What is CRIMS?

Study on effect of CRIMS data on warpage simulation
and possibility of using supplement CRIMS data
Speaker(s) Venkatesh, Aungadu Kuppuswamy – Motorola Solutions Inc
Code
SM2685-P
Class Title
Study on the Effect of CRIMS Data on Warpage Simulation and the Possibility of Using Supplemental Data
Description
This class has two parts. First, we will review a study on using corrected residual in-molded stress (CRIMS)
data to predict warpage. In this study, we selected five parts of increasing part design complexity and material
flow path. We simulated five different materials with various process parameters. From the simulated data, we
selected high-warp scenarios for molding and warpage and compared the results with CRIMS and without
CRIMS. The study showed the importance and effect of warpage. Statistically, we found that using CRIMS to
predict warpage showed a 40 times better probability of being accurate and a 24% improvement in accuracy.
Next, we will look at using supplemental data. CRIMS data is specifically measured/generated by Autodesk®
Simulation Moldflow® software by molding trails. The test method is expensive. Therefore, for limited materials,
we can supplement CRIMS with equivalent material in the existing Moldflow data base. The experiment shows
the effectiveness of method.
Learning Objectives
At the end of this class, you will be able to:

What is CRIMS

How it is important for thin walled parts warpage simulation

How can you possibility supplement CRIMS
About the Speaker
Venkatesh, Aungadu Kuppuswamy, alias AK ,Senior Staff Materials Engineer, Motorola
Solutions Inc graduated from University of Massachusetts, Lowell, MA with Master in Plastic
Engineering in 2005. Prior to his tenure at Motorola solutions, he worked as Manufacturing
Engineer at Freudenburg-NOK, Manchester, NH and Assistant Manager-Component
Development at Bajaj(Kawasaki) Auto Ltd, Pune, India. Venkatesh graduated with Bachelors
degree in Mechanical Engineering from University of Madras, India and Post Graduate Diploma
in Plastic Engineering from Central Institute of Plastics Engineering and Tooling Center
(CIPET), Chennai, India. Venkatesh currently serves as the Subject Matter Expert at Motorola
Solutions Inc for Plastic Engineering related areas and especially for Autodesk Moldflow
products. This class presentation was part of Venkatesh’s digital six sigma black belt project.
Insert Class Title as per Title Page
What is CRIMS?
CRIMS is Correct Residual In-mold stress. Autodesk-Moldflow Lab came up with a correction
factor to improved warpage prediction. CRIMS data is calculated and calibrated by Moldflow lab.
There are six factors, A1, A2,…and A6. A1 to A3 stand for parallel direction shrinkage and A3A6 stands for perpendicular direction. A1, A2, A3,and A4 are scaling factor, whereas A5 and A6
are shrinkage factors. Below slides are from the class presentation.
Current Status of Moldflow Material Data Base
• MSI uses approximately 70 plastics materials
• Remaining materials have no CRIMS data
• Only “some“ materials have CRIMS data
Lexan EXL 1433T
Lexan EXL 1414
• Cost for CRIMS data testing is expensive.
• Testing time per batch of 4 materials is 6-8 weeks
37
© 2012 Autodesk
2
Insert Class Title as per Title Page
What is CRIMS
CRIMS = Corrected Residual In-Mold Stress
Moldflow Simulation uses the following material parameters:
1. Viscosity
2. PVT
3. Thermal
conductivity
4. Specific heat
capacity
5. Shrinkage (CRIMS)
•A1, A2 and A3 coefficients modify the parallel shrinkage
A4, A5 & A6 modify perpendicular shrinkage
• A1, A2, A4 and A5 are scaling factors, where as A3 and A6 are
shrinkage values
© 2012 Autodesk
How CRIMS is important for thin walled parts warpage simulation
If a plastics part’s thickness is less than 3 mm, its considered to be suitable for dual domain
and midplane method of warpage simulation. . For thick parts, say 3mm and above, 3D
warpage simulation will be suffice.
In dual domain and midplane method ,thickness greater than 3mm cannot be well represented
in flow equation of Moldflow software,. therefore warpage calculation is not accurate
For thin wall part say less than 3mm, dual domain and midplane is recommended. CRIMS is
used dual domain and midplane mesh only. Make sure while doing dual domain warpage
simulation the selected material has CRIMS data in the database. Please see the class
presentation to understand more.
I used 5 part design of increasing complexity and five different material to show that, CRIMS
method could yield better prediction over “without CRIMS” method.
3
Insert Class Title as per Title Page
Regression Analysis
Scatterplot of No Crims, Crims vs Actual
1.2
Variable
No C rims
C rims
1.0
Y-Data
0.8
0.6
0.4
0.2
Ideal Condition
0.0
0.0
0.2
0.4
0.6
0.8
1.0
A ctual
1.2
1.4
1.6
1.8
© 2012 Autodesk
Box Plot
Boxplot of delta no crims, delta crims
1.25
Ideal Condition
1.00
0.75
Data
0.50
0.25
0.00
-0.25
-0.50
delta no crims
delta crims
© 2012 Autodesk
4
Insert Class Title as per Title Page
Probability of Good Prediction
One-Sample T: delta no crims
Test of mu = 0 vs not = 0
Variable
N Mean StDev SE Mean
95% CI
T
P
delta no crims 18 0.3003 0.4177 0.0985 (0.0926, 0.5080) 3.05 0.007
One-Sample T: delta crims
Test of mu = 0 vs not = 0
Variable
N Mean StDev SE Mean
95% CI
T
P
delta crims 18 0.0566 0.2897 0.0683 (-0.0874, 0.2007) 0.83 0.41824
There is a 40% higher probability of getting accurate
predictions by using CRIMS
© 2012 Autodesk
How can you possibility supplement CRIMS
I used physical properties, flow properties, manufacturer and family to find a suitable material
with CRIMS-data for material which did not have CRIMS. I ran simulation using the
supplemented CRIMS-data in warpage simulation. Results showed that the warpage accuracy
is better than no-CRIMS and similar to 3D method of simulation. Below slides are from the
class presentation.
5
Insert Class Title as per Title Page
Box Plot and Test of equal Variance – New part
Test and CI for Two Variances: abs-del-nocrims, abs-del-Sup
Statistics
Variable
N
abs-del-nocrims 9
abs-del-Sup
9
StDev
0.132
0.053
Variance
0.017
0.003
Ratio of standard deviations = 2.500
Ratio of variances = 6.250
95% Confidence Intervals
CI for
Distribution CI for StDev
Variance
of Data
Ratio
Ratio
Normal
(1.187, 5.264) (1.410, 27.706)
Continuous (0.478, 6.584) (0.229, 43.347)
Looking at the standard deviation of No CRIMS
shows that the data has unacceptability high
variability, hence we are discarding no CRIMS
method.
Test
Method
DF1 DF2 Statistic P-Value
F Test (normal)
8 8
6.25 0.018
Levene's Test (any continuous) 1 16
2.35 0.145
Levene test did not detect difference.
45
© 2012 Autodesk
6
Insert Class Title as per Title Page
Test and CI for Two Variances: abs-del3D, abs-del-Sup
Method
Null hypothesis
Sigma(abs-del3D) / Sigma(abs-del-Sup) = 1
Alternative hypothesis Sigma(abs-del3D) / Sigma(abs-del-Sup) not = 1
Significance level
Alpha = 0.05
Tests
Test
Method
DF1 DF2
F Test (normal)
8
8
Levene's Test (any continuous)
1
16
This shows that we can compare 3D and CRIMS
Statistic P-Value
1.99
0.350
0.70
0.416
Abs-del3D : absolute delta of 3D
Abs-del-Sup : absolute delta of supplemented CRIMS
Test of Variance between CRIMS and 3D for new part design
P-Value is great than 0.05, hence 3D
and sup-CRIMS are identical
46
© 2012 Autodesk
Anova to compare 3D and CRIMS for new part design
One-way ANOVA: abs-del3D, abs-del-Sup
Source DF
SS
MS F
P
Factor 1 0.01192 0.01192 2.86 0.110
Error 16 0.06674 0.00417
Total 17 0.07865
S = 0.06458 R-Sq = 15.15% R-Sq(adj) = 9.85%
Grouping Information Using Tukey Method
N Mean Grouping
abs-del3D 9 0.14369 A
abs-del-Sup 9 0.09223 A
Means that do not share a letter are significantly different
Abs-del3D : absolute delta of 3D
Abs-del-Sup : absolute delta of supplemented CRIMS
ANOVA analysis shows no statistical difference between
3D and supplemented-CRIMS
47
© 2012 Autodesk
7
Insert Class Title as per Title Page
Box plot for new part design
Abs-del-nocrims: absolute delta of noCRIMS (Original)
Abs-del3D : absolute delta of 3D
Abs-del-Sup : absolute delta of supplemented CRIMS
Note :Ideal response is zero
•From above data it is clear that the original method is less precise than proposed
method. This confirms with results from project1: CRIMS methods is 29% more accurate
than no-CRIMS
48
© 2012 Autodesk
If you have questions, please feel free to a send email to [email protected].
8