from localized damage modeling to regional collapse risk ... · (coleman and spacone 2001) concrete...
TRANSCRIPT
From Localized Damage Modeling to Regional Collapse Risk:
Advances in Performance Assessment of Reinforced Concrete Structures Under
Extreme EventsMaha Kenawy, Ph.D.
Postdoctoral ScholarDepartment of Civil and Environmental Engineering
University of Nevada, RenoJanuary 16, 2020
1
Acknowledgments
Supervisor: David McCallenUniversity of Nevada, RenoFinancial Support:Exascale Computing Project, Department of Energy
Previous work as Ph.D. studentFormer advisors: Sashi Kunnath, Amit KanvindeUniversity of California, DavisFinancial support: NSF Grant CMMI 1434300
2
Challenges facing simulation-based seismic performance assessment of structures
• Limitations of nonlinear structural analysis models
• Sparsity of observations on variability of earthquake ground motions across a region
3
Motivation• Assessing the seismic safety of
structures• Improving building code provisions
Key Issues
Taiwan 2018earthquake (Skynews)
Challenges facing simulation-based seismic performance assessment of structures
4
RTR variability
Post-capping stiffness
(Ibarra and Krawinkler 2005)
Contribution of uncertainty in system parameters to variance of
collapse capacity
Ductility capacity
Ground motion attribute
Structural component model attributes
Regional scale
Material and component scales
Quantifying variability is the main objective at the regional scale
5
Developing a detailed understanding of the regional scale, site-specific variation of earthquake risk to RC buildings
Using high resolution physics-based ground motions generated by the DOE project: High Performance, Multidisciplinary Simulations for Regional Scale Earthquake Hazard and Risk (EQSIM)
Tool Objective
(Rodgers et al., 2019; Rodgers, Pitarka and McCallen, 2019)
M7 fault rupture M7 fault rupture
Complex spatial distribution of structural risk over the near-fault region
0 20 40 60 80 100
X Coordinate (km)
0
20
40
Y C
oord
inat
e (k
m)
0.5
1.5
2.5
4.0
Max
imum
Inte
rsto
ry D
rift (
%)
6
0 20 40 60 80 100
X Coordinate (km)
0
20
40
Y C
oord
inat
e (k
m)
0.5
1.5
2.5
4.0
Max
imum
Inte
rsto
ry D
rift (
%)
12 story moment frame 3 story moment frame
Maximum interstory drift varies by a factor of 8 along the fault
Structural risk due to a M7 fault rupture
7
0 20 40 60 80 100
Distance Parallel to the Fault (km)
0
2
4
6
8
Max
imum
Inte
rsto
ry D
rift (
%)
2
4
6
8
10
12
14
16
18
20
Dis
tanc
e N
orm
al to
Fau
lt(km
)
0 20 40 60 80 100
Distance Parallel to the Fault (km)
0
2
4
6
8
Max
imum
Inte
rsto
ry D
rift (
%)
22
24
26
28
30
32
34
36
38
Dis
tanc
e N
orm
al to
Fau
lt(km
)
Higher structural demands and variability in the near-fault region
< 10 km normal to fault > 10 km normal to fault
Ground motions with large velocity pulses
Higher structural demands due to forward directivity effects
8
0 1 2 3 4 5 6
Peak Interstory Drift Envelope(%)
1
2
3
4
5
6
7
8
9
10
11
12
Stor
y
Station B
Max drift = 5.6%
0 1 2 3 4 5 6
Peak Interstory Drift Envelope(%)
1
2
3
4
5
6
7
8
9
10
11
12
Stor
y
Station A
Max drift = 1.2%
0 20 40 60 80 100
0.5
1.0
2.5
3.0
Classification of pulse-like ground motions by Shahiand Baker, 2014
Pulse period normalized by the building fundamental period
A B
Predicting component deterioration is crucial for assessing collapse
9
plastic hinges
Plastic hinge model
RTR variability
Post-capping stiffness
Contribution of uncertainty in system parameters to variance
of collapse capacity
(Ibarra and Krawinkler 2005)
Ductility capacity
Component moment-rotation relationship
Cord rotation
Mom
ent
Post-capping
-5 0 5
Drift (%)
-200
-100
0
100
200
Late
ral L
oad
(kN
)
Experiment
Simulation
-5 0 5
Drift (%)
-200
-100
0
100
200
Late
ral L
oad
(kN
)
Experiment
Simulation
How accurate are the underlying component models?
• Plastic hinge models are conservative in the post-peak range
• There is large uncertainty in predicting strength degradation
10
-3 -2 -1 0 1 2 3
Drift (%)
-800
-600
-400
-200
0
200
400
600
800
Late
ral L
oad
(kN
)
Experiment
Simulation
-6 -4 -2 0 2 4 6
Drift (%)
-200
-100
0
100
200
Late
ral L
oad
(kN
)
Experiment
Simulation
Modified Ibarra-Medina-Krawinkler model
Prediction equations:Haselton et al. 2015
Plastic hinge modelExperiment
Key question: how to model localized component damage?
11
• Require component-based calibration
• Model parameters difficult to estimate
• Associated numerical difficulties
• Cannot capture P-M interaction
Mesh-sensitive response in the presence of constitutive softening
Coarse mesh
Coarse mesh
Forc
eDisplacement
Fine mesh
Fine mesh
Common modeling approaches
Plastic hinge model Fiber-section model
Remedy to mesh-sensitivity: adding a length scale to the softening problem
12
Energy-based methods
Adjust input to match the mesh size length scale(Coleman and Spacone 2001)
Concrete compressive fracture energy
𝜎𝜎𝐺𝐺𝑓𝑓𝑐𝑐
𝑢𝑢
Plastic-hinge integration
Impose a fixed length scale(Scott and Fenves 2006)
𝑙𝑙𝑝𝑝𝑝𝑝 = 𝑤𝑤𝑝𝑝 𝑙𝑙𝑝𝑝𝑝𝑝 = 𝑤𝑤𝑝𝑝
𝐿𝐿
𝐼𝐼 𝐽𝐽Linear Elastic
Nonlocal models
Not so common (yet)…• Valipour and Foster
(2009)• Feng et al. (2015)• Zhang and Khandelwal
(2016)• Sideris and Salehi (2016,
2017)• Kenawy et al. (2018,
2020)
A material-based length scale
length over which the damage is measured
𝝈𝝈
𝜺𝜺
strain-softeningAveraging
domain
• Represent localized damage in an ‘averaged’ sense
• Easy to generalize13
�ε(𝑥𝑥) = �𝐿𝐿�𝑤𝑤 𝑥𝑥, 𝑟𝑟 ε 𝑟𝑟 𝑑𝑑𝑟𝑟
Spatial averaging of deformation
Nonlocal model: spatial averaging of deformation overcomes mesh-sensitivity
weight function
Integration points
The predictions of the nonlocal approach are mesh-independent
• Sensitivity of the post-peak response to the member discretization
Nonlocal model:mesh-independent
Conventional approach: mesh-sensitive
14
Quasi-static cyclic analysis
The predictions of the nonlocal approach are mesh-independent
• Sensitivity of the inelastic curvature to the member discretization
Nonlocal model:mesh-independent
Conventional approach: mesh-sensitive
3.3 x 10-4Max curvature
1.2 x 10-36.9 x 10-4 2.2 x 10-4
Max curvature
15
-4 -2 0 2 4
Drift (%)
-200
-100
0
100
200
Late
ral L
oad
(kN
)
Mesh-independent blind agreement with experimental observations
SimulationExperiment
16
Saatcioglu and Grira (1999) BG-3
Soesianawati et al. (1986) No. 3
Tanaka and Park (1990) No. 1
Saatcioglu and Grira (1999) BG-8
-4 -2 0 2 4
Drift (%)
-200
-100
0
100
200
Late
ral L
oad
(kN
)
Ang et al. (1981), No. 3
Kono and Watan-abe (2000), D1N30
How does using nonlocal modeling affect collapse assessment?
Seismic collapse fragility curve
Nonlocal modelMedian collapse
capacity𝑆𝑆𝑆𝑆 (𝑇𝑇1) = 0.5 𝑔𝑔
Local model – 5 elementsNonlocal model
17
0 0.5 1 1.5 2
SA(T1) (g)
0
0.2
0.4
0.6
0.8
1Pr
obab
ility
of c
olla
pse 38%
Probability of collapse
Local modelMedian collapse
capacity𝑆𝑆𝑆𝑆 (𝑇𝑇1) = 0.31 𝑔𝑔
0 5 10 15
Max drift ratio %
0
0.5
1
1.5
SA(T
1) (g
)
0 5 10 15
Max drift ratio %
0
0.2
0.4
0.6
0.8
1
SA(T
1) (g
)
0 0.5 1 1.5 2
SA(T1) (g)
0
0.2
0.4
0.6
0.8
1Pr
obab
ility
of c
olla
pse
Collapse assessment of RC column using plastic hinge and nonlocal models
18
44%
Plastic hinge modelNonlocal model
We do not usually have the answer.
Plastic hinge model5 0 5
Load vs. displacement
Nonlocal fiber-section model
5 0 5
Load vs. displacement
Probability of collapse
Summary
• There is large uncertainty in designing structures to resist earthquakes in both the knowledge of the earthquake loading and the resistance of structures
19
Understanding earthquake loading• Physics-based ground
motions offer opportunities to understand the variability in earthquake hazard and risk in the near-field.
• Quantifying the variability of risk will improve near-fault structural design guidelines.
Simulating structural resistance • Quantification of collapse risk requires
rigorous models to predict degradation.
• Nonlocal models predict strength degradation due to concrete softening.
• Extending nonlocal models to capturing other sources of deterioration is crucial for collapse assessment of RC structures.