You are on page 1of 6

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/326540800

Centrifuge and numerical investigations of rotated box


structures

Conference Paper · July 2018

CITATIONS READS

0 184

3 authors:

Tim Newson Osama Abuhajar


The University of Western Ontario The University of Western Ontario
111 PUBLICATIONS   1,870 CITATIONS    34 PUBLICATIONS   197 CITATIONS   

SEE PROFILE SEE PROFILE

K.J. Stone
University of Brighton
41 PUBLICATIONS   547 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Static and seismic soil culvert interaction View project

Engineering behaviour of chalk View project

All content following this page was uploaded by Osama Abuhajar on 21 July 2018.

The user has requested enhancement of the downloaded file.


Physical Modelling in Geotechnics – McNamara et al. (Eds)
© 2018 Taylor & Francis Group, London, ISBN 978-1-138-34419-8

Centrifuge and numerical investigations of rotated box structures

T.A. Newson & O.S. Abuhajar


Department of Civil and Environmental Engineering, Western University, Canada

K.J.L. Stone
School of Environment and Technology, University of Brighton, UK

ABSTRACT: Soil arching can occur due to the existence of buried structures inside a soil body. This arching
can lead to an increase or decrease in soil pressures attracted to the buried structures, which cannot be accurately
estimated by simply considering the self-weight forces generated by the prism of soil supported by the structure.
The relative stiffness between the structure and the surrounding soil is the main factor that controls these soil
pressures. This paper presents centrifuge and numerical results of two rotated box structures used to investigate
the effect of different wall thicknesses on the contribution to the overall loads attracted to the structure. Results of
comparative numerical analyses using PLAXIS are presented to aid the interpretation of the tests. Load reductions
were found to occur for the most flexible portions of the structures (up to 20%) and both the individual flexibility
of the members and the overall structure were found to be important.

1 INTRODUCTION AASHTO (2002) standard specifications for highway


bridges takes some account of arching by changing
The loads attracted to buried structures, from both vertical stresses over box culverts based on Marston-
overburden and surcharge loads, are governed by the Spangler theory. However, this approach is quite
characteristics of the soil and geometry and stiffness conservative compared to more sophisticated numer-
of the structural components. In many instances, the ical techniques, such as finite element analysis. For
redistribution of the free-field stresses as the result the case of relatively flexible buried structures, the
of the presence of a buried structure will lead to a soil-structure interaction is even more complicated,
decrease in loading over the deflecting or yielding and the problem is difficult to solve theoretically or
areas of the structure and an increase over adjoining analytically.
rigid or stationary parts. This transfer of load due to Previous work by Abuhajar et al. (2009) investi-
soil-structure interaction is known as ‘arching’. gated the differences between the loads attracted to
Whilst there have been numerous experimental and buried flexible box structures orientated in different
field studies to investigate stress distribution and arch- directions (i.e. with the sides parallel or at 45 degrees
ing (e.g. Lefebvre et al., 1976), the exact conditions to the free surface). This study found that for the same
required for this phenomenon to occur are still unclear structure the induced bending moments were higher,
and arching is often ignored in engineering design deflections smaller and attracted loads smaller for the
due to a lack of experience, and inclusion in codes rotated case. The principal aim of the current paper is
of practice is rare. A range of problems such as under- to further investigate this problem using scaled mod-
ground conduits, tunnels, trapdoors, retaining walls elling in the centrifuge and numerical analysis for
and braced cuts can all experience significant arching two 45◦ rotated square box structures. These struc-
action and theoretical analyses have been published tures have different combinations of wall and slab
on these subjects. These approaches have considered thicknesses. The results are presented in the form of
soil arching from both elastic and plastic soil states. bending moments, deflections and vertical soil pres-
In recent years, research has concentrated on scaled sures, to enable a comparison of the contribution made
physical modelling of the behaviour of buried cir- by the structural elements to the loads attracted to the
cular and square structures (e.g. Iglesia et al. 1999; structure, related to the theoretical overburden loads.
Stone & Newson 2002; McGuigan 2010; Abuhajar
et al. 2015a/b).
2 EXPERIMENTAL TESTING
Buried box section culverts and conduits are com-
monly used around transportation infrastructure, e.g.
2.1 Model preparation
to span highways. These are used to control water
flow, storm runoff, divert municipal services, allow In general, underground box structures are constructed
vehicular access and for other related activities. from short sections of reinforced concrete, which are

221
Figure 1. Model prior to embedment in sand.

joined together to form the desired cross-sections. Due


to the problems of manufacturing concrete with micro-
aggregates for scaled physical tests, model structures
(shown in Figure 1) were made from an aluminum
box section 400 mm long with an external dimension
of 101.6 mm and wall thicknesses of 2.0 and 6.35 mm.
However, correct modeling of the structural deflec- Figure 2. Diagram showing the orientation of the two
tion can be achieved in reduced scale models by structures and the properties used for analysis.
ensuring the following relationship is maintained:

where, E =Young’s modulus of the material, I = sec-


ond moment of area per unit length of the material and
N = scaling factor. The subscripts ‘m’ and ‘p’ refer to
model and prototype respectively.
To enhance the response of the model, the walls of
the aluminum section were machined to a thickness
of only 2 mm. Miniature strain gauges were bonded
at various locations on the structure to measure the
internal and external strains. The outputs were used to
determine the bending and axial strains at the location
of each gauge pair.

Figure 3. Comparison of bending moment diagrams at 80 g


2.2 Centrifuge tests (prototype scale).
The centrifuge test procedure was as follows: firstly,
a bed of sand was placed by air pluviation into a
rectangular strongbox of dimensions 500 mm wide × angle of internal friction of the sand at critical state
800 mm long × 600 mm high. The structure was is φ  = 31◦ . In this paper, tests with two different box
placed by hand into the sand to the required depth structures are presented: one with a uniform thickness
and then the pluviation was completed. The strong- of 2 mm (Test A) and the other with pairs of opposite
box was then mounted on the centrifuge platform and sides with different thicknesses – top/bottom = 2 mm
the structure placed on this layer. The strain gauges [thin] and sides = 6.35 mm [thick] (Test B). The boxes
were connected to the data acquisition system, and were orientated with their vertices rotated 45◦ into the
after checking their initial gauge output, the data logger vertical plane (as shown in Figure 2 below). The soil
was zeroed. The centrifuge was then accelerated and bed under the structure is 220 mm and the soil above is
held at the following acceleration levels, 5 g, 10 g, 20 g, 157 mm, hence the normalised embedment (H/B) was
40 g, 60 g and 80 g. Data from the strain gauges was equal to 1.14 for the modelled cases.
recorded continuously during the test. Congleton Sand Figure 3 shows a comparison between the bending
was used for all of the tests. This is a uniform sand with moment distributions for tests A and B for members
a d10 = 100 µm and maximum and minimum densities ‘1-2’and ‘2-3’(shown as triangular and circular points)
of 1.78 and 1.51 g/cm3 respectively. Air pluviation cre- at 80 g. For member 2-3, the bending moment values
ated soil beds with a relative density (Dr ) of 35%. The are very close because of the similar wall thicknesses.

222
In contrast, for member ‘1-2’, there is a clear differ-
ence in the values of the bending moments due the
different thicknesses of the two structures. It is also
seen from Figure 3 that the bending moment values at
the upper vertex (point 2) are less than those at the mid-
side vertices (points 1 and 3). For both members, the
values of the bending moments are highest at the ver-
tices and vary non-linearly along the members. Whilst
these variations are approximately parabolic, they are
asymmetric with the minima not coinciding with the
centerline of the members.

3 NUMERICAL MODELLING Figure 4. Geometry and finite element mesh for the model.

The numerical analysis results presented herein are adopts the shape of a compressed hypo-trochoidal
for the 80 g test only. These tests were designed to square, with the upper and lower vertices (points 1
investigate the interaction between the buried model and 4) moving inwards and the side vertices (points
and the surrounding soil related to the stiffness 2/3) moving outwards. The relative stiffness of the
of the soil and deformation of the model. Two- structure in the horizontal and vertical directions is
dimensional plane strain finite element analysis was approximately the same. This is similar to the behav-
conducted using the package PLAXIS® to aid inter- ior of circular culverts (e.g. Taleb & Moore 1999)
pretation of the physical model. Drained soil condi- with the crown settling and the shoulders moving out-
tions were assumed, and the modeling was carried wards. The displacements vary non-linearly along the
out using Mohr-Coulomb elastic-perfectly plastic soil. members of the structure and in common with the
The assumed material parameters were Young’s mod- bending moments, the maxima do not coincide with
ulus, E = 80 MPa, Poisson’s ratio, ν  = 0.32, effec- the centerline of the members.
tive unit weight, γ  = 16 kN/m3 (note the sand was In contrast, the deflections of the members in Test
dry), angle of internal friction, φ  = 31◦ , cohesion B [Figure 5(b)] are relatively asymmetrical. The gen-
angle, c = 1.0 kPa and angle of dilation, ψ  = 0◦ . The eral pattern of behavior is the same with the upper and
structure material (assumed to be linear elastic) was lower vertices moving inwards and the side vertices
assigned E = 70 GPa for aluminum and ν = 0.20 and moving outwards. Again, the displacements vary non-
the test was modelled as a prototype at 80 g. linearly along the members of the structure. For the
Due to the asymmetry of Test B, the whole soil- thin (2 mm) members the maxima do not coincide with
structure system was modeled. The domain was dis- the centerline of the members. For the thick (6.35 mm)
cretized using 15-noded triangular soil elements (with members, the peak displacement occurs at the vertices.
fourth order interpolation for displacements) and 5- Interestingly the displacements at the four vertices
noded beam culvert elements. Each node has three suggest that the structure is rotating marginally in the
degrees of freedom per node: two translational and anti-clockwise direction.
one rotational. Interface elements were used to pro- Figure 6 presents a comparison between the vertical
vide for possible slippage and separation between the soil pressures on members ‘1-2’ and ‘2-3’ of the two
culvert and the surrounding soil. The ‘roughness’ of models at 80 g. In common with the bending moment
the interaction is defined by a strength reduction fac- diagrams in Figure 3, the results show very similar
tor, R, which relates the interface strength to the soil vertical soil pressures on the thin members (2-3) and
strength parameters. A 33% reduction in soil strength greater variations for the thick/thin members (1-2).The
was assumed (R = 0.67) at the interface between the highest pressures are seen at the vertices of the struc-
culvert and the surrounding soil. ture and the lowest minima for the zones with the peak
The modelled geometry and a typical finite element deflections for the thinnest members (with about 20%
mesh used for the structures is shown in Figure 4. The of the pressure). For all of the members, the variation
boundary conditions used were to fix the domain sides in vertical pressures along the structure is non-linear.
in the horizontal direction (i.e. free ‘y’ movement), For the thick member (1-2) in Test B, the minima of
fix the bottom of the domain in both directions (x, y) the pressure is approximately 55% of the peak pres-
and the upper surface was free to move in both direc- sure. If the overall load attracted to the upper surface of
tions. Figure 4 shows a deformed mesh (with the vertex the two models is assessed, the Test A structure carries
numbering system used) and deflection of the model. 80% and Test B structure carries 96% of the weight of
The numerical predictions of the bending moments the overlying soil prism (from point 1 to 3). For Test
for tests A and B are also shown in Figure 3. These gen- B, 2/3 of this load is carried by the stiffer (thicker)
erally show a very good match between the centrifuge member [1-2]. Whilst the asymmetric structure (Test
model and numerical bending moments. B), conveys no advantage, the symmetrical flexible
The corresponding deflections for the Test A struc- structure (Test A) sheds 20% of the load to the stiffer
ture are shown in Figure 5(a). The loaded structure surrounding soil mass.

223
Figure 6. Comparison of vertical soil pressure diagrams for
both structures at 80 g.

Figure 5. Member deflections for (a) Test A and (b) Test B


structure at 80 g (prototype scale).

The incremental shear-strain contours for the two


models are shown in Figure 7. Test A is shown in
Figure 7(a) and indicates very little strain localisa-
tion along the middle 2/3 of the top of the structure.
However, two shear strain localisations occur close to
the side vertices (with peak magnitudes of 0.04%).
Interestingly these originate from the points of con-
traflexure (BM = 0), approximately 4/5 of the way
along the member and propagate towards the surface
of the soil mass. On the underside of the structure the
localisations are curved and span between the vertices
on these members. For Test B [Figure 7(b)], only one
localised shear strain zone was observed on the thin-
ner side of the structure with a slightly higher peak
strain of 0.12%. This has the same form as those shown
in Figure 7(a) and again originates from the point of Figure 7. Incremental shear strain for (a) Test A and
contraflexure on that member (2-3). (b) Test B.

224
Although not shown in this paper, the horizontal from the points of contraflexure on the more flexible
effective stresses were also found using Plaxis. These portions of the upper surface of the structures, sug-
indicate that significant portions of the box sides are gesting the initial formation of soil mechanisms. This
subjected to low horizontal pressures (less than the lead to some arching and load transfer to the rigid por-
active case in certain areas). The one exception is the tions of the soil mass and reductions in the overall
side vertex of the boxes, which have pressures that loads on the structure. The more rigid portions of the
begin to approach the passive pressure. structure also experienced much higher stresses and
For the symmetrical rotated box (Test A), the two bending moments. For the asymmetric case (Test B),
upper members behave like active trapdoors and the higher bending moments were found for the stiffer por-
lower members like propped walls. The structure tends tions of the structure and quite significant increases in
to reduce the total loads attracted to the top slabs, load were observed. Whilst it is unlikely this type of
whilst concentrating the loads at the vertices (corners). structure will be used in geotechnical practice, due to
This behaviour is further enhanced by the additional limitations of construction and usage, this study does
compression of the upper members acting together (in provide information on fundamental aspects of arching
a similar manner to a loaded ‘A-frame’ structure). The phenomena around buried structures.
shear strain diagram [Figure 7(a)] suggests that the
material contained within the two localisations moves
downwards with the structure and a discontinuity is REFERENCES
beginning to form with stationary soil outside of the
developing soil mechanism; this supports the observa- Abuhajar, O., El Naggar, H. & Newson, T. 2015a. Static soil
culvert interaction the effect of box culvert geometric con-
tions made regarding the loads attracted to the model
figuration and soil properties. Comput. Geotech., 69(6):
upper surfaces. 219–235.
For the asymmetrical case (Test B), the more flexi- Abuhajar, O., Newson, T., El Naggar, H. & Stone, K. 2009.
ble portions of the structure are behaving in the same Arching around buried square section structures 17th
manner as Test A, but the localisations appear to be ISSMGE Conference, Alexandria, Egypt. 1–6.
more intense and the developing soil mechanism much Abuhajar, O., Newson, T. & El Naggar, H. 2015b. Scaled
narrower, which affects the arching behavior and loads physical and numerical modelling of static soil pressures
attracted to the structure. on box culverts. Can. Geotech. J., 52(11): 1637–1648.
American Association of State Highway and Transporta-
tion Officials, Inc. (AASHTO) 2002. AASHTO Standard
Specifications for Highway Bridges, Washington, D.C.
4 CONCLUSIONS Iglesia, G.R., Einstein, H.H. & Whitman, R.V. 1999. Deter-
mination of Vertical Loading on Underground Structures
This paper has described centrifuge and numerical Based on an Arching Evolution Concept, Geotechnical
modeling used to investigate soil-structure interaction Special Publication, 90: 495–506.
of a buried rotated square box.The analyses have inves- Lefebvre, G., Laliberte, M., Lefebvre, L.M., Lafleur, J. &
tigated the contributions of the form and location of Fisher, C.L. 1976. Measurement of Soil Arching above a
the structural. The matches between the centrifuge and Large Diameter Flexible Culvert, Can. Geot. J., 13: 58–71.
numerical predictions were found to be reasonable and McGuigan, B. 2010. Earth pressures and loads on
induced trench culverts. PhD Thesis. University of New
this enabled further interpretation of the results to be
Brunswick.
conducted. The effect of member thicknesses showed Stone, K.J.L. & Newson, T.A. 2002. Arching Effects in Soil-
their important contributions to the overall structural Structure Interaction, 4th Int. Conf. Physical Modelling in
stiffness and the soil pressures attracted to the boxes. Geomechanics, St. Johns, Newfoundland, 935–939.
The shape of the soil pressure distributions was not Taleb, B. & Moore, I. 1999. Metal culvert response to
uniform and were found to be approximately parabolic earth loading: performance of two-dimensional analysis.
with high values at the edges and lower values at Transportation Research Record. 1656: 25–36.
the center. Shear localisations were found to originate

225

View publication stats

You might also like