You are on page 1of 9

Small-Scale Modeling of Reinforced Concrete Structural

Elements for Use in a Geotechnical Centrifuge


J. A. Knappett1; C. Reid2; S. Kinmond3; and K. O’Reilly4
Downloaded from ascelibrary.org by Indian Institute of Technology Bombay on 02/01/22. Copyright ASCE. For personal use only; all rights reserved.

Abstract: This paper discusses the modeling of reinforced concrete structural elements for use in geotechnical centrifuge modeling of
soil-structure interaction problems. Centrifuges are employed in geotechnical modeling so that the nonlinear constitutive behavior of soil
in small-scale models can be correctly modeled at prototype scale. Such models typically necessitate large scale factors of between 1∶20 and
1∶100, which is significantly larger than most conventional small-scale structural modeling. A new model concrete has been developed
consisting of plaster, water, and fine sand as a geometrically scaled coarse aggregate that can produce a range of model concretes with
cube compressive strengths between 25–80 MPa. Reinforcement is modeled using roughened steel wire (beams) or wire mesh (slabs).
To illustrate the validity of the modeling technique, a series of three- and four-point bending tests were conducted on model beams designed
to represent a 0:5 × 0:5 m square section prototype beam at 1∶40 scale, and model slabs designed to represent a prototype slab with plan
dimensions of 4:8 × 4:8 m and 0.4 m deep (also at 1∶40 scale). The amount of longitudinal reinforcement was varied and tests both with and
without shear reinforcement were conducted. The models were able to accurately reproduce both shear and flexural (bending) failures when
loaded transversely. The load capacity (strength), bending stiffness, and ductility were shown to be simultaneously and appropriately scaled
over a range of scaling factors appropriate for geotechnical centrifuge testing, and the technique therefore provides a significant improvement
in the ability to accurately model soil-structure interaction behavior in centrifuge models. DOI: 10.1061/(ASCE)ST.1943-541X.0000371.
© 2011 American Society of Civil Engineers.
CE Database subject headings: Centrifuge models; Concrete beams; Concrete slabs; Scale models; Size effect; Concrete structures.
Author keywords: Centrifuge models; Concrete beams; Concrete slabs; Scale models; Size effect.

Introduction 217 mm deep and 3.4 m long. The collapse loads are expressed
at prototype scale at N 2 times the load measured in the model.
Because concrete is a quasi-brittle material, care must be taken When these are normalized by the measured full-scale value
when interpreting the results of small-scale tests on reinforced con- (N ¼ 1) substantial overstrength is observed as N increases (i.e.,
crete, because a significant size effect exists. This effect manifests as the models get smaller). A power law fit to this combined data
itself as a significant overstrength when results from reduced-scale set, which is proposed herein, suggests that the overstrength
model tests are scaled up to equivalent prototype values, an effect increases as N 1=10 . A similar overstrength has previously been re-
that becomes more pronounced as the scaling factor is increased. ported for beams failing in shear (Bažant and Yu 2005). The size
Fig. 1 shows limiting transverse loads on model beams (V) from effects exhibited by reduced-scale concrete models (as shown in
test results for beams failing in flexure (i.e., containing sufficient Fig. 1) arise because the strength of reinforced concrete elements
shear reinforcement to suppress shear failure) presented by Litle typically involves fracture of the concrete, the development of
and Paparoni (1966) and Belgin and Sener (2008). In each case, which depends on the size of flaws within the concrete. The flaw
a full-scale beam was tested (1∶1 scale), followed by reduced-scale size is related to the size of the aggregate such that the flaw size in
models at scales of 1 : N using the same concrete mix with a small reduced-scale models made with similar aggregates to full-scale
maximum aggregate size and no aggregate scaling. The full-scale beams does not scale proportionately with sample dimensions
beams of Belgin and Sener were slightly larger than those of Litle (Bažant 2005). The net result of this size effect is that structural
and Paparoni, being 300 mm deep and 4.6 m long, compared to models are typically limited to scales of 1∶4 to 1∶15 (e.g., Litle
and Paparoni 1966; Bažant and Kazemi 1991; Belgin and Sener
1
Lecturer, Division of Civil Engineering, Univ. of Dundee, Dundee,
2008).
DD1 4HN, Scotland, UK (corresponding author). E-mail: j.a.knappett@ The modeling of structural behavior in reduced-scale models
dundee.ac.uk is also of interest to geotechnical engineers for the study of soil-
2
Undergraduate student, School of Engineering, Univ. of Durham, structure interaction problems. Examples of such problems involv-
South Rd., Durham, DH1 3LE, UK. ing bending include kinematic lateral loading on piles owing to
3
Undergraduate student, Division of Civil Engineering, Univ. of relative soil-pile movement during earthquakes (modeled as a
Dundee, DD1 4HN, Scotland, UK. beam), or the structural response of a retaining wall owing to an
4
Undergraduate student, Division of Civil Engineering, Univ. of adjacent excavation (modeled as a slab). Because the strength of
Dundee, DD1 4HN, Scotland, UK.
soil is dependent on the effective confining stresses owing to its
Note. This manuscript was submitted on July 29, 2009; approved on
December 27, 2010; published online on December 29, 2010. Discussion
self weight, reduced-scale models of geotechnical constructions
period open until April 1, 2012; separate discussions must be submitted for are commonly tested in the enhanced gravity field of a geotechnical
individual papers. This paper is part of the Journal of Structural Engineer- centrifuge. A 1 : N scale model tested at Ng in a centrifuge will
ing, Vol. 137, No. 11, November 1, 2011. ©ASCE, ISSN 0733-9445/2011/ have the same effective stresses acting at homologous points in
11-1263–1271/$25.00. both the model soil and the full-scale prototype soil. Under these

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / NOVEMBER 2011 / 1263

J. Struct. Eng., 2011, 137(11): 1263-1271


correctly scaled bending stiffness (EI). Owing to the difficulties in
scaling sectional properties exactly at such small scales, model sec-
tions are typically made from different materials (E) to the full-
scale prototype element; e.g., polymeric materials (Abdoun and
Dobry 2003) or aluminum alloys (Hayward et al. 2000; Knappett
and Madabhushi 2009) with a different section (I) such that EI in
the reduced-scale model is N 4 times smaller than the full-scale
prototype for similitude of bending stiffness (Table 1). The disad-
vantage of this methodology is that simultaneous scaling of both
deformation and strength behavior is not possible, owing to the
different section and material used. This may be acceptable if
Downloaded from ascelibrary.org by Indian Institute of Technology Bombay on 02/01/22. Copyright ASCE. For personal use only; all rights reserved.

only small deflection behavior is to be studied, but it is becoming


increasingly desirable to study the collapse mechanics of soil-
structure systems in which the relative soil-structure strength is also
important.
This paper presents a newly developed set of modeling materials
and techniques for modeling reinforced concrete at larger values
of N (N ¼ 40 or a scaling factor of 1∶40 is considered herein) that
will give reasonable strength and stiffness behavior at prototype
Fig. 1. Scale effects observed in previous structural modeling studies, scale under transverse loading when tested within a geotechnical
beams failing in flexure centrifuge.

conditions, a set of scaling laws has been developed for common Model Reinforced Concrete
parameters of interest, as detailed by Garnier et al. (2007), and
these may be used to fabricate reduced-scale models of structural Model Concrete
elements. The scaling laws for centrifuge modeling are given in
The same concrete cannot be used as in a full-scale concrete beam
Table 1, alongside those of a conventional structural model in owing to the size of the aggregates commonly used in concrete and
which only geometric scaling is employed (from Harris and Sabnis the high scaling factors used herein (e.g., a piece of aggregate
1999). The size of interest of most soil-structure interaction 10 mm in diameter would be of the same order of magnitude as
problems compared to the size and capacity of most geotechnical the size of a 0.5 m deep prototype beam reduced to 1∶40 scale).
centrifuges typically necessitates scales between 1∶20 and 1∶100 Gypsum-based mortars are developed here as a model concrete.
being used (e.g., a prototype consisting of a 15 m long pile sur- A review of previous work on such materials is given by Harris
rounded by soil becomes a 250 mm long model at a scaling factor and Sabnis (1999).
of 1∶60 scale and tested at 60 g). If reinforced concrete elements Four different model mixes of plaster, sand, and water are
are to be modeled at such scaling factors, Fig. 1 would suggest considered to model concretes of different strengths, as detailed
that significant overstrength would be expected in the equivalent in Table 2. Mixes 1 and 2 were based on a beta-form plaster
prototype elements. Indeed, previous attempts to model reinforced (Surgical plaster, Lafarge Prestia, France), and Mixes 3 and 4
concrete elements directly in centrifuge models (e.g., Ito et al. were based on an alpha-form plaster (Crystacal D, Lafarge Prestia,
2006) have kept scaling factors ≤ 1∶20, in line with conventional France). Alpha-form plasters are generally stronger than beta-
structural testing. This limits the size of the prototype that can be forms. In Mixes 1 and 2, the water/plaster ratio (W/P) was varied
practically modeled in a centrifuge, owing to space and payload to investigate the effect this had on strength properties, although
considerations. preliminary testing revealed that for W=P < 0:9 the model concrete
At higher scaling factors, reinforced concrete elements such as was not sufficiently workable to produce consistent reinforced
piles or walls have only been modeled at a very crude level. Before model elements. The alpha-form plaster used in Mixes 3 and 4
failure, the response of a soil-structure system is chiefly influenced required a lower W/P, as suggested by the manufacturer. In all
by the relative soil-structure flexibility. It is therefore common mixes, the aggregate/plaster ratio (A/P) was maintained as 1∶1
practice to design reduced-scale models of piles and walls to have by weight. Uniformly graded Congleton HST 95 silica sand was
used as the model aggregate because the grading curve (particle
size distribution) was broadly equivalent to a typical coarse aggre-
Table 1. Scaling Laws gate grading curve, here taken from BS 882 (BSI 1992), scaled
Conventionala Centrifuge model: down by between 1∶30 and 1∶50, as shown in Fig. 2.
Property Dimensions model: prototype prototype An average compressive strength (f c;100 ) for each mix was found
by casting six 100 × 100 mm standard cubes and performing
Stress (σ) ML1 T2 1∶1 1∶1 unconfined compression tests. Correlation of these strengths to
Strain (ε) 1 1∶1 1∶1 150 × 150 mm cube strength (f c;150 ) and 60 × 1200 (150 × 300 mm)
Young’s modulus ML1 T2 1∶1 1∶1 cylinder strength (f 0c ) may be accomplished by using the relation-
Length L 1:N 1:N ships presented by Mansur and Islam (2002):
Force ML T2 1 : N2 1 : N2
Bending stiffness (EI) ML3 T2 1 : N4 1 : N4 f c;150 ¼ 0:91f c;100 þ 3:62 ð1Þ
Bending moment (M) ML2 T2 1 : N3 1 : N3
Curvature (κ) L1 1 : N1 1 : N1
and
a
“Practical true model,” after Harris and Sabnis (1999). f 0c ¼ 0:90f c;100  6:26 ð2Þ

1264 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / NOVEMBER 2011

J. Struct. Eng., 2011, 137(11): 1263-1271


Table 2. Model Concrete Mixes
Mix ID Plaster W/P A/P f c;100 (MPa) f r (MPa) f 0c (MPa)a f c;150 (MPa)b
1 Beta 1∶1 1∶1 26.3 2.02 17.4 27.6
2 Beta 0:9∶1 1∶1 35.6 2.47 25.8 36.0
3 Alpha 0:6∶1 1∶1 50.9 3.90 39.6 49.9
4 Alpha 0:5∶1 1∶1 81.7 4.42 67.2 78.0
a
Calculated values, using Eq. (2).
b
Calculated values, using Eq. (1).
Downloaded from ascelibrary.org by Indian Institute of Technology Bombay on 02/01/22. Copyright ASCE. For personal use only; all rights reserved.

Fig. 2. Geometrical scaling of coarse aggregate

where strengths are in MPa. The modulus of rupture (f r ) was used


to characterize the tensile behavior of the model concrete, because
it was felt to be more appropriate for describing the behavior of Fig. 3. Strength properties of model concrete mixes
the concrete under bending-induced tensile stresses than the direct
tensile strength using standard sized samples that would mask any
noted by Bažant and Li (1995), the actual f r in a full-scale concrete
scale effects. Bažant and Li (1995) reported a significant size effect
section 500 mm deep is expected to be approximately 1.9 times
on the modulus of rupture, with f r increasing with reducing sample
smaller than the values from modulus of rupture tests on the con-
size. For the model concrete to provide a good approximation to
ventional smaller modulus of rupture specimens, closer to the
full-scale concrete, the modulus of rupture at the model size of
design provisions of the ACI code (ACI 2002) shown in Fig. 3.
a structural element is required to be close to that of the concrete
This suggests that the model concrete appropriately scales f r .
in the prototype/full-scale element. Ten unreinforced model con-
crete beams at the same scale as the subsequent reinforced models Model Reinforcement
(i.e., 12:5 × 12:5 × 200 mm long at model scale) were tested in
standard four-point bending to obtain an average modulus of rup- The model concrete was reinforced using stainless steel wire.
ture appropriate to the model structural elements for each concrete Two sizes were used to model discrete bars 0.58 and 0.26 mm
mix. The distance (L) between supports was set at 180 mm (with in diameter. These wires represent prototype scale bars of 23.2
the beam overhanging the supports at each end by 10 mm) and 10.4 mm at a scale of 1∶40. The larger diameter wire (grade
and symmetrical vertical shearing loads were applied at 60 and 316) was used to model longitudinal reinforcement within beam
120 mm from one of the supports (i.e., a shear span of a ¼ 60 mm). sections, whereas the smaller diameter wire (grade 304) was used
The modulus of rupture was then found using as shear reinforcement in some of the tests. A woven wire mesh was
used in slab models to simulate the welded mesh commonly used in
2V ult L slab construction. This mesh had wires with a diameter of 0.25 mm
fr ¼ ð3Þ (prototype size of 10 mm reduced by 1∶40) a center-center spacing
bh2
of 1.4 mm, and was isotropic. Tensile tests of all of the wires used
where V ult = transverse load at which rupture occurs. were performed using an Instron 1196 loading frame, the results of
The parameters f c;150 and f r for the model concrete mixes are which are shown in Fig. 4. The stress strain response showed sig-
shown in Fig. 3, where the error bars represent the maximum and nificant strain hardening, making it most closely representative
minimum strengths recorded for each mix during the tests. Also of high yield reinforcement. Yield strengths (f y ) were defined at
shown in Fig. 3 are data for normal strength concretes presented 0.2% permanent strain, giving f y ¼ 460 MPa for the 0.58 mm
by Illston and Domone (2001), using conventional modulus of rup- diameter wire, f y ¼ 380 MPa for the 0.26 mm diameter wire, and
ture specimens 150 × 150 mm in section. Fig. 3 and Table 2 show f y ¼ 380 MPa for the wire in the mesh. Although some models
that a range of concrete compressive strengths (27 MPa < f c;150 < were constructed using the smooth wire alone, most used rough-
80 MPa) can be modeled, appropriate to normal strength concretes ened wire or mesh in which Congleton HST 95 silica sand (iden-
used in many structural and geotechnical applications. The modu- tical grading to that shown in Fig. 2) was used to coat the steel using
lus of rupture for each model mix is lower than that expected for the a fast-drying epoxy resin (Araldite Rapid, Bostik Findley Limited,
full-scale prototype concrete. Considering the size effect on f r Stafford, UK). This coating was applied to give a good mechanical

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / NOVEMBER 2011 / 1265

J. Struct. Eng., 2011, 137(11): 1263-1271


Table 3. Testing Schedule of Reinforced Structural Elements and Collapse
Loads at Model Scale
Shear Collapse load, Failure
Beam ID As =bd (%) reinforcement 2V (N)a mode
0.34_R_NL_1b 0.34 None 61.7 Shear
0.34_R_NL_2 0.34 None 94.9 Shear
0.51_S_NL_1c 0.51 None 39.2 Shear
0.51_R_NL_1 0.51 None 98.1 Shear
0.51_R_NL_2 0.51 None 115.1 Shear
0.51_R_NL_3 0.51 None 106.6 Shear
Downloaded from ascelibrary.org by Indian Institute of Technology Bombay on 02/01/22. Copyright ASCE. For personal use only; all rights reserved.

0.68_R_NL_1 0.68 None 116.4 Shear


0.68_R_NL_2 0.68 None 148.3 Shear
0.85_R_NL_1 0.85 None 138.5 Shear
0.85_R_NL_2 0.85 None 145.8 Shear
0.34_R_SL_1 0.34 Smooth 81.2 Bending
0.34_R_SL_2 0.34 Smooth 83.2 Bending
Fig. 4. Stress-strain curves for stainless steel wire used to model 0.51_R_SL_1 0.51 Smooth 104.2 Bending
reinforcement 0.51_R_SL_2 0.51 Smooth 96.8 Bending
0.68_R_SL_1 0.68 Smooth 136.0 Bending
0.68_R_SL_2 0.68 Smooth 131.1 Bending
bond between the steel and the model concrete to model the inter-
0.85_R_SL_1 0.85 Smooth 123.8 Shear
locking effect of deformed bars.
0.85_R_SL_2 0.85 Smooth 123.8 Shear
0.51_R_RL_2 0.51 Rough 115.3 Bending
Model Tests of Structural Elements 0.68_R_RL_1 0.68 Rough 143.4 Bending
0.68_R_RL_2 0.68 Rough 143.5 Bending
Following characterization of the model materials, a series of 23 0.85_R_RL_1 0.85 Rough 182.7 Bending
underreinforced model beams and four model slabs were cast using 0.85_R_RL_2 0.85 Rough 172.8 Bending
the Mix 1 model concrete to determine their bending properties. Slab 1 0:44=0:44d None 702 Shear
The beams were designed to be 1∶40 scale models of a square sec- Slab 2 0:44=0:44d None 706 Shear
tion prototype beam with dimensions of 0:5 × 0:5 m and 8 m long.
Slab 3 0:44=0:44d None 764 Shear
The slabs were 1∶40 scale models of prototype sections 4:8 × 4:8 m
Slab 4 0:44=0:44d None 634 Shear
in plan area and 0.4 m deep. Details of the tested elements are given
a
in Table 3, and the elements are shown diagrammatically in Fig. 5. Collapse loads at model scale, across full width of element.
b
The first set of 10 beams contained longitudinal reinforcement at “R” in Beam ID denotes roughened longitudinal reinforcement.
c
“S” in Beam ID denotes smooth longitudinal reinforcement.
different reinforcement ratios (defined as As =bd, where As is the d
Values in two perpendicular directions (isotropically reinforced), all rein-
total area of longitudinal steel, b is the beam breadth, and d the forcement roughened.
effective depth to the reinforcement), but no shear reinforcement.
The second group of 13 beams had shear reinforcement consisting
shear capacities of the model beams at prototype scale are com-
of a continuous coil of the smaller diameter wire along the length of
pared to those of equivalent full-scale beams (predicted using codal
the beam with loops spaced at 9 mm centers (prototype spacing of
provisions and the method described in the following) in Fig. 6.
360 mm reduced by 1∶40). Beams with both smooth and roughened
The collapse loads, 2 V, in Table 3 are the total loads applied to
shear reinforcement were tested as outlined in Table 3. The beams
the sections at failure. The ultimate applied shear load across the
were tested under load control using the same four-point bending
shear span at failure (V u ¼ V) in the model tests was divided by the
apparatus that was used to determine the modulus of rupture, giving
effective area of the section (¼ bd) to give the beam shear capacity
a shear span ratio a=d ¼ 5:3. The slabs tested were simply sup-
(vu ). This is typically normalized by the square root of f 0c . The
ported at two opposite ends with a line load applied across the
equivalent full-scale behavior is represented using the relationship
entire width of the slab, midway between the supports (i.e., a three-
presented by Bažant and Yu (2005), which fits a large database of
point bending test) and were similarly cast using Mix 1 model con- test data on beams of various sizes:
crete. The span between supports was 100 mm in the slab tests with
a=d ¼ 6:25. Loading was applied under displacement control using  3=8  sffiffiffiffiffiffiffiffiffiffiffi

A d f 0c
the Instron 1196 load frame described previously. None of the slabs vu ¼ vc ¼ 1:104 s 1þ ð4Þ
contained shear reinforcement. bd a 1 þ dd0

where vu and f 0c are in MPa; a, d, and d 0 are in mm; and:


Results and Discussion
d 0 ¼ 3063ðf 0c Þ2=3 ð5Þ
Beams and Slabs with No Shear Reinforcement
The equivalent full-scale beam is assumed to be made of con-
At the shear span ratio used in the beam tests, a beam without shear crete with the same compressive strength and geometry as the
reinforcement is expected to fail in flexural shear (Kong and Evans model beams at prototype scale; i.e., f 0c ¼ 17:4 MPa [Eq. (2)] with
1987). This mode of failure was observed for all of the model b ¼ h ¼ 500 mm, d ¼ 450 mm, a=d ¼ 5:3, and f y ¼ 460 MPa.
beams and slabs tested without shear reinforcement. The measured Also shown in Fig. 6 are the predicted capacities of the equivalent

1266 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / NOVEMBER 2011

J. Struct. Eng., 2011, 137(11): 1263-1271


Downloaded from ascelibrary.org by Indian Institute of Technology Bombay on 02/01/22. Copyright ASCE. For personal use only; all rights reserved.

Fig. 5. Model reinforced concrete beam and slab elements tested (all dimensions at model scale)

As =bd, as observed in conventional (full-scale) concrete beams


(Rajagopalan and Ferguson 1968; Bažant and Yu 2005). The model
test results are lower than the equivalent full-scale values, because
the model concrete has a lower modulus of rupture than would be
expected for full-scale concrete of the same f 0c (Fig. 3).
The single beam with smooth longitudinal reinforcement at
As =bd ¼ 0:51% is included in Fig. 6 to investigate the effect of
roughening of the wire on the transfer of stress into the steel
reinforcement. Roughening of the steel has a dramatic effect, with
the smooth wire showing a capacity approximately three times
smaller than the beams with roughened reinforcement, owing to
poor stress transfer between the model concrete and the reinforce-
ment leading to pullout.

Beams with Shear Reinforcement


With the addition of shear reinforcement, the shear capacity of the
model beams was increased and flexural (bending) failure was ob-
served to become the critical failure mode for these beams. The
moment capacities and bending stiffness of the model beams at
Fig. 6. Capacity of elements in flexural shear and comparison to codal prototype scale were compared to those of equivalent full-scale
provisions beams by employing standard cracked section theory (Kong and
Evans 1987) using the stress block shown in Fig. 7 and assuming
perfect bond between the concrete and reinforcement. Using this
full-scale beam using the provisions of ACI 318 (2002) and Euro- stress block, the moment capacity is
code 2 (2004), representing current American and European prac-
tices. The ACI method does not account for size effects, unlike the  
x
other two models. The model beams and slabs show reasonable M u ¼ k 1 k 3 bdxf c;150 1  k 2 ð6Þ
prototype behavior, including the correct increase in vu =f 0c with d

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / NOVEMBER 2011 / 1267

J. Struct. Eng., 2011, 137(11): 1263-1271


Downloaded from ascelibrary.org by Indian Institute of Technology Bombay on 02/01/22. Copyright ASCE. For personal use only; all rights reserved.

Fig. 8. Comparison of moment-curvature relationship of model beams


to equivalent full-scale beams (values at prototype scale of 1∶40)

Fig. 7. Ultimate moment capacity of model beams and comparison


with equivalent full-scale capacity (values at prototype scale of 1∶40)

where  
x fy As
¼ ð7Þ
d k 1 k 3 f c;150 bd

For the work presented in this paper, the following parameters


were assumed: k 1 ¼ 0:85, k 2 ¼ 0:5, k3 ¼ 1. Fig. 7 also shows the
variation of M u with As =bd for the beams that failed in Fig. 9. Elastic bending stiffness of beam tests and comparison to
bending. The full-scale flexural capacities are replicated well. equivalent full-scale values (values at prototype scale of 1∶40)
The bending stiffness of the beam before failure for the cracked
section is given by These values are in reasonable agreement with the predicted values
 3  from Eq. (8) as shown in Figure 9, although there is significant
bx
EI ¼ E c þ ξAs ðd  xÞ 2
ð8Þ scatter in the data, and the match becomes worse as As =bd in-
3
creases. This may be attributable to poor bond between the model
where Ec = Young’s modulus of the concrete; and ξ = modular ratio reinforcing wires as the spacing becomes closer or interaction
(¼ Es =Ec , where E s is the Young’s modulus of the steel). The Ec between adjacent wires as the spacing is reduced. If the Young’s
for the equivalent full-scale beam was estimated as 25 GPa for con- modulus of the model concrete is substantially lower than the pre-
ventional concrete with f c;150 ¼ 28 after Kong and Evans (1987). dicted value for a concrete at this f c;150 , this may also account for
Fig. 8 shows the moment-curvature test data (where curvature is the lower EI. Nonetheless, the model reinforced concrete is clearly
represented by K) for the least and most heavily reinforced model able to replicate the bending stiffness of reinforced concrete struc-
beams that match the full-scale behavior well at low As =bd, becom- tural elements to better than the correct order of magnitude.
ing increasingly flexible as As =bd is increased. There is no initial
Ductility
region of high bending stiffness representing uncracked behavior,
so the model sections are not appropriate for modeling this. How- Following the tests detailed in Table 3, some additional tests were
ever, for geotechnical applications in which the strength of the undertaken on the beams under displacement control at higher a=d
models is likely to be of interest, the cracked value of EI is expected to examine the displacement ductility of the model sections. These
to be a more appropriate measure of structural stiffness for repli- additional tests were combined with the slab test results which were
cating displacements close to failure. also conducted under displacement control. Two beams were tested
Values of EI for the model beams were determined by a linear fit at the extremes of the range of As =bd shown in Table 3. The dis-
to the measured moment-curvature behavior up to M ≈ M u =3. placement ductility ratio (μΔ ) was defined as

1268 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / NOVEMBER 2011

J. Struct. Eng., 2011, 137(11): 1263-1271


Downloaded from ascelibrary.org by Indian Institute of Technology Bombay on 02/01/22. Copyright ASCE. For personal use only; all rights reserved.

Fig. 10. Comparison of observed and predicted displacement ductility


(scale independent)

Δu
μΔ ¼ ð9Þ
Δy
where Δy = yield displacement; and Δu = ultimate displacement
at which the load has dropped to 85% of the ultimate value. These
values were determined following the method outlined by Pam
et al. (2001). Equivalent full-scale values were determined using
the theoretical relationship for curvature ductility (K u =K y ) pre-
sented by Park and Paulay (1975) and recognizing that in a
four-point bending test K u =K y ≈ Δu =Δy , as demonstrated both
analytically and experimentally for singly reinforced concrete
beams by Maghsoudi and Sharifi (2009):
   2  
Δu K u 0:85k3 Es f 0c εc As 2 As As 0:5
≈ ¼ 1 þ ξ  ξ þ 2ξ
Δy K y f 2y ðbd
As
Þ bd bd bd
ð10Þ
These calculations were conducted for f 0c ¼ 17:4 MPa as used
in the model tests, and it was assumed that the concrete strain εc ¼
0:003 at failure. Fig. 10 compares these calculated values with the Fig. 11. Use of model sections to represent prototypes of different
model test data. To demonstrate that Eq. (10) is appropriate for sizes by varying scaling factor and g-level within a geotechnical
describing the ductility of reinforced concrete elements in bending, centrifuge: shear capacity
the theoretical curves are also compared with data from tests of
full-scale reinforced concrete beams with f 0c ¼ 27 MPa reported
by Pam et al. (2001).
aggregate. A good match can be observed over the full range of N.
For this particular parameter, which was underestimated by the
Extension to Other Scaling Factors model beams at 1∶40 scale (Fig. 6), the match improves at higher
scaling factors (i.e., as the models are used to represent a larger
Because the model elements in this study provide an approximately equivalent full-scale prototype beam), because the strength of
1∶1 scaling of material properties, their similitude to prototype sec- the full-scale prototype beam that is being modeled is reduced
tions should extend to any scaling factor, and not just 1∶40, as con- by a size effect [incorporated in Eq. (4)]. The figure shows that
sidered previously. As an example, Fig. 11 shows the prototype the 12.5 mm deep model beam section may be used to model
scale shear capacity for the model beam and slab sections by full-scale prototype beams between 125–1,250 mm deep by vary-
scaling the measured V u at model scale by N 2 (Table 1) for vari- ing N; i.e., the g-level in the centrifuge and the scale factor for the
ous scaling factors between 1∶10 and 1∶100, representing the rest of the model. This range of prototype sizes covers a range of
range commonly employed in geotechnical centrifuge testing. Also geotechnical elements subjected to bending loads including piles
shown in the figure are predicted values for equivalent full-scale and ground beams (beams) and retaining walls and tunnel linings
concrete beams at the different prototype sizes (i.e., the full-scale (slabs).
beams that are being modeled), calculated using Eq. (4), and as- Because only model sections with depths between 10 and
suming that all of these full-scale beams (from 125–1,250 mm 12.5 mm have had their performance validated in this paper,
in depth) would be able to use the same concrete with nonscaled centrifuge models should be designed so that the depth of the model

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / NOVEMBER 2011 / 1269

J. Struct. Eng., 2011, 137(11): 1263-1271


reinforced concrete elements in the plane of bending will be within assistance in manufacturing the testing equipment and formwork,
this size range at model scale. For a given soil-structure interaction and to Mark Truswell for his assistance with the testing programme.
problem, the desired depth of the full-scale prototype concrete The authors would also like to acknowledge the support of the
element should be used to determine the scaling factor and g-level Nuffield Foundation, UK for funding the second author’s summer
required to give a model sectional depth between 10–12.5 mm, with internship at the University of Dundee.
the rest of the centrifuge model (soil and other nonconcrete struc-
tural elements) then being modeled to this scale. If prototype con-
crete elements of small sectional depth are to be modeled, low Notation
scaling factors will be required and this will limit the depth of
the model, which may be important for problems involving deep The following symbols are used in this paper:
piling or deep excavations. Clearly, compromises will also need As = Cross-sectional area of main reinforcing steel;
Downloaded from ascelibrary.org by Indian Institute of Technology Bombay on 02/01/22. Copyright ASCE. For personal use only; all rights reserved.

to be made when modeling a system with multiple prototype b = Beam breadth;


concrete elements of different depths, because this would require c = Cover depth;
models of different depths. There is some small allowance for this d = Effective depth;
in the range of sizes described in this paper, but it is clear that E c = Young’s modulus (concrete);
further work is required to validate models that are smaller and/or EI = Bending stiffness;
larger than the 10–12.5 mm range presented herein to extend the Es = Young’s modulus (main reinforcing steel);
complexity of problems that may be modeled using the technique. f c;150 = Cube strength of concrete (150 mm × 150 mm cube);
Nonetheless, many important soil-structure interaction problems f 0c = Cylinder strength of concrete;
may be studied that do not suffer from this limitation (e.g., a group f y = Yield stress (reinforcement);
of identical piles connected by a rigid pile cap under lateral loading) h = Total depth of beam (¼ d þ c);
and it is expected that the new materials and modeling techniques K = Curvature;
described in this paper will be widely applicable to centrifuge k1;2;3 = Stress block coefficients (Fig. 7);
modeling of soil-structure interaction problems. L = Beam length;
M ðuÞ = Bending moment (ultimate);
N = Scale factor;
Conclusions V ðuÞ = Shear load (ultimate);
vu = Ultimate shear stress of concrete and main longitudinal
This paper has described a new technique for modeling reinforced steel;
concrete elements under bending loads at very high scale factors. x = Neutral axis depth;
This has direct applications in modeling soil-structure interaction Δu = Ultimate displacement;
problems in a geotechnical centrifuge such that the constitutive Δy = Yield displacement;
behavior of both the soil and the structural element (e.g., a pile or μΔ = Displacement ductility ratio;
retaining wall) will be correct at prototype scale. Gypsum-based ξ = Modular ratio (¼ Es =E c ); and
plaster is used as the cementitious binder and fine uniform labora- ρ = Reinforcement ratio (¼ As =bd).
tory sand represents a scaled coarse aggregate at scaling factor of
∼1∶40. This latter feature approximately scales the characteristic
length of the brittle fracture process, such that the modulus of rup-
ture in the reduced-scale models does not exhibit substantial over- References
strength. Four different model concrete mixes were developed and
characterized to model a range of normal strength concretes. Four- Abdoun, T., Dobry, R., O’Rourke, T. D., and Goh, S. H. (2003). “Pile
and three-point bending tests were subsequently conducted on response to lateral spreads: Centrifuge modelling.” J. Geotech.
singly underreinforced model beams and isotropically reinforced Geoenviron. Eng., 129(10), 869–878.
square model slabs with varying amounts of longitudinal reinforce- American Concrete Institute (ACI). (2002). “Building code requirements
ment, both with and without shear reinforcement. Those without for structural concrete and commentary.” ACI 318-02, 318R-02,
Farmington Hills, MI.
shear reinforcement showed classic flexural shear failure, although
Bažant, Z. P. (2005). Scaling of structural strength, Elsevier Butterworth-
ultimate shear capacities at prototype scale were underpredicted. Heinemann, Oxford, UK.
The addition of shear reinforcement led to flexural failure, with Bažant, Z. P., and Kazemi, M. T. (1991). “Size effect on diagonal failure of
moment capacities at prototype scale representing equivalent full- beams without stirrups.” ACI Struct. J., 88(3), 268–276.
scale sections well. Ductility was also found to be well replicated, Bažant, Z. P., and Li, Z. (1995). “Modulus of rupture: Size effect due to
although bending stiffness was reproduced less well at higher fracture initiation in boundary layer.” J. Struct. Eng., 121(4), 739–746.
reinforcement ratios, with greater scatter in the results. However, Bažant, Z. P., and Yu, Q. (2005). “Designing against size effect on shear
the results show that the technique allows for stiffness, strength, strength of reinforced concrete beams without stirrups: II. Verification
and ductility of structural elements under bending loads to be and calibration.” J. Struct. Eng., 131(12), 1886–1897.
simultaneously scaled and failure modes to be accurately repro- Belgin, C. M., and Sener, S. (2008). “Size effect on failure of overrein-
forced concrete beams.” Eng. Fract. Mech., 75, 2308–2319.
duced, which is not possible with current modeling techniques used
British Standards Institution (BSI). (1992). “Specification for aggregates
by centrifuge modelers. The technique will be of widespread ap- from natural sources for concrete.” BS 882, London.
plication in geotechnical modeling of soil-structure interaction Comite Europeen de Normalisation (CEN). (2004). “Eurocode 2: Design of
problems. concrete structures. Part 1-1: General rules and rules for buildings.” EN
1992-1-1:2004, Brussels, Belgium.
Garnier, J., et al. (2007). “Catalogue of scaling laws and similitude
Acknowledgments questions in geotechnical centrifuge modelling.” Int. J. Phys. Model.
Geotech., 7(3), 1–24.
The authors would like to sincerely thank Ernie Kuperus and Harris, H. G., and Sabnis, G. M. (1999). Structural modeling and
the technical staff of the Division of Civil Engineering for their experimental techniques, 2nd Ed., CRC Press, Boca Raton, FL.

1270 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / NOVEMBER 2011

J. Struct. Eng., 2011, 137(11): 1263-1271


Hayward, T., Lees, A., Powrie, W., Richards, D. J., and Smethurst, J. 1191–1204.
(2000). “Centrifuge modelling of a cutting slope stabilised by discrete Maghsoudi, A. A., and Sharifi, Y. (2009). “Ductility of high strength
piles.” Rep. 471, Transport Research Laboratory, Berkshire, UK. concrete heavily steel reinforced members.” Sci. Iran., Trans. A, 16(4),
Illston, J. M., and Domone, P. L. J. (2001). Construction materials: Their 297–307.
nature and behaviour, 3rd Ed., Taylor and Francis, London. Mansur, M. A., and Islam, M. M. (2002). “Interpretation of concrete
Ito, K., Ohno, S., and Matsuda, T. (2006). “Seismic response of under- strength for non-standard specimens.” J. Mater. Civ. Eng., 14(2),
ground reinforced concrete structure—Centrifuge model test and its 151–155.
analysis.” J. Earthquake Eng., 23(1), 117–124. Pam, H. J., Kwan, A. K. H., and Islam, M. S. (2001). “Flexural strength and
Knappett, J. A., and Madabhushi, S. P. G. (2009). “Influence of axial load ductility of reinforced normal- and high strength concrete beams.” Proc.
on lateral pile response in liquefiable soils. Part I: Physical modeling.” ICE Struct. Build., 146(4), 381–389.
Géotechnique, 59(7), 571–582. Park, R., and Paulay, T. (1975). Reinforced Concrete Structures,
Kong, F. K., and Evans, R. H. (1987). Reinforced and prestressed concrete, Wiley, New York.
Rajagopalan, K. S., and Ferguson, P. M. (1968). “Exploratory shear tests
Downloaded from ascelibrary.org by Indian Institute of Technology Bombay on 02/01/22. Copyright ASCE. For personal use only; all rights reserved.

3rd Ed., E&FN Spon, London.


Litle, W. A., and Paparoni, M. (1966). “Size effect in small scale emphasizing percentage of longitudinal steel.” J. Am. Concr. Inst.,
models of reinforced concrete beams.” J. Am. Concr. Inst., 63(11), 65(8), 634–638.

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / NOVEMBER 2011 / 1271

J. Struct. Eng., 2011, 137(11): 1263-1271

You might also like