Professional Documents
Culture Documents
Attary Et Al 2016 PDF
Attary Et Al 2016 PDF
net/publication/308892982
CITATIONS READS
48 757
5 authors, including:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Andre R. Barbosa on 07 October 2017.
Abstract: Tsunamis affect coastal regions around the world, resulting in fatalities and catastrophic damage to communities. Fragility func-
tions form the basis of most risk and resilience analyses at the individual structure level, thereby allowing physical infrastructure components
to be included at the community level. For tsunami loading, the vast majority of fragilities that have been developed are based on postevent
observations in the field, which are usually specific to the site of the event. In this paper, a methodology to generate physics-based tsunami
fragility functions is proposed, using vector intensity measures, such as tsunami flow depth and flow velocity and several combinations
thereof. The proposed methodology relies on Monte Carlo Simulation for consideration of material uncertainties and includes epistemic
uncertainties in the tsunami force calculation. The ability of different tsunami intensity measures (flow depth, flow velocity, and momentum
flux), which are common in the literature, to predict the response of structures are investigated, and a new intensity measure (kinematic
moment of momentum flux) that represents overturning moment of a structure for tsunami fragility curves is proposed. The methodology
is illustrated using an application example consisting of a steel moment frame structure and fragility functions based on the kinematic
moment of momentum flux are presented and shown to be a better predictor with less epistemic uncertainty. DOI: 10.1061/(ASCE)
ST.1943-541X.0001715. © 2016 American Society of Civil Engineers.
Author keywords: Tsunami fragility; Momentum flux; Moment of momentum flux; Physics-based fragility; Tsunami intensity measure;
Monte Carlo simulation; Structural safety and reliability.
and fill in on all sides, this force can be neglected [FEMA P-646
(2011, 2013) and Dias et al. (2009), used the data on tsunami-
(FEMA 2008a)]. Hydrostatic forces, for the special case of vertical
damaged houses to construct vulnerability curves based on
rectangular wall, can be calculated by using the following equation:
tsunami flow depth. Then, a Monte Carlo Simulation was per-
formed for a typical building subjected to tsunami loading with 1
different inundation depths to generate fragility curves. Sihombing Fh ¼ · ρ · g · B · h2max ð2Þ
2 s
and Torbol (2015), developed synthetic fragility curves for differ-
ent tsunami scenarios using advanced fluid numerical simulations where ρs = fluid density including sediment (usually assumed to be
and fluid-structure-interaction models. Although the fragility 1,200 kg=m3 ¼ 2.33 slugs=ft3 ); g = gravitational acceleration; B =
curves developed from simulation and the empirical ones provide breadth of the building in the plane normal to the direction of flow;
a reasonable estimate of damage to the structures located at a spe- and hmax = maximum water height above the base of the wall at
cific site, they would be applicable only at that specific site and for structure location. Effects of hydrostatic forces on the structure
that specific scenario. are directly related to the depth of the water.
In recent years, research has focused on developing physics- Buoyant force is a vertical hydrostatic force and is equal to the
based tsunami fragility curves (e.g., Park et al. 2012, 2013b; weight of water displaced. Buoyant forces are resisted mostly
Macabuag et al. 2014). In these studies, tsunami fragility curves through weight of the structure and can be calculated using the fol-
were developed by imposing simplified functional forms of the lowing equation:
tsunami forces on structures and performing nonlinear structural
Fb ¼ ρs · g · V ð3Þ
analysis in an effort to bridge the gap between ocean and structural
engineering. Based on fluid mechanics, numerical simulations, and where V = volume of water displaced by the building. Buoyancy
experimental studies [Neelamani et al. 1999; FEMA 2008b; Wilson forces can reduce the capacity of the structure to resist lateral loads
et al. 2009; ASCE 7-16 (ASCE 2016)], it is well understood that (Yeh et al. 2014).
tsunami forces are related to flow depth and velocity; thus, these Hydrodynamic force is similar to the drag force applied in fluid
would be logical intensity measures for fragility curve develop- dynamics and is induced by the flow of water moving at moderate
ment. Nanayakkara and Dias (2016) developed physics-based fra- to high velocity (steady flow), which can be computed as [FEMA
gility curves using inundation depth as the intensity measure and P-646 (FEMA 2008a)]
assumed that the flow velocity was a function of flow depth. It has
been observed from literature that tsunami flow depth is the most 1
Fd ¼ · ρ · Cd · B · ðhu2 Þmax ð4Þ
popular tsunami intensity measure used for fragility development. 2 s
The reason for this is the fact that tsunami flow depth can be de-
termined and estimated relatively easily after an event compared to where Cd = drag coefficient; B = breadth of the building in the
estimating the flow velocity, which usually requires more detailed plane normal to the direction of flow; h = flow depth; and u = flow
numerical simulations or recorded videos (e.g., Ngo and Robertson velocity. The term hu2 is the momentum flux per unit mass, and
2012). However, tsunami velocity measurements are now available ðhu2 Þmax is the maximum momentum flux per unit mass at any time
from the analysis of survivor videos from 2011 Great East Japan during the tsunami inundation at the site. Impulsive forces are
Earthquake and Tsunami (e.g., Ngo and Robertson 2012). Charvet caused by the leading edge of a surge of water impacting structure,
et al. (2015) looked at different possible tsunami intensity measures which can be estimated using the following equation:
for empirical fragility functions considering flow velocity. A few Fimp ¼ α · Fd ð5Þ
recent studies [e.g., Park et al. 2013a, 2014; ASCE 7-16 (ASCE
2016)], have suggested the use of momentum flux (combination where α = amplification factor with a maximum value of 1.5, based
of tsunami flow depth and velocity) as an intensity measure for on studies by Ramsden (1996) and Arnason (2005). Hydrodynamic
tsunami fragilities (because the largest component of tsunami force, forces and impulsive forces are assumed to be distributed uniformly
i.e., the hydrodynamic force and impulsive force are, in theory, lin- along the height of the building.
early related to the momentum flux). Debris impact forces are difficult to calculate and can be related
In this paper, a methodology to develop physics-based structural to maximum flow velocity but also depend on possible debris in the
fragility functions based on vector-valued intensity measures (com- region and can be estimated as
binations of tsunami flow depth and flow velocity) is proposed.
pffiffiffiffiffiffiffiffiffiffi
In addition, the ability of different tsunami IMs that are popular Fdeb ¼ Cm · umax · k · m ð6Þ
in literature, to predict the response of the structure are investigated,
and finally, an alternate IM is proposed. Different sources of un- where Cm = added mass coefficient (usually taken as Cm ¼ 2);
certainty, affecting both the structural modeling and the tsunami umax = maximum flow velocity carrying the debris; and m and
loading are considered. The proposed methodology is illustrated k = mass and the effective stiffness of the debris, respectively.
using a three-story moment frame steel structure. A Monte Carlo In order to generate tsunami fragility curves, several researchers
1
F¼ · ρ · Cd · B · Co :ðhu2 Þ ð7Þ
2 s
Modeling Uncertainties
There are many uncertainties associated with the process of esti-
mating the response of a structure subjected to tsunamis. Many
of these uncertainties are considered in the tsunami hazard analysis
phase, e.g., fault location, magnitude of the earthquake, length and
direction of the rupture, wave path along the ocean, shore slope,
and surface roughness. The statistical characteristics of the tsunami
flow depth or flow velocity at the location of the structure can be
determined by performing a tsunami hazard analysis. For a given
flow depth (h) and flow velocity (u), the total tsunami lateral force
(F) can be estimated using Eq. (7). In this study, the fluid density,
ρs , is treated as a deterministic variable because it does not vary
much (less than 10%) for a given location. The drag coefficient,
Cd , depends on the aspect ratio of the structure and is estimated
to reach a maximum of 2 for square-shaped structures. The drag
coefficient is assumed to follow a uniform probability distribution
within a range of 1.25–2 [ASCE 7-16 (ASCE 2016)]. The same
methodology can be used for α, the factor that would relate the
impulsive force to the hydrodynamic force. The deamplification
factor, Co , which accounts for the variability in the wall openings
Fig. 1. Proposed methodology to develop physics-based fragility func-
and possible wall/window collapse/breakage, is assumed to fol-
tions for structures subjected to tsunami
low a uniform probability distribution in the range of 0.7–1.0
(Fukuyama et al. 2011). Of course, all of these can be adjusted for
other structure types and conditions, but for typical commercial
building, these values are proposed. For each hj , a set of tsunami flow velocities (uk;j ) is generated con-
sidering maximum desired tsunami flow velocity (U max ).
To develop general fragility functions for tsunami hazard, both h
Proposed Methodology and u should be considered as independent variables. This may not
This methodology employs Monte Carlo Simulation for developing be the case, but there are currently no studies with enough evi-
fragilities for structures subjected to tsunami hazard. The uncertain- dence to suggest otherwise. Flow velocity and flow depth are then
ties in structural parameters are accounted for by modeling them as changed incrementally to generate a number of possible combina-
independent random variables (e.g., modulus of elasticity and yield tions of h and u at the location of the structure. A concern that might
force). The proposed methodology is summarized in the flowchart be raised in this process is that based on real events and site char-
in Fig. 1. The first step in the methodology is to generate q struc- acteristics, some of these combinations might not be realistic. In
tural models (Si ), each having random structural properties using order to resolve this issue, the Froude number, which is a dimension-
pffiffiffiffiffiffiffiffiffiffiffi
Latin hypercube sampling (LHS). These q numerical models would less number from continuum mechanics (Frk;j ¼ uk;j = g · hj )
be used to estimate the response of the structure to a particular is applied as constraint. It is observed from historical tsunamis
tsunami force. A set of tsunami flow depths (hj ) will be generated that the Froude numbers for tsunamis are typically less than 2.0
considering a maximum tsunami flow depth (Hmax ), which is se- (Matsutomi and Okamoto 2010). Hence, in this proposed method-
lected based on the height of the structure and maximum tsunami ology, the unrealistic combinations of flow depth, h, and flow veloc-
flow depth observed. It should be noted that the value of Hmax can ity, u, are not considered thereby obtaining only the combinations of
be chosen based on the desired range of flow depth, but as a starting flow depth and flow velocity that have been observed in past events.
point, using the minimum of 9 m (30 ft) (maximum flow depth ob- To calculate the total tsunami force (Fm;j;k ) for each structure
served during Indian Ocean tsunami in 2004) seems to be reasonable. (Si ) in the random set and for each combination of h and u, p
building. In order to estimate the response of the structure to the fine increments should be chosen for h and u or the momentum flux
calculated tsunami force (Fm;j;k ), a nonlinear structural (push-over itself should be incremented as well as flow depth/velocity. Using the
or time-history) analysis is then performed with the height of the four damage states mentioned earlier, the cumulative probabilities of
uniformly distributed loads equal to the selected tsunami flow damage can be calculated. It should be noted that one value of mo-
depth (hj ). Estimating the performance and damage level of the mentum flux might be a result of different combinations of h and u,
structure can be achieved using different damage states available in resulting in a possible dispersion of the response.
the literature (e.g., FEMA 2008b). In this study, maximum inter- Kinematic Moment of Momentum Flux
story drifts (ISD) are used as the engineering demand parameters Momentum flux has been proposed in several recent studies assum-
(EDP) for the damage assessment. Because the tsunami loading on ing it would have a good correlation with the structural response,
the structure is a lateral load similar to the horizontal earthquake with a focus on the total tsunami force. However, although momen-
forces induced from the inertia, the maximum interstory drift would tum flux is estimated to be linearly related to hydrodynamic forces,
give a reasonable estimate of the structural damage. This approach the magnitude of the force alone may not be a good representation
has been used in the development of other physics-based tsunami of tsunami effects on the building other than collapse. In addition to
fragility curves studies (e.g., Park et al. 2012). However, it should the magnitude of the force, the location of the resultant force and
be noted that this study uses ISD as the sole EDP and that possible distribution of the force along the height of the building are important
local failures of columns and column-foundation connections will parameters that may affect the response of the structure. For example,
not be captured. These local failures may even cause the complete one value of momentum flux might be generated from high-flow
collapse of the building and require other types of analysis, such as velocity and low-flow depth or vice-versa, and the response in these
progressive collapse analysis. In addition, new IM and EDP/damage two cases will be completely different. Thus, it would be possible
measures should eventually be defined, which requires further sig- that the data points for tsunami fragilities based on momentum flux
nificant research and implementation work but is felt to be beyond would be scattered and sometimes do not correlate well with the re-
the scope of this study and will be pursued in future studies. The sponse of the structure. Moreover, it is well understood that, by in-
maximum ISD (Drm;j;k ) is then checked against the limit state for creasing the height of the distributed load on a structure, the stiffness
each of the four damage states, namely slight, moderate, extensive and capacity of the structure will decrease. For example, a simple
and complete. This process is repeated for all q randomly generated cantilever column (neglecting the axial force) with height hb , with
tsunami forces, until uk;j and hj reach the maximum considered flow a modulus of elasticity (E), and moment of Inertia (I), subjected to a
velocity (U max ) and flow depth (Hmax ) for all p randomly generated uniform distributed load of w with a variable height of (hw , i.e., water
structures. Finally, by averaging the number of cases exceeding the flow depth), is considered. Maximum deflection of the column and
limit state from each of the four damage state limits (N j;k ) for all p equivalent stiffness of the column can be calculated theoretically.
and q cases (considering structural property uncertainties and tsu- The initial stiffness (kint ) can be calculated as
nami loading uncertainties), a cumulative probability of damage
for each damage state for every tsunami flow depth and flow velocity 24EI
can be estimated. Using these probabilities, fragility functions based kint ¼ ð8Þ
h2w · ð4hB − hw Þ
on the following intensity measures are developed.
Assuming hB ¼ 11.89 m (39 ft), Fig. 2(a), shows the variation
Intensity Measures of yield force with respect to the water flow depth hw . The simpli-
As discussed earlier, there are a number of tsunami intensity mea- fied pushover curves are obtained by assuming a simple bilinear
sures that have been used in past research studies. The progression behavior for the material, as shown in Fig. 2(b). Fig. 2(b) shows
ranging from flow depth, depth-velocity surfaces, momentum how one value of force (e.g., the dotted line shown in the figure)
flux, and finally, the proposed new intensity measure of kinematic can correspond to different flow depths, leading to a wide array of
moment of momentum flux, is examined here. possible responses.
In an effort to explore a new IM related to structural damage, a
Flow Depth combination of flow depth (h) and momentum flux (hu2 ) is pro-
Physics-based tsunami fragilities based solely on tsunami flow posed here to consider both tsunami forces and the loading height.
depth can be generated using the proposed methodology by The proposed new intensity measure, called kinematic moment of
averaging cumulative probability of damage for each depth. As momentum flux (KMMF), is expected to have a better correlation
mentioned earlier, it should be noted that using flow depth as with structural responses because for each combination of the mag-
an intensity measure has been popular in the research literature nitude of force and a particular load height, only one particular re-
(e.g., Park et al. 2012, 2013b; Macabuag et al. 2014), but it should sponse of the structure will exist. By definition, the flux of any
be noted that this IM is the most widely reported from field studies quantity per unit mass (q), across a defined surface, is the integral
following tsunamis. However, for a given flow depth, it is possible across the surface of the mass flux (ρ · u) times q. Assuming q as
to have different flow velocities, hence, different total tsunami the moment of momentum per unit mass (q ¼ z · u), the moment of
Fig. 2. (a) Relation between equivalent yield force of a cantilever column to flow depth; (b) simplified bilinear push-over curves of a cantilever
column based on different flow depths
momentum flux per unit width can be obtained using the following functions with the intensity measures described previously for a
equation (Gupta and Gupta 2012): three-story steel moment frame building. The three-story structure
Z h selected for this study is the one presented in detail in FEMA-355C
MMF ¼ ðρ · uÞ · ðz · uÞdz ð9Þ (FEMA 2000b). For this study, the three-story building, designed
0
for Seattle, Washington, according to the preNorthridge earthquake
where ρ = density; and z = height of water varying from zero at the code (i.e., without consideration of FEMA-267) is considered. The
ground till h (the maximum water height). Assuming that the flow height of the building is 11.89 m (39 ft) and width B ¼ 54.86 m
velocity would be constant along the height, the previous equation (180 ft), and it is assumed the building has its longer dimension
can be simplified as follows: facing the ocean. Some of the details of the building are presented
ρ in Fig. 3 and Table 1 (see FEMA-355C for additional details).
MMF ¼ h2 u2 ð10Þ Three-dimensional (3D) models of the structure were devel-
2
oped using A572 Grade 50 steel for all column, beam, and girder
Dividing the moment of momentum flux by the density, a sections in the design of the building. It is evident from the liter-
parameter called the KMMF can be obtained (h2 u2 =2), which is ature (Ellingwood et al. 1980; Porter et al. 2002; Kim et al. 2011),
a measurable physical quantity and it is a function of flow depth that most of the material property uncertainties for steel structures
(h) and momentum flux (hu2 ), simultaneously. arise from randomness in the yield strength (f y ). In this study, the
yield strength is assumed to follow a lognormal distribution with a
Flow Depth and Flow Velocity
mean value of 379 MPa (55 ksi) and COVof 10% (Kim et al. 2011),
As an alternative to the three IMs mentioned earlier, because
and other parameters, such as modulus of elasticity are assumed to
tsunami forces are related to both flow depth and flow velocity,
be deterministic (2 × 105 MPa ¼ 29,000 ksi). In order to capture
a vector-valued intensity measure can also be considered. The same
the material uncertainty in this study, 100 random structural models
methodology described earlier could be used to generate tsunami
fragility surfaces based on tsunami flow depth and flow velocity. were generated, using LHS, each having unique yield strength,
chosen from 100 random yield strength values.
For each of these 100 structural models, different tsunami flow
Illustrative Case Study: Three-Story Moment Frame depths (hj ), were considered. The total height of the building was
Steel Building 11.89 m (39 ft), and the flow depth for a very large tsunami was
also of a similar range; therefore, H max is assigned to be 11.89 m
The proposed methodologies are applied herein to illustrate the (39 ft) The flow depths (hj ) are incremented every 0.305 m
procedure for development of physics-based tsunami fragility (1 ft) until it reaches Hmax . For flow velocity, based on tsunami
Fig. 3. Floor plans and elevations for model buildings and floor plans showing layout of moment-resisting frames (0.305 m ¼ 1 ft) (adapted from
FEMA-355C)
Distribution
Variable Distribution Parameters Fig. 5. Sample force-displacement results of several push-over
analyses with different distributed applied load height along the height
Cd Uniform 1.25–2
of the building (4.448 kN ¼ 1 kip, 0.305 m ¼ 1 ft)
CO Uniform 0.7–1.0
fy Lognormal Mean ¼ 379 MPa (55 ksi) and COV ¼ 10%
Fig. 4. Sample plastic hinges results for distributed load push-over analysis
Results and Discussions Based on Different instead of incrementing flow velocity, momentum flux itself was
Intensity Measures incremented every 2.13 m3 =s2 (75 ft3 =s2 ) until 851.18 m3 =s2
(30,000 ft3 =s2 ) (400 points). Fig. 7 shows the cumulative probability
of damage as well as the mean and mean – standard devia-
Flow Depth
tion based on the momentum flux for the four damage states. It
Using the methodology explained earlier, fragility curves were is evident from Fig. 7 that although the mean line of the data
developed with flow depth as the sole IM. Fig. 6 shows the has the typical fragility curve trend, large standard deviation ranges
cumulative probability of damage for the four damage states, for the suggest that the cumulative probability data points are still scat-
case study structure, using tsunami flow depth as an intensity mea- tered. However, in comparison with the case of using flow depth
sure. It is clear from the figures that the data points have some scatter as an IM, the results are still much better. It should be noted that
in all damage states. This is because of the fact that most unrealistic each data points shown in Fig. 7 represents 10,000 different cases
combinations of flow depth and velocity were filtered out by impos- (based on 100 different structural models and 100 different tsunami
ing a Froude number constraint, but there are still many possible flow forces) for one combination of flow depth and momentum flux.
velocity values (and tsunami total forces) for each flow depth. In This is a total number of 1.56 × 108 simulations to create a single
order to be able to compare the results, trend lines indicating the plot and compute fragility curves.
mean values for each damage states are also provided in Fig. 6.
For example, in Fig. 6, the probability of exceeding the moderate
damage state for a flow depth of 3.96 m (13 ft) is approximately Kinematic Moment of Momentum Flux
45%. However, the possible flow velocities will be in the range The fragility curves developed for the case study structure based
of umin ¼ 0.305 m=s (1 ft=s) and umax ¼ 12.19 m=s (40 ft=s), on KMMF for the four damage states are shown in Fig. 8. Similar
Fig. 6. Cumulative probability of damage results and mean line for damage states: (a) slight; (b) moderate; (c) extensive; (d) complete, based on
tsunami flow depth as the sole IM (0.305 m ¼ 1 ft)
Fig. 7. Cumulative probability of damage for damage states: (a) slight; (b) moderate; (c) extensive; (d) complete, based on momentum flux (gray lines
show the mean – standard deviation for each damage state) (0.028 m3 =s2 ¼ 1 ft3 =s2 )
Fig. 8. Fragility curve and data points based on kinematic moment of momentum flux for damage states: (a) slight; (b) moderate; (c) extensive;
(d) complete (gray lines show the mean – standard deviation for each damage state) (0.0087 m4 =s2 ¼ 1 ft4 =s2 )
Fig. 9. Fragility surfaces based on tsunami flow depth and flow velocity for damage states: (a) slight; (b) moderate; (c) extensive; (d) complete
(0.305 m ¼ 1 ft, 0.305 m=s ¼ 1 ft=s)
to what was mentioned in the previous section, for obtaining Fig. 9. Fig. 10 shows how these four damage states would overlap,
the results based on KMMF, the same increments for flow depth showing a reasonable trend in the four damage states.
were used and KMMF was incremented every 10.82 m4 =s2 Using these surfaces, an analyst can develop cumulative prob-
(1,250 ft4 =s2 ) until reaching 4,326.83 m4 =s2 (500,000 ft4 =s2 ). abilities for different damage states for any tsunami flow depth or
Data points and the standard deviations shown in the figure suggest flow velocity. Fig. 11 shows several sample cuts through these
that the proposed new intensity measure has a better correlation surfaces for specific tsunami flow depths or flow velocities.
with structural response. It can be observed from Fig. 8 that the In order to use these surfaces more effectively, the 3D fragility
data exhibit less scatter, providing better correlation that was ob- surfaces can be presented in the form of fragility contours as well
served with momentum flux (Fig. 7). (Fig. 12). Since the response of the structure subjected to tsunami
loading is a function of both tsunami flow depth and flow velocity,
any combination of the two factors or any effort to lump their ef-
Flow Depth and Flow Velocity Simultaneously fects will result in a decrease in accuracy. In other words, consid-
Finally, if one considers tsunami flow depth and flow velocity as two ering flow depth as an IM, by using the mean trend line, 100%
independent intensity measures, fragility surfaces can be developed amount of error would be possible in comparison with using h − u
for the illustrative case study building, providing all possible results. fragility surfaces. Using momentum flux as and IM, the data points
The 3D fragility surfaces for the four damage states are shown in and standard deviations shown in Fig. 7 suggest that by using the
mean trend line, a large range (with a maximum of 55%) of error
would be possible for each specific h and u combination. On the
other hand, considering the KMMF as an IM, and again using
the mean trend lines, less error (a maximum of 20%) is observed.
This represents a significant reduction in error compared to the
other two IM’s.
Conclusion
Fig. 11. Fragility curves based on particular tsunami flow depth or flow velocity: (a) flow depth of 4.6 m (15 ft); (b) flow depth of 7.6 m (25 ft);
(c) flow velocity of 4.6 m/s (15 ft=s); (d) flow velocity of 7.6 m/s (25 ft=s), for damage states of slight, moderate, extensive, and complete
(0.305 m ¼ 1 ft, 0.305 m=s ¼ 1 ft=s)
Fig. 12. Fragility contours for damage states of (a) slight; (b) moderate; (c) extensive; (d) complete, based on flow depth and flow velocity (0.305 m
1 ft, 0.305 m=s 1 ft=s)