Meshing Techniques for Dynamic Problems
Meshing Techniques for Dynamic Problems
Chapter 9
Meshing
So far, we haven't discussed much on meshing because, for linear static problems, we usually can obtain solutions which
are acceptable both in computing time and accuracy with global mesh controls (such as Relevance Center, Relevance, and
Element Size) and some simple local mesh controls (such as Sizing). For the rest of the book, we will be dealing
with dynamic and nonlinear problems, of which the solutions are sensitive to meshing quality. With poor mesh quality, a
simulation may end up with solutions of poor accuracy or even run into convergence problems. Dynamic and nonlinear
problems require many computational resources, and poor mesh quality may aggravate the situation and result in a
lengthy computing time. For nonlinear problems, it is possible to reduce the runtime by improving the mesh quality
(because they converge easier). In contrast, it is possible that a nonlinear solution fails to converge just because of the
poor mesh quality.
@Seismicisolation
Section 9.1 Pneumatic Fingers 33 I
Pneumatic Fingers[Refl1
[I] Simulation of the pneumatic finger was previewed in Section 1.1. In this section, we will walkthrough each step.
Besides the information in I. I. I (pages 10-11), geometric details are given in the figure below. Due to the symmetry of
the cross section [2], we model only half of the finger. The mm-kg-N-s unit system is used in this section. |
80
[3] Double-click
Engineering Data.
U Modal
Static Structural
u Modal Acoustics
ш Random Vibration
QB Response Spectrum
Й Rigid Dynamics
u Static Acoustics [2] Double-click to create a
|P Static Structural
О Steady-State Thermal Static Structural system. f
tl) Thermal-Bectric
0 Topology Optimization
E? Transient Structural
C Transient Thermal
Q Turbomachinery Fluid Flow
Ш Component Systems
S Custom Systems
0 Design Explorabon
@Seismicisolation
332 Chapter 9 Meshing
Details View
Details of Extrade 1
Extrude Extrude 1
Geometry Shetchl О
Operation Add Material
Direction Vector None (Normal)
Direction Reversed О
Extent Type Fixed
FD1, Depth (>Q) 5.1 mm C
As ThinZSurface? No
Merge Topology? Yes
Geometry Selection: 1
Sketch Ishetehi [3] Extrude to
create a 3D
block. ।
©SeismicisolaHon
Section 9.1 Pneumatic Fingers 333
[5] Replicate 13 times, selecting all segments in the sketch except the
leftmost vertical segment, using the upper-left corner as the paste handle.
You may need to turn on Selection Filter: Points. |
s Geonmetxy Selectm: 1
[7] Extrude-cut to create
Sketch |Sketch2
air chambers. Close
DesignModeler. #
@Seismicisolation
334 Chapter 9 Meshing
Behavior None
S Material
Assignment PDMS
Rot
a Properties
Statistics
[3] Select PDMS. #
Geometry 1 Face
О
a Definition
Type Fixed Support
Suppressed No
Geometry 1 Face
О
a Definition
Type Fnctianless Support
Suppressed No
@Seismicisolation
Section 9.1 Pneumatic Fingers 335
Generate
[2] Select
Skewness 1for
Mesh Metric. |
[3]The default
settings result
in a poor
mesh quality
(see [4]). 1
Skewness[Refe2,3l
[4] Skewness, a measure of mesh quality, can be calculated for each element according to its geometry. Definition of
skewness can be found in the on-line documentation^5 2 3L For now, all you need to know is that it is a value ranging
from 0 to I, the smaller the better, and, as a guideline, element skewness should not be larger than 0.95. #
@Seismicisolation
336 Chapter 9 Meshing
Generate
Details of "Mesh" 9 x
s[Display
Physics Preference
Element Order Program Controlled
Element Size Default
S Sizing
Use Adaptive Sizing
Resolution
Mesh Defeatunng Yes
Defeature Size Default
Transition Fast
Span Angle Center Coarse [7] Select 7 for
initial Size Seed Assembly
Resolution. This is the
Bounding Box Diagonal 80.318 mm
Average Surface Area 29 597 mm» finest mesh we can have by
Minimum Edge Length 2 0 mm adjusting Resolution. /
Quality
Check Mesh Quality Yes. Errors
Error Limits Aggressive Mechanical
Target Quality Default (0.050000)
Smoothing Medium
Mesh Metric Skewness
|D Min 5.0302e-003
Max 0 89406
Average 0 34415
Standard Deviation 0.15781
Inflation
Advanced
Statistics [6] In many cases, it is true
[□ Nodes
that the finer the mesh, the
Elements
better the quality, but this
may not always be true
[9] A bar chart (see [10], next page) shows the distribution of
the skewness. We'll use this mesh to obtain a solution. Note
that the number of nodes here is 16573. ।
@Seismicisolation
Section 9.1 Pneumatic Fingers 337
Suppressed No
Deformation.
1.0271 Max
П-I -073279
-2.4927
- -4 2526
- -6 0126
- -7 7725
- -9 5324
- -11.292 [3]The maximum vertical deflection is 14.812 mm
-13.052
5 -14.812 Min
downward (at the finger tip). This value may not be
accurate, since we didn't consider nonlinear effect. |
@Seismicisolation
338 Chapter 9 Meshing
IBI Project
S- el Model (A4)
[- Geometry
ф Solid
Й Coordinate Systems
Mesh [6] Highlight
В . A Static Structural (A5)
Solution
Analysis Settings
ЯI Project ^*5^, Fixed Support Information. |
□ el Model (A4)
Frictionless Support
0
В
Geometry
ф Solid
Coordinate Systems
vrixi Mesh
-
Pressure
el Solution (A6)
♦ | Solution Information
V
Q
f
Directional Deformation
0 A Static Structural (A5)
Analysis Settings
[I] Highlight Analysis (Details of "Solution Information"
Fixed Support Settings. | 0 Solution Information
Fnctionless Support Solution Output Displacement Convergence Q ▼ |
i==i
\.... Pressure Newton-Raphson Residuals
Й- Ф| Solution (A6) Identify Element Violations
■4T| Solution Information
Update Interval
Directional Deformation AU Л
Display Points
В FE Connection Visibility [7] And select
[2] Turn on Auto
Details of "Analysis Settings" Activate Visibility Yes Displacement
Time Stepping. | Display All FE Connectors
Number Of Steps Convergence. |
Draw Connections Attached To AU Nodes
Current Step Number
Line Color Connection Type
Step End Time
Visible on Results No
Auto Time Stepping On
Line Thickness Single
Define By Time Select Time. Display Type Lines
Initial Time Step 5.e-00Qs
0 Solver Controls
Solver Type Program Controlled
Weak Springs Off [4] Type 0.05 (s) for Initial Time Step,
Solver Pivot Checking Program Controlled 0.01 (s) for Minimum Time Step, and
Large Deflection
0.05 (s) for Maximum Time Step. |
Inertia Relief | Off
s Restart Controls
a Noaliaear Controls
Output Controls Solve
a Analysis Data Management [5] Turn on Large Deflection. The
s Visibility simulation will include geometric
nonlinearity. We will explain these
settings further in Chapter 13. /
Displacement Convergence
[9] Each substep of a nonlinear simulation involves an iterative process. Force and displacement values are used as
convergence criteria. These concepts will be further explained in 13.1.5 (page 473). ।
@Seismicisolation
Section 9.1 Pneumatic Fingers 339
Cumulative Iteration
A: Static Structural
Directional Deformation
[10] The solution
Type: Directional Deformation(Y Axis)
Unit: mm converges in 25
Global Coordinate System iterations. |
Time: 1
1.0406 Max
-1 4941
-4.0289
-6 5637
-9.0984
-11.633
-14.168
-16.703
[12]The maximum vertical deflection is
•19.238____
-21.772 Min
21.772 mm downward, much larger than
the linear solution (9.1.9[3], page 337).
This justifies the inclusion of Large
Deflection (9.1.10[5], last page). |
Element Shapes
[13] In many cases, nonlinear simulations can be challenging. Meshing quality plays an important role in the
convergence of nonlinear solution. The mesh metric (e.g., skewness; see 9.1.7[9], page 336) is a measure of mesh
quality. Skewness often can be improved by refining elements. It sometimes needs a large number of elements to
achieve a mesh quality that is good enough to make the solution converge. In other cases, it may never achieve an
acceptable mesh quality by simply refining elements.
Another factor affecting convergence is the shapes of elements. In general, hexahedra are more efficient than
tetrahedra (see 9.3.13 and 9.3.14, page 361). In the following exercises, let's try to mesh the model with hexahedra. #
@Seismicisolation
340 Chapter 9 Meshing
Geometry 1 Body
[2] Select the solid
В Definition body. |
Method Sizing Face Suppressed No
Meshing
Method Hex Dominant
[5] Generate
mesh. |
Generate
Element Metrics
@Seismicisolation
Section 9.1 Pneumatic Fingers 341
Sweepable Bodies
[2] As mentioned in 5.3.2[2] (page 228), a simple idea for generating hexahedral elements is to mesh a face (or faces) of
a body with quadrilaterals and then "sweep" along its depth direction to the other end face (or faces) of the body. The
starting faces are called the source faces and the ending faces are called the target faces. The source or target faces can
be either manually or automatically selected.
Not all bodies are sweepable. In our case, there is only one body, and it is not sweepable.
@Seismicisolation
342 Chapter 9 Meshing
Geometry 1 Body
959.00
Number of Elements
750.00
500.00
250.00
0.00
0.20 0.30 0.40 050 062
Element Metrics
@Seismicisolation
Section 9.1 Pneumatic Fingers 343
Insert '
@Seismicisolation
344 Chapter 9 Meshing
@Seismicisolation
Section 9.1 Pneumatic Fingers 345
A: Static Structural
Directional Deformation
Type: Directional Deformation(Y Axis)
Unit: mm
Global Coordinate System
Time: 1
H 1.1116 Max
I -1.4041
-3.9198
-64355
-8 9512
-11 4«;
-13.983
-16.498
-19.014
-21.53 Min
Z
Wrap Up
[6] Save the project and exit Workbench. |
Remark
[7] As mentioned in 1. 1.8[8] (page 19), when assuming a linear material, we are also assuming the compressive behavior
is the same as tensile behavior, but this is usually not true for an elastomer under such a large deformation. (Note that
the upper portion of the finger is subject to tension, while the lower portion is subject to compression.)
Hyperelasticity, a more accurate material model for elastomer under large deformation, will be introduced in PART C
of Section 14.1 (pages 532-535). #
References
I. This exercise is adapted from an unpublished work led by Prof. Chao-Chieh Lan of the Department of Mechanical
Engineering, NCKU.
2. All Help>Meshing>Meshing User's Guide>Global Mesh Controls>Quality Group>Mesh Metric
3. All Help>Mechanical APDL>Theory Reference> 12.1. Element Shape Testing
@Seismicisolation
346 Chapter 9 Meshing
Section 9.2
Cover of Pressure Cylinder
[I] In this section, we will use the cylinder cover (Sections 4.2 and 5.2)
to demonstrate some additional meshing techniques.
The geometry of the cover is relatively complicated. It seems that a
tetrahedral mesh is the only feasible method. There is nothing wrong
with a tetrahedral mesh as long as the mesh quality is good enough.
Examining the mesh generated in 5.2.5[3-5] (page 221), we see that the
mesh quality is bad (9.2.3[I -3], next page). The mesh quality needs to be
improved. The simplest way is to adjust the relevance values. That
sometimes works, although increasing the problem size, but sometimes
fails. For a linear static simulation, problem size seems no big deal, but
for a nonlinear or dynamic simulation, the problem size should be kept
as small as possible, to maintain an acceptable computing time.
Note also that the purpose of this section is to demonstrate
meshing techniques, rather than finding the best mesh for the cylinder
cover. #
Static Structural
@Seismicisolation
Section 92 Cover of Pressure Cylinder 347
9.2.3 Increase Mesh Density Using Global Mesh Controls [7] With a mesh of
more than 360,000
nodes, some
Display
Display Style Use Geometry Setting
elements still have
[I] Highlight Mesh and bad quality. |
Physics Preference Mechanical
Element Order Program Controlled
select 4 for
Element Size Default Resolution.
Suing
Use Adaptive Sizing
Generate the mesh.
This is the mesh used in
Mesh Defeaturing
5.2.5[3-5], page 221. |
Average
[3] Some elements have
Standard Deviation large skewness. /
Inflation
Elements
[4] As we increase the mesh [5] Even when the number of [6] In this case, increasing
density by setting new relevance nodes increases to more than mesh density has little effect
values, the Max remains high. 100,000, the Max remains high. on Max value. T
@Seismicisolation
348 Chapter 9 Meshing
Method Generate
[6] Generate
[I] With Mesh highlighted, mesh. |
select Method to insert a
mesh metod. |
@Seismicisolation
Section 92 Cover of Pressure Cylinder 349
DefieiticMi
Method.
Suppoessed No
Method Tetrahedrons
Algorithm Patch Independent
[2] Select Patch
Element Order Use Global Setting Independent. / 0 Mesh
jPatch Conforming Method
Advanced
Defined By Max Element Size
@Seismicisolation
350 Chapter 9 Meshing
0 Definition Quality
Suppressed No Check Mesh Quality Yes. Errors
[9] If we further
Method Tetrahedrons Error Limits Aggressive Mechanical
increase the Target Quality Default (0 050000)
Algorithm Patch Independent
Element Midside Nodes Use Global Setting element size... Smoothing Medium
Mesh Metnc Skewness
Б Advanced
Min 2.0704e-002
Defined By Max Element Size Max 0 95501
Max Element Size Default Average 0.57351
Feature Angle 30.0“ Standard Deviation 0.
@Seismicisolation
Section 92 Cover of Pressure Cylinder 351
в Defieitin»
Suppressed No
Method Hex Dominant Q [2] Select Hex
Element Midside Nodes Use Globed Setting Dominant method. |
Free Face Mesh Type Quad/Tri
Control Messages Yes, Click To Display...
Generate
@Seismicisolation
352 Chapter 9 Meshing
- Quality
Check Mesh Quality Yes. Errors
[4] The mesh appears
Error Limits Aggressive Mechanical
Target Quality Default (0.050000) perfect... \
Smoothing Medium
Mesh Metric Skewness
Min 1 3058e-010
Max
Average 0 52958
Standard Deviation 025395
MultiZone Method
[6] As mentioned, the idea of MultiZone method is to decompose a non-sweepable body into several sweepable
bodies, and then apply Sweep method on each body. The selection of source faces can be automatic or manual. In
this case, you may try to select source faces manually.
Workbench is often smart enough to decompose the body into sweepable bodies. But, since these bodies are an
integral part, the boundaries between the decomposed bodies must be conformal (i.e,the boundaries must have the
same surface mesh); these constraints may complicate the meshing task. |
z
Wrap Up
[7] Save the project and exit Workbench. #
@Seismicisolation
Section 93 Convergence Study of 3D Elements 353
Section 9.3
Convergence Study of 3D Solid
Elements Number of Nodes
The main purpose of this section is to study 3D solid elements convergence behavior. A secondary purpose is to serve
as an exercise for mesh controls techniques. A cantilever beam of rectangular cross section is used for these purposes.
The conclusions drawn from the convergence study are crucial for CAE engineers. This section can be viewed as a
sequel of PART C of Section 3.5 (pages 161 -163).
[I] The cantilever beam is made of steel and of size 100 mm x 10 mm x 10 mm; a uniform load of I MPa is applied on
the upper face of the beam [2-4]. Convergence of three solid element shapes will be compared, namely hexahedron,
prism, and tetrahedron (1.3.3[2-3,5], page 38). \
@Seismicisolation
354 Chapter 9 Meshing
@Seismicisolation
Section 93 Convergence Study of 3D Elements 355
5 189 0.74571
4 4I6 0.74693
3 875 0.74850
2 1836 0.74980
I 12221 0.75106
Display
0.9 19097 0.75120
•ng
0.8 24696 0.75129 Defaults
Physics Preference Mechanical ]
Element Order Linear О
■ Element Size 5.0 mm
Sizing
Quality
Inflation
Advanced
s Statistics
□ Nodes 189 О
Elements 80
E
E
[4] Change
о Element Size
Ъ
ф
for each run. <-
@Seismicisolation
356 Chapter 9 Meshing
Method Tetrahedrons
for each run [3]. Resulting tip deflections are recorded
Algorithm Patch C
in the table below. The convergence curve is shown in Element Midside Nodes Use Globed S
[4]. -
4 188 0.50308
2 520 0.58585
1 2005 0.65522
Change
Element Size for
each run. <-
T ip D eflection (m m )
@Seismicisolation
Section 93 Convergence Study of 3D Elements 357
4 324 0.69584
2 2112 0.73344
1 13288 0.74671
0.75
0.72
E
E
о
0.70
0.67
@Seismicisolation
358 Chapter 9 Meshing
I 13288 0.73582
‘tails of "Mesh" 9
0.75 Display
Display Style Use Geometry Setting
Defaults
0.70
Physics Preference Mechanical
Element Order Linear о
T ip D eflection (m m )
0.50
@Seismicisolation
Section 93 Convergence Study of 3D Elements 359
S Display
[I] Delete Sweep Method. Repeat all the Display Style Use Geometry Setting
В Defaults
steps in 9.3.5 (page 355), using quadratic
Physics Preference Mechanical
elements [2] and change the element sizes as Element Order Quadratic
shown below.
3 3125 0.75071
| 0.751
2 6696 0.75108
c
О
1.5 16256 0.75136 Ъ 0.750
ф
0.748 -------------------------------------------------------------------------------
0 5000 10000 15000 20000 25000
Number of Nodes
9.3.10 Higher-OrderTetrahedra
4 971 0.74897
0.751
3 1371 0.74977 "I 0.750
2.5 1883 0.75016
.2 0 749
2 2880 0.75055
I 0.748
1.5 4127 0.75106 Ф
Q
1.2 9508 0.75110 Q- 0.747
Number of Nodes
@Seismicisolation
360 Chapter 9 Meshing
5 727 0.74815
4 1303 0.74958
3 3805 0.75049
2 9530 0.75096
5 727 0.74767
4 1303 0.74914
3 3805 0.75018
2 9530 0.75082
@Seismicisolation
Section 93 Convergence Study of 3D Elements 361
Number of Nodes
[I] The chart below is made from a collection of the convergence curves in 9.3.9 to 9.3.12 (pages 359-360) to compare
the convergence behaviors of the higher-order elements. The differences among them are not obvious but still
distinguishable. In contrast to the lower-order tetrahedron, the higher-order tetrahedron is still practically useful as
long as the mesh is fine enough. |
@Seismicisolation
362 Chapter 9 Meshing
[I] The chart below is made by a collection of the convergence curves in 9.3.5 (page 355) and 9.3.9 (page 359) to
compare the convergence behaviors between the lower- and higher-order hexahedra. It is obvious that the higher-
order hexahedral element is better than the lower-order hexahedral, but the difference is not so dramatic as the
tetrahedral element (see 9.3.16, this page). |
0.752
0.751
T 0.750
£
§ 0.749
4-J
| 0.748
О
0.747
0.746
0.745
0 5000 10000 15000 20000 25000
Number of Nodes
9.3.16 Comparison:Tetrahedra
[I] The chart below is made from a collection of the convergence curves in 9.3.6 (page 356) and 9.3.10 (page 359) to
compare the convergence behaviors between the lower- and higher-order tetrahedra. It is obvious that higher-order
element is much better than the lower-order one. Remember: NEVER use lower-order tetrahedral elements. |
@Seismicisolation
Section 93 Convergence Study of 3D Elements 363
[I] The chart below is made by a collection of the convergence curves in 9.3.7 (page 357) and 9.3.11 (page 360) to
compare the convergence behaviors between the lower- and higher-order parallel prisms. It is obvious that higher-
order element is much better than the lower-order one. Like lower-order tetrahedral, lower-order prismatic elements
are not recommended. |
0.76
0.74
0.72
0.70
0.68
0.66
0.64
0 5000 10000 15000 20000 25000 30000
Number of Nodes
[I] The chart below is made by a collection of the convergence curves in 9.3.8 (page 358) and 9.3.12 (page 360) to
compare the convergence behaviors between the lower- and higher-order perpendicular prisms. It is obvious that the
higher-order element is much better than the lower-order one. Like lower-order tetrahedral, lower-order prismatic
elements are not recommended, |
@Seismicisolation
364 Chapter 9 Meshing
z
Wrap Up
[3] Save the project and exit Workbench. #
@Seismicisolation
Section 9.4 Review 365
Section 9.4
Review
9.4.1 Keywords
Choose a letter for each keyword, from the list of descriptions
Answers:
I. ( В ) 1(C) 3. ( D ) 4. ( H ) 5. ( I ) 6. ( F 7. ( G ) 8. (J ) 9. ( A ) IO.( E )
List of Descriptions
( A ) A measure of mesh quality, calculated for each element according to its geometry. Its value ranges from 0 to I,
the smaller the better. Elements of skewness of more than 0.95 are considered unacceptable.
( В ) In nonlinear simulation, the loading is divided into substeps and applied substep by substep. By default, a substep is
said to be complete when both displacement convergence criterion and force convergence criterion are met during the
iterations.
( C ) During the iterations of a substep of a nonlinear simulation, the displacement convergence criterion is met when
the increment of displacement is less than a criterion, which is, by default, 0.5% of maximum displacement.
( D ) During the iterations of a substep of a nonlinear simulation, the force convergence criterion is met when the
unbalanced force is less than a criterion, which is, by default, 0.5% of applied force.
( E ) Sweep mesh control method can be classified into Sweep and Sweep Thin. Sweep allows a more complex
sweeping path while Sweep Thin allows only a simple sweeping path. The advantage of Sweep Thin is that it allows
multiple faces as source or target while Sweep allows only one face for both source and target.
@Seismicisolation
366 Chapter 9 Meshing
( F ) A mesh control method. It meshes all the faces of the body with triangles; the triangles then "grow" inward to
create tetrahedra. In this way, the shapes of the faces are respected (preserved).
( G ) A mesh control method. It creates tetrahedra from inside out. The outermost nodes are then projected onto the
boundary faces and the element edges are created. In this way, the mesh's outline may be different from the original
geometry.
( H ) A mesh control method. It meshes a body with Patch Conforming method first and then combines tetrahedra to
form hexahedra. It usually leaves some tetrahedra that cannot be combined to form hexahedra.
( I ) When a body is meshed with prismatic elements and the prisms are oriented such that their heights are parallel
to the bending direction, the prismatic elements oriented in this way are referred to as parallel prisms. (Note: this term
is used only in this book.)
( J ) When a body is meshed with prismatic elements and the prisms are oriented such that their heights are
perpendicular to the bending direction, the prismatic elements oriented in this way are referred to as perpendicular
prisms. (Note: this term is used only in this book.)
@Seismicisolation