You are on page 1of 16

Journal of Hydraulic Research

ISSN: 0022-1686 (Print) 1814-2079 (Online) Journal homepage: https://www.tandfonline.com/loi/tjhr20

Measurements of tsunami-borne debris impact on


structures using an embedded accelerometer

Seyedreza Shafiei, Bruce W. Melville, Asaad Y. Shamseldin, Sherif


Beskhyroun & Keith N. Adams

To cite this article: Seyedreza Shafiei, Bruce W. Melville, Asaad Y. Shamseldin, Sherif
Beskhyroun & Keith N. Adams (2016) Measurements of tsunami-borne debris impact on
structures using an embedded accelerometer, Journal of Hydraulic Research, 54:4, 435-449, DOI:
10.1080/00221686.2016.1170071

To link to this article: https://doi.org/10.1080/00221686.2016.1170071

Published online: 10 May 2016.

Submit your article to this journal

Article views: 377

View related articles

View Crossmark data

Citing articles: 17 View citing articles

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=tjhr20
Journal of Hydraulic Research Vol. 54, No. 4 (2016), pp. 435–449
http://dx.doi.org/10.1080/00221686.2016.1170071
© 2016 International Association for Hydro-Environment Engineering and Research

Research paper

Measurements of tsunami-borne debris impact on structures using


an embedded accelerometer
SEYEDREZA SHAFIEI, PhD student, Department of Civil and Environmental Engineering, The University of Auckland, Auckland,
New Zealand
Email: ssha399@aucklanduni.ac.nz (author for correspondence)

BRUCE W. MELVILLE (IAHR Member), Professor, Department of Civil and Environmental Engineering, The University of
Auckland, Auckland, New Zealand
Email: b.melville@auckland.ac.nz

ASAAD Y. SHAMSELDIN, Associate Professor, Department of Civil and Environmental Engineering, The University of Auckland,
Auckland, New Zealand
Email: a.shamseldin@auckland.ac.nz

SHERIF BESKHYROUN, EQC Research Fellow, Department of Civil and Environmental Engineering, The University of Auckland,
Auckland, New Zealand
Email: s.beskhyroun@auckland.ac.nz

KEITH N. ADAMS, Professional Teaching Fellow, Department of Civil and Environmental Engineering, The University of
Auckland, Auckland, New Zealand
Email: k.adams@auckland.ac.nz

ABSTRACT
This paper presents new experimental techniques utilizing a smart debris device for direct measurement of the impact acceleration forces associated
with tsunami-borne debris that impact inland structures. The resulting experimental data will lead to advanced predictive capabilities of such forces
for use in design guidelines. The measured debris acceleration data were used to calculate impact forces (mass × acceleration). An image processing
technique was used to detect the debris impact angle. The debris impact tests were conducted using a disc-shaped smart debris device with masses of
550, 800 and 1000 g. For calculation of the debris force it was found necessary to include the mass of entrained water. The impact acceleration was
found to be a function of debris mass, velocity, and contact duration. An equation is developed to allow estimation of the debris velocity for a known
distance between the debris pick-up location by a tsunami and the structure.

Keywords: Contact duration; debris velocity; impact acceleration; impact angle; mass; smart debris

1 Introduction threat, because field measurement of tsunami debris force is


very difficult. Experimental investigations have included the
Collision of tsunami-borne debris imposes a large force on impact of wooden debris (Haehnel and Daly, 2002, 2004; Nouri,
structures, and this can cause significant damage to inland struc- Nistor, & Palermo, 2010; Riggs et al., 2013; Yeh, Robertson, &
tures during tsunami events (NRC, 2004; Saatçioğlu, 2009; Preuss, 2005) and the impact of ice mass on stationary structures
Saatcioglu, Ghobarah, & Nistor 2006). Proper quantification (Isaacson & McTaggart, 1990; Landweber, Ettema, & Shahsha-
of the forces involved in debris impact is vital for the design han, 1992). The ice-impact studies indicated that the added mass
of structures in tsunami prone areas. Various experimental and of the water entrained by the ice-mass motion is an important
measurement techniques are needed for investigating such a term that must be included in the analysis. In general, the added

Received 20 October 2014; accepted 18 March 2016/Open for discussion until 28 February 2017.

ISSN 0022-1686 print/ISSN 1814-2079 online


http://www.tandfonline.com
435
436 S. Shafiei et al. Journal of Hydraulic Research Vol. 54, No. 4 (2016)

mass depends on the shape of the moving object, its orientation et al. (2012) is for the small time-scale acoustic response of
and its degree of submergence, which is a function of object the impact. They recommended that further research should
density (Hamilton, 2000). consider the longer time-scale gravity waves induced by the
There are three approaches to estimating the impact force impact.
of debris on structures: impulse-momentum, work-energy, and Riggs et al. (2013) reported on full scale in-air axial impact
contact-stiffness. They all consider the mass and velocity of tests using a wood utility pole and shipping container in which
debris, plus an additional parameter depending on the approach: the impact target was a reinforced concrete wall. Piran-Aghl,
the stopping time (contact duration) for the impulse-momentum Naito, and Riggs (2013) reported on the same tests, includ-
approach, the stopping distance for the work-energy approach, ing the results for a square steel tube. Piran-Aghl et al. (2013)
and the effective contact stiffness of the collision for the contact- reported that the impact force increased when non-structural
stiffness approach (Haehnel & Daly, 2002). Haehnel and Daly mass was added to a square steel tube (i.e. the tube’s stiffness
(2004) reviewed the three approaches and concluded that the was unchanged, but the total mass increased), which is contrary
additional parameters are not independent. Hence, the three to the conclusion by Paczkowski et al. (2012). Piran-Aghl et al.
approaches are, in effect, the same. Haehnel and Daly (2002, (2013) also found that the impact force was directly related to
2004) studied floating debris impact in a wave flume, under the impact velocity.
steady flow conditions and using solid wooden logs of different Riggs et al. (2013) and Ko (2013) reported on wave flume
masses. The impact orientation of the logs was either parallel or experiments using 1:5 scale models of a standard 20-foot ship-
transverse to the flow direction. They found that contact dura- ping container to investigate the effect of water on debris impact
tion of the collision depended on the mass of the debris and forces. They used an aluminium model (density 156 kg m−3 ) and
the effective contact stiffness of the collision. Haehnel and Daly an acrylic sheet model (density 145 kg m−3 ). The impact tar-
(2002) concluded that in addition to mass and velocity of the get was a narrow rigid column, supported to make it more rigid
debris, the maximum debris impact forces are affected by the than the debris. The containers’ orientation was controlled by
added mass of water and impact orientation of the debris. The guide wires. Riggs et al. (2013) concluded that the contribution
maximum impact force occurred when the long axis of the log of water (i.e. added mass effect) to the debris impact forces was a
was parallel to the flow direction and normal to the target. They secondary effect, and suggested that the added mass effect could
found that the added mass effect was not significant when the log be neglected in design. Ko (2013) reported that, for longitudi-
was parallel to the flow but was significant for other orientations. nal impact, the aluminium model’s in-water impact forces were
ASCE (2010) and FEMA (2011) design guidelines for flood up to 10% larger than those from the in-air tests and the acrylic
impact loads both use the impulse-momentum approach to esti- model’s in-water forces were up to 40% larger than those from
mate the debris impact force, but recommend widely different the in-air tests. Also, Ko (2013) observed that increasing the
values for the contact duration, t. ASCE (2010) recommends object mass increases the peak force and contact duration.
t = 0.03 s and FEMA (2011) recommends t = 0.1–1 s. This review of previous studies shows that there is no consen-
Nouri et al. (2010) studied the impact of two different sizes sus on how to estimate the impact forces from floating debris. In
and masses of solid wooden logs on a polyvinyl chloride part, contradictory conclusions and the lack of consensus reflect
tube, representative of debris impact on a cylindrical structure. the complex hydrodynamics involved in two body collisions
The structure was mounted on a multi axis load cell to mea- and the highly variable conditions under which such collisions
sure the impact force at the structure’s base. They concluded occur. Contradictory observations have been made about what
that the contact duration has a significant effect on the impact affects the contact duration of debris impact, and at what veloc-
force as it directly influences the acceleration of the impacting ity debris travels until its collision with structures, and a wide
debris. They also found that contact duration was independent range of values of t have been recommended. This study
of the log mass, contrary to the conclusion of Haehnel and Daly proposes the impulse momentum approach for estimating the
(2002, 2004), and that contact duration was independent of the impact force, but with the direct measurement of the impact
debris velocity. accelerations of the debris to overcome the problem of deter-
Paczkowski et al. (2012) developed an equation for esti- mining t and the velocity of the debris immediately before
mating debris impact force, with the same form as a contact impact. Direct measurement can be achieved using an impact
stiffness approach equation (see e.g. Haehnel & Daly, 2004). accelerometer, which is a recent development in sensing sys-
Paczkowski et al. (2012) also carried out small scale in-air tems for force monitoring. Impact accelerometers can be used
impact experiments. The impacting objects were four different in both the real environment and the laboratory and are fre-
lengths of square steel tubes. The travel of the objects was con- quently used by civil and mechanical engineers to study impact
trolled by guide rails so that they always collided at right angles in pendulum experiments, automobile collision, vibration anal-
to the target. They concluded that the maximum impact force ysis, projectile collision, etc. However, these devices have not
is independent of the total mass of the debris, which is not been used in tsunami research to study the impact of floating
in agreement with findings of Haehnel and Daly (2002, 2004) object on structures. Until recently, most high range accelerome-
and Nouri et al. (2010). The equation developed by Paczkowski ters required direct wiring to an external recording device, which
Journal of Hydraulic Research Vol. 54, No. 4 (2016) Tsunami-borne debris impact on structures 437

made them difficult to use, particularly for a moving object debris. Thus, the flow velocity (behind the leading front of the
travelling for a long distance in water. Recently developed wire- bore) remains constant until a few milliseconds before collision,
less impact accelerometers overcome such limitations and have while the debris velocity increases over time and would reach
advanced measurement techniques. The added mass should be the flow velocity over a long enough distance.
taken into consideration, as can be seen from its inclusion in Second, the debris body is accelerating during its entrainment
current design guidelines such as those for vertical evacuation and therefore is entraining water. For simplification in the debris
structures (FEMA, 2012). As discussed earlier, the added mass velocity derivation, it is assumed that there is no added mass of
of water is a function of object density (Hamilton, 2000). There- entrained water during the debris entrainment. This assumption
fore, to better clarify the in-air and in-water debris impact forces, results in overestimation of the debris velocity. However, the
further testing is required using objects with different shapes and effect of added mass is considered later when calculating the
densities. debris impact force.
This study used an impact accelerometer to experimentally Equation (1) rewritten in terms of the debris velocity ud is:
investigate the impact of tsunami-borne debris on structures, by
directly measuring the impact acceleration of the floating object. dud
md = 0.5Cd ρw (ub − ud (t))2 Ad (3)
The measurement technique used in these experiments is a new dt
method in the investigation of the tsunami-borne debris impact
Integrating Eq. (3) for the debris velocity gives:
problem. The method was used with the impulse-momentum
approach with the aim of obtaining insights into the variation of  −1
Cd ρw Ad 1
added mass of entrained water, the contact duration of impact ud (t) = ub − t+ (4)
2md ub
and the debris velocity. A debris velocity equation was also
developed, and validated using the experimental data. where t is the debris travel time. In Eq. (4), the first term in the
bracket refers to velocity of the debris and the second term to the
2 Theoretical considerations velocity of the bore carrying the debris. Integration of Eq. (4)
with respect to time gives an estimate of the distance that the
2.1 Debris velocity
debris travels (x) to reach the flow velocity:
The differential equation describing the debris impact velocity  
and the corresponding initial conditions can be written as: 2md Cd ρw Ad ub
x(t) = ub t − Ln t+1 (5)
Cd ρw Ad 2md
md ẍ = 0.5Cd ρw (ub − ẋ) Ad
2
(1)
⎧ If the distance of travel is known, Eq. (5) can be solved to esti-
⎨ẋ(0) = 0
(2) mate the time (t) for the debris to travel that distance, and the
⎩x(0) = 0 debris velocity can then be estimated using Eq. (4). In this study,
the debris was placed at a fixed distance to the structure, i.e. 2 m,
where md is the debris mass, ρ w is the water density, ub is the allowing t to be estimated from Eq. (5) and allowing the validity
tsunami bore velocity (i.e. flow velocity behind the sloping bore of Eq. (4) to be tested for this specific study. Figure 1 illustrates
front in this study), Ad is the projected area of the debris to the a schematic diagram of debris travel velocity (Eq. (4)) against
incoming bore, Cd is the drag coefficient for the debris (defining debris travel distance (Eq. (5)).
the initial motion of the debris), ẋ = dx/ dt is the debris velocity,
and ẍ = d2 x/ dt2 is the debris acceleration. Cd = 0.5, equivalent
to the drag coefficient for a short, overtopped vertical cylinder
(Heseltine, 2003; Hogben et al., 1977), was used here in the cal-
culation. The main assumptions in this study, in relation to the
debris velocity derivation, are as follows.
First, debris entrainment begins after the leading edge of the
bore passes the debris and when the depth of the bore is suffi-
cient to float the debris. Therefore, the area of interest is the part
behind the leading front of the bore. In this study, measurements
from successive pairs of wave gauges (see the experimental set-
up section) showed that the velocity in this part of the bore can
be assumed to be constant. In addition, the dimension of the
debris in relation to the bore height means that it is the average
stream-wise velocity that is acting on the debris, and this can Figure 1 Schematic diagram of the effect of debris mass on the rela-
be assumed to be constant because the depth of the bore behind tionship between debris velocity (Eq. 4) and debris travel distance
the bore front is constant during the time of entrainment of the (Eq. 5). Note the scale is arbitrary
438 S. Shafiei et al. Journal of Hydraulic Research Vol. 54, No. 4 (2016)

The physical meaning of the plot in Fig. 1 is that the oncom- effects, such as free-surface tilting, increased submergence of
ing tsunami bore picks up the initially stationary debris, giving debris, and water splash, affecting the movement of water
it an initial acceleration. Because of this initial acceleration, between the debris and the structure. Estimation of the variation
the debris velocity increases gradually as its distance from the of added mass during debris impact is a complicated hydro-
pick-up location increases. Given sufficient distance, it would dynamics problem. To work around this problem, the present
eventually reach a terminal velocity, i.e. the bore flow velocity. study used the fact that the free-surface effects between the
The distance to reach the bore velocity depends on the mass debris and the structure readily enabled water to move away
of debris. It takes longer for a heavier mass to reach the bore from the debris, and thereby cause values of added mass do
velocity than a lighter mass. not substantially increase; in effect the free-surface deformation
is type of pressure relief for water between the debris and the
structure. Consequently this study used the practical approxima-
2.2 Debris impact force
tion of assuming that the characteristics of variable added mass
Based on the assumption that the time scale of contact is can be described using the term α V . Although this approxima-
longer than the time scale of shock waves propagation in the tion requires further research beyond the scope of the present
debris (Goldsmith, 1960), a generalized form of the impulse- study (which is focused on levels of debris impact), examin-
momentum equation, Fdi = mt adi is adopted for estimation of ing the trends for α V offers insight into the variation of added
the debris impact force at the contact point, on the front wall mass in a manner that integrates the influences of debris prop-
of the structure, using the measured impact accelerations, in erties, distance to structure, and the flow conditions prevailing
which adi is the debris impact acceleration and mt is the total during debris impact. Further testing using box-shaped debris
mass of the impacting debris, which is the summation of the was carried out. The results (not reported here) showed that the
debris mass (md ) and an added mass of water (madd ). The added coefficient of α V proposed in Eq. (8) is same for both the disc-
mass depends on numerous variables defining the extent of and box-shaped debris for the conditions used in this study.
water mobilized, including proximity to the structure about to be
impacted, and free-surface presence. One variable is the degree
3 Experimental set-up
of submergence of the debris, which in turn depends on the
debris density. The degree of debris submergence was varied
3.1 Facility
by varying the debris mass, and hence the debris density.
In addition, it is hypothesized that, as the collision gap Experiments were conducted in a large wave flume, 14 m long,
narrows just before the debris impact, the degree of debris sub- 1.2 m wide and 0.8 m deep connected to a reservoir 11 m long,
mergence increases as the debris reaches a stagnation region 7 m wide and 0.6 m deep. The tsunami bore is generated using
created in front of the structure (see Fig. 9), and therefore the a 0.9 m high automatic gate which allows rapid release of the
added mass of the entrained water increases. The added mass water in the flume to generate a bore. The gate is fitted across the
is the portion of the surrounding fluid that is entrained with the full width of the flume, and consists of a vertically-sliding gate
debris (Hamilton, 2000); for a simple geometric shape it can be and a shutter gate. The sliding gate is opened using a hydraulic
estimated as: lift device, is kept open for 4 s, and is then automatically closed.
The shutter gate is opened a fraction of a second after the ver-
madd = ρw Vsub (6) tically sliding gate, and also remains open for 4 s before it is
 
ρd closed.
Vsub = αV Vd (7)
ρw
3.2 Instrumentation
in which Vsub is the submerged volume of the debris, Vd is the
total volume of the smart debris, and α V is the empirical coef- To measure the bore height and velocity, five capacitance-type
ficient taking into account the effect on the submerged volume wave gauges, with 2 mm accuracy, were placed along the flume
of the motion of floating debris immediately before and during centreline and 20 mm above its floor. The first wave gauge was
impact. Therefore, the total mass of the impacting debris is: placed 2.5 m downstream from the gate, with the remaining
gauges equally spaced 1.75 m apart over a total distance of 9.5 m
mt = md + αV ρd Vd (8) downstream from the gate. The average bore velocity was cal-
culated using the recorded time at which the bore impacted each
Equation (8) does not take into account the effects of the dis- wave gauge.
tance between the drifting debris and the structure. Landweber A disc-shaped floating debris was made from acrylic. The
et al. (1992) describe in detail the influence of distance on added circular shape of the disc creates a contact point, rather than a
mass, pointing out that estimates for added mass become com- contact area, when it strikes the structure. The point collision
plicated when free-surface effects, such as splash and waviness, reduces complications in the experiments. The disc has an outer
arise. The present impact situation involved definite free-surface diameter of 200 mm, thickness of 50 mm and mass of 510 g; it
Journal of Hydraulic Research Vol. 54, No. 4 (2016) Tsunami-borne debris impact on structures 439

Coast Data Concepts, LLC, Waveland, MS, USA). It has a


mass of 40 g and an enclosure box having dimensions of
25.6 × 26.5 × 104 mm. The accelerometer used is battery pow-
ered and USB-rechargeable and has an internal data logger. The
full-scale range of acceleration measurement is ± 250 times
the standard gravity (gn ) in the x, y and z directions, but the
sensor also has a nonlinear range beyond the rated output, giv-
ing an additional range of ± 62 gn . Thus, the full-scale range
is actually ± (250 + 62) gn , or 625 gn total. The logger can
record samples at up to 512 Hz at 14-bit resolution. The data
are internally sampled at 16 times this rate, giving an 8192 Hz
internal sample rate. The data are then summed and decimated
Figure 2 Debris with the attached accelerometer and extra mass by a factor of 16 to provide a 14-bit effective resolution, in this
case 0.0381 gn . The X250-2 typically reports less than 1 gn of
was made hollow for easier internal attachment and protection noise. The sensor responds to events with frequency up to about
of the accelerometer, and also so that it floats (Fig. 2). The wall 450 Hz. Records of events with frequency higher than this will
and bottom are 5 mm thick, and the top thickness is 3 mm. The be attenuated because the sensor cannot react fast enough.
disc is an arbitrary 1:20 model of an average-sized tsunami- The structure was mounted on a multi-axis waterproof load
borne debris, based on geometry but not on material property. cell (JR3, model 75E20A4), which was placed in a recess below
The accelerometer is tightly fastened with a clip to the inside of the flume floor to allow the base of the model structure to be
the debris to ensure it does not move during the impact event. level with the floor of the flume (see Fig. 3). The load cell mea-
The disc is well sealed to protect the accelerometer from water sures, at the structure base, forces induced by the bore and debris
damage. The initial mass of the disc smart debris (disc plus impacts. Force measurements are in the direction of the x, y and
accelerometer plus a chequerboard which is described below) z axes, with the x-axis aligned with the flow direction. The full-
is 550 g. scale ranges of force measurements are 2, 2 and 4 kN in the x,
The “structure” is a 300 × 300 × 600 mm3 square-based y and z directions, respectively. The accuracy of the load cell
prism, constructed from sheets of 5 mm thick acrylic for the is 5 N for force measurements, in each axial direction. Force
walls and 10 mm thick acrylic for the base, and placed 10 m values are calculated by applying a 6 × 6 calibration matrix
from the gate (i.e. 500 mm beyond the final wave gauge). Based (provided by the manufacturer) to the digital output signals. The
on the geometrical scale of 1:20, the structure represents a load cell data are logged at a frequency of 5 kHz.
6 × 6 m building and an oncoming bore height of about 4 m.
For approximately the same flume width, Nouri et al. (2010)
used the same structure size.
3.3 Experimental procedure
A “smart debris” device was used to measure the impact
accelerations of the debris. The smart debris device is an object To obtain different bore heights (hb ) and velocities (ub ),
with an attached impact accelerometer, which can measure the three different bore strengths were generated using differ-
acceleration before and at the instant of impact. This allows ent combinations of gate opening height (GO) and reser-
direct calculation of the forces acting on the structure at the time voir water depth (WL): B1(the weak bore): hb = 140 mm
of collision. The acceleration data also can be used to calculate and ub = 1.98 m s−1 (WL = 400 mm and GO = 200 mm),
the debris velocity during the short period ( < 1 s) immediately B2 (the moderate bore): hb = 170 mm and ub = 2.20 m s−1
before impact. As far as we are aware, such data have not previ- (WL = 500 mm and GO = 300 mm),and B3(the strong bore):
ously been obtained using a smart debris device. Accelerometers hb = 210 mm and ub = 2.45 m s−1 (WL = 600 mm and
are robust, compact, light, and stand-alone, and are also rela- GO = 300 mm. The measured bore heights (hb ) are the aver-
tively cheap compared with load cells. They are easy to calibrate age depth of water behind the sloping front of the bore. For each
and use, relatively insensitive to environmental change, are very bore strength, the bore height and the bore velocity were taken
reliable, and are stable in the long term. An accelerometer can as the mean value of five repetitions without debris. The weak,
be used in any orientation to measure acceleration along its moderate and strong bores had Froude numbers equal to 1.69,
axes. The most important advantage of an accelerometer is that 1.70 and 1.71, respectively (see Table 1).
at high frequencies, displacement and velocity drop off quickly In this study, the load cell was used to study the effect of the
into instrumental noise which often makes the acceleration the added mass of water on the debris impact force. To assess the
easiest vibration characteristics to measure. importance of the added mass effect, in-air and in-water tests
The accelerometer used in these experiments is a wire- were carried out. In-air tests were conducted using a pendu-
less tri-axial impact accelerometer and data logger (USB lum, in which the pendulum arm was made from a very flexible
Impact Accelerometer Data Logger, model X250-2, Gulf string.
440 S. Shafiei et al. Journal of Hydraulic Research Vol. 54, No. 4 (2016)

Figure 3 Coordinate systems for the flume (xyz) and the smart debris (pqr). The impact angle is about the flume y axis

Table 1 Experimental configurations experimental runs where excessive rotation occurred about the
Bore
flume x and/or z axes were repeated.
Bore height velocity Froude Debris A two-dimensional (2D) image processing technique was
Test ID (mm) (m s−1 ) number mass (g) used to detect the impact angle in the xz plane of the flume.
A high speed camera (Exilim EX-FH100, Casio Computer Co.
B1D1M1 140 1.98 1.69 550 Ltd., Tokyo, Japan) was used to film the impact, with floodlight-
B1D1M2 140 1.98 1.69 800
B1D1M3 140 1.98 1.69 1000
ing to improve the video quality. The camera has a maximum
B2D1M1 170 2.20 1.70 550 capability of 1000 frame per second (fps) of video records,
B2D1M2 170 2.20 1.70 800 giving a resolution of 226 × 168 pixels. To validate the mea-
B2D1M3 170 2.20 1.70 1000 surement of the contact duration of impact and the debris
B3D1M1 210 2.45 1.71 550 velocity, and to better understand the behaviour of the debris at
B3D1M2 210 2.45 1.71 800
impact, the camera was operated at 1000 fps. Even though the
B3D1M3 210 2.45 1.71 1000
resolution was low, enough information was obtained from the
video records. For the image processing technique (explained
Figure 3 illustrates the coordinate systems for both the flume below), the camera was operated at 210 fps, giving a resolution
(xyz) and the smart debris (pqr), in which the flume x direction of 640 × 480 pixels at 96 dpi. A chequerboard pattern (black
is in the flow direction. Because of space limitation within the and white) was attached to a vertical face of the structure. For
smart debris, the accelerometer was oriented within the disc the smart debris, a 180 × 80 × 2 mm sheet of balsa wood was
with the q axis initially aligned with the flume x axis. The glued to the top of the disc and the chequerboard pattern was
accelerometer data were resolved into the flume axes. Hereafter attached to the sheet (Fig. 2). The balsa wood sheet has 10 g
rotation of the debris about the flume y axis is defined as the mass, light enough to ensure it does not affect the balancing
impact angle, where zero is horizontal impact. of the disc during its flotation. The Camera Calibration Tool-
To investigate the effect of debris mass on impact force, an box for MATLAB, developed by Bouguet (2010), was used to
extra 250 and 450 g of mass were added to the initial mass giv- determine the impact angle of the smart debris at the instant of
ing debris with 800 and 1000 g mass, respectively. Also, the impact. The toolbox uses the sub-pixel-accurate corner extrac-
extra masses were added to change the density of the same tion method (Harris, 1987) to locate the corners of the squares of
debris to investigate the effect of density of object on the impact. the chequerboard (to a precision below 0.1 pixel) as they appear
The resulting nine experimental configurations considered in in images extracted from video films of the impact (Bouguet &
this study are shown in Table 1. The extra mass was added to Perona, 1998). The method works well as long as the corners of
the centre of the disc (Fig. 2). This arrangement was adopted, the black and white squares are detectable from the impact snap-
following trials of various arrangements of extra-mass, as that shot and rotation of the smart debris about the flume x or z axis is
which resulted in the best movement of the smart debris with less than ± 5°. An inclined and a “vertical” line can be fitted to
extra mass. The best movement is that making the impact angle the detected points from the chequerboard attached to the debris
detection possible in these experiments (the impact angle detec- and structure, respectively, and the angle between the two lines
tion method is explained below).As subsequently discussed, can be calculated from the slope of the lines. Even though the
Journal of Hydraulic Research Vol. 54, No. 4 (2016) Tsunami-borne debris impact on structures 441

structure was in a fixed position, the “vertical” line was always designed to determine the optimum parameter values. A carriage
fitted to the structure chequerboard to minimize any possible was made using an acrylic sheet mounted on ball bearings for
error in detecting the angle between the object and the structure easier movement. The carriage was connected to a rotary motor
due to the deflection of the structure at the impact instant. The through a metal arm that transformed the rotary motion of the
resolution of the photo (see Fig. 6) was determined after mea- motor to reciprocating motion of the carriage. The accelerom-
suring the spatial extent covered in the imagery (1.2 × 0.9 m, eter was tightly screwed to the carriage and subjected to a
determined on the flume wall, resulting in 1.875 mm pixel size at sinusoidal motion with peak-to-peak amplitude of 234 mm. The
this distance). The maximum error in detecting the impact angle, motor speed was adjustable from low to high frequencies. The
based on the precision of 0.1 pixel for the method used and res- time the carriage travelled the distance of 234 mm was recorded
olution of 640 × 480 pixels for images, was less than ± 1.5° in to calculate the average speed of the motion. The test was
this study. repeated for three different speeds, 0.22, 0.64 and 1.04 m s−1 .
The smart debris was freely placed on the flume floor, 2 m In addition to integration of acceleration data, the velocity
before the structure; it was then carried to the structure by the of the debris was estimated by utilizing video image analysis.
oncoming bore. The smart debris was free to move and was not The video records were processed in VirtualDub software to get
restrained in any way as it travelled to the structure. The 2 m different snapshots of the debris movement. These snapshots
distance was chosen to ensure that the debris did not often miss were used to obtain the duration of the impact event and also
the target. The debris was carried away from the structure by to approximate the average velocity of the debris immediately
the flow immediately after the impact. The impact target is the before impact (about 1 s after the debris was picked up).
centre of the structure (“contact point”); if the debris missed the
target, the test was repeated.
4 Results and discussion
Experiments were repeated 10 times for each of the nine
configurations shown in Table 1. For each repetition, the entire
4.1 Acceleration data
sequence from gate opening to backflow from the end wall of
the flume had a duration of about 7 s. For each configuration, the For the largest impact, i.e. test B3D1M3 (see Table 1), the
impact acceleration and the impact force were taken as the mean experiment was repeated 35 times to find the statistical distribu-
value of the 10 repetitions. In addition, for the largest impact, i.e. tion of the disc impact. The results of Shapiro and Wilk (1965)
the strong bore and md = 1000 g, the experiment was repeated and Jarque and Bera (1980) statistical tests both provided very
35 times to allow statistical study analysis and determination of strong evidence of normality of the 35-repetition acceleration
the impact distribution from the accelerometer data. dataset (p-value of 0.7, for a 95% confidence level). This is also
Useful information that can be derived from the use of this shown graphically using the normal probability plot (Q-Q plot)
measurement technique is the debris velocity at a few millisec- in Fig. 4. Because normality of the results can be assumed for the
onds before impact. Velocity and displacement can be obtained largest impact case, mean values can be used in all the analyses.
from acceleration by single and double integration, respectively.
In the analysis of structural dynamics, double integration of
acceleration with respect to time is the prevalent method of
obtaining displacement data. However, it is widely recognized
that velocity results cannot be obtained by simple integration
without pre-processing of the acceleration data (Boore, 2005;
Chen et al., 2010; Li et al., 2005; Stiros, 2008; Sunder & Connor,
1982); direct integration of acceleration records often causes
unrealistic “drifts” in displacements and velocities (Yang et al.,
2006). To overcome this problem, very low frequency compo-
nents need to be filtered out from the raw acceleration data (by
applying a high pass filter) before integration (Li et al., 2005).
A Butterworth filter was applied to the acceleration data
using the MATLAB function “butter”. The Butterworth filter
is the best compromise between attenuation and phase response
(Zumbahlen, 2011). It has no ripple in the pass band or the stop
band. It is characterized by a magnitude response that is max-
imally flat in the pass-band and monotonic overall. The butter
function has two parameters, “filtering order” and a “normalized
cut-off frequency”. The normalized cut-off frequency must be a
number between 0 and 1, where 1 corresponds to the Nyquist
frequency (half the sample rate). A simple apparatus was Figure 4 Normal probability plots for the disc impact acceleration
442 S. Shafiei et al. Journal of Hydraulic Research Vol. 54, No. 4 (2016)

(a)

(b)
Figure 5 Sample run of the disc smart debris device. The impact is in
the accelerometer q-direction

An example of data collected by the smart debris device is


shown in Fig. 5, for test B3D1M1 (see Table 1). The impact was
in the q direction of the accelerometer, where the accelerometer
q direction is parallel to the direction of flow (Fig. 5 is for the
raw recorded acceleration and is not converted into the flume
coordinates).
The first small change in the q-direction acceleration relates
to the deceleration of the debris in the vicinity of the structure.
Figure 6a and b show the bore front before impact and the angle
of the smart debris impact, respectively. The smart debris device
was not carried by the leading front of the tsunami bore. Initially, Figure 6 Snapshots of the debris impact: (a) before impact, and (b)
the bore front struck the structure, the bore flow slowed down, the disc impact
and water accumulated in front of the structure and splashed
upward. The oncoming debris was affected by the decelerated
4.2 Debris velocity
water in front of the structure, and decelerated over a certain
distance before it impacted the structure. The Butterworth filter was applied to the acceleration data
Figure 7 illustrates the problem that is associated with direct from the debris impact experiments to obtain the smart debris
integration of acceleration data to calculate the velocity. velocities. The smart debris velocities in some runs were also
Figure 7a shows the raw recorded acceleration data from the estimated from video snapshots and compared with velocities
apparatus used to determine optimum Butterworth filter param- from integration. There was good agreement between the two
eter values for carriage speed of 1.04 m s−1 . The data exhibit a velocities, confirming the validity of the integration method.
simple sinusoidal motion. Figure 7b shows the result of single Figure 8 shows the relationship between experimental con-
integration of the same data before application of a Butterworth figuration and debris velocities derived from single integra-
filter; and Fig. 7c shows the results of single integration after tion of the q-direction accelerations. The range bars show the
filtering. maximum and minimum values of the variations of the debris
Figure 7b illustrates that direct integration of acceleration velocities (the same concept is applied to subsequent figures).
results in a large drift (negative in this case, but it can be pos- From Fig. 8 it can be concluded that, as expected, the debris
itive). Various values of the two butter parameters were used velocity reached higher maximum velocity values with increas-
when filtering the acceleration data. The filtering order and cut- ing bore strength. Acceleration and velocity of the smart debris
off frequency were chosen to be 4 and 0.22 Hz in this study are both proportional to the kinetic energy of the bore. A faster
(based on trial and error). Figure 7c shows results of velocity bore has more kinetic energy and applies a larger force to the
after filtering, for carriage speed of 1.04 m s−1 . It was found that initially stationary smart debris; it accelerates the smart debris
the cut-off frequency value was independent of speed variation. more than a slower bore, resulting in higher velocity of the
The designed Butterworth filter is specific to the accelerometer floating smart debris. The smart debris velocity reaches lower
used in this study. maximum velocity values with increasing debris mass. Kinetic
Journal of Hydraulic Research Vol. 54, No. 4 (2016) Tsunami-borne debris impact on structures 443

(a) energy is proportional to mass; for the same amount of energy


transferred to an object, an object with larger mass will attain
a lower velocity. In addition, after the debris is detached from
the flume floor, the water flow velocity profile (having mini-
mum velocity close to the flume floor and maximum velocity
at the flow surface) affects the attained debris velocity. Debris
with greater density comes to the flow surface more slowly and
is mostly under the influence of lower velocities; consequently
it would require a longer time and distance to reach the flow
velocity.

(b)
4.3 Debris velocity equation

On average, for each experimental configuration the proposed


debris velocity equation, Eq. (4), estimates a debris velocity that
is larger than the measured velocity obtained from integration
of the acceleration data at a few milliseconds before the debris
impact. The most likely explanation for this is that Eq. (4) esti-
mates the debris travel velocity for an undisturbed flow and does
not consider the influence of the structure on the debris velocity
at a few milliseconds before the debris impact.
Figure 9 illustrates the debris entrainment from its pick-up
location to the structure. In this study, the initially stationary
(c)
debris was not immediately picked up by the oncoming bore.
The debris was initially overtopped by the approaching flow,
and was gradually accelerated by it while still submerged. After
a few milliseconds, the debris floated to the water surface and
was carried toward the structure. The bore reached the structure
before the debris collision with the structure, slowed down, and
created a stagnation region in front of the structure. The floating
debris followed the flow and consequently partially decelerated
in the stagnation region, at a few milliseconds before the impact.
Other possible reasons for overestimation of the debris velocity
using Eq. (4) are: (1) overestimation of Cd , the drag coefficient;
Figure 7 Illustration of the problem associated with integration of and (2) the assumption that there is no added mass of entrained
acceleration data and the results of application of a Butterworth filter water.
before integration. (a) Recorded acceleration from the apparatus (see On average, the measured velocity (i.e. from integration of
text). (b) Velocity obtained from direct integration of the acceleration. measured acceleration) was about 50% of the debris velocity
(c) Velocity obtained from filtered acceleration
calculated using Eq. (4). Figure 10 shows the relationship
between the measured and 50% of the calculated debris
velocities for the different experimental configurations. Dashed
lines show ± 20% deviation from the perfect fit line.
Figure 10 illustrates the good agreement between the mea-
sured debris velocity and 50% of the calculated debris veloci-
ties. Therefore, for structures close to the coastline, Eq. (4) can
be used to estimate the debris velocity at a structure, knowing
the distance between the debris pick-up location and the struc-
ture, and providing an appropriate reduction factor is applied to
the estimated values. For this study, the factor is 0.50.

4.4 Contact duration

Figure 8 Relationship between debris velocity and experimental con- Figure 11 shows the contact duration of the disc impact for each
figuration experimental configuration.
444 S. Shafiei et al. Journal of Hydraulic Research Vol. 54, No. 4 (2016)

Figure 9 Different transition steps of the debris motion until its collision with the structure

and Paczkowski et al. (2012), but is in contrast to those from


Nouri et al. (2010) (in their study the time elapsed between the
initial contact of the debris with the structure and the maximum
impact force). It was also found that debris velocity had little
effect on contact duration, which is in agreement with results
from Nouri et al. (2010). These contradictory results are because
of differences in the materials, velocities, and sizes of the debris
objects used in experiments. Haehnel and Daly (2002, 2004)
used high mass and low velocity debris, Nouri et al. (2010)
used solid wooden logs, and Paczkowski et al. (2012) used high
velocity long hollow steel tubes where their flat frontal cross-
sections collided with the target. In addition, the conclusion
Figure 10 Relationship between 50% of the calculated velocity (using
Eq. (4)) and the measured velocity (by integrating the acceleration data) from Paczkowski et al. (2012) is based on their in-air tests where
the object could easily bounce back after collision. In case of
a floating debris collision with a structure the flow applies an
additional drag force on the object at the impact time, and holds
the object on the impacted surface. Therefore, in-water debris
impact is under the influence of the level of submergence of the
debris (i.e. debris density). Debris with larger mass took longer
to accelerate and move off the structure, and accordingly had
longer contact duration.

4.5 Impact angle


As noted earlier (see also Figs 6 and 9), the smart debris device
was carried by the flow behind the leading front of the tsunami
Figure 11 Contact duration (t) versus experimental configuration bore. The floating debris followed the flow streamlines of the
bore flow that rose up in front of the structure after the initial
The average contact duration from the accelerometer record tsunami bore impact, and consequently collided with the struc-
varied between 7.5 and 11.5 ms. Video image analysis also con- ture at an angle to the horizontal (the impact angle). Figure 12
firmed these values for t. The results show that the contact shows the relationship between the impact angle (see Fig. 3),
duration increases as the mass of the object increases. This is and the experimental configurations (bore strength and debris
in agreement with results from Haehnel and Daly (2002, 2004) mass combinations). On average, the impact angle for the disc
Journal of Hydraulic Research Vol. 54, No. 4 (2016) Tsunami-borne debris impact on structures 445

(a)

(b)
Figure 12 Relationship between impact angle and experimental con-
figuration

(c)

Figure 13 Impact accelerations in horizontal (ax ) and vertical (az )


directions, for different experimental configurations

ranged from 15° to 30°. The impact angle increased with both
increasing bore strength and debris mass.
The impact angles were used to resolve the acceleration data
into the flume x direction (hereafter “horizontal”) and the flume Figure 14 Horizontal impact acceleration (ax ) related to (a) debris
z direction (hereafter “vertical”), both with respect to the flow velocity, (b) contact duration and (c) impact angle
direction in the flume. Figure 13 shows the impact accelerations
ax and az for each experimental configuration, in the horizontal
and the vertical directions, respectively. 14b and c suggest that the contact duration and the impact angle
The results show that the impact angle is a key variable are closely related. Figures 11 and 12 can also be compared.
affecting the impact acceleration. Variability in the accelera-
tion readings occurred because the debris had slightly variable
4.6 Added mass
impact angles and did not always collide with the structure at the
same contact point (for a large variation, the run was repeated). The in-air tests (for assessing the importance of added mass)
Also, the amount of water trapped between the debris and the were carried out with the debris velocity the same as for the
structure can significantly contribute to damping of the impact in-water test. Figure 15a and b show the peak acceleration
acceleration. Therefore, variation in the impact acceleration was and force measured from the smart debris and the load cell,
to be expected. respectively, for the disc (md = 550 g) impact.
Figure 14a, b and c illustrate the relationship between Figure 15a indicates that in-air and in-water tests resulted in
impact acceleration in the horizontal direction and debris veloc- approximately the same impact accelerations. But for the in-
ity (for tests B1D1M1, B2D1M1, and B3D1M1), contact water test, the force measured by the load cell at the structure
duration (for tests B3D1M1, B3D1M2, and B3D1M3) and base was about 1.5 times the force from the in-air test (Fig.
impact angle (for tests B3D1M1, B3D1M2, and B3D1M3), 15b). It can be concluded that the force difference comes from
respectively. the difference in the impacting mass. This is consistent with
Figure 14 indicates that the impact acceleration (ax ) the results of the study conducted by Haehnel and Daly (2002,
increased with increasing debris velocity, but decreased with 2004) and the recommendation of FEMA (2012) that the added
increasing contact duration and increasing impact angle. Figure mass should be taken into account.
446 S. Shafiei et al. Journal of Hydraulic Research Vol. 54, No. 4 (2016)

(a)

(b)

Figure 16 Debris impact forces: in the horizontal (FxA ) and in the


vertical (FzA ) directions

with increasing debris velocity (as is also the case for the impact
accelerations; see Fig. 13). This increase in force due to increas-
ing debris mass is in agreement with observations from Ko
(2013). The large impact angle of the disc collision caused large
vertical forces.
In order to relate these experimental results to forces in actual
Figure 15 Comparison of the (a) impact acceleration and (b) impact
tsunamis, Froude scaling (because of the importance of gravita-
force for the in-air and in-water tests at the same debris velocity
tional and inertial forces) was used to scale up the model masses
Table 2 Values of β V (ratio of Vsub to Vd ) and forces. Mass and force can each be scaled up by the factor
l3 , in which l is the length ratio of the prototype to the model
Mass (g) βV ( > 1) (Aksnes, 2010; Baker et al., 1991; Kim, 2014). The maxi-
550 0.50
mum force of the smart debris was measured to be 808 N, equiv-
800 0.70 alent to force on a geometrically similar prototype building of
1000 0.90 6464 kN.
To check the validity of the impact force calculated utilizing
From analysis of video snapshots, it was observed that the the acceleration data, the force was measured at the structure
percentage of debris submergence a few milliseconds before base using the load cell. Figure 17 compares the force measured
the impact instant was more than that obtained from consid- at the contact point using the smart debris device (FA ), with the
eration of debris density only. From inspection of video snap- force measured at the structure base, using the load cell (FL ).
shots, a value of α V = 1.4 was adopted for the use of Eqs (7) Figure 17 illustrates a good correlation between the forces
and (8) in this study. Table 2 shows the resulting values of measured at the two locations (R2 = 0.92), implying that the
β V ( = Vsub / Vd ) for various masses of the disc. The submerged measurement technique using the smart debris device is a valid
volume of the debris was found to be independent of the debris method. The impact force values measured by the load cell
velocity. In the work done by Riggs et al. (2013), the alu- at the base of the structure were smaller than those measured
minium container had smaller density than the objects used in
this study. In their experiments, only a small percentage of the
object was submerged and therefore the contribution of added
mass of water in the debris impact was negligible (for the lon-
gitudinal impact of the container). In this study a sizable part
of the debris was submerged in water and Fig. 15 clearly shows
that the effect of added mass should not be ignored in debris
impact force calculation.

4.7 Debris impact force

Using the measured impact x- and y-accelerations, the debris


impact force was calculated using Eqs (7) and (8). Figure 16
shows the debris impact forces in the horizontal (FxA ) and the Figure 17 Relationship between the debris impact forces measured at
vertical directions (FzA ), respectively. FxA and FzA both increase the contact point using the smart debris device and at the structure base
significantly with increasing debris mass and increase slightly using the load cell
Journal of Hydraulic Research Vol. 54, No. 4 (2016) Tsunami-borne debris impact on structures 447

using the smart debris devices at the contact point of the impact. study. The debris velocity is a function of the bore strength, and
This difference in the measure forces is due to various damping mass of the debris.
sources, e.g. material damping, connection damping, and water The average contact duration for the disc impact in this study
damping. ranged from 7.5 to 11.5 ms. The contact duration increases with
Debris impact is a rigid body collision that occurs at the increasing mass of the object. It was found that the contact
structure surface with a contact area that is small in comparison duration is independent of the debris velocity.
with the overall dimensions of the structure. Stresses gener- Further research is needed to obtain insight regarding the
ated in the contact area decrease rapidly with increasing radial variation of added mass during debris impact with structures.
distance from the contact point, so the internal energy of defor- The approximate estimates of added mass determined during the
mation is concentrated in a small region surrounding the contact present study may not apply to debris object shapes and densi-
point (Stronge, 2004). This method of measuring accelerations ties, or to structure dimensions, that are different from those used
using a smart debris device allows direct estimation of the force for this study.
(if the mass is known) applied to the structure surface at the
instant of collision. This force at the structure surface can con- Funding
centrate the debris impact on the cladding material and result
in a punching force that can damage a small area. The debris The study is funded by the New Zealand Ministry of Business,
impacts do not necessarily disturb the overall stability of the Innovation and Employment, through GNS Science.
structure. Instead, similar to the wind-borne debris impact, fail-
ure of the seaward wall cladding may lead to an increase in water Notation
pressure inside the building and produce failure of the principal
structural frame (Minor, 1994). adi = debris impact acceleration (m s−2 )
ax = acceleration of the debris impact in the horizontal
direction (gn or m s−2 )
5 Conclusions az = acceleration of the debris impact in the vertical direc-
tion (gn or m s−2 )
A smart debris device was used to investigate the impact force Ad = projected area of the debris to the incoming bore (m)
on a structure, at the contact point of the collision, of float- Cd = shape coefficient for the debris (defining the initial
ing objects, representative of tsunami-borne debris. The smart motion of the debris (–)
debris device was a disc-shaped object, equipped with an impact FA = debris impact force at the structure base (N)
accelerometer. The experiments were conducted with three dif- FL = debris impact force at the contact point of collision
ferent tsunami bore velocities (1.98, 2.2 and 2.45 m s−1 ), and, with structure (N)
for each velocity, for different masses of debris device (550, FxA = debris impact force in the horizontal direction (N)
800 and 1000 g).The use of the accelerometer has advantages FzA = debris impact force in the vertical direction (N)
of easy installation and calibration, and direct measurement of hb = bore height (m)
impact acceleration. The acceleration of the disc impact varied madd = added mass of entrained water (kg)
between 12 and 55 gn . The impact acceleration decreases with md = debris mass (kg)
increasing debris mass and increases slightly with increasing mt = total mass of the impacting debris (kg)
debris velocity. The experiments showed a substantial impact t = debris travel time (s)
acceleration in the vertical direction, due to large impact angles ub = bore velocity (m s−1 )
of the disc collision. ud = debris velocity (m s−1 )
In the problem associated with floating debris impact on a Vd = total volume of the debris (m3 )
structure, the added mass of the entrained water due to the Vsub = submerged volume of the debris (m3 )
debris motion is an important term that must be known when x = debris travel distance (m)
analysing the debris mass collision with structures. Equation (8) ẋ = debris velocity (m s−1 )
can be used to estimate the total mass of the impacting debris. ẍ = debris acceleration (m s−2 )
A coefficient α V value of 1.4 was found to be appropriate for the αV = empirical coefficient taking into account the effect on
conditions used in this study. the submerged volume of the motion of floating debris
The debris velocity, at a few milliseconds before the impact, immediately before and during impact (–)
is calculated by integration of acceleration data. Equation (4) βV = ratio of the submerged volume of the debris to the total
can reasonably estimate the debris impact velocity, if the dis- volume of the debris (–)
tance between the debris pick-up location and coastal structure t = contact duration of impact (s)
is known, providing the resulting velocity is reduced to allow l = length ratio of the prototype to the model debris (–)
for reduction of the bore velocity in the vicinity of the structure. ρd = debris density (kg m−3 )
A reduction factor of 0.50 was found to be appropriate for this ρw = water density (kg m−3 )
448 S. Shafiei et al. Journal of Hydraulic Research Vol. 54, No. 4 (2016)

References Heseltine, J. L. (2003). Flow around a circular cylinder with a


free end (masters thesis). University of Saskatchewan, Saska-
Aksnes, V. (2010). Model tests of the interaction between a toon, Canada. Retrieved from http://hdl.handle.net/10388/
moored vessel and level ice. Proceedings of the 20th IAHR etd-07252011-090143.
international symposium on ice. Lahti, Finland. Hogben, N., Ward, G., Carpenter, L. H., Morison, B. L.,
ASCE. (2010). Minimum design loads for buildings and other Miller, J. R., Searle, J. W., & Keulegan, G. H. (1977).
structures (ASCE/SEI 7–10). Reston, VA: American Society Estimation of fluid loading on offshore structures. In ICE
of Civil Engineers. Proceedings, 63(3), 515–562.
Baker, W. E., Westine, P. S., & Dodge, F. T. (1991). Similar- Isaacson, M., & McTaggart, K. (1990). Influence of hydrody-
ity methods in engineering dynamics: Theory and practice of namic effects on iceberg collisions. Canadian Journal of Civil
scale modeling (Rev. ed.). Amsterdam: Elsevier. Engineering, 17(3), 329–337.
Boore, D. M. (2005). Long-period ground motions from Jarque, C. M., & Bera, A. K. (1980). Efficient tests for
digital acceleration recordings: A new era in engi- normality, homoscedasticity and serial independence of
neering seismology. In P. Gülkan, & J. G. Ander- regression residuals. Economics Letters, 6(3), 255–259.
son (Eds.), Directions in strong motion instrumentation Kim, E. (2014). Experimental and numerical studies related to
(pp. 41–54). Dordrecht: Springer. the coupled behavior ofice mass and steel structures during
Bouguet, J. Y. (2010). Camera Calibration Toolbox for MAT- accidental collisions (doctoral thesis). Norwegian University
LAB. Retrieved from http://www.vision.caltech.edu/bouguetj/ of Science and Technology, Trondheim, Norway. Retrieved
calib_doc/. from http://hdl.handle.net/11250/229710.
Bouguet, J.-Y., & Perona, P. (1998). Camera calibration from Ko, H. T. (2013). Hydraulic experiments on impact forces from
points and lines in dual-space geometry. Proceedings of the tsunami-driven debris. Oregon State University, Corvallis,
5th European conference on computer vision, 2–6. OR, USA. Retrieved from http://hdl.handle.net/1957/38478.
Chen, T., Hu, W., & Sun, R. (2010). Displacement measure- Landweber, L., Ettema, R., & Shahshahan, A. (1992). A study of
ment algorithm using handheld device with accelerome- ice-mass interaction hydrodynamics (Report No. 202). Iowa
ter. 2010 Asia-Pacific conference on wearable computing City: Iowa Institute of Hydraulic Research, The University of
systems (APWCS), 122–126. Iowa.
FEMA. (2011). Coastal construction manual: principles and Li, X., Rizos, C., Ge, L., Tamura, Y., & Yoshida, A. (2005).
practices of planning, siting, designing, constructing, and The complementary characteristics of GPS and accelerom-
maintaining residential buildings in coastal areas (FEMA eter in monitoring structural deformation. Proceedings of
P-55). Washington, DC: Federal Emergency Management the Institute of Navigation, national technical meeting,
Agency. 911–920.
FEMA. (2012). Guidelines for design of structures for Minor, J. E. (1994). Windborne debris and the building enve-
vertical evacuation from tsunamis (2nd ed.) (FEMA lope. Journal of Wind Engineering and Industrial Aerody-
P646). Washington, DC: Federal Emergency Management namics, 53(1), 207–227.
Agency. Retrieved from http://www.fema.gov/media-library- Nouri, Y., Nistor, I., & Palermo, D. (2010). Experimental
data/1426211456953-f02dffee4679d659f62f414639afa806/ investigation of tsunami impact on free standing structures.
FEMAP-646_508.pdf. Coastal Engineering Journal, 52(1), 43–70.
Goldsmith, W. (1960). Impact: The theory and physical behav- NRC. (2004). Preventing earthquake disasters: the grand chal-
ior of colliding solids. London: Edward Arnold. lenge in earthquake engineering. A research agenda for the
Haehnel, R. B., & Daly, S. F. (2002). Maximum impact force Network for Earthquake Engineering Simulation (NEES),
of woody debris on floodplain structures. US Army Corps The National Academies Press.
of Engineers. Retrieved from http://ascelibrary.org/hyo/ Paczkowski, K., Riggs, H., Naito, C., & Lehmann, A. (2012).
resource/1/jhend8/v130/i2/p112_s1. A one-dimensional model for impact forces resulting from
Haehnel, R. B., & Daly, S. F. (2004). Maximum impact force of high mass, low velocity debris. Structural Engineering and
woody debris on floodplain structures. Journal of Hydraulic Mechanics, 42(6), 831–847.
Engineering, 130(2), 112–120. doi:10.1061/(asce)0733- Piran-Aghl, P., Naito, C., & Riggs, H. (2013). An experimental
9429(2004)130:2(112) study of demands resulting from in-air impact of debris. 11th
Hamilton, J. M. (2000). Vibration-based techniques for international conference on structural safety and reliability,
measuring the elastic properties of ropes and the added 16–20 June 2013. Columbia University, New York, USA.
mass of submerged objects. Journal of Atmospheric and Riggs, H., Cox, D., Naito, C., Kobayashi, M., Aghl, P. P.,
Oceanic Technology, 17(5), 688–697. doi:10.1175/1520- Ko, H. S., & Khowitar, E. (2013). Water-driven debris impact
0426(2000)017 < 0688:vbtfmt > 2.0.co;2. forces on structures: experimental and theoretical program.
Harris, C. G. (1987). Determination of ego-motion from 32nd international conference on ocean, offshore and Arctic
matched points. Alvey vision conference, 1–4. engineering, Nantes, France, V001T001A059.
Journal of Hydraulic Research Vol. 54, No. 4 (2016) Tsunami-borne debris impact on structures 449

Saatçioğlu, M. (2009). Performance of structures during the Stronge, W. J. (2004). Impact mechanics. Cambridge: Cam-
2004 Indian Ocean tsunami and tsunami induced forces for bridge University Press.
structural design. In A. Tugrul Tankut (Ed.), Earthquakes and Sunder, S. S., & Connor, J. J. (1982). A new procedure for
tsunamis (pp. 153–178). Dordrecht: Springer. processing strong-motion earthquake signals. Bulletin of the
Saatcioglu, M., Ghobarah, A., & Nistor, I. (2006). Effects of Seismological Society of America, 72(2), 643–661.
the December 26, 2004 Sumatra earthquake and tsunami on Yang, J., Li, J., & Lin, G. (2006). A simple approach to inte-
physical infrastructure. ISET Journal of Earthquake Engi- gration of acceleration data for dynamic soil–structure inter-
neering Technology, 42(4), 79–94. action analysis. Soil Dynamics and Earthquake Engineering,
Shapiro, S. S., & Wilk, M. B. (1965). An analysis of variance 26(8), 725–734.
test for normality (complete samples). Biometrika, 52(3/4), Yeh, H., Robertson, I., & Preuss, J. (2005). Development of
591–611. design guidelines for structures that serve as tsunami verti-
Stiros, S. C. (2008). Errors in velocities and displacements cal evacuation sites. Washington State Department of Natural
deduced from accelerographs: An approach based on the Resources (p. 40).
theory of error propagation. Soil Dynamics and Earthquake Zumbahlen, H. (2011). Linear circuit design handbook.
Engineering, 28(5), 415–420. Burlington, MA: Elsevier.

You might also like