Sei sulla pagina 1di 53

Prestressed Ceramics

By
Will Hawkins (Chu)

Fourth-year Undergraduate Project in Group D,


2012/2013

I hereby declare that, except where specifically indicated, the work submitted herein is my
own original work.
Signed: Date:
Will Hawkins
Will Hawkins, Churchill College 1
Prestressed Ceramics

Technical Abstract
During this project a number of prestressed ceramic structures were designed,
constructed and tested in order to investigate the effectiveness of prestressing in
improving the structural performance of common brittle materials. It was hoped that an
increase in strength and ductility could be achieved, overcoming the naturally low tensile
strength and low deformation of failure associated with ceramics.

An portable beam testing rig was designed and built, enabling testing to be carried out
quickly and with a high level of consistency. A pre-tensioned 'test' beam was constructed
using terracotta tiles, with the primary aim of assessing the feasibility of working with
and stressing aramid fibres for use as prestressing tendons. After successful fabrication,
test results showed good agreement with theoretical predictions and thus also confirmed
the accuracy of the completed testing rig.

Prestressing tendons of aramid fibre were used throughout the project, giving insight into
the behaviour of a non-typical tendon material and allowing the development of a novel
post-tensioning system. Multiple loops of aramid yarn were passed through structures
and secured at each end, before being extended by the separation of loading plates via a
pair of large screws. Hence no loading jack was required to apply the prestress. After
refinement of the initial concept, this system proved to be highly effective and paved the
way for the construction of a number of unique prestressed structures.

A beam consisting of four unbonded segments was constructed, which featured a


rectangular section and tendon at centroidal level. Each segment was made from two
pieces of porcelain tile with a central tendon duct, and the structure gained all stiffness
from the application of prestress. After a number of design iterations, a successful test
resulted in compressive failure of the tiles at the central segment interface. A theoretical
model was devised, which gave insight into the behaviour of the beam under bending
loads and the effectiveness of the prestress.
Will Hawkins, Churchill College 2
Prestressed Ceramics

After the successful construction and promising testing of a prototype, five prestressed
glass beams were produced incorporating a similar post-tensioning system. Variations in
tendon eccentricity, bond and prestressing force were investigated. The prestressing
forces were estimated from the crack reopening moments and section geometry, and
were consistently found to be within expected values. A considerable degree of ductility
was achieved by bonding the tendon to the glass. Theoretical cracked elastic analyses
were shown to accurately predict beam behaviour, subject to the limitations of the model.
The use of glass allowed excellent observation of cracking, and enabled an informed
assessment of cracked behaviour and failure modes.

Finally a prestressed ceramic truss structure was designed and built, incorporating the
same post-tensioning system and a number of fabrication techniques developed earlier in
the project. Weighing only 6.5 kg, the triangulated bridge was able to support a maximum
central point load of 3.5 tonnes over a span of 880 mm, and observation of the behaviour
suggested ways in which the design could potentially be improved upon. The bilinear
behaviour predicted by theory was observed in the measured results.

Some discrepancies were noted between the Young's Modulus values found from
preliminary tests and those estimated from measured uncracked stiffness gradients of
the structures. In some cases these could be partially explained by the simplifying
assumptions of the theoretical models, but they did also suggest some inaccuracies were
present in the results or the values assumed.

Since there was insufficient time to fully develop the concepts explored in this project, a
number of recommendations for continued investigations were made.
Will Hawkins, Churchill College 3
Prestressed Ceramics

Contents
Technical Abstract ............................................................................................................................................. 1

Nomenclature...................................................................................................................................................... 4

1. Introduction .................................................................................................................................................... 5

1.1 Motivation ................................................................................................................................................ 5

1.2 Aims ............................................................................................................................................................ 6

1.3 Previous Work ........................................................................................................................................ 6

2. Preliminary Work ......................................................................................................................................... 7

2.1 Materials ................................................................................................................................................... 7

2.2 Design & Construction of a Portable Testing Rig ................................................................... 10

2.3 Beam 1 - Pre-tensioned Test Beam.............................................................................................. 12

3. Developing a Post-tensioning System ................................................................................................ 14

3.1 Beam 2 Design ..................................................................................................................................... 14

3.2 Issues & Design Iterations............................................................................................................... 18

3.3 Theoretical Predictions .................................................................................................................... 21

3.4 Results, Discussion & Comparison with Theory ..................................................................... 23

4. Post-tensioned Glass Beams .................................................................................................................. 25

4.1 Beam 3 Design & Construction ...................................................................................................... 25

4.3 Investigated Parameters .................................................................................................................. 27

4.4 Theoretical Predictions .................................................................................................................... 28

4.5 Test Set-up and Procedure.............................................................................................................. 30

4.5 Results & Observations .................................................................................................................... 30

4.6 Discussion .............................................................................................................................................. 36

5. Post-tensioned Ceramic Truss .............................................................................................................. 40

5.1 Design Task ........................................................................................................................................... 40

5.2 Design Overview ................................................................................................................................. 40


Will Hawkins, Churchill College 4
Prestressed Ceramics

5.3 Test Set-up & Procedure .................................................................................................................. 43

5.4 Theoretical Predictions .................................................................................................................... 44

5.5. Results & Discussion ........................................................................................................................ 45

6. Conclusions .................................................................................................................................................. 47

6.1 Main Findings ....................................................................................................................................... 47

6.2 Recommendations for Further Work ......................................................................................... 48

References ......................................................................................................................................................... 50

Appendix A - Uncracked and Cracked Elastic Prestressed Beam Theory................................. 51

Appendix B - Theoretical Predictions of Truss Load-Deflection Behaviour ............................ 52

Appendix C - Risk Assessment Retrospective ..................................................................................... 52

Nomenclature
A Total effective area L2 Length of truss compression strut
Ac Area of ceramic Leff Effective length for an Euler column
At Area of tendon P Assumed prestressing force
At, eff Effective area of tendon Pi Initial prestressing force
b Total ceramic width T1 Tension in truss tension strut
d Depth to tendon T2 Compression in truss compression strut
Ec Ceramic Young's Modulus x Neutral axis depth
Et Tendon Young's Modulus Δεt Change in tendon strain
Fc Force in ceramic εc,max Maximum ceramic strain
F Total force applied to beam δ Deflection
h Total section depth θ Rotation angle between segments
I Transformed second moment of area σ Extreme fibre stress
K Tendon stiffness σb Maximum Bending Stress
l Length from beam support to load point κ Curvature
L Tendon gauge length Truss strut angle
L1 Length of truss tension strut Truss frame inclination

- All dimensions are in mm unless otherwise stated.


- Figures are not to scale
Will Hawkins, Churchill College 5
Prestressed Ceramics

1. Introduction
Prestressing is a technique commonly used to improve the structural performance of
materials with a low tensile strength. By applying compressive prestress the strength
characteristics of the material are effectively shifted, and tensile strains caused by the
applied loading can be accommodated while the bulk material remains in compression.
Furthermore, by varying the distribution of prestress within a structure its performance
can be tailored to match a specific loading requirement. Prestressing therefore provides a
range of opportunities for creating highly efficient structures using common building
materials. Prestressed concrete is the most common example of this.

Prestressing tendons must be stressed very highly in order to minimise losses due to
elastic shortening and creep. This presents a number of challenges concerning both the
application of prestress and its effective transfer to the bulk material. There are a variety
of ways of achieving this, which can be grouped into two categories; pre-tensioning,
where the bulk of the structure is formed around already stressed tendons, and post-
tensioning, where the tendons are stressed and anchored after the structure is fully
formed.

1.1 Motivation
Ceramics are non-metallic, inorganic solids covering a broad range of engineering
materials, from bricks and tiles commonly used in construction to high performance
technical ceramics such as tungsten carbide, one of the hardest and stiffest materials used
in industry. They exhibit good environmental and thermal resistance, and perform well
across a range of demanding engineering applications.

All ceramics behave in a brittle manner due to their low fracture toughness. Typically, a
ceramic’s compressive strength is ten times larger than its strength in tension. Governed
by material flaws, tensile failure is sudden, difficult to predict and usually catastrophic. By
applying large compressive forces through prestressing, these undesirable characteristics
can be overcome and a variety of new structural applications can be potentially unlocked.

One example of such an application is in gas turbine engines where the temperature of
combustion, and hence the efficiency, is limited by the thermomechanical properties of
Will Hawkins, Churchill College 6
Prestressed Ceramics

the turbine blade material. The melting temperature, corrosion resistance, stiffness and
density of technical ceramics such as silicon nitride compare favorably with the nickel
superalloys currently widely used, however safety concerns over catastrophic brittle
failures exist. These could be prevented through prestressing of the blades, eliminating
tensile stresses.

The components manufactured in this project use more commonly available ceramics.
New structural and architectural uses for these materials could be made possible through
prestressing, by increasing strength and enabling safe ductile failure modes. Safe, simple
and low cost prestressing methods are required if a practically viable technology is to be
developed.

1.2 Aims
 To investigate the effectiveness of prestress in improving the structural behaviour
of common brittle materials.
 To assess the validity of theories related to prestressed structures through
experimentation and analysis.
 To explore the possibilities of prestressed ceramic structures in terms of design,
fabrication and performance.

1.3 Previous Work


Previous projects investigating prestressed ceramic beams have been undertaken in the
department. A number of pre-tensioned beams were constructed using bathroom tiles
and glass, with steel cables as the tendon material (Causier, 2010). These beams were
constructed in a large stressing rig, and some difficulty was experienced in securing the
ends of the steel cables. Variations in prestressing force were investigated with the
tendons at centroidal level, and the strength and ductility of the beams was found to be
improved by prestressing.

Ceramic T-beams have also been constructed, using steel studding as the tension element
(Grant, 2012). Shear transfer between the web and flange required complex fabrication
methods, and problems were encountered with local crushing at loading plates and
failure of the studding (which has a much lower yield strength than cable). Again, a large
stressing rig was required to construct the beams.
Will Hawkins, Churchill College 7
Prestressed Ceramics

2. Preliminary Work

2.1 Materials

2.1.1 Ceramic Tiles & Glass


Three different brittle materials were primarily used in the structures throughout the
project. These included terracotta (white/red) as well as stronger porcelain (black/grey)
bathroom tiles, both approximately 10mm thick with a glossy top surface and rougher
textured bottom surface. These were easily shaped using a tile cutter. A number of
components were also constructed using 6mm thick annealed soda-lime glass. This
material has the advantage of being available in large sizes, although shaping is more
difficult.

Using materials commonly found in construction, rather than high specification technical
ceramics, presented a number of advantages:

 A smaller and hence more easily achievable prestressing force was required to
alter the behaviour of weaker materials.
 Potentially more widespread applications could be demonstrated by using
common materials.
 Materials could be quickly and inexpensively obtained, and shaped using standard
equipment.

Identical tests were carried out on samples of each material in order to determine key
properties for use in beams, with results shown in Table 2.1.

Young's Modulus Bending Strength


(GPa) Mean σb (MPa) Std. Deviation* (MPa)
Terracotta Tile 18.54 17.74 0.67
Porcelain Tile 72.83 76.27 2.80
Annealed Glass 59.50 66.77 7.90
Table 2.1 - Summary of Preliminary Testing Results
*approximate value based on data from three tests only.

The Young's Modulus was found from a three-point bending test. Knowing the second
moment of area I and the beam length L, the Young's Modulus was calculated from the
force-deflection gradient using the following formula:
Will Hawkins, Churchill College 8
Prestressed Ceramics

A value of the bending strength was also determined from four-point bending tests, by
assuming a linear stress distribution in order to calculate the extreme fibre stress at
failure. For each of the tiles, the glossy surface was placed on the tensile side, as shown
for the terracotta tile in Figure 2.1. Since tensile failure of ceramic materials is brittle and
governed by fracture mechanics, it is
determined by the maximum flaw size and
can hence be expected to vary between
samples. To improve reliability and assess
variability, three identical samples of each
material were tested.
Figure 2.1 - Terracotta Tile in Preliminary Testing

A correlation between bending strength and Young's Modulus was apparent in the
results. The terracotta tiles were found to be considerably less stiff and strong than the
porcelain tiles, and both types showed some variation between strength tests. An
approximate standard deviation was calculated to give some indication of variability. The
coefficients of variation for the terracotta and porcelain tiles were 3.8% and 3.7%
respectively. The consistency of these results suggests that the microstructure and
fracture processes were similar.

Typically, soda-lime glass has a Young's modulus of 68-72 GPa (Ashby, 2011) and
therefore the value of 59.50 GPa obtained is lower than expected. A greater variation in
bending strength was observed between samples for the glass than the tiles, with an
approximate coefficient of variation of 11.8% calculated. The smooth surface, sharp edges
and amorphous molecular structure of the glass is likely to make its strength highly
sensitive to flaw size, and hence more variable.

2.1.2 Aramid Fibres


Traditional prestressed structures (and those tested in previous projects) use high
strength steel tendons (cables or bars) under large tensile stresses to apply compression
to the bulk material. In order to overcome some of the issues faced in previous projects,
the idea of using high strength synthetic fibres was adopted. A key motivation for this
decision was the possibility of securing the tendons using knots, eliminating stress
concentrations caused by gripping and unlocking a variety of new potential prestressing
configurations.
Will Hawkins, Churchill College 9
Prestressed Ceramics

The material chosen was Kevlar 49, a high modulus aramid


fibre, spun into yarn as shown in Figure 2.2. A cross-
sectional area of 0.178 mm2 was calculated by weighing a
length of yarn, assuming a density of 1.44 g/cm3. A Young's
Modulus of 112 GPa is assumed throughout this report (Du
Pont, 1991). Tensile stress-strain curves for aramids are
linear up to failure, and the ultimate strength depends on
both temperature and load duration. Tensile tests on twenty
yarns similar to those used in this project showed an
average breaking load of 444.6 N, with a standard deviation
Figure 2.2 - Kevlar 49 Yarn
of 8.2 N (Gionnopolous, 2009). This corresponds to an
ultimate tensile strength of 2498 MPa.

The material's high strength derives from its molecular alignment inter-chain bonding.
Creep and relaxation is low for Kevlar 49 compared to other synthetic fibres, but can be
significant at high stresses as shown in Figure 2.3 (Ericksen, 1976). Creep-rupture is also
observed, whereby sudden tensile failure can occur after a long period of time under
sustained load. This is shown in Figure 2.4 (Du Pont, 1991), and is particularly relevant
for prestressing applications.

Figure 2.3 - Variation of Creep Strain Rate with Stress for Figure 2.4 - Creep-Rupture Charactaristics of
Aramid (Ericksen, 1976) Kevlar (Du Pont, 1991)

Aramid fibres also undergo a loss of strength with exposure to UV radiation, and
therefore those used throughout this project were stored in a dark environment
whenever possible. Steps were also taken to minimise handling of the fibres to prevent
damage.
Will Hawkins, Churchill College 10
Prestressed Ceramics

2.2 Design & Construction of a Portable Testing Rig

2.2.1 Design Requirements


It was proposed to design and construct an all-in-one, portable testing rig which would
allow beams to be tested quickly and consistently with minimal set-up time, and would be
suitable for demonstrations outside the lab. The following design requirements were
specified:

 Beams loaded in 4-point bending, accommodating a range of lengths.


 Applied moment in excess of 500Nm (based on the strongest beams tested in
previous projects).
 Precise control of loading at a variety of rates for beams of varying stiffness.
 Accurate measurement and recording of load and curvature.
 Easy viewing of beams during testing.
 Minimised risk of injury to operator and bystanders.

2.2.2 Loading & Support


The completed testing rig is shown in Figure 2.5. An existing steel frame was modified to
incorporate adjustable support platforms, allowing beams of length 750 mm - 1250 mm
to be tested. The end supports and loading points allowed full rotation, and a single
sliding support ensured the beam was simply supported.

Figure 2.5 - Completed Testing Rig


Will Hawkins, Churchill College 11
Prestressed Ceramics

Figure 2.6 shows the bending moments and shear forces arising from this particular
loading arrangement (assuming small displacements). A central region of uniform
maximum moment and zero shear was
created, enabling a simple analysis of
beam behaviour. The maximum design
load F was 5 kN. Allowing for 25 mm
overhang at each end, the length l could
vary from 200 mm to 450 mm depending
on the position of the supports, and hence
the maximum applied moment was in the
range of 500 Nm to 1,125 Nm. Figure 2.6 - Bending and Shear in Four-point Bending

The loading system was powered by a variable speed electric motor, connected to a
gearbox. Torque characteristics were determined by winching known weights. Keeping
the rate of displacement low was a key consideration, since stiffer beams could then be
tested in a controlled manner. An additional 3.7:1 reduction in rotation was achieved
using components from a bicycle drive chain. A Dyneema® rope, proof tested to 10kN,
was wrapped around the winch shaft and applied the load to the beam via the load cell
and loading bar. A synthetic rope was chosen due to its high strength and flexibility,
which allowed simple fastening with knots and the winch shaft diameter to be minimised,
further reducing the rate of displacement.

2.2.3 Instrumentation & Data Logging


A simple load cell consisting of a steel bar in three-point bending was designed,
constructed and calibrated. This is shown in
Figure 2.7. Four strain gauges (two in tension,
two in compression) formed a full Wheatstone
bridge. Noise, temperature effects and axial
strains were therefore cancelled. During
calibration, the measured strain was found to
vary linearly with load up to and beyond the
Figure 2.7 - Load Cell in Use
5kN design strength.
Will Hawkins, Churchill College 12
Prestressed Ceramics

Three wire transducers allowed the curvature of the central region to be measured.
These were attached directly to the beam at each of the loading points and at the centre.
Assuming a constant curvature a circular arc is formed. The curvature κ is hence
calculated from the central deflection δ and the length of the maximum moment region
(300 mm) as follows:

The transducer bodies were attached to the top of the frame, which remained unloaded
during testing and so provided a fixed datum from which to precisely measure distance.

The four data signals were recorded at one second intervals by a data logger, and
calibrated using Scorpio® software. This enabled moment curvature plots to be produced.

2.2.4 Performance & Accuracy


The loading mechanism applied a displacement of 100 mm in 0.8-5.5 minutes, making it
suitable for testing beams of variable stiffness and performing demonstrations. In order
to check the strength and accuracy of the completed rig, a steel bar was tested up to the
design 5 kN load. The resulting moment-curvature plot was linear as expected, and a
realistic Young's Modulus of 204 GPa was calculated from its gradient.

2.3 Beam 1 - Pre-tensioned Test Beam


In order to check the suitability of aramid yarn as a prestressing tendon material, a 'test
beam' was produced using a design copied from a previous
project, replacing the steel cables with aramid yarn.

2.3.1 Design & Construction


The 800 mm long pre-tensioned beam featured a simple
'sandwich' design consisting of a length of aramid yarn looped 54
times between two layers of terracotta tiles 40mm wide. Tile
pieces were staggered between each layer to minimise the effects
of joints. The prestressing tendons were looped around steel pins
and stressed to 10 kN using a hydraulic jack, before being bonded
to the tile pieces using Araldite® adhesive. The distribution of
tendons is shown in Figure 2.8. Once the adhesive had set, the
jack was depressurised and the load transferred to the beam. Figure 2.8 - Looped Aramid
Tendon Layout of Beam 1
Will Hawkins, Churchill College 13
Prestressed Ceramics

The aramid yarn was found to be easy to work with, and could be successfully secured
using a knot. The load appeared to be well distributed between each loop and the average
tensile force of 185.2 kN per yarn was achieved (a stress of 1040 MPa).
Ac 720 mm2
2.3.2 Theoretical Predictions
At 9.61 mm2
Using formula derived in Appendix A, predictions of cracked and b 40 mm
uncracked beam behaviour were made. The assumed section and d 10 mm
h 20 mm
material properties are shown in Table 2.1. In this case the I 26640 mm4
adhesive and tendon layer was not included in calculating section Ec 18.54 GPa
Et 112 GPa
properties. A first cracking moment of 80.5 Nm was predicted. The σb 17.74 MPa
crack re-opening moment, which depends on prestress only, was Pi 10000 N
Table 2.1 - Beam 1
predicted to be 33.3 Nm. Assumed
Parameters
2.3.3 Results & Discussion
Figure 2.9 shows the measured results along with the theoretical predictions. The initial
gradient closely matched the predicted results, suggesting that the measured tile Young's
Modulus was accurate. Crack reopening was observed at approximately 26 Nm,
corresponding to a prestress of 7.8 kN. A revised cracked prediction was therefore made,
and can be seen to more accurately fit the data. Five weeks had passed between the
beam's construction and its testing, during which some degree of relaxation of the tendon
is likely to have taken place.

Figure 2.9 -Beam 1 Results and Comparison With Theory


Will Hawkins, Churchill College 14
Prestressed Ceramics

First cracking occurred at 49 Nm, corresponding to a tensile strength in the ceramic of 8.6
MPa. This is much lower than the expected value of 17.7 MPa, due to the size dependent
strength of brittle materials. This phenomenon is described in Section 4.6 of this report.

Figure 2.10 - Beam 1 During Testing


The testing of Beam 1 can be seen in Figure 2.10. The beam failed suddenly when a
100mm long end section of the compression side of the beam sheared off, outside the
region of maximum moment. A lack of bond between the tile and Araldite® was the
probable cause.

Most importantly, the results confirmed that the aramid fibres had worked effectively as
a prestressing tendon material. The agreement between the theoretical predictions and
observed behaviour also provided further evidence of the testing rig's accuracy.

3. Developing a Post-tensioning System


After the success of Beam 1 in demonstrating the potential of the aramid yarn, it was
desired to develop a post-tensioning system which would take full advantage of its
strength, stiffness and flexibility. The system was intended to be as lightweight as
possible, incorporating a minimum amount of steel so as not to divert from ceramic as the
primary beam material. It was hoped that the system would be simple and adaptable
enough to be incorporated into a variety of prestressed ceramic components, unlocking a
multitude of opportunities for further investigations.

3.1 Beam 2 Design

3.1.1 Concept - 'Pearl-chain' Reinforcement


A potential application of small to medium scale post-tensioning using a flexible tendon is
in pearl-chain reinforcement. This recent innovation proposes the construction of 'super-
light' concrete structures, whereby high strength pre-fabricated concrete elements with
tendon ducts are held together by prestressing and establish optimised compression and
Will Hawkins, Churchill College 15
Prestressed Ceramics

tension zones within a larger body of lightweight aggregate concrete, which contributes
little strength but stabilises the reinforcement (Hertz, 2009). The pearl-chain
reinforcement therefore forms a truss or mesh, which is prestressed and inspected before
being cast into the final structure.
Complex and highly optimised mesh
shapes could potentially be achieved
by shaping the individual pre-cast
elements accordingly. An example of a
beam, with compressive arch and
Figure 3.1 - Potential Design of a Deep Beam Using
tensile tie, is shown in Figure 3.1. Pearl-chain Reinforcement (Hertz, 2009)

Ceramics are an ideal material with which to test this concept, having similar properties
to high-strength concrete. The pearl-chain concept was taken forward as inspiration for
Beam 2, with the aim of investigating the behaviour unbonded segmental prestressed
beams as well as developing the post-tensioning system.

3.1.2 Ceramic Tile Segments


It was decided that four identical 200 mm long segments would be built, making up an
800 mm long beam with a single unbonded tendon passing through a duct at centroidal
level. The simplicity of this design allowed the beam to be constructed and reconstructed
quickly, so that iterations of the
post-tensioning system could
be tested in rapid succession
during its development. Since
the segments were unbonded,
the beam had zero bending
stiffness until prestress was
applied and locked the
segments together. Figure 3.2 - Creation of a Tendon Duct

A significant practical challenge involved creating a segment with a tendon duct running
through its centre. By modifying a standard tile cutter so that the grinder did not cut
through the entire thickness of the tile, thin grooves could be cut, multiples of which
could be built up into a channel as shown in Figure 3.2. The stronger porcelain tiles
Will Hawkins, Churchill College 16
Prestressed Ceramics

proved to be better able to resist damage during this process. By gluing two identical
channeled tiles together using Araldite®, a segment with a square duct was created. A
5mm rubber shrink-wrap tube was positioned within the duct in order to provide a non-
abrasive surface against which the tendon could slide. During setting of the adhesive, this
was held open using a length of M5 studding.

3.1.3 Post-tensioning System & Beam Construction


Beams constructed in previous projects relied upon a hydraulic jack to stress the tendon,
around which the beam was constructed. The components which can be manufactured in
this way are limited to those which fit within the stressing rig.

Preliminary calculations suggested that a 10 kN prestressing force could be achieved by


extending a 30 strand aramid fibre tendon by approximately 16 mm over a 1 m beam
length. This distance was considered small enough to be contained within a single unit
attached to the beam itself.

Figure 3.3 - Dead (Left) and Live (Right) Ends of Initial Version of Post-tensioning System
A new post-tensioning system was developed, whereby a looped tendon was held in place
at either end by pins and stressed by increasing the separation distance between plates at
the live end, as shown in Figure 3.3. This was achieved by tightening a pair of M12 bolts,
which passed through threaded holes in the outer plate and rested in circular
indentations on the inner plate to ensure stability under load. Each plate contained a
central hole through which the tendon passed. The pins incorporated a deep groove
around their centre, tear-drop shaped in section, aiming to minimise pinching of the
tendon in this region. Square pins were chosen to reduce bearing stresses against the
steel end plates.

The components were designed to withstand a 10 kN prestressing force, and were made
as compact as possible to minimise steel weight and bulk. The live end plates were
Will Hawkins, Churchill College 17
Prestressed Ceramics

modelled as beams in three-point bending, and the required thickness calculated


according to first yield at the central section (reduced due to the presence of a hole). It
was conservatively assumed that the plates were mild steel with yield stress of 200 MPa.
The distance between the bolts was minimised to reduce the amount of steel required.
Consideration was also given to shearing of the threads within the outer plate, and the
bolt diameter was specified accordingly.

The tendon consisted of a single length of aramid yarn, looped 40 times to create a total
area of 7.12 mm2. To create the tendon each pin was clamped in place at the required
separation, and the yarn looped by hand. An even tension was ensured to avoid any slack.
The reel of yarn was held horizontally and allowed to rotate during this process, as
shown in Figure 3.4, since this was found to minimise accumulated twist. After the
required number of loops, the yarn was trimmed and tied securely. The completed
tendon was then threaded through each segment and the end assemblies using a piece of
wire, before being fixed in place by the insertion of the pins. The entire beam was
stiffened and held together by the application of prestress.

Figure 3.4 - Construction of a Looped Tendon

The desired 10 kN of prestress corresponded to an average force of 250 N per yarn, well
within the individual breaking limit. The unloaded length of the tendon was 850 mm. The
extension required to achieve this force was therefore calculated as follows:
Will Hawkins, Churchill College 18
Prestressed Ceramics

3.2 Issues & Design Iterations

3.2.1 Segment Alignment & Interface


Initially, small rubber pieces were placed at each segment boundary. These were
intended to reduce stress concentrations which could lead to local crushing of the tiles, as
has been previously observed in similar situations (Grant, 2012). A beam with 8 kN
prestress was constructed (premature tendon failure is discussed in Section 3.2.2).

Figure 3.5 - Prototype Version of Beam 2 During Tesing


The beam during testing is shown in Figure 3.5, and the results obtained in Figure 3.6. As
expected, it was observed that the majority of curvature occurred at the interfaces
between tiles, and due to the rubber a considerable degree of hysteresis was revealed by
the results. It was apparent that the behaviour was being largely dictated by the rubber,
and that an alternative was required.

A further issue concerned the straightness of the beam. Small variations between the
length of each tile piece resulted in some rotation at each tile interface, which the rubber
pieces were ineffective at eliminating.

Since concerns over local crushing still


existed, the beam was rebuilt in a jig
using dental plaster at the tile
interfaces to straighten the beam (see
Figure 3.7). This was successful to
some extent, but bending tests resulted
in disintegration of the plaster and
hence a high degree of flexibility in the
Figure 3.6 - Prototype Beam 2 Test Result, Showing
joints remained at low bending moments. Hysteresis
Will Hawkins, Churchill College 19
Prestressed Ceramics

The issue was eventually resolved by grinding the ends of each segment flat, using the
edge of a tile cutter blade. It was found that the tile was strong enough to avoid local
crushing at segment interfaces under full prestress without the need for rubber or
plaster. The segment interfaces also achieved maximum stiffness in this way.

3.2.2 Avoiding Premature Tendon Failure


A recurring issue was that of tensile failure of the
tendon during tensioning. Despite the average load of
250 N being only 56% of the ultimate breaking
strength (444.6 N), individual fibres would break
before the full extension was achieved. A typical failure
is depicted in Figure 3.7. This would occur usually at
the live end of the beam. A number of possibilities as to
the cause of failure were investigated before the
Figure 3.7 - Use of Dental Plaster and
problem was eventually resolved. Typical Premature Tendon Failure
Shaft Friction
It was proposed that a high level of shaft friction was causing the force in the tendon to be
non-uniform and higher at the live end. A constant shaft friction results in a linear
variation of tendon force along the beam, reducing with distance from the live end and
equal to the theoretical (zero-friction) value at the mid-point.

A simple test was performed by dragging the tendon through the beam and measuring
the frictional resistance with a spring force gauge. The resistance was measured to be
only 5 N/m, insignificant compared with the desired 10 kN prestressing force. The
frictional force per unit length was assumed not to vary with the tendon force since the
tendon duct was straight, and therefore the effects of shaft friction were concluded to be
insignificant.

Uneven Force Distribution


Tendon failure occurred only in individual yarns, suggesting an uneven loading
distribution. Since it was observed that some yarns could become slack during the
manufacture of the tendon and beam, it was proposed that this was the cause of early
tendon failure. An identical tendon to that being used (40 loops over a gauge length of
850 mm) was subject to a tensile test in an Instron® machine, with results shown in
Figure 3.8. The loop was wrapped around steel cylinders at each end.
Will Hawkins, Churchill College 20
Prestressed Ceramics

Had each yarn achieved its full


capacity of 444.6 kN, a total
strength of 17.8 kN would have
been achieved. Weakest link
theory dictates that this would
in fact be reduced for multiple
yarns acting together (Weibull,
1939), however a single yarn
failed when the total load was
only 8.8 kN, and an additional Figure 3.8 - Tensile Test Results for a 40 Yarn Tendon of Gauge Length
850 mm
sudden failure of multiple yarns occurred at 11.0 kN. Further evidence of uneven loading
is given by the approximate 3.5 mm of extension required before the full stiffness was
mobilised. This corresponds to a difference in force of 82.1 N between yarns which were
taut initially and those which only became taut after 3.5 mm of total extension.

As a potential way of eliminating slackness, the idea of forming the tendon into a rope
was explored. The yarn has a built-in direction of twist, and a 'string' can be formed by
twisting yarns together in the opposite direction.
These 'strings' can then be twisted together into a
rope. Two twisted ropes were formed, each made up
of 18 yarns, using the twisting pattern shown in Figure
3.9. These were identical apart from the number of
twists per unit length. Twisting was achieved using a
hand drill and a hooked piece of wire to hold the yarn.
Figure 3.10 shows the completed ropes.

Figure 3.9 - Stable Rope Twisting Pattern


For tensile testing, each rope was tied into a single loop thus creating a 36 yarn tendon
which could potentially be wrapped around pins and stressed using the current post-
tensioning system. The full strength of 36 yarns is 16.0 kN, however tensile failure
occurred at 3.5 kN and 2.6 kN for the less and more tightly twisted ropes respectively. For
both ropes, six yarns failed simultaneously (a single 'string'). It was concluded that
twisting the yarn into ropes causes a significant reduction in strength, because twisting
results in yarns being loaded more unevenly.
Will Hawkins, Churchill College 21
Prestressed Ceramics

Further investigations showed that the


uneven loading was a result of friction
around the pins. Despite the smoothness
of the aramid, high reaction forces around
the pins could create a considerable
change in force. Tests to measure friction
confirmed this. The problem was solved
by replacing the square pins with Figure 3.10 - Completed Ropes
cylindrical ones which could roll against the steel plate, thus redistributing load between
each side of the tendon (for the tensile test shown in Figure 3.8, the end cylinders were
constrained against rolling and hence early failure was still observed).

Abrasion & Stress Concentration


Despite the success of the cylindrical pins in redistributing force more evenly, early
tendon failure was still occasionally observed. It was decided to reduce friction, abrasion
and stress concentration by polishing any steel surface which came into contact with the
aramid fibres. The holes through the loading plates were also increased in size and
counter-sunk, as is shown in Figure 3.11.

Figure 3.11 - Use of a Round End Pin, Polishing of Steel Surfaces and Counter-sinking of Holes to Avoid
Premature Tendon Failure
The number of yarns were also increased to 44 (tendon area 7.832 mm2) in order to
further reduce the chance of tendon failure. A beam incorporating the modifications
described was finally constructed, prestressed successfully to 10 kN and tested.

3.3 Theoretical Predictions


The section of Beam 2 is shown in Figure 3.12. Considering
only the porcelain tile area, the section properties are:

Figure 3.12 - Beam 2 Section


Will Hawkins, Churchill College 22
Prestressed Ceramics

The central region contained a single segment interface at its centre. This was modelled
as a single crack over the entire depth of the section.

An unbonded prestressed segmental beam can be considered to have three distinct


regions of behaviour in bending. Initially, the largest tensile stresses resulting from the
applied moment are insufficient to overcome the initial compressive prestress. The beam
therefore acts as uncracked along its entire length, and the constant curvature can be
calculated as in Appendix A (uncracked). Using the value of Young's Modulus determined
in preliminary testing for the porcelain tiles, the bending stiffness EI is 1245.0 Nm2.
Assuming a prestressing force of 10 kN, first cracking was predicted to occur at 33.0 Nm.

Beyond this point behaviour is no longer linear. At the segment interface, the beam
behaves as cracked. The neutral axis position rises with increasing moment, and the local
curvature can be calculated as in Appendix A (cracked). Predicting the average curvature
(measured in the test) would involve an assumption of the length of the region of
influence of the central crack. This problem concerns local stress fields and was not
included in this analysis.

As the local curvature increases at the segment


interface, the neutral axis continues to rise.
Eventually, the segment interface can be modelled as
shown in Figure 3.13, where the ceramic compression
zone becomes infinitesimally thin. In this fully open
Figure 3.13 - Fully Open Segment
case, the curvature of the solid segments (still Interface Model
uncracked) is assumed to be negligible in comparison to the rotation at the segment
interface. Therefore the central deflection δ measured during the test is related to the
angle θ:

The compressive reaction force in the ceramic is equal to the prestressing force P by
equilibrium. The moment is this force multiplied by the lever arm d, initially assumed to
be half the section depth:
Will Hawkins, Churchill College 23
Prestressed Ceramics

As the angle θ increases, the total length of the prestressing tendon increases by ΔL with a
corresponding increase in tendon strain and force:

The beam therefore retains some degree of stiffness. A relationship between the moment
M and angle θ is hence found for the fully open case:

3.4 Results, Discussion & Comparison with Theory

3.4.1 Moment-Deflection Relationship


In order to make comparisons between the measured results and theory, the results are
shown in a moment-deflection plot whereby predicted curvature κ and rotation θ have
been converted to central deflection δ. This is shown in Figure 3.14.

Figure 3.14 - Beam 2 Results and Comparison with Theory


The beam was loaded and unloaded three times to increasing curvatures, before finally
being loaded to failure. The results show good agreement between loading cycles,
reflecting the fact that the beam did not undergo any damage until near failure.

Uncracked
As predicted, the initial behaviour is uncracked linear-elastic. The predicted bending
stiffness of 1245.0 Nm2 was an overestimate. If it is assumed that I was correct, this
suggests that the value of Young's Modulus used was too large. This may have been due to
experimental errors made during materials testing. It is also possible that the Young's
Will Hawkins, Churchill College 24
Prestressed Ceramics

Modulus is not uniform throughout the tile as was assumed. The dotted line shows a
revised prediction based on a Young's Modulus of 30 GPa, which can be seen to more
accurately fit the data.

First Cracking
As soon as the segment interface begins to open, a reduction in the effective section
causes a drop in stiffness. First cracking was observed at around 25 Nm (earlier than the
predicted value of 33 Nm). Since the first cracking moment depends only on the geometry
and the prestressing force, a reliable revised estimate of prestress was made. Assuming
the tendon remains at centroidal level, a first cracking moment of 25 Nm corresponds to a
prestressing force of 7.57 kN. A number of reasons why the prestressing force was lower
than the 10 kN intended were considered:

 The required tendon elongation calculated did not account for the picking up of
slack as was observed in tendon testing (Figure 3.8). Increasing the elongation by
3mm to account for slack would have resulted in a additional 3.1 kN of force.
 The tendon extension could not be measured to an accuracy greater than 1 mm, so
some degree of variability is expected in prestressing force.
 Elastic shortening of the beam may have resulted in a reduction in tendon length. A
simple calculation can estimate this:

The corresponding reduction in tendon force is 0.4 kN.


 Although the beam was stressed immediately prior to testing, some degree of rapid
relaxation of the aramid fibres could have occurred.

Fully Open Segment Interface


After first cracking, the segment interface widened and eventually approximately bi-
linear behaviour was observed. Theoretical predictions based on the revised prestressing
force of 7.57 kN initially gave a considerable overestimate of bending moment as shown
by the solid black line of Figure 3.14.

The assumption that the lever arm d (as defined in Figure 3.13) was half the section
depth at 9.5 mm was called into question. It is likely that the actual lever arm was less
than this value. Despite fitting fairly tightly within the duct, the tendon centroid is likely
to have moved upwards as a result of the curvature, by perhaps 1-1.5 mm. Also, the
Will Hawkins, Churchill College 25
Prestressed Ceramics

assumption that the compressive force in the ceramic acts at the very top of the section is
an idealisation. As can be seen in Figure 3.15, it was observed that the depth of the
compression zone did not reduce below approximately 3 mm. The dotted line shows a
revised estimate with a lever arm of 7 mm, which better fits the experimental data.

3.4.2 Failure Mode


Failure occurred suddenly and symmetrically due to compression of the tiles at the
segment interface. Some flaking of the top surface was observed prior to failure, as shown
in Figure 3.5. A large proportion of the upper tile was lost, suggesting successive failure of
the compression zone.

The improvements to the post-tensioning system had prevented tendon failure. A


maximum interface opening angle of 9.5˚ was estimated, corresponding to an increase in
tendon force of 1.02 kN (to 8.59 kN).

Figure 3.15 - Central Segment Interface Immediately Before and After Failure

4. Post-tensioned Glass Beams


4.1 Beam 3 Design & Construction
Taking forward the lessons learnt from the design, construction and testing of Beam 2, a
new beam design was developed with the primary aim of enabling a number of key
parameters and characteristics of prestressed ceramic beams to be investigated. The
beam featured a similar prestressing system to Beam 2, rebuilt to fit the new required
geometry.

Annealed soda-lime glass was chosen as the primary construction material. Being
transparent, glass allows excellent observation of cracking patterns and failure modes,
and since large pieces are available joints along the beam are unnecessary.
Will Hawkins, Churchill College 26
Prestressed Ceramics

The basic design is shown in Figures 4.1 and 4.2, and featured
two 1000 x 40 x 6 mm lengths of glass fixed together at four
points by 6 mm thick flat aluminium pieces secured using
Araldite® adhesives. These were located at each end and at the
loading points. The aluminium pieces created a gap between
each piece of glass and dictated the height of the tendon along
the beam. The tendon eccentricity could hence be optimised to Figure 4.1 - Beam 3 Section
Geometry
match the bending moment diagram.

Figure 4.2 - Beam 3 Side Elevation


A 44 yarn tendon was used for each version of Beam 3 made. Due to the open space
between the glass sides, the tendon was fully accessible after installation and stressing.
This simplified beam assembly, but most importantly allowed the tendon to be bonded to
the glass after tensioning. This was achieved by applying a layer of Araldite® adhesive to
the top and bottom of the tendon using a syringe.

A prototype, Beam 3.0, was constructed (of different dimensions to those specified in
Figures 4.1 and 4.2) in order to test the concept and refine the construction techniques.
This beam is shown in Figure 4.3. It was also tested in order to check the accuracy of the
rig for stiff beams, and to evaluate the success of the tendon bonding. Due to a failure of
the wire transducer attachment brackets, only the total central deflection was recorded.
The brackets were modified for future tests. The results (shown in Figure 4.4) were
promising, showing clearly distinct uncracked and cracked behaviour. The bond was
successful in allowing the beam to continue to function after cracking had occurred.

Figure 4.3 - Protoype Beam 3 Before Tendon Bonding


Will Hawkins, Churchill College 27
Prestressed Ceramics

Figure 4.4 - Prototype Beam 3 Test Results

4.3 Investigated Parameters


The basic design of Beam 3 allowed a number of key aspects to be easily modified. A total
of five beams were constructed and tested, incorporating variations in prestressing force,
tendon eccentricity and bond as detailed in Table 4.1.

4.3.1 Tendon Eccentricity


By varying the depth of the aluminium deviators, the tendon eccentricity in the central
region was controlled. To ensure that the tendon centroid was at the desired height, some
allowance (determined from the prototype beam) was made for the thickness of the
tendon itself. A tendon depth of 26.7 mm is of significance, since the resulting 6.7 mm
eccentricity places the tendon at the kern point. This is the maximum eccentricity at
which the prestressing force alone causes no tensile stresses in the section. For safety
reasons, it was intended to limit the eccentricity to this value to avoid handling brittle
beams with internal tensile stresses. During the construction of Beam 3.1, some slippage
of the aluminium deviators occurred as the adhesive set leading to an accidental increase
in eccentricity to 9 mm. The 17 mm eccentricity of Beam 3.4 was safe since no
prestressing force was applied.

4.3.2 Prestressing Force


Two tensile tests were conducted using an Intron® testing rig on 44 yarn looped tendons
of 1050 mm gauge length, identical to those used for Beam 3. Results are shown in Figure
4.5. Some early tendon failure was observed, for the reasons discussed in Section 3.2.2.
Will Hawkins, Churchill College 28
Prestressed Ceramics

The force-extension data was used


to determine the tendon elongation
required to achieve the desired
prestressing force. An elastic
stiffness of 759.1 N/mm was
determined, and a 3 mm offset
included to account for the picking
up of slack:
Figure 4.5 - Tensile Test Results for 44 Yarn Tendons
with Gauge Lenth 1050 mm

Thus a required extension of 16.2 mm was calculated for 10 kN of prestressing force


(Beams 3.1, 3.2 and 3.3) and 9.6 mm for 5 kN (Beam3.5).

After inserting and pinning the tendon, a judgment was required on the point from which
to measure extension, since some slack was required to insert the pin. A variability of
around 1 mm was inherent in this process. As a result, a more accurate estimation of the
slack was not deemed to be necessary.

4.4 Theoretical Predictions

4.4.1 Uncracked Behaviour


Table 4.1 shows the key characteristics of each beam, and the resulting predictions made
for the uncracked section. Appendix A details how the stress distribution and curvature
was determined for the uncracked beam. The beam was assumed to behave in a linear
uncracked manner, with bending stiffness EI, until tensile failure of the glass lead to
cracking. This was expected to occur at the bottom extreme fibre in the region of
maximum moment. Figure 4.6 shows the variation of extreme fibre stress along the beam,
which reaches a maximum (tensile) value of σbottom calculated as in Appendix A.

Figure 4.6 - Variation of Extreme Fibre Stresses (Positive in Tension)


Will Hawkins, Churchill College 29
Prestressed Ceramics

Beam Design Bonded d (mm) I (mm4) EI (Nm2) Estimated Estimated


Prestress First Cracking Crack Re-
(kN) (Nm) opening (Nm)
3.1 10  29.0 65159 3877 375 154
3.2 10  26.7 64642 3846 349 131
3.3 10  20.0 64000 3808 278 65
3.4 0  37.0 68134 4054 233 0
3.5 5  26.7 64642 3846 283 65
Table 4.1 - Beam 3 Variations, Section Properties and Uncracked Predictions
The effect of the prestressing tendon in modifying the section properties was included,
and the prestressing force P was assumed constant in the uncracked model. Values of
tensile strength and Young's Modulus were taken from preliminary testing (see Table
2.1).

4.4.2 Cracked Behaviour


The cracked-elastic model described in Appendix A assumes that the glass has zero
tensile capacity along the entire length of the cracked region. A linear strain distribution
is assumed, and the prestressing force varies from its initial value according to the local
strain. The compression zone reduces in depth as the curvature and moment increase,
and tends asymptotically towards a limiting value. Hence a linear stage of cracked
behaviour is reached.

4.4.3 Application of Theory


Two sets of theoretical predictions were made for each beam. An initial prediction of
cracked and uncracked behaviour was made based on the intended prestressing force
and the experimentally determined Young's Modulus of Glass (59.5 GPa).

Once test data was obtained, a revised value of prestress could be found based on the
observed crack re-opening moment. From the initial uncracked gradient, a revised value
of the Young's Modulus of the glass could also be calculated. A second theoretical
prediction of moment-curvature behaviour was therefore made based on these revised
values. The revised prestressing force was also used to estimate the tensile strength of
the glass, based on first cracking. Both sets of predictions are included in Figures 4.8, 4.9,
4.11, 4.13 and 4.15, with revised predictions shown as dotted lines. A summary of key
numerical results and estimated parameters is shown in Table 4.2.
Will Hawkins, Churchill College 30
Prestressed Ceramics

4.5 Test Set-up and Procedure


Each beam was tested to destruction. A number of load-reload cycles were performed
where possible in order to investigate cracked behaviour. In the following moment-
curvature plots, the curvature given is that caused by the applied moment.

Supports were placed 20 mm from the ends of the glass pieces, hence the distance l was
330 mm (referring to Figure 2.6). The weight of the load cell and loading bar was taken
into account in calculating the applied moment for all tests.

A detachable Perspex® safety screen was cut and fitted for the testing of glass beams.

4.5 Results & Observations

4.5.1 Beam 3.1


For this beam, the tendon was left unbonded to the glass. An additional pair of aluminium
deviators were included in the maximum moment region in order maintain a constant
tendon eccentricity during deflection, as can be seen in Figure 4.7. The tendon
eccentricity was marginally beyond the kern point due to the manufacturing error
previously described.

Figure 4.7 - Beam 3.1 Prior to Testing


Results and theoretical predictions are shown in Figure 4.8. The beam failed suddenly
and catastrophically when first cracking occurred at 320 Nm. Assuming 10 kN of
prestress, the tensile stress in the glass was predicted to be 49.6 MPa.

Since no crack reopening behaviour could be investigated, a revised estimate of prestress


was not made. The initial estimate of uncracked stiffness was lower than the measured
value. Assuming a Young's Modulus of glass equal to 90 GPa showed a good correlation
with measurements.
Will Hawkins, Churchill College 31
Prestressed Ceramics

Figure 4.8 - Beam 3.1 Results and Comparison with Theory

4.5.2 Beam 3.2


Beam 3.2 was similar to Beam 3.1, but with a bonded tendon which was also correctly
positioned at the kern point. Figure 4.9 shows the measured and theoretical moment-
curvature response.

Figure 4.9 - Beam 3.2 Results and Comparison with Theory

Unlike the unbonded Beam 3.1, the load was maintained beyond first cracking. The
adhesive bond was effective in maintaining the position of the tendon and preventing
significant spalling.
Will Hawkins, Churchill College 32
Prestressed Ceramics

There was a tendency for increasingly negative curvatures to be recorded between


unload-reload cycles, which is likely to have been a result of small amounts of material
loss with increased cracking. The majority of cracking occurred simultaneously in two
distinct episodes. As shown in Figure 4.10, a distinct uncracked region was present at the
top of the beam and some cracking was observed outside the region of maximum
moment.

Figure 4.10 - Beam 3.2 During Testing


A revised prestressing force of 8.78 kN was calculated based on crack reopening
observed at 115 Nm, and the Young's Modulus of glass altered to 82 GPa to match the
uncracked gradient.

Failure occurred suddenly at 273 Nm when the glass in compression appeared to burst
outwards. This did not occur at the peak moment, suggesting that the beam weakened
during loading and unloading.

4.5.3 Beam 3.3


Beam 3.3 was designed to be identical to Beam 3.2 apart from having zero eccentricity.

Figure 4.11 - Beam 3.3 Results and Comparison with Theory


Will Hawkins, Churchill College 33
Prestressed Ceramics

Figure 4.11 shows the experimental and theoretical results. After first cracking, a distinct
plateau of maximum moment was observed. Both reload curves were observed to return
to this plateau before additional cracking took place, with some loss of material as shown
in Figure 4.12. A significant increase in the amount of cracks present during the second
unload-reload cycle was observed to cause a reduction in the cracked stiffness (to near
theoretical values). This suggests that uncracked regions were contributing to the
stiffness during the first unload-reload cycle.

In contrast to Beam 3.2, an increase in


unloaded curvature was observed after
cracking. The likely cause of this was
small fragments of broken glass being
caught between cracks as they closed. No
negative curvature was measured since
the prestress applies no moment to the
beam when unloaded.
Figure 4.12 - Beam 3.3 Material Loss

The Young's Modulus of the glass was increased to 83 Gpa in order to match the
experimental data. Crack reopening at 64 Nm suggested that the prestressing force was
9.90 kN. Compressive failure in the glass was again sudden, resulting in total material loss
in the central region.

4.5.4 Beam 3.4


Since it was noted that ultimate failure in previously tested beams was in initiated in the
compressive region, it was decided that a beam with zero prestress should be tested in
order to investigate composite action alone. To maximise the efficiency of this beam the
tendon was placed at the maximum possible eccentricity of 17mm. Results are shown in
Figure 4.13 and demonstrate a distinctly different behaviour to Beams 3.2 and 3.3.

Once cracking had occurred (at the predicted moment of 233 Nm), the stiffness of the
beam was permanently reduced. By equilibrium, the neutral axis depth was predicted to
remain constant. After increasing the assumed Young's Modulus of the glass to 78 GPa,
cracked and uncracked predictions matched the data well. The positive offset in
curvature after cracking can be again attributed to cracks not fully closing on unloading.
Will Hawkins, Churchill College 34
Prestressed Ceramics

Figure 4.13 - Beam 3.4 Results and Comparison with Theory

Figure 4.14 - Beam 3.4 During Testing

Despite having zero prestress, the beam largely remained able to sustain load after
cracking. This suggests that it was the adhesive which held the cracked beam together
rather than the prestress itself. Since the tendon was at the very bottom of the beam, the
bond was able to 'hold up' cracked sections. Figure 4.14 shows Beam 3.4 after the first
instance of cracking.

4.5.5 Beam 3.5


The final beam tested featured a bonded tendon, at the kern point, with a reduced
prestressing force of 5 kN. As shown by the results in Figure 4.15, distinct uncracked and
cracked behaviour was observed.

The reduced prestress resulted in a lower crack reopening moment in both predicted and
observed results when compared to the similar Beam 3.2 (with 10 kN prestress). The
same negative curvature at zero load after cracking was again present in the results.
Will Hawkins, Churchill College 35
Prestressed Ceramics

Figure 4.15 - Beam 3.5 Results and Comparison with Theory

First cracking occurred at 283 Nm and resulted in the sudden formation of a crack
pattern over the entire central region. The beam immediately prior to failure is shown in
Figure 4.16. The cracked stiffness was therefore similar between unload-reload cycles,
since only a small amount of additional cracking occurred. The estimated prestress was
4.66 kN and the Young's Modulus was revised to 110 GPa.

Figure 4.16 - Beam 3.5 During Testing

4.5.6 Summary of Numerical Results

Table 4.2 summarises the key numerical results from testing, and includes the
parameters which were estimated using the measured data. The peak/failure moment
was taken as the maximum moment sustained by the beam. For Beams 3.4 and 3.5, this
was larger than the first cracking moment
Will Hawkins, Churchill College 36
Prestressed Ceramics

Uncracked Crack Reopening First Cracking Failure


Beam Estimated Predicted Observed Estimated Predicted Observed Estimated Peak/Failure Maximum
Ec (GPa) (Nm) (Nm) Prestress (Nm) (Nm) Tensile Moment Curvature
(kN) Strength (Nm) (m-1)
(MPa)
3.1 90 154 n/a 10000* 375 320 49.6* 320 0.055
3.2 82 131 115 8779 349 305 57.7 305 0.193
3.3 83 65 64 9895 278 244 55.6 244 0.361
3.4 78 0 0 0 233 233 66.7 246 0.279
3.5 110 65 61 4657 283 245 56.1 275 0.315
Table 4.2 - Summary of Predicted and Estimated Parameters
*For Beam 3.1, no estimate of prestress could be made since no crack reopening was observed. The
intended value of 10000 N was therefore used in estimating the glass tensile strength.

4.6 Discussion

4.6.1 Stiffness of Glass


For each beam the initial estimate of Young's Modulus had to be increased in order to
match the observed uncracked gradient. It is therefore likely that the value obtained
during preliminary testing was incorrect. This may have been due to compliance in the
testing rig, or measurement inaccuracies. As mentioned in Section 2.1, a more usual value
of stiffness for soda-lime glass is 70 GPa, however this is still lower than the estimated
values, which also show considerable variability. The glass was all ordered in one batch,
so pieces are likely to have had a similar stiffness.

The true stiffness of the beams would have been larger than predicted due to the
presence of a number of components which were not taken into account in the analysis.
These included:

 The aluminium deviators, contributing significantly to the stiffness at either end of


the central region (aluminium and glass have a similar stiffness).
 The transducer attachment brackets, visible in Figure 4.15, which were steel and
attached rigidly to the beam.
 The steel loading plates, which would have resisted curvature of the beam.
 A thick layer of Araldite® around the tendon in the four bonded beams.

Further analysis would be required to quantify the effects of each of these components in
increasing the stiffness of the beam.
Will Hawkins, Churchill College 37
Prestressed Ceramics

4.6.4 Measured Prestress


From the crack reopening moments, the estimated values of prestress were found to be
similar to the intended values. This was likely to have been a result of accounting for the
slack in the tendon. As mentioned previously, measuring accurately the tendon extension
applied while on the beam was not possible, and hence some variability would have been
expected. For example, a variation in extension of 1 mm corresponds to a change in force
of 0.84 kN for the tendon used.

The effects of elastic shortening were not taken into account, but can be quantified as in
Section 3.4.1. Assuming the glass has a Young's Modulus of 70 GPa in this case, the elastic
shortening due to a 10 kN force is 0.3 mm, corresponding to a drop in prestressing force
of 0.25 kN. This partially explains why the estimated prestressing forces were
consistently lower than design values. Additionally, some relaxation of the tendon would
also have occurred, since at least 24 hours was required between stressing and testing to
allow the Araldite® to set (for bonded beams).

4.6.2 Tensile Failure of Glass


Ignoring the value found for Beam 3.1 (since the assumed 10 kN of prestress may not
have been reliable) the average value of tensile strength was found to be 59.0 MPa, with a
standard variation of 4.50 MPa. This
compares to an average of 66.8 MPa
and standard deviation of 7.90 MPa
found in preliminary testing.

Figure 4.17 - Roughness of Glass Edges


The discrepancy between these sets of results is due to the size dependant strength
which is a feature of brittle materials. In a region of uniform maximum stress (as seen in a
four-point bending test), tensile failure occurs at the location of a maximum sized flaw. As
shown in Figure 4.17, the edge of the glass had a visible roughness which arises from the
cutting process. If the size of flaws is assumed to be distributed according to some
probability function, it is likely that a larger flaw will be present over a larger area and
hence a lower tensile strength will be achieved. The region of maximum stress in the
beam tests was 300 mm in length and contained four edges, compared to 60 mm and two
edges in the materials testing.
Will Hawkins, Churchill College 38
Prestressed Ceramics

4.6.3 Cracked Behaviour & Crack Patterns


Similarity between crack patterns was observed in each of the bonded beams. Cracks
were likely to have initiated at the
bottom face of the beam at a large
flaw, before propagating upwards
through the tensile region of the
glass. Figure 4.18 shows a typical
concentrated region of cracking,
with multiple cracks emanating from
the same point on the tension face.
Figure 4.18 - Beam 3.4 Concentrated Cracking

Some consistency in crack spacing was observed in the central region, as Figure 4.19
shows. Tensile stresses which are relieved at a crack build up again away from the crack
at a rate which depends on the bond shear stress. Another crack then forms when
stresses are sufficient. This creates a minimum crack spacing.

Figure 4.19 - Beam 3.5 Uniform Crack Spacing


Crack propagation was rapid and unstable. Usually the crack direction was observed to
change with distance from the tension face, becoming horizontal and often joining with
other cracks. The horizontal cracking may have been a result of Poisson's ratio effects in
the tension region, creating vertical tensile stresses around the neutral axis. No cracks
were observed on the compressive face, resulting in a clearly visible uncracked
compression zone with a horizontally cracked region beneath it. The depth of this
uncracked region was not observed to reduce as higher curvatures were reached, which
is in line with the theoretical prediction that the neural axis depth tends towards a
limiting value.
Will Hawkins, Churchill College 39
Prestressed Ceramics

When the central region was fully cracked, the theoretical prediction of behaviour was
reflected well in the measured results, although the predicted gradient was always
slightly lower than that measured. This can be explained by the model's assumption that
the ceramic has zero tensile strength. Whilst this is true in the cracked region, it is likely
that some tensile forces are present in the upper uncracked region of the beam (when the
neutral axis has passed above the cracks). There is hence a contribution to stiffness which
the model does not account for.

4.6.6 Failure Mode


A similar failure mode was
observed for each of the bonded
beams, which was sudden,
catastrophic and resulted in
almost total material loss in the
central region as can be seen in
Figure 4.20. The tendon always
remained intact. Figure 4.20 - Beam 3.3 Immediately After Failure

A ballpark estimate of the maximum compressive stress in the


Beam σc, fail (MPa)
glass at ultimate failure was made based on the measured
3.2 162
curvature, the approximate neutral axis depth at failure (taken 3.3 227
from the revised theoretical cracked analysis) and a revised 3.4 145
3.5 176
glass Young's Modulus of 70 GPa: Average 178
Table 4.3 - Estimated
Maximum Compressive
Stresses at Failure
Results are shown in Table 4.3 and show some degree of consistency. If it is assumed that
the compressive strength of the glass is ten times larger than its tensile strength, crushing
would be expected at approximately 590 MPa. This value was not reached, therefore is it
proposed that failure occurred due to minor axis (lateral) buckling of the uncracked
region. The curvature of the beam before failure would have encouraged downward
buckling (major axis), but this would have been restrained by the cracked glass beneath.
Localised damage to the safety screen supports the hypothesis that material was
projected laterally at high velocity.
Will Hawkins, Churchill College 40
Prestressed Ceramics

Assuming the uncracked region of each glass layer to be an Euler


column of length 300 mm and width 6 mm, simply supported at
each end by the aluminium deviators, an estimation of the
required buckling stress can be made. This is independent of the
neutral axis depth. The section under consideration is shown in
Figure 4.21.
Figure 4.21 - Assumed
Section of Euler Column

This is significantly lower than the estimated average compressive stress in the
uncracked region, which would be approximately half the maximum value (89 MPa).
Therefore it is entirely possible that minor axis buckling took place. Additional lateral
support would have been provided by the cracked material beneath, explaining the
higher stress achieved.

5. Post-tensioned Ceramic Truss

5.1 Design Task


As an extension to the investigations made and a practical design challenge, it was
proposed to construct a prestressed ceramic truss. It was hoped that the post-tensioning
system developed could be incorporated into a truss, with the aim of creating a high-
performance structure consisting primarily of brittle materials.

Matching a structural design task undertaken by Part IA students, it was desired to


construct a bridge over a span of 840 mm which would be loaded by a vertical force at its
centre. The bridge was designed to rest on flat supports at either end of this span.

5.2 Design Overview


A significant challenge was to devise a structure which could be manufactured in the
short space of time available and would be laterally stable. This required the number of
new fabrication techniques to be limited, as well as any major modifications to the post-
tensioning system which could lead to unexpected issues. A number of design concepts
were explored. The completed structure is shown in Figure 5.1, and consisted of two
triangular ceramic frames inclined towards the loading point at the top of the structure.
The horizontal tension members were post-tensioned each to 10 kN, and the mass of the
completed structure was 6.50 kg.
Will Hawkins, Churchill College 41
Prestressed Ceramics

Figure 5.1 - Completed Ceramic Truss Bridge

5.2.1 Ceramic Frames


Two identical triangular frames were constructed, using the same porcelain tiles as in
Beam 2. Each was built from 12 tiles pieces which were shaped using a water-jet cutter.
These were arranged in two layers with staggered joints as shown in Figure 5.2, designed
so that each of the four sides making up both frames were identical. The total length of
the tile frames was 920 mm, allowing for 40 mm of bearing at each end of the span.

Figure 5.2 - Ceramic Frame Tile Piece Layout

Slots were cut to allow the inclusion shrink-wrap duct, similar to the one featured in the
segments of Beam 2, to run the length of each of the horizontal tension struts. Both the
tension and compression struts are 30 mm deep, which was the smallest size considered
to be buildable. The section is therefore identical to that shown in Figure 3.12 (without
the central hole in the case of the compressive struts).
Will Hawkins, Churchill College 42
Prestressed Ceramics

The two layers of tiles were fixed together using Araldite®. During setting, these were
clamped to a wooden jig to ensure that the two frames were accurately matched and that
each was precisely flat. This was crucial since lateral imperfections would be expected to
reduce the buckling capacity of the compression struts. Again, 5 mm studding was
inserted to hold open the tendon duct, which was removed by screwing after setting.

5.2.2 Prestress Application and End Connections


The chosen design incorporated a largely unmodified version of the post-tensioning
system. Two 44 yarn tendons, each tensioned to 10 kN, were used. New loading plates
were fabricated which featured a number of modifications in order to connect the two
frames together and provide tensile resistance to prevent outward slipping of the frames.

Figure 5.3 - Dead End Loading Plates

As shown in Figure 5.3, inclined slots 3 mm deep were machined into the loading plates
into which the ceramic frames could fit. These provided a bearing surface to laterally
restrain the frames.

The idea of using single loading plates for both


frames was rejected, since a large amount of
unnecessary steel would have been required to
bridge the 150 mm gap between the centers of
each frame. Holes were drilled through the
loading plates at each end in order to allow
two lengths of M5 studding to connect the two
pieces. This also allowed minor adjustments to
be made to the separation and inclination of
each loading plate, should this have been
Figure 5.4 - Live End Assembly
required to achieve a perfect fit.
Will Hawkins, Churchill College 43
Prestressed Ceramics

Figure 5.4 shows the assembled live end connections. New smaller outer bearing plates
were constructed (which do not make contact with the ground) in order to further
minimise steel use.

5.2.3 Loading Points


The force was applied to the top of the structure via a 20mm thick loading plate, which
can be seen in Figure 5.5. It was initially intended to load the structure from beneath
using a length of M16 studding as pictured, but this was later simplified.

Figure 5.5 - Top Loading Plate Figure 5.6 - Attached Bearing Pad and Tendon Duct

It was vital to ensure that the load was applied precisely in-line with the plane of each
frame to avoid reducing the buckling capacity of the compression struts through load
eccentricities. An inclined bearing surface was hence created using a CNC process. The
loading plate was fixed to the tile frames using a thick layer of Araldite® in order to
eliminate stress concentrations.

Figure 5.6 shows one of the 40 mm long steel bearing pads which were fixed to either end
the frames. These were machined to the correct inclination to create a horizontal bearing
surface, and also to eliminate stress concentrations which may have lead to local crushing
of the tiles.

5.3 Test Set-up & Procedure


Figure 5.7 shows the structure as set up for testing. The load was applied through a ball
joint to ensure no bending could arise. A pair of rollers spaced at 880 mm provided
support. The bearing pads were placed on flat steel bars at either end to maximise contact
area.
Will Hawkins, Churchill College 44
Prestressed Ceramics

The test was displacement controlled, at a rate of approximately 0.8 mm per minute. Load
and displacement were recorded at one second intervals by a data logger. The structure
was loaded until failure.

Figure 5.7 - Test Setup

5.4 Theoretical Predictions

5.4.1 Deflection
Appendix B shows how the top deflection can be related to the applied load, assuming the
truss to be pinned and the supports free to slide outward. In the uncracked state, both the
tension and compression members were assumed to have the stiffness of the ceramic
section. Assuming the value of Young's Modulus from preliminary testing, this gives:

It was assumed that the tension members would crack when the prestress had been
overcome and the tensile strength reached. An estimate of the load at which this occurs
can be made, although this depends heavily on the tensile strength of the ceramic. Taking
this to be the bending strength of 76.27 MPa and assuming full prestress we have:

This corresponds to a total load Q of 116.7 kN, but is likely to be an overestimate since
the true tensile strength will be considerably less than the bending strength. If zero
tensile strength is assumed, the predicted cracking load is reduced to 22.7 kN.

After cracking, the stiffness of the tension members reduces to that of the tendon:
Will Hawkins, Churchill College 45
Prestressed Ceramics

Hence bi-linear behaviour was predicted, whereby the stiffness of the structure was
expected to reduce on cracking of the tension member.

5.4.2 Failure
One possible mechanism of failure is breaking of a tendon. Assuming 44 tendons with a
strength of 444.6 N each, the total tendon strength is 19.6 kN and the expected failure
load would be 44.4 kN. It is likely that individual tendons would fail before this load is
reached, as had been observed previously.

Another possible failure mode involves minor-axis buckling of a compressive strut.


Assuming the strut to be pin-ended (Leff =531.2 mm) with a second moment of area of
17,148 mm4, the Euler buckling load can be calculated:

This corresponds to a load of 85.8 kN. Eccentricities in load, material heterogeneities and
geometric imperfections can be expected to reduce the true buckling load.

5.5. Results & Discussion


Figure 5.8 shows the recorded results as well as the theoretical predictions for cracked
and uncracked behaviour.

Figure 5.8 - Truss Test Results and Comparison with Theory


Will Hawkins, Churchill College 46
Prestressed Ceramics

Some non-linearity was observed up to around 5 kN, likely due to settling of the supports.
The uncracked gradient was lower than that predicted. This may indicate that the
assumed Young's Modulus of 72.83 GPa was an overestimate, as was also suggested by
earlier results from Beam 2.

Four cracks appeared in the structure,


initiating at the inside corners of the end joints.
These are clearly visible in Figure 5.9. First
cracking was audible at approximately 18.5
kN, corresponding to a tensile force of only 8.2
kN in the horizontal struts. The first crack
became visible at 22.4 kN, at which point the
tension would have been 9.9 kN. This closely
matches the intended amount of prestress. It is
possible therefore that cracking occurred
before all the prestress had been unloaded. Figure 5.9 - Cracking and Tendon Failure

Some bending stresses would have arisen at this joint as the structure deformed, and
would have been greatly amplified by stress concentrations at the sharp inner corners.
The section was therefore deemed to be cracked prior to full unloading of the prestress in
the tension members, and the tensile capacity of the ceramic was hence never mobilised.
This hypothesis is supported by the fact that no drop in load was observed upon cracking,
as would have been expected if tensile stresses had been relieved.

After the cracks had opened, a reduction in stiffness was observed. The measured cracked
gradient was larger than predicted, possibly due to the rotational stiffness of the tiles at
the top connection.

At a load of 25.0 kN, a single yarn of one tendon failed at the live end. This can be seen in
Figure 5.9. A small drop in load is visible in the results at this point due to the reduction
in stiffness. The total force in the tendon at this point was calculated to be 11.0 kN. Since
all other yarns remained intact throughout the test, the failure is likely to have been
caused by localised damage or pinching.
Will Hawkins, Churchill College 47
Prestressed Ceramics

Ultimate failure occurred at 33.9 kN, and was sudden. Figure 5.10 shows one of the
compression struts suddenly breaking at both ends. The compression was 17.3 kN in the
strut at failure. This is significantly lower than the predicted Euler buckling load, and it is
clear that minor axis buckling has not taken place. The rotation of the piece suggests that
bending moments were present at each end. It is possible that stress concentrations (in
the same location as caused first
cracking) lead to local failure. Each
of the compression members failed
in the same locations, likely due to
progressive collapse. The end
connections and bearing pads were
successful in laterally restraining
the frames.
Figure 5.10 - Truss Immediately After Failure

Additional bending moments may have arisen at the supports due to an eccentricity in
support load. The apex of the roller support was beneath the mid-point of the bearing
pad, but the centerlines of the two struts met at the end of the ceramic. There is therefore
a 20 mm offset which would have resulted in a hogging moment, appearing to match the
failure mode.

6. Conclusions
6.1 Main Findings
Investigations throughout this project have confirmed that prestressing is effective in
improving the strength of ceramic components, and that its effects can be theoretically
modelled with a good degree of accuracy by making reasonable simplifying assumptions
with reliable material data.

The use of aramid fibres and the post-tensioning system developed was instrumental in
allowing a wide range of investigations to be conducted. This was largely because the
tendon was highly flexible and could be incorporated into structures without limitations
on size, shape or construction. The system requires no specialist tools to use and could be
a feasible solution for small-scale prestressing applications. The prestressing force could
be applied with some accuracy from measurement of the tendon extension, but this could
be improved by the incorporation of a load cell (strain gauges) into the loading plates.
Will Hawkins, Churchill College 48
Prestressed Ceramics

Results from the glass beams showed that the effects of prestress and eccentricity were in
accordance with theory. The strength was dictated by the prestress and ductility was
successfully enabled by bonding of the tendon. The use of glass enabled excellent
observation of cracking, providing insight into the behaviour of the structure. In the
majority of beams there was a marked distinction in behaviour between the cracked and
uncracked state, since extensive cracking occurred instantaneously. In this respect, the
behaviour of prestressed glass contrasts with that of concrete where tensile cracks
extend in a more stable manner. This may be because the heterogeneities found in
concrete (which help to stabilise crack growth) do not exist in ceramic materials.

The glass beams could not be considered usable in a cracked state since unloading and
reloading resulted in deterioration of strength and stiffness. The ductility resulting from
bonding could still serve a useful purpose however since ultimate failure could
potentially be avoided in an overload situation.

Prestressing of the ceramic truss created an efficient structure which performed well
during testing. Modeling the truss as pin-jointed provided a useful insight into the
observed behaviour, which also gave clear indications of design modifications which
would further improve performance.

Inconsistencies in Young's Modulus values were notable between preliminary testing and
the results from Beams 2 and 3. This could have partly been a result of the simplifying
assumptions of the models, but does suggest some inaccuracy in results. This highlights
the importance of confirming the reliability of values obtained in preliminary testing.

6.2 Recommendations for Further Work


The pearl-chain concept of unbonded segmental prestressed structures was
demonstrated to be feasible by Beam 2, and the simplified theoretical analysis provided a
reliable insight into its behaviour. This opens up an avenue of possible investigations. By
shaping the segments accordingly, beams of variable section and tendon eccentricity
could be created as well as arches, truss structures or even large prestressed meshes.
Will Hawkins, Churchill College 49
Prestressed Ceramics

Further investigation into the failure mode of the glass beams could be conducted. If the
proposed failure mode is correct, then restraining the compression zone against lateral
movement would lead to an improvement in strength and increased maximum
curvatures. This could be achieved by wrapping the beam in a membrane with a tensile
capacity, by including additional connection points between the two lengths of glass, or
by the addition of a top flange.

Improvements to the truss structure could be made by redesigning the ceramic parts to
avoid stress concentrations. Adding material at the corners and joints may prove
effective. The steel components could be reused since none were damaged during testing.

The failures of the structures tested in this


project were governed by fracture mechanics.
Interesting fracture patterns were observed in
a number of glass fragments, as is shown in
Figure 6.1. Further investigations into the
mechanisms behind this could yield
interesting results and lead to the improved
Figure 6.1 - Distinctive Glass Fracture Surface
deign of brittle structures against failure. Observed After Failure of Beam 3.3

Appendices:
A - Uncracked and Cracked Elastic Beam Theory
B - Theoretical Predictions of Truss Behaviour
C - Risk Assessment Retrospective
Will Hawkins, Churchill College 50
Prestressed Ceramics

References

Ashby, M. F. (2011). "Materials Selection in Mechanical Design", 4th Edition, Butterworth-


Heinemann

Du Pont, E. I. (1991). "Data manual for fibre optics and other cables", E. I. Du Pont de
Nemours and Co. (Inc.)

Causier, M. (2010). "Prestressed Ceramics", 4th Year Project, Cambridge University


Engineering Department

Ericksen, R. H. (1976). "Room temperature creep of Kevlar 49/epoxy composites",


Composites, pp. 189-194

Giannopolous, I. P. (2009). "Creep and Creep-Rupture Behaviour of Aramid Fibres",


University of Cambridge, PhD Thesis

Grant, A. (2012). "Prestressed Ceramics", 4th Year Project, Cambridge University


Engineering Department

Hertz, K. D. (2009). "Super-light Concrete with Pearl-chains', Magazine of Concrete


Research, 2009, 61, No. 8, October, 655–663

Weibull, W. (1939). "A Statistical Theory of the Strength of Materials",


Ingenioersvetenskapsakademiens, Handlingar., No. 151
Will Hawkins, Churchill College 51
Prestressed Ceramics

Appendix A - Uncracked and Cracked Elastic Prestressed Beam Theory

Uncracked
Use a transformed section to calculate
section properties:

Assume linear stress distributions to calculate extreme fibre stresses (positive in


compression):

Top fibre:

Bottom Fibre:

The gradient of a moment-curvature plot is calculated as:

Cracked
For the cracked section, any ceramic in
tension is assumed ineffective. The
prestress force P is assumed to vary due
to the change in local stress, Δεt:

The compressive stress in the ceramic can be resolved into a single force:

Strain can be related to curvature through compatibility:

By equating the horizontal forces, expressions can be obtained for moment and curvature in
terms of neutral axis depth:

The curvature must then be normalised (accounting for initial curvature due to prestress
eccentricity) for comparison with results.
Will Hawkins, Churchill College 52
Prestressed Ceramics

Appendix B - Theoretical Predictions of Truss Load-Deflection Behaviour

The forces in each strut are found through geometry. and . Since
deflections are small, angles are assumed constant:

The tension member and compression struts have axial stiffness and
respectively. Using the deformation mechanism shown, the the axial change in length can be
related to the force in each member:

Combining these equations gives a relationship between the total force


Q and the deflection

Appendix C - Risk Assessment Retrospective


The initial risk assessment handed in at the beginning of the project outlined a number of
hazards associated with the application of prestress, testing of brittle structures, and
handling of hazardous materials as well as working with machinery in a workshop
environment. Since this was produced prior to the development of the post-tensioning
system, there were omissions of hazards directly related to its design. Tendon failure during
stressing was a common occurrence in early iterations of the system, but since yarns only
ever failed individually no significant releases of energy could occur and no significant
hazards were in fact posed.

An additional risk assessment was conducted prior to testing of the glass beams in order to
address the specific dangers of the potentially explosive failure of highly stressed glass. A
Perspex safety screen was built and fitted to the testing rig which was effective in protecting
observers and bystanders.

Potrebbero piacerti anche