Automated Iteration with Finite Element Analysis
(HyperFEA)
1
SUMMARY APPROACH .................................................................................................................................2
1.1
1.2
2
OVERVIEW .......................................................................................................................................................2
ADDITIONAL CAPABILITIES ..............................................................................................................................4
HYPERFEA PROCESS ....................................................................................................................................7
2.1
2.2
FEM FILENAME CONVENTION ..........................................................................................................................8
SKIPPING LINES IN THE FINITE ELEMENT MODEL FILE ....................................................................................9
3
QUICK REFERENCE..................................................................................................................................... 10
4
HYPERFEA TUTORIAL ............................................................................................................................... 11
4.1 GETTING STARTED ......................................................................................................................................... 11
4.2 TUTORIAL EXAMPLE ...................................................................................................................................... 12
4.2.1
Pre-Processing the FEM ..................................................................................................................... 13
4.2.2
Submitting HyperFEA Iterations ......................................................................................................... 15
4.2.3
Re-Starting from Iteration 0................................................................................................................. 19
5
ADVANCED TOPICS ..................................................................................................................................... 20
5.1
5.2
6
SETTING UP BUCKLING FINITE ELEMENT DECKS FOR HYPERFEA ................................................................. 20
ACTIVATING DEFLECTION AND EIGENVALUE LIMITS IN HYPERFEA ............................................................. 23
REFERENCES ................................................................................................................................................. 26
APPENDIX A: LISTING OF THE BATCH FILE: NASTRAN_SUBMIT.BAT ................................................ 27
© 2008 Collier Research Corporation.
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
1
1
Summary Approach
1.1 Overview
HyperFEA automates the process of iterating between a Finite Element Analysis (FEA) and
HyperSizer to ensure that sizing loads used by HyperSizer are consistent and converged with
load-paths in the FEM. HyperFEA was created entirely using the HyperSizer Object Model,
originally developed under NASA contract for the ELVIS (Environment for Launch Vehicle
Synthesis) development program1,2.
Fig. 1 illustrates the HyperFEA’s coupling process between HyperSizer on the right and the
global FEA on the left. The process begins with development of a coarse finite element model to
represent the structural geometry. Dummy properties (e.g. min gage sheet) can be used in the
initial FEM just for check out to make sure there are no singularities. This FEM is imported into
HyperSizer which generates an initial set of properties and materials in the form of structural
element stiffnesses and thermal coefficients. These quantities are passed back to the FEA (in
updated PSHELL *, PCOMP, PBAR, MAT8, MAT2 and MAT1 cards) along with the applied
external loads, such as pressures and accelerations. After these updated properties are processed
by the FEA and the analysis has been run, the loop is completed when updated element forces
(read from the FEA results file, *.F06, as Nx, Ny, Nxy, Mx, My, Mxy, Qx, Qy) representing new
load paths are passed back to the sizing procedure for a new sizing. This iterative process
continues until the total weight of the structure converges.
HyperSizer
Fig. 1, The iteration process between HyperSizer (on the right) and the global FEA (on the
left) is automated by the HyperFEA program.
*
The FEM quantities described here apply only to the NASTRAN family of FEA programs.
© 2008 Collier Research Corporation.
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
2
Figs. 2 and 3 show typical convergence plots generated by the HyperFEA program. The weights
are broken into total panel, total beam, and total structure weights as depicted by the green, red,
and blue lines respectively in Fig. 2. In most cases, if the sizing bounds are well defined in the
HyperSizer model, the iterations converge after 4-5 iterations.
Fig 2. A typical iteration plot from HyperFEA. The plots show into panel, beam and
total weights. In most cases, the iterations are 95% converged after 4-5 iterations.
An interesting note about the iteration process is that it does not matter whether the initial
structural configuration starts with very light structure, such as minimum gage, or with very
heavy structure. In most cases, the iterations converge to the same value. In Fig. 3, the line with
magenta squares represents an initial structural weight of just over 1200 lb and after ten
iterations converges to a weight of around 700 lb. This same geometry and load configuration
was then started with a much lighter initial design, shown with the dark blue diamonds, of
around 550 lb, but again after 10 iterations has converged to a final weight of around 700 lb.
This behavior is beneficial because it means that the guessing the starting condition is not critical
to getting the final minimized weight.
This behavior is beneficial also because it means the iterative convergence is not history
dependent, therefore at any iteration in the progressive design process, the user can change the
design requirements from that given state, and move forward and get the correct answer usually
within two more iterations to dampen out the step change. A new additional requirement such as
a wing tip deflection or twist, or adding a higher fidelity analysis such as non-linear beam
column or bonded joint strength, can be inserted into, or removed from the progressive design
process at any iteration to quantify their weight impact.
© 2008 Collier Research Corporation.
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
3
Fig 3, Convergence plots from a HyperSizer-FEA iteration for a structure that was iterated
first from a very heavy starting configuration (magenta squares) and then from a much
lighter configuration (blue diamonds). In both cases, after 10 iterations the total structural
weight converges to the same final weight.
1.2 Additional Capabilities
While iteration with FEA was the original purpose, it was discovered during development that
HyperFEA has the potential to do more than just automation by introducing global failure and
optimization criteria that were previously unavailable to HyperSizer. Fig. 4 shows a recently
optimized commercial airplane wingbox structure, where HyperFEA was used for two purposes:
a) to change the pressures on the wing as a function of the wing twist; and b) to optimize the
wingbox while limiting the overall deflection and twisting angle to some prescribed limit.
The first purpose addresses the coupled structural aerodynamic response that as a wing twists
due to structural loading, the pressure on the wings is alleviated, and the center of pressure shifts
inboard toward the untwisted wing root. Because this load shift has the effect of lowering the
overall load (thereby reducing twist), the twist-load relationship is self-equilibrating and we
expect the weight to converge. This is accomplished by extracting the displacements at every
iteration at the wing tip control nodes, 2066 and 3459 (these control nodes are configurable by
the user), and then modifying the pressure loads on the FEM before each finite element analysis.
The modification to the FEM loads was performed in an external procedure, a simple fortran
program that uses the geometric relation shown in Fig. 3 to modify the PLOAD cards in the
FEM. This external procedure is either called or not called by HyperFEA depending on a
checkbox in the HyperFEA interface. This is another case where the process can easily be
modified by an HyperFEA user, by simply writing custom code to be called by HyperFEA.
© 2008 Collier Research Corporation.
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
4
Node
2066
Node
3459
b
a
Deformed
θ
wb
wa
Undeformed
δa-b
Fig. 4, Wingtip deflection and twist were determined by extracting the displacement at
two user-configurable “control-nodes”. The twist, θ, was used to modify the pressure
distribution in the finite element model through an externally called process.
To control the wing-tip twist and deflection in the structure, two “assemblies” were set up in
HyperSizer where the stiffnesses were modified during the iteration procedure.
Structural components identified for
controlling wing twist with required A33
(in-plane shear) stiffness
Structural components identified for
controlling wing tip deflection with
required A11 (membrane) stiffness for
panels, and EA stiffness for beams
Fig. 5 Two assemblies set up in the HyperSizer wing box project were used to control
wing-tip deflection (right assembly) and wing-tip twist (left assembly)
© 2008 Collier Research Corporation.
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
5
The left assembly in Fig. 5 was used to limit the wingtip twist, and the right assembly was used
to control the wingtip deflection. At each iteration, the deflection of the sized structure is
compared to the target limit. If one or both of these limits are violated by a certain percentage,
the stiffnesses of the components in these assemblies were used as constraint criteria on the next
iteration. For example, if the tip deflection limit is 13 inches, and the actual deflection is 18
inches, then the limit was violated by 38%. Therefore, HyperFEA will modify the stiffnesses in
the right most assembly and multiply them all by 1.38 and use these stiffnesses as constraints in
the next sizing iteration. If the next sizing overshoots the requirement, then the target is scaled
by this updated amount again. This strategy proved to be very effective as shown in Fig. 6.
Fig. 6, The iteration plot for the wing-box is broken into three phases. Iterations 1-8 are performed
without considering the effects of twist or deflection. Starting at Iteration 9, the external process of
modifying pressure due to wing-tip twist is activated. Predictably, this results in a net reduction in
load and a corresponding 4% reduction in weight. Starting on Iteration 17, a deflection limit of 13
inches is imposed. While this imposes a 38% decrease in the current deflection result of 18", the
overall weight is increased only 8% to meet this criteria.
© 2008 Collier Research Corporation.
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
6
2 HyperFEA Process
The HyperFEA Process is broken down into three stages with several optional operations. The
three principal stages are:
1. Pre-processing the FEM
2. Kicking off the process with an “unsized” template
3. Iteration with FEA
These operations are shown schematically in Fig. 7. The optional operations involve the calling
of an external process for FEM modification and/or changing the component stiffnesses as
described in Section 1.2 are not included in this flowchart but are described later.
FEA.dat
FEA.F06
• Create HyperSizer project
• Import initial verified FEM
• Assign all components to
desired families and apply
boilerplate sizing template
Before starting
HyperFEA
FEA_i00.dat
Split FEM into include files
FEA_i00.PM1
HyperFEA
FEM
P
i
Create new project for iteration
<project>_HyperFEA
FEA_ik+1.PM1
HyperSizer: Size entire project
with latest FEA loads
Iterations
Complete?
Yes
k=k+1
Done
No
Run FEA with updated
properties and materials
FEA_ik+1.F06
Fig. 7, Flowchart describing the HyperFEA process. After the initial setup of a HyperSizer
project, the HyperFEA process is broken into three stages: FEM Pre-processing; Kicking
off the iterations with an “unsized” template; and iteration with FEA.
© 2008 Collier Research Corporation.
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
7
2.1 FEM Filename convention
The first stage performed in HyperFEA “pre-processes” the FEM model files by removing all
elements, loads, temperatures, properties, materials, etc. into separate files, and replacing them in
the original FEM file with “INCLUDE” statements. In the process, a file naming convention is
adopted for tracking the iterations. This convention is to append “_i00”, “_i01”, “_i02”, etc. to
the FEM and FEA output file names to represent iterations 0, 1, 2, … respectively. This process
is illustrated in Fig. 8.
The original project will be unmodified
Folder Listing Original
Folder Listing after HyperFEA
Pre-processing
A newly created project will iterate with FEA
BEGIN BULK
$
PARAM,POST,-1
PARAM,OGEOM,NO
PARAM,AUTOSPC,YES
PARAM,GRDPNT,0
$
INCLUDE 'Ap1.PLOAD'
INCLUDE 'Ap1.FORCE'
INCLUDE 'Ap1.TEMP'
INCLUDE 'Ap1.SPC'
INCLUDE 'Ap1.GRID'
INCLUDE 'Ap1_i00.CP1'
INCLUDE 'Ap1_i00.PM1'
INCLUDE 'Ap1_i00.CL1'
$
ENDDATA
Fig. 8, The FEM pre-process in HyperFEA splits the original bulk data file into a
series of include files and copies and re-names the original bdf file to
<filename>_i00.dat to indicate that the bulk data file corresponds to the “zeroth”
iteration, or before any HyperSizer-FEA iterations have been performed.
The most important of the include files are the properties and materials file, or Ap1_i00.PM1, \
the CBAR file, Ap1_i00.CL1, and the Plate Element File, Ap1_i00.CP1. These files are
created by HyperSizer to be included in the FEM to generate loads for the next iteration. The
other files, such as Ap1.PLOAD, Ap1.FORCE, etc. are not modified during the iteration process.
The pre-processing step also creates a brand new project (in the same database) which uses the
split out FEM when starting the HyperSizer-FEM iteration process. The original user created
HyperSizer project and FEM files are never modified by HyperFEA.
© 2008 Collier Research Corporation.
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
8
The iteration numbers appended to the FEM filenames are used for files that feed into
HyperSizer for the current iteration. Therefore, the files that feed into HyperSizer for a
particular iteration have that iteration number. Files that are generated by HyperSizer have the
iteration number plus one. These files are then fed to the FEA for the k+1 iteration. The
HyperSizer setup form as it appears for each iteration is shown in Fig. 9 with screenshots for
iterations zero and one.
Iteration 0
Iteration 1
Iteration k
k+1
fea_i00.pm1
fea_i00.dat
FEA
HyperSizer
fea_i00.f06
fea_i01.pm1
fea_i01.dat
FEA
fea_i01.f06
HyperSizer
fea_ikk.pm1
fea_ikk.dat
FEA
HyperSizer
fea_ik+1.pm1
fea_ikk.f06
Fig. 9, The filename convention used by HyperFEA to indicate the iteration number
for a particular bulk data file, Filenames in orange boxes are input to HyperSizer.
Filenames in green boxes are output from HyperSizer.
2.2 Skipping Lines in the Finite Element Model File
Portions of the bulk data file can be skipped from the file split operation by enclosing the lines to
be skipped between the tags <HyperFEASkip> and </HyperFEASkip>. The first tag
(<HyperFEASkip>) tells HyperFEA to start skipping lines when processing the file, and the
second tag (</HyperFEASkip>) tells HyperFEA to stop skipping lines.
<HyperFEASkip> blocks can be used in the case control and/or bulk data sections, but cannot
span both of these sections - i.e. the Begin Bulk statement cannot be within the HypreFEASkip
block.
An example <HyperFEASkip> block is shown here:
$<HyperFEASkip>
{FEM Lines that are excluded from the HyperFEA file split}
$</HyperFEASkip>
© 2008 Collier Research Corporation.
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
9
3 Quick Reference
This section outlines the steps required to setup and run HyperFEA. These steps are described in
more detail in the HyperFEA Tutorial (Section 4).
Operations required before running HyperFEA for the first time:
1. Open the batch file, nastran_submit.bat (in the HyperFEA folder) and edit it to point at
the correct path and filename of the FEA executable, Nastran.exe (Section 0)
2. If desired, create a shortcut to the HyperFEA process on your desktop. (Section 0)
Operations performed before starting HyperFEA for a new project:
1. Create a new HyperSizer project
2. Assign FEM, PM1 and FEA Results filenames to the newly created project and import
the FEM. The paths and fileroots of each of these filenames must be identical.
(Section 4.2)
3. Create HyperSizer groups, assign each component in the project to a group. Groups must
be valid – that is, they must have materials assigned and each group must have at least
one candidate design.
Operations in HyperFEA:
1. Start HyperFEA and open a HyperSizer database (Section 0)
2. Select the new project from the Project dropdown list.
3. Press the Pre-Process FEM button. This process splits the original FEM file up into a
new iterative FEM file and a series of includes. It also creates a new HyperSizer project
which will be used for the iteration process (Section 4.2.1)
4. Select the number of iterations and start the iteration process (Section 4.2.2).
© 2008 Collier Research Corporation.
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
10
4 HyperFEA Tutorial
4.1 Getting Started
Setting up the Finite Element Analysis Program
Because HyperFEA does not know the location of the
Nastran process on each computer, this information must be
setup in a DOS-style batch file called nastran_submit.bat.
This file is located in the same folder as the HyperFEA
executable, which is by default:
C:\Program Files\HyperSizer\HyperFEA
Important: Before running
HyperFEA, the batch file called,
nastran_submit.bat must be set
up to point at the Nastran
Executable on your computer.
The nastran_submit.bat file can either be edited or replaced by one of the sample batch files
located in the same folder called, nastran_submit_MSC.bat, nastran_submit_NX.bat or
nastran_submit_NE.bat for MSC, NX or NEi Nastran respectively. The key setting that must be
checked in the batch file is the variable, NASTRANEXE. The value of this variable must be set
to contain the complete path and filename to the Nastran.exe executable on your computer. For
example, if running NEiNastran Version 9, the default location of this executable will be set by
changing the appropriate line in nastran_submit.bat to:
SET NASTRANEXE="C:\Program Files\NEiNastran Engine V9\Nastran.exe"
If this path is not set correctly, HyperFEA will produce an
error during the HyperSizer-Nastran iteration process. For a
full listing of the nastran_submit.bat file, see Appendix A.
For MSC or NX Nastran, set
this path to nastran.exe rather
than nastranw.exe.
Start the HyperFEA process
The HyperFEA executable file is copied by the Update program to the folder,
<HyperSizer Install>\HyperFEA, where <HyperSizer Install> is the HyperSizer installation
folder. If HyperSizer is installed to the default location, this folder will be:
C:\Program Files\HyperSizer\HyperFEA
To start HyperFEA, open Windows Explorer and browse to this folder, and double click this file
to start the HyperFEA process.
© 2008 Collier Research Corporation.
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
11
It may be helpful to create a shortcut on the desktop to the HyperFEA program. To do this, right
click on the “HyperFEA.exe” file icon and from the drop-down menu, select “Copy”. Next, go
to the desktop and right click in an empty area and select “Paste Shortcut” from the dropdown
menu. HyperFEA can now be started by double clicking this shortcut from the desktop.
4.2 Tutorial Example
An example HyperFEA problem is shipped with
the HyperFEA program with a database and
accompanying FEM files. The database can be
opened in HyperSizer to view the HyperFEA
example projects that are already set up.
Important: The HyperFEA application is
dependent on the version of HyperSizer
installed.
To get the latest version, go to
http://hypersizer.com/downloads.html
Start HyperSizer, sign-in and open the HyperSizer Database:
C:\HyperSizer Data\Projects\Collier\_HyperFEA\[Database]\HyperFEA Example 5.X.hdb.
(Note: If your HyperSizer Version is greater than the HyperFEA example database, create a new
database and import all projects from the HyperFEA database into this new database.)
This database contains two projects called “AP1 MSC” and “AP1 NE”. If using either
MSC/Nastran or NX/Nastran, open the Setup form for the project called, “AP1 MSC”. If using
NEiNastran, open the Project Setup form for the project called, “AP1 NE”.
The first thing to notice is that the FEA filenames each
have identical file paths and root filenames. This is not
required for normal HyperSizer operation, but is required
for HyperFEA operation. The filenames for these projects
are:
Important: HyperFEA requires
all FEM, FEA and PM1 filenames
to have identical paths and root
filenames. Only the file extensions
may be different.
MSC or NX Nastran Users:
HyperSizer Project: AP1 MSC
C:\HyperSizer Data\Projects\Collier\_HyperFEA\FEA\Ap1.DAT
C:\HyperSizer Data\Projects\Collier\_HyperFEA\FEA\Ap1.pm1
C:\HyperSizer Data\Projects\Collier\_HyperFEA\FEA\Ap1.F06
NEiNastran Users:
HyperSizer Project: AP1 NE
C:\HyperSizer Data\Projects\Collier\_HyperFEA\FEA\Ap1.NAS
C:\HyperSizer Data\Projects\Collier\_HyperFEA\FEA\Ap1.PM1
C:\HyperSizer Data\Projects\Collier\_HyperFEA\FEA\Ap1.ELS
Second, in order for a project to be used by HyperFEA, the project must already exist and all
components must be assigned to valid HyperSizer groups. A “valid” group is one that already
has materials assigned and shows at least one candidate design in the “Group Design Bounds and
Component Result” frame of the sizing form.
The remainder of this example will assume that the user is working with the MSC version of the
project, therefore the filenames will be “Ap1.DAT” and “Ap1.F06”. If working with
NEiNastran, substitute these filenames with “Ap1.NAS” and “Ap1.ELS” respectively.
© 2008 Collier Research Corporation.
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
12
4.2.1 Pre-Processing the FEM
1. At the beginning of the process, before any HyperFEA processing, the folder containing
the finite element model will contain only the Ap1.dat bulk data file and the Ap1.F06
results file.
2. Start the HyperFEA application (see Section 0 and select “Open Database” from the File
menu. Browse to the same HyperSizer database that contains your target project
(C:\HyperSizer Data\Projects\Collier\_HyperFEA\[Database]\HyperFEA Example
5.X.hdb)
3. HyperSizer will open the database and after a short delay will display the message,
“<Please select a project from this database>”. Use the dropdown to select your project.
© 2008 Collier Research Corporation.
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
13
4. HyperFEA will detect that this project is not yet set up for automatic iterations because
the FEM file names do not yet contain iteration numbers (e.g. _i00, _i01, etc.). At the
bottom of the HyperFEA form, the only button available should be
. Press
this button. After some processing, a text box will appear telling you the active project
has been reset. Press OK. HyperFEA is now ready to iterate.
When you press the Pre-Process FEM button, HyperFEA split up the original FEM file into
include files for elements (*.SHEL), grids (*.GRID), properties and materials (*.PM1), etc.
The bulk data file that references these include files is Ap1_i00.dat. The suffix “_i00.dat”
indicates that this is the “zero” iteration. In other words, no iterations between HyperSizer and
FEA have yet been performed.
HyperFEA has also created a new HyperSizer project called “Ap1 MSC_HyperFEA” and this
project will be the one on which HyperFEA will iterate. Since we have modified the HyperSizer
database outside of the HyperSizer application, we must refresh HyperSizer to notify it to read
the changes from the database.
5. Return to HyperSizer and on the Database Explorer Tree, press F4 (or select “Refresh
Tree” from the Options menu). This will update the tree so that you can see the newly
created project.
© 2008 Collier Research Corporation.
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
14
6. Open the Project Setup form for the new project. Note that this project is now set up
pointing to the “_i00” bulk data file and F06 file. The properties and materials file
(*.PM1) that will be created by HyperSizer when an analysis is run is pointing at
“Ap1_i01.PM1”. After analysis, this file will be used as input to the “_i01” finite
element analysis.
Note:
• The FEM Preprocess step will put all PSHELL, PCOMP, PBAR, PROD, MAT1, MAT2,
and MAT8 cards into the file *.pm1. In some circumstances, a material card (MAT1 or
MAT8) may be used by a finite element entity with which HyperSizer does not interface.
For example, suppose a PSOLID card which uses a MAT1 is included in the original rundeck. Because HyperSizer ignores all PSOLID property cards, it will not generate the
MAT1 card required by this PSOLID in the next iteration. Therefore, when the FEM is
analyzed, the PSOLID property in the main run-deck will cause an error because it will
not be able to find the required MAT1 material. If this happens, the user must manually
remove the required MAT1 from the *.PM1 file and copy it back into the main FEA rundeck.
4.2.2 Submitting HyperFEA Iterations
4.2.2.1 The First Iteration
7. After the pre-processing step, the project has been re-set to Ap1 MSC _HyperFEA (7a),
and the iteration number has been set to 0, indicating that no HyperSizer-FEA iterations
have been performed. To perform a single iteration, set the number of iterations to 1
(7b) and then press the Go button (7c). (The number of iterations is set to 1 for
instructional purposes. Usually the default value of 4 is used)
© 2008 Collier Research Corporation.
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
15
7a
7b
7c
8. Since no iterations have yet been performed, HyperFEA’s first operation is to copy the
Ap1 MSC_HyperFEA project to a project called “Ap1 MSC_HyperFEA_i00”. This new
project will be used to generate an initial PM1 file based on unsized templates for each
group. You will notice while this project is being analyzed that only a single candidate
design is being analyzed for each component.
8
© 2008 Collier Research Corporation.
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
16
9. The next operation is to run the finite element model using the first set of HyperSizer
generated properties and materials. The Nastran results (*.F06) file generated during this
run will be used as the basis for Iteration 1, and therefore will be called Ap1_i01.F06.
10. Before going on to Iteration 1, the Iteration 0 HyperSizer project is re-analyzed to obtain
the correct margins of safety based on updated FEA loads. This operation will not
change the properties or materials in the *.PM1 file, however the first time the project
was analyzed by HyperSizer (in Step 8), the FEA loads were from an unknown set of
properties.
After Step 10, HyperFEA is finished with the project called “Ap1 MSC_HyperFEA_i00”. This
project now contains weights and consistent margin of safety results for the unsized boilerplate
structure.
11. The final step is to perform the first actual “sizing” for
Iteration 1. In this step, the HyperSizer analysis is being
done based on the Ap1_i01.F06 file. After this sizing,
HyperFEA will show weight results from both Iterations 0
and 1.
© 2008 Collier Research Corporation.
Keep in mind that weight
results from Iteration 0
are from an “unsized”
structure and the weight
has no meaning compared
to the Iteration 1 “sized”
results except to show the
starting point.
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
17
After completing Iteration 1, the status of
the FEM files is shown here. Notice that
Iteration 0 and Iteration 1 are completely
finished and there are F06 FEA results for
each of these iterations. Iteration 2 has
not yet been completed. The properties
and materials file (Ap1_i02.pm1) has been
generated and therefore Iteration 2 is
ready to be analyzed by the FEA, however
an F06 file has not yet been generated.
Iteration 1 has been completed
and the finite element model has
been analyzed as shown by the
existence of the *_i01 F06 file.
The properties and materials
have been generated for
Iteration 2, however the FEM
has not been analyzed
4.2.2.2 Additional Iterations
12. At the end of the first iteration, the HyperFEA interface should appear as shown.
HyperFEA is ready for further iteration. The next step is to change to the desired number
of iterations (12a), in this case 4, and re-start the process by pressing the Go button (12b).
12a
© 2008 Collier Research Corporation.
12b
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
18
During the next four iterations, HyperSizer and the FEA analysis will be alternatively called and
the final state of the analysis is shown here. Notice that any time the HyperFEA process is started
the HyperSizer analysis is always run first, starting from the current iteration level, and it is also
run last. Therefore, every time the process is re-started, you will see a repeated iteration point as
shown at (13).
13
By re-running HyperSizer for the current iteration level at the beginning, this allows the user to
go into HyperSizer between HyperFEA runs and change the optimization bounds or any other
information in the project and immediately see the results of this change on the next iteration.
This process can be re-submitted any number of times by setting the number of iterations and
pressing the Go button.
4.2.3 Re-Starting from Iteration 0
The HyperFEA process can be re-started from any state and after any changes are made to the
HyperSizer model. In general, the iterations will re-converge after changes are made. However
in some cases you might wish to re-start the iteration process from scratch (i.e. go back to
Iteration 0). If this is the case, simply press the “Reset to Iteration 0” button on the HyperFEA
form. The following operations will be performed.
a) Re-set the FEA filenames in the <Project>_HyperFEA project to:
a. FEM Filename:
_i00.dat
b. FEA Results:
_i00.f06
c. PM1 Filename:
_i01.pm1
b) [Optional] Delete the <Project>_HyperFEA_i00 project.
c) [Optional] Compact the HyperSizer database to reduce the storage requirements and
increase the efficiency of the database.
If you choose to compact the database, make sure that this same database is not open in a
separate HyperSizer window, otherwise an error will occur. After resetting, the next time the Go
button is pressed, HyperSizer will start at Step 7. The FEM pre-processing step is not required
after this operation.
© 2008 Collier Research Corporation.
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
19
5 Advanced Topics
5.1 Setting up Buckling Finite Element Decks for HyperFEA
To perform global buckling checks as part of a HyperFEA sizing/analysis, a buckling eigenvalue
finite element deck should be created before the initial file split up. The assumption is that the
basic model (grids, elements, properties, etc) will be the same between the static analysis FEM
and the buckling analysis FEMs. The buckling analysis FEM must have the same name as the
static analysis FEM with the addition of “_BUCKLING” before the file extension. In the
example shown here, the name of the static analysis FEM is “Wing Box.NAS”. In order for
HyperFEA to see the buckling FEM, the name of the buckling analysis FEM must be,
“Wing Box_BUCKLING.NAS”.
Wing Box example files before HyperFEA File Split
Filenames that are framed by a
blue box are associated with
the static FEM. Those framed
by a green box are associated
with the buckling FEM..
The only difference between the two FEM files is in the case control where one FEM is set up
for static analysis and the other is set up for buckling analysis.
Case Control for “Wing Box.NAS”
SOL SESTATICS
CEND
FORCE = ALL
DISPLACEMENT(PLOT) = ALL
SUBCASE 1
SUBTITLE = Pressure
SPC = 1
LOAD = 3 BEGIN BULK
BEGIN BULK
Case Control for “Wing Box_BUCKLING.NAS”
SOL SEBUCKLING
CEND
FORCE = ALL
DISPLACEMENT(PLOT) = ALL
SUBCASE 1
SUBTITLE = Pressure
SPC = 1
LOAD = 3
SUBCASE 2
SUBTITLE = BUCKLING CASE
METHOD = 1
SPC = 1
BEGIN BULK
EIGRL
1
0.0
© 2008 Collier Research Corporation.
5
0
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
20
The project is set up in HyperFEA as normal. When the HyperFEA FEM split takes place, the
static FEM file is split just as with any other HyperFEA run, however, in addition to the static
FEM, the buckling FEM file is also split into a series of corresponding include files.
Wing Box example files after HyperFEA File Split
Notice that no *.CL1 (CBar Elements), *.CP1 (Plate Elements) or *.PM1 (Property and
Materials) files were created for the buckling FEM during the file split. These files are created
by HyperSizer during a sizing/analysis and will be shared by both the static and buckling finite
element analyses. A look inside the FEAs shows how this is done:
Bulk data for static FEM file, “Wing Box_i00.NAS”
BEGIN BULK
PARAM,OGEOM,NO
PARAM,AUTOSPC,YES
PARAM,GRDPNT,0
INCLUDE 'Wing Box.CORD'
INCLUDE 'Wing Box.FORCE'
INCLUDE 'Wing Box.PLOAD'
INCLUDE 'Wing Box.SPC'
INCLUDE 'Wing Box.GRID'
INCLUDE 'Wing Box_i00.PM1'
INCLUDE 'Wing Box_i00.CP1'
INCLUDE 'Wing Box_i00.CL1'
ENDDATA
Bulk data for buckling FEM File, “Wing Box_BUCKLING_i00.NAS”
BEGIN BULK
EIGRL
1
$
PARAM,OGEOM,NO
PARAM,AUTOSPC,YES
© 2008 Collier Research Corporation.
0.0
5
0
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
21
PARAM,GRDPNT,0
INCLUDE 'Wing Box_BUCKLING.CORD'
INCLUDE 'Wing Box_BUCKLING.FORCE'
INCLUDE 'Wing Box_BUCKLING.PLOAD'
INCLUDE 'Wing Box_BUCKLING.SPC'
INCLUDE 'Wing Box_BUCKLING.GRID'
INCLUDE 'Wing Box_i00.PM1'
INCLUDE 'Wing Box_i00.CP1'
INCLUDE 'Wing Box_i00.CL1'
ENDDATA
As the iterations proceed, static FEMs and buckling FEMs will be created for each iteration
number, but they will always point at the same *.PM1, *.CP1 and *.CL1 files. For example, at
the third iteration, the following files will be used in the finite element analysis:
Wing Box_i03.NAS
Wing Box_Buckling_i03.NAS
These files will have two different case controls, but will each point to the same elements,
materials and properties created by HyperSizer for the third iteration:
INCLUDE 'Wing Box_i03.PM1'
INCLUDE 'Wing Box_i03.CP1'
INCLUDE 'Wing Box_i03.CL1'
© 2008 Collier Research Corporation.
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
22
5.2 Activating Deflection and Eigenvalue Limits in HyperFEA
Sizing the structure for global deflection and/or buckling eigenvalue control is activated on the
Options dialog which is raised from the Tools menu of the HyperFEA main form.
The dialog has several options for controlling the HyperFEA process. Two of these options will
allow HyperFEA to modify stiffnesses to attempt to control global buckling and deflection
characteristics. To activate this capability for global deflection and/or twist limits press the
“Modify Stiffnesses for Deflection and Twist Limits” checkbox (A). To activate this process for
eigenvalue limits, check the “Run global eigenvalue analysis every n iterations” checkbox (B).
By default, eigenvalue analyses are performed for every iteration, you can also change the
software so that eigenvalue analyses are performed at different intervals by entering the number
of iterations(C). E.g. entering a “2” will cause the eigenvalue analysis to run every other
iteration, etc.
A
B
C
D
In addition to turning on these capabilities, you must specify the deflection and eigenvalue limits
in an ASCII file with the same name as the HyperSizer project and the extension *.UIN. This
file can be opened by pressing the “Deflection and eigenvalue limits” button (D). If this file
does not exist, pressing this button will create a template of this file for you. For the wing box
example, the *.UNI file looks like this:
© 2008 Collier Research Corporation.
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
23
Comments can be included in the UIN file by including a pound sign (#) in column 1. Each line
of the UIN file contains a constraint that will be used by HyperFEA, along with the FEA results,
to control the sizing. In the example file, there are three constraints: 1) constraint on the global
buckling eigenvalue; 2) a constraint on the deflection at a grid point; 3) constraint on twisting
deformation between two grid points. The fourth line starting with “-1” is a flag indicating that
there are no more constraints.
Each of these constraints are discussed below.
Criteria #1: Global Buckling Eigenvalue Constraint
0,
0,
0.0,
Grid 1:
Grid 2:
Deflection Limit:
Twist Limit:
Eigenvalue Limit:
ABD Controlling:
Assembly:
0.0,
1.0,
456, "ASSEMBLY: Global Buckling Example"
0
0
0.0
0.0
1.0
456 (D11, D22, D33)
"ASSEMBLY: Global Buckling Example"
Since this is a global buckling criteria, no grids are specified for deflection limits, therefore grids
1 and 2 along with the deflection and twist limits are all specified as zero. The eigenvalue limit
is 1.0 meaning that an eigenvalue from the buckling FEA of less than 1.0 indicates failure.
The ABD terms that will be modified to attempt control of local buckling are specified as 4, 5,
and 6, which correspond to D11, D22, and D33. These stiffness terms will be entered into
HyperSizer as stiffness constraints for all components that are members of “ASSEMBLY:
Global Buckling Example”.
If an eigenvalue of less than the limit is encountered, the stiffness constraints will be modified as
follows:
ABDconstraint= ABDprevious * (Eigenvalue Limit) / (Eigenvalue)
For example, if the eigenvalue limit is 1.0 and the actual encountered eigenvalue from the
buckling solution is 0.85, then the stiffness constraints will be entered into HyperSizer as:
D11,constraint
D22,constraint
D33,constraint
= D11,previous * 1.0 / 0.85
= D11,previous * 1.176
= D22,previous * 1.176
= D33,previous * 1.176
In this case, D11,constraint is a stiffness constraint that will be applied during the next sizing
iteration and D11,previous is the D11 term resulting from the previous HyperSizer sizing iteration as
reported on the Computed Properties tab of the Sizing Form.
© 2008 Collier Research Corporation.
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
24
Criteria #2: Deflection Limit
1,
2, 10.0,
Grid 1:
Grid 2:
Deflection Limit:
Twist Limit:
Eigenvalue Limit:
ABD Controlling:
Assembly:
0.0,
0.0,
1, "ASSEMBLY: Grid Deflection Example"
1
2
10.0
0.0
0.0
1 (A11)
"ASSEMBLY: Grid Deflection Example"
In this case, a deflection limit of 10 inches has been imposed for grids 1 and 2. Whichever of
these grids has a higher absolute deflection value will be used to evaluate the criteria. The twist
and eigenvalue limits have been set to 0.0 indicating that these criteria are inactive. The ABD
controlling stiffness term has been specified as 1, corresponding to the A11 stiffness. If the
deflection limit is exceeded for either grids 1 or 2, the A11 stiffness constraints for all of the
components in, “ASSEMBLY: Grid Deflection Example,” will be modified according to:
A11,constraint = A11, previous * (Deflection Actual) / (Deflection Limit)
Criteria #3: Twist Limit
3,
4,
0.0,
Grid 1:
Grid 2:
Deflection Limit:
Twist Limit:
Eigenvalue Limit:
ABD Controlling:
Assembly:
3.0,
0.0,
3, "ASSEMBLY: Grid Twist Example"
3
4
0.0
3.0
0.0
3 (A33)
"ASSEMBLY: Grid Twist Example"
In this case, a twist limit of 3 inches has been as the limit of the deflected rotation between grids
3 and 4. The twist is calculated from:
{ Twist Formula Here }
The deflection and eigenvalue limits have been set to 0.0 indicating that these criteria are
inactive. The ABD controlling stiffness term has been specified as 3, corresponding to the A33
stiffness. If the twist limit is exceeded between these two grids, the A33 stiffness constraints for
all of the components in, “ASSEMBLY: Grid Twist Example,” will be modified according to:
A33,constraint = A33, previous * (Twist Actual) / (Twist Limit)
© 2008 Collier Research Corporation.
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
25
6 References
1. “Development of HyperSizer Application Programmer Interfaces (API’s) that specifically
support the High Performance Computing and Communications Program (HPCCP)
Environment for Launch Vehicle Integrated Synthesis (ELVIS) for Reusable Launch
Vehicles (RLV) Software Design System.” In support of Prime Contract No.
GS00T99ALD0209. POCs are Dr. John Korte, P.E., and Jeff Cerro, NASA Langley
Research Center, 2001.
2. Collier Research Corp., "Using the HyperSizer Object Model for Software Integration",
HyperSizer White Paper, http://hypersizer.com/dload/downloads.htm#Papers.
© 2008 Collier Research Corporation.
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
26
Appendix A: Listing of the Batch File: nastran_submit.bat
To analyze the finite element model, HyperFEA will attempt to run a batch file located in the
same folder as the HyperFEA executable. This batch file is called, nastran_submit.bat. This
MSDOS style batch file takes in a single argument, the fully qualified path and file name for the
finite element model. The main purpose of this batch file is to parse the FEM file path and name
and submit the FEA job. The sample shown here is for submitting NEiNastran V9.0. The
biggest change that will need to be made for your own FEA runs is to change the
NASTRANEXE path and filename to the location of your own NASTRAN executable.
nastran_submit.bat
@echo off
REM This batch file is set up to start a NASTRAN process given
REM a single argument, which is the complete FEM path and file name
REM the usage is:
REM nastran_submit “<FEM_PATH>\<FEM_FILE>”
REM this batch file will strip off any leading or trailing quotes (")
REM Echo the path and name of this batch file
REM -----------------------------------------------------ECHO BatchFile: %~f0
REM
REM
REM
SET
SET
SET
SET
SET
Parse input argument (%1) into separate variables for
Drive, Path, Filename, and extension
-----------------------------------------------------FEMDRIVE=%~d1
FEMPATH=%~dp1
FEMBASE=%~n1
FEMEXT=%~x1
FEMFILE=%~nx1
REM Set a variable for the nastran executable
REM -----------------------------------------------------SET NASTRANEXE="C:\Program Files\NEiNastran Engine V9\Nastran.exe"
REM Echo all variables to the screen
REM -----------------------------------------------------ECHO FEM Drive
: %FEMDRIVE%
ECHO FEM Path
: %FEMPATH%
ECHO FEM FileBase : %FEMBASE%
ECHO FEM Extension: %FEMEXT%
ECHO Full Filename: %FEMFILE%
ECHO NASTRAN
: %NASTRANEXE%
REM Change to the drive where the FEM resides
REM -----------------------------------------------------%FEMDRIVE%
REM Change the current directory to the folder where the FEM resides
REM -----------------------------------------------------CD %FEMPATH%
REM Run Nastran sending the FEM Filename as an argument
REM -----------------------------------------------------%NASTRANEXE% %FEMFILE%
© 2008 Collier Research Corporation.
HyperSizer Documentation. Method: Automated Iteration with FEA
Date: 2008-11-25
27