improving the seismic resistance of adobe buildings yr projects 2007/jones... · section 1:...
TRANSCRIPT
I hereby declare that, except where specifically indicated, the work herein is my own original work.
Signed:
Date:
Improving the Seismic Resistance of Adobe Buildings
by
Stephen Jones (R)
Fourth-year undergraduate project in Group D, 2006/2007
i
Improving the Seismic Resistance of Adobe Buildings: Technical Abstract Stephen Jones (R), Fourth-year undergraduate project in Group D, 2006/2007
Adobe is a naturally-dried earthen brick used for building. Unfired earth construction such as
adobe is used by an estimated 30% of the global population, rising to 50% in developing countries
(Houben and Guillaud, 1994). However, adobe buildings are very vulnerable to earthquakes
because adobe is a low strength, brittle material, and so tends to suffer sudden catastrophic failure.
Further common problems include a lack of: maintenance; involvement from professional
engineers; or structural reinforcement (Tetley and Madabhushi, 2007, and Dowling, 2006).
Unfortunately, there is a strong overlap between the regions of the world where significant use of
earthen building occurs and zones of moderate or greater seismic hazard (De Sensi, 2003).
Previous research has aimed to address this problem by providing methods of reducing the seismic
vulnerability of adobe construction which are suitable for use in the low-income regions most at
risk. However, there has been insufficient large-scale take-up of these potential improvements,
because “most of the proposed systems are too complex and/or too costly to be widely used
without sustained external intervention” (Dowling, 2006).
This project firstly aimed to test scale models of unreinforced adobe walls on a 1-g shaking table
to simulate accurately the behaviour of real adobe buildings under earthquake loading by
modelling realistic collapse mechanisms. The second objective was to test scale models of adobe
walls with structural improvements to enhance seismic resistance, in order to assess the relative
effectiveness of these strengthening techniques. The improvements were designed to be simple
and cheap enough for potential widespread use in developing countries.
Models of single in-plane walls and ‘L’-shaped corner wall joints were tested at 1:5 scale and
exhibited typical real-life collapse mechanisms. A suitable material composition, supporting the
findings of Tetley and Madabhushi (2007), was 60% gravel of size 30mm - 80mm and 40% mortar
by weight. The mortar was made from 60% Kaolin clay at w = 0.45 and 40% building sand by
weight. It was necessary either to leave the completed model to dry for 30 or more days or to pre-
dry the sand in an oven to ensure the model had dried sufficiently before testing, to simulate the
drying process in real-life adobe construction.
Improving the Seismic Resistance of Adobe Buildings: Technical Abstract
ii
In the second part of the project, it was found that external vertical bamboo reinforcement, joined
through the walls of the building and tied externally with horizontal wire, can improve the seismic
resistance of the building by increasing the ductility of the walls and providing confinement for the
adobe blocks. In 1:5 scale tests on ‘L’-shaped corner joints, the capacity to resist severe structural
damage (collapse of the transverse wall) was increased by at least a factor of 2 using the method of
external bamboo reinforcement. The effect is of a similar magnitude to that observed in recent
research on a similar technique by Dowling (2006). Further research would be needed to assess if
this factor could be expected in real life if whole buildings were reinforced in this way.
A further test on a model with external vertical bamboo reinforcement with a smaller spacing
between the bamboo canes showed that seismic resistance due to confinement of the adobe is
dependent on the relative sizes of the adobe blocks and the spacing of the bamboo and wire. In real
life it is likely that there would be a compromise between the amount of confinement, the
materials available, and the number of holes which could be drilled through the walls safely.
External bamboo and wire has the potential to be used to retrofit existing houses provided holes
can safely be drilled in the walls to allow the attachment of bamboo. The technique can also be
used in new constructions. Both bamboo and galvanised wire are materials already used in the
internal reinforcement of adobe buildings and are generally low cost and widely available. Further
research is needed in the field to ensure that the technique has widespread suitability in terms of
public acceptance and ease of use. The author will work with the Salvadorean Foundation for
Reconstruction and Development in El Salvador to perform a structural survey of adobe houses in
order to assess the appropriateness of this method for strengthening existing buildings.
A third structural improvement test found that a combination of internal vertical bamboo
reinforcement and internal horizontal wire can improve the seismic resistance of the building by
increasing the ductility of the walls and increasing the connectivity at corner joints. There may be
differences in the effectiveness of this method depending on whether irregular blocks are used
which can fit around the bamboo (as in this project) or regular blocks which must be cut to fit the
bamboo into the wall and may produce weaknesses due to discontinuity problems in the adobe (as
tested by Dowling, 2006). Internal reinforcement is only suitable for newly-built constructions and
so has less widespread application.
Contents Section 1: Introduction .............................................................................................................. 1
1.1. Background ....................................................................................................................... 1 1.2. Project objectives .............................................................................................................. 2
Section 2: Literature Review..................................................................................................... 3 2.1. Seismic performance of adobe .......................................................................................... 3 2.2. Laboratory testing of model adobe walls and buildings ................................................... 3
2.2.1. Modelling adobe buildings and seismic strengthening methods................................ 4 2.2.2. Testing models using a shaking table to simulate an earthquake............................... 5
2.3. Existing guidelines on adobe construction........................................................................ 6 2.3.1. General layout and construction features................................................................... 6 2.3.2. Adobe block construction methods and materials ..................................................... 7 2.3.3. Wall reinforcement methods and materials................................................................ 7
Section 3: Experimental techniques ......................................................................................... 8 3.1. Shaking table and input motion......................................................................................... 8 3.2. Visual data recording ........................................................................................................ 8 3.3. Instrumentation and data recording................................................................................... 9
Section 4: Modelling adobe walls and seismic strengthening methods ............................... 10 4.1. Key issues and summary of model wall designs chosen................................................. 10 4.2. Size of the model walls ................................................................................................... 12 4.3. Material composition and construction of the model walls ............................................ 13 4.4. Method used to model a continuous transverse wall and boundary conditions .............. 16 4.5. Techniques used to model structural improvements ....................................................... 17
4.5.1. External vertical bamboo tied with external horizontal wire ................................... 17 4.5.2. Internal vertical bamboo tied with internal horizontal wire..................................... 18
Section 5: Behaviour of single walls........................................................................................ 20 5.1. Visual observations ......................................................................................................... 20 5.2. Summary of single wall results ....................................................................................... 24
Section 6: Behaviour of ‘L’-shaped corner wall joints ......................................................... 25 6.1. Visual observations ......................................................................................................... 25 6.2. Instrumentation data recorded......................................................................................... 29
6.2.1. Observations on accelerometer data......................................................................... 32 6.3. Summary of results for ‘L’-shaped corner wall joints .................................................... 33
Section 7: Behaviour of ‘L’-shaped corner wall joints with improvements ....................... 34 7.1. Visual observations ......................................................................................................... 34 7.2. Moisture content data recorded ....................................................................................... 39 7.3. Instrumentation data recorded......................................................................................... 39
7.3.1. Observations on accelerometer data......................................................................... 41 7.4. FFT analysis of acceleration data.................................................................................... 41
7.4.1. Observations on FFT analysis.................................................................................. 43 7.5. Summary of results for ‘L’-shaped corner wall joints with improvements .................... 45
Section 8: Conclusions ............................................................................................................. 46 8.1. Conclusions related to simulating adobe buildings using scale models.......................... 46 8.2. Conclusions related to structural improvements of adobe buildings .............................. 47 8.3. Recommendations for future research............................................................................. 48
Section 9: References ............................................................................................................... 49
1
Section 1: Introduction
1.1. Background
Adobe is a naturally-dried earthen brick used for building, usually with mortar made from the
same material as the bricks. Unfired earth construction such as adobe is used by an estimated
30% of the global population, rising to 50% in developing countries (Houben and Guillaud,
1994). Adobe’s widespread use is due to a number of advantages which include: low cost; ease
of use; wide availability; durability; high thermal capacity; and energy efficiency.
However, adobe buildings are very vulnerable to earthquakes, predominantly because adobe is
a low strength, brittle material, and so tends to suffer sudden catastrophic failure. Further
common problems include a lack of maintenance and an absence of involvement from
professional engineers (Tetley and Madabhushi, 2007), and a lack of structural reinforcement
in the majority of constructions (Dowling, 2006). Unfortunately, there is a strong overlap
between the regions of the world where there significant use of earthen building occurs and
zones of moderate or greater seismic hazard (De Sensi, 2003), as shown in Figure 1.1.
(a) (b)
Figure 1.1(a). Regions where significant use of earthen building occurs (De Sensi, 2003).
Figure 1.1(b). Zones of moderate or greater seismic hazard (De Sensi, 2003).
Considerable previous research has aimed to address this problem by providing methods of
reducing the seismic vulnerability of adobe construction which are suitable for use in the
countries at risk (discussed in Section 2.3). However, there has been insufficient large-scale
take-up of these potential improvements, evident in the high numbers of deaths caused by
recent earthquakes in regions of high adobe use. For example, major earthquakes in India, Iran
and Pakistan in the last six years, regions where adobe housing is commonly used, each killed
between 20,000 and 80,000 people (USGS, 2006).
Section 1: Introduction
2
(a) (b)
Figure 1.2(a). Adobe building failure in Pakistan earthquake, 2005 (www.bbc.co.uk).
Figure 1.2(b.) Collapse of many adobe buildings in Iran earthquake, 2003 (www.ngdir.ir).
Recent research has concluded that the reason for the lack of widespread use of adobe
improvements developed so far is that:
“most of the proposed systems are too complex and/or too costly to be widely
used without sustained external intervention” (Dowling, 2006)
Therefore it is clear that further research is needed to develop suitable methods of improving
the seismic resistance of adobe buildings which are simple and cheap enough for widespread
acceptance and use.
1.2. Project objectives
(a) To test scale models of unreinforced adobe walls on a 1-g shaking table to accurately
simulate the behaviour of real adobe buildings under earthquake loading.
(b) To test scale models of adobe walls with structural improvements to enhance seismic
resistance, which are simple and cheap enough for potential widespread use in developing
countries, to assess their relative effectiveness.
(c) To disseminate the results in appropriate useful formats to relevant organisations and
practitioners working in the field.
3
Section 2: Literature Review
2.1. Seismic performance of adobe
Experience from past earthquakes has shown the most common ways in which adobe buildings
fail under seismic loading. These are summarised in Figure 2.1. The main collapse mechanisms
are: overturning of transverse walls; vertical cracking at corner joints; and diagonal shear
cracking of in-plane walls. The diagrams are simplified to assume that earthquake motion is
parallel to one set of walls and perpendicular to the other. In reality this is not usually exactly
true so combinations of the different mechanisms are likely.
(a) (b)
Figure 2.1(a). Typical failure mechanisms of adobe buildings (from Blondet et al, 2003).
Figure 2.1(b). Typical failure mechanisms of walls only (from Tolles and Krawinkler, 1990).
The key criteria considered suitable for measuring the seismic performance of adobe buildings
(cited by Dowling, 2006, and Tolles and Krawinkler, 1990) are: no damage during minor
earthquakes; tolerable damage during moderate earthquakes; and heavy damage but no
collapse during extraordinarily severe earthquakes. The evidence of past earthquakes shows
that large numbers of adobe buildings do not currently meet these criteria.
2.2. Laboratory testing of model adobe walls and buildings
A variety of research projects which have included dynamic adobe model testing has been
undertaken over the past 30 years. The main outcomes most relevant to this project are
Section 2: Literature Review
4
summarised as follows, in two categories: the techniques used to model the adobe buildings
and any improvement methods used, and the experimental procedures used for dynamic testing
of models using a shaking table.
2.2.1. Modelling adobe buildings and seismic strengthening methods
Scale model size: Previous testing on 1:2 and 1:5 scale models has concluded that in general
the behaviour of the smaller models could accurately predict the behaviour of the larger
models, with similar failure modes and similar response to improvement methods at the two
different scales (Tolles et al, 2000). Previous tests on 1:5 scale models in Cambridge
University were considered to model typical real-life collapse mechanisms well (Tetley and
Madabhushi, 2007). However, testing at 1:20 scale did not simulate common real-life collapse
mechanisms consistently (Malton, 2005).
Modelling adobe bricks: Some previous research has used gravel to simulate adobe blocks
and simplify the modelling procedure, instead of making individual bricks as is usual in real
life. This is based on the assumption that failure planes usually pass through the mortar which
surrounds the blocks, rather than through the bricks (Tetley, 2006). However, some other
testing has observed failures through bricks themselves (Dowling, 2006). Also, the bond
strength between the mortar and the brick is affected by the brick material. This is because the
bond is a function of the amount of water absorbed into the brick, which softens the brick to
allow a stronger brick-mortar bond (Tolles and Krawinkler, 1990). A piece of gravel is able to
absorb less water than an adobe brick. The two factors of differing bond strength and possible
failure through blocks would suggest that constructing individual adobe blocks to scale would
provide a more realistic model than the use of gravel, if this was practical at the scale used.
Considering the effect of gravity: Some previous research has neglected the effect of gravity
on scale models, by assuming that gravity-induced stresses on adobe buildings (which are
usually single-storey) are much less than stresses induced by seismic loading (Tolles and
Krawinkler, 1990, Malton, 2005, and Tetley, 2006). However, other projects have noted
differences in performance in models with different gravity loads and concluded that increased
gravity loads increase diagonal cracking and subsequent damage (Tolles et al, 2000).
Considering the effect of roof loads: Some previous research has considered the effect of
varying roof loads by adding different masses to the tops of the model walls (e.g. Flores et al,
Section 2: Literature Review
5
2001). Other research has assumed that the light roofs typical of adobe buildings have
negligible effect (Dowling, 2006).
Modelling realistic failure patterns: Previous research has emphasised that adobe scale
models should exhibit real-life failure patterns (such as those described in Section 2.1) if they
are to provide useful information on the likely behaviour of real-life buildings (e.g. Dowling,
2006). Two key factors which affect failure mechanisms are the materials used to simulate the
adobe (Tetley, 2006), and the boundary conditions imposed on models which are partial
sections of a wall (usually ‘I’, ‘L’ or ‘U’ shaped) instead of complete houses (Dowling, 2006).
Reinforcement materials and techniques: Research should consider both horizontal and
vertical reinforcement. Reinforcement can be either internal or external to the walls, but only
external reinforcement is suitable for reinforcing existing houses. A wide variety of ductile
materials have been investigated for internal reinforcement but less research has focused on
external methods. These include geogrids or polymer mesh attached with polypropylene ties
through the wall (Blondet et al, 2005), bamboo tied with wire and string (Dowling, 2006) and
wire mesh nailed with metal bottle caps (e.g. Zegarra et al, 1999, and Tetley, 2006).
2.2.2. Testing models using a shaking table to simulate an earthquake
Sequence of earthquakes which are input to models: A consistent sequence of earthquakes
simulated for each model tested in a project is not always used (e.g. Yamin, 2005, and Tetley
and Madabhushi, 2007) but is desirable if possible (Dowling, 2006) to provide a more
consistent basis for comparing the results of different models.
Frequency of earthquakes which are input to models: Differing approaches have been taken
to selecting the earthquake excitation frequencies used in model testing. Dowling (2006)
proposes that for dynamic similitude, the frequency ratio used in each model test should be
identical, where the frequency ratio is defined as the “ratio of dominant input excitation
frequencies to structural frequencies (first natural frequency of each specimen)”. This requires
the input spectra to be uniquely time-scaled for each model test to change the dominant input
frequency, since each model will have a different first natural frequency. This method then
allows each model to be tested at near-resonant conditions by matching the dominant
frequencies of the input earthquake with the natural frequency of the model. This ensures that
each model experiences the most damaging conditions possible. However, this approach has
Section 2: Literature Review
6
not been followed in other research. Instead, usually an unscaled input earthquake series has
been used and each model has been tested with the same set of earthquakes. As discussed in
Section 3.1, this approach can still provide useful information about structural response at non-
resonant frequencies, provided that it is established during testing that the input frequencies are
not near to the natural frequency of the model.
Data recording: Much of the previous research has produced qualitative data only. However it
is recommended that greater quantitative data would be useful to examine more detailed
aspects of model behaviour and to provide information for the development and validation of
possible finite element models of adobe structures under earthquake loading (Dowling, 2006).
2.3. Existing guidelines on adobe construction
A variety of guidelines exists on adobe construction for improved seismic resistance. The
scope of the guidelines tends to cover the following key areas for construction: general layout
and construction features; adobe block construction methods and materials; and wall
reinforcement methods and materials.
In the existing guidelines there is little discussion of the varying cost and complexity of the
different methods proposed, nor issues such as local availability of materials or techniques for
retrofitting improvements onto existing buildings. These themes have all been proposed for
further investigation (Tetley, 2006, and Dowling, 2006).
The form of horizontal and vertical wall reinforcement is the main aspect within the scope of
this project. However, issues which are important in the other areas of construction are also
described here. This is because any wall reinforcement methods developed must take into
account the other strengthening techniques that may be used in the construction of adobe
buildings, to ensure compatibility.
2.3.1. General layout and construction features
The guidelines typically include information on site selection, building plan layout, dimensions
of walls and openings, construction of foundations and roofs, the use of buttresses or pilasters,
and the use of a ‘ring beam’ (a continuous beam, usually made of timber or concrete, which
runs around the tops of the walls and ties the walls together). Despite details on particular
Section 2: Literature Review
7
dimensions and specifications differing between the guidelines, there seems to be a broad
consensus on the factors that are most important. However, since the construction of complete
model houses was beyond the scope of this project, detailed consideration of house layout and
features was not required.
2.3.2. Adobe block construction methods and materials
Dowling (2006) observes that there is no overall agreement in the literature on the best
proportions of different soil types that should be mixed together to produce adobe blocks.
However, a simple test is available to ensure sufficient strength in whatever soil mixture has
been produced. This involves making dried balls of soil which cannot easily be crushed
between finger and thumb (IAEE, 2004). There are also tests suggested to assess the strength
of finished adobe blocks after four weeks’ drying by supporting in bending the weight of a man
(IAEE, 2004) or by checking for specified maximum crack lengths, widths and depths
(Middleton, 1987). These tests avoid the need to rely on following prescriptive mix
proportions, which may be difficult to satisfy due to limited local availability of soil types.
Overall the guidelines recommend that block size and strength are much less important than the
mortar properties because the mortar must dry in-situ between the blocks and so is more
susceptible to weaknesses from micro-cracking than the blocks, which dry separately (IAEE,
2004). This supports the argument of Tetley (2006) that adobe buildings can be modelled using
gravel to mimic adobe blocks instead of making individual blocks themselves.
2.3.3. Wall reinforcement methods and materials
A key aspect in all guidelines is the use of ductile horizontal and vertical reinforcement.
Reinforcement materials include bamboo, reeds, cane, vines, rope, timber, chicken wire,
barbed wire, or steel bars (Blondet et al, 2003). All reinforcement should be attached to the
other reinforcement and structural elements of the building (the foundations, roof, and ring
beam, if present) to provide a ‘stable matrix’ which restrains bending and therefore minimises
cracking and collapse of the walls. Horizontal reinforcement also assists in the transmission of
forces from out-of-plane transverse walls to in-plane shear walls. Vertical reinforcement also
helps minimise out-of-plane bending and in-plane shear. Most methods proposed so far use
internal reinforcement, which is a method only feasible on newly-built adobe houses.
8
Section 3: Experimental techniques
3.1. Shaking table and input motion
All earthquake simulations were undertaken on a uni-axial shaking table at the Schofield
Centre, University of Cambridge. For each earthquake simulated, one frequency (set via a
control box) and one maximum peak displacement level (set by manually changing the crank
radius of the table) were chosen. Unlike some shaking tables, the table used in this project is
not capable of modelling complex input spectra. The specification of the shaking table is
shown in Table 3.1.
Table 3.1. Specification of 1-g shaking table at the Schofield Centre.
Size of table 900mm × 450mm Nominal maximum displacement ±11mm Nominal maximum acceleration ±1.02g Nominal maximum frequency 4.8Hz
After the initial tests to establish realistic collapse mechanisms, the procedure used for all
experiments for comparing structural improvements was first to subject the model to a small
amplitude simulated earthquake at a particular frequency to establish that the test was not at
resonant frequency. (If the frequency was near the natural frequency of the model then there
would be a disproportionate response observed even at small amplitudes). The model was then
subjected to a series of increasing amplitude earthquakes at the same frequency until collapse.
This approach ensured that none of the tests was performed at near-resonant frequencies and so
allowed comparisons of the response of the different models in different tests. This was a more
suitable approach for the facilities available than changing the input frequency to ensure all
tests are at near-resonant frequencies, as proposed by Dowling (2006) and discussed in Section
2.2.2.
3.2. Visual data recording
All tests were recorded as videos using a standard Canon digital camera. The camera was also
used to take photos to show the stages of collapse of a model. A high speed Phantom video
camera (with a capacity of 1000 frames/second) was used to record video of the tests at 100
frames/second, to allow detailed observation of collapse mechanisms.
Section 3: Experimental Techniques
9
3.3. Instrumentation and data recording
A series of accelerometers and an LVDT (Linear Variable Differential Transformer)
displacement transducer were used to record the acceleration and displacement of particular
locations on the shaking table and models. These approximate locations are shown in Figure
3.1 (model design is discussed in Section 4). Each accelerometer location on the model varied
up to 25mm between different tests because accelerometers had to be fixed to a smooth, flat
section of adobe on the model wall (despite the use of wooden moulds to assist construction,
discussed in Section 4.3, the whole surface of the walls was not completely smooth).
Figure 3.1. Locations of accelerometers and LVDT on model walls and shaking table.
The accelerometer and LVDT data was recorded via junction boxes (which converted the
LVDT and accelerometer signals to volts) onto a computer using DasyLab and then
manipulated after testing using Microsoft Excel and MathWorks MatLab. The accelerometers
(DJ Birchall A/32 piezo-electric charge output model) were calibrated using a Brüel and Kjaer
Accelerometer Calibrator Type 4291.
However, the model of accelerometer used has limited accuracy at the frequencies (5Hz or
below) used in this project. As discussed in Section 6.2, there was a consistent difference found
between the recorded acceleration data from the accelerometers and the calculated acceleration
determined from the displacement and frequency measurements. This difference is likely to be
due to the inaccuracy of the accelerometers at the frequencies used. However, since the
difference was consistent it was concluded that the results obtained were still valid for
comparisons between different wall positions, earthquakes and tests.
Ac5
Ac7
Ac6 Ac3
Ac2
Ac1
Ac4
SHAKING TABLE
MODEL WALL
LVDT
Direction of earthquake motion
10
Section 4: Modelling adobe walls and seismic strengthening methods
4.1. Key issues and summary of model wall designs chosen
The main considerations in designing the model walls to be tested were the need to ensure:
• Failure of the model walls within the capacity of the shaking table so that the collapse of all
models could be observed and comparisons made of the improvements tested.
• Failure of the model walls by realistic collapse mechanisms to accurately simulate real-life
behaviour of adobe buildings under earthquake loading.
To address these needs, two types of model adobe walls were tested during this project:
• Single unreinforced model walls were tested in-plane with the motion of the shaking table
to check that the in-plane walls (and therefore also the more vulnerable out-of-plane walls)
of unreinforced corner joint models would fail within the capacity of the shaking table.
• ‘L’-shaped corner wall joints were tested (with and without structural improvements) to
observe behaviour of the corner and the out-of-plane wall section, typically the two most
vulnerable parts of adobe buildings. This allowed confirmation that typical real-life collapse
mechanisms were modelled, as discussed in Section 2.1.
(a) (b)
Figure 4.1(a). Single unreinforced in-plane model wall [Test 1].
Figure 4.1(b). ‘L’-shaped corner wall joint [Test 3].
Tetley (2006) also performed tests on corner joints, and proposed that this model could
simulate the real-life situation where there is an opening (i.e. a door or window) in the
Section 4: Modelling adobe walls and seismic strengthening methods
11
transverse wall of the building which could permit rotation of the transverse wall if cracking
occurs near the opening, as shown in Figure 4.2.
Figure 4.2. Situation represented by corner joint model (from Tetley, 2006).
However, in this project it was decided to model a more general situation of a continuous
transverse wall without existing cracking near openings. To simulate the effect of a continuous
transverse wall, a support must be provided to the free end of the transverse wall in the joint
model to prevent rotation of the whole joint. This support is shown schematically in Figure 4.3
and the actual construction of the support is described in Section 4.4.
Figure 4.3. Support provided to prevent rotation and simulate continuous transverse wall.
After determining the types of model walls to test, the design issues remaining were then:
• The size of the model walls.
• The material composition and construction of the model walls.
• The method used to model a continuous transverse wall and boundary conditions.
• The techniques used to model structural improvements.
Direction of earthquake motion
Transverse wall
In-plane wall
Support required by transverse wall to simulate continuous wall
Section 4: Modelling adobe walls and seismic strengthening methods
12
4.2. Size of the model walls
It has been shown by dimensional analysis (by Tolles and Krawinkler, 1990, p44, and Tetley,
2006, p16-17) that for creating a scale model wall to have dynamic similitude with a real-life
full-size prototype wall, the following must be fulfilled:
rr
r lE
=ρ
(4.1)
where E = Young’s modulus, ρ = material density, l = length and the subscript r denotes the
ratio of units which defines the model [m] to real-life prototype [p] relationship for one of the
relevant physical properties (i.e. Er = Em / Ep).
Based on Tetley (2006), equation (4.1) implies that for dynamic similitude, the model must
have higher density or lower stiffness than the real-life prototype. Higher density can be
achieved by the technique of ‘artificial mass simulation’, which involves adding extra masses
to the model, in a near-uniform distribution. The research of both Tolles and Krawinkler and
Tetley concluded that this method is not feasible for small-scale models. Using materials of a
significantly lower stiffness was also considered too difficult.
However, if gravitational effects are neglected then the dimensional analysis simplifies so that
material properties can be equal in the model and the prototype (i.e. density ratio ρr = 1 and
stiffness ratio Er = 1). As discussed in Section 2.2.1, the assumption of negligible effect due to
gravity has been considered valid by previous researchers because gravity-induced stresses on
adobe buildings (which are predominantly single-storey) are much less than stresses induced
by seismic loading. This simplified analysis leads to the relationship ar = 1 / lr . Therefore the
length ratio (lr) is the important ratio which determines the acceleration required in the model
test compared to the prototype.
Tetley used this relationship ar = 1 / lr to conclude that the model should be built as large as
possible to ensure that model collapse can be achieved within the acceleration capacity of the
shaking table. Previous experiments on the shaking table at the Schofield Centre by Malton
(2005) and Tetley (2006) assumed a typical real-life prototype single wall size of 2.8m × 2.2m
× 0.5m. Loaiza et al (2002) noted typical real-life wall thicknesses of 0.3m - 0.8m, and IAEE
(2004) recommends a maximum wall height to thickness ratio of 8:1 and a maximum wall
length to thickness ratio of 10:1, so this assumed prototype size matches these real-life
observations and suggested criteria.
Section 4: Modelling adobe walls and seismic strengthening methods
13
Based on these dimensions and the size of the shaking table available, the maximum single
wall model size possible was 1:5 scale, i.e. 0.56m × 0.44m × 0.10m. This was chosen as the
most suitable single wall model size based on the arguments of the other researchers outlined
above and to allow direct comparison of model tests with the work of Tetley (2006), who used
models with these dimensions. The corner joint model wall was chosen to have the same 1:5
scale and the same dimensions as the corner models used by Tetley, shown in Figure 4.4.
Figure 4.4. Dimensions of corner joint wall model (from Tetley, 2006).
4.3. Material composition and construction of the model walls
As discussed in Section 2.2.1, model adobe walls can be made either by manufacturing
individual adobe blocks (e.g. Dowling, 2006) or using gravel to simulate adobe blocks (e.g.
Malton, 2005, and Tetley, 2006). Walls are then made by joining the chosen ‘bricks’ with
mortar. For simplicity of construction, it was decided in this project to use gravel as the
‘bricks’ in the walls because constructing individual blocks at 1:5 scale would have been too
time-consuming. The possible limitations of this approach in modelling adobe walls are
discussed in Section 2.2.1. For consistency and simplicity, it was decided to use wooden
moulds, shown in Figure 4.5, to assist the construction of the walls, as used by Tetley (2006).
The gravel ‘bricks’ were laid layer by layer to imitate real-life construction, for all models.
To create a suitable mortar, a mixture of clay and sand was used. The initial composition by
weight used was 60% gravel, 40% mortar, where the mortar was 30% sand and 70% Kaolin
clay with a water content of w = 0.45. These proportions had been initially used by Tetley
(2006) with the aim of producing a mortar that enabled wall failure within the capacity of the
Schofield Centre shaking table. Tetley’s further experiments then concluded that a mortar of
Section 4: Modelling adobe walls and seismic strengthening methods
14
40% sand and 60% Kaolin clay was preferable to ensure failure. In this project it was decided
to perform tests with both 30% sand and 40% sand contents to confirm the better option.
(a) (b)
Figure 4.5(a). Construction of single wall model in mould [Test 1, unreinforced].
Figure 4.5(b). Construction of corner joint model in mould [Test 6, with internal bamboo].
Prior to testing on the shaking table, shear tests of mortars with different drying times were
performed, summarised in Table 4.1. This highlighted the importance of the length of drying
time of the mortar, and the extra shear strength due to interlocking of the gravel blocks.
Table 4.1. Summary of the results of shear tests on different mortars.
Material tested
Time allowed for drying
Maximum shear strength recorded (kPa)
Mortar of 30% sand, 70% Kaolin None (unset) 4
Mortar of 30% sand, 70% Kaolin 3 days 10
Mixture of gravel + mortar of 30% sand, 70% Kaolin 3 days 22
After each shaking table test, the wall composition and construction method was reviewed to
check the requirements of failure within capacity of the shaking table and by realistic failure
mechanisms. Table 4.2 shows the composition and construction methods used for each test.
Table 4.3 explains the developments in construction and materials as the project progressed,
based on the following materials for the model adobe walls:
• Powdered E Type Kaolin clay, mixed to water content w = 0.45.
• Hostun sand (Tests 1-2), Rugby building sand (Tests 3-7).
• Gravel of length generally 30mm - 80mm.
Section 4: Modelling adobe walls and seismic strengthening methods
15
Table 4.2. Summary of construction method and materials used.
Test Wall type
Structural improvements
Mortar sand
content
Sand type
Sand pre-drying
Model drying time before
testing
1 Single in-plane None 30% Hostun
sand Not pre-dried 4 days
2 Single in-plane None 40% Hostun
sand Not pre-dried 4 days
3 Corner joint None 40% Building
sand Not pre-dried 4 days
4 Corner joint None 40% Building
sand Not pre-dried 37 days
5 Corner joint
External bamboo (wide spacing) 40% Building
sand Pre-dried* 4 days
6 Corner joint Internal bamboo 40% Building
sand Pre-dried* 4 days
7 Corner joint
External bamboo (narrow spacing) 40% Building
sand Pre-dried* 4 days
* All pre-drying was in an oven for 24 hours at 70oC
Table 4.3. Explanation of development in construction method and materials used.
Test Development in construction materials and method
Issues arising from construction materials and method
1 Used original composition of 60% gravel, 40% mortar, where mortar was 30% sand and 70% Kaolin clay with a water content w = 0.45.
Building sand unavailable so Hostun sand used. Collapse acceleration too near shaking table capacity so needed to weaken mortar in subsequent tests.
2 Increased sand content to 40% from Test 1 to weaken mortar and decrease collapse acceleration.
Building sand unavailable so Hostun sand used. Collapse acceleration still too near shaking table capacity so still needed to weaken mortar.
3 Used 40% coarse building sand instead of 40% fine Hostun sand to increase both mortar drainage speed and amount of drying and therefore reduce both mortar ductility and collapse acceleration.
Collapse acceleration of transverse wall was high compared to equivalent test of Tetley (2006). Observation of mortar showed that there was still dampness and suction after the test (explained in Section 5.1), indicating that further drying was still needed.
4 Allowed model to dry for 37 days to avoid suction remaining.
Collapse acceleration of transverse wall very similar to test of Tetley with pre-dried sand.
5 6 7
Used pre-dried sand to avoid suction remaining in mortar, then allowed model to dry for 4 days.
Measurement of moisture content of mortar samples after drying was equivalent to the model with long drying period in Test 4 so pre-drying was confirmed as valid alternative (discussed further in Section 7.2).
Section 4: Modelling adobe walls and seismic strengthening methods
16
4.4. Method used to model a continuous transverse wall and boundary conditions
As discussed in Section 4.1, a support was needed for the free end of the transverse wall in the
corner joint model to prevent rotation of the model, in order to simulate the effect of a
continuous transverse wall. This support was provided by a wooden attachment bolted to the
base of the wooden ‘holder’ for the wall which was fixed to the shaking table.
Before testing, the support was designed to touch the transverse wall but not exert any force on
the wall. For simplicity of construction, the support was only on one side of the wall, so it only
restrained movement in one direction. A full continuous wall would provide some restraint in
both directions but the support in this project was considered a reasonable approximation for
the simple model used, and more realistic than providing no support at all.
Dowling (2006) proposed that restraint is also needed for the in-plane wall to simulate the
effect of this being a continuous wall and joining the rest of the building in real life. Dowling’s
experiments used an applied downward force at the free end of the in-plane wall to simulate
this restraint, with the aim of preventing overturning of the in-plane wall. However, Zegarra et
al (1999) did not use any restraint on the in-plane walls in model tests. Dowling argued that the
results of Zegarra’s experiments (including shear cracking of the in-plane walls) were less
consistent with real-life failure patterns. However, as discussed in Section 2.1, other research
has observed diagonal cracking of in-plane walls as a typical real-life failure mechanism. It
was decided for this project not to provide further restraint to the in-plane walls.
Figure 4.6. L-shaped joint with wooden support to simulate continuous transverse wall.
Section 4: Modelling adobe walls and seismic strengthening methods
17
4.5. Techniques used to model structural improvements
The structural improvement techniques tested were chosen to build on previous laboratory
testing and research in the field, with the aim of investigating methods that are appropriate (in
terms of cost and complexity) for the regions that require them. The methods proposed in
previous research in Cambridge University by Tetley (2006) were not considered suitable for
further investigation because they were either impractical in terms of cost (steel reinforcement
or cement), practicality of construction (using a mesh made from strips of plastic bags, which
would require prohibitive amounts of labour time in real life) or had limited scope for further
investigation using the facilities available (wire mesh).
4.5.1. External vertical bamboo tied with external horizontal wire
Previous research (especially by Dowling, 2006) has indicated that bamboo and wire offers
significant extra seismic resistance, but this method has not previously been tested in
Cambridge University. These tests were chosen to provide a useful comparison with tests
performed by other researchers, by offering further insight into both the reinforcement and the
testing process. Two tests were performed using external bamboo: with spacing of bamboo
larger than average gravel size, 100mm [Test 5], and with spacing of bamboo smaller than
average gravel size, 50mm [Test 7]. The two experiments were designed to allow observations
of the effect of the bamboo and wire in adding ductility to the structure, and also the effect of
the bamboo and wire matrix acting as a mesh to provide confinement for the gravel and mortar.
The technique of external bamboo and wire reinforcement also has potential as a retrofitting
measure for existing adobe buildings if holes can be drilled through the walls of the structure.
The method involves putting wire (or string) horizontally through the wall at various points and
then using this to attach vertical bamboo canes either side of the wall. The canes are then
linked by horizontal wires parallel to the wall, which can be looped round adjacent canes and
tied and tightened with pliers. The technique is likely to be much more difficult for buildings
made of stones (which are more difficult to drill through than adobe). However, although this
project used stones to model the adobe, the wires through the wall could be placed during
construction, within the mortar layers (as shown in Figure 4.7). The bamboo was not attached
to the wooden base (to match as near as possible the technique used by Dowling, 2006), but
could have been attached to a ring beam or other structural features if these were present.
Dimensions of bamboo and wire in Tests 5-7 are listed at the end of Section 4.5.
Section 4: Modelling adobe walls and seismic strengthening methods
18
(a) (b)
Figure 4.7(a). Construction with wires placed in mortar layers [Test 5 with external bamboo].
Figure 4.7(b). Construction after mould removed, with wires ready to attach bamboo [Test 5].
(a) (b)
Figure 4.8(a). Wall joint with some vertical bamboo attached [Test 5 with external bamboo].
Figure 4.8(b). Wall joint with all vertical bamboo attached, before wooden support attached
[Test 5 with external bamboo].
4.5.2. Internal vertical bamboo tied with internal horizontal wire
Internal reinforcement such as bamboo provides extra ductility for the structure, but does not
have the possible advantages of containment or retrofitting that external reinforcement has.
Internal bamboo or other materials are already widely recommended, as discussed in Section
2.3.3. However, research by Dowling (2006) indicated that internal bamboo could in fact cause
extra weakness due to discontinuity in the adobe, and proposed further investigation.
Wire spacing = 180mmBamboo spacing =
100mm in Test 5, 50mm in Test 7
Position of wire represented by black lines
Section 4: Modelling adobe walls and seismic strengthening methods
19
This discontinuity in real-life adobe buildings occurs because some adobe blocks will have to
be cut to shape to accommodate the reinforcement (usually, semi-circles would have to be cut
from the ends of adjacent adobe blocks to provide a circular gap for the bamboo). This also
makes construction more difficult and so makes the technique less suitable for communities in
poorer regions which have less access to training in building techniques. However in this
project it was possible to temporarily hold the bamboo canes in place using the mould, build
the wall up around the canes, and then attach horizontal internal wire within the mortar layers.
(a) (b)
Figure 4.9(a). Wall joint with internal bamboo held temporarily in mould during construction.
Figure 4.9(b). Completed wall joint with tops of internal bamboo and joining wires visible.
Summary of materials used for structural improvements:
• Bamboo of diameter in the range 5mm - 12mm, sourced from HomeBase and B&Q in
Cambridge. Dowling (2006) used bamboo of 16mm - 22mm diameter in tests of 1:2 scale
wall units, indicating that 5mm - 12mm is of the correct order of magnitude for testing 1:5
scale models. The diameter used is equivalent to 25mm - 60mm at full scale. In real life
bamboo diameters up to 100mm are common (Janssen, 1995).
• Galvanised wire of diameter 1mm, sourced from Mackays Hardware of Cambridge. Dowling
(2006) used galvanised wire of 2mm diameter in tests of 1:2 scale wall units, indicating that
1mm is of the correct order of magnitude for testing 1:5 scale models. Guidelines on the use
of internal wire reinforcement suggest that 2mm - 3mm is readily available in the field
(Nienhuys, 1999). Provided the wire has sufficient tensile strength to avoid snapping under
the loading generated, then the exact diameter is not critical.
Bamboo spacing = 100mm in Test 6
20
Section 5: Behaviour of single walls
As discussed in Section 4.1, two single unreinforced model walls were tested in-plane with the
motion of the shaking table. These tests would check that the in-plane walls (and therefore also
the more vulnerable out-of-plane walls) of unreinforced corner joint models tested later would
fail within the capacity of the shaking table. The failure mechanisms of the in-plane walls
should also be realistic compared to typical real-life mechanisms.
Dimensions of the walls are described in Section 4.2. The following tests were performed (all
accelerations referred to within the text of this section are the nominal calculated accelerations
of the shaking table, as discussed in Section 5.1):
Test 1 (13/11/06):
• Single wall in-plane. 30% sand in mortar (Hostun sand, undried), model dried for 4 days.
• Cracks began to open at 0.39g, total collapse at 1.05g.
Test 2 (20/11/06):
• Single wall in-plane. 40% sand in mortar (Hostun sand, undried), model dried for 4 days.
• Total collapse at 0.91g, no cracks observed at accelerations lower than collapse acceleration.
The next sections summarise the visual observations for the in-plane single wall tests, and then
summarise the relevance of the results in this stage of the project. No useable instrumentation
data was recorded for the in-plane single wall tests: for Test 1 the necessary equipment was
unavailable and for Test 2 there was an error in recording the data.
5.1. Visual observations
After each earthquake, the damage observed was recorded qualitatively, by digital camera
images and high speed video camera. Damage was also categorised based on the classification
system described by IAEE (2004), summarised in Table 5.1. This was adopted to provide a
consistent form of comparison between the effects of different earthquakes on different
models. The same classification system was used throughout the project, both for single in-
plane walls and corner joint models.
Section 5: Behaviour of single walls
21
Table 5.1. Categorisation of damage to buildings (from IAEE, 2004)
Damage category Description of extent of damage 0 No damage No damage. 1 Slight non-structural damage Thin cracks in plaster; falling of plaster in limited parts.
2 Slight structural damage Small cracks in walls; falling of large plaster pieces over wide areas; damage to non-structural parts e.g. chimneys.
3 Moderate structural damage Large and deep cracks in walls; widespread cracking of walls and columns; fall of chimney etc.
4 Severe structural damage Gaps occur in walls; inner and outer walls collapse; failure of ties to parts of buildings, 50% of structural elements fail; demolition or extensive restoration needed.
5 Collapse A large part or whole of the building collapses.
Table 5.2. Summary of Test 1: Single wall in-plane with 30% Hostun sand in mortar. Nominal maximum displacement of table (mm)
Nominal frequency (Hz)
Nominal calculated acceleration of table (m/s2)
Observations Damage category
2 4.00 0.13g No damage observed 0 6 4.00 0.39g Cracks begin to open up 2 8 4.00 0.52g No further cracks 2
11 4.90 1.05g ‘Rocking’ motion of wall on base followed by total collapse initiated with large diagonal cracks
5
Table 5.3. Summary of Test 2: Single wall in-plane with 40% Hostun sand in mortar. Nominal maximum displacement of table (mm)
Nominal frequency (Hz)
Nominal calculated acceleration of table (m/s2)
Observations Damage category
3 4.02 0.20g No damage observed 0 6 4.02 0.40g No damage observed 0 6 4.50 0.49g No damage observed 0 8 4.50 0.65g No damage observed 0
10 4.75 0.91g
‘Rocking’ motion of wall on base (less severe rocking than Test 1) followed by total collapse initiated with large diagonal cracks
5
In Table 5.2 and 5.3, the displacements listed are the nominal displacements measured directly
from the shaking table crank radius. The frequencies listed are the nominal frequencies
displayed on the control box. The accelerations are the nominal accelerations calculated from
the nominal displacement and frequency. When instrumentation was used to record the actual
Section 5: Behaviour of single walls
22
displacements and frequencies from Test 3 onwards it was found that these were consistently
higher than the nominal values (discussed in Section 6.2). However the nominal values were
sufficient for Test 1 and Test 2 since detailed analysis of the single walls was not required.
Figures 5.1 and 5.2 show images of the collapse of the single in-plane walls in Test 1 and Test
2. The initial rocking motion and the formation of the diagonal cracks are visible.
(a) (b)
Figure 5.1(a). Test 1: Rocking occurring as right-hand end of wall ‘lifts off’, 1.05g.
Figure 5.1(b). Test 1: Major diagonal cracks form before total collapse, 1.05g.
(a) (b)
Figure 5.2(a). Test 2: Rocking motion seems to initiate damage and diagonal cracks, 0.91g.
Figure 5.2(b). Test 2: Diagonal failure planes visible after collapse (highlighted), 0.91g.
The rocking motion observed in Test 1 was reduced in Test 2 by the use of bolts in the wooden
base which extended into the base of the wall. The diagonal failure planes observed suggested
that realistic collapse mechanisms occurred after the initial rocking effect.
‘Lift off’ due to rocking
Diagonal failure planes
Section 5: Behaviour of single walls
23
(a) (b)
Figure 5.3(a). Test 1: Significant suction still holding gravel to mortar after testing.
Figure 5.3(b). Test 2: Mortar drier due to extra sand content, but some suction still present.
Figure 5.3 shows that in both Test 1 and 2 there was suction present in the mortar, indicating
that the mortar had not fully dried out before testing. In soil that has not dried fully, there will
be negative pore pressures (suction) which gradually dissipate as the soil dries out, until pore
pressure is zero when the soil has fully dried. Negative pore pressure causes an increase in the
strength of the soil since the soil shear strength depends on effective stress, where:
effective stress (σv’) = total stress (σv) - pore pressure (u) (5.1)
From equation (5.1), it can be seen that negative pore pressure (suction) during drying causes a
higher effective stress during drying than the final effective stress after drying. This causes a
higher strength in the mortar than desired while drying is still occurring. A similar problem had
been noted by Tetley (2006). Suction could therefore be reduced either by reducing the clay
content and increasing the sand content of the mortar (since the sand was coarser than the clay
and so a higher sand content would allow faster drainage), or by increasing the model’s drying
time. As discussed in Section 4.3, it was decided after Test 2 that the fine Hostun sand used in
Tests 1 and 2 should be replaced in subsequent tests by coarser building sand, to allow faster
drainage and therefore effectively increase the drying time.
Section 5: Behaviour of single walls
24
5.2. Summary of single wall results
• The high failure acceleration of Test 1 (30% sand in mortar) and the observation of suction
remaining in the mortar after testing demonstrated the need to increase the sand content of
the mortar for subsequent tests. This would achieve a drier mortar which would fail at an
acceleration not near the acceleration capacity of the shaking table.
• The high failure acceleration of Test 2 (40% sand in mortar) compared to an equivalent
test by Tetley (2006) and the observation of some suction still remaining in the mortar (albeit
less than in Test 1) indicated that coarse building sand should be used where possible in
subsequent tests, instead of fine Hostun sand, to allow faster drainage and achieve a drier
mortar.
• The initial rocking motion in Test 1 and Test 2 before failure suggested that longer bolts
fixed into the base of the wall should be used in subsequent tests to minimise this effect and
enable observations to be made of the wall failure mechanism rather than any wall-base
interaction.
• The diagonal shear cracking observed in Test 1 and Test 2 indicated that typical real-life
collapse mechanisms for in-plane walls occurred after the initial rocking effect.
25
Section 6: Behaviour of ‘L’-shaped corner wall joints
As discussed in Section 4.1, ‘L’-shaped corner wall joints were tested (with and without
structural improvements) to observe behaviour of the corner and the out-of-plane wall section,
typically the two most vulnerable parts of adobe buildings.
Firstly, two corner wall joints were tested without any structural improvements. The aim of
these tests was to confirm that typical real-life collapse mechanisms were observed in the
models and to provide a benchmark with which to compare later models which had structural
improvements. The collapse mechanisms anticipated, as shown in Section 2.1, were
overturning of the transverse wall, vertical cracking at the corner joint, shear failure of the in-
plane wall, or a combination of these.
Dimensions of the walls are described in Section 4.2. All accelerations referred to within the
text are the calculated accelerations of the shaking table, as discussed in Section 6.1, except
where comparisons are made to previous research on the shaking table which only recorded
nominal accelerations. The following tests were performed:
Test 3 (01/12/06):
• ‘L’-shaped joint. 40% sand in mortar (building sand, undried), model dried for 4 days.
• First cracks observed at 0.56g, transverse wall began to oscillate at 0.84g.
• Total collapse of transverse wall at 1.02g, total collapse of in-plane wall at 1.12g.
Test 4 (12/01/07):
• ‘L’-shaped joint. 40% sand in mortar (building sand, undried), model dried for 37 days.
• First cracks observed and transverse wall began to oscillate at 0.29g.
• Total collapse of transverse wall at 0.47g, total collapse of in-plane wall at 0.95g.
6.1. Visual observations
The same system of observing, recording and categorising damage was used as for previous
tests, as explained in Section 5.1. In Table 6.1 and 6.2, nominal accelerations and calculated
accelerations are both shown. Nominal accelerations are presented to allow comparison with
Tetley (2006) who used nominal accelerations based on nominal displacement and frequency,
Section 6: Behaviour of ‘L’-shaped corner wall joints
26
using the same shaking table. Calculated accelerations based on the measured displacement
and frequency are presented because analysis of the LVDT and accelerometer data showed that
the measured displacement and measured frequency were consistently greater than the nominal
values (discussed further in Section 6.2).
Table 6.1. Summary of Test 3:
‘L’-shaped joint with 40% sand in mortar (building sand, undried), model dried for 4 days. Maximum displacement of table (mm)
Frequency (Hz)
Nominal acceleration of table (m/s2)
Calculated acceleration of table (m/s2)
Observations Damage category
2.50 4.30 0.13g 0.19g No damage observed 0 4.75 4.30 0.26g 0.35g No damage observed 0
7.50 4.30 0.39g 0.56g Some small pieces of mortar came away from the surface 1-2
9.10 4.30 0.52g 0.68g Some pieces of mortar came away; small cracks formed 2
9.10 4.80 0.65g 0.84g Small oscillation of transverse wall; deep crack near base 3
10.95 4.80 0.73g 1.02g Failure of transverse wall by overturning, initiated by near-vertical crack at joint
4-5
12.10 4.80 0.81g 1.12g Failure of in-plane wall, initiated by diagonal crack near the base
5
Table 6.2. Summary of Test 4:
‘L’-shaped joint with 40% sand in mortar (building sand, undried), model dried for 37 days. Maximum displacement of table (mm)
Frequency (Hz)
Nominal acceleration of table (m/s2)
Calculated acceleration of table (m/s2)
Observations Damage category
2.85 5.00 0.18g 0.29g Oscillation of transverse wall; deep crack from near base of transverse wall to top of joint
3
4.70 5.00 0.36g 0.47g Failure of transverse wall by overturning, initiated by near-vertical crack at joint
4-5
6.65 5.00 0.53g 0.67g No further visible effect on remaining in-plane wall 4-5
9.40* 5.00 0.71g 0.95g Failure of in-plane wall, initiated by diagonal crack 5
*Estimated value because the LVDT data was not useable due to wall collapse onto the LVDT.
Section 6: Behaviour of ‘L’-shaped corner wall joints
27
Figure 6.1 and Figure 6.2 show images taken from the digital camera video of the collapse
mechanisms of the transverse and in-plane wall failures in Test 3 and Test 4. The observed
failure mechanisms (overturning of the transverse wall initiated by vertical cracking at the joint
and collapse of the in-plane wall initiated by diagonal failure planes) seemed realistic.
(a) (b)
Figure 6.1(a). Test 3: Point of failure of transverse wall by overturning, 1.02g (nominal
acceleration 0.73g).
Figure 6.1(b). Test 3: Point of failure of in-plane wall, initiated by diagonal cracking, 1.12g
(nominal acceleration 0.81g).
(a) (b)
Figure 6.2(a). Test 4: Point of failure of transverse wall by overturning, 0.47g (nominal
acceleration 0.36g).
Figure 6.2(b). Test 4: Point of failure of in-plane wall, initiated by diagonal cracking,
highlighted, 0.95g (nominal acceleration 0.71g).
Overturning
Overturning Diagonal cracking
Diagonal cracking
Section 6: Behaviour of ‘L’-shaped corner wall joints
28
Collapse in Test 3 still occurred too near to the maximum acceleration of the shaking table and
at a much higher acceleration (nominal acceleration 0.73g for the transverse wall) than in the
equivalent test by Tetley (2006), in which the transverse wall collapsed at nominal acceleration
0.32g. It was observed that there was still suction in the mortar after Test 3, as discussed
previously in Section 5.1. This led to the decision to allow an extended period of drying time
(37 days) for the model in Test 4 compared to the previous tests (4 days). The difference in
mortar dryness due to this extra time is visible in Figure 6.3.
(a) (b)
Figure 6.3(a). Test 3, model dried for 4 days: Gravel still held in mortar by suction after
failure of transverse wall.
Figure 6.3(b). Test 4, model dried for 37 days: Mortar drier with much less suction after
failure of transverse wall.
The nominal collapse acceleration (0.36g) of the transverse wall in Test 4 was very similar to
the nominal collapse acceleration (0.32g) of the wall tested by Tetley (2006) which dried for 4
days but used pre-dried sand. This suggested that it is necessary to use a model without suction
present as a benchmark, and that the level of drying needed can be achieved either by pre-
drying the sand or by leaving the whole model to dry for an extended period. After Test 4, the
moisture content of a sample of mortar set aside during construction and permitted to dry under
the same conditions as the wall was measured as 0.80%, three days after testing. In subsequent
tests, there was not sufficient time available to allow a drying period of 30+ days. However, the
sand could be pre-dried in an oven. The measurement of moisture content of the mortar from
the model in Test 4 could provide a benchmark for moisture content in subsequent tests which
used pre-dried sand. Provided the moisture contents were similar, comparisons of the effect of
structural improvements could then be made. This is discussed further in Section 7.2.
Section 6: Behaviour of ‘L’-shaped corner wall joints
29
6.2. Instrumentation data recorded
For Test 3 and 4, an LVDT and accelerometers were placed on the shaking table and the model
at the positions shown on Figure 3.1, Section 3.3. An example of the LVDT and accelerometer
data recorded is shown in Figure 6.4, which displays the measured displacement and
acceleration of the shaking table in the first earthquake of Test 4.
0 1 2 3 4 5 6 7 8 9-3
0
3
time, s
disp
lace
men
t, m
m
0 1 2 3 4 5 6 7 8 9-0.5
0
0.5
time, s
acce
lera
tion,
g
(a)tabledisplacement(LVDT)
(b)tableacceleration(Ac1)
Figure 6.4(a). Measured displacement of shaking table (LVDT) during Test 4, Earthquake A
(note the difference with the nominal displacement, 2mm).
Figure 6.4(b). Measured acceleration of shaking table (Ac1) during Test 4, Earthquake A (note
the difference with the calculated acceleration, 0.29g).
Full analysis of the LVDT and accelerometer data showed that on average the measured
displacement was about 20% greater than the nominal displacement (measured on the shaking
table crank) and the measured frequency was about 3% - 8% greater than the nominal
frequency (recorded from the shaking table control box). Therefore from Test 3 onwards,
accelerations quoted in the text are the accelerations calculated from the measured
displacement and frequency, instead of the nominal accelerations as used in Test 1 and 2.
From the full set of accelerometer data, MatLab was then used to produce graphs comparing
the acceleration at each position for each earthquake in Test 3 and 4. An example of the graphs
produced comparing the accelerations for the first earthquake of Test 4 is shown in Figure 6.5.
This graph clearly shows the increasing acceleration of the transverse wall compared to the
Section 6: Behaviour of ‘L’-shaped corner wall joints
30
other positions, due to the cracking and oscillation of the transverse wall. There is greater noise
evident in the accelerometer data from the transverse wall, which may cause the difference in
acceleration to be exaggerated. However, it is still clear from the graphs that the acceleration of
the transverse wall increases throughout the test as cracking develops, and that the acceleration
of the top of the transverse wall is higher than the acceleration of the lower part of the wall.
0 1 2 3 4 5 6 7 8 9-1
0
1
0 1 2 3 4 5 6 7 8 9-1
0
1
0 1 2 3 4 5 6 7 8 9-1
0
1
0 1 2 3 4 5 6 7 8 9-1
0
1
acce
lera
tion,
g
0 1 2 3 4 5 6 7 8 9-1
0
1
0 1 2 3 4 5 6 7 8 9-1
0
1
0 1 2 3 4 5 6 7 8 9-1
0
1
time, s Figure 6.5. Measured accelerations during Test 4, Earthquake 1:
(a) Acceleration of shaking table (‘ground’) [Ac1].
(b) Acceleration of in-plane wall, corner, lower position [Ac2].
(c) Acceleration of in-plane wall, corner, upper position [Ac3].
(d) Acceleration of in-plane wall, free end, lower position [Ac4].
(e) Acceleration of in-plane wall, free end, upper position [Ac5].
(f) Acceleration of transverse wall, lower position [Ac6].
(g) Acceleration of transverse wall, upper position [Ac7].
(a)
(b)
(c)
(d)
(e)
(f)
(g)
Section 6: Behaviour of ‘L’-shaped corner wall joints
31
In order to summarise the main information visible in each graph into a form that would allow
easy comparison between positions on the model, MatLab was used to find the mean of the
peak accelerations of the different positions during each earthquake1. The ratios of the
measured accelerations at the different positions on the model to the measured accelerations on
the shaking table were also calculated (i.e. the acceleration amplication factors). An example of
a complete set of these results, for Test 3, is shown in Table 6.3 and 6.4.
Table 6.3. Test 3: Mean of the peak measured accelerations of all different positions. Acceleration of table (m/s2)
Acceleration of in-plane wall, corner (m/s2)
Acceleration of in-plane wall, free end (m/s2)
Acceleration of transvserse wall (m/s2)
Earth- quake ref.
Calculated acceleration of table (m/s2)
[Ac1] lower [Ac2]
upper [Ac3]
lower [Ac4]
upper [Ac5]
lower [Ac6]
upper [Ac7]
1 0.19g 0.24g 0.26g 0.26g 0.29g 0.31g 0.26g 0.33g 2 0.35g 0.51g 0.53g 0.55g 0.55g 0.58g 0.57g 0.71g 3 0.56g 0.72g 0.73g 0.80g 0.74g 0.81g 0.81g 1.05g 4 0.68g 0.73g 0.70g 0.73g 0.72g 0.75g 0.71g 0.92g 5 0.84g 1.27g 1.13g 1.65g 1.16g 1.59g 1.52g 1.69g 6 1.02g --- --- --- 1.29g 1.38g --- --- 7 1.12g --- --- --- --- --- --- ---
Table 6.4. Test 3: Ratios of the measured accelerations at all different positions on the model
to the measured accelerations on the shaking table (acceleration amplication factors). Acceleration of table (m/s2)
Amplication factor of in-plane wall, corner
Amplication factor of in-plane wall, free end
Amplication factor of transvserse wall
Earth- quake ref.
Calculated acceleration of table (m/s2)
[Ac1] lower [Ac2]
upper [Ac3]
lower [Ac4]
upper [Ac5]
lower [Ac6]
upper [Ac7]
1 0.19g 0.24g 1.08 1.10 1.20 1.28 1.08 1.37 2 0.35g 0.51g 1.03 1.07 1.07 1.12 1.10 1.38 3 0.56g 0.72g 1.01 1.12 1.03 1.13 1.13 1.45 4 0.68g 0.73g 0.97 1.00 0.99 1.03 0.98 1.26 5 0.84g 1.27g 0.88 1.29 0.91 1.25 1.19 1.33 6 1.02g --- --- --- --- --- --- --- 7 1.12g --- --- --- --- --- --- ---
Mean amplification factor: 0.99 1.12 1.04 1.16 1.10 1.36
It was noted for Test 3 and Test 4, and also found in Tests 5-7, that the recorded mean peak
acceleration of the shaking table was on average 1.7 times the calculated acceleration. It is
likely that this is because the frequencies used in this project are all 5Hz or below, a frequency
range in which this model of accelerometer is known to suffer inaccuracy and noise. Noise was
still present in the acceleration data even after filtering using MatLab, so all data presented here
1 All MatLab code used throughout this project was written by Ulas Cilingir, PhD student at Cambridge University, to whom the author is very grateful for his assistance.
Section 6: Behaviour of ‘L’-shaped corner wall joints
32
is unfiltered. However, since the difference between the recorded acceleration and calculated
acceleration was consistent it was concluded that the results were still valid for comparisons
between different wall positions, earthquakes and tests.
For the remainder of the report, only the data from the three key positions of the shaking table
(accelerometer ‘Ac1’), the top of the in-plane wall at the corner (accelerometer ‘Ac3’) and the
top of the transverse wall (accelerometer ‘Ac7’) is presented. This is because the main points
of interest are the possibility of any differences in behaviour between the transverse wall and
the in-plane wall at the corner, and any differences between points on the walls and the table
(‘ground’) acceleration. The accelerometers nearer the tops of the walls are more likely to
show such differences if they exist, as in the example shown for Test 3 in Table 6.4.
Table 6.5. Test 4: Summary of peak measured accelerations and acceleration amplication
factors of key positions. Earth- quake ref.
Calculated acceleration of table (m/s2)
Acceleration of table [Ac1] (m/s2)
Acceleration of in-plane wall, corner, upper [Ac3] (m/s2)
Amplication factor of in-plane wall, corner, upper
Acceleration of transverse wall, upper [Ac7] (m/s2)
Amplication factor of transverse wall, upper
A 0.29g 0.34g 0.41g 1.20 0.79g 2.31 B 0.47g 0.61g --- --- C 0.67g --- --- --- D 0.85g --- --- ---
[ ‘---’ indicates accelerometer data was not useable due to collapse of part or all of the model ]
6.2.1. Observations on accelerometer data
It can be noted in Test 3 (from Table 6.4) that for each earthquake, the acceleration of the top
of the in-plane wall is on average 1.12 to 1.16 times the acceleration of the shaking table, and
the acceleration of the top of the transverse wall is on average 1.36 times the table acceleration.
There is amplification even for the first two earthquakes, where no oscillation or damage was
visible. This suggests that there may be differences in behaviour of different parts of the model
even when there is no visible change.
In Test 4, collapse of parts of the model prevented useable data being obtained after the first
earthquake so comparisons can not be made with any confidence.
Section 6: Behaviour of ‘L’-shaped corner wall joints
33
6.3. Summary of results for ‘L’-shaped corner wall joints
• The observed failure mechanisms in Test 3 and Test 4 (overturning of the transverse wall
initiated by vertical cracking at the joint and collapse of the in-plane wall initiated by
diagonal failure planes) were typical real-life collapse mechanisms.
• The high failure acceleration of Test 3 (model dried for 4 days) and the observation of
suction remaining in the mortar after testing demonstrated the need to achieve a drier mortar
which would fail at an acceleration not too near the acceleration capacity of the shaking
table.
• Test 4 (model dried for 37 days) experienced a very similar failure acceleration compared
to an equivalent test by Tetley (2006) which used pre-dried sand. This indicated that the
level of drying necessary to avoid suction in the mortar can be achieved either by pre-drying
the sand or by leaving the whole model to dry for an extended period. The moisture content
measurement of Test 4 provided a benchmark to check that subsequent models had a similar
dryness whichever drying method was used.
• The use of a wooden support for the transverse wall was successful in preventing rotation of
the models.
• The accelerations measured in Test 3 and Test 4 demonstrated that there may be amplified
accelerations of the transverse wall compared to the ground (shaking table) acceleration,
even during earthquakes when there is no visible change in the behaviour of the model.
34
Section 7: Behaviour of ‘L’-shaped corner wall joints with improvements
In Test 4, discussed in Section 6, an unimproved corner wall joint model was developed with:
performance similar to previous research using the same shaking table; failure at accelerations
well below the shaking table capacity; and realistic collapse mechanisms. Tests 5-7 were
performed on three further models with different structural improvements which could be
compared to the benchmark of Test 4. Dimensions of the walls and material composition were
equivalent to the unimproved joint in Test 4 (‘L’-shaped joints, with 40% building sand in the
mortar, but with sand pre-dried and the model dried for 4 days). The following tests were
performed (all accelerations referred to in the text are calculated accelerations of the table):
Test 5 (26/01/07):
• Reinforced with external vertical bamboo on both sides of wall at 100mm spacing and
external horizontal wire at 180mm spacing. Bamboo joined through wall with wire.
• First cracks observed and large oscillation of transverse wall at 0.45g.
• Internal collapse of transverse wall at 0.89g; large cracking of in-plane wall at 0.89g.
Test 6 (09/02/07):
• Reinforced with internal vertical bamboo at 100mm spacing and internal horizontal
wire at 180mm spacing.
• First cracks and gaps opening observed and large oscillation of transverse wall at 0.44g.
• Failure of transverse wall by overturning at 0.89g; large cracking of in-plane wall and
toppling due to collapse of transverse wall pulling the in-plane wall over at 0.89g.
Test 7 (23/02/07):
• Reinforced with external vertical bamboo on both sides of wall at 50mm spacing and
external horizontal wire at 180mm spacing. Bamboo joined through wall with wire.
• First cracks observed and large oscillation of transverse wall at 0.44g.
• Further gaps opened but most material still confined at 0.89g without catastrophic collapse.
7.1. Visual observations
The same system of observing, recording and categorising damage was used as for previous
tests, as explained in Section 5.1. The high speed video camera was unavailable for Test 7.
Section 7: Behaviour of ‘L’-shaped corner joints with improvements
35
Table 7.1. Summary of Test 5: Reinforced with external vertical bamboo at 100mm spacing. Earth- quake ref.
Maximum displacement of table (mm)
Frequency (Hz)
Calculated acceleration of table (m/s2)
Observations Damage category
A 2.55 4.85 0.24g Some thin cracks 1
B 4.75 4.85 0.45g Large oscillation of transverse wall; large, deep cracks in transverse wall
3
C 7.00 4.85 0.66g Large gap opens near base of transverse wall 3-4
D 9.40 4.85 0.89g Internal collapse of transverse wall; large cracks in in-plane wall
4
Table 7.2. Summary of Test 6: Reinforced with internal vertical bamboo at 100mm spacing. Earth- quake ref.
Maximum displacement of table (mm)
Frequency (Hz)
Calculated acceleration of table (m/s2)
Observations Damage category
A 2.70 4.85 0.26g No visible damage 0
B 4.70 4.85 0.44g Large oscillation of transverse wall; large gap near base of transverse wall
3-4
C 6.85 4.85 0.65g Further gaps in parts of transverse wall 3-4
D 9.35 4.85 0.89g
Failure of transverse wall by overturning; in-plane wall cracks, then topples due to transverse wall movement
5
Table 7.3. Summary of Test 7: Reinforced with external vertical bamboo at 50mm spacing. Earth- quake ref.
Maximum displacement of table (mm)
Frequency (Hz)
Calculated acceleration of table (m/s2)
Observations Damage category
A 2.70* 4.85 0.26g* No visible damage 0
B 4.70* 4.85 0.44g* Large oscillation of transverse wall; large, deep cracks in transverse wall
3
C 6.85* 4.85 0.65g* Gaps open near base of transverse wall 3
D 9.35* 4.85 0.89g* Further gaps open but most material still confined; no catastrophic collapse
3-4
*A calibration error in the LVDT data for Test 7 made the displacement data not useable. The
measured table accelerations for Test 7 were very close to those measured in Tests 5 and 6 so
for the purposes of comparison of damage between earthquakes, the displacements and
calculated accelerations of the table in Test 7 were assumed to be the same as for Test 6.
Section 7: Behaviour of ‘L’-shaped corner joints with improvements
36
Figure 7.1 shows the stages of damage experienced by the model in Test 5 (reinforced with
external vertical bamboo on both sides of wall at 100mm spacing and external horizontal wire
at 180mm spacing).
(a) (b)
Figure 7.1(a). Test 5, during Earthquake B: Deep cracks (highlighted) in transverse wall at
0.44g.
Figure 7.1(b). Test 5, after Earthquake C: Large gap near base of transverse wall at 0.66g
(Note: at the corner of the wall, small bamboo pieces are used to prevent the wire cutting into
the mortar).
(c) (d)
Figure 7.1(c). Test 5, after Earthquake D: Internal collapse of transverse wall at 0.89g.
Figure 7.1(d). Test 5, after Earthquake D: Large cracks have formed in the in-plane wall but
the wall remains confined within the bamboo at 0.89g.
Section 7: Behaviour of ‘L’-shaped corner joints with improvements
37
Figure 7.2 shows the stages of damage experienced by the model in Test 6 (reinforced with
internal vertical bamboo at 100mm spacing and internal horizontal wire at 180mm spacing).
(a) (b)
Figure 7.2(a). Test 6, after Earthquake B: Large gap near base of transverse wall at 0.44g.
Figure 7.2(b). Test 6, after Earthquake C: Further gap near top of transverse wall at 0.65g,
and accelerometers Ac2 and Ac3 have fallen off.
(c) (d)
Figure 7.2(c). Test 6, during Earthquake D: Vertical cracking at corner (highlighted) and
overturning of transverse wall (indicated by arrow) at 0.89g.
Figure 7.2(d). Test 6, during Earthquake D: Cracking of in-plane wall, then toppling of in-
plane wall (indicated by arrow), due to the effect of transverse wall still attached to in-plane
wall via the bamboo and wire at the corner, at 0.89g.
Overturning Toppling
Section 7: Behaviour of ‘L’-shaped corner joints with improvements
38
Figure 7.3 shows the stages of damage experienced by the model in Test 7 (reinforced with
external vertical bamboo on both sides of wall at 50mm spacing and external horizontal wire at
180mm spacing).
(a) (b)
Figure 7.3(a). Test 7, after Earthquake B: Cracks opening on top of transverse wall at 0.44g.
Figure 7.3(b). Test 7, after Earthquake C: Further cracks opening and sections of mortar
dislodging in transverse wall at 0.65g.
(c) (d)
Figure 7.3(c). Test 6, during Earthquake D: Deep cracking (highlighted) and large oscillation
of in-plane wall (indicated by arrow) at 0.89g, but almost all material confined.
Figure 7.3(d). Test 6, after Earthquake D: Large gap open at top of transverse wall at 0.89g.
Oscillation and cracking
Section 7: Behaviour of ‘L’-shaped corner joints with improvements
39
7.2. Moisture content data recorded
To ensure that the method of pre-drying the sand was equivalent to the extended drying period
used in Test 4, moisture contents were measured of samples from each ‘batch’ of mortar made.
For practical reasons, the mortar for each model was made in 3 batches over a period of 5 days.
Therefore different layers of the wall dried for different time periods before testing. As Table
7.4 shows, the first two batches had similar moisture contents after testing to each other and to
the model in Test 4 (0.80%, as discussed in Section 6.1). However, the 3rd batch (the top 1/3 of
each wall) in each test had a higher moisture content. This may have caused greater ductility
and/or strength (due to suction, as discussed in Section 5.1) in the top part of the wall. It is not
possible to assess the significance of this effect in the performance of the models but future
research should aim for identical moisture content throughout each model tested to ensure that
material properties are consistent throughout the model.
Table 7.4. Moisture content measurements for mortar batches in Tests 5-7
Batch of mortar, drying time until testing, and moisture content (%) measured 3 days after testing
Test Structural improvements
1st batch, 7-8 days 2nd batch, 6-7 days 3rd batch, 4 days
5 External bamboo (wide spacing) Not measured 0.94 2.10
6 Internal bamboo 0.95 0.95 2.35
7 External bamboo (narrow spacing) 1.28 0.91 3.22
7.3. Instrumentation data recorded
For Tests 5-7, accelerometers were placed at the positions shown on Figure 3.1, and MatLab
was used to produce graphs comparing the accelerations at these positions as for the example
in Section 6.2, Figure 6.5. As in Section 6.2, MatLab was then used to find the mean of the
peak accelerations of the different positions during each earthquake. Only the data from the
three key positions is presented. These results are summarised in Tables 7.5 to 7.7. As
discussed in Section 6.2, it was noted that the recorded acceleration of the shaking table was on
average 1.7 times the calculated acceleration, but it was concluded that the results were still
valid for comparisons between different wall positions, earthquakes and tests.
Section 7: Behaviour of ‘L’-shaped corner joints with improvements
40
Table 7.5. Test 5: Reinforced with external vertical bamboo at 100mm spacing. Summary of
peak measured accelerations and acceleration amplication factors of key positions. Earth- quake ref.
Calculated acceleration of table (m/s2)
Acceleration of table [Ac1] (m/s2)
Acceleration of in-plane wall, corner, upper [Ac3] (m/s2)
Amplication factor of in-plane wall, corner, upper
Acceleration of transverse wall, upper [Ac7] (m/s2)
Amplication factor of transverse wall, upper
A 0.24g 0.30g 0.32g 1.06 0.46g 1.53 B 0.45g 0.53g 0.59g 1.10 0.93g 1.73 C 0.66g 0.76g 0.85g 1.12 0.81g 1.07 D 0.89g 1.15g --- ---
Table 7.6. Test 6: Reinforced with internal vertical bamboo at 100mm spacing. Summary of
peak measured accelerations and acceleration amplication factors of key positions. Earth- quake ref.
Calculated acceleration of table (m/s2)
Acceleration of table [Ac1] (m/s2)
Acceleration of in-plane wall, corner, upper [Ac3] (m/s2)
Amplication factor of in-plane wall, corner, upper
Acceleration of transverse wall, upper [Ac7] (m/s2)
Amplication factor of transverse wall, upper
A 0.26g 0.30g 0.16g* 0.55* 0.46g 1.52 B 0.44g 0.59g 0.42g* 0.71* 1.28g 2.17 C 0.65g 0.85g --- 1.25g 1.48 D 0.89g --- --- ---
*There appears to be an error in the data from Ac3 for Test 6 since the values are significantly
lower than the table acceleration.
Table 7.7. Test 7: Reinforced with external vertical bamboo at 50mm spacing. Summary of
peak measured accelerations and acceleration amplication factors of key positions. Earth- quake ref.
Calculated acceleration of table (m/s2)
Acceleration of table [Ac1] (m/s2)
Acceleration of in-plane wall, corner, upper [Ac3] (m/s2)
Amplication factor of in-plane wall, corner, upper
Acceleration of transverse wall, upper [Ac7] (m/s2)
Amplication factor of transverse wall, upper
A 0.26g 0.32 0.28 0.87 0.42 1.28 B 0.44g 0.65 0.68 1.05 1.02 1.57 C 0.65g 0.87 0.81 0.94 --- D 0.89g 1.16 1.67 1.43 ---
[ ‘---’ indicates the accelerometer data was not useable due to collapse of part of the model ]
Section 7: Behaviour of ‘L’-shaped corner joints with improvements
41
7.3.1. Observations on accelerometer data
It appears from the accelerometer data that in each of Tests 5-7 the top of the transverse wall
experiences greater acceleration than the shaking table and the top of the in-plane wall, even
when there is no visible oscillation of the transverse wall. It is also noted that this amplification
of accelerations between the table and the transverse wall seems to drop between Earthquakes
B and C in Tests 5 and 6 (the transverse wall data for Earthquake C in Test 7 was not useable
due to collapse of part of the wall). This may be related to the cracking behaviour of each
model: in each of Tests 5-7 the first cracking and oscillation of the transverse wall occurs
during Earthquake B. This may explain the higher accelerations experienced during this
earthquake.
In Tests 5 and 7, the acceleration of the top of the in-plane wall does not seem to experience
such an amplification factor except in Earthquake D of Test 7.
The data for in-plane wall acceleration in Test 6 appears to have a recording error since the
values are significantly lower than the table acceleration.
As discussed in Section 3.3, the accelerometers experience noise and inaccuracy in the
frequency range used here which means that detailed comparison of amplification factors is not
possible. The general observations of the amplified acceleration of the transverse wall and the
possible effects of cracking behaviour are still reasonable but it is not possible to draw more
detailed conclusions.
7.4. FFT analysis of acceleration data
It was noted from the acceleration graphs of Tests 5-7 that some of the acceleration data seems
to show evidence of frequency components other than the fundamental input frequency of the
shaking table (4.85Hz). To test this observation, Fast Fourier Transforms (FFTs) were
produced for each set of acceleration data using MatLab, to determine the relative magnitudes
of the signal content from the acceleration data at each nth harmonic frequency. An example of
the graphs produced by the FFT is shown in Figure 7.4.
Section 7: Behaviour of ‘L’-shaped corner joints with improvements
42
0 5 10 15 20 25 300
1000
2000
3000
4000
5000
frequency, Hz
mag
nitu
de, g
/Hz
0 5 10 15 20 25 300
1000
2000
3000
4000
5000
frequency, Hz
mag
nitu
de, g
/Hz
0 5 10 15 20 25 300
1000
2000
3000
4000
5000
frequency, Hz
mag
nitu
de, g
/Hz
Figure 7.4(a). FFT of acceleration of shaking table (Ac1) during Test 5, Earthquake A. Figure 7.4(b). FFT of acceleration of top of in-plane wall (Ac3) during Test 5, Earthquake A. Figure 7.4(c). FFT of acceleration of top of transverse wall (Ac7) during Test 5, Earthquake A.
These graphs were produced for all earthquakes in Tests 5-7. The graphs were then used to find
the ratio of the magnitude of the signal content from the acceleration data at each nth harmonic
frequency to the magnitude of the signal content from the data at the fundamental input
frequency (4.85Hz), for each accelerometer position in each earthquake. It was assumed that
the energy in the signal content was proportional to the magnitude of the signal content
because the width of each ‘spike’ on the FFT graphs is narrow enough to assume that the
height of each spike is proportional to its area. Tables 7.8 to 7.10 show these ratios.
(a)
(b)
(c)
Section 7: Behaviour of ‘L’-shaped corner joints with improvements
43
Table 7.8. Test 5: Reinforced with external vertical bamboo at 100mm spacing. Comparison of
the signal content at each nth harmonic frequency. Ratio of the magnitude of the signal content from the acceleration data at each nth harmonic frequency to the signal content from acceleration data at the fundamental input frequency (4.85Hz) Table [Ac1] Top of in-plane wall [Ac3] Top of transverse wall [Ac7]
Earth- quake ref.
3rd 2nd 3rd 4th 2nd 3rd 4th 5th A 0.03 0 0.08 0 0.09 0.32 0.04 0.24B 0.09 0 0.14 0 0 0.01 0 0C 0.05 0 0.05 0 0 0 0 0D 0.01 --- ---
Table 7.9. Test 6: Reinforced with internal vertical bamboo at 100mm spacing. Comparison of
the signal content at each nth harmonic frequency. Ratio of the magnitude of the signal content from the acceleration data at each nth harmonic frequency to the signal content from acceleration data at the fundamental input frequency (4.85Hz) Table [Ac1] Top of in-plane wall [Ac3] Top of transverse wall [Ac7]
Earth- quake ref.
3rd 2nd 3rd 4th 2nd 3rd 4th 5th A 0.06 0 0.45 0 0 0.65 0.07 0.15B 0.14 0 0.64 0 0 0.05 0 0C 0.04 --- 0 0 0 0.16D 0 --- ---
Table 7.10. Test 7: Reinforced with external vertical bamboo at 50mm spacing. Comparison of
the signal content at each nth harmonic frequency. Ratio of the magnitude of the signal content from the acceleration data at each nth harmonic frequency to the signal content from acceleration data at the fundamental input frequency (4.85Hz) Table [Ac1] Top of in-plane wall [Ac3] Top of transverse wall [Ac7]
Earth- quake ref.
3rd 2nd 3rd 4th 2nd 3rd 4th 5th A 0.10 0 0.33 0 0 0.42 0.05 0.15B 0.05 0.07 0.03 0 0.18 0.06 0 0C 0.05 0.07 0.04 0 --- D 0.03 0.14 0.05 0.07 ---
Note: Values are only displayed for harmonics of non-zero content during any earthquake.
7.4.1. Observations on FFT analysis
It is clear that for Tests 5-7, apart from the fundamental input frequency, the 3rd harmonic
frequency is the most significant contribution to the magnitude of the signal content of the
acceleration data. This suggests that the 3rd harmonic frequency could also be a harmonic
frequency of the natural frequency of the models tested. The 3rd harmonic frequency seems to
have most effect during Earthquake A, the lowest intensity earthquake.
Section 7: Behaviour of ‘L’-shaped corner joints with improvements
44
To test the possibility of the models having a natural frequency close to the 3rd harmonic
frequency of testing, the FFT results were further analysed for each earthquake to compare the
magnitude of the signal contents at the 1st and 3rd harmonics on the in-plane wall and transverse
wall to the signal contents at the 1st and 3rd harmonics measured on the shaking table. If there
was consistently greater amplification of the signal content at one harmonic than the other then
this would suggest that the models had a natural frequency nearer to that harmonic. These
results are shown in Tables 7.11 to 7.13.
Table 7.11. Test 5: Reinforced with external vertical bamboo at 100mm spacing. Comparison
of the amplification of signal content between the shaking table and the model. Ratio of the magnitude of the signal content from the acceleration data from positions on the model to the signal content from the data from the shaking table, at the 1st and 3rd harmonic frequency
Top of in-plane wall [Ac3] Top of transverse wall [Ac7]
Earth-quake ref.
1st 3rd 1st 3rd A 1.1 2.8 1.3 13.3 B 1.0 1.5 2.7 0.2 C 1.2 1.2 1.1 0.1 D --- ---
Table 7.12. Test 6: Reinforced with internal vertical bamboo at 100mm spacing. Comparison
of the amplification of signal content between the shaking table and the model. Ratio of the magnitude of the signal content from the acceleration data from positions on the model to the signal content from the data from the shaking table, at the 1st and 3rd harmonic frequency
Top of in-plane wall [Ac3] Top of transverse wall [Ac7]
Earth-quake ref.
1st 3rd 1st 3rd A 0.2 1.6 1.1 10.8 B 0.1 0.5 4.9 1.9 C --- 1.8 7.1 D --- ---
Table 7.13. Test 7: Reinforced with external vertical bamboo at 50mm spacing. Comparison of
the amplification of signal content between the shaking table and the model. Ratio of the magnitude of the signal content from the acceleration data from positions on the model to the signal content from the data from the shaking table, at the 1st and 3rd harmonic frequency
Top of in-plane wall [Ac3] Top of transverse wall [Ac7]
Earth-quake ref.
1st 3rd 1st 3rd A 0.7 2.4 0.9 4.0 B 1.1 0.7 2.2 2.8 C 0.8 0.6 --- D 1.0 2.1 ---
Tables 7.11 to 7.13 show that there is no clear pattern in the amplifications of signal content
between the shaking table and the model at the 1st and 3rd harmonic frequencies. Therefore no
further conclusions can be drawn about the natural frequencies of the models.
Section 7: Behaviour of ‘L’-shaped corner joints with improvements
45
7.5. Summary of results for ‘L’-shaped corner wall joints with improvements
• In this project the key performance measure is the capacity to resist severe structural damage
or collapse (damage category 4 or higher, equivalent to collapse of the transverse wall).
Under this measure, all three structural improvements demonstrated significantly better
performance than the benchmark unimproved corner joint in Test 4:
• In Test 4 (unimproved), the transverse wall failed during Earthquake B.
• In Test 5 (external bamboo, wide spacing), the transverse wall failed in Earthquake D,
at an acceleration about 2 times the collapse acceleration of the unimproved joint. In tests
on a similar method by Dowling (2006), the wall with external bamboo experienced its
greatest damage at an intensity about 1.7 times greater than an unreinforced wall.
Dowling measured intensity in terms of displacement not acceleration, but since in this
project displacement and acceleration were proportional, these results seem comparable.
• In Test 6 (internal bamboo), the transverse wall failed during Earthquake D, at an
acceleration about 2 times the collapse acceleration of the unimproved joint. This differs
to the most recent research by Dowling (2006), which indicated that internal bamboo
reinforcement could in fact cause extra weakness due to discontinuity in the adobe. This
problem was not evident here because gravel was used in all tests, which could be built
closely around the bamboo. Dowling noted difficulties in using adobe blocks with semi-
circles cut out for the internal bamboo tests, which caused the discontinuity problem.
• In Test 7 (external bamboo, narrow spacing), the transverse wall was damaged but
still contained during Earthquake D, the best performance of all models tested.
• All three improvement methods added to the ductility of the model. The external bamboo
also had a significant confinement effect on the adobe.
• The use of a wooden support for the transverse wall seemed successful in preventing rotation
of the models. However, in Test 7 it was noted that the transverse wall struck the support as
it oscillated. This repeated impact may have had an effect on performance which could be
analogous to the effect of an internal shear wall inadequately joined to the transverse wall.
• The accelerations measured in Tests 5-7 demonstrated that there may be amplified
accelerations of the transverse wall compared to the ground (shaking table) acceleration,
even during earthquakes when there is no visible change in the behaviour of the model.
• FFT analysis of the acceleration data showed that there were contributions to the signal
content from harmonics beyond the fundamental input frequency, most notably the 3rd
harmonic frequency. The significance of this to the results is unknown but does not appear to
invalidate conclusions made about the performance of different improvements.
46
Section 8: Conclusions
8.1. Conclusions related to simulating adobe buildings using scale models
1. Models of in-plane walls and ‘L’-shaped corner wall joints at 1:5 scale tested on a 1-g
shaking table can exhibit collapse mechanisms similar to those found in real life, provided
a suitable material composition is used to model the adobe walls.
2. A suitable material composition can use gravel to simulate adobe blocks if it is not feasible
to construct individual adobe bricks. A suitable composition for the model size and shaking
table used was 60% gravel of size 30mm - 80mm and 40% mortar by weight. Mortar was
made from 60% Kaolin clay at w = 0.45 and 40% building sand by weight. This result
supports the findings of Tetley and Madabhushi (2007). It is also necessary either to leave
the completed model to dry for an extended period of time (up to 30 or more days) or to
pre-dry the sand in an oven to ensure the mortar has fully dried before testing so that there
is no suction in the mortar. If the model is made in stages, moisture content measurements
should be taken of mortar samples to ensure that the whole model has a consistent moisture
content.
3. For model walls which represent only partial sections of adobe buildings, the boundary
conditions should be artificially adjusted to simulate the effect of the continuity present in a
complete building. In particular, to simulate a continuous transverse wall in an ‘L’-shaped
model corner joint, support should be provided to the end of the transverse wall to prevent
rotation of the model. Care must be taken to avoid the support acting as an inadequately
attached shear wall to the transverse wall which can cause repeated impacts on the
transverse wall during earthquake loading.
4. Accelerometer measurements from different positions on models tested show that parts of
model walls can exhibit amplified accelerations compared to the input shaking table
(‘ground’) acceleration, even when there is no visible damage or difference in
displacement. This suggests that quantitative data is important in making detailed
observations about the behaviour of model adobe walls and that future tests should not rely
solely on qualitative observations.
Section 8: Conclusions
47
8.2. Conclusions related to structural improvements of adobe buildings
1. External vertical bamboo reinforcement, joined through the walls of the building and tied
externally with horizontal wire, improves the seismic resistance of the building by
increasing the ductility of the walls and providing some confinement for the adobe blocks.
2. In the 1:5 scale tests on ‘L’-shaped corner joints, the capacity to resist severe structural
damage (measured as the collapse acceleration of the transverse wall) was increased by at
least a factor of 2 using external bamboo reinforcement. The effect is of a similar
magnitude to that observed in recent research on a similar technique by Dowling (2006).
Further research would be needed to assess if this factor could be expected in real life if
whole buildings were reinforced in this way.
3. The extra seismic resistance due to confinement of the adobe is dependent on the relative
sizes of the adobe blocks and the spacing of the bamboo and wire. In real life it is likely
that there would be a compromise between the amount of confinement, the materials
available, and the number of holes which could be drilled through the walls safely.
4. A combination of internal vertical bamboo reinforcement and internal horizontal wire
improves the seismic resistance of the building by increasing the ductility of the walls and
increasing the connectivity at corner joints.
5. In the 1:5 scale tests on ‘L’-shaped corner joints, the collapse acceleration of the transverse
wall was increased by a factor of 2 using internal bamboo reinforcement. Further research
would be needed to assess if this factor could be expected in real life if whole buildings
were reinforced in this way. There may be differences in the effectiveness of this method
depending on whether irregular blocks are used which can fit around the bamboo (as in this
project) or regular blocks which must be cut to fit the bamboo in and may produce
weaknesses due to discontinuity problems in the adobe (as tested by Dowling, 2006).
6. Bamboo and galvanised wire are materials already used to reinforce adobe buildings and
have sufficiently low cost and wide availability to be suitable for widespread use.
7. Internal reinforcement is only suitable for newly-built constructions. External bamboo and
wire reinforcement has the potential to be used as a retrofitting method but further research
Section 8: Conclusions
48
is needed in the field to ensure that the technique has widespread suitability in terms of
community acceptance and ease of use. Current field research related to this topic is
ongoing in locations including El Salvador and Pakistan by Dowling and others. The author
will work with the Salvadorean Foundation for Reconstruction and Development in El
Salvador to perform a structural survey of adobe houses in order to further assess the
appropriateness of external bamboo and wire reinforcement for strengthening existing
buildings.
8. Dissemination of results in formats useful to other researchers and practitioners is essential.
The results of this project will be shared via the recently set up World Adobe Forum
(www.worldadobeforum.net).
8.3. Recommendations for future research
1. Research on small-scale models should acknowledge the limitations of such testing and
investigate potential new ways of achieving results more directly relevant to real life.
2. Further research on model walls which represent only partial sections of adobe buildings
should pay particular attention to accurately simulating appropriate boundary conditions.
3. Further research using 1:5 scale models could consider the feasibility of constructing actual
adobe blocks instead of using gravel to simulate bricks. If actual adobe blocks are used,
research could include static testing of adobe prisms as well as dynamic testing of models.
4. Research could also consider the possibility of including recognised earthquake resistant
features such as ring beams into models and investigating their interaction with other
strengthening techniques.
5. All laboratory research should consider the latest updates from field investigations
regarding the suitability of different materials and techniques and should use appropriate
methods of feeding the research results back into real-life practice.
6. Seismic resistance methods investigated should have potential for retrofitting existing
structures as well as being incorporated into new buildings.
49
Section 9: References Blondet, M., Garcia, M. and Brzev, S. (2003), Earthquake-Resistant Construction of Adobe
Buildings: A Tutorial, retrieved from www.world-housing.net. Blondet, M., Torrealva, D., Villa-Garcia, G., Ginocchio, F and Madueño, I. (2005), Using
industrial materials for the construction of safe adobe houses in seismic areas, Proceedings of SismoAdobe 2005, 16-19 May 2005, PUCP, Lima.
De Sensi, B. (2003), Terracruda, La Diffusione dell’Architettura di Terra, retrieved from www.terracruda.com/architetturadiffusione.htm.
Dowling, D.M. (2006), Seismic strengthening of adobe-mudbrick houses, Faculty of Engineering, University of Technology, Sydney.
Flores, L.E., Pacheco, M.A. and Reyes, C. (2001), Algunos estudios sobre el comportamiento y rehabilitación de la vivienda rural de adobe, CENAPRED México, IEG/03/01.
Houben, H. and Guilland, H. (1994), Earth Construction: a comprehensive guide, ITDG Publishing, London.
IAEE (2004), Guidelines for Earthquake-Resistant Non-Engineered Construction, International Association for Earthquake Engineering, Tokyo, Japan.
Janssen, J.J.A. (1995), Building with Bamboo: A Handbook, ITDG Publishing, London. Loaiza, C., Blondet, M., Ottazzi, G. (2002), World Housing Encyclopedia Report: Peru Adobe
House, retrieved from www.world-housing.net. Malton, J. (2005), Earthquake Resistance of Adobe Buildings, Cambridge University
Engineering Department. Middleton, G.F. (1987), Bulletin 5: Earth-Wall Construction, Commonwealth of Australia. Nienhuys, S. (1999), Galvanised Wire-Mesh Wall Reinforcement Methodology, Aga Khan
Planning and Building Services, Pakistan. Tetley, R. (2006), Seismic Behaviour of Adobe Buildings, Cambridge University Engineering
Department. Tetley, R. and Madabhushi, S.P.G. (2007), Vulnerability of Adobe Buildings under Earthquake
Loading, Proceedings of forthcoming 4th International Conference on Earthquake Geotechnical Engineering, 25-28 June 2007, Thessaloniki, Greece.
Tolles, E.L. and Krawinkler, H. (1990), Seismic studies on small-scale models of adobe houses, John Blume Earthquake Engineering Center, Stanford University.
Tolles, E.L., Kimbro, E.E., Webster, F.A. and Ginell, W.S. (2000), Seismic stabilization of historic adobe structures, The Getty Conservation Institute, Los Angeles, California.
USGS, U.S. Geological Survey (2006), Earthquake Hazards Program: Historic Earthquakes, USGS, retrieved from http://earthquake.usgs.gov/eqcenter/historic_eqs.php.
Yamin, L.E., Phillips, C.A., Reyes, J.C., Rivero, S. and Ruiz, D. (2005) Comportamiento sísmico y alternativas de rehabilitación de edificaciones en adobe, Proceedings of SismoAdobe 2005, 16-19 May 2005, PUCP, Lima.
Zegarra, L., Quiun, D., San Bartolomé, A. and Giesecke, A. (1999), Reforzamiento de viviendas de adobe existents, Pontificia Universidad Católica del Perú.