1 Experimental analysis of a shake table test of a

2 timber-framed structures with stone and earth

3 infill
4 F. Vieux-Champagnea)bc,)) M.EERI, Y. Sieffertb)c,) S. Grangeb)c,) C. Belinga Nko’oa), E.
5 Bertrandd)
, J.C. Duccinie), C. Fayee), L. Daudevilleb)c)

6 The seismic performance of timber-framed structures filled with stones and earth
7 mortar has been analyzed by introducing the structural sub-scales (cell, wall, house)
8 at which monotonic and cyclic loadings were considered. This article aims to
9 present the dynamic behavior of a house as determined through shaking table tests.
10 Based on this experimental multi-scale analysis, this paper confirms that timbered
11 masonry structures offer effective seismic resistance; moreover, such a compre-
12 hensive analysis helps enhance understanding of the seismic resistant behavior of
13 timber-framed structures with infill. This paper also aids ongoing development of
14 a numerical tool intended to predict the seismic resistant behavior of this type of
15 structure.

16 INTRODUCTION

17 According to Gurpinar et al. (1981), the existence of timber-framed structures with infill dates
18 back several millennia, having originated in the Neo-Hittite states (northern Syria and southern
19 Turkey), where timber-framed structures filled with adobe were already widespread. In Italy, an
20 archaeological excavation campaign in Herculaneum revealed a building during Roman times
21 featuring two levels of wood-framed structure with infill. This type of structure was listed
22 under the name of Craticii or Opus Craticium by Vitruvius (Langenbach (2007)). Moreover,
23 such buildings are found throughout the world (Vieux-Champagne et al. (2014)) and still being
24 built in most countries.

25 Timbered masonry structures are popular worldwide mainly for their reduced construc-
26 tion cost, thanks to the use of local resources and know-how, as well as for their aesthetics
a) CRAterre,AE&CC Research Unit, National School of Architecture of Grenoble, France
b) Univ.
Grenoble Alpes, 3SR, F-38000 Grenoble, France
c) CNRS, 3SR, F-38000 Grenoble, France
d) CEREMA, Laboratoire de Nice, F-06300 Nice, France
e) FCBA, French Technological Institute for Forestry, Cellulose, Timber Construction and Furniture

1
27 and/or ability to achieve greater resistance to seismic forces (see Dutu et al. (2012)). Tradi-
28 tional timbered-framed structures with infill have indeed exhibited remarkable behavior during
29 recent major earthquakes (Turkey in 1999, Greece in 2003, Kashmir in 2005, Haiti in 2010 or
30 China in 2013), often sustaining very little damage. In contrast, the seismic resistant behavior
31 of new construction made of masonry blocks or concrete has generally been subpar or even
32 disastrous (Haiti in 2010). Haiti’s problems stemmed from the lack of a building code and
33 standards for designing structures, which was due to the fact that seismic forces tend not to be
34 considered in the design of most engineered buildings as well as to the presence of poor quality
35 construction and/or materials (see Paultre et al. (2013) and Dogangun et al. (2006)). Ensur-
36 ing a structural implementation that meets building codes carries with it a relatively high cost,
37 thus making modern construction techniques inaccessible to the majority of localities (Tobriner
38 (2000), Makarios and Demosthenous (2006), Langenbach (2008), Audefroy (2011), Qu et al.
39 (2014)).

40 These findings raise an issue concerning the minimal importance assigned to local architec-
41 ture by the scientific community, as well as by those responsible for reconstruction efforts. A
42 number of research projects have recently been conducted to enhance knowledge of the seis-
43 mic resistant behavior with respect to traditional wood-framed structures with infill. As regards
44 experimental works, the following authors, whose scale of investigation has ranged from the
45 connection to the shear wall, can be cited (see Figure 1):

46 • Connection: Ali et al. (2012) (”Dhajji-Dewari” construction);

47 • Elementary cell: Ferreira et al. (2012) (”Pombalino” construction);

48 • Reduced scale of a shear wall: Cruz et al. (2001) (”Pombalino” construction) and Vas-
49 concelos et al. (2013) (”Pombalino” construction);

50 • Shear wall: Meireles et al. (2012) (”Pombalino” construction), Poletti and Vasconcelos
51 (2014) (”Pombalino” construction), Ali et al. (2012), Ceccotti et al. (2004) (”Maso” con-
52 struction), Aktas et al. (2013) (”Himis” construction);

53 • Building: none, to the best of the authors’ knowledge.

54 All these shear tests are based on a quasi-static loading and corroborate the good seismic
55 resistance of such structures, even though structurally speaking they are quite distinct.

2
56 This paper will present the experimental study of a timbered masonry structure in Haiti,
57 called ”Kay peyi”, in an attempt to analyze its behavior under seismic loads. The aim here is to
58 offer a scientific assessment of the structure’s implementation. This work program is based on
59 specific buildings, as shown in Figure 2, built as part of different Haitian reconstruction projects
60 (i.e. Misereor, SC/CF projects in collaboration with the CRAterre laboratory and local partners,
61 such as the Haitian NGO ”GADRU”).

62 This dynamic structural behavior will be studied at scale 4 (building) in order to complete
63 the first part of the quasi-static experimental campaign, which encompasses the connection
64 (scale 1) and the elementary cell (scale 2) and extends to the shear wall (scale 3, Vieux-
65 Champagne et al. (2014)). This multi-scale approach is illustrated in Figure 1.

66 Such an approach provides an understanding of the behavior of the various wall components.
67 At the first scale, type-1 connections will be studied to determine the local influence of the
68 nail number under two loading directions. The elementary cell scale then serves to analyze
69 the influence of the type and presence of infill. Scale 3 (wall) will be examined as a means
70 of emphasizing the structure’s overall behavior. Lastly, scale 4 (house) provides a complete
71 understanding of the structure, as regards the dynamic behavior of: the transverse walls, the
72 in-plane wall, the roof, and the connections at the wall/roof interface.

73 For purposes of this study:

74 • A representative ground motion of Haiti’s 2010 earthquake has been designed (100%
75 deterministic), based on both the local seismic context and typical Port-au-Prince soil;

76 • A seismic signal calibrated on acceleration spectra, representative of Guadeloupe’s high

77 seismic hazard zone (Bertil et al. (2010)), has been introduced.

78 These signals yield an analysis of the structure under: the actual Haitian ground motion, a
79 ground motion that complies with current regulations, and a high-frequency range capable of
80 optimizing shake table capacity.

81 This article is divided into three parts: the first section will present the test building and
82 materials used in its construction. Next, the second section will describe the shake table test
83 program (experimental set-up, instrumentation and the simulated ground motion), while the
84 last section will report and discuss the results of this campaign.

3
 Scales studied in Vieux-Champagne (2014) Studied scale
Type-2 Type-1
Connection x2 Connection

90 cm

97,5 cm

Sh
ea
wall

rw
sv erse

all
Tran Loading
direction
Scale 1 Scale 2 Scale 3 Scale 4
Joint Elementary cell Wall Building

Figure 1. The four scales of this experimental study (house model sketched by C. Belinga Nko’o)

(a) Traditional (b) Built for a reconstruction project

Figure 2. Rural Haitian houses

85 TEST BUILDING AND MATERIALS

86 Filled wood structures are built in many countries (including France, Germany, Italy, Pakistan,
87 India and Haiti) (see Langenbach (2006), Makarios and Demosthenous (2006) and Dogangun
88 et al. (2006)), in recognition of the use in most cases of natural materials available on-site. This
89 solution therefore is environmentally friendly (a key concern in developed countries), while
90 offering simpler building conditions and a lower cost (major considerations in emerging coun-
91 tries).

92 Three main types of such structures are prevalent in Haitian rural reconstruction projects: 1)
93 bracing by means of San Andrew’s crosses (X-cross), filled with natural stones and bonded by
94 an earth mortar (see Figure 2; 2) bracing with just one diagonal and filled with adobe (handmade
95 earth brick); and 3) bracing by wattle and daub (W & D) panels (with a frame made of pieces
96 of wood or bamboo) and filled by earth mortar. The X-cross wall filled with stones was selected
97 for this study by virtue of being easiest to build and thus the most widespread. The X-crosses

4
 98 simplify filling of the structure and enhance safety in the case of falling stones thanks to smaller-
99 sized infill components. Large pieces of wood however are more difficult to procure by the local
100 population.

101 The house has a footprint of 4.65 × 4.65 m2 ; it is 3.20 m high (at the ridge) and symmetrical
102 about the N-S axis (i.e. the loading direction, see Figure 6). The shear walls have six vertical
103 posts, with a window and door respectively between posts 2-3 and 4-5 in the N-S direction. The
104 transverse walls are identical, with six vertical posts as well and one window located between
105 posts 3 and 4. The specimen was anchored into steel beams by means of 40 bolts (2 between
106 each pair of vertical posts, which were themselves bolted onto the table, see Figs. 5(b) and
107 5(c)).

108 The earth is composed of a 4/1 ratio of a 0-25 mm sifted clayey earth, a 1/2 ratio of water
109 (depending on the water content of sand) and a 1/1 ratio (very coarse measurement) of sisal
110 fibers (easily available in Haiti). To prevent the mortar from cracking, water content must be
111 limited in the earth and moreover sisal fibers need to be added.

112 These fibers serve to limit the onset of cracking and its propagation in the mortar. Plain
113 shank nails (also called ”common nails” or ”wooden nails”) are placed inside wood triangles in
114 order to improve the bond between both parts.

115 The bracings are composed of Saint Andrew’s crosses with one continuous diagonal and
116 another diagonal divided into two parts (Figure 1); they are connected to the vertical post using
117 two 70-mm long wooden nails (Figure 3(b)). These techniques facilitate construction and avoid
118 weakening the wood, as opposed to an edge half-lap joint located at the center of each diagonal.

119 Various connections have been used to build the structure described above (see Figure 3),
120 though not all of them exert the same influence on the seismic resistance of the structure. The
121 type 1 is a steel-wood nailed joint (Figure 3(a)), which consists of a punched steel strip sur-
122 rounding the wood parts (in a T-shape) and fastened by a few nails that determine the behavior
123 of the connection. The type 2 is the joint between the middle of the post, the bracings and
124 the nogging (i.e. the middle horizontal stud) (Figure 3(b)). The connection between the roof
125 and the walls is made by ligature wire, as depicted in Figure 3(c). The remaining connections
126 (horizontal diaphragm and roof structure, Figure 3(d)) are nailed joints.

127 The timber strength class is C18, with a density ρmean,C18 = 390 kg/m3 according to Eu-
128 ropean Standard EN 338 (see EN 338 (2003)). The strip is an FP30/1.5/50 from Simpson
129 Strong-Tie
R
, 30 mm wide and 1.5 mm thick. 3 × 70 mm plain shank nails were used (see

5
130 EN 10230-1 (2000)) to fasten the strip.

131 The roof was made with 28 2.5 × 0.9m2 metal sheets 0.5 mm wide (6 kg) fastened to the
132 timber structure by means of nails.

(a) Type-1 joint (b) Two type-2 joints (c) Roof/wall interface con- (d) Roof structure and a hori-
nections zontal bracing between the walls

Figure 3. Nail connections of the timber structure

133 SHAKE TABLE TEST PROGRAM

134 A dynamic test has been conducted on an entire timber-framed house with walls filled by both
135 earth and stones and whose design has been adapted from the Haitian building culture specif-
136 ically for purposes of this study (see section “acknowledgement”). This test was intended to
137 analyse the behavior of such a structure under seismic loading.

Dimensions 6×6 m2
Weight 6 tons
Max. dis. ± 0.125 m
Max. vel. 0.75 m/s
Max. acc. 4 g
Frequency range 0 - 30 Hz

Figure 4. Shake table characteristics

138 EXPERIMENTAL SET-UP

139

140 Shake table

141 These dynamic tests were performed at the uniaxial earthquake simulation facility in the Me-
142 chanical Laboratory of the French Institute FCBA. The shake table is composed of a 6 × 6m2

6
143 aluminum platform moved by a 250 kN servo-hydraulic actuator.

144

145 Shaking table/structure interface

146 Since the house contains infill panels, it had to be built directly on the table in order to prevent
147 cracks in the earth. A steel frame was designed to anchor the structure onto the table, as de-
148 picted in Figure 5. The mudsill was locked into the steel frame by means of a wooden wedge
149 (Figure 5(c)) and then bolted to it. This study is focused on seismic response of the timber
150 frame structure with infill and doesn’t take into account the foundation and its connection to the
151 wooden structure. It is assumed that the structure is perfectly connected to the foundation that is
152 considered as rigid body. Obviously, the failure location depends on the nature of the structure
153 and of the soil-structure interactions, but in terms of design, the objective is to make it happen
154 into the steel fasteners of the mud-sill to post connections.

(a) Timber structure anchored (b) Mudsill placed in the steel (c) Mudsill locked by a wooden
into the steel frame frame wedge

Figure 5. Anchorage of the structure on the shake table

155

156 Seismic test sequence (see Table 1)

157 As indicated above, this test sought to analyze the behavior of this kind of structure under
158 seismic loading. For this purpose, the seismic loading had to be representative of Haiti’s January
159 12t h 2010 earthquake. Unfortunately, this signal could not be recorded but was instead designed
160 according to the empirical Green’s function method. This simulated signal is the so-called
161 ”Haiti (HTI) 100%” in this paper (see Section 3.3.1). Next, in order to analyze the nonlinear
162 behavior of the structure, the house was subjected to the 200% and 300% signal as well as
163 to a Guadeloupe (GUA) far-field ground motion (Section 3.3.2). Between the HTI and GUA
164 signals, the structures were repaired. The reparation process consisted of hammering back the
165 extracted nails and tightening the loosened ligature wires. To assess the change in fundamental
166 frequency, the structure was subjected to low-level white noise (0.03 g and 0.5 mm RMS (Root

7
167 Mean Square)) between each seismic loading.

Table 1. Seismic test sequence

Seismic test Seismic hazard Amplitude

Ground motion Scaled PGA (g)
level level scaling factor

1 1 0.27
2 HTI - 2 0.54
3 3 0.77
Reparation operation
4 GUA 10%/50Y 1 0.32
Reparation operation
5 GUA - 3.9 1.26

168 INSTRUMENTATION

169 The measurement devices used were placed according to the layout shown in Figure 6. The
170 accelerometers allowed analyzing the natural frequency evolution, while the draw wire dis-
171 placement sensor (DWDS) and LVDT measurements were compared to results of the Digital
172 Image Correlation (DIC) produced through high-speed camera pictures, with a subsequent com-
173 parison to numerical modeling results. To the best of the authors’ knowledge, this constitutes
174 the first time a DIC has been used to analyze this kind of structure at this scale. As such, let’s
175 now focus on the set-up introduced to perform the DIC.

176 The Phantom v641 high-speed camera (with a 28 mm lens, 4 megapixels) was used to
177 capture the eastern shear wall, at a speed of 150 fps (frames per second), in order to obtain
178 the absolute displacement field of the panel by DIC. The acquisition rate was set to be higher
179 than the frequencies contained in the seismic signal (< 30 Hz, i.e. for a minimum of 30 fps).
180 The speckled pattern was generated by projecting black, gray and white paints. The average
181 speckle dot diameter was approx. 1 cm. The lighting (4,000 W, 4K alpha version) was placed
182 5 m from the wall. This powerful projector was necessary to produce a uniform and intense
183 lighting. Once the videos were recorded, kinematic field measurements could be performed by
184 means of a DIC technique with the Tracker software developed by Combe and Richefeu (2013).
185 Measurement accuracy was about 0.1 pixel or 0.22 mm across the building.

8
 DW Accelerometers
DS N (-) loa E
1 2 DW din Draw Wire
g dir
DW DS2 ect Displacement Sensor
DS io n
3 Linear Variable
DW D W Displacement Transformer
DS WD S (+)
5 S4
DW High-Speed Cameras
DW
DS DS Structure orientation and
6 N (-) E

7 W S (+)
sign convention

AC ACC5
C4
LVDT2

DW
ACC3

D S1
AC
8 C
DS 6
DW
A
AC CC8
C7
AC

LVDT1
9 C9
DS
ACC1

DW S1
0
D 1
D
W S1 AC AC
D C2 C
W 10
D

Figure 6. Measurement device position and sign convention

186 SIMULATED GROUND MOTIONS

187 For purposes of this research, a strong ground motion had to be simulated since no seismological
188 recording of the January 12th 2010 earthquake is available in Port-au-Prince. Two signals were
189 simulated, the first corresponding to a Mw 6.8 earthquake occurring 34 km west of Port-au-
190 Prince, and the second one fitting an empirical response spectrum characteristic of seismic
191 activity in the French West Indies.

192 Haiti ground motion

193 Context
194 Port-au-Prince is located at the southern edge of the Cul de Sac Valley, which is a large rift
195 valley extending east-to-west from the Haitian coast to the Dominican Republic. Most of the
196 valley is underlain by young sediments (Lambert et al. (1987)), while most of the city of Port-
197 au-Prince is underlain by Mio-Pliocene deposits including marl, sandstone, siltstone and shale
198 in alluvial fans and low foothills (Lambert et al. (1987)). The Mio-Pliocene deposits are ex-
199 pected to be relatively stiff compared to younger, less consolidated Quaternary deposits (Hough
200 et al. (2010)); however, they are also expected to be characterized by lower impedance than the
201 adjacent hills to the south, which are composed of more highly consolidated limestone, con-

9
202 glomerate and volcaniclastic rock (Lambert et al. (1987)). Some degree of sediment-induced
203 amplification is therefore expected throughout Port-au-Prince.
204

205 Methodology
206 Seismic ground motion can be predicted by numerically modeling wave propagation. Most
207 methods include the source complexity and travel path from the source to the surface. These
208 deterministic methods however are insufficient to accurately predict seismic motion above a few
209 hertz. The resulting ground motion also depends to a great extent on velocity model accuracy;
210 hence, most models do not consider amplifications due to Quaternary deposits.

211 To predict broadband ground motion, an empirical approach based on the site-specific Em-
212 pirical Green’s Function (EGF) technique (e.g. Hartzell (1978)) was adopted. This simulation
213 leads to synthetic ground motion that can be useful for the engineering community regarding
214 the seismic design and assessment of civil structures. The seismic motion, capable of being
215 generated in Port-au-Prince, was simulated in considering a scenario using a location near the
216 2010 earthquake site. This motion has been simulated for a site exhibiting strong local effects in
217 the center of Port-au-Prince nearest the targeted epicenter (HVPR, USGS-BME) (Altidor et al.
218 (2010); Hough et al. (2010)). The simulated earthquake has a moment magnitude equal to 6.8,
219 with a location 34 km from the HVPR station and a focal depth of 12 km.
220

221 The method employed is based on the hypothesis of similarity between earthquakes of dif-
222 fering magnitudes, in considering that both weak and strong earthquakes are similar processes
223 that differ by a scaling factor. The underlying principle is to simulate the recordings of a hy-
224 pothetical future earthquake using the actual recordings of a smaller one that contains rich
225 information on path and site effects. The stochastic approach proposed by Kohrs-Sansorny
226 et al. (2005) was implemented. This method is based on a two-step summation scheme. The
227 SIMULSTOC code offers the advantage of requiring only a few input parameters while rapidly
228 generating a large number of possible accelerograms. This method is also based on the works of
229 Boore (1983), Wennerberg (1990) and Ordaz et al. (1995). In practice, 500 Equivalent Source
230 Time Functions (ESTFs) was generated representing the time histories of the energy release
231 over the fault at frequencies less than the corner frequency of the small event set as the EGF.
232 A random, two-step process proposed by Ordaz et al. (1995), using two probability density
233 functions, was adopted to generate these ESTFs, whose differences can indirectly account for

10
234 the different types of ruptures and produce a large variability in ground motions (Beauval et al.
235 (2009)). Next, each ESTF was convolved with the EGF at the station corresponding to each
236 component. The higher frequency part of the spectrum (> fc) was then directly modeled by
237 the small event’s spectrum. This method yields synthetic time histories that, on average, are in
238 agreement with the ω −2 model (Aki (1967); Brune (1970)) and moreover respect a non-constant
239 stress-drop condition (Beeler et al. (2003); Kanamori and Rivera (2004)). All method details
240 can be found in Kohrs-Sansorny et al. (2005). Among the requisite input parameters, the ra-
241 tio between the static-stress drop of the target event C is the only one without an associated
242 constraint. For a practical application in the present simulation, considering a C value of 1 is
243 proposed.

244 The selected aftershock (taken as an empirical Green’s function) occurred on the 3r d of
245 May 2010 (7:21 pm). Its moment magnitude was measured to be 4.4, and the epicenter was
246 located close to Logne (18.538N, 72.643W). Since the north-south component of the EGF was
247 strongest at the HVPR station, it was used in the Mw 6.8 ground motion simulation. From the
248 500 simulations, the one with the largest loading between 0.1 and 0.5 seconds of period length
249 was selected. For low-rise structures, this simulation involves taking the worst-case scenario
250 possible among our set of synthetics. The result, in terms of elastic response spectrum, is
251 shown in Figure 7(b); Figure 7(a) displays the corresponding time history in acceleration.

252 Guadeloupe ground motion

253 Another seismic signal was introduced in order to subject the structure to:

254 • a signal calibrated on an acceleration spectrum representative of high seismic hazard

255 zones in the Antilles,

256 • a different frequency range,

257 • optimized shake table capacities.

258 The generation of this modified natural accelerogram has been adapted from the Miyagi
259 earthquake (Japan, 2003), as measured at the K-Net station and calibrated on an earthquake
260 spectrum by the relationship developed by Youngs et al. (1997) (Guadeloupe scenario, far-
261 field subduction ground motion), so as to improve spectrum representativeness. This finding is
262 correlated with a scenario of: ”strong hazard”, soil B from Eurocode 8 (EN 1998-1 (2005)), and
263 a probabilistic computational basis with a 475-year return period. This corresponding seismic

11
264 hazard level has a 10% probability of exceedance in 50 years (10%/50Y), which satisfies the
265 probability associated with the non-collapse requirement in Eurocode 8.

266 The time and frequency content of the ground motion, as obtained by the method described
267 above for the Guadeloupe far-field scenario (PGA = 0.33 g), is indicated in Figure 7.
0.5
Ground Acceleration (g)

0.25
0
−0.25 HTI
−0.5
0 5 10 15 20 25 30
0.5
0.25
0
−0.25 GUA
−0.5
0 5 10 15 20 25 30

(a) Time-history of the ground motions

1.25
Spectral Acceleration (g)

0.75

0.5

0.25
HTI
GUA
0
0 0.5 1 1.5 2
Period (s)

(b) Absolute acceleration response spectra for 5%

damping of earthquake ground motions used in seis-
mic tests

Figure 7. Time and frequency content of the Haiti (HTI) and Guadeloupe (GUA) ground motions

268 RESULTS AND DISCUSSION

269 Our results will be presented in three sections, namely: ”free vibration tests” to evaluate the
270 modal properties of the structure; ”global building response”; and ”observed damage”, as is
271 customary in articles focusing on the analysis of shake table tests at the structural scale (see
272 Filiatrault et al. (2002), Filiatrault et al. (2009), Van de Lindt et al. (2010), van de Lindt et al.
273 (2014) or Bahmani et al. (2014)).

12
274 FREE VIBRATION TESTS

275 Before the seismic tests, a modal analysis was conducted by measuring three axis acceleration
276 responses at each wood member intersection of the structure while subjected to a white noise
277 excitation. This process yielded: the first three modal shapes (Figure 8), natural frequencies,
278 and damping ratios (Table 2). The first two modes are similar and represent the transverse wall
279 modes (Figs. 8(a) and 8(b)). The small frequency difference can be attributed to the various
280 positions of openings in the shear wall as well as the intrinsic differences in the workmanship
281 of the two walls (weight, geometric dimensions, etc.). The second primary mode is the shear
282 wall vibration (Figure 8(c)).

(a) First modal shape of the first (b) First modal shape of the (c) Second modal shape of the shear
transverse wall second transverse wall wall

Figure 8. Fundamental modal shapes of the structures, as obtained through the modal analysis per-
formed before the seismic tests

Table 2. Natural frequencies and the equivalent viscous damping ratio

Pole Element Frequency Damping

(Hz) ratio
(%)

1a Transverse wall 5.2 5.2

1b Transverse wall 5.9 5.7
2 Shear wall 10.9 5.3

283 Before and after each seismic input, the structure was excited by a low-level white noise
284 base acceleration input with a uniform spectrum (0.5-25 Hz frequency band), a root mean square
285 (RMS) amplitude of 0.03 g and 0.5 mm for the purpose of evaluating both the frequency drop
286 in the structure and the destructiveness of previous input signals. An analysis of the structural
287 modification resulting from successive seismic tests was performed by computing:

13
288 • the Frequency Response Function (FRF, Figure 9) between the output acceleration at
289 node 7 (ACC7) and the input table acceleration (ACC10) obtained during each low-level
290 white noise input test (Figure 6). As a reminder, the FRF is a representation in the fre-
291 quency domain of the ratio between an output and input acceleration signals;

292 • the deterioration of the normalized equivalent lateral stiffness (NELS) in the N-S direction
293 after each seismic test (Figure 10(a)), in assuming a single-degree-of-freedom response of
294 the test building. Since the initial fundamental period T0 is known, as is the fundamental
295 period Ti measured after each seismic test, the NELS ki as a percentage of initial lateral
296 stiffness k0 can be calculated after each seismic test as:
 2
ki T0
= (1)
k0 Ti
297 The NELS remains proportional to the natural frequency of the structure;

298 • the variations in the first modal equivalent viscous damping ratio (EVDR) measured in
299 the N-S direction of the test building after each seismic test conducted (Figure 10(b)).
300 The EVDR of the test building was determined using the half-power bandwidth method
301 (see, for example, Clough and Penzien (1993)) applied to the FRF peaks.

302 By means of these graphs, it is now possible to analyze the evolution in structural damage,
303 i.e.:

304 • HTI 100%: After the ”HTI 100%” seismic test, acceleration remained constant while
305 the natural frequency (and hence the normalized stiffness) decreased by 4.8% (Figs. 9
306 and 10(a)). This outcome is due to a slight rearrangement of the wood/earth interface,
307 involving the removal of some friction interfaces between the two materials, which would
308 also explain the decrease in NELS shown in Figure 10(a).

309 • HTI 200%: After this signal, a significant reduction in both the natural frequency (5 Hz to
310 4 Hz, i.e. equivalent to a decrease in NELS of approx. 30 %) and amplitude (15.4 to 7.8)
311 are recorded, while the EVDR rises by around 6 points. These results stem from damage
312 to the steel connections (buckling of steel strips, nails pulled out, loosening of the ligature
313 wires and degradation of the wood/earth interface, Figure 20).

314 • HTI 300%: In this case, the natural frequency, acceleration amplitude and NELS all
315 followed the same trend as in HTI 200%, whereas the EVDR increased very slightly.

14
316 This finding can be explained by the fact that the same deterioration processes occurred
317 during ”HTI 200%” and ”HTI 300%” (Figs. 20(b) and 20(e)); therefore, the EVDR and
318 energy dissipation were not much higher than during the ”HTI 200%” test.

319 • Repairing: After this process, the structure returned to the level of natural frequency
320 obtained during the HTI 200% test; in contrast and as logically expected, the acceleration
321 amplitude increased and the EVDR dropped.

322 • GUA 100%: After this test, it is worthwhile to note that despite the relative harmfulness
323 of the seismic signal (in comparison with the two previous ones and the next one), the
324 energy dissipation shown in Figure 10(b) reached a maximum value due to damage to the
325 repaired connections, which is correlated to a decrease in both the NELS and acceleration
326 amplitude.

327 • GUA 390%: After this seismic level, a portion of the filling collapsed (Figure 20(h)),
328 which significantly reduced the weight of the transverse walls. Despite the increase in
329 overall structural damage, the NELS therefore remained constant while the acceleration
330 amplitude and EVDR decreased.

16
_____________________________
Transverse1wall1acceleration1Rm/s2p

Initial
Shake1table1acceleration1Rm/s2p

14 Haiti11007
Haiti12007
Haiti13007
12
Repaired
Guadeloupe11007
10 Guadeloupe13907
8

0
0 1 2 3 4 5 6 7 8
Frequency1RHzp

Figure 9. Frequency response function of the transverse wall acceleration to the shake table acceleration

331 GLOBAL BUILDING RESPONSE

332 This section aims to detail the analyses conducted relative to the global deformation and global
333 hysteretic response.

15
 Normalized Lateral Stiffness
120

Equivalent Viscous Damping

16
100 14
12
80 10

Ratio (%)
60 8
(%)
40 6
4
20 2
0 0

Seismic level Seismic Level

(a) Variations of normalized lateral stiffness (b) Variations of first modal damping ratios

Figure 10. Variation of seismic indicators

334 Global deformation

335 In order to obtain the absolute displacement in the structure, DWDS, LVDT sensors and a high-
336 speed camera were all employed. As explained above, this effort represents the first time that a
337 high-speed camera has been used at this structural scale. It was therefore necessary to validate
338 the results obtained by DIC. Figures 12 and 13 show the comparison of the displacement ob-
339 tained on the shaking table by LVDT, DWDS and by DIC for the various tests performed. The
340 correlation between results from both measurement methods is quite good, thus validating the
341 use of DIC.

342 Figure 15 indicates the maximum global deformation obtained during each seismic test.
343 This depiction was obtained by adopting the following assumptions:

344 • The deformation shape of the structure is symmetrical about both the N-S and E-W axes
345 (as illustrated in 14).

346 • The deformation shape of the transverse beams of the transverse walls can be interpolated
347 from the measures of the six DWDS.

348 Figure 15 highlights various phenomena, namely:

349 • Shear walls reach their maximum displacement almost always at the same time as the
350 transverse walls. The deformation of both are indeed correlated.

351 • Structural damage affects the time when shear walls reach their maximum deformations
352 as well as the underlying nonlinear evolution in these deformations. After each sig-
353 nal, the natural period of the building is in fact modified due to the damage sustained,

16
 Figure 11. Points whose displacement is analyzed by the DIC

40 80 150

30 60
100
Absolute displacement (mm)

Absolute displacement (mm)

Absolute displacement (mm)

20 40

50
10 20

0 0 0

−10 −20
−50
−20 −40

−100
−30 −60
LVDT LVDT LVDT
DIC DIC DIC
−40 −80 −150
5 10 15 20 25 30 35 40 45 50 5 10 15 20 25 30 35 40 45 50 5 10 15 20 25 30 35 40 45 50
Time (sec) Time (sec) Time (sec)

(a) HTI 100% (b) HTI 200% (c) HTI 300%

40 150

30
100
Absolute displacement (mm)

Absolute displacement (mm)

20
50

10

0
0

−50
−10

−20 −100
LVDT LVDT
DIC DIC
−30 −150
5 10 15 20 25 30 35 40 45 5 10 15 20 25 30 35 40
Time (sec) Time (sec)

(d) GUA 100% (e) GUA 390%

Figure 12. Comparison of the absolute shake table displacements obtained by DIC and shake table
LVDT (not represented in Figure 6)

17
 4 10 15

8
3 10

Relative displacement (mm)

Relative displacement (mm)

Relative displacement (mm)

6

2 4 5

2
1 0
0

0 −2 −5

−4
−1 −10
−6
DWDS DWDS DWDS
DIC DIC DIC
−2 −8 −15
16 17 18 19 20 21 22 23 24 25 16 17 18 19 20 21 22 23 24 25 16 17 18 19 20 21 22 23 24 25
Time (sec) Time (sec) Time (sec)

(a) HTI 100% (b) HTI 200% (c) HTI 300%

8 40

6 30
Relative displacement (mm)

Relative displacement (mm)

20
4

10
2

0
0
−10

−2
−20

−4
−30
DWDS DWDS
DIC DIC
−6 −40
16 17 18 19 20 21 22 23 24 25 16 17 18 19 20 21 22 23 24 25
Time (sec) Time (sec)

(d) GUA 100% (e) GUA 390%

Figure 13. Comparison of the top relative wall displacements obtained by DIC and DWDS1

15 15

10 10
Relative displacement (mm)

Relative displacement (mm)

5 5

0 0

−5 −5

−10 −10
DWDS1 DWDS2
DWDS2 DWDS7
−15 −15
16 17 18 19 20 21 22 16 17 18 19 20 21 22
Time (sec) Time (sec)

(a) E-W Symmetry (b) N-S Symmetry

Figure 14. Comparison of relative displacements obtained by DWDS

354 thereby varying the structural response of the subsequent signal. For instance, during
355 the HTI 100% signal, shear walls reached their maximum displacement 3.33 mm at
356 18.12 sec, while for HTI 200%, this displacement equaled 8.49 mm (i.e. a factor of
357 2.55) at 20.91 sec. This finding reveals a distinct response of the building relative to the
358 damage sustained; the present example is also true for other signals and transverse walls.

18
359 • The transverse wall deformation is much greater than that of the shear wall. This outcome
360 can easily be explained by the absence of a stiff horizontal bracing structure to prevent
361 this strong deformation and moreover by the fact that the connections between wall and
362 roof are quite flexible (ligature wires and common nails). As a reminder, this structure
363 has been adapted from Haiti’s local building culture through the use of short pieces of
364 wood and basic connection techniques. This phenomenon may be substantially mitigated
365 by constructing a floor; this kind of house has already served to build granaries above the
366 living area.

367 • The shear walls are efficient in bracing the structure in the direction of the seismic loading
368 and merely sustained little damage. The type 1 connections reached their tensile strength
369 only during the GUA 390% test (Figure 16) although no damage was visible. The max-
370 imum displacement indicated in Figure 16 was obtained from the tension tests presented
371 in Vieux-Champagne et al. (2014).

5.81 3.33 18.14 8.49

21.19 12.30

11.20 39.88
83.05
3.33 8.49 12.30
5.81 3.33 18.14 8.49 12.30
21.19

11.20 39.88
83.05

3.33 8.49 12.30

(a) HTI 100% – Dmax at 18.12 sec. (b) HTI 200% – Dmax at 20.91 sec. (c) HTI 300% – Dmax at 20.65 sec.
12.54 7.03
79.41 39.63
33.25

195.76
7.03
39.63
12.54 7.03
79.41 39.63
Initial position
33.25
195.76
Deformed shape (x5)
7.03
39.63

(d) GUA 100% – Dmax at 23.48 (e) GUA 390% – Dmax at 23.55
sec. sec.

Figure 15. Global deformations (mm) obtained at the time when the transverse wall displacement is
maximized

19
 Vertical displacement (mm)
16

14
lvdt 1 (mm)
lvdt 2 (mm)
12
dmax of the connection
10

Seismic level

Figure 16. Maximum vertical deformations (mm) experienced by type-1 joints

372 Global Hysteretic Responses

373 Figure 18 exhibits the diagram of the base shear versus the top transverse wall displacement.
374 This base shear was computed by summing the inertia forces of the infill on the basis of horizon-
375 tal acceleration recordings. Since the mass is evenly distributed along the walls, the distribution
376 shown in Figure 17 has been considered herein. For each elementary cell with infill, the weight
377 of the cell (roughly 150 kg) was evenly divided at its four corners, i.e. at the intersection of
378 the wood structure where the accelerometers had been installed. The maximum base shear and
379 displacement obtained are identified by a red circle on each graph.

380 Figure 18(f) shows the evolution in the effective lateral stiffness of the transverse wall after
381 each ground motion (called ”Global hysteresis response” (GHR) in the graph caption) along
382 with a comparison to the normalized lateral stiffness obtained from results of the Frequency
383 Response Function (FRF). The effective stiffness values have been derived by computing the
384 slope connecting the positive and negative peak base shear forces and the corresponding dis-
385 placement from the graphs in Figure 18. To compare these results, the following assumption
386 was made: the effective stiffness obtained for HTI 100% of GHR and FRF are equal.

387 The hysteretic responses of the structure recorded during the HTI and GUA series highlight
388 the trend in the case of a nonlinear overall behavior. For HTI 100%, the overall behavior is
389 linear, thus highlighting that the structure is highly resistant to a ground motion equivalent to the
390 January 12th Haiti earthquake. The slight decrease in natural frequency observed in Figure 9 for
391 HTI 100% assimilated with structural rearrangement is no longer visible here. The non-linearity
392 and dissipated energy increase as stiffness decreases (Figure 18(f)) after each HTI 200% and
393 HTI 300% signal, which reveals the structural damage (Section 4.3). Following the reparation

20
394 and during GUA 100%, the overall behavior did not fully recover its linear behavior, and its
395 effective stiffness has only risen slightly relatively to HTI 300%. This finding can be explained
396 by the irreparable damage sustained by the wood/infill interface and infill degradation, as well as
397 by the local deformation of wood at the metal connectors. The overall behavior observed during
398 GUA 390% reflects a typical hysteresis curve with good seismic resistance (to the considerable
399 seismic ground motion).

400 From Figure 18(f), it can be noted that the effective stiffness obtained with the results of
401 GHR and FRF yield similar results except for GUA 390%, whose GHR seems to provide more
402 plausible results since damage observed during this test was the most significant.

403 Figure 19 also depicts the diagram of the base shear versus the top shear wall displacement.
404 The same observations as in Figure 18 are applicable here: the connection between the behavior
405 of the shear wall and the transverse wall can again be highlighted. Figure 19(f) shows the
406 evolution in the effective lateral stiffness of the shear wall and transverse wall after each ground
407 motion and moreover indicates the ratio between the effective stiffness obtained for the shear
408 wall and that for the transverse wall. This figure underscores the similar evolution of both curves
409 with the ratio between each pair of values varying between 4.5 and 7.3. It can also be inferred
410 that the behavior of the shear wall benefits from less damage than the transverse wall; hence,
411 the reparation process exerts less influence on its behavior.
0.5

0.45

0.4

0.35

0.3

0.25

0.2

0.15

Figure 17. Assumption of the mass distribution to take into account the inertia forces

412 OBSERVED DAMAGE

413 The observed damage is described in Table 3 and Figure 20. During the HTI series, little
414 damage was noticeable, and all of it was due to deformation of the transverse walls and roof

21
 100 100 100

80 80 80
49.0 kN
60 60 60
33.8 kN 90.3 mm;
40 40 43.7 mm; 40

Base Shear (kN)

Base Shear (kN)

Base Shear (kN)

14.2 kN
20 12.2 mm; 20 20

0 0 0

−20 −13.7 kN −20 −20

−10.3 mm;
−40 −40 −33.2 kN −40
−39.4 mm; −42.5 kN
−60 −60 −60 −86.7 mm;

−80 −80 −80

−100 −100 −100

−250 −200 −150 −100 −50 0 50 100 150 200 250 −250 −200 −150 −100 −50 0 50 100 150 200 250 −250 −200 −150 −100 −50 0 50 100 150 200 250
Displacement (mm) Displacement (mm) Displacement (mm)

(a) HTI 100% (b) HTI 200% (c) HTI 300%

100 100
75.0 kN
80 80 163.6 mm;

60 60 1.4
DifferenceF(%)
1.2 GlobalFhysteresisFresponse

Effective stiffness
40 20.1 kN 40
Base Shear (kN)

Base Shear (kN)

30.7 mm; 1.0 FrequencyFResponseFFunction

(kN/mm)
20 20
0.8
0 0
0.6
−20 −16.7 kN −20 0.4
33.7
−40
−20.3 mm;
−40 0.2 15.8 18.5
0.0 1.6
−60 −60
0.0
−66.1 kN
−80 −80
−186.6 mm;
−100 −100
−250 −200 −150 −100 −50 0 50 100 150 200 250 −250 −200 −150 −100 −50 0 50 100 150 200 250
Displacement (mm) Displacement (mm) Seismic level

(d) GUA 100% (e) GUA 390% (f) Effective stiffness

Figure 18. Global hysteretic responses of the transverse wall

100 100 100

80 80 80
49.0 kN
60 60 60
33.8 kN 12.0 mm;
40 40 7.7 mm; 40
Base Shear (kN)

Base Shear (kN)

Base Shear (kN)

14.2 kN
20 2.8 mm; 20 20

0 0 0

−20 −13.7 kN −20 −20

−1.6 mm;
−40 −40 −33.2 kN −40
−5.9 mm; −42.5 kN
−60 −60 −60 −11.7 mm;

−80 −80 −80

−100 −100 −100

−50 −25 0 25 50 −50 −25 0 25 50 −50 −25 0 25 50
Displacement (mm) Displacement (mm) Displacement (mm)

(a) HTI 100% (b) HTI 200% (c) HTI 300%

100
100
75.0 kN
80 80 37.8 mm;
60 7.0
60
Ratio
6.0 GHR of shear wall
Effective stiffness

40 20.1 kN 40
Base Shear (kN)

5.0
Base Shear (kN)

5.9 mm; GHR of the transverse wall

(kN/mm)

20 20
4.0
0 0 3.0
−20 −16.7 kN −20 2.0
−4.4 mm; 1.0
−40 −40 4.9 6.2 7.3 4.9 4.5
−60 −60
0.0
−66.1 kN
−80 −80
−36.9 mm;
−100 −100
−50 −25 0 25 50 −50 −25 0 25 50
Displacement (mm) Displacement (mm) Seismic level

(d) GUA 100% (e) GUA 390% (f) Effective stiffness

Figure 19. Global hysteretic responses of the shear wall

22
415 structure. The steel strip bent and its nails were extracted by several millimeters. At the top
416 center of the transverse walls, the infill begins to fail and two stones collapse during the HTI
417 300%. Otherwise, the roof structure remained flexible thanks to the slenderness of the bracings
418 and common nail connections, which could be pulled out relatively easily (not the case during
419 HTI 100%). After the reparation, the structure sustained little damage during GUA 100%, with
420 another pull-out of the roof structure bracing. Lastly, the ground motion GUA 390% heavily
421 damaged the structure with, in particular, a collapse of two infill triangles on a transverse wall.
422 The horizontal bracing between the eastern shear wall and the northern transverse wall was
423 stripped (Figure 20(f)). The shear wall damage was visible from GUA 390%. Some infill
424 triangles began to pull out from the wood structure (Figure 20(k)), and damage in the type 1
425 connection was indeed significant, as shown in Figure 20(i) and emphasized again in Figure 16.

Table 3. Observed damage

Ground Type-1 Roof Horizontal

Infill Ligature wire
motion connection Structure bracing
HTI 100% - - - - -
Light bending
Slight
of steel strip
extraction of
Top sill/Infill and nails
HTI 200% - vertical stud -
separation extracted
and bracing
(2-3 mm)
(Figure 20(c))
(Figure 20(b))
Collapse of
two stones
Slight ligature
(Figure 20(d))
deformation
– failure of nails extracted Ligature wires Pull out of a
HTI 300% and local
the mortar at (Figure 20(e)) loosened bracing
wood
the top
deformation
sill/Infill
separation
Reparation operation
Pulled out of a
GUA 100% - - - -
bracing
Reparation operation
Damage and Buckling
bending of
collapse of Ligature wires failure of a
steel strip and Bracing
two triangle loosened and bracing –
GUA 390% nails pull off pulled out
cells of infill detached Celling joist
(Figure (Figure 20(f))
(Figure 20(h) (Figure 20(j)) pulled out
20(g),20(i))
and 20(k)) (Figure 20(l))

23
426 CONCLUSION

427 This paper has presented the last step of a multi-scale experimental program (Vieux-Champagne
428 et al. (2014)) aimed at testing the entire structure. To the best of the authors’ knowledge, this
429 seismic testing campaign is the first one ever performed on a timber-framed structure with infill.
430 For this purpose, the campaign has focused on experimental observations along with an analysis
431 based on free vibration tests and seismic tests.

432 To conduct this test, a seismic ground motion has been designed by the Nice-based CEREMA
433 (France) to subject the structure to a seismic hazard similar to the one that occurred on January
434 12t h 2010 in Haiti (due to the context of this study). Subsequently, a seismic ground motion
435 representative of high seismic hazard zones found in the Antilles was used.

436 The free vibration tests allowed detecting the first three natural frequencies of the building
437 and their evolution after each seismic ground motion. From these results, damage was iden-
438 tified; moreover, some characteristic seismic values were computed, e.g. the effective lateral
439 stiffness and the equivalent viscous damping ratio.

440 The global response of the building was analyzed through the results of displacement mea-
441 surement sensors (draw wire displacement sensor (DWDS), (LVDT) and digital image corre-
442 lation (DIC)) and accelerometers. First, the DIC was validated from DWDS results by virtue
443 of the authors’ knowledge: this was the method’s first successful implementation at this scale.
444 These databases made it possible to plot a 3D view of the global deformation and to better
445 understand the overall structural behavior. The transverse wall deformation was especially im-
446 portant for its dimensions due to the lack of horizontal bracing in the structure, whose design
447 had been based on existing Haitian structures. On the other hand, the transverse wall did not fail
448 after the 5 seismic ground motions; however, two infill triangles collapsed, which could have
449 been harmful for dwelling occupants. This issue would not be any more serious for multi-story
450 buildings since the floor, rigid in comparison to the transverse wall, would limit the effect and
451 prevent against deformation accumulation. The shear walls underwent only limited deformation
452 (at the top of the walls) and adequately braced the structure. For instance, the maximum dis-
453 placement was 45 mm for GUA 390%, whereas a modern timber-framed wall (OSB12 panels
454 with 1.5 tons at the top of the wall, Humbert et al. (2014) and Boudaud et al. (2014)) reaches
455 44 mm for GUA 320% (and in similar experimental configurations: ground motion, shake table,
456 etc.). The global hysteretic response has been analyzed through the base shear versus the top
457 displacement of both the transverse wall and shear wall. These results served to highlight the

24
458 similar shape of the hysteretic responses of both walls, which experienced significant deforma-
459 tion and dissipated a large amount of energy. The lateral effective stiffness was computed on the
460 basis of hysteretic response results, in demonstrating the evolution of damage after each ground
461 motion. An interesting correlation has been drawn between the results obtained from the GHR
462 and FRF, revealing that FRF can provide significant results as regards the evolution in structural
463 behavior after each ground motion. The seismic vulnerability of the transverse wall was then
464 pointed out by comparing its effective stiffness evolution to that of the shear wall.

465 The last section focused on damage observed during the experimental campaign. It was
466 revealed that the main damage was located in the steel connectors and infill. After each ground
467 motion, the nailed connections did in fact began to pull out, especially at the top of the trans-
468 verse walls and on the roof structure. The damaged infill panels confirm that a substantial
469 amount of energy was dissipated both in the panels and at their interface with the wood struc-
470 ture. In the case of the shear wall, the infill damage was limited. The behavior of infill during
471 seismic ground motion therefore is particularly valuable since it enhances the lightweight wood
472 structure’s efficiency in resisting highly harmful ground motion while sustaining only minor
473 overall damage due to fiber reinforcement of the earth. It must be pointed out however that for
474 the ground motion GUA 390% (Figure 20(k)), two infill panels were pulled-out out-of-plane,
475 which is a key issue in preventing injury to dwelling occupants, even though this risk is techni-
476 cally easy to mitigate (e.g. a wire mesh nailed onto the inner face of the wall).

477 This article has confirmed that timber-framed structures with infill exhibit relatively high
478 seismic resistance and, thanks to the shake table test, provide direct proof to the scientific
479 community as well as to those leading reconstruction efforts in Haiti and elsewhere that this
480 type of structure, in relying on local knowledge, can be used in new construction. Moreover,
481 this article enhances the state of knowledge of seismic resistant behavior relative to traditional
482 wood-framed structures with infill.

483 This experimental campaign at the 4-scale of the structure contributes relevant results to
484 completing not only the analysis performed at scales 1, 2 and 3 (Figure 1) but also the exper-
485 imental multi-approach that allows developing a model to predict the behavior and assess the
486 vulnerability of timber-framed structures with infill under seismic loading conditions.

25
487 ACKNOWLEDGEMENT

488 The authors wish to thank and acknowledge the support of the French National Research
489 Agency (ANR) for the ReparH project under reference code ANR-10-HAIT-003 (coordinated
490 by CRAterre in collaboration with UJF-3SR, and the AE&CC research Unit of ENSAG and
491 the Haitian NGO GADRU), the participating associations of the PADED platform and all lo-
492 cal partners for their involvement and participation, contributing to this research project. This
493 work is supported by a public grant overseen by the French National Agency as part of the
494 ”Investissements d’Avenir” program (reference: ANR-10-LABX-0083). The laboratory 3SR is
495 part of the LabEx Tec 21 (Investissements d’Avenir - reference: ANR-11-LABX-0030). The
496 authors wish to thank and acknowledge Sadrac St Fleur and Franoise Courboulex who partici-
497 pated to the simulation of the Haitian ground motion. The authors also thank the technical staffs
498 of the FCBA for the tests they conducted. Gratitude is extended to Simon Pla for his valuable
499 assistance during the experimental program.

500 REFERENCES

501 Aki, K., 1967. Scaling law of seismic spectrum. Journal of Geophysical Research 72, 1217–1231.
502 Aktas, Y. D., Akyüz, U., Türer, A., Erdil, B., and Sahin Güçhan, N., 2013. Seismic Resistance Eval-
503 uation of Traditional Ottoman Timber-Frame Himis Houses: Frame Loadings and Material Tests.
504 Earthquake Spectra .
505 Ali, Q., Schacher, T., Ashraf, M., Alam, B., Naeem, A., Ahmad, N., and Umar, M., 2012. In-Plane
506 Behavior of Full Scale Dhajji Walls (Wooden Braced with Stone Infill) under Quasi Static Loading.
507 Earthquake Spectra 28, 835–858.
508 Altidor, J., Dieuseul, A., Ellsworth, W., Given, D., Hough, S., Janvier, M., Maharrey, J., Meremonte, M.,
509 Mildor, B., Prepetit, C. et al., 2010. Seismic Monitoring and Post-Seismic Investigations following
510 the 12 January 2010 Mw 7.0 Haiti Earthquake. In AGU Fall Meeting Abstracts, vol. 1, p. 07.
511 Audefroy, J. F., 2011. Haiti: post-earthquake lessons learned from traditional construction. Environment
512 and Urbanization 23, 447–462.
513 Bahmani, P., van de Lindt, J. W., Gershfeld, M., Mochizuki, G. L., Pryor, S. E., and Rammer, D., 2014.
514 Experimental Seismic Behavior of a Full-Scale Four-Story Soft-Story Wood-Frame Building with
515 Retrofits. I: Building Design, Retrofit Methodology, and Numerical Validation. Journal of Structural
516 Engineering .
517 Beauval, C., Honoré, L., and Courboulex, F., 2009. Ground-motion variability and implementation of
518 a probabilistic–deterministic hazard method. Bulletin of the Seismological Society of America 99,
519 2992–3002.
520 Beeler, N., Wong, T.-F., and Hickman, S., 2003. On the expected relationships among apparent stress,
521 static stress drop, effective shear fracture energy, and efficiency. Bulletin of the Seismological Society
522 of America 93, 1381–1389.
523 Bertil, D., Rey, J., and Belvaux, M., 2010. Projet ANR SISBAT – Modélisation de l’action sismique.

26
524 Rapport final. BRGM/RP-58886-FR.
525 Boore, D. M., 1983. Stochastic simulation of high-frequency ground motions based on seismological
526 models of the radiated spectra. Bulletin of the Seismological Society of America 73, 1865–1894.
527 Boudaud, C., Humbert, J., Baroth, J., Hameury, S., and Daudeville, L., 2014. Joints and wood shear walls
528 modelling II: Experimental tests and {FE} models under seismic loading. Engineering Structures
529 pp. –.
530 Brune, J. N., 1970. Tectonic stress and the spectra of seismic shear waves from earthquakes. Journal of
531 geophysical research 75, 4997–5009.
532 Ceccotti, A., Faccio, P., Nart, M., and Simeone, P., 2004. Seismic behavior of wood framed buildings in
533 Cadore mountain regioni - Italy. In 13th World Conference on Earthquake Engineering, p. 14p.
534 Clough, R. W. and Penzien, J., 1993. Dynamics of structures. New York: McGraw-Hill.
535 Combe, G. and Richefeu, V., 2013. TRACKER: A particle image tracking (PIT) technique dedicated
536 to nonsmooth motions involved in granular packings. In AIP Conference Proceedings, vol. 1542, p.
537 461.
538 Cruz, H., Moura, J. P., and Machado, J. S., 2001. The use of FRP in the strengthening of timber
539 reinforced masonry load-bearing walls. Historical constructions, Guimarães, Portugal 847.
540 Dogangun, A., Tuluk, O., Livaoglu, R., and Acar, R., 2006. Traditional wooden buildings and their
541 damages during earthquakes in Turkey. Engineering Failure Analysis 13, 981–996.
542 Dutu, A., Ferreira, J., L., G., Branco, F., and Goncalves, A., 2012. Timbered masonry for earthquake
543 resistance in Europe. Materiales de construccion 62, 615–628.
544 EN 10230-1, 2000. Steel Wire Nails - Part 1: Loose Nails For General Applications.
545 EN 1998-1, 2005. Design of structures for earthquake resistance - General rules, seismic actions and
546 rules for buildings.
547 EN 338, 2003. Structural timber - Strength classes.
548 Ferreira, J., Teixeira, M., Duţu, A., Branco, F., and Gonçalves, A., 2012. Experimental Evaluation and
549 Numerical Modelling of Timber-Framed Walls. Experimental techniques .
550 Filiatrault, A., Christovasilis, I. P., Wanitkorkul, A., and van de Lindt, J. W., 2009. Experimental seismic
551 response of a full-scale light-frame wood building. Journal of structural engineering 136, 246–254.
552 Filiatrault, A., Fischer, D., Folz, B., and Uang, C.-M., 2002. Seismic testing of two-story woodframe
553 house: Influence of wall finish materials. Journal of Structural engineering 128, 1337–1345.
554 Gurpinar, A., Erdik, M., and Ergunay, O., 1981. Siting and structural aspects of adobe buildings in
555 seismic areas. In Earthen buildings in seismic areas. Proceedings of the international workshop
556 held at the university of new mexico, albuquerque, may 24-28, 1981, pp. 139–183. National science
557 foundation.
558 Hartzell, S. H., 1978. Earthquake aftershocks as Green’s functions. Geophysical Research Letters 5,
559 1–4.
560 Hough, S. E., Altidor, J. R., Anglade, D., Given, D., Janvier, M. G., Maharrey, J. Z., Meremonte,
561 M., Mildor, B. S.-L., Prepetit, C., and Yong, A., 2010. Localized damage caused by topographic
562 amplification during the 2010 M [thinsp] 7.0 Haiti earthquake. Nature Geoscience 3, 778–782.
563 Humbert, J., Boudaud, C., Baroth, J., Hameury, S., and Daudeville, L., 2014. Joints and wood shear
564 walls modelling I: Constitutive law, experimental tests and FE model under quasi-static loading. En-
565 gineering Structures 65, 52–61.
566 Kanamori, H. and Rivera, L., 2004. Static and dynamic scaling relations for earthquakes and their

27
567 implications for rupture speed and stress drop. Bulletin of the Seismological Society of America 94,
568 314–319.
569 Kohrs-Sansorny, C., Courboulex, F., Bour, M., and Deschamps, A., 2005. A two-stage method for
570 ground-motion simulation using stochastic summation of small earthquakes. Bulletin of the Seismo-
571 logical Society of America 95, 1387–1400.
572 Lambert, M., Gaudin, J., and Cohen, R., 1987. Carte Geologique D’Haiti. Feuille Sud-Est: Port-au-
573 Prince 1.
574 Langenbach, R., 2006. Preventing pancake collapses: Lessons from earthquake-resistant traditional con-
575 struction for modern buildings of reinforced concrete. In International Disaster Reduction Conference
576 (IRDC), Davos, Switzerland.
577 Langenbach, R., 2007. From Opus Craticium to the Chicago Frame: Earthquake-Resistant Traditional
578 Construction. International Journal of Architectural Heritage 1, 29–59.
579 Langenbach, R., 2008. Learning from the past to protect the future: Armature Crosswalls. Engineering
580 Structures 30, 2096–2100.
581 Makarios, T. and Demosthenous, M., 2006. Seismic response of traditional buildings of Lefkas Island,
582 Greece. Engineering structures 28, 264–278.
583 Meireles, H., Bento, R., and Cattari, S., S. Lagomarsino, 2012. A hysteretic model for frontal walls in
584 Pombalino buildings. Bulletin of Earthquake Engineering 10, 1481–1502.
585 Ordaz, M., Arboleda, J., and Singh, S. K., 1995. A scheme of random summation of an empirical Green’s
586 function to estimate ground motions from future large earthquakes. Bulletin of the Seismological
587 Society of America 85, 1635–1647.
588 Paultre, P., Calais, É., Proulx, J., Prépetit, C., and Ambroise, S., 2013. Damage to engineered structures
589 during the 12 January 2010, Haiti (Léogâne) earthquake 1. Canadian Journal of Civil Engineering
590 40, 1–14.
591 Poletti, E. and Vasconcelos, G., 2014. Seismic behaviour of traditional timber frame walls: experimental
592 results on unreinforced walls. Bulletin of Earthquake Engineering pp. 1–32.
593 Qu, Z., Dutu, A., Zhong, J., and Sun, J., 2014. Seismic Damage of Masonry Infilled Timber Houses in
594 the 2013 M7. 0 Lushan Earthquake in China. Earthquake Spectra .
595 Sieffert, Y., Vieux-Champagne, F., Grange, S., Garnier, P., Duccini, J.-C., and Daudeville, L., 2016.
596 Traditional Timber-Framed Infill Structure Experimentation with Four Scales Analysis (To Connec-
597 tion from a House Scale), pp. 287–297. Springer International Publishing, Cham. ISBN 978-3-319-
598 39492-3. doi:10.1007/978-3-319-39492-3 24.
599 Tobriner, S., 2000. Wooden architecture and earthquakes in Turkey: a reconnaissance report and com-
600 mentary on the performance of wooden structures in the Turkish earthquakes of 17 August and 12
601 November 1999. In International Conference on the Seismic Performance of Traditional Buildings,
602 Istanbul, Turkey, Nov, pp. 16–18.
603 van de Lindt, J. W., Bahmani, P., Mochizuki, G., Pryor, S. E., Gershfeld, M., Tian, J., Symans, M. D., and
604 Rammer, D., 2014. Experimental seismic behavior of a full-scale four-story soft-story wood-frame
605 building with retrofits. II: shake table test results. Journal of Structural Engineering .
606 Van de Lindt, J. W., Pei, S., Pryor, S. E., Shimizu, H., and Isoda, H., 2010. Experimental seismic
607 response of a full-scale six-story light-frame wood building. Journal of Structural Engineering 136,
608 1262–1272.
609 Vasconcelos, G., Poletti, E., Salavessa, E., Jesus, A. M., Loureno, P. B., and Pilaon, P., 2013. In-plane
610 shear behaviour of traditional timber walls. Engineering Structures 56, 1028 – 1048.

28
611 Vieux-Champagne, F., Sieffert, Y., Grange, S., Polastri, A., Ceccotti, A., and Daudeville, L., 2014.
612 Experimental analysis of seismic resistance of timber-framed structures with stones and earth infill.
613 Engineering Structures 69, 102 – 115.
614 Wennerberg, L., 1990. Stochastic summation of empirical Green’s functions. Bulletin of the Seismolog-
615 ical Society of America 80, 1418–1432.
616 Youngs, R., Chiou, S.-J., Silva, W., and Humphrey, J., 1997. Strong ground motion attenuation relation-
617 ships for subduction zone earthquakes. Seismological Research Letters 68, 58–73.

29
 1
2
3
4
5

6
7
8

(a) Damage location

(b) HTI 200%: (c) HTI 200%: (d) HTI 300%: Fill- (e) HTI 300%:
Connection – Loca- Roof bracing – ing – Location 7: Connection – Loca-
tion 6: North transv. Location 1: South South transv. wall tion 7: South transv.
wall transv. wall wall

(f) GUA 390%: Bracing – (g) GUA 390%: (h) GUA 390%: Fill- (i) GUA 390%:
Location 2: North transv. Connection – Loca- ing – Location: South Connection – Loca-
wall tion 7: South transv. transv. wall tion 4: West shear
wall wall

(j) GUA 390%: Ligature (k) GUA 390%: (l) GUA 390%: Celling joist
wire – Location 8: West Infill – Loca- – Location 3: Roof
shear wall tion 5: East shear
wall 30

Figure 20. Damage