Sunteți pe pagina 1din 78

Model Checkout Procedures Using Nastran

2
Why do I need to check my model?

• A model may appear normal from the


outside, but a model checkout may
identify underlying issues
• Unintentional constraints
• Incorrectly connected CELASi
• Ill-conditioned MPC’s
• Mechanisms
• Poorly shaped elements

• If my model passes checks is it good?


• Mathematically, yes
• Physically, maybe
• A model is only as good as the assumptions.

3
Your Model Checkout Requirements

• You should follow the finite element model checkout


procedures provided by your company, organization,
or program
• These model checkout procedures are put in place to
ensure model quality by requiring a set of basic
“sanity” checks
• This presentation is not intended to replace your
model checkout procedures. It is intended to provide
insight into these sanity checks to help you
understand how they work

4
Model Checkout in MSC.Nastran

• MSC.Nastran provides a number of diagnostic tools for


checking out your model
• Some of these are provided as standard output such as Epsilon and
Maximum Matrix Diagonal to Factor Diagonal Ratio
• Others are available upon request such as GROUNDCHECK and
MODEL_CHECK
• This presentation will cover a number of diagnostic
tools which you can use (right out of the box – no DMAP
Alters!) to perform quantifiable sanity checks on your
model

5
Basic Model Checkout tools in MSC.Nastran
• Equilibrium check with 1G loading
• Grid Point Singularity check
• Max Pivot Ratio check
• Roundoff Error check
• Free-Free Normal Modes check
• Grounding check
• Mass check
• Free Thermal Expansion check
• Element Geometry check

6
Equilibrium Check with 1G loading

7
Equilibrium Check
• Why
• To detect modeling problems such as ill-conditioned MPC equations,
incorrectly defined spring elements, etc.
• The symptom is that the total applied load does not equal to the total
reaction

• How
• Run a linear static analysis with 3 orthogonal unit gravity subcases –
1G in X, 1G in Y, 1G in Z
• For each subcase, the SPC force resultant should be equal and
opposite to the OLOAD resultant
• You can compare them manually or use the case control command
EQUILIBRIUM to ask MSC.Nastran to do the comparison for you

8
Equilibrium Check – sample deck
• 3 Unit gravitational load cases

9
Equilibrium Check – Manual Comparison

Compare

10
Equilibrium Check – Let Nastran do it
• Case Control Command EQUILIBRIUM=YES

11
Grid Point Singularity Check

12
Grid Point Singularity Check
• Why
• By default, MSC.Nastran AUTOSPC will catch degrees of freedom
with little or no stiffness and constrain them in a number of solution
sequences
• This is reported in the Grid Point Singularity Table
• AUTOSPC may mask modeling problems – its output needs to be
inspected and understood

• How
• Run a linear static analysis
• Inspect the Grid Point Singularity Table if it is present
• Be able to explain why AUTOSPC has kicked in to constrain these
reported degrees of freedom

13
Grid Point Singularity Check - Example
• Review the Grid Point Singularity Table
• Understand why Nastran has constrained the degrees of freedom
shown below for you
• Is this what you intended or is this due to modeling errors?

14
Max Pivot Ratio Check

15
Max Pivot Ratio Check
• Why
• A model can be ill-conditioned due to modeling problems such as
• Rigid body motions due to inadequate constraints
• Mechanisms due to DOF mismatch between dissimilar element types
• Very stiff elements near flexible elements
• Very distorted elements
• Unreasonable material properties
• Etc.
• How
• Run a linear static solution
• MSC.Nastran automatically detects these ill-conditioning problems
during matrix decomposition by computing Matrix Diagonal to Factor
Diagonal Ratios and keeping track of the maximum ratio.
• When this maximum pivot ratio becomes too large (>107) or when
there are negative factor diagonal terms, the solution will stop and a
fatal error message is issued.

16
Max Pivot Ratio Check

• How (cont.)
• The MSC.Nastran fatal error message is accompanied by a table of
offending degrees of freedom and their pivot ratios. This table
provides clues to the cause of the problem.
• PARAM,BAILOUT,-1 may be used VERY CAREFULLY as a
debugging tool only. It forces the solution to proceed to completion
by ignoring the ill-conditioned matrix. This is intended to provide you
with additional solution output to help debug the model. In the final
production model, this parameter must be turned off.

17
Max Pivot Ratio Check - Examples
• A successful run

• A failed run

This table
provides clues to
modeling
problems

18
Roundoff Error Check

19
Roundoff Error Check

• Why
• Numerical roundoff errors can result from modeling problems
• These roundoff errors affect the accuracy of your analysis

• How
• Run a linear static analysis with 3 unit gravity load cases
• Inspect the value of Epsilon for each load case

20
Roundoff Error Check – What is Epsilon?

• In a static analysis, the equation

Ku  P

is solved for the displacements u


• Subsequently, residual loads can be computed from

ΔP  Ku  P

• Because of numerical roundoff, the residual loads will not be zero

21
Roundoff Error Check – What is Epsilon?
• For each subcase, a normalized value of the residual load is given by

u T ΔP
e
uTP

• A value of e less than about 10-9 is generally considered acceptable.

• Large values of e indicate an ill-conditioned model, possibly due to rigid


body motions or mechanisms.

22
Roundoff Error Check – Example

• Sample Epsilon output in .f06

23
Free-Free Normal Modes Check

24
Free-Free Normal Modes Check
• Why
• Your model might be “grounded” due to unintentional constraints,
incorrectly defined springs, problematic MPC equations, etc.
• The symptom is that the unconstrained (free-free) model will not have
clean rigid body modes

• How
• First remove the constraints so your model is free-free
• Perform a normal modes analysis and request for enough modes to
cover the rigid body modes (6 plus any intentional mechanisms) and
a few flexible modes
• The rigid body modes should have clean zero frequencies. Look for
at least 3 to 4 orders of magnitude frequency separation between the
last rigid body mode and the first flexible mode

25
Free-Free Normal Modes Check - Examples
• Model 1 - Clean rigid body modes with good separation between rigid/flex modes

• Model 2 - Mode 6 is grounded due to a bad MPC

26
Grounding Check

27
Grounding Check

• Why
• Unintentional constraints, incorrectly connected spring elements, ill-
conditioned MPC’s, and very poorly shaped elements are some of
the ways by which a model gets “grounded”
• The grounded model will lead to incorrect results
• We can detect this grounding problem by moving the model in a rigid
body fashion
• A clean model will exhibit true strain-free rigid body motion
• A grounded model, however, will develop internal strains and forces as it
is moved as illustrated by the “Gumby Stretch” on the following slides

28
The World According to Gumby
• You should be able to pick up an unconstrained object and move it
across the room without stressing it…

Free at last

29
What is a “Gumby Stretch”? (skyhook)

• A "Skyhook" is any constraint on what is a supposedly "free"


structure…

30
What is a “Gumby Stretch”? (skyhook)

• A "Skyhook" is any constraint on what is a supposedly "free"


structure…
Youch!

31
Grounding Check – Some Basics
• Statement of Fact
• Applying a uniform displacement field to a free structure will generate no
internal forces or strain energy
• In MSC.Nastran, a uniform displacement field can be calculated for all 6 dof
in a structure. These are called Rigid Body (RB) Modes
• Force Calculation RB Translation
• F = Kii * RB
• Strain Energy Calculation:
• S.E. = RBT * Kii * RB
• Result should be ~0

RB Rotation
Unit z
Unit y

Unit x

32
Grounding Check
• The grounding check is requested by the GROUNDCHECK case
control command
• Example: GROUNDCHECK(SET=ALL,DATAREC=YES)=YES

33
Grounding Check

• Rigid Body Strain Energy Check can be performed at


several DOF set levels
• G-SET [Kgg]
• Form stiffness matrix [K] from GRIDs & structural elements
• N-SET [Knn]
• Reduce out RIGID elements/MPCs
• F-SET [Kff]
• Apply user-specified SPCs
• A-SET [Kaa]
• ASET/SE boundary reduction

34
Grounding Check – G-SET Check

What causes S.E. > 0 for KGG?


• Check before rigid elements/MPCs and SPCs are processed
• CELASi with only 1 GRID listed
• CELASi between non-coincident GRIDs
• CELASi with non-colinear DOFs
• DMIG import errors (GRID locations, roundoff)
• CSUPER attach errors (GRID locations, order)
• REALLY bad elements...

Unit z
Unit y

Unit x

35
Why Do My CELAS Require Coincident Grids?

• CELAS elements are unidirectional springs bound to


obey small displacement theory:

x
Rotated
B Position

Initial
kx
Position
x = z * (LAB)  0
z SE = ½ kxx2  0
Frb = kxx  0
X
A
Y

36
Other CELAS Issues
• CELAS with only 1 GRID Not Specified
• CELAS2, K, G1, C1 , G2, C2
• {F} = [K] {}
• NO SPCFORCE !!!

• CELAS with non co-linear DOFs


•CELAS2, K, A, 1, B, 2 Cross Coupled dof
•  Ax connected to By
B’
B

A A’
37
Other CELAS Issues (cont.)
• CELAS with non co-linear DOFs
• Watch out for same DOF but different output coordinate
systems (CDs) !
•CELAS2, K, A, 1, B, 1
GRID, A, , 0.,0.,0.
GRID, B, , 0.,0.,0., 45
CORD2R,45, 0, …

B’
B

A A’
38
Example of Improper Spring (CELAS)

$.......2.......3.......4.......5.......6.......7.......8.......9.......0
$
$ Beam Made of Plates Model
$
$ 201 202 203 204 205 206 207 208 209 210 211
$ Y *----*----*----*----*----*----*----*----*----*----*
$ ^ | | | | | | | | | | |
$ | | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 |
$ | | | | | | | | | | | |
$ +--->X *----*----*----*----*----*----*----*----*----*----*
$ 101 102 103 104 105 106 107 108 109 110 111
$
$.......2.......3.......4.......5.......6.......7.......8.......9.......0

CELAS2 501 1.e3 111 1 211 1

X-Direction Spring Improperly Defined


Between Grids 111 and 211

39
KGG SE for Improper Spring

*** USER INFORMATION MESSAGE 7570 (GPWG1D)


RESULTS OF RIGID BODY CHECKS OF MATRIX KGG (G-SET) FOLLOW:
PRINT RESULTS IN ALL SIX DIRECTIONS AGAINST THE LIMIT OF 9.084249E-05
DIRECTION STRAIN ENERGY PASS/FAIL
--------- ------------- ---------
1 1.236913E-10 PASS
2 5.093170E-10 PASS
3 4.147296E-10 PASS
4 9.640644E-11 PASS
5 2.501110E-09 PASS
6 5.000000E+02 FAIL
SOME POSSIBLE REASONS MAY LEAD TO THE FAILURE:
1. CELASI ELEMENTS CONNECTING TO ONLY ONE GRID POINT;
2. CELASI ELEMENTS CONNECTING TO NON-COINCIDENT POINTS;
3. CELASI ELEMENTS CONNECTING TO NON-COLINEAR DOF;
4. IMPROPERLY DEFINED DMIG MATRICES;

40
KGG Forces for Improper Spring
•DATAREC=YES
•Recommended option on GROUNDCHECK

DIRECTION 6
G R O U N D C H E C K F O R C E S ( G - S E T )

POINT ID. TYPE T1 T2 T3 R1 R2 R3


111 G 1.000000E+03 0.0 0.0 0.0 0.0 0.0
211 G -1.000000E+03 0.0 0.0 0.0 0.0 0.0

Reaction Forces for


Rigid Body Modes.

Points directly to offending DOF!

41
Grounding Check – N-SET Check

What causes S.E. > 0 for KNN?


• Process rigid elements/MPCs, then check S.E.
• Improperly defined RBEi or MPC equations
• MPCs that do not follow rigid body motion
• Dis-similar DOFs (e.g. Dx(101) = Dy(102) )
• Different displacement coordinate systems
• RBE3s where the independent DOFs cannot describe all 6 rigid
body modes
• Co-linear GRIDs with 123 DOFs only

42
Example of Improper MPC

$.......2.......3.......4.......5.......6.......7.......8.......9.......0
$
$ Beam Made of Plates Model
$
$ 201 202 203 204 205 206 207 208 209 210 211
$ Y *----*----*----*----*----*----*----*----*----*----*
$ ^ | | | | | | | | | | |
$ | | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 |
$ | | | | | | | | | | | |
$ +--->X *----*----*----*----*----*----*----*----*----*----*
$ 101 102 103 104 105 106 107 108 109 110 111
$
$.......2.......3.......4.......5.......6.......7.......8.......9.......0

MPC 345 111 1 -1. 211 2 1.

Dof 111X tied to Dof 211Y

43
KNN SE for Improper MPC

*** USER INFORMATION MESSAGE 7570 (GPWG1D)


RESULTS OF RIGID BODY CHECKS OF MATRIX KNN (N-SET) FOLLOW:
PRINT RESULTS IN ALL SIX DIRECTIONS AGAINST THE LIMIT OF 9.084249E-05
DIRECTION STRAIN ENERGY PASS/FAIL
--------- ------------- ---------
1 2.271062E+05 FAIL
2 2.271062E+05 FAIL
3 4.147296E-10 PASS
4 9.640644E-11 PASS
5 2.501110E-09 PASS
6 4.598901E+06 FAIL
SOME POSSIBLE REASONS MAY LEAD TO THE FAILURE:
1. MULTIPOINT CONSTRAINT EQUATIONS WHICH DO NOT SATISFY RIGID-BODY MOTION;
2. RBE3 ELEMENTS FOR WHICH THE INDEPENDENT DEGREE-OF-FREEDOM CANNOT DESCRIBE
ALL POSSIBLE RIGID-BODY MOTIONS.

44
KNN Forces for Improper MPC
DIRECTION 1
G R O U N D C H E C K F O R C E S ( N - S E T )

POINT ID. TYPE T1 T2 T3 R1 R2 R3


110 G 2.619048E+05 0.0 0.0 0.0 0.0 0.0
111 G 0.0 1.785714E+05 0.0 0.0 0.0 0.0
210 G 2.875458E+05 -1.785714E+05 0.0 0.0 0.0 0.0
211 G -9.523810E+04 -4.404762E+05 0.0 0.0 0.0 0.0

DIRECTION 2
G R O U N D C H E C K F O R C E S ( N - S E T )

POINT ID. TYPE T1 T2 T3 R1 R2 R3


110 G -2.619048E+05 0.0 0.0 0.0 0.0 0.0
111 G 0.0 -1.785714E+05 0.0 0.0 0.0 0.0
210 G -2.875458E+05 1.785714E+05 0.0 0.0 0.0 0.0
211 G 9.523810E+04 4.404762E+05 0.0 0.0 0.0 0.0

MPC 345 111 1 -1. 211 2 1.

45
Grounding Check – F-SET Check
What causes S.E. > 0 for KFF?

• KFF check is after AUTOSPC and applied SPCs


• If model is supposed to be constrained (user-specified
SPCs), then there should be significant S.E.!
• If no user-specified SPCs, that leaves AUTOSPC’d
DOFs to check

46
Grounding Check – A-SET Check

What causes S.E. > 0 for KAA?


• Check S.E. after condensation to A-set
• A-set reduction present in SuperElements and/or
Reduced Stiffness models
• If model (or SuperElement component) is supposed to be
constrained (user-specified SPCs), then there should be
significant S.E.!

50
Mass Check

51
Mass Check
• Why
• Your FEM mass and CG should be in close agreement with the
actual structural design as reported in the mass properties report

• How
• Use the GRDPNT parameter to request for the mass properties
output

52
Mass Check
• Mass properties are computed by the MSC.Nastran grid
point weight generator (GPWG)

• The grid point weight generator computes


• the total mass
• the moments of inertia
• the location of the center of mass

• It is activated with parameter GRDPNT

53
Mass Check – PARAM,GRDPNT

PARAM, GRDPNT, grid

grid = -1 do not compute mass properties (default)

grid = 0 compute mass properties relative to the origin of the


basic coordinate system

grid > 0 compute mass properties relative to grid point with


identifier grid.
If this grid point does not exist, compute mass
properties relative to the origin of the basic coordinate
system

54
Output from Grid Point Weight Generator

C
D
E
F

55
Rigid Body Mass Matrix

• Output A shows the rigid body • M33 is the mass matrix. It usually is
mass matrix Mrr with respect to the a diagonal matrix with the mass on
reference point. the diagonal.
• It relates the inertia loads to the • R33 is the matrix of static moments
rigid body accelerations about the about the reference point.
reference point. • J33 is the matrix of moments of
• It can be subdivided into inertia about the reference point.

M 33 R 33 
M rr   T
 R 33 J 33 

56
Transformation from
Principal Mass Axes

• Output B shows the transformation matrix from


the principal mass axes to the basic coordinate
system.
• If no directional scalar masses have been defined
this should be an identity matrix.

57
Mass and
Center of Mass

• Output C shows the masses in the three directions and


the locations of the center of mass.
• The locations are the distances, in the basic coordinate
system, of the center of mass from the reference point.
• If no directional scalar masses have been input, all
masses should be identical.

58
Moments of Inertia

• Output D shows the moments of inertia about the center


of mass with respect to the basic coordinate system.
• Output E shows the transformation matrix from the
principal directions to the basic coordinate system.
• Output F shows the principal moments of inertia.

59
Free Thermal Expansion Check

60
Free Thermal Expansion Check
• Why
• Rigid elements, offsets, and MPC equations can cause problems in
your model if you plan to use it for thermal structural analysis
• Before applying actual thermal loads to your model, you should
perform a free thermal expansion check to help you identify
potential problem areas
• How
• Replace the model constraints with a set of statically determinate
constraints
• Set all the coefficients of expansion to a single value
• Apply a uniform T to the model. If the model is “clean”, then it
should be strain free.
• If strains develop in the model, they need to be investigated and
corrected by removing the artificial constraints that caused them.

61
Free Thermal Expansion Check
• Tips
• Editing the coefficient of thermal expansion field on all MAT cards and
switching between the two sets of data for production and checkout
runs is tedious and prone to error
• Instead of editing the coefficient of thermal expansion field on all MAT
cards, you can temporarily set them to a single value in a checkout
run
• This is done using the executive control statement MODEL_CHECK
• For example, MODEL_CHECK MAT_TECO=1.0E-6 will temporarily
update the thermal expansion direct coefficient for all MAT1, MAT2,
MAT3, MAT8 and MAT9 cards to 1.0E-6 for the duration of the run

62
Free Thermal Expansion Check – Example 1

A clean checkout run

63
Free Thermal Expansion Check – Example 2

A problematic checkout
run - Stress built up due
to an RBE2

64
Element Geometry Check

65
Element Geometry Check
• Why
• Poorly-shaped elements will deteriorate solution quality
• Very poorly-shaped elements can casue numerical difficulties such as
high pivot ratios and grounding

• How
• MSC.Nastran automatically performs the element geometry check and
flags elements which exceed the tolerance limits defined on the
GEOMCHECK executive control statement
• Set the tolerance values based on your program model checkout
requirement or best practice
• You can also set the amount of message and message type

66
The GEOMCHECK Statement

GEOMCHECK keyword  value, MSGMLIMIT  n,

  FATAL 
 MSGTYPE  INFORM , SUMMARY,NONE
  
  WARN 
 

• Examples:
GEOMCHECK Q4_SKEW = 15.0, MSGLIMIT=50
GEOMCHECK MSGLIMIT = 500
GEOMCHECK Q4_TAPER, MSGTYPE=FATAL
GEOMCHECK SUMMARY

67
The GEOMCHECK Statement (cont.)

keyword Keyword associated with the particular element


geometry
value Tolerance value to be used for the specific test
n Maximum number of messages produced for each
element type (default: 100)
FATAL Geometry checks that exceed the tolerance
produce a fatal message
INFORM Geometry checks that exceed the tolerance
produce an informative message (default)
WARN Geometry checks that exceed the tolerance
produce a warning message

68
The GEOMCHECK Statement (cont.)

SUMMARY A summary table of the geometry tests performed


is printed. No individual element information
messages are output.
NONE None of the optional element geometry tests will be
performed.

• If MSGTYPE is set to FATAL the job will abort after the element
matrix generation module has executed.

69
QUAD4 and QUADR Checks

70
QUAD4 and QUADR: Angles

• Skew Angle:
b4
a min  a  90 

b3
a
• Interior Angles:

b min  b i  b max b1
b2

71
QUAD4 and QUADR: Taper and Warping

• Taper Ratio: • Warping:

2A 2 zi
 i  i  1   max wi   wmax
A d1  d 2
z4
A
d2
z3
A1 d1
z1
z2

72
TRIA3 and TRIAR Checks

• Skew Angle:

a min  min b1 , b 2 , b 3  b3

b2
• Interior Angles: b1

b i  b max

73
Solid Element Checks
• Aspect Ratio: longest edge length divided by the shortest
edge length
• Edge Point Length Ratio:

min d1 , d 2 
 tol
e2 maxd1 , d 2 
d2
e i  30
e1
d1

74
Solid Element Checks (cont.)
• Face Warping Coefficient: cosine of the angle between the
normal vectors of diagonally opposite corner points

75
TETRA/HEXA/PENTA Element Checks

76
BEAM Element Checks

• The offset length ratio is the offset divided by the


length of the element.

77
GEOMCHECK: Output Example

78
GEOMCHECK: Output Example (cont.)

79
GEOMCHECK: Output Example (cont.)

80
Thank You

• Model Checkout in MSC.Nastran

S-ar putea să vă placă și