Page 1 of 255

NX Nastran Dynamic Analysis

Transient, Frequency Response,

Random Response, and Response
Spectrum Analysis Types
 Page 2 of 255

Review of Fundamentals

NX Nastran Dynamic Analysis

 Page 3 of 255

Dynamic Analysis Process

Overview
NX Nastran is a general purpose Finite Element Analysis
solver capable of simulating a broad range of
engineering problems in many different industries.
The solver can be run as a stand-alone solver using an
existing NASTRAN input file or in conjunction with EDSs
Finite Element Analysis Pre and Post-Processor,
FEMAP.
NX Nastran 1.0 is analogous with MSC NASTRAN 2001,
release 9 and runs on Windows NT, 2000, XP, and 2003
server. Versions of the stand-alone solver are available
on various UNIX (HP, Sun, SGI, and IBM) and LINUX
platforms.
 Page 4 of 255

Dynamic Analysis Process

Overview
Dynamic System
Environment Design

Finite Element Interfacing

Model Media

Modal
Analysis

Dynamic
Response
Analysis
 Page 5 of 255

Single DOF System

Dynamic equation of motion:
p(t)
+ku(t) = p(t)
m(t) + bu(t) u(t)

m = mass (inertia)
m
b = damping (energy dissipation)
k b
k = stiffness (restoring force)

p = applied force

u = displacement of mass

= velocity of mass
u

= acceleration of mass

and p are time varying in general.

u, u,
m, b, and k are constants
 Page 6 of 255

Units
+ku(t) = p(t)
m(t) + bu(t)
Fundamental units
- Length, L (inch, meter)
- Mass, M (slug, kilogram)
- Time, T (second)

Fundamental and derived units

-m M
-b MT-1
-k MT-2
-p MLT-2
-u L

- u LT-1
- LT-2
 Page 7 of 255

Units
Variable Dimensions English Metric
Length L in m
Mass M lb-sec2/in kg
Time T sec sec
Area L2 in2 m2
Volume L3 in3 m3
Velocity LT-1 in/sec m/sec
Acceleration LT-2 in/sec2 m/sec2
Rotation - rad rad
Rotational Velocity T-1 rad/sec rad/sec
Rotational Acceleration T-2 rad/sec2 rad/sec2
Circular Frequency T-1 rad/sec rad/sec
Frequency T-1 cps; Hz cps; Hz

L = Length
M = Mass
T = Time
- = Dimensionless
 Page 8 of 255

Units
Variable Dimensions English Metric
Eigenvalue T-2 rad2/sec2 rad2/sec2
Phase Angle - deg deg
Force MLT-2 lb N
Weight MLT-2 lb N
Moment ML2T-2 in-lb N-m
Mass Density ML-3 in-sec2/in4 kg/m3
Youngs Modulus ML-1T-2 lb/in2 Pa;N/m2
Poissons Ratio - - -
Shear Modulus ML-1T-2 lb/in2 Pa;N/m2
Area Moment of inertia L4 in4 m4
Torsional Constant L4 in4 m4
Mass Moment of inertia ML2 in-lb-sec2 kg-m2
Stiffness MT-2 lb/in N/m
Viscous Damping Coeff. MT-1 lb-sec/in N-sec/m
Stress ML-1T-2 lb/in2 Pa: N/m2
Strain - - -
 Page 9 of 255

Single DOF System Undamped Free

Vibrations
Dynamic equation
m(t) + ku(t) = 0
Solution
u(t) = A sin nt + B cos nt
k
n = m = natural frequency (rad/sec)
n
fn = = natural frequency (cycles/sec)
2
Initial conditions
B = u (t = 0)

u(0) and u(0) are known
u
A= (t = 0)
n
Finally
u(0)
u(t) = sinnt + u(0)cosn
n
 Page 10 of 255

Single DOF System Undamped Free

Vibrations

Displacement

Time
SDOF Oscillator Nonzero initial conditions
 Page 11 of 255

Single DOF System Damped Free

Vibrations
Dynamic Equation
+ku(t) = 0
m(t) + bu(t)
Critical Damping
bc = 2 km = 2mn
Initial conditions
b
=
bc
The amount of damping determines the form of the solution
Underdamped
b < bc

u(t) = e-bt/2m(A sindt + B cosdt)

Damped Natural Frequency:

d = n 1 2
 Page 12 of 255

Single DOF System Damped Free

Vibrations
Critically damped
b = bc
No oscillation occurs
u(t) = (A + Bt)e-bt/2m
Overdamped

b > bc

No oscillation occurs. The system gradually returns to Equilibrium (at rest,

undisplaced) position.
The usual analysis case is underdamped
Structures have viscous damping in the 0-10% range
 Page 13 of 255

Single DOF System Damped Free

Vibrations

Displacement

Time (seconds)
 Page 14 of 255

Single DOF System Undamped Sinusoidal

Vibrations
Dynamic Equation
m(t) +ku(t) = P sint
Where = forcing frequency
Solution
P/k
u(t) = A sindt + B cos dt + sint
1- / n
2 2

Initial Conditions Steady-State

where
u(t=0) P/k
B = u(t=0) A= -
n (1-2/ n2)n
Steady-state Solution
P/k is the static response
1 is the dynamic magnification factor
1-2/ n2
 Page 15 of 255

Magnification factor

Frequency ((0.5)*Hz)
 Page 16 of 255

Single DOF System Damped Sinusoidal

Vibrations
Dynamic Equation
+ku(t) = P sin t
m(t) + bu(t)
Transient solution decays rapidly and is of no interest

Steady-state Solution
sin (t + )
u(t) = P/k
(1-2/ n2)2 + (2/n)2

2/n
= - tan-1
(1-2/ n2)

is phase lead
 Page 17 of 255

Single DOF System Damped Sinusoidal

Vibrations
For >> 1
n

Magnification factor 1(Static Solution)

Phase angle 360(Response is in phase with the force)

For << 1
n

Magnification factor 0 (No Response)

Phase angle 360(Response has opposite sign of force)

For 1
n

Magnification factor 1/2

Phase angle 270
 Page 18 of 255

Multi-Degree of Freedom System

The equation becomes:

+ [K]{u} = {P}
[M]{} + [B]{u}

Where: {u} = Displacement

[M] = Mass Matrix
[B] = Damping Matrix
[K] = Stiffness Matrix
{P} = Forcing function
 Page 19 of 255

Classification of Dynamic
Environments
Deterministic Random

Periodic Transient Stationary Nonstationary

Simple Shock Ergodic

Harmonic Spectra
 Page 20 of 255

NX Nastran Dynamic Excitations

Pulse Sinusoidal

Transient Random
 Page 21 of 255

Finite Element Dynamic

Modeling Considerations
Frequency Range

Grid Points/constraints/elements

Linear versus nonlinear behavior

Interaction with adjacent media

Test/measured data integration

Damping
 Page 22 of 255

Dynamics Modeling Input

NX Nastran Dynamic Analysis

 Page 23 of 255

NX Nastran Input File Setup

FMS and NASTRAN Statements File allocation and system call

Executive Control Section Solution type, time allowed, system

diagnostics
CEND Required Delimiter Entry

Case Control Section Output requests, selects certain Bulk Data

items

BEGIN BULK Required Delimiter Entry

Bulk Data Section Structural Model definition, solution conditions

ENDDATA Required Delimiter Entry

 Page 24 of 255

Finite Element Analysis

Real World is not comprised of only SDOF systems!
Finite Elements are used to model the mass, damping, and stiffness
of complex systems and structures.
Degrees of freedom (DOF) are independent coordinates that
describe the motion of the structure at any instant in time.
GRIDs are used to model the continuous structire as a discrete
entity.
Each GRID may have six DOFs: 3 in translation directions (X, Y,
and Z) and 3 in rotation about the X, Y, and Z axes.
y Time 2
Time 1

x 2
z 1
Book-keeping is done via the matrices that define the relationships
between the DOFs.
 Page 25 of 255

Commonly Used Elastic Elements

One-Dimensional Geometry Number of DOFs
ROD Pin-ended rod 4
BAR Prismatic beam 12
BEAM Straight beam with warping 12
BEND Curved beam, pipe, or elbow 12
Two-Dimensional Geometry
TRIA3 Triangular plate 15
QUAD4 Quadrilateral plate 20
SHEAR 4-sided shear panel 8
TRIA6 Triangular plate with midside nodes 30
QUAD8 Quadrilateral plate with midside nodes 40
Three-Dimensional Geometry
HEXA Solid with six quadrilateral faces 24-60
TETRA Solid with four triangular faces 12-30
PENTA Solid with two triangular faces and three 18-45
quadrilateral faces
Zero-Dimensional Geometry
ELAS Simple spring connecting two degrees of freedom 2
 Page 26 of 255

Commonly Used Elastic Elements

Use the Model -> Property
command and in the Define
Property dialog box, click
Elem/Property Type Button. The
desired Element/Property Type
can then be chosen from the
dialog box.
 Page 27 of 255

Coupled vs. Lumped Mass

Coupled mass is generally more accurate than lumped mass.
Lumped mass is preferred for computational speed in dynamic
analysis
User-selectable coupled mass matrix for elements
PARAM, COUPMASS, 1 to select coupled mass
The Default is lumped mass
Elements which have either lumped or coupled mass:
BAR, BEAM, CONROD, HEXA, PENTA, QUAD4, QUAD8, ROD, TETRA,
TRIA3, TRIA6, TRIAX6, TUBE
Elements which have lumped mass only:
CONEAX, SHEAR
Elements which have coupled mass only:
BEND, HEX20
 Page 28 of 255

Coupled vs. Lumped Mass

Lumped mass contains only diagonal, translational components (no
rotational components).
Coupled mass contains off-diagonal translational components, as
well as, rotations for BAR (no torsion), BEAM, and BEND elements.
 Page 29 of 255

ROD Finite Element

2 4 Length, L
1 3 Area, A
Torsional Constant, J
Youngs Modulus, E
L Shear Modulus, G

AE -AE
Stiffness Matrix: 0 0
L L
GJ -GJ
0 0
L L
k= -AE AE
0 0
L L
-GJ GJ
0 0
L L

Classical consistent Mass: 1/3 0 1/6 0

I I
0 3A 0 6A
m = AL 1/6 0 1/3 0
I I
0 6A 0 3A
 Page 30 of 255

ROD Finite Element

NX Nastran Lumped Mass: 1/2 0 0 0

0 0 0 0
m = AL
0 0 1/2 0

0 0 0 0

NX Nastran Coupled Mass: 5/12 0 1/12 0

0 0 0 0
m = AL
1/12 0 5/12 0

0 0 0 0

The translational terms represent the average of lumped mass and classical consistent
mass. This average is best for ROD and BAR elements.
 Page 31 of 255

Justification of Nastran
Coupled Mass Convention
Consider a fixed-free rod:

1 u(t)

L Single Element
2

Exact quarter-wave natural frequency:

E/ E/
1/4 = = 1.5708
2L L
 Page 32 of 255

Justification of Nastran
Coupled Mass Convention
Different approximations

- Lumped Mass
E/ E/
L = 2 = 1.414 (-10%)
L L

- Classical consistent mass

E/ E/
L = 3 = 1.732 (+10%)
L L

NX Nastran

- Coupled Mass
E/ E/
L = 12/5 = 1.549 (-1.4%)
L L
 Page 33 of 255

Mass Units
NX Nastran assumes consistent units. BE CAREFUL!!!!
Weight units may be input instead of mass units if this is more
convenient. The PARAM,WTMASS must then be used to convert
them to mass.
Weight-to-mass conversion:
Mass = (1/G)* Weight (G=Gravity Acceleration)
Mass Density = (1/G) * Weight Density
PARAM,WTMASS, factor performs conversion with factor = 1/G
The default value for the factor is 1.0
Example
Input RHO () = 0.3 for weight density
Use PARAM,WTMASS,0.00259 (1/386.4) to multiple 0.3 for G = 386.4
in/sec2
PARAM,WTMASS is used once per run and multiples all
weight/mass input (including MASSi, CONMi, and nonstructural
mass input). Use all mass of all weight inputs (Do not mix!!)
 Page 34 of 255

Mass Units
Material Density
MATi entries
1 2 3 4 5 6 7 8 9 10
MAT1 MID E G NU RHO A TREF GE

Select Model->Material and this dialog box will appear

 Page 35 of 255

Mass Units
Nonstructural mass
Mass input on element property entry which is not associated with geometric
properties of the element. Input as mass/length for line elements and mass/area
for elements with 2-D geometry
Select Model->Propertyclick Elam/Property Type button. Select plate element
and this dialog box will appear.
 Page 36 of 255

Mass Units
Scalar Mass
CMASSi, PMASS
Grid point mass
CONM1 (6 x 6 mass matrix) User defines half of the terms, symmetry is
assumed.
Select Model->Propertyclick Elem/Property Type button.
Select Mass Matrix and this dialog box will appear
 Page 37 of 255

Mass Units
CONM2 (concentrated mass) M
M SYM
M
Select Model->Propertyclick I11
Elem/Property Type button. -I21 I22
-I31 -I32 I33
Select Mass and this dialog
box will appear:
 Page 38 of 255

Basic NX Nastran Set Operations

See the NX Nastran Quick Reference Guide
Grid Set (G) = N + M
M Multipoint Constraints

Independent DOF (N) = F + S

S Single Point Constraints

Unconstrained DOF (F) = A + O

O Static Condensation, Guyan or CMS

Analysis Set A = L + R
R Free-Body Partitioning

Solve A-Set Modes

Reverse Process for Data Recovery to G-Set

 Page 39 of 255

Tips on Model Verification

PARAM,GRDPNT,V1 (V1 > 0)

Grid point weight generator

PARAM,USETPRT, V1 (V1 = 0,1,2)

NX Nastran set tables

As always, engineering judgment

 Page 40 of 255

Normal Modes Analysis

NX Nastran Dynamic Analysis

 Page 41 of 255

Natural Frequencies and Normal Modes

Determine the dynamic characteristics of the structure
- For instance, suppose a piece of rotating machinery, such as a motor, is
mounted on a structure. The running motor will produce a frequency that
may be close to one of the natural frequencies of the structure. The motors
frequency may excite the structure and create excessive vibration.

- Sometimes static loads can be subject to dynamic amplification

- Natural frequencies and normal modes are often used as a base or a guide
to subsequent dynamic analysis (transient response, response spectrum),
such as what should be the appropriate t for integrating the equation of
motion in transient analysis.

- Transient analysis can also take advantage of normal modes and natural
frequencies using a technique known as modal expansion

- Can be used to guide placement of accelerometers in physical testing

 Page 42 of 255

Theoretical Results
Start with this equation:
[M] { x } + [K] { x } = 0 (1)

Assume a harmonic solution:

{ x } = {}eit (2)

(Physically, this means that all the coordinates are in synchronous

motion and the system configuration does not change shape during
motion, only amplitude)

From the harmonic solution equation:

{ x } = - 2{}eit (3)

 Page 43 of 255

Theoretical Results
Through substitution of (2) and (3) into equation (1)
-2 [M] {}eit + [K] {}eit = 0

That equation simplifies to:

([K] - 2 [M]) {} = 0 (4)

This is an Eigenvalue problem

 Page 44 of 255

Theoretical Results
From equation (4) there are two possible solutions:
If det ([K] - 2 [M]) 0, the only possibility, from equation (4), is

{} = 0

which produces a so-called trivial solution and will not reveal

anything about the behavior of the system from a physical
perspective

If the det ([K] - 2 [M]) = 0, there is a non-trivial solution for {}

The Eigenvalue problem reduces to solving:

det ([K] - 2 [M]) = 0 or det ([K] - [M]) = 0

where = 2
 Page 45 of 255

Theoretical Results
If the structure has N dynamic degrees of freedom (with mass),
there are N number of s that are solutions to the Eigenvalue
problem.
- These s (1, 2, , N) are the natural frequencies of the
structure, sometimes referred to as:
Characteristic Frequencies
Fundamental Frequencies
Resonance Frequencies

The eigenvector {j} associated with the natural frequency j is

called the normal mode or mode shape
- The normal mode corresponds to deflated shape patterns of the
structure

When a structure is vibrating, its shape at any time is a linear

combination of its normal modes.
 Page 46 of 255

Theoretical Results
Example
Simply Supported Beam

Example plots of the first three modes:

- Mode 1 1

- Mode 2 2

- Mode 3 3
 Page 47 of 255

Facts Regarding Normal Modes

When [K] and [M] are symmetric and real (this is true for all the
standard structural finite elements), the following orthogonality
property holds:

{i}T[M]{j} = 0 If ij
and
{i}T[K]{j} = 0 If ij
also
{i}T[K]{j}
j2 = {i}T[M]{j}

The natural frequencies (1, 2, ) are expressed in

radians/second. They can also be expressed in hertz
(cycles/second) using:
j (radians/second)
fj (hertz) =
2
 Page 48 of 255

Facts Regarding Normal Modes

- For Example the system below is unconstrained and
has a rigid-body mode

x1 x2

m m

1
1 = 0 {1} =
1

When a structure is not fully constrained, meaning it will exhibit a

rigid-body mode (stress free mode) or a mechanism, at least one
natural frequency will be zero.
 Page 49 of 255

Facts Regarding Normal Modes

Scaling of normal modes is arbitrary. For instance:

m x1

m x2

1 300 0.66
{1} = , {1} = , and {1} =
0.5 150 0.33

All represent the same mode of vibration

 Page 50 of 255

Facts Regarding Normal Modes

FFor practical considerations, modes should be normalized by a
chosen convention. In NX Nastran there are three normalization
choices (except when using the Lanczos method)

TThe unit value of generalized mass (default)

{i}T [M]{i} = 1.0

TThe unit value of the largest A-set component in each mode

TThe unit value of a specific component (not recommended)

IIn Lanczos method, normalization is to a unit value of generalized

mass or to a unit value of the largest component.
 Page 51 of 255

Additional Modal Properties

Since strains, internal loads, and stresses develop when a
structure deforms, additional useful modal information can be
recovered by utilizing:
sStrain-displacement relationships
{} = [Ku] {u}

SStress-strain relationships
{} = [K] {}

SStatic force-displacement relationships

{Pst} = [K] {u}

EElement strain energy relationships

Ve = 1/2{ue}T[Kee]{ue}
 Page 52 of 255

Methods of Computation
NX Nastran provides the user with three types of
methods for eigenvalue extraction
- Tracking Methods
Natural Frequencies (Eigenvalues) are determined one at a time using
an iterative approach
Two variations of the inverse power method are provided using INV
and SINV
This approach is more convenient when a small number are to be
determined
- In general, SINV is more reliable than INV
- Transformation Methods
The original eigenvalue problem
([K] - [M]){} = 0
is transformed to the form:
[A]{} = {} where [A] = [M]-1[K]
 Page 53 of 255

Methods of Computation
Then, the Matrix [A] is transformed into a tri-diagonal matrix using
either the Givens technique or the Householder technique
Finally, all eigenvalues are extracted at once using the QR algorithm
Two variations of both the Givens and the Householder methods are
provided for use:
- GIV
- MGIV
- HOU
- MHOU
These methods are more efficient when a large number of
eigenvalues are needed to be extracted
- Lanczos Method
The newest method, Lanczos is a combined tracking-
transformation method
- This method is most efficient for computing a few eigenvalues of
large, sparse problems
 Page 54 of 255

Sturm Sequence Theory

CChoose .
FFactor [K iM] into [L][D][LT].
TThe number of negative terms on the factor diagonal is the
number of eigenvalues below .

No. Neg. No. Neg.

Terms=7 Terms=8

0.0
8 (must be in range)
 Page 55 of 255

Lanczos Method
Block, shifted, inverted Lanczos

Random starting vectors

Automatic shift logic

Partial and selective othogonalization

Sturm sequence diagnosis

Givens plus QL eigensolution

Can be used for both buckling and normal modes analysis

Mass and largest component normalize only

 Page 56 of 255

Creating Analysis Set with Analysis Set

Manager
All Analysis cases should be set-up using the Analysis Set Manager. The
Analysis Set Manager is accessed using the Model->Analysis command.
Note: For users of previous versions of FEMAP, the analysis set manager was
added for version 8.0, it is the recommended way to create an input file.

Analysis Set Manager Main Window

Copies Existing Case

Delete Existing Case

Loads Saved Set

Creates New Set

Edits Existing Set Information

 Page 57 of 255

Create Analysis Set

In Analysis Set Manager dialog box, click the New button.
The Analysis Set dialog box will appear:
Select 2..Normal Modes/Eigenvalue
from the Analysis Type drop-down
menu. Then click Next button

In the NASTRAN Executive and

Solution Options dialog box,
Executive Control options can be
specified, such as problem ID,
diagnostics, restarts, or output
directories. Also, Solution options
such as using the iterative solver
or multiple processors can be
selected. Click Next button.
 Page 58 of 255

Create Analysis Set

The NASTRAN Bulk Data Options dialog box will appear:
In the NASTRAN Bulk Data
Options dialog box, important
PARAMs for dynamic analysis can
be specified such as GRDPNT
and WTMASS. Also, input file
formats, plate and beam element
types (for instance, CQUADR for
plates with drilling degree-of-
freedom), and manual additions to
the Bulk Data.

PARAMs not in the PARAM

portion of the dialog box may be
added to the analysis set using
the Manual Control section of
the dialog box. For example, the
PARAM,USETPRT can be added
to the NX Nastran analysis deck
Click the Next button
by clicking the Start Text button,
and entering PARAM,USETPRT
in text format.
 Page 59 of 255

Creating Analysis Set

NASTRAN GEOMCHECK Diagnostic specific Dialog Box

Type of Message to
alert user in .f06 file
(Fatal, Warning, or
Information)

Tolerance values
(different for each
Test, defaults listed)

Message Limit (If

more than this many
occur, wont be
written to .f06 file)
Click the Next button Summary Report
All Tests on/off toggle
 Page 60 of 255

Special Case Control using FEMAP

NASTRAN Model Check Dialog Box (Weight Check)
WEIGHTCHECK is a Case Control
command used at each stage of the
mass matrix reduction, compute rigid
body mass, and compare it with the
rigid body mass of the G-set

The DOF sets can be chosen from the

letters G, F, A, V, N, or N +AUTOSPC.
For further information on DOF sets
see NX Nastran Quick Reference
Guide Section about DOF Set
Definition.
CGI requests output of center of gravity and mass moment of inertia (Default:CGI=NO)

Ref Node refers to GRID point used for calculation of the rigid body motion. (Default is
the origin of the basic coordination system.)

Units selects the output in units of weight or mass (Default=WEIGHT)

 Page 61 of 255

Special Case Control using FEMAP

NASTRAN Model Check Dialog Box (Ground Check)
GROUNDCHECK is a Case Control
command used to perform grounding
check analysis on stiffness matrix to
expose unintentional constraints by
moving the model rigidly.

The DOF sets can be chosen from the

letters G, F, N, A, or N +AUTOSPC.
For further information on DOF sets
see NX Nastran Quick Reference
Guide Section about DOF Set
Definition.

DATAREC refers to data recovery of grounding forces. (Will Print top n% of forces)

Ref Node refers to GRID point used for calculation of the rigid body motion.

Max Strain Energy states what the maximum strain energy that passes the check
(Default value is largest term in the stiffness matrix divided by 1.E10)
 Page 62 of 255

Create Analysis Set

The NASTRAN Dynamic Analysis dialog box will appear:
In this dialog box, the Modal
Solution Method can be chosen.

The Frequency range of interest,

the number of modes to be
retrieved, the normalization
method, and the type of mass
(Lumped or Coupled) can be set.

Also, the Solution type can be set

to either Direct or Modal for more
advanced dynamic analysis.

Click the Next button

 Page 63 of 255

Create Analysis Set

The NASTRAN XY Output for Modal Analysis dialog box will appear:
The NASTRAN XY Output for Modal Analysis
dialog box appears if you pick the Modal solution
type on the NASTRAN Dynamic Analysis dialog
box for the following solution types: Normal
Modes/Eigenvalue, Random, and Buckling. It also
applies to Transient Dynamic/Time History and
Frequency/Harmonic response when the system
modes are calculated. This dialog box controls the
type of modal participation information that is
written to the PRINT output file (*.f06).
Click the Next button 3
times
If you enter a Reference Node, Nastran will use it
for the calculation. If you leave the value as 0,
Nastran will use the origin of the global rectangular
coordinate system. FEMAP will read the output
information into a FEMAP function. In FEMAP, you
can display this data as an XY plot.
 Page 64 of 255

Create Analysis Set

The Boundary Conditions dialog box will appear:

The Boundary Conditions dialog box is

used to choose the Constraint and Load
sets to be used in the active analysis set.

Other boundary condition sets such as

initial conditions, constraint equations,
and other DOF sets can be chosen in
this dialog box as well.

Click the Next button

 Page 65 of 255

Create Analysis Set

The Nastran Output Requests dialog box will appear:
The Nastran Output
Requests dialog box allows
the user to select what output
NX Nastran should create for
post-processing.

Element corner (Guass

Points) results and Output
modes can be specified.

Types of results files to

create (print f06 files, post-
process op2 files, punch
pch files, or a combination
of these output files can be
selected with the Results
Destination drop-down menu. Click the OK button
 Page 66 of 255

Problem #1

Modal Analysis of a Flat Plate

 Page 67 of 255

Problem #1: Modal Analysis

of a Flat Plate
For this problem, use Lanczos method to find the first ten natural frequencies and
mode shapes of a flat rectangular plate. Build a finite element representation of
the rectangular plate. The plate will be 5 inches by 2 inches and the material and
element properties are on the next page. The left side of the plate will be fixed.
 Page 68 of 255

Problem #1: Modal Analysis

of a Flat Plate

Length 5 in
Height 2 in
Thickness 0.100 in
Weight Density 0.282 lbs/in3
Mass/Weight Factor 2.59E-3 sec2/in
Elastic Modulus 30.0E6 lbs.in2
Poissons Ratio 0.3
 Page 69 of 255

Problem #1: Modal Analysis

of a Flat Plate

Use these results for comparison:

Mode 1 133.1684 Hz
Mode 2 648.7171 Hz
Mode 3 821.3796 Hz
Mode 4 2043.021 Hz
Mode 5 2277.875 Hz
Mode 6 2357.667 Hz
Mode 7 3704.534 Hz
Mode 8 4343.623 Hz
Mode 9 4762.875 Hz
Mode 10 5569.165 Hz
 Page 70 of 255

Reduction in Dynamic
Analysis
NX Nastran Dynamic Analysis
 Page 71 of 255

Introduction to Dynamic Reduction

Definition
Dynamic Reduction means reducing a given dynamic math model to
one with fewer degrees of freedom.

Why Reduction for Dynamics?

The model may be to large to solve without reduction.
The model has more detail than required.
Dynamic reduction is cheaper then the precise analysis of a large
model. In other words, each analysis will take less time when reduction
is put to use.
Dynamic reduction is more accurate (and will likely solve in les time)
than constructing a separate, smaller dynamic model.
 Page 72 of 255

Reduction Methods for Dynamics

Available with NX Nastran
Guyan reduction (static condensation)

Modal reduction
 Page 73 of 255

Static Condensation
(Internal Calculation)
Let {uf} be the set of the unconstrained (free) structural coordinates
Partition
{uo}
{uf} =
{ua}

where
ua = analysis set
uo = omitted set

Degrees of freedom removed during

User-Selected dynamic
 Page 74 of 255

Static Condensation
(Internal Calculation)
Form a static equation for uf and partition the stiffness matrix into the
O-set and the A-set.

Koo Koa uo Po
=
KoaT Kaa ua Pa

Assume Po is zero and solve for uo in terms of ua

{uo} = [Goa] {ua}

[Goa] = -[Koo]-1 [Koa]
 Page 75 of 255

Static Condensation
(Internal Calculation)
Transformation from the A-set to F-set is:

uo Goa
{uf} = = {ua}
ua I

O-set is dependent upon the A-set. The motion of the O-set is a
linear combination of the A-set motions. The columns of Goa are the
static shape vectors
The equations of motion for the F-set are written in terms of the
A-set
TMf{a} + TBf{u a} + TKf{ua} = TPf
or
Maaa + Baau a + Kaaua = Pa
 Page 76 of 255

Static Condensation
(Internal Calculation)
Dynamics problems are solved in terms of the reduced coordinates
(A-set). O-set components are recovered.

O-set mass, damping and stiffness is spread to the A-set.

The largest cost (increased run time) is associated with the

formulation of Maa and Baa, particularly for nondiagonal (coupled
mass) Mff.

The resulting Kaa, Baa, and Maa are small and dense (i.e. matrix
bandedness is destroyed).
 Page 77 of 255

Static Condensation
(Internal Calculation)
Summary
Separate free degrees of freedom (uf) into the omitted set (uo) and
the analysis set (ua) by means of OMIT entries or ASET entries.
Retain only a small fraction of the DOFs (typically 10% or less) in
the analysis set because the computer costs for static condensation
increase rapidly with the size of the analysis set. Otherwise, retain
all of the DOFs.
Retain DOFs with large concentrated masses in the analysis set.
Retain DOFs that are loaded (in transient and frequency response
analysis)
Retain DOFs to adequately describe deflected shape or modes of
interest.
 Page 78 of 255

Solution Control for Guyan Reduction

Executive Control Section

Any SOL

Case Control Section

No special commands required

Bulk Data section

ASET (optional* - specifies A-set)

OMIT (optional* - specifies O-set)
Using the Analysis Set Manager in
FEMAP, the A-Set or O-set can be
selected as boundary condition sets
in the lower portion of the Boundary
Conditions dialog box.
 Page 79 of 255

Solution Control for Guyan Reduction

Creating an A-Set:
First, create a new constraint set using the Model->Constraint->set command.
Next, use the Model->Constraint->Nodal command

Choose the nodes wanted for the A-Set, then Click OK.
Now, choose which DOFs for the nodes should be in the A-Set.

Finally, click OK, then Cancel.

 Page 80 of 255

Difficulties with Guyan Reduction

User effort in selecting A-Set points
Accuracy depends on users skill in selecting A-Set points
Regardless of users skill, high accuracy requires a large number of
A-Set points (increases run-time) 2 to 5 times the number of
modes desired
Stiffness reduction is exact; mass and damping reductions are only
approximations
No loss in accuracy of modes occurs when omitting mass-less
degrees of freedom
Errors are most pronounced at higher modes
Local modes may be missed altogether
Not generally recommended, except when performing test-analysis
correlation
 Page 81 of 255

Difficulties with Guyan Reduction

The static condensation approximation may miss the local dynamic
effects

0
{uo} = [Goa] {ua} + {uoo} Local Dynamic Effect

Physical Variables
Static Transformation

{uoo} = [Koo-1] {Po}

Loads on O-set Components

 Page 82 of 255

Modal Reduction
All NX Nastran linear dynamic solutions have two versions:
Direct The solution is solved in terms of the A-Set
Modal The solution is solved in terms of modal coordinates (H-set)

In the modal solution sequences the A-Set coordinates are written in

terms of modal coordinates
{ua} = [a] { }

Modal Coordinates
Matrix of Mode Shapes

Modal vectors (mode shapes) are solutions to the undamped

eigenvalue problem (A-Set coordinates)
[Maa] {a} + [Kaa] {ua} = 0
 Page 83 of 255

Modal Reduction
Equations of motion for the A-Set are written in terms of modal
coordinates (H-set notation, modal coordinates are handled
internally) Note: E-Set DOFs are not shown here for clarity.
[a]T[Maa][a][] + [a]T[Baa][a][] + [a]T[Kaa][a][] = [a]T[Paa]
If [] is mass normalized and there are no K2PP, M2PP, B2PP, or TF, then:

[ I ]{} + []T[Baa][a]{} + [i2][] = []T[Pa]

Note: A-Set matrices may be reduced matrices from Guyan Reduction or GDR.
Transformation from model coordinates to the F-Set would require two transformations.

{uf} = [] {ua}
{ua} = [a] {}
{uf} = [] {a} {}
 Page 84 of 255

Solution Control for Modal Reduction

Executive Control Section

Any modal dynamic

analysis SOL

Case Control Section

METHOD (required
selects Bulk Data EIGR
or EIGRL entry)

Bulk Data section

EIGR or EIGRL (required selects parameters for eigenanalysis)

 Page 85 of 255

Solution Control for Modal Reduction

Using the Analysis Set Manager:
Choose 4..Frequency/Harmonic Response from the Analysis Type
drop-down menu

Click the Next button 5 times until

the NASTRAN Dynamic Analysis
dialog box appears. Click Modal in
the Solution Type portion of the
dialog box. Then click OK.
 Page 86 of 255

Problem #2

Normal Modes Analysis using

Guyan Reduction
 Page 87 of 255

Problem #2: Normal Modes Analysis of a

Flat plate using Static Reduction
For this problem, use Guyan Reduction to reduce the model used in
Problem #1. Then find the first five natural frequencies and mode shapes
using the Automatic Givens method. Use the nodes indicated on the next
page for the A-Set.
 Page 88 of 255

Problem #2: Modal Analysis of a Flat

Plate Using Static Reduction

Create a NEW constraint set to represent the A-Set using the

Model->Constraint->Set command and give it the title A-set to
avoid confusion. Then use the Model->Constraint->Nodal
command and choose the nodes that are circled in the picture
above (Nodes:3,5,7,9,11,25,27,29,31,33,47,49,51,53,55) to
include in the A-set. After the nodes are chosen, click OK and
in the Create Nodal Constraints/DOF dialog box click the Fixed
button to include all six DOFs for each node that was selected.
 Page 89 of 255

Problem #2: Modal Analysis

of a Flat Plate
Use these results for comparison:
Mode 1 133.1727 Hz
Mode 2 649.0864 Hz
Mode 3 822.2924 Hz
Mode 4 2054.124 Hz
Mode 5 2295.341 Hz
Mode 6 2360.133 Hz
Mode 7 3764.084 Hz
Mode 8 4432.393 Hz
Mode 9 4836.664 Hz
Mode 10 5560.032 Hz
 Page 90 of 255

Rigid Body Modes

NX Nastran Dynamic Analysis

 Page 91 of 255

Rigid Body Modes

A structure has the ability to displace without developing internal
loads or stresses if it is not sufficiently grounded.
Examples:
(a) No

(b) Partial

(c) Mechanism

P
In cases (a) and (b), the structure can displace as a rigid body.
 Page 92 of 255

Rigid Body Modes

The presence of rigid body and/or mechanism modes is evidenced
by zero frequency values in the solution of the eigenvalue problem.

[K]{} = [M] {}

On the assumption that the mass matrix [M] is positive definite, zero
eigenvalues result from a positive semi-definite stiffness, i.e.:

{}RIGT[M]{RIG} > 0
{}RIGT[K]{RIG} = 0

SUPORT does not constrain the structure. It simply defines the R-

Set components. In normal modes analysis, rigid body modes are
calculated using the R-Set as reference degrees of freedom.
 Page 93 of 255

Calculation of Rigid Body Modes

If R-Set is present, rigid body modes are calculated in NX Nastran
by the following method:
Step 1: a-set partitioning
ul
{ua} =
ur
Step 2: Solve for ul in terms of ur.

Kll Krl ul 0
=
Krr Krl ur Pr
Note: Pr is not actually applied!!!

{ul} = [Dm] {ur}

Where [Dm] = -Kll-1Klr
Dm
[RIG] =
Ir
 Page 94 of 255

Calculation of Rigid Body Modes

Step 3: Mass matrix operations
T
Dm D
[Mr] = [Maa] m
Ir Ir
where [Mr] is not diagonal in general

Using Gram-Schmidt orthogonalization (in the READ module), the

matrix [Mr] is orthogonalized by the transformation [ro], that is:

T
[Mo] = [ro][Mr][ro]
 Page 95 of 255

Calculation of Rigid Body Modes

Step 4: Rigid body mode construction

Dmro
[a]RIG =
ro

with the property:

[a]RIGT[Kaa][a]RIG = Krr 0

[a]RIGT[Maa][a]RIG = [Mo]
 Page 96 of 255

Selection of SUPORT
Degrees of Freedom
Care must be taken when selecting SUPORT DOFs.
SUPORT DOFs must be able to displace independently without
developing internal stresses (statically determinate)

2 5 2 5
1 4 1

3 6 3 4

Bad Selection for SUPORT Good Selection for SUPORT

(The independent displacement of 1 and 4
may produce internal stress)
 Page 97 of 255

Selection of SUPORT
Degrees of Freedom
In NX Nastran for FEMAP, the
SUPORT degrees of freedom
can be entered by creating a
constraint set with the desired
nodes and degrees of freedom
(much like the way the A-set was
created for Guyan Reduction in
the last section), then selecting
that constraint set from the
Kinematic (SUPORT) drop down
menu in the Boundary
Conditions dialog box of the
Analysis Set Manager
 Page 98 of 255

Checking of SUPORT
Degrees of Freedom
NX Nastran calculates internal strain-energy (work) for each rigid
body vector.

Kll Krl D
[X] = [DTl] = =
Krr Krl I

[X] = DTKllD + Krr

Rigid Body Vectors
Strain Energy Matrix, Diagonals Printed

If actual rigid body modes exist, the strain-energy is 0

Note that [X] is also the transformation of the stiffness matrix [Kaa] to
R-Set coordinates, which by definition of rigid body (zero frequency)
vector properties, should be null.
 Page 99 of 255

Checking of SUPORT
Degrees of Freedom
NX Nastran also calculates the rigid body error ratio

[X]
=
Krr
Where means Euclidian norm of the matrix

= xij2
i j

Only one value of is calculated using [X] and [Krr] based on all
SUPORT DOFs
 Page 100 of 255

Checking of SUPORT
Degrees of Freedom
Except for round-off errors, the rigid body error ratio and the strain
energy should be zero if a compatible set of statically determinate
supports are chose by the user. The quantities may be non-zero for
any of the following reasons:
Round-off error accumulation
The ur set is over-determined leading to redundant supports (high energy
strain).
The ur set is underspecified leading to a singular reduced stiffness matrix
(high rigid body error ratio).
The multipoint constraints are incompatible (high strain energy and high rigid
body error ratio)
There are too many single point constraints (high strain energy and high
rigid body error ratio)
Krr is null (unit value for rigid body error but low strain energy). This is an
acceptable condition and may occur when generalized dynamic reduction is
used.
 Page 101 of 255

Rigid Body Modes

In NX Nastran, flexible body modes associated with the A-Set mass
and stiffness matrices are calculated. The first N modes calculated
by the eigenanalysis (where N is the number of DOFs in the R-Set)
are discarded. The N rigid body modes are substituted in their
place.
RIG
{ua} = [aFLEX] =
FLEX

Note: NX Nastran does not check that discarded modes are rigid body modes
(i.e., =0)
 Page 102 of 255

Rigid Body Modes

When this transformation is applied to the dynamic system and
modes are the unit mass normalized, we obtain:

IRIG 0 RIG RIG 0 0 RIG

+ [TB] +
0 FLEX
2
0 IFLEX FLEX FLEX FLEX

RIGTP RIGT
= + {N + Q}
FLEX TP FLEX T
 Page 103 of 255

Rigid Body Modes

As a result of the transformation, the following consequences occur:
Constraint forces are not externally active, i.e.,
RIGT
{Q} = {0}
FLEXT
If damping elements are not connected to ground, then:
[RIGT][B] = [0]

Thus,
RIGT 0 0
[B][RIG FLEX] =
0 FLEX [B] FLEX
T
FLEX T

If damping is proportional, then:

RIGT 0 0
[B][RIG FLEX] =
FLEX T 0 2ii
 Page 104 of 255

Dynamic Matrix Assembly

NX Nastran Dynamic Analysis

 Page 105 of 255

Dynamic Matrix Assembly

NX Nastran provides direct and modal methods for
performing transient and frequency response and
complex mode analysis.

The dynamic matrices are assembled differently

depending on the analysis and method.
 Page 106 of 255

Damping
Damping represents energy dissipation observed in structures

Damping is difficult to accurately model since damping results from

many mechanisms:
Viscous effects (shock absorber, dashpot)
External friction (slippage in structural joints)
Internal friction (characteristic of material type)
Structural nonlinearities (plasticity)
Analytical conveniences used to model damping
Viscous damping force Structural damping force

fv = bu fs = igku where i = -1

b = viscous damping coefficient g = structural damping coefficient

m + bu
+ ku = p m + (1 + ig)ku = p
 Page 107 of 255

Structural Damping versus

Viscous Damping
Assume sinusoidal response:
u = eit

Then u = ieit = -2eit

Viscous damping:
m + bu + ku = p(t)
m(-2eit)+ b(ieit)+ keit = p(t)
-2meit + ibeit+ keit = p(t)

Structural damping:
m + (1 +ig)ku = p(t)
m(-2eit)+ (1+ ig)keit = p(t)
-2meit + igkeit + keit = p(t)
 Page 108 of 255

Structural Damping versus

Viscous Damping
Both equations are identical if:
gk
gk = b b=
Therefore, if structural damping g is to be modeled using viscous damping
b, then the equality holds at only one frequency 3 (or 4)
gk
b=

if
k
= n =
m
gk
b= = gmn
n
but
bc = 2mn
 Page 109 of 255

Structural Damping versus

Viscous Damping
then
b gmn g
= = =
bc 2mn 2

= critical damping ratio (percent critical damping)

1
g= = structural damping ratio
Q
Q = quality factor or magnification factor
 Page 110 of 255

Structural Damping versus

Viscous Damping
(Constant Displacement)
Viscous Damping,
= bu
fv = bu Structural Damping
fs = gku

Damping

Equivalent Viscous Damping

b = gk/3(or 4)

3 (or 4)
Viscous and structural damping are equivalent at frequency 3 (or 4)
 Page 111 of 255

Damping Summary
Viscous damping force proportional to velocity

Structural damping force proportional to displacement

Critical damping ratio b/bc

Quality factor Q inversely proportional to energy dissipated per cycle

of vibration

At resonance =~ n
= g/2
Q = 1/(2)
Q = 1/g
 Page 112 of 255

Damping in NX Nastran for FEMAP

Damping Direct Modal Direct Modal

Type Transient Transient Frequency Frequency

Viscous Force Force Force Force

b b b b
Element Velocity Velocity Velocity Velocity
kge kge
Material b b GE GE
4 4

Overall kg kg
Structural b b PARAM,G PARAM,G
3 3

Modal N/A TABDMP N/A TABDMP

 Page 113 of 255

Damping Input
Structural Damping
MATi Bulk Data Entries
1 2 3 4 5 6 7 8 9 10
MAT1 MID E G NU RHO A TREF GE

PARAM,G,factor (default = 0.0)

Overall structural damping coefficient to multiply entire system stiffness matrix
PARAM,W3,factor (default = 0.0)
Converts overall structural damping to equivalent viscous damping
PARAM,W4,factor (default = 0.0)
Converts element structural damping to equivalent viscous damping
Units for W3,W4 in NX Nastran for FEMAP are in Hertz (Hz)

If PARAM,G is used, PARAM,W3 must be given a setting greater

than zero; otherwise, PARAM,G is ignored in transient response
analysis
 Page 114 of 255

Damping Input
Setting the PARAMs for G, 3, and 4 in NX Nastran for FEMAP is
accomplished using the Model->Load->Dynamic Analysis
command. The Load Set Options for Dynamic Analysis dialog box:

4
 Page 115 of 255

Damping Input
Modal Damping can be set using the Model->Function command.
Choose the type of damping desired from the Type drop-down menu
in the Function Definition dialog box:

Damping
function types:

6..Structural vs.
Freq

7..Critical Damp
vs. Freq

8..Q Damping
vs. Freq
 Page 116 of 255

Damping Input
Scalar viscous damping

CDAMP1 Scalar damper between two

DOFs with reference to a
property entry

CVISC Element damper between

two grid points; references a
property entry (PVISC)

Damping values are assigned through

spring properties
 Page 117 of 255

Damping Input
DOF Spring Property creates CDAMP1
 Page 118 of 255

Damping Input
SPRING Property creates PVISC Property card

Also, used to create the PBUSH property card for Nastran CBUSH elements
 Page 119 of 255

Direct Methods
The general dynamic equation used in the direct method is:
[Mddp2 + Bddp + Kdd] {ud} = {Pd}
where p = a derivative operator
ud = the union of the analysis set ua and extra points ue
For frequency response and complex eigenvalue analysis, the
dynamic matrices are:

[Kdd] = (1 + ig)[K1dd] + [K2dd] + i[K4dd]

[Bdd] = [B1dd] + [B2dd]
[Mdd] = [M1dd] + [M2dd]

For transient response, the dynamic matrices are:

[Kdd] = [K1dd] + [K2dd]

[Bdd] = [B1dd] + [B2dd] + g [K1dd] + 1 [K4dd]
3 4
[Mdd] = [M1dd] + [M2dd]
 Page 120 of 255

Dynamic Matrix Definitions

[K1dd] is the reduced structural stiffness matrix plus the reduced direct input
K2GG (symmetric).
[K2dd] is the reduced direct input matrix K2PP plus the reduced transfer
function input (symmetric or unsymmetric).
[K4dd] is the reduced structural damping matrix obtained by multiplying the
stiffness matrix [Ke] of an individual structural element by an element
damping factor ge and combining results for all structural elements
(symmetric).
[B1dd] is the reduced viscous damping matrix plus the reduced direct input
B2GG (symmetric).
[B2dd] is the reduced direct input matrix B2PP plus the reduced transfer
function input (symmetric or unsymmetric).
[M1dd] is the reduced mass matrix plus the reduced direct input M2GG
(symmetric).
[M2dd] is the reduced direct input matrix M2PP plus the reduced transfer
function input (symmetric or unsymmetric).
g, 3, 4 are the constants specified by the user.
 Page 121 of 255

Modal Methods
The general dynamic equation used in the modal method is:
[Mhhp2 + Bhhp + Khh] {uh} = {Ph}
where p = a derivative operator
uh = the union of the modal coordinates I and extra points ue
The transformation between I and ua is:
{ua} = [ai]{i}
where [ai] is the matrix of eigenvectors obtained in real eigenvalue analysis
The transformation from uh to ud is obtained by augmenting [ai] to
include the extra points.
{ud} = [dh] {uh}
ai 0
where [dh] =
0 Iee

i
{uh} =
ue
 Page 122 of 255

Modal Methods
For frequency response and complex eigenvalue analysis, the
dynamic matrices are:
[Khh] = [ki] + [dh]T(ig[K1dd] + [K2dd] + i[K4dd]) [dh]
[Bhh] = [bi] + [dh]T([B1dd] + [B2dd]) [dh]
[Mhh] = [mi] + [dh]T[M2dd][dh]

where [mi] = a diagonal matrix with terms mii = [ai]T[Maa][ai]

[bi] = a diagonal matrix with terms bii = ig(i)mii is the radian
frequency of the i-th normal mode and g(i) is a damping factor
obtained from interpolation of a user-supplied table (TABDMP1)
[ki] = a diagonal matrix with terms kii = i2mii
If parameter
KDAMP = -1, then
mii = mii
bii = 0
kii = (1 + ig(i))kii
g(i) is a damping factor obtained from the interpolation of a user-supplied
table (TABDMP1)
 Page 123 of 255

Modal Methods
[mi], [bi], and [ki] are expanded by the addition of zeros to the rows
and columns corresponding to the extra points (ue).

For transient response the dynamic matrices are:

[Khh] = [ki] + [dh]T([K2dd][dh]

g 1 1 4
[Bhh] = [bi] + [dh]T([B1dd] + [B2dd] + [K dd] + [K dd]) [dh]
3 4

[Mhh] = [mi] + [dh]T[B2dd][dh]

If only [mi], [bi], and [ki] are present in any modal dynamic analysis,
then the modal dynamic equations are uncoupled.
 Page 124 of 255

Transient Response Analysis

NX Nastran Dynamic Analysis

 Page 125 of 255

Introduction to Transient
Response Analysis
Compute response to time-varying excitation.
Excitation is explicitly defined in the time domain. All of
the applied forces are known at each instant in time.
Computed response usually includes nodal
displacements and accelerations, and element forces
and stresses.
Two categories of analysis direct and modal
 Page 126 of 255

Direct Transient Response

Dynamic equation of motion

[M] {(t)} + [B] {u(t)} + [K] u(t) = {p(t)}

Response solved at discrete times with fixed t

Using central finite difference representation for {u(t)} and {(t)} at
discrete times

1
{u n} = 2t {un+1 un-1}

1
{n} = {u 2un + un-1}
t2 n+1
Note: These equations are also used by NX Nastran to compute velocity and
acceleration output.
 Page 127 of 255

Direct Transient Response

Numerical integration (Newmark Beta type method)
(except smear force over 3 adjacent time points)

m b k
(un+1 2un + un-1) + (un+1 un-1) (un+1 + un + un-1)
t2 2t 3
1
= (Pn+1 Pn + Pn-1)
3
un+1 + un + un-1
3 Time p(t)
pn+1 + pn + pn-1 Average
3
t t

Alternate methods: Wilson , Hughes , Bathe

 Page 128 of 255

Direct Transient Response

Solution

[A1] {un+1} = [A2] + [A3]{un} + [A4]{un-1}

Where [A1] = [M/t2 + B/2t + K/3] Dynamic Matrix

[A2] = 1/3 {Pn+1 + Pn + Pn-1} Applied Force
[A3] = [2M/t2 - K/3] Initial Conditions, from
[A4] = [-M/t2 + B/2t - K/3] previous Time Step

Except that {P(t)} is averaged over three time points and [K] is modified such
that the dynamics equation of motion reduces to a [K]{un} = {Pn} if no [M] or [B]
 Page 129 of 255

Direct Transient Response

Solve by decomposing A1 and applying it to the right-hand side of the
above equation.
Similar to classical Newmark-Beta
method, except:
Applied force is averaged over 3 adjacent time points.
The stiffness K is similarly averaged to get the correct static solution for
M,B = 0.
M,B, and K do not change with time.
A1 needs to decomposed only once if t is unchanged throughout the
entire solution. If t is changed, A1 must be re-decomposed (which
may increase run times dramatically).
The output time interval may be greater than the solution time interval
(example: use solution t of 0.001 second and output results every
fifth time step or with output t of 0.005 second).
 Page 130 of 255

Damping in Modal Transient Response

If damping matrix B exists, then the assumption is made that it is not
diagonalized by :
TB diagonal

The coupled problem is solved using modal coordinates utilizing the

direct transient response Newmark-Beta type numerical integration.

[A1] {n+1} = [A2] + [A3]{n} + [A4]{n-1}

where
[A1] = []T[M/t2 + B/2t + K/3][] Dynamic Matrix
[A2] = 1/3[]T{Pn+1 + Pn + Pn-1} Applied Force
[A3] = []T[2M/t2 - K/3][] Initial Conditions, from
[A4] = []T[-M/t2 + B/2t - K/3][] previous Time Step
 Page 131 of 255

Damping in Modal Transient Response

If modal damping is used, then each mode has damping bi.

The equations of motion become uncoupled

mi + bi +ki = pi(t)
 Page 132 of 255

Damping in Modal Transient Response

Use Duhamels integral to solve for modal response as decoupled
SDOF systems.
Duhamels integral:

(t) = e-bt/2m o cos dt + o + (b/2m)o +

d
t

e-bt/2m
1
md e
0
-bt/2mp()sin (t
d ) d

p(t)

t
d
t
 Page 133 of 255

Damping in Modal Transient Response

Most efficient to use modal damping ratios since equations are
decoupled
TABDMP1 Bulk Data entry defines the modal damping ratios.
Example: For 10% Critical Damping

g = 0.2 (6)

CRIT = 0.10 (7)

Q = 5.0 (8)

These are the

three options for
modal damping
 Page 134 of 255

Data Recovery in Modal

Transient Response
Recover physical response as the summation of the modal
responses

u = [] {}

Not as large a computational penalty for changing t in modal

transient response as in direct. However, the constant t is still
recommended.

The output interval may be greater than the solution time interval
 Page 135 of 255

Mode Truncation
May not need all of the computed modes. Often only the lowest few
will suffice for dynamic response calculation.
PARAM,LFREQ gives the lower limit on the frequency range of
retained modes.
PARAM,HFREQ gives the upper limit on the frequency range of
retained modes.
PARAM,LMODES gives the number of the lowest modes to be
retained.
Truncating high-frequency modes truncates high-frequency response.
 Page 136 of 255

Mode Truncation
Mode truncation PARAMs can be entered in FEMAP using Model->
Load->Dynamic Analysis command. The Load Set Options for
Dynamic Analysis dialog box will appear (Modal Transient or Modal
Frequency must be selected as the Solution Method):
Solution Methods
(make Response
Based on Modes
fields active)

PARAM,LMODES
PARAM,LFREQ
PARAM,HFREQ
 Page 137 of 255

Transient Excitation
Define force as a function of time using the Model->Function
command and choosing 1..vs Time as the Type

Choose 1..vs Time from

the Type drop-down menu

Enter values for the Excitation in the Y-field and values for Time in the X-field.
Also, a linear ramp or other equation can be used to create values automatically.
Once the values are input, Click OK.

Note: Tabular information in an Excel spreadsheet (2 columns maximum) or

comma separated table can be pasted into FEMAP from the clipboard to create a
function using the Get button
 Page 138 of 255

Transient Excitation
Create a functionally dependent load representing dynamic excitation
by using the Model->Load->(Nodal or Elemental) command. The
Create Loads on (Nodes or Elements) dialog will appear:

Enter a unit value in the

direction of the dynamic
excitation and choose a
loading function from the
Function Dependence drop-
down menu.
 Page 139 of 255

Transient Excitation Considerations

The 1/3 smearing of applied loads must be taken into account. This
will smooth the force and decrease apparent frequency content.
Avoid discontinuous forces. These may cause different results on
different computers.

Force A-
A

B
C+
C

Time
If Nt causes a solution at ABC, then NX Nastran should select the average force B.

However, due to numerical round-off, Nt on one computer may be at time A- and will
give force A. On another computer, Nt may be at time C+ and will give force C.

The integration results will differ depending on whether the force at Nt is A, B, or C.

 Page 140 of 255

Transient Excitation Considerations

Smooth a discontinuous force over one t

Force

= original force
= smoothed force

Time
 Page 141 of 255

Initial Conditions
May impose initial displacement and/or velocity in direct transient
response via the TIC Bulk Data entry. Initial Conditions are not
available in standard modal transient.
The IC Case Control Command selects the TIC entry.
Attention initial conditions for unspecified DOFs are set to Zero.
Initial conditions may be specified only for A-set DOFs.
Initial conditions may be specified only in direct transient response. In
modal transient response, all initial conditions are set to zero.
Initial conditions are used to determine the values of {u0}, {u-1}, {P0},
and {P-1} used in calculating {u1}. The acceleration for all points is
assumed to be zero for time, t 0 (constant velocity).
0}t
{u-1} = {u0} {u
0}
{P-1} = [K]{u-1} [B]{u
The load specified by the user at t = 0 is replaced by:
{P0} = [K]{u0} [B]{u 0}
 Page 142 of 255

Initial Conditions
The recommended practice for any type of dynamic excitation is to
use at least one time step of zero excitation prior to applying the
dynamic force

Force

Time
 Page 143 of 255

TSTEP Entry
Select integration time step for direct and modal transient response.
Integration errors increase with increasing natural frequency.
Recommended t is to use at least eight solution time steps per
period (cycle) of response.
The TSTEP Bulk Data entry controls solution and output t, and is
selected by the TSTEP Case Control command.

The cost of integration is directly proportional to the number of time

steps when t is constant.

Use adequate length of time to properly capture long-period (low-

frequency) response.
 Page 144 of 255

TSTEP Entry
Creating the TSTEP entry in FEMAP is accomplished using Model->
Load->Dynamic Analysis command. The Load Set Options for
Dynamic Analysis dialog box will appear:
Portion of dialog box
for creating the Ni,
DTi, and NOi,
entries on the
TSTEP Bulk data
card.
Ni, Number of Time
Steps of value DTi

DTi, Time increment

NOi, Skip factor for

output. Every NOi-th
step will be saved for
output (default =1)
 Page 145 of 255

Problem #3

Direct Transient Response

 Page 146 of 255

Problem #3: Direct Transient Response

For this problem, use the direct method to determine the transient response of the flat
rectangular plate, created in problem #1, subject to a time-varying excitation. The
structure is excited by a 1 psi pressure load over the total surface of the plate
varying at 250 Hz. It is also excited with a 50 lb force in the lower right corner (node
11) which is also varying at 250 Hz, but 180 out-of-phase with the pressure load.
Both time-dependent dynamic loads are applied for a duration of 0.008 seconds only.
Use structural damping of g = 0.06 and convert this damping to equivalent viscous
damping at 250 Hz. Continue the analysis to 0.04 seconds.

1 psi over entire surface

50.00 lb
 Page 147 of 255

Problem #3: Direct Transient Response

T3 Translation for Node 11 over complete duration of analysis
 Page 148 of 255

Problem #3: Direct Transient Response

Use these results for comparison:
Time 0.0024 T3 Displacement
Node 11 -0.26051 in
Node 33 -0.287 in
Node 55 -0.31076 in

Time 0.0052 T3 Displacement

Node 11 0.27895 in
Node 33 0.31823 in
Node 55 0.3528 in

Time 0.02 T3 Displacement

Node 11 0.038693 in
Node 33 0.038876 in
Node 55 0.038945 in
 Page 149 of 255

Problem #4

Modal Transient Response

 Page 150 of 255

Problem #4: Modal Transient Response

For this problem, use the direct method to determine the transient response of the flat
rectangular plate, created in problem #1, subject to a time-varying excitation. The
structure is excited by a 1 psi pressure load over the total surface of the plate
varying at 250 Hz. It is also excited with a 25 lb force in the lower right corner (node
11) which is also varying at 250 Hz, but starting 0.004 seconds after the pressure
load. Both time-dependent dynamic loads are applied for a duration of 0.008
seconds only. Use a modal damping table of = 0.03 for all the modes. Continue
the analysis to 0.04 seconds.

1 psi over entire surface

25.00 lb
 Page 151 of 255

Problem #4: Modal Transient Response

T3 Translation for Node 11 over complete duration of analysis
 Page 152 of 255

Problem #4: Modal Transient Response

Use these results for comparison:
Time .0068 T3 Displacement
Node 11 0.14765 in
Node 33 0.16062 in
Node 55 0.17344 in

Time .0092 T3 Displacement

Node 11 -0.16902 in
Node 33 -0.18492 in
Node 55 -0.20055 in

Time .022 T3 Displacement

Node 11 -0.038703 in
Node 33 -0.03752 in
Node 55 -0.036237 in
 Page 153 of 255

Frequency Response Analysis

NX Nastran Dynamic Analysis

 Page 154 of 255

Introduction to Frequency
Response Analysis
Compute response to oscillatory excitation.
Excitation is explicitly defined in the frequency domain.
All of the applied forces are known at each forcing
frequency.
Computed response usually includes nodal
displacements and element forces and stresses.
The computed responses are the complex numbers
defined as magnitude and phase (with respect to forcing)
or as real and imaginary components
Two categories of analysis direct and modal
 Page 155 of 255

Direct Frequency Response

Dynamic equation of motion

[-2M + iB + K] {u()} = {P()}

PARAM,G and GE on MATi entry do not form a damping matrix. They

form a complex stiffness matrix
K = (1+ iG)K1 + iGEkE
where K1 = global stiffness matrix
G = overall structural damping coefficient (PARAM,G)
kE = element stiffness matrix
GE = element structural damping coefficient (GE on MATi entry)

Solve the equation by inserting to form a complex left-hand side, and

then solve it similar to a Statics problem (using complex arithmetic)
 Page 156 of 255

Modal Frequency Response

Convert to modal coordinates and solve as decoupled SDOF
systems
Pi
i =
-mi+ ibi+ ki

Much quicker to solve this equation than in direct method

Decoupled procedure can be used only if either no damping is

present or if modal damping alone (via TABDMP1) is used.
Otherwise, use the less efficient direct approach (on smaller modal
coordinate matrices) if non-modal damping (VISC,DAMP) is present.
 Page 157 of 255

Excitation Definition
Define force as a function of time using the Model->Function
command and choosing 3..vs Frequency as the Type

Choose 3..vs Frequency

from the Type drop-down
menu

Enter values for the Excitation in the Y-field and values for Frequency in the X-
field. Also, a linear ramp or other equation can be used to create values
automatically. Once the values are input, Click OK.
 Page 158 of 255

Excitation Definition
Create a functionally dependent load representing dynamic excitation
by using the Model->Load->(Nodal or Elemental) command. The
Create Loads on (Nodes or Elements) dialog will appear:

Enter a unit value in the

direction of the dynamic
excitation and choose a
loading function from the
Function Dependence drop-
down menu.
 Page 159 of 255

Frequency Response Considerations

Exciting an undamped (or modal damped) system at 0.0 Hz gives the same
results as a static analysis. Therefore, if the maximum excitation frequency is
much less than the lowest resonant frequency of the system, a static analysis
is sufficient.
Very lightly damped structures exhibit large dynamic responses for excitation
frequencies near resonant frequencies. A small change in the model (or
running it on another computer) may give large changes in such response.
Use a fine enough frequency step size (f) to adequately predict peak
response. Use at least 5 points per half-power bandwidth.

Peak
Response
Peak/ 2 = Half power point

f1 f2 Frequency
Half-Power
For maximum efficiency, use an uneven frequency step size. Use smaller f
in regions of resonant frequencies and larger f in regions removed from
resonant frequencies
 Page 160 of 255

Solution Frequencies
Define a table of solution frequencies using the Model->Function
command. Then use the Model-> Load->Dynamic Analysis command
to choose the Modal Frequency Table. The Load Set Options for
Dynamic Analysis dialog box will appear:
If a normal modes
analysis has already
been performed and the
results read into
FEMAP, then Modal
Frequency Table can be
created automatically
from those modal
results by pressing the
Modal Freqbutton.
Some additional factors
need to be specified to
create the table but in
general the defaults in
FEMAP are usable.
 Page 161 of 255

Problem #5

Direct Frequency Response

 Page 162 of 255

Problem #5: Direct Frequency Response

For this problem, use the direct method to determine the frequency response of the
flat rectangular plate, created in problem #1, subject to a frequency-varying excitation.
The structure is excited by a unit load (1.0) at the lower right corner (node 11). Use a
frequency step (f) of 20 Hz between a range of 20 and 1000Hz. Use Structural
damping g = 0.06.

1.00
 Page 163 of 255

Problem #5: Direct Frequency Response

T3 Translation for Node 11 (Magnitude) over complete duration of analysis (Rectilinear graph)

D
i
s
p
l
a
c
e
m
e
n
t

Frequency
 Page 164 of 255

Problem #5: Direct Frequency Response

T3 Translation for Node 11 (Magnitude) over complete duration of analysis (Y-axis log scale)

D
i
s
p
l
a
c
e
m
e
n
t

Frequency
 Page 165 of 255

Problem #5: Direct Frequency Response

T3 Translation for Node 11 (Phase) over complete duration of analysis (Rectilinear graph)

D
i
s
p
l
a
c
e
m
e
n
t

Frequency
 Page 166 of 255

Problem #5: Direct Frequency Response

T3 Translation for Node 33 (Magnitude) over complete duration of analysis (Y-axis log scale)

D
i
s
p
l
a
c
e
m
e
n
t

Frequency
 Page 167 of 255

Problem #5: Direct Frequency Response

T3 Translation for Node 33 (Phase) over complete duration of analysis (Rectilinear graph)

D
i
s
p
l
a
c
e
m
e
n
t

Frequency
 Page 168 of 255

Problem #5: Direct Frequency Response

T3 Translation for Node 55 (Magnitude) over complete duration of analysis (Y-axis log scale)

D
i
s
p
l
a
c
e
m
e
n
t

Frequency
 Page 169 of 255

Problem #5: Direct Frequency Response

T3 Translation for Node 55 (Phase) over complete duration of analysis (Rectilinear graph)

D
i
s
p
l
a
c
e
m
e
n
t

Frequency
 Page 170 of 255

Problem #5: Direct Frequency Response

Use these results for comparison (Magnitude):
Frequency 140 T3 Displacement
Node 11 0.067867 in
Node 33 0.068504 in
Node 55 0.068888 in

Frequency 480 T3 Displacement

Node 11 0.00076559 in
Node 33 0.0003570 in
Node 55 0.0013882 in

Frequency 680 T3 Displacement

Node 11 0.0078558 in
Node 33 0.00026242 in
Node 55 0.0074482 in
 Page 171 of 255

Problem #5: Direct Frequency Response

Use these results for comparison (Phase):
Frequency 140 T3 Displacement
Node 11 212.265
Node 33 211.938
Node 55 211.655

Frequency 480 T3 Displacement

Node 11 348.688
Node 33 184.776
Node 55 176.419

Frequency 680 T3 Displacement

Node 11 296.914
Node 33 338.782
Node 55 113.105
 Page 172 of 255

Problem #6

Modal Frequency Response

 Page 173 of 255

Problem #6: Modal Frequency Response

For this problem, use the modal method to determine the frequency response of the
flat rectangular plate, created in problem #1, subject to a 0.1 psi pressure load over
the entire surface and a unit load (1.0lb) at the lower right corner (node 11) lagging
45. Use a modal damping of = 0.03. Use a frequency step (f) of 20 Hz
between a range of 20 and 1000Hz. In addition, specify five evenly spaced
excitation frequencies between the half-power points of each resonant
frequency between the range of 20 1000 Hz (Modal Frequency table). Run modal
analysis first to determine resonant frequencies between 20 -1000 Hz.

0.1 psi over the entire surface

1.00
 Page 174 of 255

Problem #6: Modal Frequency Response

T3 Translation for Node 11 (Magnitude) over complete duration of analysis (Y-axis log scale)

D
i
s
p
l
a
c
e
m
e
n
t

Frequency
 Page 175 of 255

Problem #6: Modal Frequency Response

T3 Translation for Node 11 (Phase) over complete duration of analysis (Rectilinear graph)

D
i
s
p
l
a
c
e
m
e
n
t

Frequency
 Page 176 of 255

Problem #6: Modal Frequency Response

T3 Translation for Node 33 (Magnitude) over complete duration of analysis (Y-axis log scale)

D
i
s
p
l
a
c
e
m
e
n
t

Frequency
 Page 177 of 255

Problem #6: Modal Frequency Response

T3 Translation for Node 33 (Phase) over complete duration of analysis (Rectilinear graph)

D
i
s
p
l
a
c
e
m
e
n
t

Frequency
 Page 178 of 255

Problem #6: Modal Frequency Response

T3 Translation for Node 55 (Magnitude) over complete duration of analysis (Y-axis log scale)

D
i
s
p
l
a
c
e
m
e
n
t

Frequency
 Page 179 of 255

Problem #6: Modal Frequency Response

T3 Translation for Node 55 (Phase) over complete duration of analysis (Rectilinear graph)

D
i
s
p
l
a
c
e
m
e
n
t

Frequency
 Page 180 of 255

Problem #6: Modal Frequency Response

Use these results for comparison (Magnitude):
Frequency 180 T3 Displacement
Node 11 0.011813 in
Node 33 0.012332 in
Node 55 0.012829 in

Frequency 440 T3 Displacement

Node 11 0.00034835 in
Node 33 0.00083596 in
Node 55 0.0016477 in

Frequency 720 T3 Displacement

Node 11 0.0045846 in
Node 33 0.00036933 in
Node 55 0.0045729 in
 Page 181 of 255

Problem #6: Modal Frequency Response

Use these results for comparison (Phase):
Frequency 180 T3 Displacement
Node 11 154.131
Node 33 153.283
Node 55 152.48

Frequency 440 T3 Displacement

Node 11 233.878
Node 33 155.416
Node 55 143.213

Frequency 720 T3 Displacement

Node 11 174.215
Node 33 260.31
Node 55 345.321
 Page 182 of 255

Enforced Motion

NX Nastran Dynamic Analysis

 Page 183 of 255

Introduction to Enforced Motion

Used to analyze constrained structures with base input acceleration,
velocity, and displacement.

Common examples include earthquakes (for transient analysis) and

swept-sine shaker test simulation (for frequency response analysis).

For many years, no automatic method existed in Nastran for

applying displacements, velocities, or accelerations to the base. At
that time, there was a need to convert applied forces on equivalent
unconstrained structure to enforced motion of constrained structure.
Now accelerations, velocities, and displacements can be directly
applied to the base node in what is known as the Direct Method.
The Direct Method will be explained in a later chapter.

Several methods of Enforced Motion exist: large mass and large

stiffness.
 Page 184 of 255

Enforced Motion in Transient Response

(Large Mass and Large Stiffness methods)
The utilization of the Large Mass or Large Stiffness methods was
necessary to specify acceleration, velocity, or displacement for
dynamic analysis that involved enforced

For the indirect method, Nastran can only apply forces {P(t)} to the
structure. If enforced motion is selected, for this section, it is
assumed that the user is imposing the motion on a large mass.
Therefore, the force to move the large mass is proportional to
acceleration.
F = m
or
F
=
m
 Page 185 of 255

Using Large Mass Method in

Transient Response

Large Mass
Structure
m of Interest
M = (103-108)m

Rigid Body

For enforced acceleration b

1
b P ML = large mass
ML
Larger mass should be 103 to 108 times the Structure of Interest
mass
 Page 186 of 255

Using Large Stiffness Method in

Transient Response
u Structure
m of Interest
Large Spring

For enforced motion displacement ub

1
ub = K P KL = large spring
L

Apply P to spring/structure interface to obtain desired ub. Spring KL

is entered on spring entries. Scale factor KL is entered on force
entries.

The large spring should be 102 to 104 times c2m where c is the
cutoff frequency of the excitation.
 Page 187 of 255

Using Large Stiffness Method in

Transient Response
Advantages in special cases: Avoids round-off errors in
differentiation and avoids rigid body drift when enforced motion is
applied to statically indeterminate points.

Disadvantages in all cases: difficult to estimate a good value for large

stiffness, and the required modes (in a modal formulation) are high
frequency ones that are not likely to be included in retained modes.
 Page 188 of 255

Enforced Motion in Frequency Response

(Large Mass and Large Stiffness methods)
The utilization of the Large Mass or Large Stiffness methods was
necessary to specify acceleration, velocity, or displacement for
dynamic analysis that involved enforced motion.

For these methods, Nastran can only apply forces {P(t)} to the
structure. If enforced motion is selected, for this section, it is
assumed that the user is imposing the motion on a large mass
unless otherwise stated. Therefore, the force to move the large
mass is proportional to acceleration.

The easiest method to use to define enforced motion with

Frequency Response is the Large Mass method
 Page 189 of 255

Using Large Mass Method in

Frequency Response
In frequency response the input and response are assumed to be
sinusoidal functions

P = P()eit

u = u()eit

resulting in a simplified dynamic equation of motion:

[-2M + iB + K] {u()} = {P()}

 Page 190 of 255

Using Large Mass Method in

Frequency Response
As in transient response, the force required to move a large mass
is:
P = ma
but the acceleration is assumed to be a sinusoidal function:

a = a()eit

Therefore, to impose an acceleration, apply:

P() = ma()
 Page 191 of 255

Using Large Mass Method in

Frequency Response
To impose displacement:

u = u()eit

A force needs to be applied that results in the desired displacement.

Differentiating the displacement twice produces an acceleration:

a() = -2u()eit
The applied force is:

P() = ma() = -m2u()

 Page 192 of 255

Using Large Stiffness Method in

Frequency Response
Using the Large Stiffness method is similar to the large mass
method except the applied force is:

P = Ku

To impose a displacement:

P() = Ku()

To impose an acceleration:
1
P() = Ka()
- 2
 Page 193 of 255

Recommendations for Enforced Motion

(Large Mass and Large Stiffness methods)
Use Large Mass method.

Large mass should be at least 103 times structure mass for accuracy,
but no more than 108 times structure mass (any higher causes
numerical errors).

Retain rigid body modes for analysis.

Be careful with units many times enforced acceleration is specified

in terms of g (acceleration constant) rather than in direct units (such
as in/sec2).

Use a small model to verify solution procedure.

 Page 194 of 255

Specifying Enforced Motion

(Large Mass Method)
Using the Model-> Load->Dynamic Analysis command to choose the
Enforced Motion button. This will guide the user through creating enforced
motion via the Large Mass method. The Load Set Options for Dynamic
Analysis dialog box will appear:
Click the Enforced
Motion button after
selected the desired
Solution Method at the
top of the dialog box.

The other parameters

can be filled in after the
enforced motion
process has been
completed.
 Page 195 of 255

Specifying Enforced Motion

(Large Mass Method)
Select Coordinates for Base Mass. Any point in 3-D space can be chosen and
existing node or point potions may be used as well. Click OK

Select Nodes on the Base. At least one node must be chosen to connect the
structure to the Base Mass. Once these nodes are chosen, FEMAP will
automatically create a rigid element from the newly created node at the Base
Mass location (independent node) to the nodes chosen to represent the base of
the structure (dependent nodes). An example of proper selection of Base
Nodes would be the points at the bottom of a High-Voltage tower that attach
the tower to the ground or its foundation. Click OK
 Page 196 of 255

Specifying Enforced Motion

(Large Mass Method)
Select acceleration or rotational acceleration from the list of available loads.
Give a value for the selected load in the forcing direction, and a forcing function.
With Large Mass Method, this
acceleration will be applied as
a force which is exciting the
base mass (a = F/M). Click
OK.

The mass size and mass factor

will be selected automatically
in the next dialog box, but can
also be altered for a
customized enforced motion
effect. Click OK
 Page 197 of 255

Enforced Motion in Transient Response

(Direct method)
While enforced motion with Nastran was accomplished for years by
either the Large Mass approach or the Lagrange Multipliers
technique, direct enforced motion capabilities are available in all
versions of NX Nastran.

Both the Large Mass and Lagrange Multipliers methods are both
theoretically valid, they can also be very cumbersome to implement.

Disadvantages:
Large Mass Method often leads to computational and numerical problems
due to round off errors and pseudo-rigid body modes.

Lagrange Multipliers Method require a version specific DMAP ALTER

command.
 Page 198 of 255

Enforced Motion in Transient Response

(Direct method)
The Direct method allows for direct specification of displacements,
velocities, or accelerations using SPC/SPC1/SPCD data, eliminating
the need to employ Large Mass or Lagrange Multipliers.

NX Nastran directly utilizes this enforced motion information in the

equations of motion, partitioning and integrating them (i.e., transient
analysis) in accordance with the type of motion specified.

Direct enforced motion is available in direct and modal frequency

analysis (Solutions 108 and 111), direct and modal transient
analysis (Solutions 109 and 112), and design optimization (Solution
200)

Field 3 of the TLOADi and RLOADi cards has been changed from
DAREA to EXCITEID.
 Page 199 of 255

Direct Method in Frequency Response

In frequency response the transient effects are assumed to be
negligible, and the time-dependent nature of the loading can be
expressed in terms of harmonic forcing functions.

Force response of the structure to these harmonic loads occurs at

the same frequency, and in proportion to the magnitude of the
applied loads.

When an enforced motion is applied instead of a harmonic force, the

effect is similar, creating a response with proportional forces of
constraint at the same frequency as that of the enforcing motion.

Any one of the enforced displacement, velocity, or acceleration must

uniquely determine the other two (differ only by multiples of
frequency), with resultant forces of constraint derived from a solution
of governing equations.
 Page 200 of 255

Direct Method in Frequency Response

To illustrate this, an applied harmonic forcing function of the form:

P(t) = P()eit
will lead to the in-plane displacement:

u(t) = U()eit
With the corresponding velocity and acceleration:

= iU()eit
u(t)
and

(t) = -2U()eit

yielding the equations of frequency response.

 Page 201 of 255

Direct Method in Frequency Response

Going into further detail, the frequency response equations are
written after multipoint constraint partitioning operations have been
performed leaving just the free (f-set) and constrained (s-set)
degrees of freedom:

Mff Mfs Bff Bfs Kff Kfs Uf Pf

-2 + i + = Eq. 9-1
Msf Mss Bsf Bss Ksf Kss Us Ps + qs
Where Ps are the external loads applied to the S-Set and qs are the corresponding
forces of constraint.

If the constraints specify zero motion (Us = {0}), the solution for the free degrees-of-
freedom may be obtained directly from the upper part of this equation:
(-2Mff + iBff + Kff) Uf = Pf Eq. 9-2
And the corresponding constraint forces from the lower part:

qs = -Ps + (-2Msf + iBsf + Ksf) Uf Eq. 9-3

 Page 202 of 255

Direct Method in Frequency Response

If enforced displacements, velocities, or accelerations are applied, Us {0} and the
free degrees-of-freedom, from Eq. 9-1 are:

(-2Mff + iBff + Kff) Uf = Pf - (-2Msf + iBsf + Ksf) Us Eq. 9-4

which corresponds with constraint forces:

qs = -Ps + (-2Msf + iBsf + Ksf) Uf + (-2Mss + iBss + Kss) Us Eq. 9-5

Comparing 9-4 and 9-5 with 9-2 and 9-3 shows that the enforced motion modifies
the force applied to the f-set degrees-of-freedom.

Had an enforced velocity been applied instead, the resultant applied displacement
would differ by a factor of (1/(i)).

Had it been an applied acceleration, it would differ by (1/2).

 Page 203 of 255

Direct Method in Transient Response

When the transient dynamic equations of motion are written in terms
of displacement and its higher order derivatives, the solution of
these equations must be performed in a stepwise integral fashion for
very time step, t.

As with Frequency response, any enforced displacement, velocity,

or acceleration must uniquely define the other two quantities for that
degree-of-freedom with resultant forces of constraint derived from a
solution of governing equations of motion at that particular time
step(s) of interest.

Unlike Frequency response, higher-order displacement derivatives

for the enforced degrees-of-freedom must be determined by finite
difference, with lower-order quantities (for example, displacements
and velocities for an applied acceleration) determined by numerical
integration.
 Page 204 of 255

Direct Method in Frequency Response

Going into further detail, the transient dynamic equations of motion
are written after multipoint constraint partitioning operations have
been performed leaving just the free (f-set) and constrained (s-set)
degrees of freedom:

Mff Mfs f B B u f K K uf Pf(t)

+ ff fs + ff fs = Eq. 9-6
Msf Mss s Bsf Bss u s Ksf Kss us Ps(t) + qs(t)

In the case of zero constrained motion, us = us = us = {0} and the solution for the free
degrees-of-freedom may be obtained directly from this equation:

Mff f + Bff u f + Kff uf = Pf(t) Eq. 9-7

with corresponding forces of constraint from:

f + Ksf uf)
qs(t) = -Ps(t) + (Msf f + Bsf u Eq. 9-8
 Page 205 of 255

Direct Method in Frequency Response

If enforced displacements, velocities, or accelerations are applied, us, u s, s {0}, and
the solution for the free degrees-of-freedom from Eq. 9-6, are:

Mff f + Bff u f + Kff uf = Ps(t) + (Msf s + Bsf u

s + Ksf us) Eq. 9-9

with constraint forces:

f u u
qs(t) = -Ps(t) + [Msf Mss] + [Bsf Bss] f + [Ksf Kss] f Eq. 9-10
s u s us

As with frequency response, the effect of enforced motion is to modify the loads on
the f-set, and the s-set forces of constraint
 Page 206 of 255

Problem #7

Direct Transient Response with

Enforced Acceleration
 Page 207 of 255

Problem #7: Direct Transient Response

with Enforced Acceleration
For this problem, use the direct method to determine the transient response of the flat
rectangular plate, created in problem #1, subject to a unit acceleration sine pulse of
200 HZ applied to the base (node 23) in the z-direction. A large mass of 1000 lb is
applied to the base. Use a structural damping coefficient of g = 0.05 and convert this
damping to equivalent viscous damping at 200 Hz.

Large Mass
1000 lbs
 Page 208 of 255

Problem #7: Displacement Results

T3 Translation for Node 23 (Base node) over duration of analysis

D
i
s
p
l
a
c
e
m
e
n
t

Time
 Page 209 of 255

Problem #7: Displacement Results

T3 Translation for Node 33 over complete duration of analysis

D
i
s
p
l
a
c
e
m
e
n
t

Time
 Page 210 of 255

Problem #7: Velocity Results

T3 Velocity for Node 23 (Base node) over duration of analysis

Velocity

Time
 Page 211 of 255

Problem #7: Velocity Results

T3 Velocity for Node 33 over complete duration of analysis

Velocity

Time
 Page 212 of 255

Problem #7: Acceleration Results

T3 Acceleration for Node 23 (Base node) over duration of analysis

A
c
c
e
l
e
r
a
t
i
o
n

Time
 Page 213 of 255

Problem #7: Acceleration Results

T3 Acceleration for Node 33 over complete duration of analysis

A
c
c
e
l
e
r
a
t
i
o
n

Time
 Page 214 of 255

Problem #7: Direct Transient with

Enforced Motion
Use these results for comparison (Node 23):
Time T3 Displacement
0 0.0 in
0.02 2.525E-6 in
0.04 2.525E-6 in

Time T3 Velocity
0 0.0
0.02 -1.35554E-7
0.04 -7.25978E-8

Time T3 Acceleration
0 0.103
0.02 0.00016233
0.04 -0.00013677
 Page 215 of 255

Problem #7: Direct Transient with

Enforced Motion
Use these results for comparison (Node 33):
Time T3 Displacement
0 0.0 in
0.02 4.59E-6 in
0.04 7.1984E-7 in

Time T3 Velocity
0 0.0
0.02 0.0012328
0.04 0.00065997

Time T3 Acceleration
0 -0.0043797
0.02 -1.47664
0.04 1.24428
 Page 216 of 255

Problem #7a

Direct Transient Response with

Enforced Acceleration
(Direct application of acceleration)
 Page 217 of 255

Problem #7a: Direct Transient Response

with Enforced Acceleration (Direct)
For this problem, use the direct method to determine the transient response of the flat
rectangular plate, created in problem #1, subject to a unit acceleration sine pulse of
250 HZ applied to the base (node 23) in the z-direction. No large mass is not required
for this example. Use a structural damping coefficient of g = 0.06 and convert this
damping to equivalent viscous damping at 250 Hz. Compare with results of Ex.7.
 Page 218 of 255

Random Response Analysis

NX Nastran Dynamic Analysis

 Page 219 of 255

Classification of Dynamic
Environments
Deterministic Random

Periodic Transient Stationary Nonstationary

Simple Shock Ergodic

Harmonic Spectra
 Page 220 of 255

Introduction to Random Response

Random Vibration is vibration that can be described only in a
statistical sense. Its instantaneous magnitude at any time is not
known; rather, the probability of its magnitude exceeding a certain
value is given.

Examples include earthquake ground motion, ocean wave heights

and frequencies, wind pressure fluctuations on aircraft and tall
buildings, and acoustic excitation due to rocket and jet engine noise.

NX Nastran performs random response as post-processing to

frequency response. Inputs include the output from frequency
response, as well as, user-supplied loading conditions in the form of
auto- and cross-spectral densities. Outputs are response power
spectral densities (PSDs), autocorrelation functions, number of zero
crossings with positive slope per unit time, and the RMS values of
response.

Reference: Random Vibration in Mechanical Systems, by S. H.

Crandall and W. D. Mark, Academic Press, 1963
 Page 221 of 255

Introduction to Random Response

There are several conventions used to define random analysis
quantities. Care must be taken to use NX Nastran random response
capabilities properly.

NX Nastran random analysis assumes ergodic processes, which

means the excitations are stationary with respect to time.

The concepts of autocorrelation, auto-spectrum (power spectrum),

cross-correlation, and cross-spectrum must be defined.

The mean square value and apparent frequency are the key
statistical quantities to be gotten from Random Response.
 Page 222 of 255

Autocorrelation and
Autospectrum
Autocorrelation function:

T
Rj() = lim
1
u (t)uj(t - )d
T T 0 j

Rj(o) is the mean-square value of

Autospectrum function:

T 2
Sj() = lim
2
T T 0
uj(t)eitd
 Page 223 of 255

Autocorrelation and
Autospectrum
Mean square value:

uj(t)2 = Rj(o) =
1
S ()d
2 0 j

Apparent frequency N0 (zero crossings):

0
(/2)2Sj()d
N 02 =
0 Sj()d
 Page 224 of 255

Calculation of Linear System Response

to Ergodic Random Excitation
Uj()

Fa()
From frequency response analysis:
uj() = Hja() Fa()
where Hja() is the frequency response or transfer function relating output uj
to input Fa.
If there are several inputs, then:

uj() = Hja()Fa() + Hjb()Fb() +

 Page 225 of 255

Definition of Multiple Input-Output Spectral

Relationship for a Linear System
In matrix form:
Fa()
uj() = [ Hja() Hjb() ] Fb()

The output autospectrum is:
Fa() H*ja()
Sujuj = T[ Hja Hjb ] Fb() [F*a() F*b() ] H*jb()

The individual input spectra are:
TFa() F*a() = Saa()
TFa() F*b() = Sab()
TFb() F*b() = Sbb()
 Page 226 of 255

Definition of Multiple Input-Output Spectral

Relationship for a Linear System
The multiple input-output spectral relationship is therefore:

Saa Sab
Sujuj() = [ Hj]T Sba Sbb [ H*j]

where [Hj]T = [ Hja Hjb ]

H*ja
[H*j] = H*jb

 Page 227 of 255

Definition of Multiple Input-Output Spectral

Relationship for a Linear System
The input cross-spectral matrix is:

Saa() Sab()
[S]IN = [ Hj]T Sba() Sbb()

It has the special properties:

Sab() = S*ab()

Saa(), Sbb() = real 0

 Page 228 of 255

Definition of Multiple Input-Output Spectral

Relationship for a Linear System
Commonly used special cases:
Single input analysis (fully correlated inputs)

Sujuj() = Hja() 2Saa()

Uncorrelated multiple inputs

Sujuj() = Hja() 2Saa() + Hjb() 2Sbb() +

 Page 229 of 255

Random Analysis as
Implemented in NX Nastran
It is assumed that the output from the frequency response
calculations is Hja(). It does not calculate:

Hja() = uj()/Fa()

If Hja() is desired, use F() = 1.0.

 Page 230 of 255

Random Analysis as
Implemented in NX Nastran
Use the Model->Load->Dynamic Analysis command, the Load Set
Options for Dynamic Analysis dialog box will appear:

The Power Spectral

Density data should be
created in functional
form and then selected
from the PSD drop-down
menu in the Random
Analysis Options portion
of the dialog box.
 Page 231 of 255

Random Analysis as
Implemented in NX Nastran
Use the Model->Analysis command to create a Random Response
analysis set using the Analysis Set Manager :

Choose 6..Random Response

as the analysis type.

Determine what type of output is desired for Nodes and/or Elements.

Choose Power Spectral Density Functions or Autocorrelation Functions
or Both.
 Page 232 of 255

Random Response XY Output

Once the type of Output is determined, Choose existing group of
the specific Output requests are made nodes to retrieve requested
in this dialog box. output. (If Nodal results have
been requested)

Choose existing group of

elements to retrieve requested
output. (If Elemental results
have been requested)
 Page 233 of 255

Correlation Table
Only available when there are multiple subcases with multiple Load Sets. Choose
real and imaginary component factors of specific PSD functions to create a
correlation table between the Excited Load Set and the Applied Load Set. This
Dialog box can only be found by in the tree structure of the Analysis Set Manager

An Example of when this might be useful is when correlating between PSD load sets
from four wheels of a vehicle riding on a rough surface.
 Page 234 of 255

Random Analysis as
Implemented in NX Nastran
Many times, no other
Output needs to be
requested other than the
PSD functions and the
Autocorrelation Functions
for Random Response
Analysis.
 Page 235 of 255

Random Analysis
Recommendations
Most spectra are given as a log function. Use equation
features on the function form if the PSD is given in log scale.

Always generate the output PSD at the input location if

possible

Plot the output PSD. Do not use the summery output blindly.

Use several frequencies in the vicinity of each mode.

For low frequencies (<20 Hz), use many frequencies since the
displacement spectra is changing rapidly for a constant input
acceleration.
 Page 236 of 255

Problem #7a

Direct Transient Response with

Enforced Acceleration
(Direct application of acceleration)
 Page 237 of 255

Problem #7a: Direct Transient Response

with Enforced Acceleration (Direct)
For this problem, use the direct method to determine the transient response of the flat
rectangular plate, created in problem #1, subject to a unit acceleration sine pulse of
250 HZ applied to the base (node 23) in the z-direction. No large mass is not required
for this example. Use a structural damping coefficient of g = 0.06 and convert this
damping to equivalent viscous damping at 250 Hz. Compare with results of Ex.7.
 Page 238 of 255

Shock and Response

Spectrum
NX Nastran Dynamic Analysis
 Page 239 of 255

Response Spectrum
Response spectrum depicts the maximum response of a SDOF
system as a function of its resonant frequency for base
excitation.
This Graph:

Response

Resonant Frequency (Hz)

is generated from:
x(t)
FN = 0.0 12 2.0 FMAX

UB(t)

Point on Larger, vibrating structure

Response computed for Transient Analysis
 Page 240 of 255

Response Spectrum
The peak of each SDOF oscillator is calculated from its X(t). The
oscillator base motion UB is derived from the force or base
excitation applied to a larger structure.

An example: An earthquake time history is applied to a power plant.

Response Spectra are calculated at the locations of the floors to be
used in the design of components (machinery, piping systems, etc.)

An implicit assumption is that the oscillators mass is very small

relative to the larger, vibrating mass. Therefore, no dynamic
interaction occurs between the two. (Consequently, the response
spectrum analysis is decoupled from the transient analysis).
 Page 241 of 255

Response Spectrum
Analysis is repeated for several damping values to generate a
family of curves.

1
1 damping = 0% critical
Response 2 2 damping = 3% critical
3
3 damping = 5% critical

Resonant Frequency (Hz)

Damping applies to each oscillator, not the vibrating structure

 Page 242 of 255

Response Spectrum
Maximum displacement response from X(t) is calculated for each
oscillator. The maximum relative displacement between each
oscillator and its base (a point on the vibrating structure) is also
computed.
X = maximum inertial (absolute)
Xr = maximum relative

Relative velocity and absolute acceleration are approximately

related to the relative displacements by

= X
Xr r
= Xr
X 2

, and X.
For design, useful variables are Xr, X Design spectra are
r
usually in terms of these variables.
 Page 243 of 255

Response Spectrum
Response Spectra may be generated in any transient solution
(SOLs 109, 112).

The transient response for selected DOFs in model is used as the

input time history for the generation of the response spectra curves.
 Page 244 of 255

Response Spectrum Generation

Define Functions using Model->function command

Function to Define
Oscillation Frequencies

Function to Define
Oscillation Dampings
 Page 245 of 255

Response Spectrum Generation

Use the Model->Load->Dynamic Analysis command, the Load Set
Options for Dynamic Analysis dialog box will appear:

Function to Define
Oscillation Frequencies

Function to Define
Oscillation Dampings
 Page 246 of 255

Response Spectrum Generation

Response Spectrum Specific dialog boxes using the Analysis set
manager, Modal->Analysis command.

Define Analysis Type as

5..Response Spectrum

Request NASTRAN Output

for Response Spectrum Choose existing group of
Analysis nodes to retrieve requested
output.
 Page 247 of 255

Applying Spectra
Available in Solution 103:

Poor Mans Transient. The input spectra are used to determine

the peak response of each mode.

These peak modal responses are combined to obtain the system

response (timing of each modes peak is not known).

Three methods of combining the modal responses are available

(ABS, SRSS, NRL)
 Page 248 of 255

Applying Spectra
Procedure:

A model of the structure to be analyzed is created with the input

points identified as SUPORT DOFs.

A large mass (usually 103 to 106 times the structural mass) is

attached to the SUPORT DOFs.

System modes are obtained for the modal (including 0.0 Hz modes)
with the SUPORT DOFs unconstrained.

This approximates the cantilevered modes of the model attached

to the exciting structure.
 Page 249 of 255

Applying Spectra
The 0.0 Hz modes (Dm) approximate the static motion the model
experiences when the supporting structure moves statically.

Participation Factors (PF) are calculated using the following

expression:
= TMDm

PF is used in conjunction with the spectra described as shown in the

input section to calculate the peak response for each mode.

Data recovery quantities (displacements, stresses, forces, etc.) are

then calculated for each mode based on its peak motion.

These quantities are then combined for the modes using the
selected method (ABS,SRSS,NRL, NRLO) and the results are
printed.
 Page 250 of 255

Applying Spectra
Xr response of a single DOF oscillator due to the base motion is
calculated as follows:

+ gX + 2X = (t)
Xr r r r

The actual transient response at a physical point is

uk(t) = ikirr (i, gi, t)

i r
Absolute Value Rule (ABS) option:
~
uk = ik irri (i, gi)
i r
where ri (i, gi) = max ri (i, gi, t)

and i represents a mode

and r represents a direction
 Page 251 of 255

Applying Spectra
Square Root of the Sum of the Squares Rule (SRSS) option:

uk ~
= (iki)2
i

Where the average peak modal magnitude, i is

~
i = (irr (i, gi))2
r

U.S. Navy Shock Design Modal Summation Convention (NRL) option:

uk ~
= jkj + (iki)2
ij

where jkj is the peak magnitude

 Page 252 of 255

Applying Spectra
NRLO refers to the NRL method used in version 69 of MSC Nastran.
The NRL was updated slightly in version 70 to adhere to NAVSEA-
0908-LP-000-3010 specification.

The ABS rule is the most conservative it assumes that the modal
responses all achieve their peak response at the same time and with
the same with the same phase (this rarely happens), therefore it
usually over-predicts the response.
 Page 253 of 255

Applying Spectra
Modes that are close in frequency may have their peak response
occur at the same time (and with the same phase). The SRSS and
NRL methods contain a provision to sum modal responses via ABS
method for modes that have closely spaced natural frequencies.
Close natural frequencies are defined by frequencies that meet the
following in equality:
fi+1< CLOSE * fi

The value of CLOSE is defined by PARAM,CLOSE (the default is

1.0).

The modal summation option is set via PARAM,OPTION (ABS is

default). Both PARAM,OPTION and PARAM, CLOSE may be set in
any subcase allowing for summation by several conventions in a
single run.
 Page 254 of 255

Response Spectrum Application

Create a Function vs. Damping function using the results of the
Shock Spectrum Generation
Use these functions to create
a Function vs. Damping
relationship for the Spectrum
Function ID.
 Page 255 of 255

Response Spectrum Application

Response Spectrum Specific dialog boxes using the Analysis set
manager, Modal->Analysis command.
Define Analysis Type as
2..Normal Modes/Eigenvalue

Choose the type of Spectrum and

a Spectrum function ID (a scale
factor can also be added)
Choose a modal combination
method (ABS, SRSS, NRL,
NRLO) and a closeness factor.
Chose a constraint set to be
the SUPORT set and
Choose a damping function
for modal damping