Professional Documents
Culture Documents
MSC.Software Corporation
MA*2003*Z*Z*Z*SM-MAR103-NT1
Europe
MSC.Software GmbH Am Moosfeld 81829 Mnchen, GERMANY Telephone: (49) (89) 431 987 0 Fax: (49) (89) 436 1716
Asia Pacific
MSC Japan Ltd. Entsuji-Gadelius Building 2-39, Akasaka 5-chome Minato-ku, Tokyo 107-0052, JAPAN Telephone: (81) (3) 3505 0266 Fax: (81) (3) 3505 0914 Part Number: MA*2003*Z*Z*Z*SM-MA103-NT1
Worldwide Web
www.mscsoftware.com
Disclaimer
THE CONCEPTS, METHODS, AND EXAMPLES PRESENTED IN THE DOCUMENTATION ARE FOR ILLUSTRATIVE AND EDUCATIONAL PURPOSES ONLY, AND ARE NOT INTENDED TO BE EXHAUSTIVE OR TO APPLY TO ANY PARTICULAR ENGINEERING PROBLEM OR DESIGN. USER ASSUMES ALL RISKS AND LIABILITY FOR RESULTS OBTAINED BY THE USE OF THE COMPUTER PROGRAMS DESCRIBED HEREIN. IN NO EVENT SHALL MSC.SOFTWARE CORPORATION BE LIABLE TO ANYONE FOR ANY SPECIAL, COLLATERAL, INCIDENTAL, INDIRECT OR CONSEQUENTIAL DAMAGES ARISING OUT OF, RESULTING FROM, OR IN CONNECTION WITH USE OF THE CONTENTS OR INFORMATION IN THE DOCUMENTATION. MSC.SOFTWARE CORPORATION ASSUMES NO LIABILITY OR RESPONSIBILITY FOR ANY ERRORS THAT MAY APPEAR IN THE DOCUMENTATION. THE DOCUMENTATION IS PROVIDED ON AN AS-IS BASIS AND ALL EXPRESS AND IMPLIED CONDITIONS, REPRESENTATIONS AND WARRANTIES, INCLUDING ANY IMPLIED WARRANTY OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, ARE DISCLAIMED, EXCEPT TO THE EXTENT THAT SUCH DISCLAIMERS ARE HELD TO BE LEGALLY INVALID. MSC.SOFTWARE CORPORATION RESERVES THE RIGHT TO MAKE CHANGES IN SPECIFICATIONS AND OTHER INFORMATION CONTAINED IN THE DOCUMENTATION WITHOUT PRIOR NOTICE.
Trademarks
MSC, Dytran, MARC, and Patran are registered trademarks of MSC.Software Corporation or its subsidiaries in the United States and/or other countries. MSC., MSC.Dytran, MSC.Marc, and MSC.Patran are trademarks of MSC.Software Corporation. NASTRAN is a registered trademark of the National Aeronautics and Space Administration. MSC.Nastran is an enhanced proprietary version developed and maintained by MSC.Software Corporation. All other trademarks are the property of their respective owners.
Government Use
Use, duplication, or disclosure by the U.S. Government is subject to restrictions as set forth in FAR 12.212 (Commercial Computer Software) and DFARS 227.7202 (Commercial Computer Software and Commercial Computer Software Documentation), as applicable.
MA*2003*Z*Z*Z*SM-MAR103-NT1
Table of Contents
Contents
1 3
7 Course Objective: FEA & Laboratory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 Course Schedule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 About MSC.Marc Products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 About Axel Products, Inc. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 Data Measurement and Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 Typical Properties of Rubber Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 Important Application Areas. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 The Macroscopic Behavior of Elastomers . . . . . . . . . . . . .
Microscopic Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Temperature Effects, Tg . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Time Effects, Viscoelasticity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Curing Effects (Vulcanization) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Damage, Early Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Damage, Fatigue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Damage, Chemical Causes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Deformation States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
CHAPTER 2
21 22 23 24 26 27 28 28 29 31 32 33 35 36 38 40 41 43 45 46 47 48 49 51 52
3
CHAPTER 3
Engineering Materials and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Neo-Hookean Material Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Neo-Hookean Material Extension Deformation . . . . . . . . . . . . . . . . . . . . . . . . . . Neo-Hookean Material Shear Deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Neo-Hookean Material Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A Word About Simple Shear. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-Constant Mooney Extensional Deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . Other Mooney-Rivlin Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ogden Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Foam Models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Viscoelastic Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Model Limitations and Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Determining Model Coefficients. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
CHAPTER 4
Laboratory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Lab Orientation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Experimental Elastomer Analysis
Contents
Basic Instrumentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Measuring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . What about Shore Hardness? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Testing the Correct Material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Tensile Testing in the Lab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Compression Testing in the Lab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Equal Biaxial Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Compression and Equal Biaxial Strain States . . . . . . . . . . . . . . . . . . . . . . . . . . . . Volumetric Compression Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Planar Tension Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Viscoelastic Stress Relaxation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dynamic Behavior Testing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Friction Test. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Data Reduction in the Lab. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Model Verification Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Testing at Non-ambient Temperatures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Loading/Unloading Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Test Specimen Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Fatigue Crack Growth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Experimental and Analysis Road Map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
CHAPTER 5
Major Modes of Deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Confined Compression Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hydrostatic Compression Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . General Guidelines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mooney, Ogden Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Visual Checks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Material Stability. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Future Trends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Adjusting Raw Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Consider All Modes of Deformation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Three Basic Strain States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Curve Fitting with Mentat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
CHAPTER 6
Workshop Problems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Some Mentat Hints and Shortcuts. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Model 1: Uniaxial Stress Specimen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Model 1: Uniaxial Curve Fit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4
Contents
Model 1C: Tensile Specimen with Continuous Damage . . . . . . . . . . . . . . . . . . Model 1: Realistic Uniaxial Stress Specimen . . . . . . . . . . . . . . . . . . . . . . . . . . . Model 2: Equi-Biaxial Stress Specimen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Model 2: Equi-Biaxial Curve Fit. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Model 2: Realistic Equal-Biaxial Stress Specimen. . . . . . . . . . . . . . . . . . . . . . . Model 3: Simple Compression, Button Comp. . . . . . . . . . . . . . . . . . . . . . . . . . . Model 4: Planar Shear Specimen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Model 4: Planar Shear Curve Fit. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Model 4: Realistic Planar Shear Specimen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Model 5: Viscoelastic Specimen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Model 5: Viscoelastic Curve Fit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
123 135 139 143 155 158 166 170 185 188 190 203 204 206 207 208 209 210 210 211 212 213 215 217 218 219 220 220 221 222 223 224 225 226 228 228 229
CHAPTER 7
Contact Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Definition of Contact Bodies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Control of Rigid Bodies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Contact Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bias Factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Deformable-to-Deformable Contact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Potential Errors due to Piecewise Linear Description: . . . . . . . . . . . . . . . . . . . . Analytical Deformable Contact Bodies: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Contact Flowchart . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Symmetry Body . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Rigid with Heat Transfer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Contact Table. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Contact Areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Exclude Segments During Contact Detection. . . . . . . . . . . . . . . . . . . . . . . . . . . Effect Of Exclude Option:. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Contacting Nodes and Contacted Segments . . . . . . . . . . . . . . . . . . . . . . . . . . . . Shell Thickness is taken into account according to: . . . . . . . . . . . . . . . . . . . . . . Friction Model Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Stick-Slip Friction Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Coulomb (Sliding) Friction Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Shear (Sliding) Friction Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Coulomb Sliding Friction Model use Stresses or Forces . . . . . . . . . . . . . . . . . . Glued Contact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Release Option . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Interference Check / Interference Closure Amount . . . . . . . . . . . . . . . . . . . . . . Forces on Rigid Bodies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Contents
231 232 236 239 240 240 245 247 248 253 257 258 262 263 264 265 267 268 271 277 278 281 282 283 286 289 290 291 292 293
CHAPTER 1
Introduction
This course is to provide a fundamental understanding of how material testing and finite element analysis are combined to improve your design of rubber and elastomeric products. Most courses in elastomeric analysis stop with finite element modeling, and leave you searching for material data. This experimental elastomer analysis course combines performing the analysis and the material testing. It shows how the material testing has a critical effect upon the accuracy of the analysis.
Chapter 1: Introduction
Right Brain
Laboratory Experimental Subjective Intuitive
W = NkT ( I 1 3 ) 2 W = C1 ( I1 3 ) + C2 ( I2 3 )
N
W =
n=1
n ----- [ ( 1 n + 2 n + 3 n ) 3 ] n
1 1 2 2 W = G -- ( I 1 3 ) + --------- ( I 1 3 ) + 2 20N
Chapter 1: Introduction
Course Objective
Discuss the TEST CURVE FIT ANALYSIS cycle specific to rubber and elastomers. Limit scope to material models such as Mooney-Rivlin and Ogden form strain energy models. Material Model (curve fit)
Analyze Specimen
? n Correlatio ?
Analyze Part
Test Part
Chapter 1: Introduction
Why arent ASTM specs always the answer? What should I do about pre-conditioning? Why are multiple deformation mode tests important? How can I judge the accuracy of different material models? How do I double check my model against the test data?
10
Course Schedule
Chapter 1: Introduction
Course Schedule
DAY 1 Begin 9:00 10:30 12:00 1:00 3:15 5:00 End 10:15 12:00 1:00 3:00 5:00 Topic
Introduction, Macroscopic Behavior of Elastomers Laboratory Orientation Lunch Tensile Testing Tensile Test Data Fitting FEA of Tensile Test Specimen Adjourn
Chap. 1, 2, 3 4
5 6
DAY 2 - Chapter 6 + Lab Begin 9:00 10:45 12:00 1:00 3:15 End 10:30 12:00 1:00 3:00 5:00 Topic
Equal Biaxial Testing, Compression, Volumetric Equi-Biaxial Test Data Fitting, Comp., Volumetric FEA of Biaxial Specimen, Comp., Volumetric Lunch Planar Shear Testing Planar Shear Test Data Fitting Data Fitting with All Test Modes FEA of Planar Test Specimen Adjourn
5:00
11
Chapter 1: Introduction
Course Schedule
Chap. 6
3:15 5:00
5:00
Keep Involved:
Tell Me and Ill Forget Show Me and Ill Remember Involve Me and Ill Understand
12
Chapter 1: Introduction
13
Chapter 1: Introduction
14
Chapter 1: Introduction
Analysis
Analysis of continuum mechanics using FEA techniques introduces certain assumptions and approximations that lead to uncertainties in the interpretation of the results. This simply means that: FEA Results depend upon the approximations
Together
This course combines performing the material testing and the analysis to understand how to eliminate uncertainties in the material testing and the finite element modeling to achieve superior product design.
15
Chapter 1: Introduction
Tension/Compression
PA E = ------------------ ( L ) L P, L
Torsion
Tc J E = 2 ( 1 + ) ------------ T,
Bending
PL E = --------3I
3
P,
Wave Speed
E = c
2
Do you expect all of these Es to be the same for the same material?
16
Chapter 1: Introduction
17
Chapter 1: Introduction
600
18
Chapter 1: Introduction
19
Chapter 1: Introduction
20
CHAPTER 2
Elastomers (natural & synthetic rubbers) are amorphous polymers, random orientations of long chain molecules. The macroscopic behavior of elastomers is rather complex and typically depends upon: Time (strain-rate) Temperature Cure History (cross-link density) Load History (damage & fatigue) Deformation State
21
Microscopic Structure
Microscopic Structure
22
Temperature Effects, Tg
Typical values of Tg (in oC) are: -70 for natural rubber, -55
for EPDM, and -130 for silicone rubber.
23
Time Effects, Viscoelasticity Temp. & Time effects derive from long molecules sliding
along and around each other during deformation.
24
Time Effects, Viscoelasticity (cont.) Different types of tests can be used to evaluate the
short-time and long-time stress-strain behavior.
25
Curing Effects (Vulcanization) Curing creates chemical bonds cross-linking. Cross-link density directly affects the stiffness. Cross-link density effect for Natural Rubber:
Be careful that real parts and test specimens share the same
curing history, thus same stiffness.
26
Be careful that real parts and test specimens share the same
load history, Preconditioning.
27
Damage, Fatigue
Damage, Chemical Causes Many other chemicals are known to damage elastomers
and degrade the mechanical behavior: Ozone Oxidation Ultraviolet Radiation Oil, Gasoline Brake Fluid Hydraulic Fluid
28
Deformation States
Deformation States Shearing vs. Bulk Compression Shearing Modulus, G , typical ~ 1 - 10 MPa Bulk Modulus,
p K = ----------------- , typical ~ 2 GPa V V 0
K hence --- 10 3 G
and --
1 2
29
Deformation States
30
CHAPTER 3
It is useful to know the historical evolution of rubber material models. We will cover Neo-Hookean, Mooney, Mooney-Rivlin, and Ogden material models. Each model is based on the concept of strain energy functions, which guarantees elasticity.
31
Metal
Copper, Gold Bronze Casting Iron Age Damascus Steel
Elastomer
Analysis
32
Incompressibility:
1 2 3 = 1
33
Hence:
1 2 2 W = -- G + -- 3 2
Engineering Stress:
1 = dW d = G ---- = 2 1 = G 1 + ------------------ 2 (1 + )
2 1 True Stress: t = --------- = = G -- 1
Simple, one parameter material model Positive G guarantees material model stability
34
experiment
Engineering stress (N/mm2) (N/mm
2
4.0
2.0
theory
0.0
35
Y atan
X 1 If 1 = , then 2 = -- and 3 = 1
36
1.2
experiment
0.8
0.4
0.0
37
1
2 2 1 + ---- + 1 + ---2 4
2 1
1 2
3
2 1
1 2
Simple Shear
2 2 1 + ---- 1 + ---2 4
Neo Hookean
1 2 2 2 W = -- G ( 1 + 2 + 3 3 ) 2 = W = ( ) = W = G
direct stresses
shear stress
Note: Shear Stress-Strain Relation is the same for Hookean and Neo Hookean.
38
0 2
G= 2{ 1 + (1 + ) } {1 + (1 + ) } {1 + (1 + ) }
2 3 5
Biaxial
Planar Shear
2 2
Uniaxial
5.0
Hookean Biaxial Hookean Planar Shear Hookean Uniaxial New Hookean Biaxial Neo Hookean Planar Shear Neo Hookean Uniaxial
0.0
-5.0
-10.0 -1.0
-0.5
0.0
0.5
1.0
Engineering Strain
39
2 = 2
2 2 1 + ---- 1 + ---2 4
( 3 = 1)
into
1 1 2 2 2 2 W = -- G ( 1 + 2 + 3 3 ) = -- G 2 2
and then
= W = G
atan
40
Simple shear:
2 1 2 W = ( C 1 + C 2 ) 1 + ---- 2 = ( C 1 + C 2 ) 2 1 = dW d = 2 ( C 1 + C 2 )
Hence G = 2 ( C 1 + C 2 )
C2 C2 1 = 2 ---- C 1 + ----- or ----------------------------- = C 1 + ---- 2 2 2( 1 )
41
G F
0.3
E D C A
0.2
0.1 0.5
0.6
0.7
0.8
0.9
1.0
1/ 1/
42
I 3 = 1 , so that:
W = C 10 ( I 1 3 ) + C 01 ( I 2 3 ) (Mooney-Rivlin notation)
43
44
Ogden Models
W =
n=1
n and n are material constants, K is the initial bulk modulus, and J is the volumetric ratio, defined by J = 1 2 3
The order of magnitude of the volumetric changes per unit volume should be 0.01 Usually, the number of terms taken into account in the Ogden models is N = 2 or N = 3 .
45
Foam Models
W =
n=1
n ----- [ 1 n + 2 n + 3 n 3 ] + n
n=1
n ----- ( 1 J n ) n
46
Viscoelastic Models
Viscoelastic Models MSC.Marc has the capability to perform both small strain and large
strain viscoelastic analysis. The focus here will be on the use of the large strain viscoelastic material model.
where W ( E ij ) is a standard Mooney-Rivlin or Ogden form strain energy function for the instantaneous deformation. And R ( t ) is a relaxation function in Prony series form:
N
R(t) = 1
n=1
( 1 exp ( t ) )
47
Model Limitations and Assumptions This material model assumes that the rate of relaxation is independent
of the load magnitude. For instance, for relaxation tests at 20%, 50%, and 100% strain, the percent reduction in stress at any time point should be the same.
When used with an Ogden model, the material may be made slightly compressible.
48
Determining Model Coefficients This material model requires two different types of tests be
conducted and two separate curve fits be performed.
49
50
CHAPTER 4
Laboratory
Need to know: What are the actual tests used to measure elastomeric properties. The limitations of common laboratory tests. How to specify a laboratory experiment as required by your product requirements. Lets understand the specimen testing better to achieve better correlation and confidence in our component analysis.
51
Chapter 4: Laboratory
Lab Orientation
Lab Orientation
Laboratory Dangers
High Pressure Hydraulics Class II Lasers Instrument Crushing
52
Basic Instrumentation
Chapter 4: Laboratory
Basic Instrumentation
Electromechanical Tensile Testers
Servo-hydraulic Testers
53
Chapter 4: Laboratory
Basic Instrumentation
Aging Instrumentation
54
Measuring
Chapter 4: Laboratory
Measuring Force
Strain Gage Load Cells
Position
Encoders and LVDTs
Strain
Clip-on Strain Gages Video Extensometers Laser Extensometers
Temperature
Thermocouples
55
Chapter 4: Laboratory
Measurements
Measurements Force, Position, Strain, Time, Temperature Testing Instrument Transducers Load Cell (0.5% - 1% of Reading Accuracy in Range) Position Encoder (Approximately +/- 0.02 mm at the Device) Position LVDT (Between +/- 0.5 to +/- 1.0% of Full Scale) Video Extensiometer (Function of the FOV) Laser Extensiometer (+/- 001 mm) Time (Measured in the Instrument or at the Computer) Thermocouple
56
Chapter 4: Laboratory
What about Shore Hardness? Perhaps the Most Common Rubber Test Useful in General Easy to Perform at the Plant Generally Useless for Analysis!
57
Chapter 4: Laboratory
58
Chapter 4: Laboratory
59
Chapter 4: Laboratory
60
Chapter 4: Laboratory
61
Chapter 4: Laboratory
62
Chapter 4: Laboratory
63
Chapter 4: Laboratory
3 =
2 =
1 2
1 =
1 2
1 =
3 =
2 =
64
Chapter 4: Laboratory
Volumetric Compression Test Direct Measure of the Stress Required to Change the Volume of an Elastomer Requires Resolute Displacement Measurement at the Fixture Initial Slope = Bulk Modulus
Typically, only highly constrained applications require an accurate measure of the entire PressureVolume relationship.
Base Data Set
300
VOLCOMP_B
250 200
Pressure (MPa)
150
2
Bulk Modulus = 2.1 GPa
100
50
Volumetric Strain
65
Chapter 4: Laboratory
0.5
Planar Tension
0.4 0.3 0.2 0.1 0.0 0.0 0.2 0.4 0.6 0.8 1.0
Engineering Strain
66
Chapter 4: Laboratory
67
Chapter 4: Laboratory
7 6
0.8
Strain
Stress (MPa)
Strain = 50 %
Stress (MPa)
5 4 3 2
Strain = 30 %
0.0
2000
4000
6000
8000
Time (s)
Time (Seconds)
68
Chapter 4: Laboratory
69
Chapter 4: Laboratory
70
Friction Test
Chapter 4: Laboratory
Friction Test
Friction is the force that resists the sliding of two materials relative to each other. The friction force is: (1) approximately independent of the area of contact over a wide limit and (2) is proportional to the normal force between the two materials. These two laws of friction were discovered experimentally by Leonardo da Vinci in the 13th century, and latter refined by Charles Coulomb in the 16th century.
Friction Test
Friction Force
Coulomb performed many experiments on friction and pointed out the difference between static and dynamic friction. This type of friction is referred to as Coulomb friction today.
Position
In order to model friction in finite element analysis, one needs to measure the aforementioned proportionally factor or coefficient of friction, . The measurement of is depicted here where a sled with a rubber bottom is pulled along a glass surface. The normal force is known and the friction force is measured. Various lubricants are placed between the two surfaces which greatly influence the friction forces measured.
71
Chapter 4: Laboratory
Adjusted Data
72
Chapter 4: Laboratory
Data Reduction in the Lab (cont.) Data Reduction Considerations for Data Generated using Cyclic Loading
1. Slice out the selected loading path. 2. Subtract and note the offset strain. 3. Divide all strain values by (1 + Offset Strain) to account for the new larger stabilized gage length. 4. Multiply all stress values by (1+ Offset Strain) to account for new smaller stabilized cross sectional area. 5. The first stress value should be very near zero but shift the stress values this small amount so that zero strain has exactly zero stress. 6. Decimate the file by evenly eliminating points so that the total file size is manageable by the particular curve fitting software.
73
Chapter 4: Laboratory
74
Chapter 4: Laboratory
75
Chapter 4: Laboratory
Testing at Non-ambient Temperatures Testing at the Application Temperature Measure Strain at the Right Location Perform Realistic Loadings
Elastomers Properties Can Change by Orders of Magnitude in the Application Temperature Range.
76
Loading/Unloading Comparison
Chapter 4: Laboratory
Loading/Unloading Comparison
77
Chapter 4: Laboratory
78
Chapter 4: Laboratory
79
Chapter 4: Laboratory
TABLE 5. Analysis Workshop Models Model 1 2 3 4 5 6 7 8 9 10 Uniaxial Biaxial Planar Shear Compression Button Viscoelastic Volumetric Compression Friction Sled Foam Tube Crush Damage Planned Planned Planned Planned Planned mu=0.4 Description Notes
80
CHAPTER 5
The experimental determination of elastomeric material constants depends greatly on the deformation state, specimen geometry, and what is measured.
81
2 1 3
3 1 2
82
Major Modes of Deformation (cont.) Planar Tension, Planar Shear, Pure Shear
1 = 2 = 1 3 = 1
2 1 3
Simple Shear
83
Confined Compression
F
Hydrostatic Compression
F
84
Stress State:
1 = 2 = 3 = F Ao = p
p p K = ----------------- = ----------------L L 0 V V 0
200.0
MSC.Marc Mentat uses the pressure, 1 V V - 3 0 p , versus a uniaxial equivalent of the volumetric strain namely, 1 V V , to determine the bulk -0 3 modulus as shown on the right. Take care to divide the volumetric strain by 3, because V V 0 = L L 0 you may forget.
100.0 0.0 0.000 0.010 0.020 Equivalent Uniaxial Strain [1] 0.030
0.040
85
F, L
Stress State:
1 = 2 = 3 = F Ao = p
and since
= 1 + L L0
V0 .
Again MSC.Marc Mentat uses the pressure, p , versus a uniaxial -equivalent of the volumetric strain namely, 1 V V 0 , to determine the 3 bulk modulus.
86
General Guidelines
General Guidelines
Its just curve fitting!
No Polymer physics as basis Dont use too high order fit Remember polynomial fit lessons (high school?)
87
Yeoh (1995)
Rivlin and Saunders (1951) have pointed out that the agreement between experimental tensile data and the Mooney-Rivlin equation is somewhat fortuitous. The Mooney-Rivlin model obtained by fitting tensile data is quite inadequate in other modes of deformation, especially compression.
Using only uniaxial tension data is dangerous! Mooney model in MSC.Marc allows no
compressibility
Ogden model does allow compressibility
88
Visual Checks
Visual Checks
Extrapolations can be dangerous Always visually check your models predicted
response
Check it outside of the datas range (see below) Check it outside the tests scope
d d > 0
Predicted Response Predicted Response Real Material
d d < 0
DATA
Real Material
89
Material Stability
Material Stability
Unstable material model -> numerical difficulties
in FEA
Druckers stability postulate, d d > 0 Graphically:
d 11 d 11 > 0 d 11 d 11 < 0
strain range
90
Future Trends
Future Trends
Statistical Mechanics Models
Based on single-chain polymer chain physics Build up to network level using non-gaussian statistics
multiple modes
91
1.0
0.5
0.0 0.0
0.2
0.8
1.0
92
= ( p ) ( 1 + p )
1.0
= ( 1 + )
0.5
0.0 0.0
0.2
0.8
1.0
then needs to be shifted such that the curve passes through the origin. Remember hyperelastic models must be elastic and have their stress vanish to zero when the strain is zero. This shift changes the apparent gauge length and original cross sectional area.
93
Fit to upload
0 0
uniaxial/experiment
decide upon the most appropriate way to adjust the data prior to fitting the hyperelastic material models.
94
95
1.5
1.0
0.5
0.0 0.0
0.2
0.8
1.0
increasing stress for the same strain, respectively. Knowledge of this and the actual shape above where say at a strain of 80%, the ratio of equal biaxial to uniaxial stress is about 2 (i.e. 1.3/0.75 = 1.73) will be come very important as we fit this data with hyperelastic material models. Furthermore, this fit reduces the 10,000 data points taken from the laboratory to just a few constants.
96
97
Notice that the uniaxial, biaxial, planar shear and simple shear modes are shown, where the uniaxial mode matches the material input. To turn some modes off, or make other display modifications go to PLOT OPTIONS.
PLOT OPTIONS SIMPLE SHEAR PLANAR SHEAR RETURN SCALE AXES (this toggles it off) (this toggles it off)
98
99
squares fitting
The least squares error is given by either:
Ndata
error
i Ndata
or
error =
( measured calc )
The error R and error A are relative or absolute respectively Ndata is the total number of data points
calc is the calculated stress measured is the measured engineering stress
i i
100
method
Downhill-simplex method is an iterative method
Uses a number of start points
max min - tol Continues until: --------------------------------------------------------------- < ------
tol is set using CONVERGENCE TOLERANCE error min can be set with the ERROR LIMIT button
101
squares fitting
A Prony series (exponential decay) is fit to the
test data
The least squares error is given by:
Ndata
error
102
CHAPTER 6
Workshop Problems
These problems are to provide self paced examples to develop skills in performing elastomer material curve fitting and finite element analysis using MSC.Marc and MSC.Marc Mentat. Please note the directory hierarchy of: eea/wkshops_A/ or eea/wkshops_B/ uniaxial biaxial planar comp visco test_data (raw data)
103
104
This will produce and run a uniaxial stress model. Please familiarize yourself with this model. Look at the BCs, the material specification, the contact bodies and contact table, and the loadcase.
105
After the procedure file is finished the final picture on your screen will look like this. Here is a brief summary of the uniaxial model we have created: A single brick element, full integration, Herrmann. Boundary conditions on x=0 & y=0 faces to prevent free translation in space. Material model is neo-Hookean with C10 = 0.5 Rigid contact surfaces are used to impose deformation. lower rigid body, cbody2, is stationary. upper rigid body, cbody3, is moved so as to first push, then pull, the brick element. Loading is performed in 40 equal time increments. Increment 10 is full compression of 50%, increment 30 is full extension of 200%, increment 40 returns the brick to its original configuration. Now lets look at the results of this analysis before curve fitting our uniaxial test data.
106
All of the postprocessing functions are accessed from RESULTS, which is located on the topmost MAIN menu. We are especially interested in deformed shape plots and XY plots of stress vs. strain.
MAIN RESULTS OPEN DEFAULT DEF & ORIG SKIP TO INC 10 <cr> PLOT SURFACES WIREFRAME REGEN RETURN CONTOUR BAND SCALAR Displacement Z, OK SCALAR PLOT SETTINGS #LEVELS 5 <cr>, RETURN SKIP TO INC 30 <cr> FILL REWIND MONITOR
107
Now lets generate the stress-strain plot that the MSC.Marc analysis has calculated. When we curve fit the actual test data, this analysis stressstrain curve should match the curve fit response exactly.
HISTORY PLOT COLLECT GLOBAL DATA NODE/VARIABLES ADD GLOBAL CRV. Pos Z cbody3 Force Z cbody2 FIT, RETURN
Since the original area is one, and since the original length in the zdirection is one, the above plot is the engineering stress versus the engineering strain for a uniaxial stress specimen with neo-Hookean behavior. We use the Body 2 force just to get the sign correct. Another way of getting engineering stress-strain output is to use the user subroutine PRINCA.F. This is a plotv routine that calculates principal values of engineering stress & strain as well as principal stretch ratio. If available try re-running this analysis with the princa.f routine. Q: Why is it ok to use a one element model for this problem? A: ____________________________________________________
RETURN, CLOSE, SHORTCUTS SHOW MODEL OK, MAIN
108
109
Make the table type experimental_data, and associate this data with the uniaxial button. Your screen should look similar to the one below, and we are ready to start curve fitting the data.
TABLE TYPE experimental_data, OK, RETURN UNIAXIAL table2
110
Choose the neo-Hookean curve fitting routine and base the curve fit on just uniaxial data. The compute button will compute the model coefficients. By default, responses for many modes are plotted. The single neo-Hookean coefficient, C10, is 0.265.
ELASTOMERS NEO-HOOKEAN UNIAXIAL COMPUTE, OK SCALE AXES PLOT OPTIONS SIMPLE SHEAR, RETURN (this turns off simple shear)
111
Comments: We have just fit a neo-Hookean model using only uniaxial data. MSC.Marc Mentat by default shows the models response in all major modes of deformation. This is very important. You should always know your models response to each mode of deformation. Look again at the previous stress-strain plot. Notice the relative magnitude of the responses. Uniaxial is the lowest magnitude, the planar shear is higher, and the biaxial response is the highest. This is typical of most elastomers. See, for example, the stress-strain plot on the front cover of these notes. Always start fitting with simple models first. If a simple model captures the curvature of the test data, use it! Proceed to higher order and more complex models only as needed. Go back and use the EXTRAPOLATION feature and replot the neoHookean results from -0.5 to 2.0 strain. It is very important to look at the models response over a wide range of strain, including both tension and compression. We are looking for stability limits (maxima in the stressstrain curve). Mooney form models with all positive coefficients guarantee stability in all modes, for all strain. The simpler the material model, the higher probability it will be stable over a wider strain range. Later, after curve fitting several choices of models and selecting the best one, we will re-run our simple analysis.
112
Heres how to use the extrapolation feature to extend the strain range over which we plot the models response. We see that our neo-Hookean model is stable for all deformation modes.
NEO-HOOKEAN EXTRAPOLATION EXTRAPOLATE LEFT BOUND, enter -0.5, <cr> RIGHT BOUND, enter 2.0, <cr>, OK COMPUTE, OK SCALE AXES
113
Now fit a Mooney 2-term material model. Turn the extrapolation feature off for now. The Mooney coefficients are C10 = 0.074 and C01 = 0.280. Positive coefficients guarantee stability. Notice the relative magnitudes now the biaxial stiffness is about 4 times the earlier material model. Of course, the fit to the uniaxial data is better, with more terms this model can capture a higher curvature in the stress-strain data.
MOONEY(2) EXTRAPOLATION EXTRAPOLATE, OK COMPUTE, OK SCALE AXES
114
Now fit a Mooney 3-term material model. The Mooney coefficients are C10 = -0.735, C01 = 1.21, and C11 = 0.194. The uniaxial response is fantastic! The presence of a negative coefficient means that the material model might be unstable. We need to visually determine the stability range of the model. Note that the peak stress for the biaxial response has gone from 1.0 (neo-Hookean), to 4.5 (Mooney 2-term), to 36 (Mooney 3-term). Which one is correct?
MOONEY(3) COMPUTE, OK SCALE AXES
MOONEY(3), EXTRAPOLATION EXTRAPOLATE, OK COMPUTE, OK SCALE AXES (after viewing this turn extrapolate back off)
115
Comments: Which biaxial fit is correct? Well, we dont know because we havent (yet) performed a biaxial test. This is the great difficulty with the Mooney form and Ogden form material models they are just curve fits. There is no rubber physics embedded in these equations. As we see here, a curve fit to uniaxial data will have a good response for that mode of deformation. But the responses for the other modes of deformation are all over the map. A rule of thumb based on observations of natural rubber and some other elastomers is that the tensile equi-biaxial response should be about 1.5 to 2.5 times the uniaxial tension response. We have seen many instances of higher order Mooney and Ogden models (using only uniaxial data) returning biaxial responses that are far too high. These are clearly bad material models. Try playing with the POSITIVE COEFFICIENTS option to see how much the responses change. For the curve fitting examples, you may need to toggle certain things on & off to better view and understand the computed fit. Keep these features in mind throughout all of these exercises: EXTRAPOLATION on/off PLOT OPTIONS, PREDICTED MODES (select subsets of UNIAXIAL, BIAXIAL, PLANAR SHEAR) PLOT OPTIONS, LIMITS, YMAX, etc. (you may need to set plot limits by hand for better viewing)
116
Now fit an Ogden 2-term material model. The uniaxial response is very good, but the biaxial response is now even higher than the Mooney 3term. Ogden coefficients come in pairs, the moduli are i and the exponents are i . If each i and i have the same sign then stability is guaranteed. If a i is positive and its corresponding i is negative (or vice versa) then the material model might be unstable. Thus we may need to visually determine the stability range of the model.
OGDEN COMPUTE, OK
This plot is to the same scale (ymax) as the Mooney 2-term plot.
117
Comments: We are now finished with the curve fitting portion of this uniaxial exercise. We see that the Mooney 3-term and Ogden 2-term fit the uniaxial test data very well. However, we are concerned (or should be!) that the equi-biaxial response for some models (M 3-term, O 2-term) are too high and could make the material model overly stiff if that mode of deformation exists in our analysis. We need equi-biaxial test data to get a better fit to that mode. Lets run this uniaxial analysis with the Ogden 3-term model. We select the curve fit model by pressing the APPLY button. Now go back and view the material model. Submit the analysis, then we will postprocess and show the analysis calculated stress-strain curve.
OGDEN # OF TERMS = 3, OK COMPUTE, APPLY, OK PLOT OPTIONS (turn off all leave uniaxial only) > XY (sends to generalized xy plotter) RETURN (thrice) MECH. MATERIALS TYPE, MORE OGDEN (look at the material properties) OK FILES SAVE AS ogden3, OK MAIN JOBS RUN SUBMIT1 MONITOR
118 Experimental Elastomer Analysis
Now go to postprocessing and generate the engineering stress-strain curve (we did this earlier with the original model). We will also save the analysis generated stress-strain curve to an external file for comparison to the test data.
MAIN RESULTS OPEN DEFAULT HISTORY PLOT COLLECT DATA 14 30 1 <cr> (this collects just the tensile part) NODE/VARIABLES ADD GLOBAL VAR. Pos Z cbody 3 Force Z cbody 2 FIT, RETURN > XY (send results to generalized xy plotter) SAVE type ogden3.tab
This last command saves the table to an external file named ogden3.tab (.tab is just to remind us that it is table data).
119
To compare the two stress-strain curves, we will use MSC.Marc Mentats generalized plotter feature.
UTILS GENERALIZED XY PLOT FIT SHOW IDS = 0
120
Zoom in and tilt the plot and you will notice three curves: the data, the fit, and the response of our model. Note that the model must follow the hyperelastic material model (Ogden(3)) exactly.
tress Engineering S
at a
O gd ns e en (3 )f it
es po
Engin
tra ring S ee
in
121
One may also use xmgr to read the file ogden3.tab that was generated in MSC.Marc Mentat. From a terminal window type: xmgr st_18.data ogden3.tab A graphics screen will appear in which the experimental data is shown in black and the analysis generated stress-strain curve is shown in red. Of course, the test data only extends to about 100% strain whereas we performed our analysis out to 200% strain.
122
1.0
0.8
0.6
0.4
0.2
0.0 0.0
0.2
1.0
This damage can be due to polymer chain breakage, multi-chain damage, and detachment of filler particles from the network entanglement.
123
In this workshop problem, we will simulate this behavior using the continuous damage model discussed in Appendix B. To clarify the behavior lets plot the peak stress versus the cycle number as shown below.
Tensile Data
Continuous Damage for Engineering Strain = 1.00
1.10
1.05
1.00
0.95
0.90 0.0
2.0
4.0
6.0
8.0
10.0
Cycle Number
If our application experiences, this kind of behavior then we may wish to simulate this continuous damage. We would start by doing any normal hyperelastic curve fit. However, we would use the 1st cycle of the stress strain curve, not the steady state behavior in the file st_18.data which was for the 10th cycle shown above. We are now ready to begin modeling this continuous damage. In a terminal window, use the cd command to move to the wkshops_A/uniaxial or the wkshops_B/uniaxial directory.
124
From NT (Windows 2000) just click on the uni_neo05.proc file or from unix Type mentat to start the MSC.Marc Mentat program, then starting from the main menu proceed as follows:
UTILS PROCEDURES EXECUTE pick the file named uni_neo05.proc OK MAIN
This will produce and run a uniaxial stress model. Using this model file, we will go to the material definition stage and redefine the material by reading the uniaxial data, filename st_1st.tab, damage data, st_cont.tab, loading data st_load.tab and proceed to re-run the problem using an Ogden 1-term fit with continuous damage.
MATERIAL PROPERTIES EXPERIMENTAL DATA FITTING TABLES READ NORMAL FILTER: type st* pick file st_1st.tab, OK (different data from st_18.data)
125
ELASTOMERS MORE CONTINUOUS DAMAGE CONSTANT NUMBER OF TERMS = 2 FREE ENERGY = 1.07 (this is just the 1st peak stress) COMPUTE APPLY, OK, RETURN
OGDEN UNIAXIAL NUMBER OF TERMS = 1 COMPUTE, APPLY, OK SCALE AXES PLOT OPTIONS SIMPLE SHEAR (this turns off simple shear)
Experimental Elastomer Analysis 127
RETURN (twice)
Lets review the material properties to check that the curve fit has been properly applied to the selected material.
MAIN MATERIAL PROPERTIES MORE OGDEN, DAMAGE EFFECTS - RUBBER, OK OK
128
Now we can complete the model and run the analysis. The remaining item to finish is to attach a table to the contact body to cycle the loading several times from a strain of 0 to a strain of 1.
MAIN CONTACT CONTACT BODIES EDIT (pick cbody3) RIGID POSITION (Z) TABLE (pick table st_load) OK (twice) MAIN LOADCASE MECHANICAL
129
STATIC TOTAL LOADCASE TIME = 940 # STEPS = 20 OK MAIN FILES SAVE AS ogden_damage OK MAIN JOBS RUN, SUBMIT1, MONITOR, OK MAIN RESULTS OPEN DEFAULT HISTORY PLOT COLLECT DATA 1 19 2 NODE/VARIABLES ADD GLOBAL VAR. Time Force Z cbody 2 FIT, RETURN > XY (send results to generalized xy plotter) SAVE type ogden_damage.tab
130
Here we see the peak engineering stress drop upon subsequent applications of the prescribed displacements. Lets run this same example but increase the number of load cycles by using the BEGIN/END SEQUENCE feature of MSC.Marc. This can be done by closing the post file, going to jobs, editing the input file to MSC.Marc then executing the edited input file.
MAIN RESULTS CLOSE MAIN JOBS RUN ADVANCED JOB SUBMISSION EDIT INPUT
Here we need to locate the first occurrence of the auto load keyword.
Experimental Elastomer Analysis 131
Now locate the second occurrence of the keyword continue and insert after it the following:
end sequence
Now delete all input records after the end sequence record inserted. The tail end of the input data set will look like:
begin sequence,100, auto load 1 0 time step 4.700000000000000+1 motion change 2 2 0 0.000000000000000+0 3 -1 0.000000000000000+0 continue auto load 1 0 time step 4.700000000000000+1 motion change 2 2 0 0.000000000000000+0 3 -1 0.000000000000000+0 continue end sequence
10
10
This change to the input file will run with 100 repetitions of the load sequence above.
132
Save the input file and run the job by selecting the execute button, namely:
OK RUN, EXECUTE1, MONITOR, OK MAIN RESULTS OPEN DEFAULT HISTORY PLOT COLLECT DATA 1 1999 2 NODE/VARIABLES ADD GLOBAL VAR. Time Force Z cbody 2 FIT, RETURN
133
We now see that the engineering stress asymptotically approaches a value of 0.972 [Mpa] from its initial value of 1.114 [Mpa]. As shown below, the peak stress drops by about 13% from the initial load to an infinite number of repeated loadings. Although this drop may not appear to be large, other materials may demonstrate larger drops in peak stress upon repeated loadings and be more worthy of damage modeling.
Tensile Simulation - Continuous Damage
1-Term Ogden and Original Data
1.0
0.8
0.6
0.4
0.2
0.0 0.0
0.2
0.4
0.6
0.8
1.0
Should one desire to use a Mooney material model, the model would have to be converted to an updated Lagrangian formulation, by changing to element type 7, and choosing the LARGE STRAIN-UPDATED LAGRANGE rubber elasticity procedure. Finally, the hyperelastic fit above can be made better by simultaneously using other deformation modes as we shall see in subsequent exercises.
134
This problem is in the subdirectory named ./uniaxial/big. Geometry is 55 L x 4 H x 2 W (mm) between grips, 10 mm length under grip. Elements are 1 mm x 1mm in XY, 0.5 mm in Z. Read model from file uni_f4.mud or uni_f6.mud. Grips are modeled as discrete rigid surfaces, with a friction coefficient of 0.4 (uni_f4) and 0.6 (uni_f6). Run analysis with Mooney 1-term model and plot engineering stressstrain, compare with original test data. Use princa.f usersub if possible.
135
L d A=4mm^2
Mooney 1 Term C = 0.265
1 Element
Engineering Stress
600 Element
Engineering Strain
136
Results: uni_f4.mud
Model 1 Stress Strain Diagram
2.0 Specimen, Data, Test Element Mooney C1=0.265 Specimen (Grip Force)/4 versus (Grip Disp)/27.5 Uniaxial Data Test Element
1.0
0.5
0.0
-0.5
-1.0
0.0
0.5
1.5
2.0
137
138
This will produce and run a biaxial stress model. Please familiarize yourself with this model. Look at the BCs, the material specification, the contact bodies and contact table, and the loadcase.
139
After the procedure file is finished the final picture on your screen will look like this. Here is a brief summary of the biaxial model we have created: A single brick element, full integration, Herrmann. Boundary conditions on y=0 face to prevent free translation in space. Material model is neo-Hookean with C10 = 0.5 Rigid contact surfaces are used to impose deformation. cbody2 & cbody5 are stationary. cbody3 & cbody4 are moved so as to impose displacements in the Z & X directions respectively. Loading is performed in 30 equal time increments. Increment 10 is biaxial compression of 50% (compression in X & Z), increment 30 is biaxial extension of 200%(extension in X & Z). Now lets look at the results of this analysis before curve fitting our biaxial test data.
140
All of the postprocessing functions are accessed from RESULTS, which is located on the topmost MAIN menu. We are especially interested in deformed shape plots and XY plots of stress vs. strain.
MAIN RESULTS OPEN DEFAULT DEF & ORIG SKIP TO INC 10 <cr> CONTOUR BAND SCALAR Displacement Z, OK SCALAR PLOT SETTINGS #LEVELS 5 <cr>, RETURN SKIP TO INC 30 <cr> REWIND MONITOR
141
Now lets generate the stress-strain plot that the MSC.Marc analysis has calculated. When we curve fit the actual test data, this analysis stressstrain curve should match the curve fit response exactly.
HISTORY PLOT SET NODES (pick node 8 shown) END LIST COLLECT DATA 0 30 1 <cr> NODE/VARIABLES ADD VARIABLE Displacement Z Force Z cbody 2 FIT, RETURN RETURN CLOSE, MAIN
Pick
Since the original area is one, and since the original length in the zdirection is one, this plot is the engineering stress versus the engineering strain. We use the Body 2 force just to get the sign correct. Notice how much different compression is for biaxial than uniaxial behavior. Of course, biaxial compression is very hard to simulate with a physical test, and only tension is usually done.
142
143
Since we will be reading more than one set of test data, lets name the datasets. Then make the table type experimental_data, and associate this data with the uniaxial button.
NAME uniaxial TABLE TYPE experimental_data, OK, RETURN UNIAXIAL uniaxial
Repeat the above sequence to read in the file eb_18.data and name this dataset biaxial. Associate this dataset with the biaxial button. Your screen should look similar to the one below and we are ready to start curve fitting the data.
144
Choose the neo-Hookean curve fitting routine and base the curve fit on all the data. The compute button will compute the model coefficients. By default, responses for many modes are plotted. Turn off the plotting of simple shear.
ELASTOMERS NEO-HOOKEAN USE ALL DATA COMPUTE, OK SCALE AXES PLOT OPTIONS SIMPLE SHEAR, RETURN (this turns off simple shear)
145
Comments: We have just fit a neo-Hookean model using both uniaxial and biaxial data. MSC.Marc Mentat by default shows the models response in all major modes of deformation. This is very important. You should always know your models response to each mode of deformation. Compare this plot with the uniaxial only stress-strain plot on (page 111). Both plots are very similar. The uniaxial only C10 was 0.265, while the new material model based on both uniaxial and biaxial data gives C10 = 0.280. These neo-Hookean coefficients are quite close, telling us that the earlier model was pretty good. We would prefer to use the latest model since it is based on more information and gives a better fit to the biaxial test data. If you can accept the differences between the test data and fitted response, this material model is quite adequate (and stability is guaranteed because the coefficient is positive). For scoping analysis and the initial stage of an analysis, this model is sufficient.
146
Now fit a Mooney 2-term material model. Make sure extrapolation is off. The Mooney coefficients are C10 = 0.247 and C01 = 0.0270. Notice the relative magnitudes now the biaxial response is much different than before (page 114) and the coefficients are much different as well. (Uniaxial coeffs were C10 = 0.074 and C01 = 0.280). This confirms our suspicion that the earlier Mooney 2-term model based on only uniaxial data misrepresented the biaxial behavior.
MOONEY(2) COMPUTE, OK SCALE AXES
147
To see the old (uniaxial data only) fit response plotted along with the new data, use the EVALUATE feature.
MOONEY(2) EVALUATE type in the old coeffs as prompted at the command line ENTER C10: 0.074 <cr> ENTER C01: 0.280 <cr> All coefficients entered. Continue? y <cr>
So this is the uniaxial only model response. Notice how overly stiff the biaxial model response (yellow/light grey line) is compared to the actual biaxial test data (yellow/light grey line with squares).
148
Now fit a Mooney 3-term material model. The Mooney coefficients are C10 = 0.246, C01 = 0.029, and C11 = -0.0004. This is essentially the same as the Mooney 2-term material model from the previous page. The biaxial data is adding additional constraint to the fit. The third term is almost zero, thus the fit has not changed. One would not choose this model over the Mooney 2-term fit.
MOONEY(3) COMPUTE, OK SCALE AXES
149
Now fit an Ogden 2-term material model. The uniaxial and biaxial model responses are slightly better than the Mooney models. However, the first pair of coefficients (modulus term of -2.55E-6 and exponent of -10.5) only contribute to the response at high strains. Set the NUMBER OF TERMS to 1 and re-fit the data.
OGDEN COMPUTE, OK
150
Comments: We are now finished with the curve fitting portion of this uniaxial and biaxial data exercise. As you saw, the addition of biaxial information was very valuable. The earlier Mooney and Ogden uniaxial only fits were way off base! However, it is interesting to note that the earlier neo-Hookean fit was pretty decent. This gives more merit to keeping the material as simple as possible. Lets run this biaxial analysis with the Mooney 2-term model. Go back to MOONEY(2) and fit it again, press the APPLY button. Submit the analysis, then we will postprocess and show the analysis calculated stress-strain curve.
MOONEY(2) COMPUTE APPLY, OK PLOT OPTIONS, > XY, RETURN RETURN (twice) MECHANICAL MATERIALS TYPE, MORE MOONEY look at the material properties OK FILES SAVE AS moon2, OK RETURN (twice) JOBS RUN SUBMIT1 MONITOR
151
Now go to postprocessing and generate the engineering stress-strain curve (we did this earlier with the original model). We will also save the analysis generated stress-strain curve to an external file for comparison to the test data.
MAIN RESULTS OPEN DEFAULT HISTORY PLOT SET NODES (pick node 8 shown) END LIST COLLECT DATA 14 30 1 <cr> NODE/VARIABLES ADD VARIABLE Displacement Z Force Z cbody2 FIT, RETURN > XY SAVE type moon2.tab
Pick
This last command saves the table to an external file named moon2.tab (.tab is just to remind us that it is table data).
152
To compare the two stress-strain curves we will use MSC.Marc Mentats generalized plotter feature.
UTILS GENERALIZED XY PLOT FIT SHOW IDS = 0
Biaxial Response
Uniaxial Fit
Engineering Strain
153
To compare the two stress-strain curves we will use XMGR. From a terminal window type: xmgr eb_18.data moon2.tab A graphics screen will appear in which the experimental data is shown in black and the analysis generated stress-strain curve is shown in red. Of course, the test data only extends to about 100% strain whereas we performed our analysis out to 200% strain.
154
This problem is in the subdirectory named ./biaxial/big. Geometry is 86 Dia x 2 Thick (mm), 16 Grips around full circumference (22.5 deg). Mesh uses symmetry at X=0, Y=0, and Z=0. Read model from file bi_glue.mud. Grips are modeled as discrete rigid surface, Grips are 10 mm in dia., placed on a 71 mm dia., friction coefficient is infinite. Run analysis with Ogden 3-term model and plot engineering stress-strain, compare with original test data. Use princa.f usersub if possible.
155
Results: bi_glue.mud
Model 2 Equal-Biaxial
2.0
1.0
0.5
0.0 0.0
0.2
0.4
0.6
0.8
1.0
Engineering Strain
156
157
158
CONTACT CONTACT BODIES DEFORMABLE, OK ELEMENTS ADD (pick top elements) NAME, uniaxial NEW DEFORMABLE, OK ELEMENTS ADD (pick bottom elems) NAME, button NEW RIGID DISCRETE, OK SURFACES ADD (pick z=30 surface) NAME, uni_bot NEW RIGID DISCRETE, OK SURFACES ADD (pick z=43 surface) NAME, uni_top ID BACKFACES
z=43 z=30
z=13 z=0
(Make sure gold side of surfaces touch the deformable brick. If not flip surfaces until this happens, otherwise, continue.)
SAVE
159
CONTACT (contd) NEW RIGID DISCRETE, OK SURFACES ADD (pick z=0 surface) NAME, but_bot NEW RIGID DISCRETE, OK SURFACES ADD (pick z=13 surface) NAME, but_top
(Make sure gold side of surfaces touch the deformable brick. If not flip surfaces until this happens, otherwise, continue.)
EDIT uni_top, OK RIGID VELOCITY PARAMETERS VELOCITY Z=-6 OK (twice)
160
Construct your contact table to look like the one below. Notice that the elements uniaxial touch uni_bot and uni_top, while elements button are glued to but_bot and but_top. All separation forces are zero. Return to the MAIN menu.
Make elements uniaxial touch uni_bot and uni_top, while elements button are glued to but_bot and but_top.
OK, MAIN
161
LOADCASES MECHANICAL STATIC STEPPING PROCEDURE FIXED PARAMETNERS # OF STEPS=12 <cr>, OK (twice), MAIN JOBS MECHANICAL lcase1 ANALYSIS OPTIONS LARGE DISPLACEMENT, OK JOB RESULTS CAUCHY STRESS TOTAL STRAIN, OK OK INITIAL LOADS xsym ysym CONTACT CONTROL INITIAL CONTACT CONTACT TABLE ctable1 OK (3 times) JOBS SAVE RUN SUBMIT1 MONITOR OK, MAIN
162 Experimental Elastomer Analysis
RESULTS OPEN DEFAULT DEF & ORIG SKIP TO INC 12 <cr> PLOT SURFACES WIREFRAME REGEN RETURN
163
Construct time history of Pos Z uni_top vs. Force Z uni_top. This is the true uniaxial response. Construct the same for Pos Z but_top vs. Force Z but_top. This is response that mixes shearing and bulk compression (remember bulk, or hydrostatic, compressive stiffness is many times higher than the shear stiffness)
164
POST PROCESSING HISTORY PLOT COLLECT GLOBAL DATA NODES/VARIABLES ADD GLOBAL CURVE POS Z UNI_TOP FORCE Z UNI_TOP ADD GLOBAL CURVE POS Z BUT_TOP FORCE Z BUT_TOP
165
Objective: To model an elastomeric material under a planar shear stress deformation mode. To focus on curve fitting elastomeric test data, a fully runnable procedure file is provided that will build and (and run) an initial model. However, the model contains only a trivial neo-Hookean material model with C10 = 0.5. It will be your job to modify the model by reading in the test data, and curve fitting it using various material models. In a terminal window, use the cd command to move to the wkshops_A/planar or the wkshops_B/planar directory. Type mentat to start the MSC.Marc Mentat program, then starting from the main menu proceed as follows:
UTILS PROCEDURES EXECUTE pick the file named ps_neo05.proc OK OK
This will produce and run a planar shear stress model. Please familiarize yourself with this model. Look at the BCs, the material specification, the contact bodies and contact table, and the loadcase.
166
After the procedure file is finished, the final picture on your screen will look like this. Here is a brief summary of the planar shear model we have created: A single brick element, full integration, Herrmann. Boundary conditions on y=0 face to prevent free translation in space. Material model is neo-Hookean with C10 = 0.5 Rigid contact surfaces are used to impose deformation. cbody2, cbody4 & cbody5 are stationary. cbody3 is moved so as to impose displacement in the Z direction. Loading is performed in 30 equal time increments. Increment 10 is compression of 50% (compression in Z), increment 30 is extension of 200% (extension in Z). Now lets look at the results of this analysis before curve fitting our planar shear test data.
167
All of the postprocessing functions are accessed from RESULTS, which is located on the topmost MAIN menu. We are especially interested in deformed shape plots and XY plots of stress vs. strain.
MAIN RESULTS OPEN DEFAULT DEF & ORIG SKIP TO INC 10 <cr> CONTOUR BAND SCALAR Displacement Z, OK SETTINGS #LEVELS 5 <cr>, RETURN SKIP TO INC 30 <cr> REWIND MONITOR
168
Now lets generate the stress-strain plot that the MSC.Marc analysis has calculated. When we curve fit the actual test data, this analysis stressstrain curve should match the curve fit response exactly.
HISTORY PLOT SET NODES (pick node 8 shown) END LIST COLLECT DATA 0 30 1 <cr> NODE/VARIABLES ADD VARIABLE Displacement Z Force Z cbody2 FIT, RETURN RETURN
Pick
Since the original area is one, and since the original length in the zdirection is one, this plot is the engineering stress versus the engineering strain. We use the Body 2 force just to get the sign correct. You will usually see this test performed only in tension, but some labs will perform a plane strain compression test.
CLOSE, MAIN
169
170
Since we will be reading more than one set of test data, lets name the datasets. Then make the table type experimental_data, and associate this data with the uniaxial button.
NAME uniaxial TABLE TYPE experimental_data, OK, RETURN UNIAXIAL uniaxial
Repeat the above sequence to read in the file eb_18.data and name this dataset biaxial. Associate this dataset with the biaxial button. Repeat again to read in the file ps_18.data and name this dataset planar. Associate this dataset with the planar shear button.
171
Choose the neo-Hookean curve fitting routine and base the curve fit on all the data. The compute button will compute the model coefficients. By default, responses for many modes are plotted. Turn off the plotting of simple shear.
ELASTOMERS NEO-HOOKEAN USE ALL DATA COMPUTE, OK SCALE AXES PLOT OPTIONS SIMPLE SHEAR, RETURN (this turns off simple shear)
172
Summary of neo-Hookean fits: We have just fit a neo-Hookean model using three sets of data, uniaxial, biaxial, and planar shear. MSC.Marc Mentat by default shows the models response in all major modes of deformation. This is very important. You should always know your models response to each mode of deformation. Compare this plot with the uniaxial only stress-strain plot on (page 111), and the uniaxial+biaxial fit on (page 145). All the plots are very similar. The uniaxial only C10 was 0.265, the uniaxial and biaxial data gives C10 = 0.280, and the fit of all three sets of data simultaneously gives C10 = 0.276. These neo-Hookean coefficients are quite close, telling us that all of the neo-Hookean models are pretty good. We would prefer to use the latest model since it is based on more information and gives a better fit to all the test data. If you can accept the differences between the test data and fitted response, this material model is quite adequate (and stability is guaranteed because the coefficient is positive). For scoping analysis and the initial stage of an analysis, this model is sufficient.
173
Now fit a Mooney 2-term material model. Make sure extrapolation is off. The Mooney coefficients are C10 = 0.244 and C01 = 0.0270. Compare these results to those of the uniaxial+biaxial fit on page 147. There is very little difference in the fit and the coefficients have changed only slightly.
MOONEY(2) COMPUTE, OK SCALE AXES
174
Summary of Mooney 2-term fits: We have now completed a series of Mooney 2-term fits that used progressively more information as the basis for the curve fitting. The table below summarizes the coefficients calculated in each case. The conclusion is that adding biaxial data had a big influence on the quality of the fit and changed the coefficients greatly. Adding the planar shear data did not cause further big changes. Mooney 2-term Fitting Summary Uniaxial Uniaxial + Biaxial Uniaxial+Biaxial+Planar Data Data Shear Data C10 C01 0.074 0.280 0.247 0.027 0.244 0.027
175
Now fit a Mooney 3-term material model. The Mooney coefficients are C10 = 0.239, C01 = 0.035, and C11 = -0.0015. This is essentially the same as the Mooney 2-term material model from the previous page. The third term is almost zero, thus the fit has not changed. One would not choose this model over the Mooney 2-term fit.
MOONEY(3) COMPUTE, OK SCALE AXES
176
Summary of Mooney 3-term fits: We have now completed a series of Mooney 3-term fits that used progressively more information as the basis for the curve fitting. The table below summarizes the coefficients calculated in each case. The conclusion is that adding biaxial data had a big influence on the quality of the fit and changed the coefficients greatly. Adding the planar shear data did not cause further big changes. Mooney 3-term Fitting Summary Uniaxial Data C10 C01 C11 -0.735 1.21 0.194 Uniaxial + Biaxial Uniaxial+Biaxial+Planar Data Shear Data 0.246 0.029 -0.0004 0.239 0.035 -0.0015
Mooney 2-term Fitting Summary Uniaxial Data C10 C01 0.074 0.280 Uniaxial + Biaxial Uniaxial+Biaxial+Planar Data Shear Data 0.247 0.027 0.244 0.027
177
Now fit an Ogden 2-term material model. The fit is similar to the earlier one based on just uniaxial and biaxial data. Indeed, adding the planar shear data has caused the biaxial fit to be worse.
OGDEN COMPUTE, OK
178
Just for fun, try fitting an Ogden 3-term material model to just the uniaxial and planar shear data. You will have to clear the table associated with the biaxial button to do this. The results should look like the figure below. Removing the biaxial data is like removing a constraint. The uniaxial and planar shear response improve quite a bit. However, the biaxial fit response is very bad, with a stability point at about 30% strain.
179
Summary of Ogden 2-term fits: We have now completed a series of Ogden 2-term fits that used progressively more information as the basis for the curve fitting. The table below summarizes the coefficients calculated in each case. We know the uniaxial only data fit had too little information as its basis, and its biaxial response was very bad. The last two fits, however, were relatively similar and yet their coefficients are markedly different. We see this in many Ogden fits and it is attributed to the many local minima that exist in the Ogden equation set. Ogden 2-term Fitting Summary Uniaxial Data
1 1 2 2
180
Comments: We are now finished with the curve fitting portion of this exercise. The further addition of planar shear data did not change the material models very much. Lets run this planar shear analysis with the Mooney 2-term model. Go back to MOONEY(2) and fit it again, press the APPLY button. Submit the analysis, then we will postprocess and show the analysis calculated stress-strain curve.
MOONEY(2) COMPUTE APPLY, OK RETURN (twice) MECHANICAL MATERIALS TYPE, MORE MOONEY (look at the material properties) OK FILES SAVE AS moon2, OK RETURN (twice) JOBS RUN SUBMIT1 MONITOR, OK
181
Now go to postprocessing and generate the engineering stress-strain curve (we did this earlier with the original model). We will also save the analysis generated stress-strain curve to an external file for comparison to the planar shear test data.
MAIN RESULTS OPEN DEFAULT HISTORY PLOT SET NODES (pick node 8 shown) END LIST COLLECT DATA 14 30 1 <cr> NODE/VARIABLES ADD VARIABLE Displacement Z Force Z cbody 2 FIT, RETURN > XY SAVE type moon2.tab
Pick
This last command saves the table to an external file named moon2.tab (.tab is just to remind us that it is table data).
182
For the moment, we shall use the generalized xy plotter to compare the response of the model to the curve fit.
MAIN RESULTS CLOSE, RETURN UTILS GENERALIZED XY PLOT DATA FIT FIT, FILL
183
To compare the two stress-strain curves we will use XMGR. From a terminal window type: xmgr ps_18.data moon2.tab A graphics screen will appear in which the experimental data is shown in black and the analysis generated stress-strain curve is shown in red. Of course, the test data only extends to about 100% strain whereas, we performed our analysis out to 200% strain.
184
This problem is in the subdirectory named ./planar/big. Geometry is 75 L x 12 H x 2 W (mm) between grips. Read model from file pt_45.mud. Grips are modeled as discrete rigid surfaces, with glue. Run analysis with Mooney 1-term model and plot engineering stressstrain, compare with original test data. Use princa.f usersub if possible.
185
Results: pt_45.mud
1.0
0.5
0.0 0.0
0.2
0.4
0.6
0.8
1.0
Engineering Strain
186
187
This will produce a uniaxial stress model. Please familiarize yourself with this model. Look at the BCs, the material specification, the contact bodies and contact table, and the loadcase.
188
After the procedure file is finished the final picture on your screen will look like this. Here is a brief summary of the uniaxial model we have created: A single brick element, full integration, Herrmann. Boundary conditions on x=0 & y=0 faces to prevent free translation in space. Material model is neo-Hookean with C10 = 0.5, no viscoelastic properties are included. Rigid contact surfaces are used to impose deformation. lower rigid body, cbody2, is stationary. upper rigid body, cbody3, is position controlled and moves +0.5 in the Z direction at time zero to achieve 50% strain. Seven loadcases are used to mirror the test data sampling times. This problem is not run in this trivial form since no viscoelastic properties have been added yet. We will now read in the material data and perform the curve fit(s).
189
190
Make the table type experimental_data, and associate this data with the uniaxial button. Your screen should look similar to the one below, and we are ready to start curve fitting the data.
TABLE TYPE experimental_data, OK, RETURN UNIAXIAL table2
191
Choose the neo-Hookean curve fitting routine and base the curve fit on just uniaxial data. The compute button will compute the model coefficients. By default, responses for many modes are plotted. The single neo-Hookean coefficient, C10, is 0.542. Notice that the models uniaxial response does not exactly match the data.
ELASTOMERS NEO-HOOKEAN UNIAXIAL COMPUTE, APPLY, OK SCALE AXES
192
Comments: For simplicity, we have fit a neo-Hookean model using only uniaxial data. All of the previously discussed issues regarding using only one mode of deformation still apply here! We are simply ignoring them for purposes of this exercise. We have used the first data point from the stress relaxation test to define our instantaneous or short time behavior. We could have used data from a separate simple tension test (non-relaxation), but this would add to our uncertainty. Test sample differences (cure, preconditioning, etc.), test strain-rate differences, and other such influences may cause correlation difficulties. We have based our neo-Hookean model on both 30% and 50% strain data. If we wanted near perfect correlation between one test and one analysis, we could have based the neo-Hookean model on just the 50% strain test. Now we are ready to read in one set of relaxation test data, curve fit, and run our uniaxial stress relaxation analysis.
193
For the 50 durometer silicone rubber we have been using in this class, we will perform 2 stress relaxation tests one at 30% strain and at 50% strain. For completeness, we show these two sets of data below.
194
Remember, that a key limitation of this large strain viscoelastic material model is that it assumes the relaxation rate (and thus overall stress relaxation at any time) is independent of the imposed strain. It would be reasonable to check our test data to see if this material satisfies this assumption. We do so by normalizing each dataset (the 30% and 50% strain stress relaxation datasets) and plotting both. This has been done and is shown below. Our 50 durometer silicone rubber satifies this assumption nicely within this range of strain. Q: What to do if your material shows markedly different relaxation rates at different strain levels?
195
Continuing from our previous location in the menu hierarchy, we will now read in one set of stress relaxation data. Choose the 50% strain data (we have set up the analysis for 50% strain loading).
RETURN TABLES READ RAW FILTER: type *.data pick file 50percent.data, OK TABLE TYPE experimental_data, OK, RETURN ENERGY RELAX. table3 ELASTOMERS ENERGY RELAXATION RELAXATION COMPUTE, OK SCALE AXES
We have done this initial fit with the default of two terms in the prony series. This is a pretty crude fit. A rule of thumb is to use as many terms as there are time decades of data. We have 5 decades of data. Re-fit the data using 3, then 4, then 5 terms and watch especially the relaxation time values. Notice that finally you will have a relaxation time value in each decade.
196
The final 5 term prony series fit will look like this. Note the coefficients in the upper right portion of the screen. We are happy with this fit and are ready to APPLY it to the current material definition. From the menu shown below, do the following:
APPLY, OK
197
Comments: We are now finished with the curve fitting portion of this viscoelastic exercise. Lets save our changes to the model and run the analysis.
SAVE MAIN JOBS RUN SUBMIT1 MONITOR OK (when finished)
198
Now go to postprocessing and generate the engineering stress-time relaxation curve. We will also save the analysis generated stress-time curve to an external file for comparison to the test data.
MAIN RESULTS OPEN DEFAULT HISTORY PLOT COLLECT DATA 1 60 1 <cr> NODE/VARIABLES ADD GLOBAL CRV Time Force Z cbody 2 FIT, RETURN > XY RETURN SAVE type visco50.tab
This last command saves the table to an external file named visco50.tab (.tab is just to remind us that it is table data).
199
We can now use the GENERALIZED XY PLOTTER to compare the response with the data.
CLOSE UTILS GENERALIZED XY PLOT DATA FIT (this get the data fit curves)
200
To compare the two stress-strain curves we will use XMGR. From a terminal window type: xmgr 50percent.data visco50.tab A graphics screen will appear in which the experimental data is shown in black and the analysis generated stress-time curve is shown in red (and dashed). You will not see all the text labels. Q: Why is there a difference between the two lines?
201
Q: Why is there a difference between the two lines? A: Recall from (page 197) that the 5 term prony series fit the data extremely well, the fit and data lines were virtually indistinguishable. So why does the MSC.Marc result also not lie directly on top of the test data? The difference is caused by the error in the instantaneous neo-Hookean model. Remember (page 192) that the neo-Hookean model with C10 = 0.542 did not pass exactly through the 50% strain point. This error causes all the difference in the stress-time plot shown on the previous page. To achieve a better correlation of MSC.Marc result to the 50% strain test data, base the neo-Hookean fit on just the 50% strain data. Doing so gives a C10 = 0.554 and the MSC.Marc results will now match the relaxation test data very closely.
202
CHAPTER 7
Contact Analysis
This features allows for the automated solution of problems where contact occurs between deformable and rigid bodies. It does not require special elements to be placed at the points of contact. This contact algorithm automatically detects nodes entering contact and generates the appropriate constraints to insure no penetration occurs and maintains compatibility of displacements across touching surfaces.
203
204
205
Prescribe translational and/or rotational velocity as a function of time using a time table. Prescribe position/rotation as a function of time. Prescribe force on rigid body as a function of time: define force on additional node connect node to rigid contact body motion of rigid contact body is in direction of applied force; motion in perpendicular direction is constrained
206
Contact Procedure
Contact Procedure
Deformable to Rigid Body Contact
Case 1: Contact not detected when u A n < D d Cases 2, 3: Contact detected when u A n d D Case 4: Penetration detected when u A n > D + d u A Rigid Body (set of curves or surfaces)
A
Deformable Body (set of elements)
D D 3 4
Case 1 2
d
with:
u A :incremental displacement vector of node A n : unit normal vector with proper orientation D :contact distance (Default = h/20 or t/4) F s :separation force (Default = Maximum Residual)
Case 1:Node A does not touch, no constraint applied. Case 2:Node A is near rigid body within tolerance, contact constraint pulls node to contact surface if F < Fs . Case 3:Node A penetrates within tolerance, contact constrain pushes node to contact surface. Case 4:Node A penetrates out of tolerance and increment gets split (loads reduced) until no penetration.
Experimental Elastomer Analysis 207
Bias Factor
Bias Factor
By default, the contact tolerance is equally applied to both sides of a segment; this can be changed by introducing a bias factor B ( 0 B 1 ):
D contact D contact D contact ( 1 B ) D contact ( 1 + B )
reduce increment splitting, since the distance to cause penetration is increased improve accuracy, since the distance below which a node comes into contact is reduced:
default
208
Deformable-to-Deformable Contact
Deformable-to-Deformable Contact
Discrete deformable contact (default) is based on piecewise linear geometry description of either 2-node edges in 2 dimensions or 4-node faces in 3 dimensions on the outer surface of all contacting meshes. actual geometry finite element approximation contacting body contact tolerance
y
A
x
contacted body Then the contact constraint: defines tying relation for displacement component of contacting node in local y -direction applies correction on position in local y -direction
209
210
Contact Flowchart
Contact Flowchart
Input Initial set up of contact bodies Incremental data input Check on contact Set up of contact constraints begin increment Apply distributed loads Assemble stiffness matrix; include friction Apply contact constraints Solve set of equations Recover stresses Update contact constraints Split increment No Yes Yes No Converged solution? Yes Separation? No Penetration? No Last increment? Yes Stop
begin iteration
211
Symmetry Body
Symmetry Body
Symmetry bodies often provide an easy way to impose symmetry conditions; they may be used instead of the TRANSFORMATION and SERVO LINK options that WILL CAUSE PROBLEMS in contact as these nodes come into contact. A symmetry plane is characterized by a very high separation force, so that only a movement tangential to the contact segment is possible The symmetry plane option can only be invoked for rigid surfaces
deformable_body
symmetry_plane_1
symmetry_plane_2
none
212
channel
none
MARC element 10
213
billet
channel
none
deformable-rigid (coupled a
channel
MARC element 10
none
MARC element 10
214
Contact Table
Contact Table
3 1 4
Single-sided Contact:
Contact Table
Contact tables must be turned on initially in contact control, or during any loadcase to become active. With no contact tables active, all bodies can come into contact including self contact.
216
Contact Areas
Contact Areas
Very useful for defining certain nodes of a body that may enter contact.
Like contact tables, contact areas must be turned on initially in contact control, or during any loadcase to become active. With no contact areas active, all nodes of all bodies can come into contact. Both contact table and contact areas can reduce the amount of node to segment checking and can save compute time.
217
Options to influence search for contact include: Contact table: define which bodies can potentially come into contact (defined per loadcase) Contact node: define which nodes of a body can potentially come into contact (defined per loadcase) Single-sided contact: searching for contact is not done with respect to bodies with a lower body number (defined for the whole analysis) Exclude: define which segments of a body can never be contacted (defined per loadcase)
218
Standard contact
excluded segments
219
220
221
F n
F n
2 2 u t
with:
F t F n static , F t F n kinetic u t :incremental tangential displacement : slip to stick transition region (default 1 10 : coefficient multiplier (default 1.05) e : friction force tolerance (default 0.05) : small constant, so that 0 (fixed at 1 10 )
6 6
222
with:
C : relative sliding velocity below which sticking is simulated
(Default = 1.0!)
223
However, the friction force depends upon a fraction of the shear strength of the material, not the normal force:
vr Y 2 - t ------ -- atan --- C 3
with:
C : relative sliding velocity below which sticking is simulated
(Default = 1.0!)
224
Nodal forces: (Always used for shell elements) define nodal load F t in opposite direction of relative sliding velocity, F t = F n , where, , friction coefficient and, F n , normal force in contact point
vr Ft Ft
225
Glued Contact
Glued Contact
Sometimes a complex body can be split up into parts which can be meshed relatively easy: * define each part as a contact body * invoke the glue option (CONTACT TABLE) to obtain tying equations not only normal but also tangential to contact segments * enter a large separation force
body 1
body 2
Z
226
Glued Contact
(pure shear) rubber specimen being pulled by two grips. The grip force versus displacement curve is directly available on the post file and can be compared directly to the force and displacement measured.
227
Release Option
Release Option
The release option provides the possibility to deactivate a contact body: upon entering a body to be released, all nodes being in contact with this body will be released. Using the release option e.g. a spring-back effect can be simulated. Releasing nodes occurs at the beginning of an increment. Make sure that the released body moves away to avoid recontacting.
228
This makes rigid bodies useful to monitor the force versus displacement behavior as shown at the right.
Body 3 Force Y
229
230
APPENDIX A
The macroscopic behavior of elastomers depends greatly upon the deformation states because the material is nearly incompressible.
231
Deformation States
Deformation States
L1 L3 L2 t1 t2 t3 1 L1 3 L3 2 L2 t1 t3 t2
Stretch ratios:
L i + L i i = ------------------- = 1 + Li engineering strain = ( L i L i ) =
Incompressibility:
1 2 3 = 1
Eliminate 3 :
1 2 2 1 W = -- G 1 + 2 + ----------- 3 2 2 2
1 2
232
Deformation States
Two-dimensional extension:
F2
L2
F1
dL 2
L1
dL 1
F1
F2
Engineering stresses i : forces per unit undeformed area True stresses t i : forces per unit deformed area
233
Deformation States
Two-dimensional extension:
t1 = 1 ( 2 3 ) = 1 1
or:
t1 = G ( 1 3 )
2 2
and:
t2 = G ( 2 3 ) t3 = 0
2 2
Constant volume implies that a hydrostatic pressure p cannot have an effect on the state of strain, so that the stresses are indeterminate to the extent of the hydrostatic pressure
234
Deformation States
1 K -- , hence --- 2 G
Ordinary solid (e.g. steel): G and K are the same order of magnitude. Whereas, in rubber the ratio of G to K is of the order 4 10 ; hence the response to a stress is effectively determined solely by the shear modulus G
235
236
Invariants of C ij :
I 1 = C ii 1 I 2 = -- ( C ii C jj C ij C ij ) 2 I 3 = det C ij
237
Zero deformation:
W 0 S ij = 2 ------ I 1
0
W + 4 ------I 2
+ 2h ij
hence:
W p = 2 ------I 1
0
W 0 4 ------- 2h I 2
so that the stresses can be expressed in terms of displacements and the hydrostatic pressure
238
Sij Eij dV Qi ui dV Ti ui dA + ( I3 1 ) dV = 0 V A V
N ( Xi )ui
and
p ( Xi ) =
h ( Xi )p
The incremental stresses are related to the linear strain increment by: S ij = D n E kl p ( Cn ) ijkl ij
1
+K
(1)
] [ H ] u [0] p
[ H ]
= PR g
with:
[K [K
(0) (1)
[ H ] : the nodal pressure coupling matrix P : nodal load vector R : internal stress vector g : vector quantity representing the incompressibility constraint
Experimental Elastomer Analysis 239
W ( E ij ,Q ij ) = W ( E ij )
n=1
Q ij E ij
240
where E ij are the components of the Green-Lagrange strain tensor, Q ij internal variables and W 0 the elastic strain energy density for instantaneous deformation. In MSC.Marc, it is assumed that W 0 is the energy density for instantaneous deformations is given by the third order James Green and Simpson form, or the energy function as defined by Ogden. The components of the second Piola-Kirchhoff stress then follow from:
W W S ij = --------- = --------- E ij E ij
0 N
n=1
Q ij
The energy function can also be written in terms of the long term moduli resulting in a different set of internal variables T n : ij
N
W ( E ij, T ij ) = W ( E ij ) +
n=1
T ij E ij
where W is the elastic strain energy for long term deformations. Using this energy definition the stresses are obtained from:
) S ij = W ( E - + ------------------E ij
N
n=1
T ij
241
Observing the similarity with the equations for small strain viscoelasticity the internal variables can be obtained from a convolution expression:
t
T ij =
n . n S ij ( ) exp [ ( t ) ] d 0
where S ij are internal stresses following from the time dependent part of the energy functions.
n S ij
= W ---------E ij
W = W +
n=1
W exp ( t )
Observing the difficulty in finding accurate expressions for the multiaxial aspect of the elastic energy in time independent rubber a further simplification is used. We assume that the energy expression for each term 0 is of similar form to the short time elastic energy W and only different by n 0 a scalar multiplier W = W . This equation can now be rewritten as:
N
W = W +
n=1
W exp ( t )
242
where n is a scalar multiplier for the energy function based on the short term values. The stress strain relation is now given by:
N
S ij ( t ) = S ij ( t ) +
Tij ( t )
n n=1
S ij
T ij =
Analogue to the derivation for small strain visco-elasticity, a recurrent relation can be derived expressing the stress increment as a function of the strain increment and the internal stresses at the start of the increment:
N
S ij ( t m ) = S ij ( t m ) +
n=1
S ij ( t m )
S ij ( t m ) = S ij ( t m ) S ij ( t m )
N
S ij ( t m ) = ( h ) [ S ij ( t m ) S ij ( t m h ) ]
n=1
( h )S ij ( t m h )
243
The functions and are a function of the time step h in the time interval[ t m 1, t m ] :
( h ) = 1 exp ( h ) ( h ) = ( h ) ---h
n n n n n
The equations above are based on the long term moduli. Since in the MSC.Marc program always the instantaneous values of the energy function are given on the MOONEY option, the equations are reformulated in terms of the short time values of the energy function:
N
S ij ( t m ) = 1
N
n=1 n n
[ 1 ( h ) ]
S ij ( t m ) S ij ( t m h )
n=1 n
S ij ( t m h )
n n 0 n
S ij ( t m ) = ( h ) [ S ij ( t m ) S ij ( t m h ) ] ( h )S ij ( t m h )
n n
It is assumed that the visco-elastic behavior in MSC.Marc acts only on the deviatoric behavior. The incompressible behavior is taken into account using special Herrmann elements.
244
W(t) = W +
n=1
W exp ( t )
where W is the energy function for very slow processes. W is an extra amount of energy necessary for time dependent processes. To each amount W n , a characteristic time is associated. At time zero (or for time processes: t < ), the elastic energy reduces to:
N n
W(0) = W = W +
n=1
If we assume that the energy function for each time dependent part is different only by a scalar constant:
W = W
n n 0
W = W +W
n=1
or
N n 0 = 1 W n=1
245
W(t) = W W
n=1 N
+W
n=1
exp ( t )
= W 1
n=1
( 1 exp ( t ) )
= ( 1 )W
0
0 n
W ( t ) = W [ 1 ( 1 exp ( t ) ) ]
246
APPENDIX B
Under repeated application of loads, elastomers undergo damage by mechanisms involving chain breakage, multi-chain damage, micro-void formation, and micro-structural degradation due to detachment of filler particles from the network entanglement. Two types of phenomenological models namely, discontinuous and continuous, exists to simulate the phenomenon of damage.
247
Strain History
For Discontinuous Damage 1.0
0.6
0.4
0.2
0.0 0.0
0.5 Time
1.0
The higher the maximum attained strain, the larger is the loss of stiffness upon reloading. Hence, there is a progressive stiffness loss with increasing maximum strain amplitude. Also, most of the stiffness loss takes place in the few earliest cycles provided the maximum strain level is not increased. This phenomenon is found in both filled as well as natural rubber although the higher levels of carbon black particles increase the hysteresis and the loss of stiffness.
248
determines the evolution of the discontinuous damage. The reduced form of Clausius-Duhem dissipation inequality yields the stress as:
W S = 2K ( , ) --------C
0
Mathematically, the discontinuous damage model has a structure very similar to that of strain space plasticity. Hence, if a damage surface is defined as:
= W0
The loading condition for damage can be expressed in terms of the KuhnTucker conditions:
0 0 = 0
249
The parameters required for the damage model can be obtained using the experimental data fitting option MSC.Marc Mentat. To calibrate the Kachanov factor for the discontinuous damage mode, one measures at a stretch amplitude 0 , the stress level. A loading history is thus:
4 5 3
1
3 2
0 1
time
The model is hyperelastic and assumes that unloading from say state 2 to the undeformed state, and subsequent reloading, occur along the same path. Viscoelastic effects tend to cause the reloading path to reside above the unloading path. Secondary damage effects tend to cause the reloading path to reside below the unloading path. We will now examine the stressstrain plot closely.
250
A procedure to get the discontinuous damage increasing strain table is shown below. The bottom curve is used to compute the damage parameters in MSC.Marc Mentat using a Prony series.
3a na
2a
1 2 n
1 = -- 1
1 = w ia ia
1a
- 1- ia ia ,S i = 2 - 1- ia ia ,S i = 2
1, 2, 3n 1, 2, 3n
1
2
--
1 --
w1 a -------- 1 w1
w 2a --------- 1 w1
wn a --------- 1 w1
251
The results from the analysis show how the damage model works below.
0.4165
0.6
252
1.0
0.8
0.6
0.4
0.2
0.0 0.0
0.2
1.0
This model can be used to simulate fatigue behavior. More realistic modeling of fatigue would require a departure from the phenomenological approach to damage. The evolution of continuous damage parameter is governed by the arc length of the effective strain energy as:
t
------- W s
0
( s ) ds
253
Hence, accumulates continuously within the deformation process. The Kachanov factor K ( , ) is implemented in MSC.Marc through both an additive as well as a multiplicative decomposition of these two effects as:
2
K ( , ) = d +
n=1 2
d n exp ----- + n
m=1
d m exp ----- m
K ( , ) = d +
n=1
+ n d n exp ------------------ n
To calibrate the Kachanov factor for the continuous damage mode, one applies the following loading history to get the input file shown.
1 2 3 4
1 2
tim e
1 2
W1
For the MSC.Marc Mentat implementation, the user needs to know the value of the Free Energy Function at point 1, W1.
1 1 2 2
254
Below is a sample of the continuous damage simulation using a 1-term Ogden model superimposed onto the original data.
Tensile Simulation - Continuous Damage
1-Term Ogden and Original Data
0.8
0.6
0.4
0.2
0.0 0.0
0.2
0.8
1.0
The above damage model is available for deviatoric behavior and is flagged by means of the OGDEN and DAMAGE model definition options. If, in addition, viscoelastic behavior is desired, the VISCELOGDEN option can be included. Finally, a user subroutine UELDAM can be used to define damage functions different from the above.
255
256
APPENDIX C
Elastomeric foams (e.g. rubber foam) are widely used in industry. They exhibit linear elasticity at low stress followed by a long collapse plateau, truncated by a regime of densification in which the stress rises steeply. Furthermore, when loading is compressive, the plateau is associated with the collapse of the cells by elastic buckling. Unlike conventional rubber, foam can undertake large amounts of volumetric compression.
257
Theoretical Background
Theoretical Background
Foams and convention rubber behave differently in tension and compression, with foams have a much larger difference as shown in the figure below:
cr
Plateau (Elastic Buckling) Densification
Elastomer foams are modeled as a compressible Ogden Model with the strain energy density of:
N
W =
n=1
n ----- [ 1 n + 2 n + 3 n 3 ] + n
n=1
n ----- ( 1 J n ) n
258
Theoretical Background
The last term of the strain energy equation is the volumetric change, which can be as high as 90% engineering strain for foams in compression. For i = 0 , there are no lateral effects. For the general theory of isotropic elasticity to be consistent with the classical theory in the linear approximation, the strain-energy function W = W ( 1 , 2 , 3 ) must satisfy:
W ( 1 ,1 ,1 ) = 0 W ( 1 ,1 ,1 ) = 0 i
2
,i = 1, 2, 3 ,( i, j ) = 1, 2, 3
W ( 1 ,1 ,1 ) = + 2 ij i j
Where , and are Lames constants. The initial bulk modulus K and the shear modulus G can be derived from the above as:
1 K = -- i ( i 3 i ) 3
i
1 G = -- i i 2
i
259
Theoretical Background
Blatz and Ko proposed a material model for rubber foams with the strain energy function defined as:
-------------f 1 2 1 2 W = ---- I 1 3 + -------------- I 3 1 + 2 ------------- (1 f) 1 2 1 2 1 ------------------ I 2 3 + -------------- I 3 2 2 2
where:
I1 = 1 + 2 + 3 I2 = 1 + 2 + 3 I3 = 1 2 3
2 2 2 2 2 2 2 2 2
By using the two-term MSC.Marc foam model, the generalized compressible Ogden model can be reduced to the Blatz-Ko model. For temperature effects, the thermal principal stretches follow the temperature and the isotropic thermal expansion coefficient. The thermal principal stretches are defined as:
T i = 1 + T ,i = 1, 2, 3
260
Theoretical Background
The total Lagrange method with conventional elements is used in MSC.Marc for the foam model. The virtual work equation can be formulated as:
Sij Eij dV Qi ui dV T i ui dA V A
= 0
where S ij , is the second Piola-Kirchhoff stress tensor, E ij is the GreenLagrange strain tensor, Q i is the body force per unit undeformed volume, and T i is the prescribed surface tractions per unit undeformed area. All elements in MSC.Marc except Herrmann elements and be used in the foam model.
261
J 0 = i ---- 1
i
1 -2
The specimen should be measured at different load levels. This makes a table of stress, strain, and cross sectional area for these load levels.
262
APPENDIX D
263
Abstract
Abstract
Constitutive models for hyperelastic materials may require multiple complimentary strain states to get an accurate representation of the material. One of these strain states is pure compression. Uniaxial compression testing in the laboratory is inaccurate because small amounts of friction between the specimen and the loading fixture cause a mixed state of compressive, shear, and tensile strain. Since uniaxial compression can also be represented by equibiaxial tension, a test fixture was developed to obtain compressive strain by applying equibiaxial tensile loads to circular sheets while eliminating the errors due to friction. This paper outlines an equibiaxial experiment of elastomeric sheets while providing analytical verification of its accuracy.
264
Introduction
Introduction
Constitutive models for hyperelastic materials are developed from strain energy functions and require nominal stress vs. nominal strain data to fit most models available. In general, it is desirable to represent the three major strain states which are: uniaxial tension, uniaxial compression, and pure shear. If compressibility is a concern, then bulk compressibility information is also recommended. The uniaxial tension strain state is easily obtained and the pure shear test can be performed using a planar tension test with excellent, repeatable accuracy. However, the uniaxial compression test is difficult to perform without introducing other strain states that will affect the accuracy. The main cause of the inaccuracy is the friction between the specimen and the loading platens. The friction can also vary as the compressive load (normal force) increases. To characterize the friction effect, an analysis of a standard ASTM D395, type 1 button under uniaxial compression loading was performed. A plot of compressive stress vs. compressive strain with varying coefficients of friction shows the variation caused by friction (see Attachment A: Compression Analysis on page 278). The analysis of the standard button indicates that for small levels of friction the deviation from the pure uniaxial compressive strain state causes significant errors. An equibiaxial testing fixture is examined to determine if a pure compressive strain could be obtained accurately because an equibiaxial tension state of strain is equivalent to an uniaxial compressive strain.
265
Introduction
The equibiaxial straining apparatus described in this paper also has other advantages with respect to specimen availability and load control. These advantages include: 1. Achieving a strain condition equivalent to simple compression while avoiding the inherent experimental errors associated with compression. 2. Being able to perform strain and load control experiments as well as look at equilibrium behavior. 3. Testing on readily available test slabs. 4. Performing a test at the loading rates equivalent to tension and shear loading rates. Several other experimental approaches to the biaxial straining of elastomers have been developed. In general, two approaches have been used. The first involves the expansion of a thin elastomer membrane using air pressure. Strain control is difficult to obtain with this procedure making it difficult to create conditions that compliment the other strains states required to get a full set of data for fitting hyperelastic constitutive equations. The other problem is that the thickness of the sheets needs to be much thinner than the typical sheet thickness that is created. The second approach involves the gripping of a rectangular specimen around the perimeter and stretching the specimen with multiple arms or cable bearing systems. This approach has been used with great success by several investigators. Difficulties arise with the measurement of strain and the calculation of stress. The advantage of this approach is that while somewhat complex, it allows the
266 Experimental Elastomer Analysis
Overall Approach
investigator to examine elastomer deformation in unequal biaxial deformation states. Since the objectives herein do not involve the need for unequal biaxial straining, the mechanical aspects of the experimental approach can be greatly simplified and the relations between forces and stresses in the specimen can be ascertained with greater certainty by restricting the apparatus to equal biaxial straining.
Overall Approach
The overall approach is to strain a circular specimen radially. Constant stress and strain around the periphery of the disk will create an equibiaxial state of stress and strain in the disk independent of thickness or radial position.
267
In the apparatus, 16 small grips mechanically attach to the perimeter of an elastomer disk using spring force attachment. The grips are moved radially outward by pulling with thin flexible cables which are redirected around pulleys to a common loading plate (Figure 1 on page 264). When the loading plate is moved all of the attachment points move equally in a radial direction and a state of equal biaxial strain is developed in the center of the disk shaped specimen, Figure 2.
268
The Specimen
The actual shape of the specimen is not a simple disk as shown in Figure 3. There are radial cuts introduced into the disc specimen so that there are no tangential forces between the grips. This is necessary because the grips are not attached to the outer edge of the specimen. They are attached to the top and bottom surfaces of the specimen which does not allow material to flow within the grip. Small holes are introduced at the ends of the radial cuts so that the specimen is less likely to tear.
269
Strain Measurement
The relationship between grip travel and actual straining in the center area of the specimen is not known with certainty because of the unknown strain field around the grips and the compliance that may exist in the loading cables and the material flowing from the grips. To determine the strain, a laser non contacting extensometer is used to measure the strain on the surface of the specimen away from the grips.
Force Measurement
The total force transmitted by the 16 grips to the common loading plate is measured using a strain gage load cell. Relating Force Measured to Stress The nominal equibiaxial stress contained inside the specimen inner diameter (Di) is calculated as follows:
= F ( D i t )
where: Di = Diameter as measured between punched holes F = Sum of radial forces t = Original thickness = Engineering stress
270
Analytical Verification
Analytical Verification
Once the closed form solution has shown that a circular disk pulled with a uniform circumferential load produces a biaxial stress and strain field we then need to verify that pulling the disk from 16 discrete grip locations is an acceptable approximation. The following analytical procedure will examine the effects of the boundary conditions imposed by the experimental approach on the ideal closed form solution. The experimental aspects of concern are: A. The specimen is not gripped continually around the circumference. B. Cuts are introduced between the grips that alter the strain field. C. The relationship between force and stress is based on the inside diameter indicated in Figure 3. First finite element analysis is used to verify the closed form solution on a representative specimen model. The following steps will show how the proposed specimen will be compared to the closed form solution.
271
Analytical Verification
The disk specimen finite element model used to verify the closed form solution is shown in Figure 4. Radial loads are applied at every node around the perimeter.
Figure 4. FEA model of uncut specimen with radial loads applied at every perimeter node.
The nominal finite element stress calculated within each element was compared to the stress calculated with the formula below and found to be equivalent.
= F ( Dt )
where: D = Original outside diameter F = Sum of radial forces t = Original thickness = Engineering stress This formula can now be used in a testing environment since all the parameters are known.
272
Analytical Verification
The next step needs to show that using a cut specimen with 16 grips (FEA model shown in Figure 5) will accurately represent the ideal loading condition of the previous finite element analysis.
Figure 5. FEA model of specimen with slits and punched holes, radial loads applied at 16 grip locations.
The original outside diameter used in the above stress formula will be equal to the diameter measured at the inside edges of the punched holes at the ends of the radial slits between the grips. For the proposed configuration this dimension is 50 mm.
273
Analytical Verification
A nominal stress vs. nominal strain comparison of this configuration vs. FEA closed form results is shown for two hyperelastic material representations.
274
Analytical Verification
The first (Figure 7) represents a simple 2nd order polynomial approximation and the second (Figure 8) represents an Ogden 5-term approximation. Both show excellent correlation between the proposed test configuration and the theoretical results.
275
Analytical Verification
Summary
The equibiaxial experiment as proposed in this paper does an excellent job of obtaining the pure strain state required for hyperelastic constitutive models. The error due to the boundary condition approximations are small but consistent as opposed to the uniaxial compression test where the experimental error depends on friction which is unknown and varies as a function of the test material and the normal force. The testing done in this manner can provide excellent consistent and accurate compression strain states while using standard ASTM slabs and a minor amount of specimen preparation to perform.
276
References
References
1. Kao, B. G. and Razgunas, L.,On the Determination of Strain Energy Functions of Rubbers, SAE Paper 860816, (1986) 2. Treloar, L. R. G., Stresses and Birefringence in Rubber Subjected to General Homogeneous Strain, Proc. Phys. Soc., London, 60, 135-144 (1948) 3. Rivlin, R. S. and Saunders, D. W., Large Elastic Deformations of Isotropic Materials, VII, Experiments on the Deformation of Rubber, Phil. Trans. Roy. Soc., London, 243 (Pt. A), 251-288 (1951) 4. Zapas, L. J., Viscoelastic Behaviour Under Large Deformations, J. Res. Natl. Bureau of Standards, 70A (6), 525-532 (1966) 5. Blatz, P. J. and Ko, W. L., Application of Finite Elastic Theory to the Deformation of Rubbery Materials, Trans. Soc. Rheol., 6, 223-251 (1962) 6. Ko, W. L., Application of Finite Elastic Theory to the Behavior of Rubberlike Materials. PhD Thesis, California Ins. Tech., Pasadena, California (1963) 7. Hutchinson, W. D., Becker, G. W. and Landel, R. F., Determination of the Strain Energy Function of Rubberlike Materials, Space Prams Summary No. 37-31, Jet Propulsion Laboratory, Pasadena, California, IV, 34-38 (Feb. 1965) 8. Becker, G. W., On the Phenomenological Description of the Nonlinear Deformation Behavior of Rubber-like High Poymers, Jnl Polymer Sci., Part C (16), 2893-2903 (1967) 9. Obata, Y., Kawabata, S. and Kawai, H., Mechanical Properties of Natural Rubber Vulcanizates in Finite Deformation, J. Polymer Sci. (Part A-2), 8, 903-919 (1970) 10. Burr, A., Mechanical Analysis and Design, Elsevier, New York, 1981, p.315 11. Timoshinko, S.P., Goodier, J.N., Theroy of Elasticity, p 69, 3rd Ed, McGraw hill, New York, 1951 12. ABAQUS v5.8 Users Manual Vol. 1, Section10.5.1
Experimental Elastomer Analysis 277
278
A coefficient of friction value of zero corresponds to a perfect state of simple uniaxial compression (Figures A1 and A2). From the analysis, one can conclude even very small levels of friction significantly effect the measured stiffness and this effect is apparent at both low and high strains.
279
280
APPENDIX E
ACE/gr is a 2D plotting tool for X Window System. It uses an Motif based user interface, which is the reason why its also known as Xmgr. For more detail see: http://plasma-gate.weizmann.ac.il/Xmgr/
281
Features of ACE/gr
Features of ACE/gr
User defined scaling, tick marks, labels, symbols, line styles, colors. Batch mode for unattended plotting. Read and write parameters used during a session. Regressions, splines, running averages, DFT/FFT, cross/autocorrelation, . . . Support for dynamic module loading. Hardcopy support for PostScript, HP-GL, FrameMaker, and InterLeaf formats. An example of ACE/gr is shown below:
282
Using ACE/gr
Using ACE/gr
The use of ACE/gr or xmgr will be to read in from a file existing xy data (Block Data) and overlay plots. To read in block data click on File, and select Read, then Block Data. This brings up the file browser below:
Here you can select the data you have stored from test data or MSC.Marc Mentat history plots. Lets suppose that we have two Block Data files that look like:
file1 0 1.66667 3.33333 5 1 3.77778 12.1111 26 0 1.66667 3.33333 5 file2 1.1382 3.39864 10.1483 30.3025
283
Using ACE/gr
Using the file browser, select file1 and identify from which column you want x and y to come from in the menu below:
Clicking Accept will bring in the first curve then autoscale by picking the icon below:
284
Using ACE/gr
Y-Axis Area
X-Axis Area To place symbols on the plot, simply click on a curve and select a symbol desired. To place a Title or Axis Labels, click in the Title area or Axis area and fill in the menu.
285
Menus:
286
Axis Summary:
Symbol Summary:
287
Log Plots:
Bar Charts:
288
APPENDIX F
The purpose of this appendix is to provide pages for notes and the course critique.
289
Notes
Notes
290
Notes
Notes
291
Notes
Notes
292
Course Critique
Course Critique
Please use this form to provide feedback on your training program. Your comments will be reviewed, and when possible included in the remainder of your course.
Lecture Materials Is the level of technical detail appropriate? Are the format and presentation correctly paced? Are the discussions clear and easy to follow? What changes do you suggest? excellent average poor
What additional information would you like? ____________________________________________________________________________ ____________________________________________________________________________ Workshop Are the available problems relevant? Was the technical assistance prompt and clear? Was the equipment satisfactory? What changes do you suggest? excellent average poor
What additional information would you like? ____________________________________________________________________________ ____________________________________________________________________________ Laboratory Are the available specimens relevant? Was the technical assistance prompt and clear? Was the equipment satisfactory? What changes do you suggest? excellent average poor
What additional information would you like? ____________________________________________________________________________ ____________________________________________________________________________ General How would you change the balance of time spent on theory, workshop, and laboratory
no change
more theory
more workshop
more laboratory
293
Course Critique
294