evaluating structural engineering finite element analysis data
TRANSCRIPT
Matija Radovic and
Dr. Jennifer McConnell
Evaluating Structural
Engineering Finite Element Analysis Data
Using Multiway Analysis Results
Background
Introduction
Methodology
Conclusion
Finite Element Analysis (FEA) • Common tool in structural engineering • Predicts structural behavior • Based on a discretization of structural parts
into geometric shapes (elements) • The number of elements in a typical model
could vary anywhere from hundreds to millions
Results
Background
Introduction
Methodology
Conclusion
FEA in Current Practice • Only a small fraction of this available data
(such as min. and max. stresses) are quantitatively analyzed
• Big data techniques provide opportunity for more holistic analysis
• Likely to be advantageous for comparing differences in competing design options
Load increments
Stre
sses
(p
si)
Results
Background
Introduction
Methodology
Conclusion
Goal: • To explore the use of multiway data
analysis techniques in analyzing structural engineering FEA output
Scope: • Propose a new procedure for interpreting
FEA data in structural engineering • Propose using multiway method (Tucker3
tensor decomposition) in evaluation of FEA data
• Make recommendations for future use of multiway tools in structural engineering FEA
Results
Background
Introduction
Methodology
Conclusion
Tucker3 Tensor Decomposition • Type of higher order singular value
decomposition • Decomposes 3D array into sets of scores
that describe the data in a more condensed form
Results
Background
Introduction
Methodology
Conclusion
FEA Subject Bridge
Results
Background
Introduction
Methodology
Conclusion
0 LPF5 LPF
10 LPF17
0
6000
12000
18000
24000
30000
36000
0
0.05
0.1
0.15
0.2
0.25
0.3
Results
Background
Introduction
Methodology
Conclusion
Data Preprocessing for Tensor Decomposition, cont.
0 LPF5 LPF
10 LPF17
0
6000
12000
18000
24000
30000
36000
0
0.05
0.1
0.15
0.2
0.25
0.3
psi
0 LPF5 LPF
10 LPF17 L
0
6000
12000
18000
24000
30000
36000
0
0.2
0.4
0.6
0.8
0 LPF5 LPF
10 LPF17
0
6000
12000
18000
24000
30000
36000
0
0.2
0.4
0.6
0.8
G1 BF G4 BF
XG1 XG2
Results
Background
Introduction
Methodology
Conclusion
Data Preprocessing for Tensor Decomposition • To carry out Tucker3 decomposition, the data
must be organized in a 3-way (3-mode) format
• 1st mode- 7 element groups (4 girder groups and 3 cross-frame groups)
• 2nd mode - 51 stress ranges (51 stress histogram bins)
• 3rd mode -17 loading increments (LPFs).
Results
Background
Introduction
Methodology
Conclusion
Determining Appropriate Tucker3 Model Fitting procedure based on: • Percent variance explained
• Optimal model complexity
Results
Background
Introduction
Methodology
Conclusion • High negative scores in Component 1 -narrowest spread in stress distribution
• High positive scores in Component 2 -
widest spread in stress distribution
Results: Element Group Loading Scores
-0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
G1G2
G3 G4
XG1
XG2
XG3
Component 1
Com
pone
nt 2
15 16 14 13 12 17 11 10 9 8 7 6 5 1 2 3 40.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
LPFsResults
Background
Introduction
Methodology
Conclusion
• High positive scores in Component 1 and high negative scores in Component 2 low LPFs
• High positive score in Component 2 high LPFs.
Results: LPF Loading Scores
0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
1
23
4
5
6
78
910
11121314151617
Component 1
Com
pone
nt 2
Results
Background
Introduction
Methodology
Conclusion
Comparing Tucker3 Results to Experimental Results
15 16 14 13 12 17 11 10 9 8 7 6 5 1 2 3 40.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
LPFs
Results
Background
Introduction
Methodology
Conclusion
• Innovative method of FEA data interpretation
• Possible ability to highlight latent behavior
of bridge components subjected to increasing load
• Ability to quantify and differentiate the
stress profiles of different bridge components
Conclusions