Evaluating Structural Engineering Finite Element Analysis Data ...
Post on 27-Dec-2016
218 Views
Preview:
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
top related