Linear Dynamic System Analyses with Creo Simulate

– Theory & Application Examples, Capabilities, Limitations –

9th SAXSIM 2017, TU Chemnitz, 28.03.2017

Rev. 1.2 | 07.04.2017
Dr.-Ing. Roland Jakel
 At a glance: Altran, a global leader

20+
Countries

30,000+
Innovation Makers

EUROPE ASIA AMERICAS

5 — — —
Austria, Belgium, France, China, India, Canada,
Industries Germany, Italy, Luxembourg, Malaysia, Middle East United States of America,
Netherlands, Norway, Portugal, Mexico
Spain, Sweden, Switzerland,
United Kingdom
€ 2.120bn
2016 Revenues

2
 What do we do? The industries

Our presence in the main business sectors enables us to partner with key players in
the market:

24% 25% 23% 16% 12%

AEROSPACE, DEF ENCE ENERGY, INDU STRY AU TOMO TI VE, TELECO MS F INANCIAL
& RAILWAY & LIF E SCIENCES INF RAST RU C TU RE & MEDIA SERVICES
& TRANSP O R T A T I O N & GOVERN ME N T

3
 Table of Contents

Slide:
1. Introduction to dynamic analysis theory in Creo Simulate 5-11
1.1 Basic equations for dynamic analyses 5
1.2 Solution method coded 6
1.3 Damping 7
1.4 Limitations of the solution coded in Creo Simulate 9
1.5 Result quality assurance when performing dynamic analysis 10

2. Modal analysis 12-20

2.1 Standard modal analysis 12
2.2 Modal analysis with prestress 18
2.3 Hints for application 20

3. Dynamic analysis 21-63

3.1 Classification of the supported dynamic analysis types 21
3.2 Dynamic time analysis 22
3.3 Dynamic frequency analysis 33
3.4 Random response analysis 42
3.5 Dynamic shock analysis 50
3.6 Examples for dynamic analyses of big system models 57

4. Feedback to the software developer PTC 64-67

4.1 Missing functionality & enhancement requests to PTC for dynamic analysis 64
4.2 Known issues in dynamic analysis and possible workarounds 67

5. References 69-70

4
Rev. 1.2 | 07.04.2017
 1. Introduction to dynamic analysis theory in Creo Simulate

1.1 Basic equations for dynamic analyses

Basic equation for dynamic systems

• Creo Simulate can only solve dynamic problems which can be described with help of
the following linear differential equation (DEQ) of second order:
M x C x K x  F (t )
• Herein, we have: [𝑀]=mass matrix, [𝐶]=damping matrix, [𝐾]=stiffness matrix,
{𝐹}=force vector, {𝑥}=displacement vector and its derivatives with respect to time

Modal analysis as basis for all dynamic studies

• In order to determine the fundamental frequencies of a mechanical structure, first a
modal analysis is performed before any subsequent dynamic studies are carried out
• The equation which is solved here is a special case of the above differential equation:
M x K x  0
• Hence, for fundamental frequency determination in Creo Simulate, no damping 𝐶 is
taken into account, so the real mechanical structures to be computed may only contain
little damping to keep the error small
• Damping is taken into account only during subsequent dynamic analysis like shown on
the next slide

5
Rev. 1.2 | 07.04.2017
 1. Introduction to dynamic analysis theory in Creo Simulate

1.2 Solution method coded

Solution Sequence:
• In Creo Simulate, the mentioned linear differential equation of second order,
M x C x K x  F (t )
is not solved directly in physical coordinates, but in the following way:
• Before any dynamic analysis is performed in Simulate, the damping-free modal
analysis, M x K x  0 , is carried out to obtain the modal base (eigenvector
matrix) for the modal transformation
• The system is then transformed from physical space 𝑥 to modal space 𝜉 by
replacing the physical coordinates with modal coordinates: 𝑥 = 𝜙 𝜉
• Herein, 𝜙 is the eigenvector matrix, and 𝜉 modal coordinates; 𝜙 has a number of
rows equal to the DOF in the model, and columns equal to the number of modes;
𝜉 has one column and rows equal to the number of modes
• In a subsequent dynamic analysis, in which modal damping [𝐶] = 2𝛽 [𝑀]𝜔 and a
forcing function is added, we have [𝑀], [𝐶] and [𝐾] as diagonal matrices now in modal
coordinates!
• After the solution is performed, the solution is transformed back into physical space
for post-processing
Remark: This solution method is used in many FEM codes for linear, small damped
dynamic systems because of its computational efficiency (only diagonal matrices) and
various practical advantages, e.g. different dynamic analysis types and damping values
can be rapidly executed on base of the existing modal analysis!
6
Rev. 1.2 | 07.04.2017
 1. Introduction to dynamic analysis theory in Creo Simulate

1.3 Damping

Let’s now look at the modal damping [𝐶] = 2𝛽[𝑀]𝜔 mentioned on the previous slide:
• For a simple, linear damped one-mass-oscillator (=harmonic oscillator with mass 𝑚,
velocity proportional damping constant 𝑐 and spring stiffness 𝑘), the damping ratio
𝛽 (in German “Lehrsches Dämpfungsmaß”) is
c c c
  
2m0 2 mk ccrit

• Herein, 𝑐𝑐𝑟𝑖𝑡 is the so called “critical damping”, 


 0 .0
 0 .5
leading to the aperiodic limit case 𝛽 = 1, in which the   1 .0

oscillator just does not overshoot (red curve right)

  1 .5

• The damping ratio 𝛽 is often expressed in % (like in Simulate), so we have

  100 % : very strong damping (creeping case)
  100 % : aperiodic limit case (critical damping = no overshooting)
  50 % : max. damping supported in Creo Simulate (green curve in the diagram)
  1...4 % : typical values used for many small damped, real mechanical structures
  0 %: no damping
• Remember:
The undamped and damped fundamental angular frequencies are related as follows:
k
0  2f 0  ;   0 1   2
m
7
Rev. 1.2 | 07.04.2017
 1. Introduction to dynamic analysis theory in Creo Simulate

1.3 Damping

There are many methods to measure damping, like e.g.:

• Logarithmic decrement 𝛿 (relative damping):
x1 1 x
  ln  ln 1
x2 n xn 1
• 𝛿 can simply be obtained from the decrease of amplitudes 𝑥𝑖 in an experiment
evaluated in the time domain where the structure is dying out over the time with its
natural frequency
• 𝛿 can then be transferred into 𝛽 with help of the following equation (for small 𝛽):
2
  2 xout
1  xin Q
1
2
• Another method for an experiment   0 .0
evaluated in the frequency domain Q
  0 .1
is measuring the bandwidth B (in 2   0 .2
  0 .3
German “Halbleistungsbandbreite”)
  0 .5
or the magnification factor Q of the
  1 .0
oscillator (for 𝛽 <<1):
1 f 1 B
Q  0   
2 B 2Q 2 f 0
• This is exemplarily depicted right
for 𝛽 =0.1 f
B
f0

8
Rev. 1.2 | 07.04.2017
 1. Introduction to dynamic analysis theory in Creo Simulate

1.4 Limitations of the solution coded in Creo Simulate

Limitations of the solution method

• We have only a linear system (all matrices are constant), that means no nonlinearities
can be taken into account like
 Contact
(therefore unknown force-vs-time curves of impact problems cannot be computed,
but have to be assumed and then applied as external force function vs. time [4])
 Change of constraints
(all dynamic analyses use the constraints defined in the modal analysis!)
 Nonlinear material
 Nonlinear damping (e.g. from friction, hydraulic devices,…)
• Only modal damping can be applied to keep the damping matrix diagonal and
therefore run times short (this damping is called in German language “Bequem-
lichkeitshypothese“ – “hypothesis of comfort”)
• A severe limitation is that no discrete damper, not even a linear one, is supported
(discrete linear springs are supported in dynamic analysis!)
• A discrete linear damper can only be approximated, this means those mode shapes
which are damped by a discrete damper may be taken into account with a higher,
individual modal damping
• Therefore, in Creo Simulate the damping can be applied in three ways:
 Constant for the complete frequency domain
 As function of frequency
 For individual modes

9
Rev. 1.2 | 07.04.2017
 1. Introduction to dynamic analysis theory in Creo Simulate

1.5 Result quality assurance when performing dynamic analysis

• The solution method is of approximative nature even for ideal linear structures:
 Continuum mechanical structures have an infinite number of natural modes, but for
computation only a finite number of modes can be taken into account by the FEM code
 Therefore, the modal base is cut after a certain number of modes, but in theory, the
exact solution can only be obtained by superposing all modes to the total response!
 Depending on this number of modes taken (or better not taken) into account, the
analysis results may become pretty inaccurate!
• Therefore, it is in the responsibility of the user to assure that a sufficient number of
modes is taken into account to obtain results of the required accuracy!
• There are a couple of methods how the result quality can be assured:
 Compare the results with an analytical solution (if existing!)
 Repeat the dynamic analysis with an increasing number of modes and see if the results
converge (typically done in analyses with force excitation)
 Only for analyses with base excitation: Check if the sum of the effective masses 𝑚𝑖, 𝑒𝑓𝑓
of all modes taken into account is close to the total mass of the structure
 A rule of thumb is to take into account all modes with eigen frequencies until at least
the double value of the excitation frequency, but often even this may not be sufficient
(sometimes the author had to use >4x the max. excitation frequency)
 Check that for an excitation frequency of Zero Hz, the results match the results of a
separately performed linear static analysis undertaken with the same model!
Note: There will always be a difference at the location of force introduction!
 …

10
Rev. 1.2 | 07.04.2017
 1. Introduction to dynamic analysis theory in Creo Simulate

1.5 Result quality assurance when performing dynamic analysis

Creo 2.0

Participation factors and effective masses:

Creo 3.0
• For all dynamic analyses with base point excitation, the
code allows to compute mass participation factors and
effective masses:
 The effective mass 𝑚𝑖, 𝑒𝑓𝑓 , multiplied with the base point acceleration, reflects the share
this mode has to the total base point reaction force!
 Note the effective mass of the mode depends on the excitation direction!
 The sum of the squares of the participation factors is the total absolute mass of the
structure!
• A Simulate example output for a 2-mass oscillator with a total mass of 2 kg:
(=0.002 t; analysis was performed in mm, t, s unit system!)
Mode frequency part. factor eff. mass tot. mass
---- ------------ ------------ --------- --------- m1,eff = (4.351645e-02)2 t = 1.894 kg = 94.7 %
1 5.445276e+01 4.351645e-02 94.7% 94.7% m2,eff = (1.027535e-02)2 t = 0.106 kg = 5.3 %
2 1.425516e+02 1.027535e-02 5.3% 100.0% Total mass = m1,eff + m2,eff = 2.000 kg = 100.0 %

• Note: Modes with an effective mass of Zero cannot be excited over the base points
(=the interface the structure is mounted to), but of course they may be by another
external force directly acting on certain points of the structure!
• Effective masses are therefore just of importance for base point excited structures
and not for force excited structures!
• Since Creo Simulate 3.0, the mass participation factors can also be requested in a
modal analysis (output very comfortably for all three translations and rotations!)

11
Rev. 1.2 | 07.04.2017
 2. Modal analysis

2.1 Standard modal analysis

2.1.1 Introduction
• Like mentioned in chapter 1, the modal analysis just solves
M x K x  0
• Note again damping [𝐶]{𝑥} is not taken into account in the
Simulate modal analysis, so fundamental frequencies may
appear in reality at slightly lower frequencies than predicted
• For typical damping around 1-4 %, this influence is negli-
gible, but for the max. modal damping supported in the
subsequent dynamic analysis (50 %), it may be up to 13.4%:
  0 1  0.52  0  0.866
• The user has the following choices to request modes:
 Number of modes (always starting at Zero Hz)
 Al modes in frequency range (with arbitrary min. and max.
frequency)
• The code supports constrained and unconstrained (“free-
free”) modal analysis with rigid mode search
• Displacements (mode shapes) are always output as result;
optional are stresses, rotations (for beams and shells), local
stress errors, and new in Creo 3.0 mass participation factors
• Until Creo 2.0, mass participation factors could only be
requested in subsequent dynamic analysis with base point Creo 3.0

excitation

12
Rev. 1.2 | 07.04.2017
 2. Modal analysis

2.1 Standard modal analysis

2.1.1 Introduction
• All known convergence methods are supported, with
the exception of multi-pass adaptive convergence on
measures (as supported in static analysis)
• Note the plotting grid must be set in the modal
analysis, too, and cannot be changed in the
subsequent dynamic analysis!

Mode shape output:

• Per default, the eigenvector displacements (mode
shapes) are output unit normalized (=max. disp.
magnitude scaled to 1), but the user may request
mass normalization acc. to the equation
xi T  M  xi   1
• This is sometimes advantageous since modal stress
(if requested as result) is always output for mass
normalized mode shapes, and for meaningful modal
stress evaluation it does not make sense to use
different normalizations for displacements and
stresses (see chapter 3.2)

13
Rev. 1.2 | 07.04.2017
 2. Modal analysis

2.1 Standard modal analysis

2.1.2 Example

A long, slim drive shaft

• Steel shaft length 500 mm,
diameter 13 mm
(E=190 GPa, =0.3, =7.85g/cm3)
• Simply supported
• 1st fundamental frequency:
f0  100 Hz

Analysis as
1. simple 2 p-beams model
(=much faster)
2. volume model with help of a
mapped mesh
(=better visualization of results)
• Request for mass-normalized
displacement output:

14
Rev. 1.2 | 07.04.2017
 2. Modal analysis

2.1 Standard modal analysis

2.1.2 Example

Modal analysis results:

Beam model (2 p-beams only!): Volume model (420 p-solids):
(CPU time 0.47 s) (CPU time 150.45 s)

Rigid body mode (shaft rotation)

1st bending mode

2nd bending mode

3rd bending mode

4th bending mode

axial “pumping”
5th bending mode

1st torsional
6th bending mode

7th bending mode

2nd torsional
8th bending mode

Note: Modes with same frequency, respectively, occur because of the rotational
symmetric structure (bending may appear in any lateral direction)!

15
Rev. 1.2 | 07.04.2017
 2. Modal analysis

2.1 Standard modal analysis Virtual shaft thickness increase under torque just
2.1.2 Example because of linearized theory and displacement scaling!

Mode shapes (with mass normalized displacement output!)

1st 2nd 3rd 4th

5th 1st 2nd

1st axial
torsional torsional

16
Rev. 1.2 | 07.04.2017
 2. Modal analysis

2.1 Standard modal analysis

2.1.2 Example

Modal von Mises stress (always computed for mass normalized displacement output!)

17
Rev. 1.2 | 07.04.2017
 2. Modal analysis

2.2 Modal analysis with prestress

2.2.1 Introduction

• Allows to take into account fundamental frequency

changes from preloads:
 Tensile stresses in slim structures increase the
fundamental frequency (e.g. a music instrument
string or a turbine blade under centrifugal loads)
 Compression stresses in slim structures decrease
the fundamental frequency
 Bending preloads do not significantly change
fundamental frequencies, since tensile and
compressive stress loaded regions of the structure
are balanced and compensate each other
• Basis of a modal analysis with prestress is a linear
static analysis that defines the preloaded state created
by the preload force {𝐹𝑃 }. From this preloaded state,
the stress stiffness matrix [𝐾𝜎 ] is computed for each
integration point of each element
• The modal analysis with prestress then solves the
following equation:
M x K   K Fp x  0
Note the force applied in the
previous static analysis defining the
• Note that unlike in a Creo Simulate static analysis with preloaded state can be optionally
scaled, so it does not need to be
prestress, the static prestress cannot be combined re-run if another preload is applied!
with dynamic stress in subsequent dynamic analysis
output!

18
Rev. 1.2 | 07.04.2017
 2. Modal analysis

2.2 Modal analysis with prestress

2.2.2 Relationship with other analyses taking into account preloads

• Linear buckling analysis (“Eigenvalue buckling”): It is

also based on a previous static analysis defining the
preloaded state, but solves the equation:
K    K FP x  0
  is also called the “buckling load factor” BLF
 The linear buckling analysis does not take into
account large displacements (it is assumed that the
geometry is not significantly changing under load),
so no tangential stiffness matrix 𝐾𝑇 = 𝐾 + 𝐾𝜎 + 𝐾𝐿
with 𝐾𝐿 as stiffness matrix for large displacements is
taken into account [5]
• Static analysis with prestress:
K   K FP x  F 
Like the modal analysis with prestress, this analysis
takes into account weakening or stiffening effects from
preloads!
 Both preload analyses may fail with the misleading
error message “insufficiently constrained”, if the
applied preloads are above the critical buckling load!
 Static analysis with prestress outputs raw stresses
and no superconverged stresses [9], so unlike in all
other analyses in the postprocessor you can smooth
these stresses (unsmoothed raw stress output allows
to check for meshing quality)
19
Rev. 1.2 | 07.04.2017
 2. Modal analysis

2.3 Hints for application

• Mass normalization can be requested via config.pro-

option “sim_massnorm_modes” or with the engine
command line option “-massnorm”
• With the “plotting grid” setting the RAM and hard disk resource
consumptions can be highly influenced: E.g. for volume structures
with no interest in a detailed stress computation in a later dynamic
analysis, a plotting grid of 2 is sufficient!
(for beam models, a high plotting grid up to 10 should be used!)
• Also note requesting modal stress in the modal analysis needs a lot of RAM, so huge
system models running into memory limits in the modal analysis may successfully
run with deactivating the modal stress request
• If a static analysis fails with the error message “insufficiently constrained”, the rigid
mode search in a constrained modal analysis can be successfully used for detecting
the under constrained part or degree of freedom of the model (pretty useful for big
system models with many parts/subassemblies).
Remark: If the modal analysis fails with the same error message even though, the
reason is usually a free rotating point of the structure (e.g. a spring end point)!
• A modal analysis may also be used for checking mechanism modes, so for assuring a
correct force flow in a static analysis!
• Measurement points should be defined before meshing (e.g. use hard point Auto
GEM control or directly define the measure before meshing of the modal analysis
model) – otherwise, the mesh cannot be reused in the subsequent dynamic analysis
and the modal analysis has to be performed again!
20
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.1 Classification of the supported dynamic analysis types

• Depending on the forcing function {F} on the right side of the differential equation (DEQ), four
different linear dynamic analysis types are supported in Creo Simulate:

Periodic system (steady state) Transient system

Exact Dynamic frequency analysis: Dynamic time analysis: *)
excitation Only harmonic excitation 𝐹 = 𝐹 cos 𝜔𝑡 + 𝜑 - The excitation function is accurately known and
function neglecting transient effects - is present; the force is applied as function of time; the solution
results are
is done in
𝜔
applied as function of frequency (𝑓 = ) and all the
evaluated
time domain
in the time domain
(deter- 2𝜋
ministic) results are evaluated in the frequency domain

Stochastic
Non- Random response analysis: Dynamic shock analysis: **)
excitation
deter- The forcing function is obtained by statistical means The excitation force typically
is typically
is a short random
ministic (usually with help of measurements) and applied excitation (e.g. earthquake, pyrotechnic shock), for
typically as acceleration density function which an SRS (shock response spectrum) has to be
excitation (=acceleration density vs. frequency) computed as input for the FEM analysis. The
analysis just offers an image of the “worst case”
state with low computational effort!

*) Of course, in a dynamic time analysis also a harmonic excitation function can be applied, but
unlike in dynamic frequency analysis, which just regards the particular solution of the DEQ,
also the transient state will then be computed (homogeneous solution) before the steady state
is reached
**) Strictly speaking, a dynamic shock analysis can be performed for any type of excitation for
which an SRS can be obtained (e.g. also deterministic functions like half sine shocks, impulse
functions, even harmonic excitation), but often it is performed for the mentioned transient
examples

21
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.2 Dynamic time analysis

3.2.1 Introduction External file
data import
• The dynamic time analysis is the
most universal and most simple to
understand dynamic analysis in
Creo Simulate, but also the
computationally most intensive
• If other dynamic analyses fail for
certain coding limitations, it is a
good idea to try this analysis type!
• Nearly any arbitrary force vs. time
function can be applied, either
Selection between base point
 in form of any analytic excitation or external force
function like e.g. excitation (=“load functions”)
𝐹 = 𝐹 cos 𝜔𝑡 + 𝜑
(for all programmable
functions, see right)
 or as a tabular function (e.g.
as file input from a given Default time-dependent function
force-vs.-time for dynamic time analysis is the
impulse function (“Dirac-impact”
measurement) of infinite short impact duration)
• Alternatively, any base point
(interface) acceleration can be
defined in a similar way

22
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.2 Dynamic time analysis

3.2.2 Examples

A half sine base point acceleration shock

• We want to apply a half sine wave shock
with 50 g peak acceleration and 10 ms
1 1
duration (𝑓 = = = 50 𝐻𝑧)
𝑇 20 𝑚𝑠
• The required analytic expression takes
advantage of the “if”-function shown right
• Correct coding can be checked by
graphical visualization of the function

A base point step (jump) function

• We want to apply a 1 g step function for
time t=0
• Very simple to code like shown right
• Always note the constant scale factor and
units in the analysis definition dialogue!

A harmonic sine excitation

• We want to apply a force of 10 N with
f=100 Hz, starting with 0 N at t=0

23
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.2 Dynamic time analysis

3.2.2 Examples

A long, slim drive shaft with unbalance operated at

its resonance frequency (100.3 Hz)
• Example model of chapter 2.1.2
• Rotational speed 6000 rpm (100 Hz)
• Modal damping 2 %
• Static unbalance mass u=1 gram at unbalance
radius 6.5 mm in the middle of the shaft (no
dynamic unbalance assumed in the example, but
this would be simple to simulate, too)

Questions of interest:
• How big is the displacement (shaft bending)
under this operating condition?
• How big is the max. shaft acceleration due to
vibration created by the unbalance?
• How long does it take until the shaft swings up?
• Are the stresses in the shaft still low enough so G  e 
U ur  mm 
    balancing quality 
that it can be safely operated even though it is m m  s 
running in resonance? U = unbalance u.r [g mm]
u = unbalance mass
• Is the foreseen balancing quality G sufficient, e.g. r = unbalance radius of u
to obtain the required swinging velocity? m = total mass of the rotor
e = excentricity of m
 = angular velocity
24
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.2 Dynamic time analysis

3.2.2 Examples

Before we run the dynamic time analysis, let’s see what we

can do with some simple estimations and the information
we already have:
• Unbalance force: 𝐹𝑢 = 𝑢 ∙ 𝑟 ∙ 𝜔2 = 2.566 N
• Static deformation under this force: 0.025 mm
(obtained in a simple, linear static analysis shown right
or with help of a formulary)
• Magnification factor for the given damping:
Q=1/2=25
• Therefore the expected deformation at 6000 rpm
(very close to the resonance peak) will be:
 0.625 mm
• From the modal analysis of chapter 2.1.1 we know that
the modal von Mises stress is 3019 MPa for a mass
normalized displacement of 61.95 mm
• Since with the central unbalance force at 100 Hz we
predominantly excite just the first mode, we can
estimate the real stress by scaling the modal stress
from mode 1 to the estimated displacement of
0.625 mm by rule of three: We obtain approx.
30 MPa in resonance from this!
x MPa 0.625 mm
  x  30 MPa
3019 MPa 61.95 mm
25
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.2 Dynamic time analysis

3.2.2 Examples

Setting up the dynamic time analysis

• The 100 Hz rotating unbalance is defined by
using a sine and a cosine forcing function as
time dependent functions for the two unit
forces applied under 90 °, respectively

𝐹𝑢 = 𝑢 ∙ 𝑟 ∙ 𝜔2 Cosine function for 100 Hz

26
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.2 Dynamic time analysis

3.2.2 Examples

Displacement animation [mm] of the swing up process (scale 100:1, 0.5 s duration)

max. displacement 0,61 mm

27
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.2 Dynamic time analysis

3.2.2 Examples

Von Mises stress animation [MPa] of the swing up process (scale 100:1, 0.5 s duration)

max. von Mises stress 29.62 MPa

28
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.2 Dynamic time analysis

3.2.2 Examples
Measures for displacement magnitude [mm] (left) and Y-acceleration [g] (right)

29
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.2 Dynamic time analysis

3.2.2 Examples

Using the dynamic analysis output functionality

• Dynamic analysis can typically be computed
with automatic output or user defined output
steps
 For the first case, only measures are output
and the code automatically assures a
suitable (time or frequency) stepping
 Since there are no system default measures
for dynamic analysis, the user always has
to define measures before a dynamic
analysis with automatic steps (better
already before the modal analysis)
 For user-defined output, the user can
select stepping and request full results
(=colorful PP images) for all or only those
steps of certain interest
• Often it is a good idea to run the first analysis
with automatic output (and meaningful
measures!) and then run a second analysis with
user defined output for further evaluation
• We will subsequently explain this at the shaft
example

30
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.2 Dynamic time analysis

3.2.2 Examples

Animating only the particular solution of the DEQ for

the unbalanced shaft in a dynamic time analysis
• The measurement output for e.g. the Y-
acceleration shows that the system has practically
reached its steady state at the end of the
computed time span of 0.5 s
 So we will request full output now just for the
last period of the analysis, this can then be
repeatedly displayed in the postprocessor

31
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.2 Dynamic time analysis

3.2.2 Examples

Von Mises stress animation [MPa] of the steady state (scale 100:1, one period of 0.01 s)

32
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis The independent variable in dynamic
frequency analysis is the frequency

3.3 Dynamic frequency analysis

3.3.1 Introduction

• The dynamic frequency analysis computes the

structure’s responses to pure harmonic
excitation (cosine/sine function with one
frequency at the same time) and disregards any
homogeneous solution (transient oscillation);
just the steady state (particular solution) is taken
into account
Default amplitude
• So for a given excitation frequency 𝑓 = 𝜔/2𝜋, the function for dynamic
exciting force (if force excitation is present) has time analysis is the
uniform function (no
the following form: change of amplitude vs.
frequency)
F (t )  Fmax cos(t   )
• The amplitude 𝐹𝑚𝑎𝑥 and phase 𝜑 of the excitation
vs. frequency function can be input by analytic
functions or as tabular input (in analogy to the
tools used in dynamic time analysis)
• Like in dynamic time analysis, in addition to load
functions also base excitation is supported
• A typical application is a sine sweep test with a
very low sweep rate (strictly speaking: an
infinitely low sweep rate!)
Default phase function is Zero

33
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.3 Dynamic frequency analysis

3.3.2 Examples

0.5 g constant base point acceleration in X

between 10 and 2000 Hz (sine sweep)
• Very simple to define, see right
• Note the possibility to request output
relative to “ground” or “supports”-option
for base point excitation!

10 mm frequency independent (constant)

base point displacement amplitude
• We have to define a frequency dependent
acceleration function for this as shown
below
x  xmax sin(t )
x  xmax cos(t )
x  xmax 2 ( sin(t ))

Important to activate for non-constant tabular

input only, otherwise the automatic frequency
stepping generator misses those frequencies!

34
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.3 Dynamic frequency analysis

3.3.2 Examples

A long, slim drive shaft with unbalance

operated at its resonance frequency
(100.3 Hz)
• Example model of chapters 2.1.2
and 3.2.2:
 Nominal rotational speed
6000 rpm (100 Hz)
 Modal damping 2 %
 Static unbalance mass
u=1 gram at an unbalance
radius of 6.5 mm in the
middle of the shaft
• In addition, we also want to know
the response of the shaft vs.
frequency between 0 and 2000 Hz
• Necessary analysis definition is
shown on the right side
• Additional analysis with full output
request at the fundamental
frequencies for results animation
within the postprocessor

35
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.3 Dynamic frequency analysis

3.3.2 Examples

Frequency response curve for displacement magnitude of the central shaft point
•

First bending mode at 103 Hz: Third bending mode at 896 Hz

0.62 mm (same result as in
dynamic time analysis!)
No magnification at the second (400 Hz) and fourth
(1584 Hz) bending mode, since the excitation force is
applied to a nodal point of these modes!
Note: The frequency stepping generator automatically
refines stepping around the resonances!

36
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.3 Dynamic frequency analysis

3.3.2 Examples

Von Mises stress animation [MPa] for 100 Hz (scale 100:1)

•

Comparison of elapsed times:

Dynamic frequency analysis requesting full output for 7 frequencies: 5.71 s
Dynamic time analysis with full output just for the steady state at 100 Hz: 224 s
37 Dynamic time analysis for the complete swinging up process at 100 Hz: 664 s
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.3 Dynamic frequency analysis

3.3.2 Examples

Von Mises stress animation [MPa] for 896 Hz (scale 100:1)

•

38
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.3 Dynamic frequency analysis

3.3.2 Examples

Von Mises stress animation [MPa] for 2454 Hz (scale 50:1)

•

39
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.3 Dynamic frequency analysis

3.3.3 Remarks for application

Note SPR 2875703 when evaluating dynamic frequency results in the postprocessor:
• Unfortunately, it may happen that the phase result is not taken into account correctly
in the postprocessor. In latest tests the following occurred:
 Displacement, velocity and acceleration magnitude fringe plots just give correct results
when animated, but wrong when not!
 Displacement, velocity and acceleration component fringe plots just give correct results
when not animated (but never the “Max. Disp.” value display as shown below)
• Be also careful when evaluating other results (phases…)

Wrong fringe
color coding
Correct colors
during animation

Phase not taken into

account when vector sum Correct colors
of Y- and Z-displacement during animation
is computed

40
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.3 Dynamic frequency analysis

3.3.3 Remarks for application

Note SPR 2875703 when evaluating dynamic frequency results in the postprocessor:
• Furthermore, to obtain a “smooth” animation, use as many frames n as possible
(>20), since the PP erroneously divides the deformed shape by n-1 (=first and last
frame have identical shape; this error can be reduced by using many frames!)
• In general, do not activate deformed shape without activating animation for dynamic
frequency analysis results evaluation (this results in a meaningless shape just taking
into account the amplitude maximum in positive coordinate direction, but no phase)

Frame 1
Frame 2

Frame 3 Frame 4 =
Frame 1

41
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.4 Random response analysis

3.4.1 Introduction

• Until now, we have just looked at excitations where the excitation of the structure
can be predicted for any time t with help of a deterministic mathematical function
• In reality, we have also excitations for which an accurate input prediction in a
deterministic sense cannot be done. Examples for this may be
 Jet or rocket engine noise

𝑥
 Turbulent fluid flow
t
 Ground acceleration during an earthquake
• Such excitations are described with statistical methods, and usually a lot of
measurements of the excitation in the time domain must be done, evaluated and
edited in order to obtain a reliable acceleration or force spectral density function
 For more details, refer to suitable technical literature
• With this random response spectrum finally on hand, the code computes in this
analysis type the answer of the mechanical structure when subjected to this
excitation
[6]
• In probably >90% of the application cases this will
be an acceleration spectral density function introduced
into the structure’s base points, exemplarily shown right
(note also force excitation is supported in Simulate)
• In the following chapter we will therefore exemplarily
describe how this works with help of a simple one-mass-
resonator, for which an analytical solution exists

42
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.4 Random response analysis

3.4.2 Examples

Base point excitation of a one-mass-oscillator with white noise:

• Given is a white noise acceleration density spectrum with a constant acceleration
density of Win=0.2 g2/Hz for 1-2000 Hz
• The effective input acceleration is the square root below the acceleration density
curve, also called RMS- (root mean square) or 1-acceleration:
g in , RMS  0.2(2000  1)  19.995 g RMS

• With help of the Miles formula (valid for ideal white noise
with an infinite frequency span from 0 Hz to  Hz!)

g out , RMS   Q  f 0  Win
2
we obtain for a resonator with Q=25 (=2 %) and a
fundamental frequency of 700 Hz the 1-output acceleration:
 g2
g out , RMS   25  700 Hz  0.2  74.15 g RMS
2 Hz
• If the momentary values of the output are Gauss distributed, the expected peak value
is usually defined as the 3 value, which means that only 0.3 % of the momentary
values of the output acceleration are greater than 222,5 gRMS, and that for 99,7 % of
the time the acceleration is below!
• This 3-value is typically used for estimation of the max. resonator acceleration and
therefore for evaluating the risk of forced rupture, even though 0.3 % of the peak
values are not covered: Practical experience shows this usually works fine!
43
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.4 Random response analysis

3.4.2 Examples

Base point excitation of a one-mass-oscillator

with white noise:
• Oscillator properties: 𝑚 = 4 𝑘𝑔, 𝐾 = 77377.7 𝑁/𝑚𝑚
• Fundamental undamped frequency:
1 𝐾
𝑓0 = = 700 Hz
2𝜋 𝑚

• With 2% damping: 𝑓 = 𝑓0 1 − 𝛽2 = 699.86 𝐻𝑧

• Note: The bandwidth B for harmonic excitation (see

chapter 1.3) and the “effective bandwidth” 𝐵𝑒𝑓𝑓 for
white noise excitation are related as follows:
  f0
Beff  B
2 2 Q
• So we obtain for harmonic excitation 𝐵 = 28 𝐻𝑧 and for
white noise excitation 𝐵𝑒𝑓𝑓 = 44 𝐻𝑧
1 f 1 B
Q  0   
2 B 2Q 2 f 0

44
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis
Measures to
3.4 Random response analysis compute the
Measures to
3.4.2 Examples compute the effective
acceleration acceleration
of resonator
Measure definitions distribution
and base
function of
resonator and (gRMS-value,
Measures to output in the
base (for
compute the engine
PP evaluation)
acceleration report-file)
density function
of resonator
and base (for
PP evaluation)

45
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.4 Random response analysis gRMS measure output in the

3.4.2 Examples engine report:
196150.9 mm/s2 = 19.995 g
Fringe plot and report file results 739829.1 mm/s2 = 75.4158 g

Fringe plot
of the local
gRMS-value
distribution

46
Rev. 1.2 | 07.04.2017
 Refined output
3. Dynamic Analysis with manual
stepping

3.4 Random response analysis

3.4.2 Examples

Measure results for automatic frequency stepping

acceleration density
function p of the base

acceleration distribution
function P of the base

acceleration density acceleration distribution

function p of the resonator function P of the resonator

47
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis
Bandwidth B: The power to
maintain the oscillation
3.4 Random response analysis grows with the square of the
3.4.2 Examples amplitude and is at the
border of the bandwidth
approximately half of the
Bandwidth results in comparison to harmonic excitation max. resonance power

Harmonic excitation with 1 g input Random excitation

1
Q  25
2

Q
 17,7
2

f0
B  2 f0   28 Hz
Q 
Beff  B  44 Hz
2

48
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.4 Random response analysis

3.4.3 Remarks for application

• Always define suitable measures at a base point to

check if the given input acceleration density was
correctly applied!
• Take care of output settings (relative to ground or
supports)!
Note: Output relative to supports means that the
results are expressed in what an observer sitting on
a base point (=support) sees when looking at the
excited structure, but not what he feels!
• The automatic frequency stepping generator may
be fooled out if the lower frequency bound is
requested to be 0 Hz, better use e.g. 1 Hz.
Otherwise, it may happen that a finer stepping
around the natural mode(s) does not correctly
take place!
• Therefore, always check stepping of the output
acceleration around the resonance frequencies with
a suitable measure
• Do not forget to toggle on “Use frequency steps
from table function” for non-uniform random
spectra to accurately capture function values at
these frequencies

49
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.5 Dynamic shock analysis

3.5.1 Introduction

• Dynamic shock analysis significantly differs from the other

three supported dynamic analysis types: At first sight the
user can observe that neither force excitation, nor any
modal damping  can be defined!
• The reason is that this analysis type only computes a single
image of the expected “worst-case” structural response
with help of a given SRS (Shock Response Spectrum), in
which damping already is included (this analysis type does
not provide any animations like e.g. dynamic time analysis!)
• An SRS is defined as the maximum response of an SDOF
(Single Degree Of Freedom) system with a given damping 
and variable natural frequency to base point excitation.
• So first, prior to the dynamic shock analysis, an SRS has to
be computed by the user for the existing base point
excitation (described on the next slide)

What is the advantage of dynamic shock analysis? Remark: Dynamic shock

analysis was proposed for
• Very low computational resources required (=very quick),
the first time in 1932 in the
since the DEQ does not need to be solved again! doctoral thesis of the
• The evaluation of the shock spectra for different excitation Belgian-American scientist
functions allows to judge immediately which excitation Maurice Anthony Biot
creates the worst-case loads!
50
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.5 Dynamic shock analysis

excitation function
3.5.1 Introduction

Given base point

• The process to generate an SRS for
a given base point excitation is
depicted in the right image
• The SRS can be expressed in

excitation (shown for three T0 examples)

Obtained SRS, shown in Disp. response of the SDOF system with
displacements 𝑑 (shown right),
“pseudo-velocity” 𝑣 or “pseudo-

varying eigenfrequency to the base

acceleration” 𝑎 vs. 𝑇0 or better, as
Note damping 
required in Simulate, vs. 𝑓0 is included here!
• These spectra are linked by the so
called “spectra response relation”:
s a  s v   2 s d
• Important: When computing the
SRS, note the different required
output references [3]:
 Displacements: relative
supports

expressed in acceler-
max. disp. vs. natural

(Pseudo-)acceleration: rel.
ation response a vs.
 (typically, the SRS is
vibration period T0

eigenfrequency f0)
ground
 Pseudo-velocity: has to be
computed by the spectra
response relation from
𝑠𝑑 or 𝑠𝑎

51
Rev. 1.2 | 07.04.2017
 independent SDOFs real MDOF system
3. Dynamic Analysis

3.5 Dynamic shock analysis

3.5.1 Introduction
≠
Modal superposition method
• For independent SDOF-systems, the dynamic shock analysis
always delivers per definition an accurate answer (since the
response spectrum was created with such a system)
• Unfortunately, for real MDOF systems this is not the case!
• Indeed, the dynamic response of a linear elastic MDOF
system can be expressed as linear combination of its
fundamental modes, but the information at which time the
modes answer with their maximum, respectively, is lost
during the creation of the shock response spectrum
• So two pragmatic (empirical) approaches are offered in
Simulate to compute the maximum structural response:
 The modes are simply combined with their absolute sum:
 very conservative approach, since in reality the modes
do not answer with their maximum at the same time
 “Square Root of the Sum of the Squares” (averaged
superposition, expected to be more realistic)
• For this superposition, the individual max. modal response
magnitude is read out from the shock response spectrum,
and of course the individual mass participation factors are
taken into account (base point excitation!)
• For more details about dynamic shock analysis, see [1-3]

52
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.5 Dynamic shock analysis

3.5.2 Examples

• A shock response spectrum can be obtained in several different ways, e.g.

 From Literature (typically normalized diagrams are given)
 By rules and standards (e.g. for civil or military engineering)
 With help of a Creo simulate global sensitivity study
 With Mathcad as shown in [7], [8]
 By other suitable software
 …
• In [3], a method is shown how to use the Simulate global sensitivity study for this:
 Create a simple SDOF model with just a point mass and a discrete spring
 Enter the base excitation function of interest in a dynamic time analysis, also enter the
damping present in the real structure you want to compute afterwards
 Define a measure for the max. system response over the complete analysis time
(use max. displacement output rel. supports, or max. acceleration relative ground, but
do not use a max. velocity response measure; see also [3], p. 23!)
 Define a property parameter to vary the fundamental frequency of the SDOF system (i.e.
spring stiffness or mass)
 Sweep this parameter in a global sensitivity study referencing the dynamic time analysis
in a way that the complete frequency span is covered for which you need the SRS.
Note: For practical reasons, you may need more than one analysis with different step
width and edge frequencies for this, since at lower frequencies you have less sample
points (f~ 𝐾)
 After the study, draw the max. displacement or acceleration response measure vs. SDOF
frequency (MS EXCEL) – this is the SRS you need!

53
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.5 Dynamic shock analysis

3.5.2 Examples

• By following this procedure, you should obtain the following shock response
spectrum for an acceleration half sine shock of 50 g, 6 ms duration ( f=83.3 Hz)
and no damping, see [3]

54
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.5 Dynamic shock analysis

3.5.2 Examples

• Now, take this spectrum and import it into the response

spectrum form sheet of the dynamic shock analysis
definition dialogue
• Run the dynamic shock analysis e.g. with the 700 Hz-
SDOF-system used in chapter 3.4
• A max. relative displacement result of 0.028 mm will be
computed (see next side left). Note: Displacement output
in a Simulate shock response analysis is always given
relative to supports, never relative to ground!
• Run two dynamic time analysis for cross-checking: One
relative to ground and one relative to supports, with the
measures shown below – you should obtain similar
results (units mm, t, s)
• For an SDOF-system, you can simply compute the max.
acceleration out of the shock analysis displacement result
with help of the spectra response relation (see next slide)

Remark: Using the spectra response relation for velocities may deliver a very inaccurate result compared to
𝑚𝑚 𝑚𝑚
the real rel. velocity, see also [3]: 𝑣𝑚𝑎𝑥 = 𝜔 ∙ 𝑑𝑚𝑎𝑥 = 2 ∙ 𝜋 ∙ 700𝐻𝑧 ∙ 0.027995 𝑚𝑚 = 123.1 ≫ 25.97
55 𝑠 𝑠

Rev. 1.2 | 07.04.2017

 3. Dynamic Analysis

3.5 Dynamic shock analysis

3.5.2 Examples

• Response comparison of the 700 Hz SDOF-resonator for the different analysis types
(50 g, 6 ms half sine shock, =0)

Displacement response from Acceleration response from

a dynamic time analysis a dynamic time analysis
relative to supports relative to ground
Displacement result of the
dynamic shock analysis

max. -2.798673e-02
max. 55.187 g

Alternatively, use the

spectra response relation:
𝑎𝑚𝑎𝑥 = 𝜔2 ∙ 𝑑𝑚𝑎𝑥
= 2 ∙ 𝜋 ∙ 700 𝐻𝑧 2
∙ 0.027995 𝑚𝑚 = 55.2 𝑔

56
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.6 Examples for dynamic analyses of big system models

• For Zero-G experiments, Airbus DS in Bremen develops various payloads for the ESA
MAXUS and TEXUS sounding rockets
• Such experiment platforms are subjected to very high dynamic loads during launch
Movie of a MAXUS
rocket launch
(Kiruna, Sweden)

MAXUS & TEXUS ballistic sounding missiles

simplified flight profile

57
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.6 Examples for dynamic analyses of big system models

• PERWAVES: An experiment that examines combustion processes in Zero G environment

• This experiment contains mechanisms and several fragile glass tubes: Their resistance
against various flight & operational loads had to be proved

Mode 8 showing
turn drive/
carousel
oscillation Mode 1
(uncritical, since showing
not excited over elastic
the interface!) experiment
support for
dynamic load
decoupling

58
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.6 Examples for dynamic analyses of big system models

• GRADECET – A high temperature furnace that melts metallic specimens at 1700 °C and
recrystallizes them convection free in Zero-G environment on MAXUS flights
• Improvements had to be found and analyzed to reduce dynamic random loading within
the heating section

Specimen

Exemplary harmonic excitation of the first 1-displacement response of initial (left) and improved design (right)
mode of the tantalum “finger”, covering a
ceramic tube around the specimen

59
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.6 Examples for dynamic analyses of big system models

• Computation of the force transfer functions of a tire test rig for high speed
uniformity measurements (HSU 5 from ZF Test Systems, Passau, Germany)

Animation shows
system response
at the first lateral
eigenfrequency

60
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.6 Examples for dynamic analyses of big system models

• Behavior of a new ZF highly-dynamic half axis test rig in hexapod architecture

 One goal of the examination was to estimate the response function for harmonic pulsing of the
two tire load cylinders with an amplitude of 20 mm
 This was done with help of the so called “seismic mass” concept

System response for Response for 20 mm

20 mm pulsing pulsing close to the
within the operating system eigenfrequency
frequency (Scale 2:1) (Scale 2:1)

61
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.6 Examples for dynamic analyses of big system models

• Random response analysis of a thermal imaging system to compute optical surface

displacements under specified acceleration spectral density (Hensoldt, Oberkochen)

62
Rev. 1.2 | 07.04.2017
 3. Dynamic Analysis

3.6 Examples for dynamic analyses of big system models

• Shock response behavior of an imaging system for visible and short wave infrared
light (Hensoldt, Oberkochen)

63
Rev. 1.2 | 07.04.2017
 4. Feedback to the software developer PTC

4.1 Missing functionality & enhancement requests to PTC for dynamic analysis

• Missing discrete damper:

 add a linear, discrete damper that connects two points at least with 𝐶𝑥
a simple translational and/or rotational damping constant C
• Missing support for rotating machinery:
 Support gyroscopic effects in modal analysis: Whereas for slim
rotors (e.g. the shaft example) this effect can be neglected, for
massive, disk-shaped rotors (fly-, gearwheels) it is of fundamental
importance for correct prediction of their rotordynamic behavior!
 Support centrifugal softening (also called “spin softening”): This
effect leads to a decrease of fundamental frequencies since the
elastic stiffness K has to be replaced by the effective stiffness: [Wikipedia]
𝐾𝑒𝑓𝑓 = 𝐾 − 𝜔𝑠2𝑚
(can be obtained by balancing spring and centrifugal force). m
This is of importance if the radial displacement u under centrifugal
force cannot be neglected compared to the radius r (e.g. for fast
spinning turbine blades). In case 𝐾 < 𝜔𝑠2𝑚: Instability appears!
• Improved functionality for dynamic shock analysis:
 Support more modal combination methods from literature, like e.g.
1st mode absolute, all higher modes as SRSS
 Support not only displacement, but also velocity and especially
acceleration fringe plots in the postprocessor (similar for measures)
 Since translational excitation can just be defined in the world
coordinate system, also support user-defined coordinate systems
? (Workaround: Assemble the model into a higher-level assembly
with suitable orientation!)
64
Rev. 1.2 | 07.04.2017
 4. Feedback to the software developer PTC

4.1 Missing functionality & enhancement requests to PTC for dynamic analysis

• Better support of results superposition in the post processor:

 Currently, results superposition is only supported within a
single linear static analysis containing at least 2 load sets
 It would be pretty useful if the user could superpose and scale
results of (any) different analyses (even, although risky, any
load step of a nonlinear analysis), as long as the models have
similar geometry, mesh and plotting grid
 This could e.g. allow to combine the static (constant)
prestress state of a prestress modal analysis e.g. with the
varying stress states of a dynamic time or frequency analysis
based on a prestress modal analysis (enables mean- and
deflection stress evaluation, which is currently not possible)
• Missing measures: Unfortunately, several measures are not
supported in various dynamic analysis types, like e.g.:
 No phase output for spring forces in dynamic frequency
analysis (workaround is using displacements and phase of
displacements of the spring end points, but this is by far not
accurate enough for very stiff springs!)
 No constraint or resultant force/moment measure output in
dynamic analysis at all
 No velocity and acceleration output in dynamic shock analysis
(just displacement and stress output, even for fringe plots!)
 No support of any fastener measures in random response
analysis
 …
65
Rev. 1.2 | 07.04.2017
 4. Feedback to the software developer PTC

4.1 Missing functionality & enhancement requests to PTC for dynamic analysis

• Improved support for analyzing very large system models:

 Experience won in many industrial projects shows that the code works very fine for
analyzing large CAD-assemblies (see chapter 3.6): The excellent CAD-FEM integration
and associativity of the analysis model with the CAD model allows a very quick
modification of the design which can immediately be re-meshed and re-analyzed in the
p-FEM environment
 Also all idealizations (beams, shells, springs, discrete masses, rigid and weighted links)
are supported and are well suitable to decrease model size
 However, the level of detail which can be taken into account in the model is usually
driven by the modal analysis with the requested number of modes, the chosen plotting
grid and the optional modal stress request. For big models, a lot of RAM is required so
that the engine does not crash!
 To decrease RAM resource consumptions, it would be useful that subsequent dynamic
analyses (which typically need much less RAM compared to the modal analysis) can
reference a series of modal analyses with split frequency band requests, respectively,
e.g. from 0-1000 Hz, 1000-2000 Hz and 2000-3000 Hz, and not only one single
modal analysis requesting all modes of the complete frequency domain (0-3000 Hz)
 Multi-threading should be supported in dynamic analysis, too (in modal analysis, the
code already addresses many CPU automatically). Until now, the user can only obtain a
significant speed increase in dynamic analysis by using an SSD hard drive!
• Implementation of superelements for model size reduction:
 For very large system models, a submodeling technique would be very useful, that
means certain subassemblies can just be represented as a “superelement” to decrease
hardware resources and increase analysis speed

66
Rev. 1.2 | 07.04.2017
 4. Feedback to the software developer PTC

4.2 Known issues in dynamic analysis and possible workarounds

Project work with the product uncovered the following issues to be fixed by PTC R&D:
• SPR 4714483: Off-diagonal terms in the spring stiffness matrix of advanced springs
lead to incorrect results in fundamental frequency computation (no workaround
available, currently do not use off-diagonal terms at all!)
• SPR 2875703: In dynamic frequency analysis with force excitation and phases
between the exciting forces, wrong animations/PP plots for certain result components
may appear (try with the shaft example in chapter 3.3.2 and compare with dynamic
time results; you may use dynamic time analysis as workaround!)
• SPR 2847768: In a random response analysis, Simulate may compute wrong von Mises
stress hot spots (this may appear especially for very big system models)
• SPR 2867898: Mixed models with shell-solid links may deliver wrong results in a
modal analysis (locking appears especially at high p-levels).
Workaround: Try to prevent shell-solid links by suitable modeling/meshing!
• SPR 4461169: In a model containing thin solid elements (wedges and bricks); Simulate
does not correctly detect rotational rigid body modes with standard settings in a
modal analysis (wedges and bricks – unlike tetrahedrons - need very high p-levels to
correctly detect rigid body modes at a frequency of zero!). Workaround: Use
tetrahedrons or enforce high accuracy in the modal analysis to use high p-levels!
• SPR 4967524: In a model containing variable thickness shells; msengine hangs up
during mass calculation. Workaround: Replace variable thickness shells with volumes
or with stepwise varying constant thickness shells!

67
Rev. 1.2 | 07.04.2017
 Any Questions?

68
Rev. 1.2 | 07.04.2017
 5. References

[1] Shock Response Spectrum – A Primer

J. Edward Alexander, BAE Systems, US Combat Systems
Minneapolis, Minnesota; J. Sound & Vibration, June 2009
http://www.sandv.com/downloads/0906alex.pdf
[2] An Introduction to the Shock Response Spectrum
Tom Irvine, Rev. R, July 29, 2010; www.vibrationdata.com
[3] Berechnung von Schockspektren und praktische Anwendung
der dynamischen Stoßanalyse in Creo Elements / Pro Mechanica
Roland Jakel, Presentation at the 3rd SAXSIM, TU Chemnitz, 19.04.2011, Rev. 1.0
www.saxsim.de, https://www.ptcusercommunity.com/blogs/RSS
[4] Linear and nonlinear Analysis of High Dynamic Impact Events with Creo Simulate
and Abaqus/Explicit; Roland Jakel, Presentation at the 7th SAXSIM, TU Chemnitz,
31.03.2015, Rev. 1.1
www.saxsim.de, https://www.ptcusercommunity.com/blogs/RSS

69
Rev. 1.2 | 07.04.2017
 [6]
5. References

[5] Zienkiewicz, O.C., „The Finite Element Method“;

3. Edition, McGraw-Hill, 1977, page 500-526
[6] https://femci.gsfc.nasa.gov/random/randomequations.html
[7] Berechnung von Schockantworten in Mathcad;
Dr. Wigand Rathman; Universität Erlangen-Nürnberg,
18. Nov. 2010; Vortrag zum 10. Simulationsanwender-
treffen im Rahmen der Plant PTC live (www.saxsim.de)
[8] Berechnung von Schockantworten für gemessene
Anregungen; Dr. Wigand Rathman; Universität
Erlangen-Nürnberg, 18. April 2011; Vortrag zum
Mathcad-Workshop im Rahmen des 3. SAXIM
(www.saxsim.de)
[9] Zienkiewitz, O and Zhu, J: A simple error estimator and
adaptive procedure for practical engineering analysis;
Int. J. Numer. Methods Engr. 24 (1987); 337-357

70
Rev. 1.2 | 07.04.2017
Dr.-Ing. Roland Jakel
Senior Consultant, Structural Simulation

Phone: +49 (0) 421 / 557 18 4111

Mobile: +49 (0) 173 / 889 04 19
[email protected]

Henning Maue
Global aircraft engineering architect ASDR group
Head of Engineering ASDR Germany
Technical Unit Manager Airframe Structure

Phone: +49 (0) 40 35 96 31 96 122

Mobile: +49 (0) 172 83 63 335
[email protected]