Version of Record: https://www.sciencedirect.

com/science/article/pii/S0950061816313939
c280564cb446e94c1b5197d458654f7b
1

1 Selection, Production, and Testing of Scaled Reinforced Concrete Models and their
2 Components
3
4 Debra F. Laefer, Ph.D. (Corresponding Author)
5 Associate Professor and Head, Urban Modelling Group (UMG), School of Civil Engineering
6 and Earth Institute, University College Dublin (UCD), Dublin 4, Ireland
7 Office: +353-1-716-7276
8 Fax: +353-1-716-7399
9 E-mail: debra.laefer@ucd.ie
10
11 Aykut Erkal, Ph.D.
12 Assistant Professor, Department of Civil Engineering, School of Engineering and Natural
13 Sciences, Istanbul Kemerburgaz University, 34217, Bagcilar, Istanbul, Turkey
14 Mobile: +90 532 470 84 12
15 Office: +90 212 604 01 00 Extension: 4014
16 Email: erkala@hotmail.com
17

18 Abstract

19 This paper presents considerations and procedures for the selection, production, and testing of

20 small-scale, reinforced concrete frames to maintain geometric and kinematic similitude. To

21 verify the proposed solutions, 1/10th scaled models were subjected to adjacent

22 excavation-induced settlements in 1-gravity, soil-structure experiments. Material scaling

23 strategies were verified by comparing the resultant surface soil settlements with published

24 studies and by comparing the model building response with numerical simulation, as well as

25 the extent of the damage with previously established thresholds.

26

27 Key Words: Modelling; Similitude; Material Scaling; Kinematic Scaling; Reinforced

28 Concrete Model

© 2019. This manuscript version is made available under the Elsevier user license
https://www.elsevier.com/open-access/userlicense/1.0/
 2

29 1. Introduction

30 Full-scale testing is performed to describe real phenomena in engineering. However, due to

31 the relatively high cost and related resources needed to conduct full-scale soil-structure

32 interaction tests, many researchers work at a scale less than full size. The difficulty of such

33 an approach is minimizing (and when possible eliminating) scale effects. When material

34 properties are not properly scaled, elements may not behave in a manner representative of the

35 full-scale problem. These potential negative influences of such arrangements are called scale

36 effects. These scale effects may emerge in terms of geometric, kinematic, and/or dynamic

37 factors. Geometric scaling is relatively well understood and usually satisfied in most

38 experimental work. Yet, kinematic and/or dynamic scaling is less well understood and often

39 ignored, thereby potentially compromising the test results. To fill this gap, the current study

40 describes the material scaling in production, assembling, and testing of reinforced concrete

41 (RC) structural frame models in 1-gravity, soil-structure experiments. The work presented

42 herein was part of a larger study on the effects of adjacent excavations on existing structures

43 [1]. As details of the unreinforced brick masonry scaling have already been published [2], this

44 paper will restrict itself to the RC work.

45

46 2. Previous Studies and Scale Model Considerations

47 Experimental results from a small-scale system can be used to predict the response of a

48 full-scale system, if complete similitude is attained. To achieve this, the model should be

49 geometrically, kinematically, and dynamically similar to the full-scale prototype. This can also

50 be expressed as geometrical, material, and process-related similitude [3]. Geometric similitude

 3

51 entails replicating the precise shape of the full-scale model. Kinematic and dynamic similitude

52 require all velocity and force vectors in the model to have the same direction, respectively, as

53 those of the full-scale model, with the corresponding magnitudes related by a single scale

54 factor [4, 5]. Dimensional analysis helps satisfy these conditions through the production of

55 dimensionless groups (also called π-sets) composed of dimensional variables for the scale

56 model and prototype [6]. Matching the dimensionless groups for scale model and prototype

57 facilitate the determination of the model parameters. This technique was first proposed by

58 Buckingham (1914) [7] and is often called the “Buckingham pi theorem”. The approach was

59 further developed by many others [8, 9, 10].

60

61 Since its early days, scale models have been used in many disciplines such as hydraulics,

62 structural engineering, naval architecture, and even meteorology and geophysics [11, 12, 13].

63 Scaling in practice, however, often poses problems because while the models (e.g. footings,

64 buildings) are reduced geometrically (e.g. 1/8th scale), unless the material properties (e.g.

65 compressive strength, tensile strength) are re-engineered, they retain the same behavioral

66 characteristics of the full-scale materials [14,15]. Thus, failure to scale material properties in

67 an experiment can result in non-representative models [6]. However, in practice in scaled

68 structural engineering experiments there is no consistency in adherence to these requirements.

69 In fact, many researchers seem to treat material scaling as an optional activity. For example,

70 while Datin and Prevatt [12] in their 1/3rd scale tests of a three-dimensional (3D), light-framed,

71 wood structure to wind loading applied material scaling by assigning a π-set as Load/(Young’s

72 modulus*length2) to account for the static elastic behavior of the structure, Anil & Altin (2007)
 4

73 [16] working at the same scale to investigate cyclic loading on partially infilled reinforced

74 concrete frames did not incorporate material scaling. Consequently, their overly strong models

75 precluded direct comparisons to the prototype leaving only comparisons between the various

76 scaled models [17]. Similarly, in testing of 3m high, 2 story models of an RC space frame

77 under harmonic base excitations, De-la-Colina et al. [18] were limited to reporting only

78 qualitative observations. Arguably, exclusion of material scaling could be more critical for RC

79 models than those comprised of a single-material, because problems in RC models can arise

80 from insufficient bond strength between small diameter bars and concrete, excessive aggregate

81 size in the model concrete, and incompatible strength levels in the model [19, 20].

82

83 Table 1 chronologically presents some recent studies of RC modelling. Material strength

84 similitude in Table 1 describes the application of strength alteration to model materials to

85 capture the prototype behavior. Table 1 implies that material scaling is arguably adopted more

86 often for dynamic loading than static loading and for smaller-scale models starting around a

87 scale of 1/6 [21]. In the studies where material strength similitude was adopted (see Table 1),

88 exclusion of material scaling might have critical effects. Primary amongst them is that heavier

89 loading would have to be applied to fail these overly strong, reduced-scale models. This is a

90 widely observed phenomenon in laboratory testing [17]. In this case, excessive loading can

91 cause non-representative stress distributions, failure mechanisms, and damage patterns.

92 Additionally, under excessive loading, local imperfections within the scale model or local

93 behavior of its structural elements would result in inaccurate structural behavior. Large

94 additional masses may also be needed to increase the inertial forces high enough to cause
 5

95 failure of the structural system, thereby excessively increasing the dead load. In static and

96 pseudo-static, soil-structure interaction experiments the problems is even more critical.

97 Specifically, because a soil’s stiffness is proportional with its depth, if full loading is applied to

98 a structural model, there would be a strong likelihood of overloading the soil. This point will be

99 discussed further in the modeling section of this paper.

 6

100 Table 1. Main characteristics of reinforced concrete scale models used in experiments
Testing Model Overall Geometrical Material strength Max reinforcing Max reinforcing Max concrete Reference
dimensions of the scale similitude bar diameter steel strength strength (MPa)
largest model (mm) (mm) (MPa)
Cyclic loading RC slab with opening 5400x4400x1800 1/5 No 10 407 32.5 Hirai et al. 1995 [22]

Earthquake time 2-bay 6-story RC frame 2460x3300x50 1/5 No 6 610 45 Skjærbæk et al. 1997 [23]
history loading

Simulated One-story, one bay RC 1829x1422x152 1/2 No 9.5 438 28 Klinger et al. 1997 [24]
earthquake motions frames

Cyclic loading RC column 4500x500x500 1/4 No 8 194 48.4 Ohtaki 2000 [20]

Varying sinusoidal Masonry infilled RC 1540x1280x810 1/4 Yes 6 250 13.9 Žarnić et al. 2001 [17]
Loading frame buildings

Earthquake time Multistory RC frame 2370x3660x600 1/5.5 No 6 430 30 Lu 2002 [25]

history loading

Cyclic loading RC cantilever wall 4200x1350x125 1/2 No 16 517 60 Holden et al. 2003 [26]

Cyclic loading Slender RC structural 3658x1219x1118 1/4 No 4.8 414 27.4 Thomsen & Wallace 2004 [27]
wall

Cyclic loading RC cantilever wall 1200x1800x100 1/2.5 No 16 678 75 Antoniades et al. 2005 [28]

Earthquake time RC column 400x80x60 1/5 Yes 4.2 400 20.6 Kim et al. 2009 [29]
history loading

Increasing Geogrid reinforced 1500x4000x800 1/214 Yes na na 0.03 Liu et al. 2014 [13]
sinusoidal dynamic cemented rock-fill dam
loading

Earthquake time RC subway structure 705x577x1216 1/30 Yes 1.4 1190 7.5 Guoxing et al. 2015 [30]
history loading
 7

101 3. Prototype Description

102 The first step in creating a scaled model is the selection of the prototype. Defining the full-scale

103 structure involves choosing all of the component parts, their physical properties, and the

104 anticipated level and distribution of loading. Since the majority of adjacent excavation-related

105 problems occur in urban areas, a representative edifice was identified as a modern four-story,

106 RC frame structure without infill. To facilitate comparison with the unreinforced masonry

107 structures being tested as part of the larger program, a RC frame was selected. In this

108 arrangement, the floor loads were transferred to the supporting transverse and longitudinal

109 beams in two directions as schematically illustrated in Figure 1. A “strong-column

110 weak-beam” approach was used where the moment capacity of the column was 1.2 that of the

111 beam. Prototype loads were selected from Table 2 of the Uniform Building Code [31].

112

113 A concrete mix having a compressive strength of 34.5 MPa at 28 days was selected in

114 accordance with standard, non-high-rise construction. The prototype concrete’s tensile

115 strength was taken as 1/10th of that of the compressive capacity, as a function of the Young’s

116 modulus, which was selected as 27,717 MPa based on guidance from the American Concrete

117 Institute [32]. The Poisson’s ratio of the concrete was assumed to be 0.15. Deformed bars of

118 grade 60 steel with a modulus of elasticity of 200,000 MPa were selected for the prototype,

119 thereby complying with ACI specifications [32].

 8

120
121 Figure 1. Three-dimensional floor plan of RC slab prototype with tributary area
122

123 The position of the scale models in the testing chamber is shown in Figure 2. Close-up pictures

124 of the 1/10th scale models and their geometry for the prototype are shown in Figure 3. The

125 experimental design and procedures for model material selection and construction

126 considerations are explained in the following sections.

127
128

129
130 Figure 2. Test setup
 9

131

132
133 Figure 3. Model RC buildings
134
135
136 Table 2. Loading applied to the RC prototype
Loading Type Loading Level (Pascal)
Wind load 486.5 - 658.8
Superimposed dead load 1675.8
Live load 2394
Live load (roof) 1197
Concrete self-weight 7182
137

138 4. Experimental Scaling Considerations

139 Geometrical, material, and process-related similitude were adhered to closely in this study,

140 despite constrictions in geometrical requirements, material availability, and constructability.

141 Geometric similitude was primary, as it was controlled by the size of the testing chamber,

142 which limited the maximum size to a 1/10th scale for representation of the prototype based on a

143 maximum allowable footprint so as to not generate boundary condition problems (as

144 established through a Boussinesq analysis).

145

146 In a reduced-scale, soil-structure model, confinement stress in the uppermost part of the soil

147 profile is usually minimal. Therefore, the stress applied to soil surface must be restricted to
 10

148 prevent overloading. Consequently, the stress applied to the building materials is, thus,

149 necessarily reduced. For the models to behave as the prototype under this reduced stress, the

150 material capacities in terms of strength and stiffness must also be reduced, as well. This is

151 established formally through the creation of the dimensionless π-sets to determine the specific

152 material properties that need to be altered. Since the strains in structural elements are the

153 function of stresses, as well as of the material stiffness, there needs to be strain parity (Equation

154 1), as previously demonstrated by Tomaževič and Klemenc [33].

155  p   m (1)

156 where  p and  m are normal strains in the prototype and scale model, respectively (the

157 subscripts p and m denote the prototype and scale model designations, respectively). By

158 assuming that a linear relationship between stress and strain exists within the elastic range of

159 materials (Hooke’s law), strain parity can be represented as Equation 2

   
160     (2)
 E  p  E m

161 where  and E are the normal stress and the material’s modulus of elasticity for the prototype

162 and the scale model, respectively. Owing to the above-mentioned soil loading constraints and

163 to the fact that soil stresses acting on the structure can be considered directly proportional with

164 soil depth, the scale factor of 1/10th is adopted to reduce applied stresses in the model (Equation

165 3). Žarnić et al. [17] adopted the same strategy in their scaling of masonry infilled RC

166 buildings.
1
167  p   m (3)
10
168
 11

169 In the envisioned soil-structure experiment, considerations for the soil and the frame building

170 were the same, since the applied stresses to the building were transferred directly to the soil. To

171 maintain strain parity, the modulus of elasticity of the model was reduced, as well as the tensile

172 and compressive stresses (Equation 4 and 5).

  10    
173      (4)
 E 10  p  E m
1
174  E  p  Em (5)
10
175

176 As a result, all frame dimensions, cross-sections, loads, and material properties were scaled to

177 1/10th; the larger testing program is discussed in detail in Section 5. If only geometric

178 similitude was considered the models would have to be loaded to their full-scale capacity. This

179 would result in an overloading of the soil by an order of magnitude.

180

181 5. Experimental Design: Material Production and Scale Model Construction

182 Within the scope of a more comprehensive research program [1], two tests (Test 2 and 6) were

183 performed on RC frames without infill to investigate their response to adjacent excavation.

184 Each test included a pair of 1/10th RC frames, which were set on discrete, shallow foundations,

185 without a grade beam and placed in dry sand perpendicular to the adjacent excavation (Figure

186 3). The difference between Test 2 and Test 6 were the foundation depths, the excavation depth

187 (121.92 cm versus 109.22), and the location of the first support. Foundation dimensions of Test

188 2 were 11.9x12.7x12.7 cm, while those of Test 6 were twice as deep (11.9x12.7x25.4 cm)

189 (Figure 3). Model embedment depth was the full foundation height (12.7 cm for Test 2 and

190 25.4 cm for Test 6). The model buildings were 4-story high (30 cm each) and composed of 3
 12

191 equidistant 0.6m long bays (Figure 2 & 3). The embedded structures were set 0.15m from the

192 excavation wall. For each building, all beam cross sections were 30x46 mm, and all column

193 cross sections were 46x46 mm.

194

195 5.1. Testing Chamber

196 Experiments were conducted in a purpose-built, reconfigurable testing chamber in the

197 Schnabel Laboratory at the University of Illinois at Urbana-Champaign (UIUC). The chamber

198 was 4.9x4.3 m2 in plan and 3.0m in depth (Figure 2). Concrete blocks were post-tensioned

199 vertically and laterally to construct the chamber with a high level of rigidity and safety. The

200 interior walls of the chamber were lined with a dual layer of plastic sheeting to minimize

201 boundary related friction effects. Poorly graded sand, weighing 1400kN (mostly rounded and

202 sub-rounded particles) was pluviated into the chamber to form the model soil.

203

204 5.2. Experiment Soil

205 The sand was a washed and kiln-dried river deposit from a quarry near Pekin, Illinois. Since a

206 mass of 40 tons of sand was needed to operate the testing system, no scalable material for the

207 sand was found that could be handled and placed reliably in such quantities (see [34] for a

208 further discussion of this point). Soil properties are shown in Table 3.

209

210 In situ measurements of average mass density and void ratio were obtained as 1619 kg/m3 and

211 0.51, respectively. The results suggest that void ratio of the medium dense model sand would

212 likely be slightly larger than that of the prototype sand. Thus, it can be assumed that model sand
 13

213 would dilate less at the low testing stress levels, since dilatancy is a function of density and

214 stress intensity [13]. The in situ measured friction angle (35-45°) combined with the low level

215 of confinement would suggest that the model sand can be considered as representative of a

216 medium-dense to dense prototype sand.

217
218 Table 3. Physical properties of the model soil
Sand Property Measurement
Unified Soil Classification SP
Uniformity coefficient (Cu) 2.64
Curvature coefficient (Cc) 1.12
CPT tip resistance range 2.76 - 4.14 MPa
Dense 1826 kg/m3
Mass density (ρ)
Medium 1618 kg/m3
Loose 1586 kg/m3
Dense 0.48
Void ratio, (e)
Medium 0.67
Loose 0.71
Friction Angle (ø) 32-55°
219

220 5.3. Excavation Wall

221 The tied back anchor excavation wall was 1.52 m deep by 4.9 m wide constructed of three

222 pieces of 2.35 mm thick sheet metal. The pieces were attached by a continuous wale consisting

223 of a 2.5 cm x 2.5 cm square tube at each tie back level. The excavation wall was set 0.76 m

224 from the front of the testing chamber (Figure 2). The excavation wall was supported by three

225 levels of anchors spaced equidistantly in the vertical and horizontal directions (Figure 2a). The

226 anchors were installed 15° from horizontal and made of 6.4mm diameter stainless steel rods.

227 The unbonded portion of each was encased in a PVC tube; a discussion of the scaling and

228 manufacturing of the anchors is available elsewhere [35].

229

230
 14

231 5.4. Concrete

232 A large variety of concrete mixes were tested to achieve a 3.45 MPa compressive strength with

233 a modulus of elasticity of 206.8 MPa. These included varying the water/cement ratio, the

234 aggregate content and types, and the additives. Their performance characteristics were

235 investigated at 7, 14, and 28 days. Table 4 shows trial compressive strength, (fc) and modulus of

236 elasticity, (E) values. The modulus of elasticity values were obtained directly from the

237 compressive testing of cylinders. This crude approach was used as a quick screening method to

238 narrow the field of possible mixes to a few that warranted more extensive investigation.

239

240 The main difficulty in obtaining a weak concrete mix was that most of the additives that were

241 effective in lowering strength had an unduly extreme impact on the modulus of elasticity or

242 failed to keep the particles in suspension. In many cases, lowering the strength by one order of

243 magnitude caused the modulus of elasticity to decrease by two orders of magnitude. Because of

244 the lyocratic nature of bentonite, initially it appeared as an attractive additive in controlling

245 bleeding, but it too resulted in an excessive reduction in the modulus of elasticity. Excessive

246 bleeding was also typical of many of the trial mixes. In order to combat all of these problems, a

247 variety of aggregates and additives were tries including granite, quartz, feldspar, plagioclase,

248 alumina, and kaolinite. Ultimately limestone was selected. The limestone was ground to

249 cement particle size as the aggregate, which alleviated the need for a separate admixture. The

250 limestone was obtained from a ready-mix concrete distributor as material typically used for

251 local concrete mixes. The limestone was crushed in the laboratory to create a two-part

252 fine/very-fine mix. The fine limestone passed the #20 sieve and was retained by the #30 sieve.
 15

253 The very fine limestone passed the #30 and was retained by #50 sieve. The selected

254 composition is shown in Table 5. As part of the mix design, masonry cement was used instead

255 of a Portland cement to help control strength gain.

256

257 Table 4 Sampling of some of the trial concrete mixes

Water Binder Aggregate 7-days test 14-days test 28-days test
Specimen
percentage percentage percentage
number fc (MPa) E (MPa) fc (MPa) E (MPa) fc (MPa) E (MPa)
(%) (%) (%)

1 22.90 9.92 67.18 3.52 438.3 - - 3.90 575.4

2 23.19 9.42 67.39 2.00 431.2 1.86 476.0 1.09 60.0
3 23.19 9.42 67.39 1.86 201.1 2.45 497.9 1.76 205.9
4 21.53 8.75 69.72 1.07 194.8 - - 1.21 292.8
5 20.41 10.20 69.39 1.72 379.5 1.55 375.9 1.17 265.6
6 19.20 10.40 70.40 1.28 198.4 1.07 178.4 0.97 227.6
7 20.69 8.97 70.34 0.86 172.4 0.79 152.8 0.79 212.0
8 20.69 8.96 70.35 0.29 233.7 0.40 114.9 0.26 57.9
9 23.15 9.40 67.45 2.07 382.9 2.10 425.0 1.90 434.7
10 14.43 6.56 79.01 2.24 391.2 2.24 368.0 1.72 298.7
11 19.01 9.15 71.84 0.69 155.4 0.38 82.1 0.45 100.8
12 17.27 9.35 73.38 0.86 241.2 0.62 165.8 0.45 97.8
13 15.44 9.56 75.00 1.00 384.2 0.59 201.9 0.57 150.0
14 19.01 9.15 71.84 0.93 208.4 - - 0.83 248.7
15 17.27 9.35 73.38 1.52 341.8 1.03 176.9 0.94 203.5
16 15.44 9.56 75.00 1.41 242.6 1.55 329.2 1.21 204.5
17 19.31 10.32 70.37 0.63 75.25 0.57 59.4 - -
18 22.90 9.92 67.18 1.79 297.5 - - - -
19 22.90 9.92 67.18 1.90 333.1 - - - -
258
259
260 Table 5. Concrete mix
Constituents Quantity
Water 480 g
Cement 325 g
Fine limestone 1100 g
Very fine limestone 1100 g
Superplasticizer 14 ml
261

262 Despite concerns that the non-traditional composition of the concrete mix would have

263 insufficient tensile strength (one-tenth of that of the compressive strength), this was not a
 16

264 problem. Laboratory tests on a variety of these non-traditional compositions confirmed the 1 to

265 10 ratio, which is the traditional assumption.

266

267 5.5. Reinforcing Steel

268 To produce model reinforcement based on similitude was also especially challenging, as

269 previously identified by Lee and Woo [36]. When Lee and Woo [36] created a 1/5th scale,

270 3-story RC frame, they relied upon matching the yield forces to achieve the similitude for

271 reinforcement instead of yield stresses. Specifically this was done by modifying the

272 cross-sectional area of the reinforcing.

273

274 In the study herein, lead wires 12.2mm in diameter were chosen as a material closely matching

275 1/10th the tensile strength and modulus of elasticity of reinforcing steel. Mechanical properties

276 were established through tensile testing using a modified version of the tensile testing

277 apparatus. This was done by attaching alligator clips to the free ends of the lead. Displacement

278 was measured checking wire elongation with a rigidly attached scale positioned behind the

279 lead. A single strand was used for testing. The material had a modulus elasticity of 20684.3

280 MPa and an average tensile yield stress of 19 MPa (Figure 4).

281

282 The concrete/reinforcing bond was the next issue. Due to the relatively low tensile capacity of

283 lead, a direct pull-out test of the embedded wire could not be effectively conducted to discern

284 the effects of different wire configurations within the concrete. Instead, experiments were

285 undertaken using the model concrete mix and steel wire of the same gauge as the lead.
 17

286 Replicating the Knurland effect of standard steel reinforcing was seen as important to generate

287 the expected mechanical interlock.

288

289
290 Figure 4. Stress-strain behavior of lead
291

292 Bond testing was done by embedding steel wire in concrete cylinders 50.8mm high and

293 25.4mm in diameter. In one specimen, a single strand of wire was placed directly in the

294 concrete with no modification. The second sample also included a single strand, but this one

295 had been twisted around a 12.7mm diameter threaded rod and tested after the rod was removed.

296 The third sample used two wire strands twisted using an electric drill. In the first sample, the

297 wire pulled out of the cylinder under only 10 N, the other two samples exceeded the capacity of

298 the testing equipment at 220 N.

299

300 The actual models were produced by mechanically twisting single strands together to create

301 either double or triple strands (depending upon the area required in the specific cross section

302 based on the scaling of the full-sized building design) [Figure 5]. The stirrups in the beams and

303 spiral reinforcing of the columns were comprised of continuous, single, non-twisted members.
 18

304 Allowing the wire reinforcing to develop a light film of rust through exposure in a cure room

305 (to assist with surface bonding) was not an option since lead does not rust.
306

307
308 Figure 5. Model beam and column reinforcement of twisted lead
309
310

311 5.6. Reinforced Concrete Testing

312 Four-point bending tests were conducted at 14 days on a minimum of 3 beam and 3 column

313 sections (Figure 6). Figure 7 compares the ratio of allowable to design moment compared to the

314 ratio of the allowable to the design curvature. This was achieved by the close match between

315 the strength and stiffness having 1/10th of those of the prototype materials.

316

317 5.7. Geometrical Divergences

318 Despite these efforts, some scaling divergences remained even with respect to the geometry.

319 For example, instead of a single knurland bar, two wire strands whose cumulative area was

320 equivalent to a scaled single bar were used (Figure 5). This decision was made due to the

321 unavailability of small diameter lead rods and the inability to achieve a consisted twisted

322 profile with a single wire.

 19

323
324 Figure 6. Four-point bending test setup
325

326
327 Figure 7. Comparison of the ratios of allowable to design moment and curvature for the model
328 and prototype
 20

329 6. Model Construction and Installation Procedures

330 6.1. Model Construction

331 To minimize the variability of concrete quality, material sources were limited whenever

332 possible. All binder was taken from the same bag of masonry cement, all aggregate was

333 obtained from the same distributer, and an attempt was made to produce the aggregate from the

334 same portion of a larger aggregate pile (selecting all the needed aggregate for the entire testing

335 program in a single instance). Concrete mixing was standardized through strict procedural

336 standards.

337

338 The stirrups and spiral reinforcing were connected to the other metal pieces with a light

339 soldering instead of using tie wire (Figure 8a). This technique was also used for chair

340 placement (using small straight pieces of steel fencing wire) to ensure adequate concrete cover

341 in the beams (Figure 8b). The reinforcing for each column was initially assembled around a

342 25.4mm diameter wooden dowel (Figure 8a). When the soldering of the vertical pieces to the

343 spiral was complete, the dowel was used to relocate the column reinforcing into the formwork

344 and then removed (as discussed below).

345

346 The concrete formwork was made of Plexiglas pieces assembled with hot glue (Figure 9). The

347 transparent framework enabled the identification voids and honeycombing in the beams and

348 the columns and the easy disassembly of the formwork so as to prevent undue stress on the

349 models during formwork removal.

350
 21

351
352 Figure 8. Column and beam reinforcement configuration
353
354

355
356 Figure 9. Erection of the RC model
357

358 After the formwork was erected, the column reinforcing was inserted, from the top. After the

359 column reinforcing was properly located within the column’s height, the dowel was removed

360 and beam reinforcing was started.

 22

361 The stirrups, like the column spiral was formed from a continuous piece of lead that had been

362 wrapped around a rectangular, stainless steel form of appropriate size. Each discrete piece

363 represented the necessary area of reinforcing for the stirrups for a single beam. The wound

364 piece was centered in the empty beam form. Through predrilled holes located on the exterior

365 side of the outside columns, twisted double strands were threaded by hand from outside of one

366 column through the first beam, then through the next column and beam progressively, until the

367 set of paired strands finally emerged from the last column, also through pre-drilled holes in the

368 Plexiglas. Installation was done one paired strand at a time working from the bottom of the

369 model upwards. Care was taken not to cause any crossovers of these paired wires. This was

370 critical to facilitate the subsequent insertion of copper tremie tubes through the columns for the

371 concrete placement. After all the top and bottom beam reinforcing was threaded through the

372 models, the stirrup wire in each beam was distributed across the length of the beam at an even

373 spacing. The density was based on the maximum reinforcing needed at the beam-column

374 interfaces. The result was a slightly over-reinforced beam for shear at the mid-span.

375

376 Once all of the reinforcement was positioned, the soldering of the beams began. Each of the top

377 pieces was soldered to the stirrups at every fourth or fifth interval. Steel wires (acting as chairs)

378 were placed as needed. When all of the assembly reinforcing was complete, Plumbers putty

379 was spread on the exterior of all the Plexiglas joints to help prevent leakage. The bottom of the

380 form at the four column locations was left open to accommodate a 6.4mm threaded, steel rods,

381 which served as a shear keys between the model frame and the footing. During concrete

382 pouring, temporary plugs of Styrofoam were used to close the bottoms of the column. The steel
 23

383 rods were inserted prior to pouring. To avoid cold joints, the entire four-story structure was

384 poured pretty much continuously. The columns had the concrete tremied to the bottom using

385 temporary copper piping, which was lifted as the form filled. Each beam level was poured from

386 the top of its respective level, screeded, and then covered with a Plexiglas top piece to ensure

387 an even surface finish for later dead load application. At each beam level, a commercial

388 vibrator was applied to the outside of the formwork. Total pour time ranged from 4 to 8 hours.

389 No special curing agents or conditions were used. Cylinders were taken and broken in

390 compression when the final composite testing in the chamber occurred. This was targeted to

391 occur at 28 days.

392

393 6.2. Model Installation and Loading

394 To place the models into the testing chamber, they were lifted by crane. To prevent any damage

395 during transport, temporary wooden false work surrounded the Plexiglas frame, which

396 permitted an extremely rigid framework for lifting (Figure 9). The models were inserted into

397 the previously placed footings. Attachment was done with the introduction of Quickcrete in the

398 slightly overdrilled footing holes just prior to the model insertion. After the Quickcrete cured

399 but before the dead load was applied, the model was white-washed with two to three layers of a

400 plaster/water mix to create a stress-coat, which assisted in identifying crack development.

401 Staged lateral bracing of the model enable temporary access for plastering, as necessary. The

402 lateral bracing was located below the beams and consisted of short copper tube sections placed

403 over long, greased metal dowels, thereby permitting only in-plane motion (Figure 10).

404
 24

405

406 Figure 10. Uniform loading on RC frames held in plane by lateral bracing

407

408 Applied dead weight consisted of cut steel bar stock. Each piece was attached to a Plexiglas

409 sleeve 44.5mm in width. A very slight space was left between each sleeve to prevent the

410 unintentional reinforcing of the beam. The pieces of steel bar stock allowed for uniform beam

411 loading to account for the missing mass of the slab. Additional weight was added onto the

412 columns based on the tributary area, as a percentage of the combined dead and live load used as

413 the design load. Because of access restraints at the column-beam joints, the column loading for

414 each floor had to be relocated to the top of the structure in a cumulative fashion. These weights

415 were steel plates, each with a small hole drilled vertically through the center of the plate to

416 enable mounting onto the 6.4mm threaded rod that was cast into the concrete columns at the

417 time of the initial pour for connection to the footings. The loaded structure is shown in Figure

418 11.a, and the rods are in Figure 11.b. Throughout the column lengths, these rods were isolated

419 from the surrounding concrete with a flexible greased sleeve.

420
 25

421

422 Figure 11. Loaded RC frame

423

424 7. Instrumentation and Testing

425 Tests 2 and 6 were conducted to evaluate the response of RC structures to adjacent excavation.

426 Frames in each test were differentiated by East (E) and West (W) designations, depending on

427 their location within the testing chamber (Figure 2 & 3) and had different beam configurations.

428 The structural movements were measured by a series of dial gauges and linear voltage

429 displacement transducers (LVDTs) installed on the model buildings. This combination of

430 analog and electronic instrumentation was used to measure frame displacements.

431

432 The excavations were done progressively. In each monitoring stage, a depth of approximately

433 102mm of soil was removed from in front of the “wished in place” tieback excavation wall.

434 At pre-designated depths on the excavation wall, the tieback anchors that were in the testing

435 chamber (but until that point disengaged) were then loaded. Test readings of all elements in

436 the experiments were taken before and after loading, as well as after any delays of 2 hours or

437 more in the testing cycle. A total of 15 stages (excavation and anchor stressing) were

438 conducted to reach the final design grade. The results of the 3rd and 15th stages are presented
 26

439 in detail, herein. The 15th stage represents the final design grade excavation stage for Test 2

440 and 6 (1.00Hdg2 and 1.00Hdg6), and the 3rd stage represents excavation to 0.25Hdg2 (30.48 cm)

441 for Test 2 and 0.16Hdg6 for Test 6 (17.48 cm)

442

443 8. Results and Discussion

444 The results and their respective validation can be considered in four distinct areas of the

445 testing: the ground movements, the column base displacements, the story-level

446 displacements, and the resulting damage.

447

448 8.1. Ground Movements

449 The surface soil settlements in the free field are shown in Figure 12 multiplied by the scale

450 factor of 10 and overlaid atop field data collected by Peck [37]. The experimental data appears

451 in the middle of Region I for surface movement in sand with average workmanship.

452 Additionally, the maximum vertical displacement of free field (Figure 2) was around 0.7% H,

453 where H is the excavation depth. Furthermore, the trough profile shape was similar to that

454 proposed by Peck [37]. These results are, thus, well validated by the published literature.

455
 27

456

457 Fig. 12. Scaled free field ground movements overlain on field results reported by

458 Peck [37]

459 8.2. Column Base Displacements

460 The column base displacements just prior to the first post-tensioning (stage 3) (0.25Hdg2 and

461 0.16Hdg6) and at the end of excavation (stage 15) (1.00Hdg2 and 1.00Hdg6) are plotted in Figure

462 13; data not directly obtainable from specific instruments during experimentation were

463 derived by linear interpolation of the surrounding instruments. The displacements are (as

464 expected) smaller at the columns further from the excavation, except at the end of the T2E

465 testing for column b (Figure 13a). This resulted from the lack of lateral constraint caused the

466 absence of the 1st, 2nd and 3rd floor beams in the middle bay. Table 6 shows the column

467 displacements at the end of testing and the resulting maximum differential vertical gradients

468 for each bay (as calculated from the differential vertical displacement of adjacent columns).

469
 28

470

471 Figure 13. Column base displacements at two excavation stages

472

473 Table 6. Horizontal (x) and vertical (y) displacements of column foundations at final
474 excavation stage (mm) and resulting maximum gradients
Test-frame Column a Column b Column c Column d Maximum Maximum Maximum
gradient gradient gradient
x y x y x y x y
between between between
a and b b and c c and d
T2E 3.51 10.59 6.25 5.36 3.65 2.78 1.04 0.20 0.009 0.004 0.004
T2W 4.80 10.72 4.01 6.88 3.49 3.57 2.97 0.25 0.006 0.006 0.006
T6W 1.19 4.70 1.00 2.46 0.81 0.64 0.71 0.08 0.004 0.003 0.001
475 Note: Bolded values were obtained by interpolation

476

477 As shown in Figure 14, the vertical displacement results align well with results reported by

478 Castaldo and De Iuliis [38] for their numerical study of a 26.6m high, 8-story, 2-span (each

479 6.30m), RC plane-frame building on sand soil. Their maximum gradient was only 0.001 versus

480 0.004-0.0009 seen in the experimental work. This is as expected for a structure located more

481 than 10 times further from the excavation [16.5 m versus 1.5 m (prototype)].

482
 29

483

484 Figure 14. Comparison of maximum gradient results with the study of Castaldo and De Iuliis

485 [38]

486 8.3. Story-Level Displacements

487 For further comparisons, analytical models were generated in the structural analysis modeling

488 package Sofistik [39], as reported in detail in Laefer et al. [40]. Analyses were conducted for

489 each of the three frames for multiple excavation stages (Table 7); in Test 2, building

490 displacements were recorded only at the base of the columns and at the fourth story, while in

491 the subsequent Test 6, the instrumentation was expanded to record displacements at all story

492 levels. Generally the experimental results aligned well with the numerical analyses results. The

493 RC frames tended to conform with the ground, and the lack of a grade beam allowed a two-part

494 behavior: the upper stories exhibited fairly uniform swaying that was not reflected along the

495 lowest level, where the base of the bottom columns displaced laterally with minimal tilt.

496 Notably, the data recorded at the end of excavation (1.00Hdg2 and 1.00Hdg6) were more

497 consistent with the numerical results than that recorded early in the process (0.25Hdg2 and
 30

498 0.16Hdg6), because of instrumentation sensitivity and the very small movements that were

499 generated at that testing stage.

500
501 Table 7. Maximum displacements of analyzed frames due to applied soil profiles (mm)
Applied Story 1 Story 2 Story 3 Story 4
Test Test step
Displacements X Y X Y X Y X Y
Experimental NM NM NM NM NM NM 0.58 1.54
0.25Hdg2
Test 2 Column 0.48 1.95 0.73 2.13 0.99 2.29 1.58 2.35
East Experimental NM NM NM NM NM NM 8.28 12.40
1.00Hdg2
Column 3.15 10.8 4.94 11.04 6.88 11.18 8.73 11.25
Experimental NM NM NM NM NM NM 0.66 1.54
0.25Hdg2
Test 2 Column 0.26 1.97 0.54 2.21 0.75 2.37 1.15 2.44
West Experimental NM NM NM NM NM NM 9.68 10.24
1.00Hdg2
Column 4.39 11.0 6.03 11.28 7.68 11.44 9.56 11.51
Experimental 0.03 0.10 0.05 0.10 0.08 0.08 0.13 0.10
0.16Hdg6
Test 6 Column 0.04 0.15 0.06 0.15 0.07 0.15 0.09 0.15
West Experimental 1.47 4.72 2.21 4.75 3.10 4.75 4.37 4.75
1.00Hdg6
Column 1.17 5.02 1.90 5.26 2.61 5.41 3.51 5.48
502 NM denotes that the data was not measured during the experiment
503

504 8.4. Damage

505 At the end of excavation, the maximum relative gradients between the columns (calculated

506 from deflections and presented in Table 6) were 0.009 for T2E, 0.006 for T2W, and 0.004 for

507 T6W. These amounts all exceeded the previously published damage-inducing thresholds of

508 0.006 for structural damage [41, 42] and 0.003 for cracking [41]. Commensurate damage

509 levels were recorded in the actual models (Figure 15). In Figure 15, the notations next to the

510 cracks show the stage of the crack, for example crack 0.95Hdg2 at the top of the middle bay of

511 T2W is the crack the formed at 95% of the excavation depth. The direction of the arrows

512 below the frames indicates the excavation location; these images show the damage for all 15

513 experimentation stages. There was relatively good agreement between the gradient limits and

514 the extent of the damage. Deviations from the predictions were mostly attributable to

515 nonlinear material behavior, changes in stiffness during the excavation process, continuously
 31

516 changing load patterns, and the heterogeneous nature of concrete. The above-described tests

517 show the overall efficacy of the modeling strategy for small-scale reinforced concrete frames

518 under 1-gravity loading.

519

520
521 Figure 15. Locations and stages of structural cracks in the frames with directional arrow
522 showing excavation position (note that only structural cracking was visible due to model size)
523

524 9. Conclusions

525 To ensure the reliability of testing of reinforced concrete structural systems in 1-gravity,

526 soil-structure experiments less than full scale, material scaling in addition to geometrical

527 scaling requirements must be met. Conforming to these requirements and application processes

528 regarding construction and loading of models facilitate the generation of considerably

529 small-scale (1/10th scale) RC models and their behavior akin to actual full-scale structures.

530 This paper provides useful theoretical and practical considerations for soil-structure

531 experimentation to simulate small-scale RC building models. Alteration of the mechanical

532 properties of model materials is needed to capture the prototype behavior. Otherwise,

533 inappropriate materials used in reduced scale experiments can cause departures from complete
 32

534 similitude. The relative success of the developed materials was verified by comparing surface

535 settlement of model soil, reinforced building response (column and story movements), and

536 cracking patterns with trends seen by other researchers in the field and in numerical simulation.

537

538 Acknowledgments

539 The experimental work, conducted under the supervision of Dr. James H. Long and Edward J.

540 Cording, was funded by National Science Foundation grant CMMI9713854 and UIUC and

541 realized through the tireless dedication of several dozen undergraduates. The Schnabel

542 Laboratory was made possible through generous funding by the Schnabel Foundation

543 Company, donated technical assistance of Mr. Keith Brandau of Frauenhoffer Associates, and

544 institutional leadership from Dr. Tony Graziano of UIUC.

545

546 References

547 [1] Laefer DF (2001) Prediction and assessment of ground movement and building damage

548 induced by adjacent excavation. PhD Thesis, University of Illinois at Urbana-Champaign,

549 USA.

550 [2] Laefer DF, Hong LT, Erkal A, Long JH & Cording EJ (2011) Manufacturing, assembly,

551 and testing of scaled, historic masonry for one-gravity, pseudo-static, soil-structure

552 experiments. Construction and Building Materials, 25, pp.4362–4373.

553 [3] Zlokarnik M (2001) Scale-up of processes using material systems with variable physical

554 properties. Chemical and Biochemical Engineering Quarterly, 15, pp.43-47.

 33

555 [4] Shaughnessy EJ, Katz IM & Schaffer JP (2005) Introduction to fluid mechanics. Oxford

556 University Press, Oxford, England.

557 [5] Wood DM (2004) Geotechnical modelling. Spon Press, Oxfordshire, UK.

558 [6] Van Geem KM, Žajdlík R, Reyniers MF & Marin GB (2007) Dimensional analysis for

559 scaling up and down steam cracking coils. Chemical Engineering Journal, 134, pp.3–10.

560 [7] Buckingham E (1914) On physically similar systems: Illustrations of the use of dimensional

561 equations. Physical Review, 4, pp.345-376.

562 [8] Bridgman PW (1922) Dimensional Analysis. Yale University Press, New Haven, CT.

563 [9] Lanchester FM (1940) The theory of dimensions and its applications for engineers.

564 Crosby-Lockwood, London.

565 [10] Langhaar HL (1951) Dimensional analysis and theory of models. John Wiley & Sons,

566 Inc., New York.

567 [11] Schuring DJ (1977) Scale models in engineering: Fundamentals and applications.

568 Pergamon Press, Oxford, England.

569 [12] Datin PL & Prevatt DO (2013) Using instrumented small-scale models to study

570 structural load paths in wood-framed buildings. Engineering Structures, 54, pp.47–56.

571 [13] Liu J, Liu F, Kong X & Yu L (2014) Large-scale shaking table model tests of aseismic

572 measures for concrete faced rock-fill dams. Soil Dynamics and Earthquake Engineering,

573 61-62, pp.152-163.

574 [14] El Sawwaf MA (2007) Behavior of strip footing on geogrid-reinforced sand over a soft

575 clay slope. Geotextiles and Geomembranes, 25, pp.50–60.

 34

576 [15] Bailey CG & Toh WS (2007) Small-scale concrete slab tests at ambient and elevated

577 temperatures. Engineering Structures, 29, pp.2775–2791.

578 [16] Anil Ö & Altin S (2007) An experimental study on reinforced concrete partially infilled

579 frames. Engineering Structures, 29, pp.449–460.

580 [17] Žarnić R, Gostič S, Crewe AJ & Taylor C (2001) Shaking table tests of 1:4 reduced-scale

581 models of masonry infilled reinforced concrete frame buildings. Earthquake Engineering

582 and Structural Dynamics, 30, pp.819-834.

583 [18] De-la-Colina J, Valdés-González J & González-Pérez CA (2013) Experiments to study the

584 effect of foundation rotation on the seismic building torsional, response of a reinforced

585 concrete space frame. Engineering Structures, 56, pp.1154–1163.

586 [19] Lu Y, Vintzileou E, Zhang G-F & Tassios TP (1999) Reinforced concrete scaled columns

587 under cyclic actions. Soil Dynamics and Earthquake Engineering, 18, pp.151-167.

588 [20] Ohtaki T (2000) An experimental study on scale effects in shear failure of reinforced

589 concrete columns.12th World Conference on Earthquake Engineering; Auckland, New

590 Zealand, 30 January - Friday 4 February 2000.

591 [21] Kim W, El-Attar A & White RN (1988) Small-scale modeling techniques for reinforced

592 concrete structures subjected to seismic loads. Technical Report NCEER-88-0041, Cornel

593 University.

594 [22] Hirai T, Itoh M, Nishikawa T & Takei K (1995) Study on strength of thick reinforced

595 concrete slab with opening. Nuclear Engineering and Design, 156, pp.l-16.
 35

596 [23] Skjærbæk PS, Taşkin B, Kirkegaard PH & Nielsen SRK (1997) An experimental study of

597 a midbroken 2-bay, 6-storey reinforced concrete frame subject to earthquakes. Soil

598 Dynamics and Earthquake Engineering, 16, pp.373-384.

599 [24] Klinger RE, Rubiano NR, Bashandy TR & Sweeney SC (1997) Evaluation and analytical

600 verification of shaking table data from infilled frames. The Masonry Society Journal, 15,

601 pp.33-41.

602 [25] Lu Y (2002) Comparative study of seismic behavior of multistory reinforced concrete

603 framed structures. Journal of Structural Engineering, 128, pp.169–178.

604 [26] Holden T, Restrepo J & Mander JB (2003) Seismic performance of precast reinforced and

605 prestressed concrete walls. Journal of Structural Engineering (ASCE), 129, pp.286-296.

606 [27] Thomsen JH and Wallace JW (2004) Displacement-based design of slender reinforced

607 concrete structural walls—Experimental verification. Journal of Structural Engineering

608 (ASCE), 130, pp.618–630.

609 [28] Antoniades KK, Salonikios TN & Kappos AJ (2005) Tests on seismically damaged

610 reinforced concrete walls repaired and strengthened using fiber-reinforced polymers.

611 Journal of Composites for Construction (ASCE), 9, pp.236–246.

612 [29] Kim NS, Lee JH & Chang SP (2009) Equivalent multi-phase similitude law for

613 pseudodynamic test on small scale reinforced concrete models. Engineering Structures, 31,

614 pp.834-846.

615 [30] Guoxing C, Su C, Xi Z, Xiuli D, Chengzhi Q & Zhihua W (2015) Shaking-table tests and

616 numerical simulations on a subway structure in soft soil. Soil Dynamics and Earthquake

617 Engineering, http://dx.doi.org/10.1016/j.soildyn.2014.12.012

 36

618 [31] Uniform Building Code (UBC) (1997) Proc., UBC 97—Int. Conf. of Building Officials,

619 Vol. 2, International Conference of Building Officials, Baltimore.

620 [32] American Concrete Institute (ACI) (1995) Building code requirements for structural

621 concrete (ACI 318-95) and commentary (ACI 318R-95), ACI, Detroit.

622 [33] Tomaževič M. & Klemenc I (1997) Verification of seismic resistance of confined

623 masonry buildings. Earthquake Engineering & Structural Dynamics, 26, pp.1073–1088.

624 [34] Laefer DF, Erkal A, Cording E, Long J & Truong-Hong L (2010) Theoretical solutions for

625 strength-scaled unreinforced masonry for scaled soil-structure experimentation. Journal of

626 Testing and Evaluation (ASTM), 38, pp.449-457.

627 [35] Laefer DF (2014) Solutions for model-scale, tied-back anchors and sheet pile walls.

628 Geo-Congress 2014 Technical Papers: Geo-Characterization and Modeling for

629 Sustainability. pp.2578-2587, doi: 10.1061/9780784413272.24.

630 [36] Lee HS & Woo SW (2002) Seismic performance of a 3-story RC frame in a

631 low-seismicity region. Engineering Structures, 24, pp.719–734.

632 [37] Peck RB (1969) Deep excavation and tunneling in soft ground. Proc., 7th International

633 Conference on Soil Mechanics and Foundation Engineering, Sociedad Mexicana de

634 Meccannica de Suelos, Mexico, pp.225-290.

635 [38] Castaldo P & De Iuliis M (2014) Effects of deep excavation on seismic vulnerability of

636 existing reinforced concrete framed structures. Soil Dynamics and Earthquake Engineering,

637 64, pp.102–112.

638 [39] Sofistik. Structural analysis modeling package program, (V10.31-23)

639 http://www.sofistik.com/en/ Last access date: 1 February 2015.

 37

640 [40] Laefer DF, Ceribasi S, Long JH & Cording EJ (2009) Predicting RC frame response to

641 excavation-induced settlement. Journal of Geotechnical and Geoenvirontal Engineering,

642 135, pp.1605-1619.

643 [41] Skempton AW & MacDonald DH (1956) The allowable settlement of buildings. ICE

644 Proceedings: Engineering Divisions, Vol. 5, pp.727-768.

645 [42] Bjerrum L (1963) “Discussion session IV.” Proc., European Conference on Soil

646 Mechanics and Foundation Engineering, Wiesbaden, Germany, Vol. 2, pp.135–137.

647

648 Figure Captions

649 Figure 1. Three-dimensional floor plan of RC slab prototype with tributary area

650 Figure 2. Test setup

651 Figure 3. Model RC buildings

652 Figure 4. Stress-strain behavior of lead

653 Figure 5. Model beam and column reinforcement of twisted lead

654 Figure 6. Four-point bending test setup

655 Figure 7. Comparison of the ratios of allowable to design moment and curvature for the model

656 and prototype

657 Figure 8. Column and beam reinforcement configuration

658 Figure 9. Erection of the RC model

659 Figure 10. Uniform loading on RC frames held in plane by lateral bracing

660 Figure 11. Loaded RC frame

661 Fig. 12. Scaled free field ground movements overlain on field results reported by
 38

662 Peck [37]

663 Figure 13. Column base displacements at two excavation stages

664 Figure 14. Comparison of maximum gradient results with the study of Castaldo and De Iuliis

665 [38]

666 Figure 15. Locations and stages of structural cracks in the frames with directional arrow

667 showing excavation position (note that only structural cracking was visible due to model size)

668

669 Table Captions

670 Table 1. Main characteristics of reinforced concrete scale models used in experiments

671 Table 2. Loading applied to the RC prototype

672 Table 3. Physical properties of the model soil

673 Table 4 Sampling of some of the trial concrete mixes

674 Table 5. Concrete mix

675 Table 6. Horizontal (x) and vertical (y) displacements of column foundations at final

676 excavation stage (mm) and resulting maximum gradients

677 Table 7. Maximum displacements of analyzed frames due to applied soil profiles (mm)