panels

49

still be modelled in order to examine

the influence of the glue to the
buckling load. The adhesive layer
covers the honeycomb core and
stabilizes it. This overlap is simulated
by expanding the shell model of the
honeycomb core with two layers of
constant thickness, which have the
properties of the adhesive (Figure
10). The simulations showed that the
influence of the glue to the buckling
load is less than 10%.

Core indentation (Buckling

of the core)

Figure 4: Three-point bending setup used for fatigue testing

modelled with solid95 elements,

volume elements with 20 nodes with
3 degrees of freedom each. The
contact condition between the roll
and the sample is modelled with the
elements conta174 and targe170.
These elements have 8 nodes and
are placed on the surface of the shell
elements. Contact occurs when the
surface of a conta174 element
penetrates one of the targe170
elements.
To make the simulation as realistic
as possible, several imperfections
are introduced (Figure 9):

Roll not centred (load inserted on

ribbon or next to ribbon)
( <1mm)
Roll rotated around the x-axis, so
that the device wont be loaded

evenly ( <0.2)
Roll rotated around the z-axis
( <0.5)
Cells are not regular hexagons
(all the coordinates are moved by
a small random value)
( <0.3mm)
Cells not planar (Small forces
(Fi<0.5N) are inserted into the
simulation, which dent the walls)

The simulations showed that a

rotation of the roll around the x-axis
( ) has a big influence. A horizontal
displacement of the roll ( ) can
move the force application from a
ribbon to a free wall, which also has
an influence on the results. All other
imperfections are quite insignificant.
The failure of the glue is not the main
subject of this article, but it should

The experiments have shown that the

samples, loaded with the hard roll,
failed in the mode of core
indentation. Physically, core
indentation of honeycomb panels
means that the cell walls are
buckling (but usually they can still
carry loads). The buckling process
induces bending stresses in the cell
walls, including high tensile stresses.
These tensile stresses influence the
fatigue behaviour of the core very
negatively, so that the crack initiation
phase gets much shorter (Figure 8).
In most cases however, these local
cracks barely influence the strength
of the structure.
Some simulations showed however,
that it would be too time-consuming
to simulate the growth of the cracks
(which is very sensitive to
imperfections) within this project, in
order to see which cracks lead to
failure and which cracks not.
Therefore, it was assumed that a
cyclic buckling of the honeycomb
cells is not tolerable if a part is
dimensioned against fatigue. In this
case, the stresses are distributed
more uniformly and the crack
growing process is not so important.
This assumption does not lead to a

Figure 5a: Hard load application made of steel

50

Figure 6: Fatigue Shear failure

big oversizing of the part, because

the fatigue curve of the core
indentation in Figure 8 is very flat.
Therefore, it does not make a big
difference if the part is dimensioned
for 100,000 cycles or one million
cycles (it is assumed, that for 1M
million cycles, no buckling is
occurring). In the field, core
indentation is usually avoided by
reinforcing the panel at the position
of the load application.
For this reason, honeycomb
sandwich structures should be
dimensioned so that no buckling
occurs, because only in this case can
good results be achieved. The
buckling load can be calculated by a
Finite Element Method in two
different ways. First, by a buckling
analysis, that calculates the
theoretical buckling load for a
perfect elastic system (Euler
analysis). Alternatively, if
nonlinearities have to be considered,
the buckling load can be evaluated
out of a nonlinear simulation. The
contact of the hard load application
(Figure 5a) is not very load
dependant, so that the linear
buckling analysis is possible.

However, the soft load application

(Figure 5b) causes a nonlinear
contact condition, so that in this case
a nonlinear simulation is necessary.

occurring, the most damaging stress

component in the core is the shear
stress. Therefore, the critical location
can be determined from a shear
stress contour plot. At this location, a
fatigue prediction can be
accomplished using the FKMguideline [4].
In a 3-point bending test, away from
the load application, the shear stress
can be checked analytically:

Shear failure
After proving that no buckling is
occurring, a normal static analysis
can be accomplished. In this case,
the stress state in the core is quite
homogeneous and it will be possible
to do a fatigue prediction with this
analysis. Just under the load
application, the compression stress
dominates, and next to the load
application, the shear stress
dominates like shown in Figure 3.
When no buckling of the core is

Figure 7: Fatigue buckling failure for W and L specimens

In this formula, it is assumed that

the shear stress is distributed
uniformly over the honeycomb cells.
In the element coordinate system,
the angle of the cell walls does not
appear in the formula of the shear
stress. However, the number of cell
walls across the width is important,
which implies that the L-samples are
much more stable against shear than
W-samples (n much bigger for Lsamples than for W-samples).
These approximate formulas are just
used to understand the influence of
the parameters and to check the
simulations. In order to have the
exact stresses with all local effects,
the Finite Element simulations are
still needed.

51

Figure 8: Fatigue strength diagrams for L und W samples

Fatigue analysis of examined

specimens
The procedure of the fatigue analysis
for the core structure of an
aluminium honeycomb sandwich
should be as follows:
Determine the buckling load of
the core. Applied load must not
exceed this value

Determine the stresses in a static

Finite Element Analysis
Locate the critical points (e.g. in a
contour plot of shear stresses
because these stresses are
predominating)
Calculate the lifetime of the
honeycomb core, using the FKMguideline [4]

Confirm the results when

possible by tests

Buckling loads
The buckling loads of different
specimens and different load
applications are shown in Table 3.
The buckling loads of the soft load

Figure 9: Imperfections included in the simulations (illustrated exaggeratedly)

52

Figure 10: Glue on Honeycomb cells modelled by different layers

Table 3: Comparison: buckling load / fatigue limit

Table 4: Life comparison: FKM / experiment

application cannot be compared

with the test results, because in
these cases, the failure mode is not
core indentation, and therefore no
buckling is occurring. These
buckling loads are higher than the
failure loads in Figure 8, so the
failure is not due to buckling
effects, as it is also shown in the
experiments.
The buckling load of the hard load
application is exactly in the area of
the fatigue limit found in the
experiments. It was assumed that
at the fatigue limit no more
buckling is occurring, and so in the
experiments the load at the fatigue
limit is exactly the buckling load. In
these cases, the endurance limit
can be predicted with an accuracy
of approximately 10%.
In Table 3, nonlinear analysis
means, that the contact surface is
changing with load and that the
deflections can grow at the
buckling load nonlinearly.
Shear failure
When no buckling is occurring, a
fatigue analysis is performed using
the FKM-guideline [4].
The hard load application leads to
core indentation: here, only the soft

load application is examined. In

Table 4, the results of an L and a Wspecimen are compared with the
test results.

The differences between the

predicted and the tested lifetimes
are only 10% based on stress and
only a factor of two in real-life.

The lifetime predictions for the two

cases examined in Table 4 are
conservative. The durability found in
the tests was higher than the
prediction by a factor of two. This
can be considered to be a good
prediction [5].

Acknowledgement

Conclusions
Two different failure modes of the
honeycomb core structure were
examined: core indentation and
shear failure. Core indentation
induces buckling of several
honeycomb cells. This results in
high tensile forces, which will
quickly initiate cracks. In practice,
components should be designed so
that no buckling occurs. The
buckling load can be calculated
easily with a Finite Element
simulation.
The shear failure mode can be
analysed by doing a static Finite
Element Analysis. Afterwards a
lifetime analysis can be done using
the FKM-guideline. There were only
small differences between the
fatigue predictions and the
experiments.

The materials were sponsored by

Eurocomposites, Echternach,
Luxembourg. Many thanks to them.

References
[1] Sharma, N.; Gibson, R.F.; Ayorinde, E.O.:
"Fatigue of Foam and Honeycomb Core
Composite Sandwich Structures: A
Tutorial", Journal of Sandwich Structures
and Materials 2006 8: pp 263-319
[2] Blitzer, T.: "Honeycomb Technology,
Materials, design, manufacturing,
applications and testing", first edition,
Chapman & Hall, 1997
[3] Bauer, J.: "Ermittlung von
Schwingfestigkeitseigenschaften fr
Leichtbaupaneele mit Wabenstruktur",
University of Luxembourg, Faculty of
Science, Technology and Communication,
PhD-FSTC-5-2008, 2008
[4] Hnel, B.; Haibach, E.; Seeger, T.:
"Rechnerischer Festigkeitsnachweis fr
Maschinenbauteile aus Stahl, Eisengussund Aluminiumwerkstoffen FKMRichtlinie", 5th edition, VDMA Verlag
GmbH, 2003
[5] Schijve, J.: "Fatigue of Structures and
Materials", first edition, Kluver Academic
Publishers, 2001

53

Coupling 1D and 3D CFD

The Challenges and
Rewards of Co-Simulation
Vincent Soumoy of EURO/CFD and David Kelsall of Flowmaster Ltd, both members
of the NAFEMS CFD Working Group, provide an overview of the recent NAFEMS
UK seminar on coupling 1D and 3D.

he benefits of coupling 1D and

3D CFD codes have long since
been r0ecognised.
Automotive and aerospace
companies have used 1D codes to
gain a better understanding of
system performance (such as fuels
systems), whilst 3D codes are used
to analyse detailed behaviour within
and around key components. With
that in mind, the NAFEMS CFD
Working Group recently arranged a
seminar at the Heritage Motor
Centre in Gaydon UK to understand
the benefits of such links and
assess the current state of the art.
Approximately 40 interested parties

54

from across the NAFEMS

membership attended to hear a
number of interesting and thoughtprovoking presentations from
various speakers.
Darren Morrison started the
technical presentations by sharing
an interesting view on the subject
from the perspective of a large
aerospace company (AIRBUS).
Validation is seen as desperately
important, so that much of their
work is to prove that any couplings
are producing realistic and
reasonably accurate predictions. In
designing fuel systems, much of

the analysis is done with 1D codes

for reasons of computational
economy but sometimes the
passages and fluid interactions are
so complex that only a 3D
treatment is felt appropriate.
Hitherto results have been passed
manually from 1D to 3D analyses.
There is a desire for such couplings
to be automatic but without
compromising the integrity of the
analysis.
Representing a vendors
perspective, Domonik Sholz from
ANSYS Germany called for
participating codes to develop a

common infrastructure so that they

could support a wide range of
multi-physics applications. Using
the example of tracer transport in a
pipe network, he showed how cosimulation between ANSYS CFX and
LMS AMESim gave excellent
agreement with experiment, for
flows in- and around- pipe
junctions. The inter-code coupling
was partially enabled by ANSYS
CFD codes (CFX and Fluent)
providing direct links to several 1D
Codes (including AMESim,
Flowmaster and GTPower). Further
examples included:
a vehicle thermal
management model
simultaneously running
Fluent, GTPower and
Flowmaster which giving
temperature results to within
2% of experimental
observations, and
an exhaust gas recycle (EGR)
featuring CFX and GTPower.
LMS Internationals R&D Manager
Roberto dIppolito then
demonstrated an exciting
application of 1D-3D coupling:
optimisation. 3D CFD on its own is
currently too computationally
intensive to be used in conjunction
with optimization analyses for large
industrial systems. 1D codes can
be used to approximate the
essential features of 3D CFD
predictions so that meaningful
optimization analyses can be
performed in conjunction with CFD
analyses. Using the example of a
water jacket for a 5-cylinder in-line
turbo-diesel, dIppolito
demonstrated a practical
methodology to optimise the design
of the cooling holes of the head
gasket. This is a multi objective
optimisation problem with a need to
maximize the minimum velocity
through the holes and to minimize
the related pressure losses
between the cylinder head and
crank-case in the context of a
complicated flow topology. Even
with 1D analyses simplifying the
fluid dynamic calculations, about
250 CPU-days of CFD computations
where used to optimise the
configuration.

Large aircraft system co-simulation

Picking up on some of the concepts

raised by ANSYSs Sholz, Sreenadh
Jonnavithula from CD-adapco
discussed the motivations for
coupling 1D and 3D CFD drawing
on experience gained within CDadapco. (In fact, these struck a
chord with most participants in the
meeting.) He showed how
couplings to SPT Groups
multiphase flow code OLGA,
Gamma Technologies GT-Power
and Ricardos WAVE have been
implemented in CD-adapcos
newest CFD code, STAR-CCM+.
Jonnavithula used automotive and
oil industry case studies to
demonstrate the generic coupling
capabilities of STAR-CCM+ together
with specific interfaces to 3rd party
products, including:
a coupling to OLGA to facilitate
the design of an oil company
slug-catcher (to capture a
large plug or slug of liquid that
might be projected from a
multiphase oil pipeline), and

a coupling with GT-POWER to

facilitate the design of auto
engine intake and exhaust
systems, with GT-POWER
modelling exhaust pipes and
ducts, whilst STAR-CCM+
simulated detailed flows within
the manifolds.
As a complete contrast to the
bespoke couplings offered by
ANSYS and CD-Adapco, Pascal
Bayrasy of the Fraunhofer Institute
for Algorithms and Scientific
Computing (Fraunhofer SCAI)
presented the neutral coupling
interface server, MpCCI( Meshbased parallel Code Coupling
Interface). MpCCI was originally
developed as a multi-physics
coupling application. It facilitates
coupling and data exchange
between, for example, a finite
element (FE) stress analysis code
and a CFD flow analysis codes for
Fluid Structure Interaction (FSI)
calculations and has recently been
enhanced to allow 1D-3D couplings.

55

Full vehicle thermal management

MpCCI addressed some of the

challenges inherent in cosimulation - complex hardware
environments and challenging
software engineering requirements
- by using adapters (developed for
each software vendor) to establish a
direct connection between the
MpCCI Coupling Server and the 1D
or 3D CFD code. Currently coupling
adaptors exist for Abaqus, ANSYS,
Fine/HEXA, Fine/TURBO,
Flowmaster, Fluent, Flux, ICEPAK,
MSC.Marc, Permas, STAR-CD and
RadTherm amongst others. In
principle, MpCCI offers the
potential of even more complex
couplings than bi-lateral ones
between 1D and 3D CFD codes.
Nevertheless Bayrasy
demonstrated the attention to detail
that has been necessary to ensure
that MpCCI produces stable,
convergent, conservative and
consistent co-simulation solutions.
Flowmasters David Kelsall then
illustrated how a 1D code might be
coupled to a 3D code (Fluent, STARCD or STAR-CCM+) using MpCCI as
a coupling adaptor. Using the
example of an aircraft
environmental control system (ECS)
to manage passenger cabin
climate, Flowmaster was used to
model the equipment and ducting
within the ECS supply, whilst 3D
CFD codes were used to model a
partial section of the cabin (to
minimize CFD run-times). MpCCI
was used as a coupling adapter.
The overall model allowed various

56

what-if scenarios to be tested.

Changes within the ECS supply
network were shown to have a
demonstrable effect on passenger
comfort within the aircraft cabin.
The example showed that realistic
simulations are possible and
provided further scope for
development and optimization. The
presentation discussed some of the
challenges overcome in coupling 1D
and 3D models and demonstrates
that a methodical approach
promotes convergence. With the
MpCCI coupling adaptor it was a
relatively straightforward exercise
to swap the CFD codes between
STAR-CD, Fluent and STAR-CCM+
The final session of the day was
dedicated to different aspects of the
1D-3D coupling challenges.
Francesca Iudicello from the ESDU
Fluid Mechanics Group reminded
the meeting of the importance of
using fully validated data and
correlations, particularly when 3D
calculations are approximated as
1D processes. ESDU has a rich
history in developing methods for
the design of internal flow systems
for over 40 years, using validated
experimental data and 1D analytical
methods. Their methods now
include the use of 3D CFD
predictions to supplement and
support experimental data.
Iudicello emphasized the
importance of understanding:

the type of averaging to use for

the flow parameters at the
inlet and outlet boundaries,
and
the sensitivity of the CFD
solutions to the location,
magnitude, profile and
turbulence entity of the
boundary conditions
Much of ESDUs experience is now
captured in CFD Best Practice
Guidelines for modelling pressure
loss and flow characteristics.
The final talk of the day came from
David Burt of MMI Engineering. He
showed a multiply coupled problem
featuring buoyancy driven flow in a
complex ventilation system. It
related to a nuclear facility where
no contaminants could be allowed
to escape from a process building.
The modelling involved coupling a
3D CFD model (for the building
space), a 1D model (for the
ventilation system) and MATLAB to
define some of the key components
within the overall model. Much of
the coupling was achieved
manually, and whilst this gave
acceptable results it limited the test
scenarios, use cases and failure
cases that could be assessed. Burt
felt that an automatic coupling
capability (between the computer
applications) would have led to an
improved understanding of the
influence of each model on any of
the others.

The links with 1D software are fully integrated in the STAR-CCM+ user interface

The presentations of the day clearly

demonstrated that there is a
significant interest in the coupling
of 1D-3D CFD.

the specific combination of 1D-3D

applications that specific vendors
already support otherwise the
development costs may be
significant if a new coupling
adaptor needs to be developed.

The type of organisation

undertaking coupled solution would
seem to be capital intensive
industries (such as automotive,
aerospace, and oil) where
significant gains may accrue from
improved understanding of system
interactions.

Fraunhofer-SCAI are pursuing a

different strategy. They provide a
neutral interface for simulation
code coupling and already provide
coupling adaptors to a wide range
of FE, 3D and 1D CFD and other
simulation tools.

Concluding Remarks

Developers and vendors are clearly

responding to customer needs
because many 3D CFD developers
(e,g, ANSYS, CD-adapco) are
developing bespoke coupling
solutions for their own products,
linked to specific 3rd party
applications. However many users
will be lucky if they happen to have

During the day and in the questions

time after the presentations there
were a number of lively
discussions, with some useful
insight into the different
perspectives of the vendors and
users in a range of different
industries.

There are clearly many issues still

to be addressed before coupling
and co-simulation become
universally stages of the analysis
process. But the current state of
the art (and the competing offerings
from developers and suppliers)
would seem to suggest that this
technology will develop and improve
over the coming years. It is an area
that NAFEMS will continue to
monitor and make information
available to members.
Thanks are recorded to members of
the NAFEMS CFD working group
who organised this event and
especially to Jo Davenport (of
NAFEMS) for organizing the venue
and ensuring the day ran so
smoothly and David Kelsall as
technical champion.

57