FEMCI – The Book

Table of Contents
FEMCI – The Book....................................................................................................................................... 1
Adhesive Modeling in Bonded Joints ....................................................................................................... 4
Coefficient of Thermal Expansion ............................................................................................................ 7
Craig Bampton Models ........................................................................................................................... 11
Background ......................................................................................................................................... 11
Craig-Bampton Theory ....................................................................................................................... 12
How to Create a C-B Model ............................................................................................................... 16
Load Transformation Matrices (LTMs)............................................................................................... 18
Checking C-B Models and LTMs ....................................................................................................... 21
Development of Fixed Base Shake Transfer Functions .......................................................................... 22
Eigenvalue Extraction Methods in NASTRAN ...................................................................................... 28
Honeycomb Plates .................................................................................................................................. 34
Properties ......................................................................................................................................... 34
PSHELL Card ..................................................................................................................................... 37
Isogrid Plates........................................................................................................................................... 39
Modal Weight Calculations ..................................................................................................................... 42
Method for Recovering Modal Mass Participation Output ..................................................................... 44
Preloads ................................................................................................................................................... 47
Random Vibration ................................................................................................................................... 50
Nastran Input Deck Example .............................................................................................................. 50
Specification Magnitude Equations .................................................................................................... 55
Calculating GRMS .............................................................................................................................. 58
Creating a Test Specification .............................................................................................................. 62
Miles Equation .................................................................................................................................... 64
Definition of Decibels ......................................................................................................................... 68
Sine Vibration ......................................................................................................................................... 70
Nastran Input Deck Example .............................................................................................................. 70
Double Amplitude ............................................................................................................................... 73
Semi-Rigid Connections ......................................................................................................................... 75
Unit Consistency ..................................................................................................................................... 81
Validity Checks ....................................................................................................................................... 82
Unit Enforced Displacement and Rotation ......................................................................................... 85
Free-Free Dynamics with a Stiffness Equilibrium Check ................................................................... 90
Unit Gravity Loading in All Three Translational Directions .............................................................. 93
 Unit Temperature Increase .................................................................................................................. 97
Yield Stress Notes ................................................................................................................................... 98
Adhesive Modeling in Bonded Joints

A practical finite element approach has been developed to model the adhesive in a bonded joint.
Numerical examples have shown good agreement with classical solutions.

The method uses a gap the thickness of the adhesive,

two rigid elements, and three zero- length spring
elements between coincident nodes. One rigid element
stretches from one adherend to a node at the center of
the gap, while the second rigid element stretches from
the other adherend to a coincident node also at the
center of the gap.

The spring elements connect the two rigid elements

between the coincident nodes. The figure shows the
layout. For clarity, the coincident nodes are shown
separated. No rotational springs are used in this
modeling technique.

Forces, stresses, strains, etc., can be recovered

directly for the adherend elements. Recovery of
spring forces and deformations can in turn be used to
determine the stresses and strains in the adhesive.

The various spring stiffness values are determined using the following equations. Note that there are
different values depending on whether the spring elements are internal to, along the edge of, or at a corner
of the finite element mesh.

Ael = Element area

G = Shear modulus of the adhesive
η = Adhesive thickness
E = Young's modulus of the adhesive
ν = Poisson's Ratio

Shear spring constant, ks, in both lateral directions,

Peel spring constant, kp, through the thickness,

where

where

where
Output results for the adhesive can be calculated using the following equations:

where
i = Grid number
x, y = Shear directions
z = Peel direction

= Spring deflection

= Peel strain

= Spring force

= Shear strain

= Peel stress

= Shear stress
Coefficient of Thermal Expansion

Dr. William Case

April 14, 1994

By definition, thermal strain at temperature T is the change in length of a member, due to a change in
temperature, ( T - Tref ), divided by the original length l of that member. Denoting thermal strain at
temperature T as eT

The thermal strain is zero at the reference temperature, Tref .

Also, by definition, the coefficient of thermal expansion (CTE or α) at temperature T is defined as

Therefore,

where t is a dummy variable of integration. Define α(T) to be the average value of alpha over the
temperature range from Tref to T.

Then,
Then, the thermal strain can be written as:

Generally, thermal expansion data is given as a curve of vs. T.

Based on equations (2) and (5), it is seen that:

Whereas

NASTRAN requires that be input, not . However, the two are equal when is constant.

An example of the calculation of , along with the NASTRAN material data input, is given below.
Thermal Expansion of Magnesium
(Delta L/L and Alpha data from NBS Monogram 29: Thermal Expansion of
Technical Solids at Low Temperatures)

Temp, deg 103 Delta 106 106 Alpha-

K L/L Alpha bar
0 -4.90 0.0 16.7
10 -4.90 0.1 17.3
20 -4.90 0.4 17.9
30 -4.89 1.4 18.6
40 -4.86 3.3 19.2
50 -4.83 5.7 19.9
60 -4.75 8.1 20.4
70 -4.66 10.3 20.9
80 -4.54 12.2 21.3
90 -4.41 13.9 21.7
100 -4.27 15.4 22.1
120 -3.93 17.6 22.7
140 -3.56 19.4 23.3
160 -3.16 21.0 23.8
180 -2.73 22.2 24.2
200 -2.27 23.2 24.4
220 -1.80 23.9 24.7
240 -1.32 24.4 24.9
260 -0.83 24.8 25.1
280 -0.33 25.2 25.3
300 0.18 25.5 25.4
NASTRAN Temperature Dependent Material Data Input for Magnesium
(using degrees Kelvin for temperature)

Dr. William Case

April 14, 1994
Craig Bampton Models

Scott Gordon
NASA Goddard Space Flight Center
May 6, 1999
(updated January 2008)

Background

Who is Craig Bampton?

Coupling of Substructures for Dynamic Analysis
by Roy R. Craig Jr. and Mervyn C. C. Bampton
AIAA Journal, Vol. 6, No. 7, July 1968

What is the Craig-Bampton Method?

• Method for reducing the size of a finite element model.
• Combines motion of boundary points with modes of the structure assuming the boundary points are
held fixed.
• Similar to other reduction schemes

Guyan Reduction

Modal Decoupling

Craig-Bampton
Why is the C-B Method Used?

• Allows problem size to be reduced.

• Accounts for both mass and stiffness (unlike Guyan reduction)
• Problem size defined by frequency range.
• Allows for different boundary conditions at interface (unlike modal decoupling)

Craig-Bampton Theory

Equation of Motion (ignoring damping)

The Craig-Bampton transform is defined as:

Where:

ub = boundary DOFs

uL = internal (leftover) DOFs

= rigid body vector

= fixed base mode shapes

 q = modal DOFs

The 2X2 matrix in Equation 2 is the Craig-Bampton transformation Matrix.

Combining equations (1) and (2) and pre-multiplying by

Define the C-B mass and stiffness matrices as

Write equation (3) using equations (4) and (5)

Where input forces are applied at the boundary only (F=0)

Important properties of the C-B mass and stiffness matrices:

• Mbb = Bounday mass matrix -- total mass properties translated to the boundary points

• Kbb = Interface stiffness matrix -- stiffness associated with displacing one boundary DOF
while other are held fixed.
o If the boundary point is a single gird (i.e. non-redundant) then Kbb = 0

• If the mode shapes have been mass normalized (typically they are) then

We can finally write the dynamic equation of motion (including damping) using the C-B transform as
Where,
= modal damping
= % critical damping
Summary of C-B Theory

• C-B Mass and Stiffness Matrices fully define system

Dynamics problem solved using C-B DOFs
• C-B boundary DOFs provide location to apply BC’s and Forces or to couple with another
structure
• C-B transform is used to calculate physical responses from C-B responses.
How to Create a C-B Model

$ C-B Output File

ASSIGN,OUTPUT4='cb_ex1.kmnp',STATUS=unknown,UNIT=31
ID GORDON,CB_EX1
$ Normal Modes Solution
SOL SEMODES
TIME 5
DIAG 8,14
$ C-B DMAP
INCLUDE 'path/cb103_2005r2b.dmp'
CEND
$
TITLE = Craig-Bampton Example 1
ECHO = NONE
METHOD = 1
$ SPC = 1
$ Request modal effective mass output
MEFFMASS(ALL) = YES
$
$ Request Output Grid Point Forces, Element Forces, and Element Stresses
$ This is recovered in the .F06 file
SET 100 = 209,210
SET 101 = 110,114
GPFORCE = 100
ELFORCE = 101
ELSTRESS = 101
$
BEGIN BULK
$
PARAM,AUTOSPC,YES
PARAM,GRDPNT,0
PARAM,WTMASS,0.00259
$
$ Print G-set (U1 in the DMAP) & R-set internal order
PARAM,USETSEL,0
PARAM,USETSTR1,U1
PARAM,USETSTR2,R
PARAM,USETPRT,0
$
$ Acceleration Output
$ Recover accleration response at GRID 11
$ This is output in the .kmnp file and processed in FLAME
USET1 U1 123456 11
$
$ Define number of modes to recover (or frequency range)
EIGRL 1 10 0 MASS
$
$ Boundary (R-set) Defined on SUPORT cards
$ 1 boundary point x 6 DOFs
$ = 6 physical boundary points
SUPORT 999 123456
SPC 1 999 123456 0.
$
$ Don't forget the rest of your bulk data
INCLUDE 'bulkdata.bdf'
$ ENDDATA
What is created?
• File cb_ex1.kmnp which contains C-B stiffness and mass matrices (k,m), net CG ltm (n), and the
C-B transformation matrix (phig)
• The .kmnp file is in NASTRAN binary output4 format
• K and M size is [C-B DOFs (boundary + modal) x C-B DOFS]
• phig size is [G-set rows x C-B DOFs]
• Net CG LTM recovers CG accelerations and I/F Forces, Size is [6+boundary DOFs x C-B
DOFs]

How do you use this?

• Solve dynamics problem for C-B DOF response using the K and M matrices
• Transform C-B responses using phig to get physical responses
Load Transformation Matrices (LTMs)

LTM is a generic term referring to the matrix used to transform from C-B DOFs to physical DOFs
(also referred to at OTMs, ATMs, DTMs …)

In its simplest form, the LTM is simply the phig matrix

(Only the rows corresponding to the physical DOFs of interest are needed)

There are other useful LTMs that can be created:

• Interface (I/F) forces

• Net CG accelerations
• Stress and force LTMs

I/F Force LTM (created by C-B DMAP)

IF Force =

(If boundary is non-redundant, then Kbb = 0)

Net CG LTM (created by C-B DMAP)
Net CG acceleration =

where

is the mass matrix about the CG (6 x 6)

and

is the rigid body transform from I/F to CG (BDOF x 6)

PHIZ LTM
Allows physical displacements to be calculated from C-B accelerations:

• Same as modal acceleration approach in NASTRAN

• Useful in calculating relative displacements between DOFs
• Also used to calculate stresses and forces which are a function of displacements
• Calculated from C-B dmap using PARAM,PHZOUT,1
LTM’s can be created using FLAME, MATLAB or using DMAP
LTM’s can (and usually do) contain multiple types of responses

LTM’s can be used to recover responses for nested C-B models

Where

Is the row partition of CB1 DOFs from the CB0 PHIG matrix.
Checking C-B Models and LTMs

C-B Models and LTMs should be verified to make sure that they have been created correctly (especially
for complicated LTM’s or nested C-B models)
C-B Mass and stiffness matrices can be checked by computing free-free and fixed-base modes C-B
boundary Mass matrix can be transformed to CG and compared with NASTRAN GPWG
LTMS can be checked by applying unit acceleration at the boundary:

Where

Is the boundary rigid body vector (b x 6)

• Each response column represents acceleration in a single direction
• Accelerations should be in correct directions
• Forces should recover weight or correct moments
• Unit acceleration applied to PHIZ can be checked by gravity run with physical model and
comparing displacements
Development of Fixed Base Shake Transfer Functions

Timothy Carnahan
Goddard Space Flight Center
January 23, 1996

The block diagram below is for a typical feedback servo control system in an instrument.

The plant, H(s), will generate some error signal of a sensor. In motor systems this is typically an error
between the rotor and the stator. The controller senses these errors and applies a correction load. For
rigid systems the external load would be applied to the mass of the system to generate a disturbance
acceleration. The more complex systems are flexible. This could mean that the flexibility is in a
rotational shaft or between the point of application of an external load and the instrument. For the CIRS
scan mechanism a flexible plant model was desired to couple spacecraft disturbances at the base of the
instrument with sensor errors between the reference interferometer and the scan mirror. Since the S/C is
supplying a disturbance without concern for the response of the instrument, the plant model assumes the
instrument is fixed to ground. Further, the S/C generates disturbance forces and moments at the base of
the instrument.

The following is the methodology and derivations developed to provide the controls analyst with H(s).
This is expressed as a transfer function of coefficients to a numerator/denominator pair.

The general equation of motion that governs an undamped systems without external forces is:

For a system with fixed base motions acceleration,

Equation (1) becomes:

If we define ur = u - uB and
Equation (2) becomes:

Expressing ur in model coordinates we get

Where is a fixed base mode shape for mode i and qi are the model coordinates.

Converting equation (3) to model coordinates and pre-multiplying by

For the modal mass, m =

The modal stiffness, k =

And including generalized modal damping of

We get:
Solving equation (6) for q(s) and using

Using equation (7) in equation (4) to solve for ur

Solving for u
For relative rotor/stator T.F. we want the difference between the two or, ustator - urotor = durotor-stator
which expands to

For absolute stator T.F. we want a complete solution to u. Remember that that equations of motion, (2),
was in terms of acceleration and therefore
Equations (11) and (12) needs the inputs:

m - the modal mass for mode i

- the mode i cantilevered shapes.

- the eigenvalue for mode i

- generalized modal damping

and AO.

For mass normalized eigenvectors m is unity and can be eliminated from the problem.

AO is defined as

The 'motion' program processes the output of a normal modes run, rigid format 3, and generate the H(s)
T.F. for Matlab. The DMAP for UAI NASTRAN version 11.8 to solve for AO and output to an
OUTPUT4 file is shown below.

ASSIGN AOout=aOout.ot4,NEW,USE=OUTPUT4,TYPE=FORMATTED
ALTER 188 $
MATGEN BGPDT,EQEXIN,CSTM/UB/102/0/ $ GENERATE Ub
MPYAD MGG,UB /FB/0/ -1/0/ $
MPYAD PHIG,FB, /AO/-l/l/O/ $
OUTPUT4 A0,,,,//*AOout*/0 /*FULL*/ $
ENDALTER $
The eigenvalues and eigenvectors are read from a punch file. In order for the post-processing program to
work the following guidelines must be observed:
1. Because m is not exported to a file the eigenvectors must be mass normalized. (Therefore, m is
the density matrix.)
2. Do not use DYNRED, as it changes the set definitions and the DMAPS MPYAD does not use the
correct size arrays.
3. Use DISP(PUNCH)=set id to output the grid points of interest. The punch files are much smaller
when only a set, the rotor/stator gid and the input gid, are in the punch file. The punch file includes the
eigenvalues.
The 'motion' program is written in "C" to read the NASTRAN data and convert it into MATLAB m-files
in the form of MATLAB numerator and denominator transfer functions. This program is located in
/home/tim/bin and was written for SGI machines. Motion has been ported to an IBM RS6000. The source
code is in the directory /home/tim/bin/SRC/NASTRAN. The data used by the "motion" program will
come from a punch file which contains the eigenvectors and eigenvalues and the OUTPUT4 file of the A0
matrix. With these the program asks for the grid points and dof for the rotor and the stator and the
direction of input disturbance. When the desired output is to know absolute position, stator only, the same
grid id is entered for rotorr and the slator. Motion will add the needed MATl,AB commands to generate
both the position and acceleration transfer functions for absolute displacement. The MATLAB variable
names of these absolute transfer functions is pos_num and pos_den for position transfer functions and
acc_num and acc_den for acceleration transfer functions. In the case of relative transfer functions only the
position transfer functions are generated and the MATLAB variable names are plant_num and plant_den.
The translational units of the transfer function are in whatever units the NASTRAN model was in. The
rotational units are in degrees.
The 'motion' program is coded with the following maximums: Number of eigenvalues = 230. Note:
matlab can not accurately handle such a large number of eigenvectors/T.F.'s. A method of mode selection
is needed to provide only significant modes for controls work. The generalized modal damping, ~, is set
to 0.1% of critical. This damping values depend on a number of mechanical characteristics from the
magnitude of the applied structure to the type of assembly. For typical micro- gravity disturbances on
small, < 100 lbs, a value of 0.1% is considered conservative.

Tim Carnahan
Eigenvalue Extraction Methods in NASTRAN

Jeff Bolognese
NASA Goddard Space Flight Center
November 29, 1995

Introduction

This document will discuss 3 of the commonly used eigenvalue extraction techniques used in
NASTRAN: Modified Givens, Inverse, Lanczos. In addition, 2 methods of automated model dynamic
reduction will also be investigated: AUTOOMIT(static condensation or Guyan Reduction) and
Generalized Dynamic Reduction.
Since it's beyond the scope of this report (and my scope!) to discuss the theories of the various eigenvalue
extraction techniques, this report will stick mainly to a general discription of the technique, the pluses and
minuses, and other considerations to be aware of. Following the detailed descriptions of these methods
will be a summary table.

Methods

This section will discuss three of the NASTRAN methods used to extract eigenvalues for a dynamics
analysis. These methods can be used alone, or in conjunction with dynamic reduction techniques.
Modified Givens: (MGIV) This method is, for most problem sizes, the slowest method of eigenvalue
extraction. MGIV will solve for ALL eigenvalues in the NASTRAN model (eigenvectors will only be
calculated for the range of frequencies specified). As a result, this method is most useful if all the
eigenvalues are of equal interest. For most problems, MGIV is only practical if the model is small, or if an
ASET or OMIT set is provided in the model in order to decrease the number of dynamic degrees of
freedom (DOF). When using MGIV, NASTRAN will, by default, use AUTOOMIT to remove all massless
degrees of freedom from the ASET DOF in order to compact the matrices.
As stated earlier, for practical purposes, MGIV requires that the model be reduced in size. This can be
done by either specifying which degrees of freedom to keep in the model (ASET), or which degrees to
remove from the model (OMIT). The problem with reducing the model is that it introduces error into the
results, particularly for higher modes, as not all the structural mass and stiffness is taken into
consideration. This can result in the shifting of mode frequencies as well as missing frequencies
completely. Deciding what degrees of freedom to keep is an art unto itself. UAI/NASTRAN has
automated the process somewhat with the use of AUTOOMIT.
Inverse: (SINV) This method is faster than MGIV for eigenvalue extraction, but still uses the entire mass
and stiffness matrices of the model. SINV is very efficient for extracting a specific range of frequencies.
Unlike MGIV, SINV only solves for the eigenvalues in the frequency range of interest. Also unlike
MGIV, SINV does not need to remove massless DOF's in order to run efficiently. In fact, SINV takes
advantage of the sparseness of the matrices caused by massless degrees of freedom. As a result, using
AUTOOMIT to remove massless degrees of freedom from an SINV run will actually Increase model run
time. Since SINV uses all the DOF's in the model and does not require ASET's for reasonable run times
the solution accuracy is very good and, for similar run times, is more accurate than an MGIV model run
with an ASET.
Lanczos: (LANCZ) The Lanczos method is, for most applications, the fastest, accurate solution method in
NASTRAN. LANCZ is faster than both MGIV and SINV and is just as accurate for the models tested.
LANCZ is also a sparse matrix solver and, like SINV, will run slower if the matrices are compacted by
removing massless DOF. This method of extraction seems the best choice for most NASTRAN models.
Even with compacted matrices, LANCZ can be faster than any of the other methods mentioned. However,
if you have large, dense matrices (i.e.. very few massless DOF), using the MGIV method along with
Generalized Dynamic Reduction may be the fastest method.

Dynamic Reduction Techniques

Dynamic reduction is a way of reducing the number of dynamic degrees of freedom in a finite element
model and therefore reducing the size of the problem to be solved. This is particularly important when
using MGIV as its solution time is dramatically decreased when the ASET can be reduced. One
consequence of dynamic reduction, however, is the introduction of error. Since dynamic reduction
removes degrees of freedom from the ASET it can cause mode frequencies to shift, and also may miss
modes altogether.
AUTOOMIT: AUTOOMIT is a tool for automatically performing dynamic reduction on a model.
AUTOOMIT is a case control command that creates an OSET (OMIT cards) automatically, given various
parameters. It will automatically remove all massless DOF when using the MGIV method. AUTOOMIT
can also be used to remove other DOF's from the ASET. AUTOOMIT eliminates DOF based on their
mass to stiffness ratio. A high ratio means a DOF with high mass and relatively low stiffness. In general,
this is more likely to be a DOF that will contribute to a lower frequency mode than a DOF that has low
mass and high stiffness. AUTOOMIT can be requested to omit DOF's based on a value for the mass to
stiffness ratio ("AUTOOMIT(EPS=x) =YES"), based on DOF mass ("AUTOOMIT(MASS=y) =YES"),
or a certain percent of DOF's can be requested to be retained ("AUTOOMIT(KEEP=z) =YES").
AUTOOMIT works best in models with good, discreet mass distribution (i.e. - use of concentrated
masses). For a model with a more homogeneous mass and stiffness distribution, AUTOOMIT will not
produce an ASET with points distributed nicely around the entire model. This can increase errors as some
more localized modes may be missed. In general, AUTOOMIT is a good tool to start generating an ASET
for a model. In order to improve accuracy of the results, additional DOF's may have to be added later.
Generalized Dynamic Reduction: (DYNRED) Generalized Dynamic reduction (as the name implies) is
not a method of eigenvalue extraction, but rather a method to reduce the size of the model other than
using ASETs or OMITs. It also produces results much faster than using static condensation. In essence,
the way that DYNRED works is by approximating the mode shapes of the model first, and then solving
for the actual modes. It generates scalar points, based on the number of modes desired, and uses those
scalar points to approximate mode shape. These represent the generalized coordinates for the mode shape
approximation. In UAI NASTRAN they are created automatically.
DYNRED, used in conjunction with SINV or MGIV, is generally a fast and accurate aid in extracting
eigenvalues. For a sparse matrix LANCZ is still faster, but DYNRED is still no slouch. Since it
approximates mode shape to begin with, it does not generate answers as accurately as SINV or LANCZ,
or MGIV by itself. Generally, the first modes are very accurate, with accuracy tailing off toward the
higher frequencies. Even at the higher modes, the accuracies are still rather good, especially when
compared to MGIV with ASET or OMIT cards.
In UAI NASTRAN, using DYNRED requires a DYNRED card in the case control to specify the
DYNRED ID, and a DYNRED card in the bulk data deck to specify the frequency range of interest.
DYNRED still requires a METHOD card in the case control, and an EIGR card.

Analysis Results

To better understand and compare the various eigenvalue extraction methods, several NASTRAN runs
were made with a simple plate model. The model was a square plate held along one edge and held at one
of the opposite corners. The model mesh started as a 5x5 mesh, but meshes of 6x6, 8x8, 10x10, 15x15,
20x20, 30x30, and 50x50 were also created from the same geometry. All these models were run using the
previously mentioned eigenvalue extraction methods to calculate the first 20 modes. The results were then
compared. All runs were made in UAI NASTRAN v8 and run on an IBM RS6000-410 computer.
Two specific pieces of data were compared: solution CPU time, and solution accuracy. For the sake of
these comparisons, the SINV results were the baseline, as this method used the entire stiffness and mass
matrices. All the methods, except for the MGIV with DYNRED runs, produced exactly the same modal
frequencies. The error in the runs with DYNRED occurred only in the higher modes, and those errors
were negligible. The main differences among the runs were the CPU times. The following table compares
the times for the various methods based on the number of DOF in the model ASET.
As the problem size increases, the SINV and LANCZ methods begin to lose their appeal. The DYNRED
method times are, essentially, unchanged.
Among dynamic reduction techniques, DYNRED generalized dynamic reduction is much more time
efficient, and generally more accurate. The graph below compares AUTOOMIT static condensation to
DYNRED for various percentages of DOF retained by AUTOOMIT. This chart is for the 6th mode of the
MLS NASTRAN model. The percentage of frequency error for a given percentage of DOF kept by
AUTOOMIT is shown in one line, with the CPU time shown on the other. The DYNRED solution time
(which had no frequency error) is shown on the CPU time axis. For comparison, the Lanczos time is also
shown:
Conclusion

For most model sizes and applications, the Lanczos eigenvalue extraction method is the most efficient.
For smaller models, or models where all the eigenvalues are to be extracted, the MGIV method may be
more useful. In addition, for models that do not have sparse matrices, the LANCZ and SINV methods
may actually be slower than using other methods. For those cases, the MGIV method with DYNRED may
be the best choice.
In addition to the example problems elaborated above, several other, more detailed NASTRAN models
were run using the various extraction methods. These were models that had been used for real-world
structural analysis. Although there have been some concerns that some solution extraction techniques may
miss mechanism or duplicate modes, this was not seen in any of the problems run.
The DYNRED method is also a more acurate and faster aid to model dynamic reduction than using
standard static condensation with AUTOOMIT. However, in cases where it's necessary to develop an
ASET, or OSET for a model, such as when deciding where to place accelerometers for a vibration test,
AUTOOMIT can be a very useful tool. It is, however, most useful for models with discrete, lumped
masses and stiffness differences. For more homogeneous mass and stiffness distributions, it is not as
efficient at selecting an ASET.
The following table summarizes the eigenvalue extraction methods detailed above:

Performance Time vs
Model Size
Method small med. large Accuracy Best Applications
MGIV good fair poor excellent small models or when all eigenvalues are desired
small to moderate size models when a range of
SINV good good (1) fair (2) excellent
frequencies are desired, and matrices are sparse
excellent excellent
LANCZ good excellent moderate to large models with sparse matrices
(3) (3)
MGIV moderate to large models, particulary with dense
good good good good
w/DYNRED matrices

(1) For dense matrices, performance goes to fair

(2) For dense matrices, performance goes to poor
(3) For dense matrices, performance goes to good
Honeycomb Plates
Properties

Define the following terms as

t = total face sheet thickness (both sheets together)

d = total core depth
E = Young's modulus of face sheets
nu = Poissons ratio (Greek letter)
g = material density of face sheet material
gc = material density of core
I' = bending mom. of inertia/width of plate
K = membrane stiffness of plate (composite)
D = bending stiffness of plate (composite)
w = weight per unit area of plate (composite)

From classical plate theory, considering only the face sheets to carry in-plane stresses:

(NOTE: Shear flexibility of the plate is not included.)

Where I' is the bending moment of inertia of the honeycomb plate per unit width:

For d >> t, I' simplifies to:

The weight per unit area of the plate is:

Thus, the stiffness to weight ratios (membrane and bending) are:

Define nondimensional stiffness/weight ratios:

Membrane

Bending
PSHELL Card

Information needed:
T - Total facesheet thickness - this should include both facesheets (the value "t" in the figure above)
D - Total core thickness (the value "d" in the figure above)
Facesheet - Material properties
Honeycomb - Material properties, cell size, and gage

Format of the PSHELL Card:

PSHELL and PID are self explanatory.

MID1, T
These terms are for in plane load of a honeycomb panel. Assume that the facesheets take all the load.
Therefore:
MID1 references the material for the facesheets.
T is the total thickness of the facesheets.
MID2, 12I/T3
These terms are for the bending load of a honeycomb panel. Assume the facesheets take all the load.
Therefore:
MID2 references the material for the facesheets.
12I/T3 is the inertia of a rectangular plate
I for Honeycomb is approximately TD2/4 for thin facesheets and a thick core.
If the facesheets are thicker relative to the core, then I is [2/3][(D/2)+(T/2)]3-[D3/12] T is the total
thickness of the facesheets.
D is the thickness of the honeycomb core.
MID3, TS/T
These terms are for the shear load of a honeycomb panel. Assume that the honeycomb core takes all the
load. Therefore:
MID3 references the material for the honeycomb core. The shear modulus and the poisson's ratio
is required on the MATERIAL card for MID3 and is obtained from the HEXEL manuals. You
need to know the cell size and gage, or the desired nominal density.
TS is the shear thickness, or for the case of honeycomb, it is the core thickness D.
T is the total thickness of the facesheets.

NSM
The non structural mass is required for honeycomb panels. The density (rho) is ignored on the MID3 card,
however, the mass of the honeycomb core and bonding material must be included. The NSM, which is
given in mass per unit area, is determined by:
NSM = (rhonominal)(D),
where D is the thickness and rhonominal is often given in pcf, pounds per cubic foot. Be careful of
your units when calculating NSM!

Jim Loughlin
Isogrid Plates

An Isogrid panel is a plate (or face sheet) with triangular integral stiffening ribs (often called stringers).
The triangular pattern was found to be very efficient because it takes advantage of the fact that triangular
trusses are very efficient structures. The term isogrid is used because the structure acts like an isotropic
material.
The first figure shows the rib pattern for an isogrid plate. While other types of rib patterns can also be
designed, e.g., square or rectangular, the triangular pattern is the only true isogrid.

The figure below shows the top view of a single isogrid cell. The flange is shown as a hidden line through
the face sheet. The terms shown are defined as:
b - rib thickness
c - flange thickness
d - rib height
h - cell height
t - face sheet thickness
w - flange width
With these values it is possible to calculate the overall section properties of the isogrid plate and
NASTRAN PSHELL input values. The derivation of the necessary equations is not included here. If you
are interested in the math, please contact Dr. William Case at the e-mail address provided at the top and
bottom of this page.
A Microsoft Excel spreadsheet is provided to calculate the section properties and PSHELL input values.
(This spreadsheet will work with Excel versions 5, 6, and 7, and may work with other versions as well.)
In the spreadsheet the values shown in the figure below are required (except z, which is called zbar in the
spreadsheet).
If you have a plugin that reads Excel spreadsheets, then clicking on the link will launch the plugin; if you
want to save the file to your hard drive, then hold down the shift key and click on the link.

Download the MS Excel spreadsheet ISOGRID.xls. (26k)

Dr. William Case, retired

Jeff Bolognese
Modal Weight Calculations
Method for Recovering Modal Mass Participation Output

MSC/NASTRAN version 2001 and beyond has the capability to extract modal mass participation directly
from nomal modes analysis. No DMAP is required and it can be used with the normal modes solution
sequence. The following is a sample executive and case control deck for a normal modes run with modal
mass extraction. The cards that are specific for the modal mass extraction are shown in bold.

Sample Input
ID XYZ,MODALWT
SOL 103
TIME 10 APP DISP CEND
TITLE=XYZ TEST VERIFIED MODEL (english - renumbered) (08/96)
SUBTITLE=FIXED BASE DYNAMICS W/ASET
SPC=1
echo=none
SPCFORCE=ALL
METHOD=100
MEFFMASS(ALL)=YES
BEGIN BULK

Sample Output
Preloads

Applying Preloads with the Differential Stiffness Approach

Differential Stiffness Theory

In geometric non-linear problems, displacements are large. Therefore, the equilibrium equations must be
written with respect to the deformed geometry. This theory assumes that the geometric stiffness matrix
can be added to the structural (or differential) stiffness matrix; therefore, the change in geometry of the
structure is only reflected in the stiffness matrix.
Differential Stiffness Matrix
A matrix that is necessary to account for the change in potential energy associated with rotation of
continuum elements under load. (stiffness effects that depend linearly on displacements)
Fundamental Assumption
External loads do not change in magnitude or direction as the structure deflects. However, the loads do
travel with the grid points.
Properties of KDGG (Differential Stiffness Matrix)
1. Symmetric
2. Independent of elastic properties
3. Depends on element geometry, displacement field, and state of stress
4. Indefinite and cannot be inverted

Setting Up the NASTRAN Data Deck

Using the differential stiffness approach to applying preloads to a NASTRAN finite element model
consists of two executions of the model. The first model is set up to obtain the differential stiffness matrix
(KDGG) as a results of the preload. The second model is where the differential stiffness matrix is added
to the geometric stiffness matrix and desired results are requested.
The DMAPs shown are for use in MSC/NASTRAN Version 67.
The First Model
Executive Control
Sol 5
Compile sol5, souin=mscsou
Alter 313
Output4 KDGG//-1//11 $(creates a fort.11 file of KDGG)
Exit $

Case Control
Title = 'title'
Subtitle =
SPC = 'boundary points'
$'Output requests' - probably want to output elements in which the
$ preload is being applied to verify the magnitude.
Subcase 1
Deform = 'id'
Subcase 2
Bulk Data
Add the necessary number of 'DEFORM' bulk data cards. The 'DEFORM' card requests the deformation
of a one dimensional element, i.e. BAR, ROD, TUBE. For a preload on a bar, the deformation is:
delta = PL / EA
where P is the applied preload. You can verify that the correct preload has been applied to the element by
reviewing the BAR, ROD, or TUBE, element forces.

The Second Model

Executive Control
Sol 24
Compile sol24, souin=mscsou
Alter 90
Inputt4 /KDGG,,,,/1/11/-1 $ Add KGG, KDGG/K1X $
Equiv K1X,KGG/Always $
Endalter

Case Control and Bulk Data decks should be set up as any other run would be.
NOTE: The differential stiffness matrix can be very big. Therefore, you should be careful to ensure that
the matrix is not sent across the network. This will slow the job execution substantially. If this is a
problem, use the ASSIGN card to write the matrix to the scratch directories on the computer that is
executing the NASTRAN job.
Cherie Congedo
May 18, 1995
Random Vibration
Nastran Input Deck Example

Following is an example of an MSC.Nastran v. 2001 random vibration analysis run. All the NASTRAN
cards necessary to perform a random run are described here. Comments follow the card (or cards) being
described. The random-specific cards are in bold. The model is a simple cantilever beam with 48 elements
and 49 nodes. It lies along the Y-axis with the first node, 101 at the origin. The downloadable data file can
be found at the bottom of this page.
NOTE: This method will only work for MSC's version 2001 or later of NASTRAN because they (finally!)
did away with the large mass needed to do frequency response runs. This method is extremely similar to
UAI's version of frequency response runs. (See notes below for the differences.)
Input Data Deck
$EXECUTIVE CONTROL DECK
ID BEAM,RANDOM
SOL 111

Modal Frequency Response Solution Number

$
TIME 10 CEND
$
$CASE CONTROL DECK
TITLE = RANDOM VIBRATION EXAMPLE MODEL
SUBTITLE = 48-ELEMENT CANTILEVER BEAM ON THE X-AXIS
LABEL = Y-DIRECTION RANDOM INPUT OF 0.01 G^2/HZ FROM 20-2000HZ
$ Beam is along the X-axis and is 48 inches long
$
SPC = 1
METHOD = 1
DLOAD = 10
SDAMPING = 20
RANDOM = 30
FREQ = 40

SPC refers only to the SPC1 card (not the SPCD card);
METHOD refers to EIGR card;
DLOAD refers to RLOADx card;
SDAMPING refers to TABDMP1 card;
RANDOM refers to RANDPS card;
FREQ refers to all FREQx cards used
$
ECHO = NONE
$
$SET 99 = 101
$SPCFORCES(PHASE) = 99

The preceding two cards are used for shaker point force recovery only (remove $ if needed). SET 99
sets the shaker GRID 101 output;
SPCFORCES requests force recovery at SET 99 (see OUTPUT and SPCD cards below)
$
OUTPUT(XYPLOT)
XYPEAK,XYPUNCH,ACCE,PSDF/ 101(T2),149(T2)
Cards used for plot output.
NOTE: THIS IS THE ONLY WAY TO GET RANDOM OUTPUT!
OUTPUT card is required to use XYPLOT commands;

XYPEAK recovers RMS and other peak responses in the .prt file (required for RMS output);
XYPUNCH creates a .pch file of response vs. frequency;
ACCE requests acceleration output (FORCE requests force output);
PSDF requests output as PSD;
GRID point 101 and 149 responses recovered; T1=X, T2=Y, T3=Z, as well as R1, R2, and R3
$
BEGIN BULK
$ PARAM,AUTOSPC,YES
PARAM,GRDPNT,0
PARAM,RESVEC,YES

PARAM,RESVEC,YES is REQUIRED for accurate results when NOT using the seismic mass. It
computes residual vectors.
GRID 101 0 0. 0. 0. 0

Shaker grid point. Often this is an RBE element attached to many GRID points with the
independent GRID being the shaker point.
$
SPC1 1 123456 101

The Shaker point is SPC’d in all DOF

$
RLOAD2,10,11,,,12,,ACCE

RLOAD defines frequency response dynamic loading;

LOAD = SPCD * TABLED1 [= A * B(f)];
11 IS SPCD card;
12 IS TABLED1 card;

ACCE refers to type of dynamic excitation, enforced acceleration in this case.

Other options are LOAD for applied load, DISP for enforced displacement, and VELO for enforced
velocity.
$
SPCD,11,101,2,1.0

Y-axis input at GRID 101;

2 refers to input direction (1=X, 2=Y, 3=Z);
1.0 is any scale factor;
$
$SPCD,11,101,2,386.4

Use this card instead when recovering forces or displacements (assuming you're using the
inch/pound system), such as at the shaker point. If you're using a metric based system, such as
mm/gram, then 386.4 becomes 9807. (You can also change the 1.0 values in the TABLED1 card
below to the acceleration value and get the same result. But do not change both.)
TABLED1 12 LOG LOG +
+ 20.0 1.0 2000.0 1.0 ENDT

TABLED1 is a tabular function defining the frequency-dependent portion of RLOAD2

The table is given as X1,Y1, X2,Y2, X3,Y3, etc.;
20.0 is the beginning frequency (X1); 1.0 is the magnitude (Y1);
2000.0 is the ending frequency (X2); 1.0 is the magnitude (Y2);
The Y values are used as scale factors; the actual Y values are entered as random input in
TABRND1 below (not frequency input);
The LOG LOG entries refer to how the data is interpolated on each axis.

NOTE: All tables MUST end with an ENDT card.

Also, tables are not always compatible between UAI/NASTRAN and MSC/NASTRAN. The tables
used in the older version of this example would not work under Nastran 2001.

TABDMP1 20 +
+ 20.0 0.1 2000.0 0.1 ENDT

Defines modal damping of Q=10;

The damping is 1/Q over the range of frequencies; in this case Q is constant. Q can vary over
frequency if the proper data is available.
$
RANDPS,30,1,1,1.0,0.0,31

Defines random PSDF loading;

The important value here is the 31, which refers to the TABRND1 card
TABRND1 31 LOG LOG +
+ 19.999 0.01 20.0 0.01 2000.0 0.01 2000.1 0.01 +
+ ENDT

Defines PSDF from 20 to 2000 Hz on a log-log scale; First value must be slightly lower than actual
frequency; Format is Frequency, PSD,
Final value must be slightly higher than actual frequency
$
FREQ1,40,20.0,2.0,140
FREQ1,40,300.0,5.0,140
FREQ1,40,1000.0,20.0,50
FREQ3,40,20.0,2000.0

FREQx cards input frequencies where responses are recorded;

FREQ1 gives beginning frequency, freq. increment size and number of increments;
FREQ3 adds each natural frequency FROM 20 TO 2000 Hz
$
EIGRL 1 0. 2000. 0 MASS

EIGRL determines normal modes from 0 to 2000 Hz

$
INCLUDE 'randbulk.dat'

Bulk Data can be elsewhere in a different file.

$ ENDDATA
Eigenvalue Output

The eigenvalue and frequency table from the .F06 file is given below. It gives the frequency range
requested in the EIGRL card, 0-2000Hz. Note that the frequencies listed in this table under CYCLES
correspond with the peaks in the plot at the bottom of the page.

Example random output of peak values.

The PSD output for GRIDs 101(T2) and 149(T2) from the run's .F06 file is given below. The output is
considered to be in units of G2/Hz, while the Grms value is in G. The PSD values are actually
(in/sec2)2/Hz, but because frequency is dimensionless, we can pretend that G and in/sec2 are equal, i.e.,
an input of 1 G comes out the same as an input of 1 in/sec2.

If force, displacement or stress output is desired however, the acceleration must be multiplied by 386.4
in/sec2 in the SPCD card to obtain pounds. (If you are using metric units of mm, then you must multiply
by 9807 mm/sec2.)

The Grms values are highlighted in the fourth column. (MSC could probably do a better job of formatting
their output tables!) Also of note are the minimum and maximum Y values (G2/Hz) and their
corresponding X values (frequency). You will probably have to scroll to the right to see everything.
Plot of the Example Random Vibration Data

Below is a plot of the data given in the punch file (random_test.pch). It shows the input and the response.
The input is a straight line at .01 G2/Hz. The response shows the peaks corresponding to the modes of the
beam.
Specification Magnitude Equations

When performing a random vibration analysis, an input spec is generally given in a form such as the log-
log plot in the figure or written in the table below. The problem is what to do with such information. We
cannot input these values directly into NASTRAN because it will not accept a slope in dB/Oct. Individual
points in G2/Hz vs. Hz are required. This page details how to get from the input to graphical points
needed.

(Note that regardless of popular opinion, G2/Hz is actually an acceleration spectral density (ASD), not a
power spectral density (PSD). PSD refers to the actual plot generated during testing, which simply reads
the power output from the accelerometers.)
In tabular form, the input may be given in this form (beginning and ending frequencies are not always
necessary if a continuous line is assumed):
Because no beginning or ending frequencies, Fl and FH, are given in this table, they must first be decided
upon. This is generally project specific. However, the frequency range is usually 20Hz to 2000Hz. From
the graph, Fl = 20Hz and FH = 2000Hz. Fh and FL are 100Hz and 600Hz, respectively.
First, determine the number of octaves between the two frequencies. Keep in mind that an octave is the
doubling of the frequency. So going from 1Hz to 2Hz is an octave and going from 1000Hz to 2000Hz is
also an octave. Thus, the number of octaves could be estimated from the graph above. The equation to
calculate the exact number is:

where FH is the higher frequency and FL is the lower frequency.

Second, determine the number of dB by multiplying the number of octaves by the slope, making sure to
use the correct sign (positive or negative) for the slope:

The previous equation also shows the definition of dB, where ASDH and ASDL are the acceleration
spectral densities for the higher and lower frequencies respectively (NOT for the higher and lower ASD
values! That is, ASDL can be greater than ASDH whereas FH is always greater than FL).

Finally, solve for the ASD at the desired frequency:

Or, for those of you who want a more expanded and complete version (where m is the slope in dB/Oct):
And a final, simplified version:

I'll leave it up to you to use the example given in the graph above to prove that the equations work.
Thanks to Bob Coladonato and Dr. Bill Case, both retired from Goddard, and Jaap Wijker of University of
Technology Delft in the Netherlands for their assistance with this page.
Calculating GRMS

It is very easy to describe the Grms (root-mean-square acceleration, sometimes written as GRMS or Grms
or grms or grms) value as just the square root of the area under the ASD vs. frequency curve, which it is.
But to physically interpret this value we need to look at Grms a different way. The easiest way to think of
the Grms is to first look at the mean square acceleration.
Mean-square acceleration is the average of the square of the acceleration over time. That is, if you were to
look at a time history of an accelerometer trace and were to square this time history and then determine
the average value for this squared acceleration over the length of the time history, that would be the mean
square acceleration. Using the mean square value keeps everything positive.
The Grms is the root-mean-square acceleration (or rms acceleration), which is just the square root of the
mean square acceleration determined above.
If the accelerometer time history is a pure sinusoid with zero mean value, e.g., a steady-state vibration, the
rms acceleration would be .707 times the peak value of the sinusoidal acceleration (if just a plain average
were used, then the average would be zero). If the accelerometer time history is a stationary Gaussian
random time history, the rms acceleration (also called the 1 sigma acceleration) would be related to the
statistical properties of the acceleration time history (you may have to refresh your probability and
statistics knowledge for this):

• 68.3% of the time, the acceleration time history would have peaks that would not exceed the +/- 1
sigma accelerations.
• 95.4% of the time, the acceleration time history would have peaks that would not exceed the +/- 2
sigma accelerations.
• 99.7% of the time, the acceleration time history would have peaks that would not exceed the +/- 3
sigma accelerations.
There is no theoretical maximum value for the Gaussian random variable; however, we typically design to
3 sigma since it would only be theoretically exceeded 0.3% of the time. In addition, from a practical point
of view, we know that it would be physically impossible to achieve unreasonably high sigma values.
Below is presented the method to calculating the root-mean-square acceleration (Grms) response from a
random vibration ASD curve.

Typical random vibration response curve:

Grms values are determined by the square root of the area under a ASD vs. frequency response curve. The
Acceleration Spectral Density values are in g2/Hz and the frequencies are in Hz.
The figure above shows a bandwidth of 10 Hz, which will be used as an example for calculating Grms.

First, calculate the number of octaves. From the plot, FL = 20Hz and FH = 30Hz. The equations below
gives #octaves = 0.58 .

Second, calculate the dB value. For FL = 20Hz, ASDL = 1.0 g2/Hz, while ASDH = 1.1 g2/Hz. The
calculated value is 0.41 dB. (The definition of dB is also provided.)

Third, calculate the slope, m, of the segment between the frequencies FL and FH. Dividing the number of
dB by the number of octaves gives m = 0.71 dB/oct.

Fourth, calculate the area under the curve between the frequencies FL and FH. In our example, A = 10.53
g2.

NOTE - The above equation is invalid if the slope m = -10log(2) because you would be dividing by zero.
If m = -10log(2), use the following equation for the area:

(L'hôpital's Rule is used to solve the equation as the limit of [1 + m/(10log2)] goes to zero -- or some such
nonsense.)
Finally, take the square root of the area for the Grms value. To finish our example, the acceleration = 3.24
Grms.

In order to calculate the Grms value for the entire curve, sum up all the areas (A1 + A2 + A3 + ... + An =
A) and take the square root of the sum.

NOTE: 3dB is a factor of 2 for ASD curves (g2/Hz) while 6dB is a factor of 2 for Grms values. For
example, reducing a peak ASD value of 12g2/Hz by -3dB would give you 6g2/Hz; reducing a value of 12
Grms -3dB results in a value of 9 Grms and reducing it -6dB results in a value of 6 Grms. This tends to be
confusing for people new to random vibration.

Thanks to Bob Coladonato and Bill Case, both now retired from Goddard, and Jaap Wijker of University
of Technology Delft in the Netherlands for their assistance with this page.
Creating a Test Specification

One purpose for performing a random vibration analysis is to create a component test level specification.
With a finite element model, a random vibration analysis can be performed to predict acceleration
responses from 20 Hz to 2000 Hz. This response is in turn used as a template to derive a test level
specification. Use these guidelines to create the spec.

1. From NASTRAN random vibration response data (or whatever analysis software you're using),
create a log-log plot of the response of ASD (acceleration spectral density) vs. frequency.
NOTE This will work using any type of ASD response curve. For example, if an engineering
model (EM) is being tested, responses from accelerometers on the EM can also be used.

2. Add in Component Minimum Workmanship to the plot. For components less than 100 lbs (45.4
kg), NASA uses these values from GEVS, Table 2.4-5:

3. Enclose response curve inside test spec curve using the following rules:
• Slopes should be less than + 25 dB/oct or greater than - 25 dB/oct (this depends on your
shaker table, consult your test engineer for exact slope limits)
• Frequency bands should be greater than 10 Hz.
• Don’t plot spec below Minimum Workmanship
• Sharp peaks can be cut off at about 1/2 their height (-3 dB) *e.g., a sharp peak of 0.52 g2/Hz
can be cut off at > 0.26 g2/Hz
• Drop spec curve into valleys that are large
• Try to keep the overall grms level of the spec to be no more than 1.25 times the overall grms
level of the response curve.
The following plot is an example of a spec to be used for component random vibration testing.

* NOTE: 3dB (3.01dB to be exact) is a factor of 2 for ASD curves (g2/Hz) while 6dB is a factor of 2 for
grms values. For example, reducing a peak ASD value of 12g2/Hz -3dB would give you 6g2/Hz; reducing
a value of 12grms -3dB results in a value of 8.48grms and reducing it -6dB results in a value of 6.0grms.
This tends to be confusing for people new to random vibration.

Thanks to Bob Coladonato, retired from Goddard, for his assistance with this page.
Miles Equation

Ryan Simmons
NASA Goddard Space Flight Center
May 2001

The following equation is attributed to John W. Miles.

Where

GRMS = Root Mean Square Acceleration in G's (sometimes given as ÿRMS).

= Natural Frequency

= Transmissibility (or amplification factor) at

where is the critical damping ratio.

[ASDinput] = Input Acceleration Spectral Density at in units of .

In 1954, Miles developed his version of this equation for GRMS as he was researching fatigue failure of
aircraft structural components caused by jet engine vibration and gust loading. Miles simplified his
research by modeling a system using one degree of freedom only. He also applied statistical advances that
had been made at the time. While his goal was to analyze the stress of a component, the equation can be
rearranged and used to determine, among others, displacement, force, and, in our case, acceleration.
Single Degree of Freedom System
Miles' Equation is derived using a single degree of freedom (SDOF) system (lightly damped), consisting
of a mass, spring and damper, that is excited by a constant-level "white noise" random vibration input
from 0 Hz to infinity. Miles' Equation is thus technically applicable only to a SDOF system.

This figure shows a typical representation of a SDOF oscillator. The mass (m) is
attached to the spring (stiffness k) and the damper (damping c). The system is
forced by the random vibration function (F) in the y-direction only.

Derivation of Miles' Equation is left up to you. The references at the bottom of the
page provide different methods of deriving the equation, so reviewing each method
could provide insight into Miles' Equation.

SDOF System Response Plot

The plot below shows the input and response of a SDOF system. Miles' Equation calculates the square
root of the area under the response curve, providing us with the GRMS value.
Response Parameters
Miles' Equation can also be used to predict other response parameters such as stress or displacement. For
example, the displacement equation for YRMS is given here. Displacement can in turn be used to
calculate the stress.

NOTE: In this form of the equation, since we want units of length, the ASDinput must be in units of
[length/sec2]2/[Hz].
You will most likely have to convert your ASD data from g2/Hz to something else.

USES OF MILES' EQUATION

Design
During the design of a part, if enough analysis has been performed to determine the part has a
predominant resonant frequency, then Miles' Equation can be used to estimate the loads due to random
vibration. Just calculate the GRMS value and multiply it by 3. That's the "three-sigma" load. Keep in
mind, this is conservative.
Testing
Accelerations due to random vibration at resonant frequencies in a multiple degree of freedom system can
be approximated using Miles' Equation. This will indicate how much of the overall RMS acceleration is
occurring at a resonant peak of interest compared to the complete frequency spectrum.

MISUSES OF MILES' EQUATION

Miles' Equation Does NOT Work in Reverse
Accelerations cannot be determined during random vibration testing using Miles' Equation. An upper
bound on loads can be calculated using the 3-sigma value, but that's about it. In other words, a part
designed to 3-sigma equivalent static loads will survive a random vibration test. However, Miles'
Equation cannot be used to predict the failure of a part designed to less than 3-sigma levels.
Miles' Equation Does NOT Give an Equivalent Static Load
Calculating the GRMS value at a resonant peak after a random vibration test and multiplying it by the test
article mass does not mean that the test article was subjected to that same, equivalent static load. It simply
provides a statistical calculation of the peak load for a SDOF system. The actual loading on a multiple
DOF system due to random input depends on the response of multiple modes, the mode shapes and the
amount of effective mass participating in each mode. Static testing must still be done.
Miles' Equation may not be Conservative for a Shaped Input Spectrum
Miles' Equation is based on the response of a SDOF system subjected to a flat random input. If Miles'
Equation is used to calculate the GRMS response to a shaped input spectrum, the result will not
accurately reflect the rigid body response below the resonant frequency. If the shaped input spectrum has
high ASD levels below the resonant frequency, then Mile's Equation may significantly underpredict the
GRMS response. The following plot shows this condition. Note that the Miles' assumption has a response
of 31.8 GRMS while the actual SDOF response is much greater at 39.7 GRMS. This demonstrates that
Miles' Equation is best used when the input ASD is flat or nearly so.

REFERENCES
John W. Miles, On Structural Fatigue Under Random Loading, Journal of the Aeronautical Sciences, pg.
753, November, 1954.
Dave S. Steinberg, Vibration Analysis for Electronic Equipment, Wiley-Interscience, New York
Thomas P. Sarafin (editor), Spacecraft Structures and Mechanisms: From Concept to Launch,
Microcosm, Inc., Torrance, CA, 1995

Thanks to Scott Gordon and Bob Coladonato for their invaluable assistance in the creation of this page.
Ryan Simmons
May 2001
Definition of Decibels

The decibel is a unitless method of expressing the ratio of two quantities. The expression is in terms of the
logarithm to base 10 of the ratio instead of the raw ratio. This is done for convenience in expressing the
ratio of numbers many magnitudes apart with decibel numbers that are not as large.

If G1 and G2 are two vibration acceleration amplitudes, then their decibel ratio (or decibel difference) is:

For example, if G1 is twice G2, then

Or G1 is about 6 dB greater than G2. If G1 were half as large as G2 then:

Or G1 is about 6 dB less than G2. It is seen from the dB equation that each factor of 10 in the ratio results
in 20 dB difference in amplitudes.

Note from equation 1 that:

That is, for squares of amplitudes (power) each factor of 10 in the ratio results in 10 dB difference in
power. Thus, a factor of 1000 in the power ratio is a 30 dB difference in power. Equation 2 is the normal
way of defining decibels, and equation 1 results from it.
Also for convenience, decibels are used rather than Bels. From Equation 2, a Bel would be defined as:

Seeing the equation written out in this form makes the definition much clearer.
Thanks to Dr. Bill Case for the definition.
Sine Vibration
Nastran Input Deck Example

Following is an example of a NASTRAN swept sine frequency response analysis run. All the NASTRAN
cards necessary to perform a frequency response run are described here. Comments follow the card (or
cards) being described. The frequency response-specific cards are in bold. The model is a simple
cantilever beam with 48 elements and 49 nodes. It lies along the Y-axis with the first node, 101 at the
origin. The downloadable data file can be found at the bottom of this page.
NOTE: This method will only work for MSC's version 2001 or later of NASTRAN because they (finally!)
did away with the large mass needed to do frequency response runs. This method is extremely similar to
UAI's version of frequency response runs. (See notes below for the differences.)

$EXECUTIVE CONTROL DECK

ID SINE,TEST
SOL 111

Modal Frequency Response Solution Number

$
TIME 10
CEND
$
$CASE CONTROL DECK
TITLE = SINE VIBRATION EXAMPLE MODEL
SUBTITLE = 48-ELEMENT CANTILEVER BEAM ON THE X-AXIS
LABEL = Y-DIRECTION SINE INPUT
$ Beam is along the X-axis and is 48 inches long

$
ECHO = NONE
$
SPC = 1
METHOD = 1
DLOAD = 10
SDAMPING = 20
FREQ = 40

METHOD refers to EIGRL card;

DLOAD refers to RLOADx card;
SDAMPING refers to TABDMP1 card;
FREQ refers to all FREQx cards used

$
$SET 98 = 101
$SPCFORCES(PHASE) = 98

For force recovery, these two cards create SET1 for the shaker point, GRID 101,
then request forces (as magnitude/phase) at the shaker point (see SPCD card below)

$
SET 99 = 101, 149
Defines GRIDs for response output
$
ACCELERATION(SORT2,PRINT,PUNCH,PHASE) = 99

Requests frequency response output for SET 99;

Acceleration output given as magnitude/phase angle;
Output is written to the .F06 and punch files;
SORT2 sorts output as GRID1, all frequencies; GRID2, all frequencies; etc.
(SORT1 sorts output as frequency 1, all GRIDs; frequency 2, all GRIDs; etc.)

$
OUTPUT(XYPLOT)
XYPEAK,XYPUNCH,ACCE/ 101(T2),149(T2)

Cards used for plot output.

XYPEAK recovers peak responses in the .prt file;
XYPUNCH creates a .pch file of response vs. frequency;
ACCE requests acceleration output (FORCE requests force output);
GRID point 101 and 149 responses recovered; T1=X, T2=Y, T3=Z, as well as R1, R2, and R3
(NOTE: the punch file will contain both ACCELERATION and OUTPUT data.)

$
BEGIN BULK
$
PARAM,AUTOSPC,YES
PARAM,GRDPNT,0
PARAM,RESVEC,YES

PARAM,RESVEC,YES is REQUIRED for accurate results when NOT using the seismic mass. It
computes residual vectors.

$
GRID 101 0 0. 0. 0. 0

Shaker grid point.

$
SPC1 1 123456 101

The Shaker point is constrained in all DOF. Often this is an RBE element attached to many GRID
points with the independent GRID being the shaker point.

$
RLOAD2,10,11,,,12,,ACCE

RLOAD defines frequency response dynamic loading;

LOAD = SPCD * TABLED1 [= A * B(f)];
11 IS SPCD card;
12 IS TABLED1 card;
ACCE refers to type of dynamic excitation, enforced acceleration in this case.

$
SPCD,11,101,2,1.0

Y-axis input at GRID 101;

2 refers to input direction (1=X, 2=Y, 3=Z);
1.0 is any scale factor;

$
$SPCD,11,101,2,386.4

Use this card instead when recovering forces or displacements, such as at the shaker point. If you're
using the Metric system, for example mm, then use 9807 instead of 386.4.

$
$
TABLED1 12 LOG LOG +
+ 2.0 0.25 500.0 0.25 ENDT

TABLED1 is a tabular function defining the frequency-dependent portion of RLOAD2

This table gives the sine vibration input spec;
The table is given as X1,Y1, X2,Y2, X3,Y3, etc.;
2.0 is the beginning frequency (X1); 0.25 is the magnitude (Y1);
500.0 is the ending frequency (X2); 0.25 is the magnitude (Y2);
If your input is written as double amplitude (D.A.), a conversion is available.

All tables MUST end with an ENDT card.

$
TABDMP1 20 +
+ 20.0 0.1 2000.0 0.1 ENDT

Defines modal damping of Q=10;

The damping is 1/Q over the range of frequencies; in this case Q is constant.

$
FREQ1,40,2.0,2.0,250
FREQ3,40,2.0,500.0

FREQx cards input frequencies where responses are recorded;

FREQ1 gives beginning frequency, freq. increment size and number of increments;
FREQ3 adds each natural frequency up to 500 Hz

$
EIGRL 1 0. 500. 0 MASS

EIGRL determines normal modes up to 500 Hz

$
INCLUDE 'sinebulk.dat'

Bulk Data can be elsewhere in a different file.

$
ENDDATA
Double Amplitude

When you input your frequency and acceleration data into a sine vibration frequency response
NASTRAN deck, you may run into that cryptic notation, 0.32 inch D.A. from 10-20Hz. This is what is
known as a vibrational amplitude. But since NASTRAN does not know what D.A. is, and you probably
don't either, you need to convert that value into something everyone understands, such as a vibrational
acceleration, which could be measured in g's or in/sec2 or m/s2. Fortunately, this conversion is not
difficult to do.
Start with the fact that acceleration is the second time derivative of length:

At maximum amplitude (ignore the minus sign):

Therefore:

By definition xo is the single amplitude:

Also, we know:
Finally we get the acceleration in terms of the length you are using (length/sec2):

If you are working in an inch/pound system and you want your answer in g, here is the final equation:
Semi-Rigid Connections

One of the basic assumptions of conventional structural analysis is that joints are either perfectly rigid or
perfectly hinged. Therefore, when analyzing a structure, joints are idealized as either fixed or hinged.
However, in actual structures, typical connections do not behave in either a perfectly rigid or a perfectly
hinged manner. The different types of connections that are commonly used fill the entire flexibility
spectrum from flexible "hinge like" connections to semi-rigid connections to rigid connections. These
non-ideal connections are often referred to as flexible connections, semi-rigid connections, spring hinged
joints, or sticky hinges.

To achieve a more accurate analysis of a structure it would be advantageous to include the true behavior
of the joints. For example, there is a substantial variation in the distribution of bending moment for a
beam with hinged, semi-rigid, and fixed joints (Figure 1). (W = force/unit length)

There is also a substantial difference in the minimum natural frequency and mode shape of a beam with
hinged, semi-rigid, and fixed joints (Figure 2).

E = Young's Modulus; I = moment area of inertia; L = length of beam; W = force/unit length

Semi-rigid joints can be modeled using a rotational spring which separates a member from it's
surrounding environment. The rotational spring is characterized by it's moment-rotation curve, where the
rotational stiffness of the spring k is equal to the slope of the moment-rotation curve (Figure 3}. The value
of k for a given connection detail can be determined by testing.

The magnitude of the stiffness of the spring hinge k does not determine if a joint will behave in a rigid or
flexible manner. Rather it is the ratio of the spring stiffness k to the flexural stiffness of the member EI/L
that determines what behavior will be exhibited. This ratio is often called the "joint stiffness" and is
defined as kL/EI.

In order to visualize this phenomenon consider some "medium sized" spring with spring stiffness K-
medium (Figure 4). If a very flimsy, spaghetti-like member is connected to it, even though the spring is
only "average sized" it will still keep the flimsy member from rotating at the support, just like a rigid
connection would. However, if a very stiff member were connected to our "average sized" spring, it
would appear as if the spring had little to no effect in restraining the member.
This now raises the question of what specific values of the stiffness ratio will result in hinged behavior,
semi-rigid behavior, or fixed behavior. To answer this question the dynamic and static behavior of a
prismatic beam with uniform load was investigated.

A finite element model of a typical sized beam with spring hinges at the supports was created (Figure 5
inset). A NASTRAN normal modes analysis was performed to obtain the minimum natural frequency for
various values of the spring stiffness k. The results of this analysis are shown in Figure 5.
As k gets smaller the solution approaches the pin-pin solution. as k gets larger the solution approaches the
fixed-fixed solution. Furthermore, there is a discernable range in which the behavior changes from pined
to fixed.
The results shown in Figure 5 are for a specific beam. For the results to be of use they must be applicable
to any beam. Therefore, the behavior of a second beam with increased mass (100 times greater than the
initial test case) and the behavior of a third beam with increased flexural stiffness (100 times greater than
the initial test case) was investigated, the results are shown in Figure 6.

When the mass was increased the natural frequency dropped but the transition range remained the same.
Therefore it appears that the mass has no effect on the joint stiffness. When the flexural stiffness was
increased the natural frequency increased and the transition range shifted also, demonstrating that the
flexural stiffness of the beam effects the joint stiffness.
The previous results can be non-dimensionalized. The natural frequency can be non-dimensionalized
using the natural frequency of the pin-pin case as zero and the natural frequency of the fixed-fixed case as
1. And instead of the spring stiffness the joint stiffness ratio of K/ EI/L will be used. The results plotted in
this manner are shown in Figure 7.
When plotted in this manner all three test cases fall on the same line. From this it can be seen that if the
joint stiffness is less than 1 the natural frequency is within 10% of the pin-pin condition and if the joint
stiffness is greater than 100 the natural frequency is within 10% of the fixed-fixed condition.

Conclusion
Rotational springs can be used to model semi-rigid restraint conditions. The restraint conditions can be
characterized by a quantity called the joint stiffness. The joint stiffness depends on the ratio of the spring
stiffness k to the flexural stiffness EI/L of the beam and is defined as kL/EI. NASTRAN analysis results
confirm that the mass of a beam does not effect the joint stiffness, only the flexural stiffness of the beam
does.
It was also found that when the joint stiffness is less than 1.0 the beam behaves as if it is hinged, while if
the joint stiffness is greater than 100 the beam behaves as if it is fixed. Therefore it may not be worth the
effort to model a connection with a rotational spring if the resulting joint stiffness being used is less than 1
or greater than 100. Furthermore, if semi-rigid behavior is desired a joint stiffness between 1 and 100
should be used. Figure 7 can help in determining what joint stiffness should be to used to achieve a
particular behavior.

Written by Dr. John McGuire

Unit Consistency

Have you ever wondered why your results look a bit odd sometimes: are those pound forces or pound
masses?!... will I get Newtons if I input millimeters?! You complete your analysis only to determine your
units were wrong. If so, then this table of consistent units might help you.

NOTE:
t = Metric ton = 1000 kg = 1 Mega gram (Mg)
Validity Checks

After a finite element model is created and before results are used from that model, Code 542 performs
several standard validity checks on the model. This document will describe these standard validity checks.

There are four mathematical validity checks. They are:

Unit Enforced Displacement and Rotation
Free-Free Dynamics with a Stiffness Equilibrium Check
Unit Gravity Loading
Unit Temperature Increase

The first three checks should be performed on all models used for either statics or dynamics analyses. For
models developed for thermal distortion analyses, the Unit Temperature Increase check should also be
performed.

Computer-Aided Accuracy Checks

In addition to the mathematical validity checks given above, there are many other model checks that can
and should be performed. The checks listed below require a pre/post processor, such as FEMAP,
Pro/Engineer, or Patran. Furthermore, you should be aware of any warnings or potential errors your solver
gives you, such as grid point singularities. This is not an exhaustive list, just some suggestions.
Free Edge - Use a free-edge check to make sure elements are connected properly, particularly where 1D,
2D, and 3D elements are connected to zero-length elements (springs).
Coincident Nodes - Check for erroneous coincident nodes; but be careful not to merge intentional
coincident nodes.
Coincident Elements - Check for coincident elements; this can happen if you are not careful when
creating multiple meshes.
Bar/Beam Orientation - Use orientation visualization to determine if bar and beam elements are oriented
properly.
Plate Element Distortion - Check elements for problems like aspect ratio, internal angles, warping.
Plate Element Normals - Check that plate normals are in the same direction where necessary.
Mass - Check to be certain the model's mass is reasonable and accurate.
Common Sense Accuracy Checks
The following checks should speak for themselves. They are just simple, common sense checks we
should make on every model. If you've been bitten by a common sense check that's not on this list, e-mail
it to us and we'll consider putting it on the list.
Material Properties - Make sure the property values are correct, match the units you're using elsewhere,
and include all the information you need. For example, thermal expansion coefficient, reference
temperature, density, etc. And know if you're using mass density or weight density. It could cause
problems when retrieving forces.
Element Properties - Verify beam cross sections and corresponding area moment of inertias, plate
thicknesses and element offsets. Verify mass and inertia properties for concentrated mass elements and
stiffnesses for elastic elements.
Units - Are your units consistent? NASTRAN doesn't care what units you use and won't give you any
warnings. Some pre-processors will also let you use inconsistent units. Take particular care with strange,
mixed units (e.g. - mm and Kg). Are you using mass or weight units? You may need a PARAM,WTMASS
card. For inch/pound models, the WTMASS value is 0.00259.
Dimensions - Check to make sure that your overall geometry is correct. Check your FEM vs. drawings or
CAD. This is especially important if you've transformed units.
Grounding Conditions - Make sure your constraints are in the correct locations and DOFs. Check
NASTRAN SPC forces to make sure only your mount points are listed.
Proper Element Use - Are you using the right element for the job? Are you using it correctly? For
example, RBE3's aren't really "rigid" and shouldn't be used as structural elements. They are primarily for
calculating average motion or distributing loads. CELAS elements should be between coincident grids or
may generate artificial forces during rigid body rotation.
MPC's and RBE's - Are your MPC's formulated correctly? Are you using the appropriate RBE? For
example, is that component you're modeling with that RBE2 really stiffer than the structure you're
attaching it to? Will you artificially lock up your structure with a big RBE? Avoid over-constraining your
structure with RBE's.
Coordinate Systems - If you're using multiple input and output coordinate systems, make sure that you
know which ones your elements and constraints are referencing. For example, RBE elements will use the
grid output coordinate system when assigning DOFs.
Understand singularities - If you have singularity errors, are they because of rotational degrees of
freedom from plates and solids? Are those removed by AUTOSPC? If you are unsure, then you should try
to remove the singularities and re-run the model. If some still exist, they may indicate grounding
problems.
Validity checks, i.e., validation of the finite element model, are defined as checks that ensure the model is
mathematically accurate. Validity checks do not ensure the accuracy of the model in representing a
physical system, just that the model will give mathematically correct results.
Accuracy checks help the modeler to determine and ensure that the model represents the physical system
he or she is modeling.
- Contrast these checks with the verification of a model, which is generally done by correlating a model
with actual mechanical test results.
Unit Enforced Displacement and Rotation

Purpose
The Unit Enforced Displacement and Rotation check verifies that no illegal grounding (such as incorrect
SPC or MPC) is present in the model. The model should move as a rigid body when it is translated one
unit or rotated one radian. This check verifies displacements, element forces, and grid point forces. The
check is performed in all three translational and all three rotational directions. This check outputs lots of
data.
Output Results
The displacements results from the three translational subcases should be 1.0 in the input direction and
0.0 in the other five directions. From the rotational subcases, the rotation in the input direction should be
1.0 and the other two rotations should be 0.0. The three translational displacements from the rotational
subcases will not be zero. The element forces and grid point force balances should be zero with a
maximum of 0.1 lb force and 1.0 in-lb moment. The grid point held should not have other SPC's, and, for
best results, should be near the center of gravity of the model.

Input Required
EXECUTIVE DECK must include the following information:

SOL 101

CASE CONTROL must include the following information:

SPCFORCES=ALL
DISP=ALL
GPFORCES=ALL
ELFORCES=ALL
SPC=100

(only for normal plate or solid rotations and the structure must be free at base)

SUBCASE 1
SUBTITLE=UNIT ENFORCED DISP IN X DIRECTION
LOAD=100
SUBCASE 2
SUBTITLE=UNIT ENFORCED DISP IN Y DIRECTION
LOAD=200
SUBCASE 3
SUBTITLE=UNIT ENFORCED DISP IN Z DIRECTION
LOAD=300
SUBCASE 4
SUBTITLE=UNIT ENFORCED ROTA ABOUT X AXIS
LOAD=400
SUBCASE 5
SUBTITLE=UNIT ENFORCED ROTA ABOUT Y AXIS
LOAD=500
SUBCASE 6
SUBTITLE=UNIT ENFORCED ROTA ABOUT Z AXIS
LOAD=600

BULK DATA must include the following information: (The displaced grid is grid point 17104 in this
example.)

BEGIN BULK
SPCD 100 17104 1 1.0
SPCD 200 17104 2 1.0
SPCD 300 17104 3 1.0
SPCD 400 17104 4 1.0
SPCD 500 17104 5 1.0
SPCD 600 17104 6 1.0

$ SPC FOR UNIT ENFORCED DISP CHECK

SPC1 100 123456 17104
FORCE 100 17104 0 0.0 1.0 0.0 0.0
FORCE 200 17104 0 0.0 0.0 1.0 0.0
FORCE 300 17104 0 0.0 0.0 0.0 1.0
MOMENT 400 17104 0 0.0 1.0 0.0 0.0
MOMENT 500 17104 0 0.0 0.0 1.0 0.0
MOMENT 600 17104 0 0.0 0.0 0.0 1.0
Unit Enforced Displacement or Rotation with Different Output Coordinate Systems
EXECUTIVE DECK must include the following information:

SOL 24

$ CAUSES GRID DISPACEMENTS TO BE OUTPUT IN THE BASIC COORD. SYSTEM

$ !WARNING! OUTPUT ONLY DISPLACEMENTS
$ *********FORCES AND SUCH MAY NOT BE CORRECT*********
COMPILE SOL24,SOUIN=MSCSOU
ALTER 160 $
PARAM //'ADD'/ALWAYS/-1/0 $
VECPLOT UGV,BGPDT,EQEXIN,CSTM,CASECC,/UGVBASIC//0/1 $
EQUIV UGVBASIC,UGV/ALWAYS $
ENDALTER

CASE CONTROL must include the following information:

$SPCFORCES=ALL
DISP=ALL
$GPFORCES=ALL
$ELFORCES=ALL
SPC=100 $(only for normal plate and solid rotations
$ and the structure must be free at base)
SUBCASE 1
SUBTITLE=UNIT ENFORCED DISP IN X DIRECTION
LOAD=100
SUBCASE 2
SUBTITLE=UNIT ENFORCED DISP IN Y DIRECTION
LOAD=200
SUBCASE 3
SUBTITLE=UNIT ENFORCED DISP IN Z DIRECTION
LOAD=300
SUBCASE 4
SUBTITLE=UNIT ENFORCED ROTA ABOUT X AXIS
LOAD=400
SUBCASE 5
SUBTITLE=UNIT ENFORCED ROTA ABOUT Y AXIS
LOAD=500
SUBCASE 6
SUBTITLE=UNIT ENFORCED ROTA ABOUT Z AXIS
LOAD=600

BULK DATA must include the following information:

BEGIN BULK
SPCD 100 17104 1 1.0
SPCD 200 17104 2 1.0
SPCD 300 17104 3 1.0
SPCD 400 17104 4 1.0
SPCD 500 17104 5 1.0
SPCD 600 17104 6 1.0
$ SPC FOR UNIT ENFORCED DISP CHECK
SPC1 100 123456 17104
MOMENT 400 17104 0 0.0 1.0 0.0 0.0
MOMENT 500 17104 0 0.0 0.0 1.0 0.0
MOMENT 600 17104 0 0.0 0.0 0.0 1.0
FORCE 100 17104 0 0.0 1.0 0.0 0.0
FORCE 200 17104 0 0.0 0.0 1.0 0.0
FORCE 300 17104 0 0.0 0.0 0.0 1.0
Free-Free Dynamics with a Stiffness Equilibrium Check

Purpose

The Free-Free Dynamics with a Stiffness Equilibrium check verifies that the model will act as a rigid
body when it is unconstrained. It also checks the stiffness matrix to verify that it doesn't contain any
grounding effects, such as illegally specified SPC or rigid elements. To perform this check, remove all
external constraints (usually SPCs) and perform standard normal modes run recovering at least the first
six modes.

Output Results

The model should obtain a minimum of six rigid body modes. These modes should have frequencies less
than or equal to 1.0E-04 Hz. The GROUNDCHECK case control card should be used as a stiffness
equilibrium. Strain energy filters are set by the THRESH parameter in the card. By default, the threshold
is calculated by dividing the largest term in the stiffness matrix by 1.0e+10. The output will indicate
whether the check is passed or failed.

The equilibrium check gives possible grounding in the model. In the output file look G-SET, N-SET, F-
SET, and A-SET. The maximum strain energy for each DOF will be shown in these tables as well as
whether or not it passes or fails. The following descriptions give possible problems from a particular
table. Note that an error in a given set will cause errors to appear in subsequent sets, regardless of if any
exist.

G-set
• CELAS problems. Non-coincident grids or misaligned output coordinate systems (most
common).
• DMIG or DMAP inserted matrices associated with GRIDs. For the latter they should obviously
be inserted before the check.
• Bad NASTRAN elements. This is rare but possible. Also small errors from a lot of elements will
eventually add up, so very large models often have relatively poor equilibrium.

N-set
• Incorrect or imprecise MPC equations. Usually similar to CELAS problems. NASTRAN runs
with improperly inserted Craig-Bampton models will show errors in this table if the boundary, as
well as modes, were placed at SPOINTs. This is because this is the first set where geometry is
associated with the boundary.

F-set
• Incorrect SPC's. Note that a constrained model will obviously show grounding.

A-set
• The model is constrained to ground (and not intended to be Free-Free).
• The reference GRID point in the GROUNDCHECK command is located too far from the model
CG. The GRID parameter is optional.
• PARAM,AUTOSPC,YES constrains near-singular DOF.
Input Required

EXECUTIVE DECK must include the following information:

SOL 103 for MSC

CASE CONTROL must include the following information:

GROUNDCHECK(SET=ALL)=YES
$ (this will print out all the set info - sample output below)
SPCFORCES=ALL
DISP=ALL
SPC=100
$ (only for normal plate rotations, structure must free at base)
METHOD= 20

BULK DATA must include the following information:

EIGR, 20, $ (Use any method you want to obtain all rigid bodies modes and the
$ first flexible mode. We recommend LANCZOS.)

Sample GROUNDCHECK Output

*** USER INFORMATION MESSAGE 7570 (GPWG1D)

RESULTS OF RIGID BODY CHECKS OF MATRIX KGG (G-SET) FOLLOW:
PRINT RESULTS IN ALL SIX DIRECTIONS AGAINST THE LIMIT OF 9.216514E+00
DIRECTION STRAIN ENERGY PASS/FAIL
--------- ------------- ---------
1 2.098354E-05 PASS
2 6.123831E-06 PASS
3 3.367564E-08 PASS
4 1.326509E-03 PASS
5 9.882150E-04 PASS
6 4.239914E-04 PASS

SOME POSSIBLE REASONS MAY LEAD TO THE FAILURE:

1. CELASI ELEMENTS CONNECTING TO ONLY ONE GRID POINT;
2. CELASI ELEMENTS CONNECTING TO NON-COINCIDENT POINTS;
3. CELASI ELEMENTS CONNECTING TO NON-COLINEAR DOF;
4. IMPROPERLY DEFINED DMIG MATRICES;
.
.
.
*** USER INFORMATION MESSAGE 7570 (GPWG1D)
RESULTS OF RIGID BODY CHECKS OF MATRIX KNN (N-SET) FOLLOW:
PRINT RESULTS IN ALL SIX DIRECTIONS AGAINST THE LIMIT OF 9.216514E+00
DIRECTION STRAIN ENERGY PASS/FAIL
--------- ------------- ---------
1 2.098167E-05 PASS
2 6.120336E-06 PASS
3 3.724166E-08 PASS
4 1.326376E-03 PASS
5 9.887038E-04 PASS
6 4.241481E-04 PASS

SOME POSSIBLE REASONS MAY LEAD TO THE FAILURE:

 1. MULTIPOINT CONSTRAINT EQUATIONS WHICH DO NOT SATISFY RIGID-BODY MOTION;
2. RBE3 ELEMENTS FOR WHICH THE INDEPENDENT DEGREE-OF-FREEDOM CANNOT DESCRIBE
ALL POSSIBLE RIGID-BODY MOTIONS.

.
.
.
*** USER INFORMATION MESSAGE 7570 (GPWG1D)
RESULTS OF RIGID BODY CHECKS OF MATRIX KFF (F-SET) FOLLOW:
PRINT RESULTS IN ALL SIX DIRECTIONS AGAINST THE LIMIT OF 9.216514E+00
DIRECTION STRAIN ENERGY PASS/FAIL
--------- ------------- ---------
1 2.098167E-05 PASS
2 6.120336E-06 PASS
3 3.724166E-08 PASS
4 1.326376E-03 PASS
5 9.887038E-04 PASS
6 4.241481E-04 PASS

SOME POSSIBLE REASONS MAY LEAD TO THE FAILURE:

1. CONSTRAINTS WHICH PREVENT RIGID-BODY MOTION.
.
.
.
*** USER INFORMATION MESSAGE 7570 (GPWG1D)
RESULTS OF RIGID BODY CHECKS OF MATRIX KAA1 (A-SET) FOLLOW:
PRINT RESULTS IN ALL SIX DIRECTIONS AGAINST THE LIMIT OF 9.216514E+00
DIRECTION STRAIN ENERGY PASS/FAIL
--------- ------------- ---------
1 2.098167E-05 PASS
2 6.120336E-06 PASS
3 3.724166E-08 PASS
4 1.326376E-03 PASS
5 9.887038E-04 PASS
6 4.241481E-04 PASS

IF THE MODEL HAS PASSED THE PREVIOUS CHECKS FOR THE G-SET AND N-SET,
THEN SOME POSSIBLE CAUSES ARE:
1. THE MODEL IS NOT INTENDED TO BE FREE-FREE WHICH INDICATES THAT THE MODEL IS
PROPERLY CONSTRAINED TO GROUND;
2. THE REFERENCE GRID POINT (GRID=GID ON THE GROUNDCHECK COMMAND) IS LOCATED
TOO FAR FROM THE MODEL'S CENTER OF GRAVITY. IT IS RECOMMENDED THAT THE
REFERENCE GRID POINT BE LOCATED AS CLOSE AS POSSIBLE TO THE MODEL'S CENTER
OF GRAVITY OF THE MODEL (SEE THE GRID POINT WEIGHT GENERATOR OUTPUT);
3. PARAM,AUTOSPC,YES CONSTRAINS NEAR-SINGULAR DEGREES-OF-FREEDOM. WHEN A
FINITE ELEMENT MODEL WITH AUTOSPC FAILS THE A-SET CHECK, IT IS NOT EVIDENT
THAT GROUNDING HAS OCCURRED. THE USE OF PARAM,SNORM WILL NOT ELIMINATE THE
SPURIOUS FAILURE.
Unit Gravity Loading in All Three Translational Directions

Purpose
The Unit Gravity Loading validity check verifies that the model will provide accurate displacements and
reactions forces under gravity loading. This is a good check to perform if your model will be used for
quasi-static loads analysis.

Output Results
No large displacements should be found as a result of this validity check. In addition, forces and moments
at only the mounting locations should be obtained when SPCFORCES=ALL is used in the CASE
CONTROL. The forces should add up to the total weight of the model in the input loading direction and
the forces in the other directions should be zero.

Input Required

EXECUTIVE DECK must include the following information:

SOL 24 or 101 for MSC

CASE CONTROL must include the following information:

SPCFORCES=ALL
DISP=ALL
SPC=100 (both normal plate rotations and structure held at base)
SUBCASE 1
LABEL=UNIT GRAVITY LOADING IN X DIRECTION
LOAD=139997
SUBCASE 2
LABEL=UNIT GRAVITY LOADING IN Y DIRECTION
LOAD=139998
SUBCASE 3
LABEL=UNIT GRAVITY LOADING IN Z DIRECTION
LOAD=139999

BULK DATA must include the following information:

PARAM GRDPNT 0
GRAV 139997 0 386.4 1. 0. 0.
GRAV 139998 0 386.4 0. 1. 0.
GRAV 139999 0 386.4 0. 0. 1.
$ (386.4 in/sec2, use PARAM WTMASS .00259 in BULK DATA)
Unit Temperature Increase

Purpose
The Unit Temperature Increase validity check verifies that the model will provide accurate displacements
due to temperature loading. This check should be performed on all models used for thermal distortion
analyses.

Output Results
The displacements should be compared to the theoretical solution of alpha*l*deltaT. This comparison can
be done easily when a sorted bulk data deck of the model is run through the program THDEF, located on
"merger" in MERGER::SYS$SYSDISK:[USER,THDEF]. The program will provide a file containing the
theoretical displacements for the model. The displacements from NASTRAN should be the same as the
displacements from the THDEF program. The model must contain a kinematic mount to run the THDEF
program. Also, the NASTRAN model should contain a kinematic mount for this validity check. A
limitation to the THDEF program is that it does not handle local coordinate systems.
The grid point force balance should be zero for all grid points and only forces at the mounting locations
should be output when SPCFORCES=ALL is included in the CASE CONTROL. Look at the
displacements of the model in your post-processor to make sure it expands as expected.

Input Required
It is suggested you create a kinematic mount for your model, e.g., one GRID point constrained in all 6
DOFs.

EXECUTIVE DECK must include the following information:

SOL 24 or 101 for MSC

CASE CONTROL must include the following information:

SPCFORCES=ALL
DISP=ALL
GPFORCES=ALL
SPC=100 (both normal plate rotations and kinematically mounted structure)
TEMP=50

BULK DATA must include the following information:

TEMPD 50 21. (reference temperature + 1.)

RBE2s that are not joined by coincident nodes should not be used.
Change all MAT1, MAT8, MAT9, etc., to contain the same thermal expansion coefficient and the same
reference temperature (20. in this example).
Yield Stress Notes

A Few Definitions
Yield Stress = sigma y
Ultimate Stress = sigma u

Sigma y and sigma u are determined from uniaxial tensile tests. Since the exact point at which yield
occurs may not be visually obvious it must be defined in some other consistent manner. For most ductile
metals the following convention has been widely accepted. The yield point is determined using a line
parallel to the linear portion of the curve and originating at epsilon = 0.002 (The 0.2% offset method).

Since several tests of the same material will result in slightly different values for sigma y and sigma u, a
statistical approach must be used to determine the exact value that will be used. For example: stress
values listed in the military handbook (MIL-HDBK-5F) are catagorized as A-Basis and B-Basis. Quoting
from the hand book:

• A-Basis. -- At least 99 percent of the population of values is expected to equal or exceed the A
basis mechanical property allowable, with a confidence of 95 percent.
• B-Basis. -- At least 90 percent of the population of values is expected to equal or exceed the B
basis mechanical property allowable, with a confidence of 95 percent.

From this it can be seen that just determining what value to use for the yield stress requires some thought.

A Few More Definitions (for bi-axial stress)

Normal Stress in x direction - sigma x
Normal Stress in y direction - sigma y (not the same as yield stress)
Shear Stress in xy plane - tau xy
Principal Stresses - sigma 1 and sigma 2
For a 2D plate the state of stress at a point can be described using any arbitrary axis. The state of stress
using this axis will be described by the values sigma x, sigma y, and tau xy.

There will exist some orientation of the axis at which when the state of stress is described, the shear stress
will be zero. The axis when oriented at this angle is called the principal axis. The normal stresses for this
special orientation are defined as the principal stresses, sigma 1 and sigma 2.
Tresca Yield Criteria
(Max Shear Stress Yield Criteria)

Tresca postulated that a material under a multi-axis state of stress will yield when the Max Shear Stress
reaches some critical value. Using this assumption the yield envelope for a bi-axial state of stress is as
shown in Figure 4. (Note: sigma y in Figure 4 is the yield stress)

Why the yield envelope takes this shape can be illustrated by inspecting Morh's circle for various states of
stress. First, consider the uniaxial tensile test at the moment yield occurs (Figure 5.). For this case one of
the principal stresses will be equal to the yield stress and the other principal stress will be zero. This
condition defines the critical value of the shear stress at which yield occurs. According to Tresca, yield
will occur only if the shear stress exceeds this value. It should be noted that this value is half the value of
the yield stress.
If either of the principal stresses exceed the yield stress, the out of plane shear will exceed the Max Shear
value (see Figure 6.). This is because for 2D problems the out of plane principal stress (sigma 3) is
assumed to be zero. Therefore, neither sigma 1 nor sigma 2 may exceed sigma y.

In the case where one of the principal stresses has the opposite sign of the other (i.e. one in compression
the other in tension. Quadrant 2 and 4 of the Tresca yield envelope) yield will occur before either of the
principal stresses reaches the yield stress. As illustrated in Figure 7. if sigma 2 becomes more negative
sigma 1 must also decrease (Other wise the shear stress will increase beyond the max value). Essentially
you may slide a constant size circle along the axis between positive and negative sigma y.
Von Mises Yield Criteria

Von Mises postulated (1913) that a material will yield when the distortional energy at the point in
question reaches a critical value. The distortional energy written in terms of the 2D principal stresses and
the yield stress is as follows.

Figure 8. shows the Von Mises Yield envelope superposed over the Tresca yield envelope. The scatter of
test data is such that either criteria is as applicable as the other. For hand calculations Tresca is easy to
implement; for a computer program Von Mises is convenient because the entire envelope can be
represented with a single equation.