Professional Documents
Culture Documents
Publisher
South African Institution of Civil Engineering
Block 19, Thornhill Office Park,
Bekker Street, Vorna Valley, Midrand, South Africa
Private Bag X200, Halfway House, 1685, South Africa
Tel +27 (0)11 805 5947/48, Fax +27 (0)11 805 5971
http://www.saice.org.za
civilinfo@saice.org.za
Contents
Editor-in-chief
Prof Gerhard Heymann
University of Pretoria
Tel +27 (0)12 420 3627
gerhard.heymann@up.ac.za
joint Editor-in-chief
Prof Chris Clayton
University of Southampton
2 Characterisation of rigid polyurethane foam‑reinforced
c.clayton@soton.ac.uk ballast through cyclic loading box tests
MANAGING Editor
Verelene de Koker R F du Plooy, P J Gräbe DOI: 10.17159/2309-8775/2017/v59n2a1
Tel +27 (0)11 805 5947, Cell +27 (0)83 378 3996
verelene@saice.org.za
11 Fatigue behaviour in full‑scale laboratory tests of
journal editorial Panel
Prof G Heymann – University of Pretoria a composite deck slab with PBL reinforcement
Prof CRI Clayton – University of Southampton
Prof Y Ballim – University of the Witwatersrand P Lu, X Zhan, R Zhao DOI: 10.17159/2309-8775/2017/v59n2a2
Prof W Burdzik – University of Pretoria
Dr P Day – Jones & Wagener (Pty) Ltd
Prof J du Plessis – University of Stellenbosch 19 The influence of foundation stiffness on the
Prof GC Fanourakis – University of Johannesburg
Prof M Gohnert – University of the Witwatersrand
behaviour of surface strip foundations on sand
Prof PJ Gräbe – University of Pretoria
Dr C Herold – Umfula Wempilo Consulting H E Lemmen, S W Jacobsz, E P Kearsley DOI: 10.17159/2309-8775/2017/v59n2a3
Prof A Ilemobade – University of the Witwatersrand
Prof SW Jacobsz – University of Pretoria
Prof EP Kearsley – University of Pretoria
28 Revised regional sediment yield prediction
Prof C Roth – University of Pretoria methodology for ungauged catchments in South Africa
Prof W Steyn – University of Pretoria
Mr M van Dijk – University of Pretoria V C Msadala, G R Basson DOI: 10.17159/2309-8775/2017/v59n2a4
Prof JE van Zyl – University of Cape Town
Prof C Venter – University of Pretoria
Prof A Visser – University of Pretoria 37 The two-points condensation technique (TPC) for
Dr E Vorster – Aurecon South Africa (Pty) Ltd
Prof J Wium – University of Stellenbosch detection of structural damage due to vibration
Prof A Zingoni – University of Cape Town
Prof M Zuidgeest – University of Cape Town A N Al-Qayyim, B Ö Çağlayan DOI: 10.17159/2309-8775/2017/v59n2a5
Peer reviewing
The Journal of the South African Institution of 49 The application of continuous surface wave testing
Civil Engineering is a peer-reviewed journal
that is distributed internationally for settlement analysis with reference to a full-scale
Design and reproduction
Marketing Support Services, Ashlea Gardens, Pretoria
load test for a bridge at Pont Melin, Wales, UK
Printing G Heymann, J Rigby-Jones, C A Milne DOI: 10.17159/2309-8775/2017/v59n2a6
Fishwicks, Pretoria
Papers for consideration should be submitted 59 Effect of the minimum void ratio on the vertical
online at: http://journal.saice.org.za
intercept of the steady state line of non-plastic soils
The South African Institution of Civil Engineering accepts no
responsibility for any statement made or opinion expressed L A Torres-Cruz, S Geyer, P R Mackechnie DOI: 10.17159/2309-8775/2017/v59n2a7
in this publication. Consequently, nobody connected with
the publication of this journal, in particular the proprietor,
the publisher and the editors, will be liable for any loss or
damage sustained by any reader as a result of his or her action
upon any statement or opinion published in this journal.
© South African Institution of Civil Engineering
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 1
TECHNICAL PAPER Characterisation of rigid
Journal of the South African
Institution of Civil Engineering
polyurethane foam‑reinforced
ISSN 1021-2019
Vol 59 No 2, June 2017, Pages 2–10, Paper 1486 ballast through cyclic
RUDOLPH DU PLOOY obtained a BEng degree
loading box tests
in Civil Engineering from the University of
Pretoria. He followed that up with a BEng (Hons)
in Transportation Engineering in 2014 and an R F du Plooy, P J Gräbe
MEng (Transportation) in 2016, also from the
University of Pretoria.
As train speeds and heavy haul axle loads constantly increase due to market demands, so do the
stresses and strains experienced by track structures. This is especially true for track transitions
Contact details: that generate high dynamic forces on both the track and vehicles because of poor vertical
Transnet Freight Rail Chair in Railway Engineering track geometry and/or differing track stiffness values on either side of the track transition.
Department of Civil Engineering
Reducing differential settlement between the two track structures at a track transition is one
University of Pretoria
Pretoria 0001 method of improving the life of the track, and increasing maintenance intervals. In this study,
South Africa rigid polyurethane foam was used to reinforce ballast. Tests were conducted using a dynamic
T: +27 83 288 7669 hydraulic load frame and a single sleeper in a large ballast box subjected to heavy haul axle
E: rudidu@gmail.com loads. Unreinforced, reinforced and 50% reinforced ballast layers of 300 mm depth were tested
to approximately 5 000 000 load cycles. The results showed that rigid polyurethane foam-
PROF HANNES GRÄBE (Pr Eng, FSAICE) is a civil reinforced ballast exhibited in the order of 60% less settlement for a fully reinforced layer, and
engineer with experience in the fields of track
42% less settlement for a half reinforced layer. The use of rigid polyurethane foam (RPF) to
technology, geotechnology, advanced
laboratory testing, field investigations, reinforce ballast has a number of benefits that could result in better track geometry and longer
maintenance models and numerical analysis of maintenance cycles, in turn resulting in lower life cycle costs.
track structures. He is Associate Professor:
Transnet Freight Rail Chair in Railway
Engineering at the University of Pretoria, where
INTRODUCTION sample material tests were also conducted
he lectures under- and post-graduate courses in railway engineering. He is
also responsible for railway research, as well as continuing professional The objective of this study was to study to compare the material behaviour of rigid
development (CPD) in the form of short courses presented to industry. the settlement behaviour and other track polyurethane foam and ballast reinforced
parameters through long-term, large-scale, with rigid polyurethane foam.
Contact details:
Transnet Freight Rail Chair in Railway Engineering cyclic loading box tests subjected to heavy
Department of Civil Engineering haul axle loads (30 tons). It was hypothesised
University of Pretoria that the reinforcement of ballast using rigid BACKGROUND
Pretoria 0001
polyurethane foam would lead to a reduction The most widely used track structure world-
South Africa
T: +27 12 420 4723
in ballast layer settlement, an increase in wide is the conventional or ballasted track
E: hannes.grabe@up.ac.za stiffness, reduced ballast breakage and an structure. Typically, it consists of two main
overall improvement in track performance. parts – the superstructure (i.e. rail, fastening
Two main objectives of the study were, system and sleepers) and the substructure
firstly, to characterise the properties of the (i.e. ballast, subballast and subgrade) as
ballast, polyurethane foam and the ballast- defined by Selig and Waters (1994).
polyurethane-foam composite material and, Ballastless track structures are track
secondly, to characterise the settlement structures that have been developed to
behaviour of the ballast and polyurethane mitigate the problems relating to the slow
foam-reinforced ballast layer. deterioration and subsequent settlement of
Only a ballast layer was modelled in the the ballast material as a result of traffic load-
box test with no other foundation layers. ing. Ballastless track structures are typically
Various test configurations were prepared, divided into two main categories, namely
ranging from completely unreinforced bal- ballastless track that provides continuous
last to fully reinforced ballast. Each sample support to the rail, and ballastless track
was subjected to 5 000 000 load cycles at the that provides discrete support to the rail.
specified load. Ballastless track structures can result in
Settlement prediction models were maintenance costs that are 20% to 30% lower
compared to actual ballast settlement behav- than ballasted track (Esveld 2001).
iour, and the validity of these settlement Track transitions occur where a bal-
prediction models for use with polyurethane lasted track section changes to a ballastless
reinforced ballast samples was examined. track system or a ballasted track system
Resilient sleeper deflection, ballast layer on a structure. The abrupt change in track
Keywords: ballast, concrete sleepers, ballast settlement, ballast breakdown, stiffness and ballast layer strain were all support that occurs at these locations has
rigid polyurethane foam, cyclic loading determined and compared. A number of often been associated with accelerated rates
2
Du Plooy RF, Gräbe PJ. Characterisation of rigid polyurethane foam-reinforced ballast through cyclic loading box tests.
J. S. Afr. Inst. Civ. Eng. 2017:59(2), Art. #1486, 9 pages. http://dx.doi.org/10.17159/2309-8775/2017/v59n2a1
of track geometry and component degrada- 1
tion, which in turn can lead to poor ride A A d95i
quality and increased maintenance demand BBI =
(Read & Li 2006). When maintenance at B A+B
ge
The problems at track transitions can be
ka
ea
attributed to three primary factors (Sasaoka
br
um
Fraction passing
& Davis 2005):
im
ax
■■ Differential settlement
fm
■■ Differences in settlement characteristics
yo
ar
■■ Discrepancies in track damping proper-
nd
ou
ties between adjacent sections.
yb
ar
Nicks (2009) categorised remedial measures
it r
rb
aimed at reducing track transition impact
A
forces as follows:
■■ Limit approach track structure settlement.
■■ Decrease modulus on bridge deck/ballast
less track structure.
■■ Increase the modulus for the approach Initial PSD
track.
■■ Reduce ballast wear and movement. Final PSD
■■ Increase damping on bridge deck/ballast-
0
less track structure. 0 2.36 53
When ballast has been in service for a sig- Sieve size (mm)
nificant period of time, it becomes damaged
and contaminated, resulting in a change in Figure 1 Definition of ballast breakage index (Indraratna et al 2005)
the grading of the ballast which in turn leads
to a reduction in performance. The process Ballast breakdown can be quantified The variation in settlement behaviour
through which this occurs is known as by using a parameter known as the Ballast between uncompacted and compacted
fouling. Selig and Waters (1994) list the five Breakage Index (BBI), and the ballast breakage ballast samples was compared, with the
causes of ballast fouling as: can be quantified by comparing the plots of uncompacted ballast settling significantly
■■ Ballast breakdown particle size distributions. As the number more than the compacted ballast, along
■■ Infiltration from ballast surface of breakages increase, the particle size dis- with the uncompacted ballast having greater
■■ Sleeper wear tribution shifts to the left of the graph (see first cycle axial strain development (2.8%)
■■ Infiltration from underlying granular layers Figure 1), and the area between this new line compared to 0.58% for the first cycle of the
■■ Subgrade infiltration. and the original particle size distribution is compacted ballast. Each additional cycle of
Ballast fouling prevents the ballast from considered as the breakage zone. Indraratna loading caused an increment of plastic strain,
fulfilling its functions, and the extent of this et al (2005) state that the breakage potential is but at a diminishing rate, varying from 0.24%
loss in function is dependent on the amount considered to be the area between the original after the first cycle to 0.001% after approxi-
and size of the fouling material. As the mass particle size distribution and an arbitrary mately 50 cycles (Indraratna et al 1997).
of sand and fine-gravel-sized fouling particles reference line connecting the point between A number of equations have been pro-
(0.075 mm to 19 mm) increases, the resilience the intersection of d95 of the largest sieve size posed to describe the settlement behaviour
to vertical deformation, as well as the void and the minimum particle size of 2.36 mm. of ballast. Selig and Waters (1994) compared
space, decreases. Fouled ballast is gap-graded. Figure 1 shows the definition of the BBI a number of settlement prediction equations
This results in a reduction in drainage and parameters which are used in Equation 1. and found that, of the semi-log, hyperbolic,
could make surface lining (i.e. tamping) parabolic and power relationships, the best
operations difficult. As the void space is filled, A overall trend was represented by the power
BBI = (1)
the density of the ballast material increases. A+B relationship of the form shown in Equation 2
Should this material then be tamped, a higher and the logarithmic relationship shown in
rate of ballast settlement can be expected The permanent deformation behaviour of bal- Equation 3.
after tamping. An increase in the mass of clay last under cyclic loading is usually in the form
and silt-sized particles (particle sizes smaller of settlement. The settlement behaviour can SN = S1 N b(2)
than 0.075 mm) also results in reduced drain- be both elastic (such as initial settlement due
age, leading to ballast erosion and subgrade to compaction) and plastic (due to breakage of SN = S1(a log N + 1) (3)
attrition. When mixed with water, fine ballast particles). The number of load cycles
particles may form a slurry that is abrasive to also has an effect on the permanent deforma- Where:
the ballast material. Fouled ballast with high tion, with the permanent deformation being SN = settlement after cycles
water content leads to higher rates of plastic a function of the logarithm of the number of S1 = settlement after the first cycle
strain in the fouled ballast. cycles (Shenton 1974). a and b are shape parameters.
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 3
Granular materials, such as railway ballast, (i.e. ballast, subballast and subgrade) and no diisocyanate (IsoPMDI 92140). The material
experience problems over poor formations, negative effect on the overall elastic response properties of the rigid polyurethane foam
due to the development of tensile strains under loading due to the change in stiff- (RPF) and the ballast material were deter-
at the formation interfaces. The continued ness of the polyurethane reinforced areas. mined prior to testing. An initial grading
rotation of the principle stresses causes bal- Specific material properties and structural analysis and a loose bulk density determina-
last densification, and hence settlement. This components of the track system appear to tion were conducted on the ballast. RPF
increases over soft formations as a result of have far greater influence on substructure cylinders, and RPF reinforced ballast cylinders
low track stiffness, which then induces bal- elastic strain and track modulus than the were cast and their load-deflection behaviour
last plastic strains. One method of improving integration of polyurethane reinforcement was tested in a universal testing machine. The
the ballast layer, and reducing the amount of into the ballast layer. ballast material properties are summarised
plastic strain and settlement is to reinforce Keene et al (2012b) report that, while in Table 1, with a summary in Table 2 of the
the ballast layer (Woodward et al 2009). rigid-compact polyurethane in rail infrastruc- various foam material properties.
A method of reinforcing or stabilising the ture has been used, very few experimental and
ballast layer is the use of polyurethane foam. empirical methods have been developed for Table 1 Ballast material properties
Keene et al (2012a) conducted a study into ascertaining the mechanical properties and
Ballast Material Quartzite
the characterisation of rigid polyurethane life cycle characteristics of rail substructures
foam (RPF) reinforced ballast. RPF is an that have been stabilised with polyurethane. ACV (%) 14.5
expanding polyurethane which is placed The investigation into the injection of RPF Water absorption (%) 0.3
in the ballast layer (either through pouring into the rail substructure is an uncharted
Loose bulk density (kg/m3) 1 511
or injection) and is then allowed to fill the area. Standard laboratory tests for the fabrica-
ballast layer voids, preventing the infiltra- tion and characterisation of polyurethane Void ratio (loose) 0.857
tion of water and other contaminants. The stabilised ballast (PSB) are aimed at contribut-
Porosity (loose) 0.461
RPF would further protect the structure ing to research infrastructure. The overall
from rearrangement and settlement. RPF objective is to reduce maintenance life cycle
is a cross-linked, closed-celled, thermoset costs, increase rail freight load capacity and Table 2 Rigid polyurethane foam properties
material with a low density, and a number of provide maintenance techniques that are not
RPF density (kg/m3) 36.9
the RPF material’s properties are dependent disruptive to railroad traffic.
on the density. RPF is typically supplied in PRB density (kg/m3) 1 135.5
two separate components – an A-component RPF Compressive strength 138 kPa – 140 kPa
(polyether polyol) and a B-component (poly- EXPERIMENTAL WORK
Mixing ratio (A:B) 100:107
isocyanate) are then mixed and applied. The characteristics of the foam and ballast
Keene et al (2012b) summarised the materials that were subjected to cyclic load-
mechanical property results of the work ing box tests and the details of the experi- Cyclic loading ballast box tests
on polyurethane stabilised ballast (PSB) as mental work are described in the following A hydraulic MTS load frame with a maxi-
follows: sections of the paper. mum actuator capacity of 500 kN was used
■■ PSB outperforms other track-substruc- to apply the cyclic loading to the ballast
ture materials. Material properties samples. A large steel box was constructed,
■■ PSB typically had higher elastic deforma- The rigid polyurethane foam that was used to with internal dimensions of 2 400 mm long,
tional behaviour. reinforce the ballast was Elastopor® H 1311/1. 600 mm wide and 400 mm high. Each test
Furthermore, Keene et al (2012b) concluded The foam is supplied in two separate com- consisted of a 300 mm ballast layer with a
the following with regard to the feasibility ponents, namely a Polyol-component (A) and PY-sleeper for ballast loading. Linear variable
of using polyurethanes to stabilise track an Iso-component (B). The polyol component differential transducers (LVDTs), with a full
substructure: is a mixture of polyetherpolyolls, stabiliser, scale of 20 mm, were mounted onto the test
■■ Stabilisation does not have a negative catalyst, flame retardant and water. The iso- frame at each sleeper end for local sleeper
impact on elastic response. component is polymeric diphenylmethane displacement measurement in addition to the
■■ Injection methods that are cur-
rently employed are feasible for track
stabilisation.
■■ PSB can greatly increase the track mecha-
nistic life cycle.
Keene et al (2013) conducted numerical
modelling into the effect of polyurethane
stabilisation on rail track response. The
numerical model was used to determine the
effects of various parameters, such as loca-
tion, thickness and polyurethane properties,
on the ballast layer and how resilient behav-
iour is influenced. A larger range of PSB
moduli than observed in the laboratory was
also incorporated into the numerical model.
The simulations by Keene et al (2013)
showed that there are minimal changes
in the strain of each substructure layer Figure 2 B
allast box test setup before being placed into the hydraulic load frame
4 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
internal actuator displacement measurement.
The setup of a single sleeper on a 300 mm
ballast layer in a steel box, simulating a sec-
tion of track, is shown in Figure 2.
Support column Each test was divided into primarily two
phases, namely the initial phase and the
consolidation phase. During the initial phase,
Crosshead large settlement values were recorded as a
result of the settlement of the uncompacted
Load frame swivel/hinge
ballast and the tests had to be paused to
Hydraulic actuator readjust the LVDTs to their measuring range.
The loading during the initial phase consist-
Load cell
ed of a number of sub-phases where the load
Loading beam
and loading frequency were incrementally
LVDT for sleeper displacement
measurement increased during the initial 57 000 cycles.
Loading block The final test setup is shown in Figure 3.
LVDT support frame The second phase of the test consisted of
Ballast box the consolidation phase where a sinusoidal,
cyclic 260 kN load (30 ton/axle loading) was
applied to the sleeper for a total of 5 000 000
cycles. The magnitude of this cyclic load
was obtained using the AREMA Manual of
Railway Engineering standard specification.
Load frame test bed A maximum rail seat load of 60% of the
axle load is recommended for a pre-stressed
Hydraulic ram to raise/lower
crosshead concrete sleeper at 760 mm sleeper spacing
(Figure 30.4.1 in AREMA 2010). A dynamic
factor of 1.46 was applied to the static load of
147 kN. A total of four tests were conducted,
Figure 3 T est apparatus and test setup all at a test frequency of 10 Hz. The complete
test record, summarising the various tests, is
Table 3 Test record summary shown in Table 3.
During initial sample preparation it was
Test Type Comment Frequency (Hz) Cycles (No)
discovered that, without restricting the
1 Reinforced ballast Free expansion 10 5 058 308 movement of the ballast, the foam would
expand and push the ballast particles apart.
2 Reinforced ballast Expansion limited 10 5 040 036
In Test 1, the 300 mm ballast layer was
3 Unreinforced ballast – 10 4 787 750 reinforced using RPF, and the expansion of
the ballast and foam was not limited after
4 Reinforced ballast (50%) Expansion limited 10 4 967 957
pouring. During the preparation of Test 2,
the expansion of the RPF in the ballast layer
was limited by the placement of a sleeper on
top of the ballast layer during the pouring
process. Test 3 was a ballast test without
any reinforcement applied. Test 4 was
conducted using 50% reinforcement (i.e.
bottom 150 mm reinforced and top 150 mm
unreinforced).
Figure 4 shows the complete void filling
of the RPF of the ballast layer to produce the
reinforced ballast composite structure.
Ballast settlement
The settlement results from the initial phase
of testing are shown in Figure 5. A significant
amount of settlement occurred in the initial
phase of testing, due to the ballast layer being
uncompacted at the start. From Figure 5 it
Figure 4 F ully reinforced ballast layer sample (upside down) after testing, showing full void filling can be observed that the unreinforced ballast
by RPF (UR) experienced the greatest amount of
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 5
settlement, as the particles were free to settle
and undergo reorientation. The reinforced 0
–20
expansion, where a total combined settle-
ment of 63.0 mm was observed. The unrein-
–30
forced ballast layer exhibited the second larg-
est combined settlement of 55.6 mm, the vast
majority of which occurred during the initial –40
phase (92.9%). Figure 7 shows the combined
settlement results, with the 50% reinforced –50
ballast layer showing the second lowest total
settlement (32.0 mm). The reinforced ballast –50
layer with the RPF expansion limited showed
the lowest combined settlement, with a total 0 1 × 10 6 2 × 10 6 3 × 10 6 4 × 10 6 5 × 10 6 6 × 10 6
settlement of 22.3 mm. Cycles
The use of RPF as a means of reducing
ballast settlement is a possibility, provided Figure 7 T otal settlement results of the polyurethane stabilised ballast samples
6 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
70 that the expansion of the foam into the voids
of the ballast does not result in significant
60 ballast particle uplift which could remove
the inter-particle contact between the stones.
50
Settlement (mm)
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 7
power functions were not adequate and the Table 4 Comparison of R2 values for different test phases (consolidation and combined phases)
logarithmic prediction function gave better
Initial and
correlation. The lowest R2 value for the total Initial Phase
Consolidation Phase
Test
settlement logarithmic prediction functions
Power Logarithmic Power Logarithmic
was 0.77, and for the same test the R2 value
for the power function was 0.22. A summary Fully reinforced – free expansion 0.81 0.63 0.38 0.88
of these results is provided in Table 4.
Fully reinforced – expansion limited 0.77 0.43 0.64 0.96
From these results, and from those in the
previous section, it would appear that when Unreinforced 0.78 0.31 0.22 0.77
there is significant initial settlement in the
Semi-reinforced 0.97 0.42 0.29 0.88
early load cycles (such as in uncompacted
ballast), the logarithmic prediction function
would be the ideal relationship to use, as it
provides the closest estimate in these cases.
For test samples where initial settlement 220
is not that significant, the power function
tended to provide superior results. For this 200
reason, the prediction function chosen
should be selected based on the present
Resiliant modulus (MPa)
180
testing conditions and behaviour of the
test sample. 160
8 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
1.4 120 with time could be used to aid engineers
in the design of track transitions or as
1.3
a solution to problem track transition
110 sections.
■■ The power and logarithmic function
1.2
Resiliant sleeper deflection (mm)
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 9
Indraratna, B, Lackenby, J & Christie, J 2005. Effect Keene, A, Tuncer, E, Fratta, D & Tinjum, J 2013. Selig, E T & Waters, J M 1994. Track Geotechnology and
of confining pressure on the degradation of ballast Modeling the effect of polyurethane stabilization on Substructure Management, 1st ed. London: Thomas
under cyclic loading. Geotechnique, 55(4): 325–328. rail track response. Proceedings, Geo-Congress 2013, Telford Publications.
Keene, A K, Edil, T B, Tinjum, J M & Brown, R W 3–7 March 2013, San Diego, CA. Shenton, M J 1974. Deformation of railway ballast
2012a. Characterization of polyurethane-stabilized Nicks, J 2009. The bump at the end of the railway under repeated loading conditions. In: Kerr, A D
ballast. Paper presented at the 3rd International bridge. PhD thesis, College Station, TX: Texas A&M (Ed.), Railroad Track Mechanics and Technology, 1st
Conference on New Developments in Soil Mechanics University. ed. Proceedings of a Symposium held at Princeton
and Geotechnical Engineering, Nicosia, North Read, D & Li, D 2006. Design of track transitions. University, NJ, 21–23 April 1975.
Cyprus. Transportation Research Cooperative Program: Woodward, P, Kennedy, J & Medero, G 2009. Three-
Keene, A, Tinjum, J, Tuncer, E & Brown, R 2012b. Research Results Digest, 79(October): 1–36. dimensional track reinforcement using polymer
Railway substructure stabilization with polyurethane Sasaoka, C & Davis, D 2005. Implementing track geocomposites. Proceedings, Annual Convention
injections. Paper presented at the Mid-Continent transition solutions for heavy axle load service. of the American Railway Engineering and
Transportation Research Forum 2012, 6–7 Proceedings, AREMA Annual Conference, 25–28 Maintenance-of-Way Association (AREMA),
September 2012, Madison, WI. September 2005, Chicago, IL. Chicago, IL.
10 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
Fatigue behaviour in TECHNICAL PAPER
full‑scale laboratory tests Journal of the South African
Institution of Civil Engineering
of a composite deck slab ISSN 1021-2019
Vol 59 No 2, June 2017, Pages 11–18, Paper 1298
known as a perfobond (PBL) shear connector. Since the more serious effects of fatigue load- Contact details:
A perforated-stiffened-plate composite bridge ing have yet to appear, the fatigue problem has Faculty of Civil Engineering and Architecture
Zhejiang University of Technology
deck slab combines the advantages of steel not been sufficiently addressed, nor systemati-
Hangzhou 310014
and concrete in a general composite deck slab. cally researched. Fatigue damage and failure PR China
For bridge construction, a steel plate is used decrease the reinforcement provided by the T: +86 0571 88320153
as permanent formwork. This design saves steel in a new bridge deck, which reduces the E: zhanxl@zjut.edu.cn
time during the installation and removal of superstructure stiffness while increasing slip-
the scaffolding setup formwork, and reduces ping. These effects seriously affect the vehicle PROF RENDA ZHAO is at the School of Civil
the costs in relation to the construction cost capacity of the bridge structure. Therefore, Engineering, Southwest Jiaotong University,
China. He is also a vice-president of the China
of the whole steel beam structure. The steel an assessment of the fatigue performance of a
Institute of Civil Engineering. He obtained his
plate and concrete slab are connected by PBL PBL-shear-connected steel and concrete com- BSc degree from Chongqin University, and his
shear connectors to improve the composite posite slab at highway loads has important MSc from Southwest Jiaotong University. He has
effect of the slab, enhance the slab’s stiffness academic and design potential, and will serve a broad interest in structural behaviour and
and mechanical performance, and reduce as a basis for future strategies for preventing experimental work.
deformation and slippage. The composite or reducing damage due to fatigue loading. Contact details:
bridge deck slab – in conjunction with PBL Few studies of perforated-stiffened-plate School of Civil Engineering
shear connectors – is widely used in the deck slabs have been published. Ryu et al Southwest Jiaotong University
Chengdou 610011
construction of new bridges, maintenance (2007) conducted a full-scale model test of
PR China
activities and the reconstruction of existing a two-span continuous bridge deck slab to T: +86 028 87601333
bridges. Moreover, the composite slab has study the crack development of a fatigue- E: rendazhao@163.com
technical and economic advantages, and loaded composite deck slab. The results
improves material performance and fast indicated that the cracks in a region with a Keywords: steel truss arch bridge, steel-concrete composite deck slab,
construction compared to other promising negative bending moment were controlled fatigue performance
materials. However, despite the increasing use within an allowable crack width for a specific
of PBL-shear-connected steel and concrete fatigue load. The steel shuttering described
composite slabs, the bridges in which they are in their paper was a profiled sheet; compared
used are not immune to the long-term effects with steel-concrete bridge deck slabs, the
of moving loads. In particular, compared with perforated stiffened plates had a different
11
Lu P, Zhan X, Zhao R. Fatigue behaviour in full-scale laboratory tests of a composite deck slab with PBL reinforcement.
J. S. Afr. Inst. Civ. Eng. 2017:59(2), Art. #1298, 8 pages. http://dx.doi.org/10.17159/2309-8775/2017/v59n2a2
arrangement. To verify the composite effect on the fatigue life. Liu et al (2012) proposed
of the steel shuttering and the perforated a new composite bridge deck system and Region of model experiment
steel plate combined with concrete, a test of a performed a theoretical analysis including a
full-scale model of a perforated-stiffened plate full-scale model test of a bridge deck. Their
deck slab was conducted by Kim and Jeong results showed that after 2 million cycles of
(2006). Their results showed that the perfo- fatigue loading at a load frequency of 5 Hz,
rated and stiffened plate effectively enhanced no new cracks appeared in the composite
the composite effect. Their study primarily deck, nor had the initial cracks expanded.
analysed the effect of the perforated steel Additionally, the stiffness of the test beam had
plate for the entire composite bridge deck slab. not decreased, indicating that this composite Figure 1 E levation of bridge, bridge floor
Leitão (2011) developed a numerical model for bridge deck system had satisfactory fatigue system and location of model test
the dynamic analysis of composite highway properties. Zong and Che (2000) performed
bridges using a finite-element method with a fatigue test on a simply-supported continu- of this study have important theoretical value
mesh-refining technology, and achieved ous pre-stressed composite beam. They also and practical value for the optimisation,
satisfactory results when assessing fatigue analysed pre-stressed composite beams design and theoretical analysis of composite
behaviour. To evaluate cracking in the vertical constructed from different types of concrete deck slabs.
and horizontal joints of a composite bridge at different magnitudes and orders of pre-
deck, Chang and Shim (2001) performed stressing. They drew some valuable conclu-
fatigue behaviour tests of the composite sions by summarising the fatigue test results TEST MODEL
connection points of a continuous composite of the pre-stressed composite beams, discuss-
bridge. Their study also discussed methods ing the maintenance of the steel members and Engineering background
of longitudinal pre-stressing. Although they stud connectors, and proposing a principle for for the fatigue tests
did not conduct a thorough analysis of the the corresponding fatigue strengths. The main span of the Dongping Bridge
effects of fatigue at the connection points on For the fatigue of steel-concrete composite (43.5 m + 95.5 m + 300 m + 95.5 m + 43.5 m)
the entire composite bridge deck slab, their bridge decks, the theoretical analyses and in Guangdong is a half-through steel truss
results provide a reference for the study of the experimental results provided by the above- arch bridge with a full length of 1 322.2 m.
fatigue behaviour of perforated steel plates, mentioned studies are in agreement regarding The steel boxes and concrete slabs are con-
including the junctions of the underside fatigue-loaded steel-concrete composite nected by PBL shear connectors. The grid
and side steel plates. Allahyari et al (2014) bridge decks. However, the individual behav- beams consist of three main longitudinal
investigated the behaviour of bridge decks iour of various combinations of steel-concrete girders, secondary longitudinal girders, main
for a static load applied to the centre of the composite bridge decks is quite different, i.e. beams and secondary beams. Perforated-
deck. To evaluate the dynamic properties of different combinations of materials and forms stiffened-plate composite bridge deck slabs
the decks, they experimentally investigated lead to different forms of fatigue behaviour. were erected on the grid beams. Each slab
the dynamic properties of exodermic bridge Additionally, fatigue test results are discrete. has a minimum thickness of 12 cm and a
decks with alternative PBL shear connectors. Therefore, many tests are needed to explore maximum thickness of 20 cm. The overall
Millanes et al (2014) investigated the design the mechanical behaviour of perforated, stiff- layout of the Dongping Bridge and the test-
of a composite steel-concrete deck for a long ened composite slabs. There are relatively few ing zone of the composite bridge deck are
railway bridge. Gara et al (2013) investigated studies of fatigue performance and the cor- shown in Figure 1.
the effectiveness of various casting techniques responding design for novel composite bridge The Dongping Bridge uses a composite
used to control the tensile stresses in the slab deck slabs. Many theoretical and technical deck-binding system with a space grillage
during the construction of continuous steel problems still have not been solved, such as design. Under the sustained action of moving
and concrete composite bridge decks. Leitão the failure mechanism, mechanical behaviour, loads, the longitudinal shear performance
et al (2013) carried out a fatigue analysis and load-bearing capacity, and deformation of of the perforated and stiffened composite
predicted the lifetime of composite highway novel composite bridge deck slabs under bridge deck slabs will be significantly
bridge decks under traffic loading conditions. fatigue-loading conditions. Therefore, con- degraded, and the degradation in the per-
Wang and Jiang (2007) reviewed the sidering the variety of steel-concrete bridge formance of the PBL shear connectors will
fatigue problem of composite structures and decks and fatigue problems, the present study reduce the composite effect of the composite
the direction of future studies on composite aims to evaluate the fatigue performance of bridge deck slabs. As a consequence, the
structures. They argued that future studies composite bridge decks under highway load- load-bearing capacity and stiffness of the
should continue to focus on the mechanism ing conditions using the Dongping Bridge in composite bridge deck may be reduced,
of fatigue damage accumulation, damage Guangdong as a case study. which will affect the mechanical behaviour.
identification, and fatigue reliability. Yang et In this study, we conducted a model test Under such conditions, fatigue failure of the
al (2012) performed mechanical performance and simulation analysis to evaluate the fatigue composite bridge decks will occur once the
tests of a fatigue-loaded composite beam of a performance of a perforated stiffened deck fatigue damage has escalated to a certain
composite bridge deck. Their results indicated slab at a highway load. We also propose some point. To assess the fatigue performance of
that the fatigue failure mode of the composite important indicators of fatigue performance, the composite decks of the Dongping Bridge,
beam specimens with a positive bending provide a reference for composite materials we performed fatigue experiments and a
moment was the crushing of the concrete and forms, and assess their superior mechani- numerical simulation in an area comprising
in the compression zone resulting from cal behaviour to validate the effectiveness positive and negative bending moments.
fatigue damage to the lowest steel beam. The of the proposed deck system for bridge
fatigue life was directly related to the range applications. Our study contributes to the Actual bridge model selection
of fatigue stresses; however, the upper and understanding of fatigue-induced damage or Domestic and international research indi-
lower limits of fatigue loading had little effect failure of bridge deck structures. The results cates that the stress amplitude Δσ and the
12 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
bending moments. The elevation and plan
Steel Headed stud PBL shear Concrete
bottom shear connector connector slab of the model structure described above are
shown in Figures 2(a) and 2(b).
p p
p and the top flange of the secondary beams
Loading position
was 706 mm.
2 400
1 940
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 13
According to comparative results of the Table 1 Equivalent moment range (kN.m)
four principles, the size of Model B was
China’s China’s China’s
determined to be 2 146 mm. The size of bridge design bridge design bridge design BS5400
Fatigue load BS5400 AASHTO
Model A is 2 400 mm × 5 000 mm, and the specifications specifications specifications Axle load
200 kN 300 kN 550 kN
size of Model B is 2 400 mm × 6 000 mm.
The cross-sections of Models A and B The equivalent
and the loading positions are shown in bending moment 88.2 124.45 86.09 120.30 154.50 68.36
value
Figures 2(c) and 2(d).
3
DESIGN OF THE FATIGUE
MODEL TEST
2
Determination and theoretical
analysis of fatigue loads 1
The stress conditions of the actual concrete
plate were the focus of the present study. First, 0
we ensured that the stress conditions of the
Stress (MPa)
14 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
0.10 model was simplified as a four-span continu-
ous beam. The amplitude of the equivalent
bending moment of the middle cross-section
for the second span is listed in Table 1.
0.05
According to a comparison of the results for
different fatigue loads in Table 1, the 30 t
fatigue vehicle load specified by China was
Stress (MPa)
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 15
following numbers of fatigue loading cycles: 16
10 000, 50 000, 250 000, 500 000, 1 million,
1.5 million and 2 million. The test programme
14
of Model B was similar to that of Model A.
The measuring points and numbers for
Models A and B are listed in Table 4. 12
10
FATIGUE TEST RESULTS
Stress (MPa)
Fatigue test results for Model A 8
The basis for the test number was the proxim-
ity to the maximum deflection of 0.645 mm 6
for a load of 83.66 kN during the static load-
ing stage. A deflection analysis shows that the
4
maximum deflection occurred at measuring
point #15 under all working conditions.
Measuring point #15 is located immediately 2
below a loading point. The maximum deflec-
tion of the composite deck was 0.515 mm 0
for a load of 63.52 kN after 2 million cycles. 0 20 40 60 80 100
It should be noted that after 2 million cycles (kN)
of fatigue loading, Model A had yet to show 0 2 million times 3 million times
evidence of fatigue cracking. Additionally, the 10 000 times 2.1 million times theoretical values
stiffness was not significantly reduced. To
further investigate the fatigue performance Figure 5 Curve of load stress for PBL
of Model A, a load of 78.28 kN was applied to
the model up to 100 000 fatigue cycles. After
3.5
100 000 cycles, the maximum deflection of
the composite bridge deck plate was 0.63 mm,
which is close to the maximum deflection 3.0
after 2 million cycles. Therefore, we increased
the fatigue limit load to 83.66 kN, and
again applied it up to 900 000 cycles. After 2.5
2.0
deflection (0.645 mm) at the static loading
stage (83.66 kN). The results showed that the
maximum deflection of the composite deck 1.5
was 0.63 mm for a load of 78.28 kN after
100 000 cycles. The maximum deflection of 1.0
the composite deck was 0.65 mm under a
load of 83.66 kN after 900 000 cycles loading.
Hence, the stiffness of the composite deck did 0.5
not decrease significantly.
A stress analysis showed that for a load of 0
83.66 kN after 900 000 cycles, the steel plate 0 20 40 60 80 100
surface, PBL connector and top concrete sur- (kN)
face had maximum principal tensile stresses 0 2 million times 3 million times
of 8.15 MPa, 14.28 MPa and 3.18 MPa, with 10 000 times 2.1 million times theoretical values
maximum principal compressive stresses of
11.01 MPa, 6.78 MPa and 0.81 MPa, respec- Figure 6 Curve of load stress for concrete slab top
tively. Each plate remained in an elastic
stress state, and none of the stress levels of Figure 6 shows the change in the maxi- ANSYS theoretical results were greater than
the plates were high, except for the concrete mum longitudinal stress of the top concrete those of the static load test.
slabs. Figure 5 shows the change in the maxi- surface for different numbers of fatigue load- Testing indicated that after 10 000
mum longitudinal stress in the shear con- ing cycles. Figure 6 shows that, after 10 000 cycles with an upper limit of 63.52 kN, four
nectors with PBL strips based on a test using cycles, the stress was greater than that before longitudinal micro-cracks appeared in the
different numbers of fatigue loading cycles. fatigue loading; additionally, after 2 million maximum transverse tensile stress zone of the
After 2 million cycles, the measured stress cycles, the measured stress was basically top concrete surface. Meanwhile, a transverse
increased somewhat; however, the results the same as that after 10 000 cycles. Since crack appeared along the direction of the PBL
were mostly unchanged when compared with the effect of the upper steel rebar was not connector for 100 mm from the mid-span
those after 10 000 cycles. considered in the theoretical calculation, the of the model beam; the maximum width of
16 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
the longitudinal cracks was 0.04 mm. As
A–6 A–5 A–4 A–3 A–2 A–1 B–2 B–3 B–4 B–5 B–6
the number of fatigue load cycles increased,
new cracks continuously emerged, and the
existing cracks continuously expanded. The
19
widths of the longitudinal and transverse
22 25 cracks away from the mid-span section also
8 10
11 23 increased, whereas the transverse cracks near
24 26
2 7 14
the mid-span section did not increase in size.
13 12 16
15
28
29
17 After 2 million cycles, the maximum width
21
5 1 4 of the longitudinal cracks was 0.05 mm, and
3 20
27 the maximum width of the transverse cracks
was 0.10 mm. After 3 million cycles of fatigue
loading, the maximum width of the longitudi-
nal cracks was 0.055 mm, and the maximum
A’–6 A’–5 A’–4 A’–3 A’–2 A’–1 B’–2 B’–3 B’–4 B’–5 B’–6 width of the transverse cracks was 0.09 mm.
Details of the crack pattern of Model A are
Figure 7 Crack pattern detail for Model A shown in Figure 7.
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 17
cycles, the maximum deflection of Model A was 19% greater than before loading. Allahyari, H, Dehestani, M, Beygi, Morteza, H A, Neya,
was 0.515 mm, and the maximum deflection However, the growth rate was only 2.6% B N & Rahmani, E 2014. Mechanical behavior of
of Model B was 0.8 mm. The deflection of from 10 000 to 2 million cycles. This steel–concrete composite decks with perfobond
Model B was greater than that of Model A indicated that the plate had good overall shear connectors. Steel and Composite Structures,
after 2 million cycles. The stress level for all performance, and its stiffness was not 17(3): 339–358.
types of loads for Model B was not greater attenuated after some initial loss of stiff- BS (British Standard) 1980. BS 5400 Part 10: Code
than when considering the stress at each ness. The stress levels across all slabs of Practice for Fatigue. London: British Standards
element in the model structure. For Model were not high, and in the same sections Institution.
A, none of the stress levels at the elements the horizontal and longitudinal strain Chang, S P & Shim, C-S 2001. Continuous composite
were significant, with the exception of the curves of the components appeared to be bridges with precast decks. Steel Structures, 1:
concrete slabs. However, the stress levels consistent for different numbers of cycles. 123–132.
were slightly greater than those for Model B. During cyclic fatigue loading, the strain Gara, F, Leoni, G & Dezi, L 2013. Slab cracking control
Concerning the cracks, before carrying out in the steel increased with the number of in continuous steel–concrete bridge decks. Journal of
500 000 cycles of fatigue loading, the number loading cycles, and cracks in the concrete Bridge Engineering, 18(12): 1319–1327.
of cracks for Models A and B increased slab formed or expanded. GB (Chinese Standard) 2003. GB 50017-2003. Code
linearly. After 2 million cycles, the number ■■ The fatigue tests for Models A and B for Design of Steel Structures. Beijing, China: China
of cracks for Model A increased slightly, indicated that the steel-concrete compos- National Institute of Standardization.
but gently. The cracks in the bottom of the ite bridge deck of the Dongping Bridge JTT (Chinese Industry Standard) 1986. JTT 025-86
concrete slabs of Model B indicated that new performed well overall, and the fatigue Specifications for Design of Steel Structure and
cracks continuously emerged. From the above performance met the Chinese Code for Timber Structure Highway Bridges and Culverts.
analysis, the test results for Model A were the Design of Steel (GB 2003). Under Beijing, China: China Communications Press [in
different than those for Model B with regard normal conditions, the composite bridge Chinese].
to the deflection, stress and cracks, because deck of the Dongping Bridge will not Kim, H-Y & Jeong, Y-J 2006. Experimental investigation
of the different types of model study and the experience fatigue damage. on behaviour of steel–concrete composite bridge
fatigue loading, which led to differences in ■■ The results of the fatigue tests of Models decks with perfobond ribs. Journal of Constructional
the mechanical behaviour of the positive and A and B indicate that the state of the Steel Research, 62(5): 463–471.
negative bending moment regions. loading levels are different for Models A Leitão, F N, da Silva, J G S, da S. Vellasco, P C G,
and B. Model A was located in a nega- de Andrade, S A L & de Lima, L R O 2011. Composite
tive bending moment area. For a load of (steel–concrete) highway bridge fatigue assessment.
CONCLUSIONS 83.66 kN and after 3 million cycles, the Journal of Constructional Steel Research, 67(1): 14–24.
The mechanical behaviour of the Dongping maximum deflection of Model A was Leitão, F N, da Silva, J G S & de Andrade, S A L 2013.
Bridge deck structure was studied using 0.65 mm, which is close to the deflec- Fatigue analysis and life prediction of composite
experiments and finite-element analyses. tion of 0.645 mm before cyclic loading. highway bridge decks under traffic loading. Latin
The following conclusions were reached after Model B was located in a positive bending American Journal of Solids and Structures, 10(3):
evaluation of the experimental and finite- moment area. For a load of 168.2 kN 505–522.
element analyses of the positive and negative and after 2 million cycles, the maximum Liu, M, Shao, X, Zhang, Z & Hu J 2012. Experiment
bending moment regions of the composite deflection was 0.80 mm, resulting in an on flexural fatigue performance of composite
bridge decks. increase of 19% compared to the deflec- deck system composed of orthotropic steel deck
■■ The test results for Model A revealed a tion of 0.67 mm before cyclic loading. and ultra-thin RPC layer. Journal of Highway and
maximum deflection of 0.65 mm, a maxi- Transportation Research and Development, 29(10):
mum principal tensile stress in the bottom 46–53.
surface of the T-steel beam of 8.15 MPa, ACKNOWLEDGEMENTS Millanes, F, Bordo Bujalance, E, Mansilla Dominguez,
a maximum principal tensile stress in The authors gratefully acknowledge financial J L & Martin Suarez, J 2014. Archidona Viaduct:
the PBL connector of 14.28 MPa, and a support provided by the Science Foundation Composite steel–concrete deck for a long railway
maximum tensile stress in the longitudinal of China (Grant No 2016M60035), the bridge. Structural Engineering International, 24(1):
rebar of 13.53 MPa. Additionally, the stress Science Foundation of the Ministry of 122–126.
level was not high in the steel structure, Housing and Urban-Rural Development of Ryu, H-K, Kim, Y-J & Chang, S-P 2007. Crack control
with the entire structure remaining in the the People’s Republic of China (Grant No of a continuous composite two-girder bridge with
elastic stress state. At less than 500 000 2012-K2-6), the Science and Technology prefabricated slabs under static and fatigue loads.
cycles, the number of cracks increased Agency of Zhejiang Province (Grant No Engineering Structures, 29: 851–864.
linearly. However, after 500 000 cycles, 2015C33222), the Science Foundation of Wang, P & Jiang, S 2007. Overview of fatigue damage
there was only a small increase in the Shanghai (Grant No 13R21421100) and for composite structure (bridge deck). Journal of
number of cracks. The overall stiffness of Wenzhou City Science And Technology Harbin Institute of Technology, 39(2): 694–698.
the composite deck exhibited no apparent Projects (Grant No G20140017). Yang, Y, Zhou, X & Xue, J 2012. Experimental study
attenuation after 2 million cycles of fatigue on fatigue behavior of composite girders with steel
loading, and the maximum width of the plate–concrete composite bridge decks. China Civil
longitudinal cracks in the top concrete REFERENCES Engineering Journal, 45(6): 123–131.
surface was 0.05 mm. AASHTO (American Association of State Highway and Zong, Z & Che, H 2000. Fatigue behavior of pre-
■■ The test results for Model B revealed a Transportation Officials) 2005. LRFD Bridge Design stressed composite steel–concrete beams. Journal of
maximum deflection of 0.80 mm, which Specifications, 3rd ed. Washington, DC: AASHTO. the China Railway Society, 22(3): 92–95.
18 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
The influence of TECHNICAL PAPER
foundation stiffness on Journal of the South African
Institution of Civil Engineering
the behaviour of surface ISSN 1021-2019
Vol 59 No 2, June 2017, Pages 19–27, Paper 1423
19
Lemmen HE, Jacobsz SW, Kearsley EP. The influence of foundation stiffness on the behaviour of surface strip foundations on sand.
J. S. Afr. Inst. Civ. Eng. 2017:59(2), Art. #1423, 9 pages. http://dx.doi.org/10.17159/2309-8775/2017/v59n2a3
study of concrete pavements he presented bearing capacity of the soil. The strength Where:
a mathematical model to compute the and behaviour of the footing are also con- Ks = System stiffness
stresses in a concrete slab by assuming that sidered to remain constant regardless of the Eb = Young’s modulus of the footing
the slab acts as an isotropic, homogeneous, applied load. Es = Secant modulus of the soil
elastic solid, and that the subgrade reaction Footings are often analysed and designed d = Depth of the footing
is perfectly vertical and proportional to using the Winkler hypothesis, where the l = Width of the footing
the size of the slab. He assumed that the soil is modelled as a continuous system of
subgrade was an elastic medium, where non-connected discrete springs. Due to the Table 1 S tiffness limits for foundation systems
the stiffness of the subgrade could be simplicity of the Winkler model, it is used (Arnold et al 2010)
considered as the force which would cause extensively to solve soil-structure interac-
Ks range System stiffness
unit deflection if it were spread over a unit tion problems and has produced satisfactory
area. Westergaard (1937) also proposed a results in many cases (Morfidis & Avramidis 0 Absolutely flexible
measure of the stiffness of the subgrade 2002). The Winkler approach yields relatively
0–0.01 Semi-flexible
compared to the stiffness of the founda- accurate results if the structure does not
tion, termed the radius of relative stiffness. yield at any point. However, it becomes less 0.01–0.1 Semi-stiff
The work done by Westergaard has been applicable if the load keeps on increasing
0.1–infinity Stiff
refined over the years through analytical beyond yielding (Baumann & Weisgerber
models and experimental procedures, and it 1983). The Winkler method is a very basic
still remains the basis of a large number of simplification of the behaviour of the soil Arnold et al (2010) concluded that the
design procedures (Hemsley 2000). under loading conditions (Morfidis & stress distribution below a footing on sand
For the design of footings from a struc- Avramidis 2002). As load transfer occurs is dependent on the magnitude of the load
tural engineer’s point of view, the footing between the footing and the soil, the foun- applied to the footing. Initially the behaviour
is often considered as a rigid body resting dation deforms, causing internal forces to of a footing represents that of a stiff system,
on an elastic medium. This leads to the change within the structure, thus creating and changes to flexible with an increase in
assumption that the vertical settlement of a complex soil-structure interaction. The the applied load. The use of a single system
the footing on the elastic medium must significance of this interaction is important stiffness parameter is therefore not adequate,
have a planar distribution, because a rigid for both dynamic and static loading condi- as it does not take the applied load into
structure remains planar when it settles. tions. The most important aspect of the soil- account (Arnold et al 2010). The constant
The planar distribution theory is based on structure interaction is the contact stress stiffness typically assumed for a reinforced
the assumption that the ratio of pressure to distribution beneath the footing (Conniff & concrete structure is also not valid, as its
settlement remains constant, and that the Kiousis 2007). The relationship between the stiffness is significantly influenced by non-
stress distribution below the rigid footing stiffness of the footing and the soil controls linearity in the concrete’s material properties
is uniform. However, these assumptions the contact stress distribution beneath the associated with cracking. The stiffness of a
are not necessarily true, as experimental footing (Aiban & Znidarčic 1995; Arnold footing changes when the concrete starts to
studies have shown that soil exhibits plastic et al 2010). crack (Skorpen & Dekker 2014).
behaviour, footings have a finite stiffness and The structural behaviour of a footing can The aim of the current investigation was
the distribution of pressures below a footing largely be attributed to the stiffness thereof. to determine how changes in the stiffness
varies with time. The pressure distribution It is important to consider the stiffness of of strip footings affect the deformation of
in the soil is a function of the type of soil the entire foundation system, and not just such footings and the resulting contact stress
(i.e. sand or clay), as well as the stiffness of the stiffness of the footing. Arnold et al distribution underneath during loading. This
the footing and superstructure (Algin 2007). (2010) conducted a series of centrifuge model was investigated by testing a range of foot-
Currently footing design is based on an experiments to determine the effect of the ings of various depths in a physical model
iterative process where the required size of system stiffness on the stress distribution study. The study modelled strip footings
the footing often relies on the engineer’s below the footing. From their analysis they on dense sand. In further discussion in this
educated guess. The size of the footing is found that the stress distribution below the paper the term “footing” refers to the con-
then checked to see whether it satisfies footing is dependent on the stiffness of the crete slab bearing on the soil and the term
the serviceability conditions and that the footing, as well as the footing settlement for “foundation” refers to the combination of the
bearing capacity of the soil is not exceeded similar soil conditions. Whether or not a footing and the underlying soil.
(usually not critical). As soon as the size of foundation behaves in a flexible or rigid fash-
the footing has been confirmed as adequate, ion can be assessed using a dimensionless
the depth of the footing is determined to parameter Ks that incorporates a representa- EXPERIMENTAL SETUP
ensure that it has adequate flexural, shear tive Young’s modulus for the soil, combined
and punching shear strength to resist the with terms representing the stiffness of the Description of centrifuge model
applied loading conditions (Algin 2007). It footing in order to determine an equivalent A physical model was developed in which
is assumed that sufficient structural depth system stiffness (Canadian Foundation footings of various depths on dense sand
at the column will provide the footing with Engineering Manual 2006). The stiffness could be loaded under plane-strain condi-
adequate shear, diagonal tension, and bend- parameter Ks is calculated using Equation 1 tions. Due to the non-linear stress-strain
ing moment capacity (Pisanty & Gellert (Arnold et al 2010). The stiffness of the sys- properties of soils, the model was tested in
1972). The current design procedure for tem can be classified based on the ranges for a geotechnical centrifuge at the appropriate
surface footings does not take the stiff- Ks given in Table 1. acceleration to correct the stress distribution
ness of either the footing or the soil into to that of the full-scale prototype following
account, but rather focuses on the strength 1 Eb d 3
Schofield (1980). Physical models examining
Ks = (1)
of the reinforced concrete and the allowable 12 Es l the behaviour of shallow foundation in the
20 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
are similar (Knappett et al 2010). The only
Applied load
physical property of the aluminium other
IDTs than density necessary for this work is the
Young’s modulus which was measured
at 70 GPa.
Because the majority of structural foot-
ings comprise reinforced concrete, the varia-
tion in the pre- and post-cracking behaviour
of reinforced concrete complicates behav-
iour. It was therefore decided to carry out
Model footing one additional test on a reinforced concrete
150 mm model footing to compare behaviour with
that of the much more ductile aluminium
footings. The compressive strength of the
300 mm
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 21
The maximum and minimum relative
Plan view Side view
densities were determined using ASTM
D4253 (ASTM 1996a) and ASTM D4254
(ASTM 1996b) respectively at 1 669 kg/m3
75 mm
and 1 392 kg/m3. The specific gravity of the 16 mm
grains was 2.67.
150 mm
As discussed earlier, the structural behav- 14 mm
iour of a footing is dependent on the proper-
ties of the soil on which it is constructed. In
order to characterise the relative stiffness of
a foundation, a representative secant Young’s H
modulus is required for the soil (in this case,
the sand). Representative Young’s moduli for
each experiment were back-calculated from 150 mm 67 mm
elastic solutions by Milovic et al (1970) based
on the load-settlement response observed Figure 2 Model footing geometry
during the tests. Milovic’s solution applies
to rigid strip footings underlain by an elastic
100
layer of finite thickness and is given below:
90
P
E= ω 0(2)
ρz 80
70
Where: E is the Young’s modulus, P the load
Percentage passing (%)
22 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
6 (i.e. for footing depths of 16 mm and higher),
the load-settlement response could not be
improved by making the footing any stiffer,
5
le
Normalised settlements δ/B (%)
x ib
Settlement rate
e
Careful examination of the 16 mm deep
-f l
decreases with
mi
4 iff
increasing stiffness i- s t footing showed, however, that it did not
Se
S em St i f f behave entirely rigidly. Figure 5 presents the
3 settlement measured across the width of
the footing at third points, showing slight
2 n variation and hence bending of the footing at
uc t i o f ne s
s
r a t e re d t i n g s t i f larger loads. This suggests that the classifica-
e n t f o o
em ng
s e t t l i nc r e a s i tion system by the Canadian Foundation
1 it o f
L i m v a ble by
ac h i
e Engineering Manual (2006) could perhaps be
0
refined somewhat at the boundary separating
0 5 10 15 20 25 30 35 semi-stiff and stiff footings.
Applied load (kN)
Model footing depth: 3 mm 10 mm 25 mm 50 mm Load-settlement response of the
6.7 mm 16 mm 35 mm reinforced concrete footing
In addition to the investigation of the influence
Figure 4 Normalised load-settlement response for aluminium footings of varying depth of footing rigidity using aluminium footings,
one test was carried out using a footing made
from scaled-down reinforced concrete. The
3.5 depth of the reinforced concrete footing was
chosen so that, for an uncracked section, the
3.0 foundation system classified as stiff (Ks value of
Normalised settlements δ/B (%)
3.0 Broken line shows settlement rate for which the footing was unable to accept
semi-flexible footing from Figure 4 additional load and just continued to deform.
2.5
Cracking of the footing resulted in the foot-
2.0 First onset of cracking ing effectively changing from stiff to much
more flexible. The settlement rate for the
1.5 3 mm deep footing, which classified as semi-
flexible, is presented against the settlement
1.0 record of the concrete footing measured
Load settlement response of 20 mm at the column edge for comparison. This
0.5
deep aluminium footing with similar illustrates that, after cracking, the load-
Ks value as concrete footing
0
settlement response of the concrete footing
0 5 10 15 20 effectively became semi-flexible.
Applied load (kN)
Column edge 25 mm from centre 50 mm from centre Footing edge Contact stress distribution
The contact stress distribution underneath
Figure 6 Normalised settlement of reinforced concrete footing each model footing was measured using a
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 23
Front of footing
Edge of footing
Active part of pressure sensor
24 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
rapidly with increasing foundation stiffness
Footing depth
and then stabilises, becoming practically
6.7 mm
50 mm
25 mm
35 mm
10 mm
16 mm
independent of foundation stiffness once the
3 mm
foundation classifies as stiff. This implies
6 that, as soon as the stiffness of the footings
was sufficient for the foundation to classify
5 as stiff, the behaviour of the footings was
Normalised settlement δ/B (%)
50 mm
25 mm
35 mm
10 mm
16 mm
3 mm
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 25
Canadian Foundation Engineering Manual
Footing depth
(2006) as a Ks value of 0.1, the 16 mm footing
6.7 mm
50 mm
25 mm
35 mm
10 mm
16 mm
(Ks value of 0.15) did not exhibit the same
3 mm
behaviour as the stiff footings. Therefore, it
Ks =0.3
is suggested that the proposed lower bound-
100 ary to which a uniform pressure distribution
is valid is adjusted upward from a Ks of 0.1
Edge contact stress / centre contact stress (%)
90
to a value of 0.3. Even at conventional loads
80 (equivalent to 5 kN (222 kPa) or 10 kN
70
(444 kPa)) semi-stiff footings still showed
reduced stress under the edges, not behaving
60 as stiff as shown in Figure 12.
50
40
log Ks CONCLUSIONS AND
c= – 1.26
30 2 RECOMMENDATIONS
The following conclusions from the model
20
study are presented:
10 ■■ The relative stiffness of a foundation sys-
0
tem affects the deflection of and contact
0.001 0.01 0.1 1 10 stress distribution beneath the footing.
Relative foundation stiffness Ks The contact stress distribution beneath
Applied load: 5 kN 10 kN 15 kN 20 kN 25 kN 30 kN footings in a stiff foundation on dense
sand is approximately uniform. A reduc-
Figure 12 Effect of footing bending stiffness on stress distribution under footings tion in stiffness results in the footing
bending during loading, which reduces
in Figure 12 and extrapolating suggest that, in Figure 12. This relationship is described the contact stress at the footing edges.
for perfectly stiff conditions, the relative by Equation 3: ■■ The contact stress distribution beneath
stiffness Ks should exceed 0.3. a footing plays an important role in
log Ks not only the settlement and deflection
c= – 1.26 (3)
Predicting bending moment 2 behaviour of the footing, but it affects the
in the footing bending moments in the footing that the
Stress measurement underneath the model Let a represent the maximum contact stress structural engineer needs to design for.
footings showed that, for stiff foundation under the footing centreline and b the foot- ■■ The results presented suggest that the
systems (Ks > 0.3), the contact stress under- ing width. The maximum bending moment optimum depth of a footing is that depth
neath the footings can be taken as approxi- per unit length occurring under the centre- resulting in the foundation just classify-
mately uniform. As the relative stiffness line of the footing, assuming a trapezoidal ing as stiff, i.e. corresponding to a rela-
reduces into the semi-stiff range, the contact distribution (see Figure 9b), is then: tive foundation stiffness Ks of 0.3. For
stress underneath the footing edge as a this, and higher stiffnesses, the contact
percentage of the maximum reduces. When ab2 stress distribution under a strip footing
M= (1 + 2c)(4)
the stiffness reduces further to semi-flexible, 24 can be assumed to be approximately
the contact stress at the edge approaches uniform.
zero and the edge may even lift up. It would For a stiff footing the pressure distribution ■■ The behaviour of an initially stiff con-
be beneficial to be able to predict the contact is uniform and c = 1 so that the bending crete footing will become more flexible
stress distribution as a function of the rela- moment M = ab2/8. With a reduction in after the onset of cracking. This can
tive foundation stiffness, as the contact stress relative stiffness to semi-flexible, c reduces occur well below the allowable bearing
distribution affects the bending moment that to zero, but under the same vertical load a capacity of the foundation. The settle-
the footing has to be designed for. increases by a factor of 2 so that the bend- ment undergone by a column bearing
Figure 12 shows that the contact stress ing moment becomes M = ab2/12. The on such a footing will then depend on
at the edge of the footing, as a percentage of bending moment in a footing forming part both the deformation of the underlying
the maximum stress, does not change much of a semi-flexible foundation can therefore soil and the deformation of the footing.
during loading. It also shows that, across a be as little as two-thirds of the moment in The contact stress at the footing edges
range of relative stiffness values, the contact footing which is part of a stiff foundation. In will reduce so that the maximum stress
stress at the footing edge as percentage of the addition, using the correct stress distribu- under the foundation will increase if
maximum varies approximately linearly with tion, the deflected shape of the footing can the footing still has to support the same
the logarithm of the relative foundation stiff- be calculated more accurately than a result load. This situation appears less optimal
ness Ks. Let the ratio between the contact simply based on the assumption of a uniform than the behaviour of a stiff foundation.
stress under the edge of the footing versus stress distribution. The recommendation is therefore that
the maximum value occurring under the foundations be designed to have a relative
centre be represented by c. It is proposed that Relative foundation stiffness foundation stiffness classifying as “stiff”,
c varies linearly with the logarithm of the “stiff” versus “semi-stiff” i.e. a Ks value of 0.3. It does not appear
relative foundation stiffness from zero at a Although the boundary between stiff foot- that increasing the stiffness beyond this
value of 0.003 to 1 at a value of 0.3 as shown ings and semi-stiff footings is given by the value is beneficial.
26 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
REFERENCES Brown, R, Valsangkar, A J & Schriver, A B 2004. Meyerhof, G G 1963. Some recent research on
Aiban, SA & Znidarčic, D 1995. Centrifugal modeling Centrifuge modeling of surface footings on a sand the bearing capacity of foundations. Canadian
of bearing capacity of shallow foundations on sands. layer underlain by a rigid base. Geotechnical and Geotechnical Journal, 1(1): 16–26.
ASCE Journal of Geotechnical Engineering, 121(10): Geological Engineering 22: 187–198. Milovic, D M, Touzot, G & Tournier, J P 1970.
704–712. BSI (British Standard) 2015. BS 8004:2015: Code of Stresses and displacements in an elastic layer due
Algin, H M 2007. Practical formula for dimensioning Practice for Foundations. London: British Standards to an inclined eccentric load over a rigid strip.
a rectangular footing. Engineering Structures, 29(1): Institution. Geotechnique, 20(3): 231–252.
1128–1134. Canadian Foundation Engineering Manual, 4th Morfidis, K & Avramidis, I E 2002. Formulation of
Archer, A 2014. Using small-strain stiffness to predict ed. 2006. Richmond, BC, Canada: Canadian a generalised beam element on a two-parameter
the settlement of shallow foundation on sand. MEng Geotechnical Society. elastic foundation with semi-rigid connections
dissertation, University of Pretoria. Conniff, D E & Kiousis, P D 2007. Elastoplastic medium and rigid offsets. Computers and Structures, 80(1):
Arnold, A, Laue, J, Espinosa, T & Springman, S M 2010. for foundation settlements and monotonic soil– 1919–1934.
Centrifuge modelling of the behaviour of flexible raft structure interaction under combined loadings. Pisanty, A & Gellert, M 1972. Automatic design of
foundations on clay and sand. In: Springman, S M, International Journal for Numerical and Analytical sloped spread footings. Building Science, 7(1): 53–59.
Laue, J & Seward, L (Eds.), Physical Modelling in Methods in Geomechanics, 31(1): 789–807. Schofield, A N 1980. Cambridge geotechnical
Geotechnics. Proceedings of the 7th International Haigh, S K, Houghton, N E, Lam, S Y, Li, Z & centrifuge operations. Geotechnique, 30(3): 227–268.
Conference on Physical Modelling in Geotechnics Wallbridge, P J 2010. Development of a 2D servo- Skorpen, S & Dekker, N 2014. The application and
(ICPMG 2010) held in Zurich. London: Taylor and actuator for novel centrifuge modelling. In: interpretation of linear finite element analysis results
Francis, 679–684. Springman, S M, Laue, J & Seward, L (Eds.), Physical in the design and detailing of hogging moment
ASTM (American Society for Testing and Materials) Modelling in Geotechnics. Proceedings of the 7th regions in reinforced concrete flat plates. Journal of
1996a. ASTM D4253 –93: Test Methods for International Conference on Physical Modelling in the South African Institution of Civil Engineering,
Maximum Index Density and Unit Weight of Soils Geotechnics (ICPMG 2010) held in Zurich. London: 56(1): 77–92.
Using a Vibratory Table. West Conshohocken, PA: Taylor and Francis, 239–244. TekscanTM 2014. Sensor Model 5105. Available at:
ASTM International. Hemsley, J A 2000. Design Application of Raft http://www.tekscan.com/5101-pressure-sensor.
ASTM (American Society for Testing and Materials) Foundations. London: Thomas Telford. (Accessed on 29 October 2014).
1996b. ASTM D4254 –91: Standard Test Method Knappett, J A, O’Reilly, K, Gilhooley, P, Reid, C & Westergaard, H M 1925. Stresses in concrete pavements
for Minimum Index Density and Unit Weight of Skeffington, K 2010. In: Springman, S M, Laue, J & computed by theoretical analysis. Public Roads, 7: 25.
Soils and Calculation of Relative Density. West Seward, L (Eds.), Physical Modelling in Geotechnics. Westergaard, H M 1937. What is known of stresses.
Conshohocken, PA: ASTM International. Proceedings of the 7th International Conference Engineering News Record, 118(1): 26–29.
Baumann, R A & Weisgerber, F E 1983. Yield-line on Physical Modelling in Geotechnics (ICPMG White, D J, Take, W A & Bolton, D 2003. Soil
analysis of slabs-on-grade. Structural Engineering 2010) held in Zurich. London: Taylor and Francis, deformation measurement using particle
Journal, 109(11): 1553–1567. 141–146. image velocimetry (PIV) and photogrammetry.
Brinch Hansen, J 1968. A revised extended formula Kusakabe, O 1995. Foundations. In: Taylor, R N (Ed.), Geotechnique, 53(7): 619–631.
for bearing capacity. Danish Geotechnical Institute Geotechnical Centrifuge Technology, London: Blackie
Bulletin, No. 28. Brabrand, Denmark. Academic and Professional, 118–165.
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 27
TECHNICAL PAPER Revised regional sediment
Journal of the South African
Institution of Civil Engineering
yield prediction methodology
ISSN 1021-2019
Vol 59 No 2, June 2017, Pages 28–36, Paper 1326 for ungauged catchments
DR VINCENT MSADALA holds PhD and
in South Africa
MSc degrees in Hydraulic Engineering from
Stellenbosch University. He has worked as a
lecturer at the University of Malawi in the V C Msadala, G R Basson
Department of Civil Engineering. His
research interests are in hydraulics, hydrology
and water resources development. Vincent This paper presents the research on the revision of the regional probabilistic method
is a member of the South African Institution
(Rooseboom et al 1992) for the prediction of catchment sediment yields in South Africa. The
of Civil Engineering (SAICE).
determination of sediment yields using probabilistic or empirical methods is suitable for
Contact details: ungauged catchments in the absence of observed data. The prediction of sediment loads is a
Department of Civil Engineering
key component in the quest to deal with reservoir and river sedimentation which is a potential
Stellenbosch University
Private Bag X1 threat to the sustainability of water resources in southern Africa.
Matieland The revision was necessitated by increased sediment data availability and improved
7602 data analysis tools. Ten new sediment yield regions were demarcated in South Africa and
South Africa Lesotho. Two analytical approaches were developed, namely probabilistic and empirical.
T: +265 888 817 470 / +265 882 750 550
The probabilistic approach is applicable in sediment yield Regions 3, 6 and 9. The empirical
E: vmsadala@gmail.com
approach is applicable in sediment yield Regions 1, 2, 4, 5, 7 and 8. The estimation of
sediment yields in Region 10 (Lesotho Highlands) needs to be based on direct measurements
PROF GERRIT BASSON is Head of the Water
Division, Department of Civil Engineering, and locally observed data since no adequate analysis of sediment loads was possible due to
Stellenbosch University. He specialised in river limited data. GIS and electronic portable document file (pdf) copies of maps were produced
hydraulics, sedimentation and design of for the retrieval of catchment data.
hydraulic structures. He worked for ten years at
consulting engineers in South Africa, then
joined the University of Pretoria in 1997,
and since 2000 he is based at Stellenbosch INTRODUCTION OBJECTIVES OF THIS PAPER
University. He has worked on projects in 21 countries and has 30 years’ Sediment affects the water quantity and This paper discusses the work carried out
experience as hydraulic engineer. He recently served the dam engineering quality in rivers, lakes and reservoirs. The recently to revise the sediment yield predic-
industry as Vice-President of the International Commission on Large Dams analysis of reservoir sedimentation rates in tion method by Rooseboom et al (1992).
(ICOLD), from 2012 to 2015. Prof Basson is also a Member of the South
South Africa indicated an estimated mean The detailed information on the research
African Institution of Civil Engineering (SAICE).
annual loss in original storage capacity of project is contained in a report to the Water
Contact details:
0.4% per year. Although the estimated global Research Commission of South Africa
Department of Civil Engineering
Water Division
reservoir sedimentation rate is 0.8% per year (Msadala et al 2012). The objectives of the
Stellenbosch University (ICOLD 2009), the sedimentation rate in research project on the revision of the 1992
Private Bag X1 South Africa is quite significant considering methodology were to use the most recent
Matieland the extent of increased water demand across reservoir sedimentation and river sediment
7602
southern Africa. Almost 25% of the total transport data, and to evaluate a regional
South Africa
T: +27 21 808 4355
number of reservoirs that were analysed in empirical stream power approach in addition
E: grbasson@sun.ac.za South Africa had lost between 10% and 30% to the original probabilistic sediment yield
of their original storage capacity. This calls prediction methodology, in order to improve
for increased attention towards reservoir the prediction accuracy.
sedimentation problems.
There have been extensive studies with
regard to sediment transport in rivers and SEDIMENT YIELD PREDICTION
reservoirs in South Africa dating back some APPROACHES
50 years. Some of the studies resulted in In practice, the sediment yield at any refer-
the development of the sediment yield map ence point is found by applying the following
of southern Africa and a treatise on sedi- four major approaches:
ment transport in rivers and reservoirs by ■■ Direct measurements from reservoir sedi-
Rooseboom et al (1992). The latter resulted ment deposit surveys
in what was considered as a basic handbook ■■ River suspended sediment sampling
for sediment yield prediction. With time it ■■ Catchment sediment yield mathematical
became apparent that the methods contained modelling
in the book needed to be updated to take ■■ Analytical methods such as sediment
into account the additional observed sedi- yield maps.
Keywords: sediment yield, analytical, probabilistic method, empirical method, ment data and technologically advanced data Two analytical approaches were evaluated,
ungauged catchments analysis tools. namely the probabilistic method (Rooseboom
28
Msadala VC, Basson GR. Revised regional sediment yield prediction methodology for ungauged catchments in South Africa.
J. S. Afr. Inst. Civ. Eng. 2017:59(2), Art. #1326, 9 pages. http://dx.doi.org/10.17159/2309-8775/2017/v59n2a4
et al 1992) and the empirical method. The 78
probabilistic method was based on statisti-
cal analysis of observed data, particularly 77
newly calculated sediment yield data, and the
revised erosion hazard classes. The empirical 76
method was based on the total input stream
5
method (Rooseboom et al 1992) was imple-
mented. As such, the probabilistic method is
4
applicable to sediment yield Regions 3, 6 and
9, while the empirical method is applicable
3
to sediment yield Regions 1, 2, 4, 5, 7 and
8. The estimation of sediment yields in
2
Region 10 (Lesotho Highlands) should be
based on direct measurements and locally
1 y = 3.4678ln(x) – 7.0333
observed data since no adequate probabilistic R2 = 0.9807
or deterministic analysis of sediment loads 0
and predictor variables was possible due to 1 10 At 50 years 100
limited data. Period (year)
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 29
20°E 25°E 30°E 35°E
potential consolidation characteristics of WATER RESEARCH COMMISSION (2010): REVISED SEDIMENT YIELD MAPS
SOUTH AFRICA AND LESOTHO
fine sediment (clay and silt). A logarithmic SEDIMENT YIELD BY RIVER GAUGE AND DAM CATCHMENT
trend line was fitted through the data points
to determine the 50-year sediment volume
from the equation of the trend line. This was
based on the assumption that a logarithmic
relationship exists between the sediment
25°S
25°S
Legend
deposited in a reservoir and the time period
Rivers
(Rooseboom et al 1992). The 50-year sedi-
Regions
ment volume and the effective catchment Sediment yield (t/km2.a)
area were used to compute the sediment 1–50
50.000001–100
yield or load at the dam.
100.000001–150
A map of observed sediment yields is 150.000001–200
30°S
given in Figure 3. It should be noted that it 200.000001–250
30°S
250.000001–300
was not an objective of this study (Msadala 300.000001–350
et al 2012) to develop a sediment yield map 350.000001–400
Datum: WGS 84
ments in South Africa.
35°S
Projection: Albers Equal Area 1 000.000001–2 000
0 100 200 300 400 500 600 700 800 900 1 000
Central Meridian: 24°E Kilometres
The revised sediment yield values were Standard Parallels: 18°S and 32°S 2 000.000001–2 500
sometimes higher for similar dams when 20°E 25°E 30°E 35°E
9–15
16–20 Low
Yield region
THE PROBABILISTIC APPROACH Catchment
boundaries
Background
The probabilistic approach was an updated
methodology that was based on the previous
report’s (Rooseboom et al 1992) fundamental
assumptions. The main underlying assump-
tion in the probabilistic approach is that the
fine cohesive sediment transport is depend-
ent on sediment availability, and in turn
sediment availability is influenced by the
soil erosion hazard. This was a rather crude
assumption, considering the dominant role
of other significant sediment-controlling fac-
tors, such as river network density and floods
that could also affect the sediment yield Based on classification by Prof E Verster, UNISA
30 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
20°E 25°E 30°E
Erosion Index
area that was covered by each individual
1 – Very low erosion hazard class. Each observed sedi-
2
3
ment yield value was associated with its
4 – Moderate corresponding single dominant erosion
5 1
6
hazard class (Erosion Index – EI) within the
2
25°S
25°S
7 – High
8
9
hazard class per catchment was based on the
10 – Extremely high weighted average erosion hazard class that
Regions was converted to the nearest integer in the
4
5 range of 1 to 10. The sediment yield values
for dam and river catchments with similar
10 dominant erosion hazard classes were
3
30°S
30°S
6
grouped together. The median values of the
sediment yield values with similar dominant
erosion hazard class were computed.
7
Probabilistic analysis was done to obtain
9 sediment yields for all the required prob-
8
abilities of exceedance within a region. The
0 200 400 600 800 Kilometres
plotting of sediment yields data on a graph
35°S
35°S
15°E 20°E 25°E 30°E at the required exceedance probability was
Figure 5 N
ew sediment yield regions showing erosion hazard classes (Msadala et al 2012) done manually using Microsoft Excel. Three
distribution parameters were employed to
et al (2008), whereby newly improved cover in the previous methodology (Rooseboom statistically analyse the data, namely the
and topography factor maps were applied. et al 1992) was found to be not representa- Log Normal (LN), the Log Pearson III (LP3)
Ten erosion hazard classes were identified tive enough with respect to the observed and General Extreme Value using Mean
using Geographical Information Systems erosion potential in the characteristically Moments (GEVmm). The LN distribution was
(GIS) spatial data analysis tools. According diverse homogeneous regions. Hence, the found to be best suited to data for Regions 3
to Figure 5, very low erosion hazard was current approach proved to be a significant and 6, while data for Region 9 was based on
taken as Class 1, and Class 10 represented improvement in the prediction accuracy of the average between the LN and Log GEVmm
extremely high erosion hazard. The deci- sediment yields. distributions’ values, which gave a better fit.
sion to use ten erosion hazard classes was The main assumption in the probabilistic
necessitated by the high variability in erosion The probabilistic methodology analysis was that the standardised average
hazard potential across southern Africa. The The probabilistic analysis applied the pro- yield (SYsty) for each region, could be taken
use of three groups of erodibility indices portion of area out of the total catchment as the sediment yield at 50% exceedance
10 000
Region 3 plot
Region 6 plot
Region 9 plot
Region 3 data points
Region 6 data points
1 000 Region 9 data points
Sediment yield (t/km 2 .a)
100
10
1
0.998 0.995 0.99 0.98 0.95 0.9 0.8 0.5 0.2 0.1 0.05 0.02 0.01 0.001 0.0001
Probability of exceedance
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 31
probability from the statistical analysis plots
in Figure 6.
The probabilistic approach was designed
to use sediment potential factors to relate the
standardised average yield (SYsty) per region
to the estimated sediment yield SYest for an
ungauged catchment within a homogeneous
sediment yield region. The sediment potential
factor was taken as the ratio of the median of
each group of observed sediment yield with
similar dominant erosion hazard class and the
standardised average sediment yield.
For any ungauged catchment within these
three regions, Equation 1 is used to predict
or estimate the sediment yield SYest.
A1 A2 A3
SYest = SYsty F1 + F2 + F3
AT AT AT
A4 A5 A6 A7 Figure 7 I llustration of data query per quaternary catchment (Msadala et al 2012)
+ F4 + F5 + F6 + F7
AT AT AT AT
Acrobat Reader. As shown in Figure 7, the factor of 1 is applied at 50% probability of
A8 A9 A10 user is provided with sediment-related qua- exceedance, such as when Equation 1 is used
+ F8 + F9 + F10 (1)
AT AT AT ternary data (as shown on the left pane) upon without considering confidence limits. The
clicking within any quaternary. multiplication factors for selected confidence
Where: Equation 1 only calculates the estimated limits for Regions 3, 6 and 9 can be obtained
SYest = estimated median sediment yield median sediment yield value (SYest). In order from the appendices in the Water Research
value (t/km2.a) to get the predicted sediment yield (termed the Commission Report (Msadala et al 2012).
SYsty = standardised average sediment yield ‘factored estimated sediment yield’ (SYfest)),
value (t/km2.a) for a specific region. the estimated median sediment yield (SYest) Verification of results
value should be multiplied by a special mul- The predictive accuracy of the probabilistic
The standardised average yield for each sedi- tiplication factor. The multiplication factor, method was evaluated by using the discrep-
ment yield region is shown in Table 1. F1 to which is dependent on the preferred confi- ancy ratio test at 50% exceedance probability.
F10 are sediment potential factors, of which dence limit, catchment area size and sediment The technique compares all the predicted
the values are also given in Table 1. A1 to yield region is obtained from sediment yield sediment yields against all the observed sedi-
A10 in Equation 1 are catchment areas that confidence limits graphs, which were specifi- ment yields in terms of the discrepancy ratio
are covered by erosion hazard classes 1 to 10 cally prepared for each of the three sediment xi whereby each predicted value is divided by
respectively within the total catchment AT. yield regions. The relationship is shown below: the corresponding actual observed value. In
The determination of the area covered mathematical terms the discrepancy ratio is
by each of the erosion hazard classes within SYfest = Multiplication Factor × SYest(2) given by the following relationship:
the ungauged catchment to be applied in
Equation 1 is illustrated in Figure 7, which Where: SYsim
= xi(3)
shows a sample of an electronic portable SYfest = Factored estimated median sediment SYobs
document file (pdf) copy for sediment yield yield value (t/km2.a).
Region 1. The quaternary catchments are Where:
labelled based on the WR90 nomenclature, Multiplication factors and SYsim = predicted sediment yield
and in this particular case it relates to A24H. confidence limits SYobs = observed sediment yield.
The area covered by each of the erosion The confidence limits are required to act as
hazard classes within a quaternary catch- envelope values on the regional standardised The predicted sediment yields were calcu-
ment can be displayed upon querying each average yield upon application of Equation 2. lated using Equation 2 at 50% exceedance
of the individual quaternaries within the A 50% exceedance probability within a given probability. The results were as follows:
catchment boundary. The querying is done region indicates that 50% of the predicted val-
using the ‘Object Data Tool’ functionality ues could be lower, or 50% of the predicted val- 0.33 ≤ xi ≤ 3 ( 81% of the data was in this
in the ‘Model Tree’ viewing mode in Adobe ues could be higher. By default a multiplication range)
3 40 0.027 0.417 1.202 1.110 0.575 0.000 0.000 0.000 0.000 0.000
6 622 0.000 0.000 0.000 0.000 0.000 1.428 1.042 1.002 0.998 0.000
9 50 0.196 1.824 1.871 0.000 0.000 0.000 0.000 0.000 0.000 0.000
32 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
0.5 ≤ xi ≤ 2 ( 68% of the data was in this Table 2 Discrepancy ratio results for the probabilistic method (Msadala et al 2012)
range)
Percentage of data in this range
0.67 ≤ xi ≤ 1.5 (43% of the data was in this Region Obs n
range) 0.67 ≤ xi ≤ 1.5 0.5 ≤ xi ≤ 2 0.33 ≤ xi ≤ 3
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 33
Table 3 E mpirical equations based on GIS maps using a similar procedure that was Significance of the variables in
regression analysis explained under Figure 7. The electronic the empirical equations
version of these maps were prepared as part
Region Proposed equation
of the WRC project (Msadala et al 2012) and Weighted erosion hazard class
1 0.98 S –0.38 R 0.13 A 0.10 EI 1.22
Qs = 15Q10 0 nd e w were included as part of the report appen- The weighted erosion hazard class basi-
2 0.32 S 0.19 R 0.63 A 0.95 EI –1.44
Qs = 6.4Q10 dices of maps. The weighting of the erosion cally provides a quantitative measure of
0 nd e w
hazard classes is with respect to the catch- the following parameters: climate, soil
4 0.53 S 0.43 R 1.08 A 0.88 EI –1.04
Qs = 0.2Q10 0 nd e w ment area covered by each erosion hazard profile, relief, vegetation, land use and land
5 1.18 S 0.58 R –1.27 A 0.54 EI –0.30
Qs = 1 208Q10 0 nd e w
class. The 10-year ARI flood and effective management practices based on the Revised
catchment area can be calculated using Universal Soil Loss Equation (RUSLE) model
7 0.43 S 0.6 R 0.17 A 0.5 EI 0.76
Qs = 40Q10 0 nd e w standard hydrological methods. Similarly, of Renard et al (1991).
8 –0.24 S 1.27 R 1.52 A 1.26 EI –0.61
Qs = 0.004Q10 0 nd e w the river network density can be obtained
from the electronic pdf maps in km/km2 per River network density
Where:
quaternary based on the data, such as that Drainage density is a measure of the length
Qs = sediment load (t/a)
Q10 = a flood of a recurrence interval of 10 years
given in Figure 7. It should be noted that of stream channel per unit area of basin.
(m3/s) electronic pdf maps give river network densi- River network drainage density can be clas-
Rnd = river network density (m/km 2) ties in km/km2 units, and these values need sified as one of the factors that determine
Ae = effective catchment area (km 2) to be converted to m/km2 prior to applying the catchment’s sediment yield according
EIW = weighted erosion hazard class according to in any of the equations in Table 3. Using the to Strand and Pemberton (1982). It can be
sub-catchment areas
area for each quaternary, the river length per assumed that, with all factors equal for the
So = average river slope (%)
quaternary can be computed by multiplying same catchment area, longer lengths of
the total area and the corresponding river river channel per unit area must be able to
The equations in Table 3 can be used to network density, and then summing up all transport more sediment than shorter river
predict the sediment yield or load in an river lengths for all quaternaries to give the channels within the catchment.
ungauged catchment as long as accurate total river length in the whole catchment.
data with respect to the specified variables is Finally, the river network density for the Recurrence interval flood
applied. The weighted erosion hazard class whole catchment can be computed by divi In the original application of the unit stream
can be determined using data obtained from ding the total river length against the total power, the instantaneous discharge was used
electronic portable document files (pdf) or catchment area under consideration. in the relationship to describe the sediment
15 Mzhelele Dam 530 5.00 1.1160 832 7.00 101 315 230 847
16 Olifantsnek Dam 510 2.36 0.1026 499 3.93 119 249 50 513
18 Vaalkop Dam 770 1.39 0.1244 3 917 3.33 183 235 231 148
34 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
Table 4(b) Split-sample analysis for Region 2
1:10-year
River network Erodibility Calculated Observed
recurrence Average slope Effective area
ID Station name density (Rnd) index/hazard sediment sediment
interval flood (So) (river) (%) (Ae) (km 2)
(m/km 2) class (EIw) load (t/a) load (t/a)
(Q10) (m3/s)
14 Massingir Dam 5 309 2.62 0.1223 63 350 4.81 4 786 006 15 520 750
15 Middel Letaba Dam 715 2.92 0.1331 1 051 5.84 48 752 547 048
16 Nooitgedacht Dam 245 1.46 0.1220 1 583 4.07 140 983 193 161
20 Rust de Winter Dam 125 1.83 0.1128 1 127 4.84 76 287 27 955
24 Vygeboom Dam 460 5.06 0.1477 1 541 5.59 109 221 84 761
transport. However, when considering Table 5 Results of split-sample predictive accuracy analysis
sediment load over a long period of time,
Obs Percentage of the data in this range
an effective discharge passing through a Region
point along a river or a reservoir would best N 0.67 < xi < 1.5 0.5 < xi < 2.0 0.33 < xi < 3.0
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 35
The empirical equations in Table 3 Table 6 Discrepancy ratio results for the empirical method
were derived from all the observations in
Percentage of data in this range
the given region. The predictive accuracy Region Obs n
checks per region were done by way of 0.67 ≤ xi ≤ 1.5 0.5 ≤ xi ≤ 2 0.33 ≤ xi ≤ 3
inspection of the discrepancy ratio. Table 6 1 18 56 83 94
shows the results of the discrepancy ratio
test for each region. 2 25 48 68 88
36 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
The two-points condensation TECHNICAL PAPER
technique (TPC) for Journal of the South African
Institution of Civil Engineering
detection of structural ISSN 1021-2019
Vol 59 No 2, June 2017, Pages 37–48, Paper 1400
37
Al-Qayyim AN, Çağlayan BÖ. The two-points condensation technique (TPC) for detection of structural damage due to vibration.
J. S. Afr. Inst. Civ. Eng. 2017:59(2), Art. #1400, 12 pages. http://dx.doi.org/10.17159/2309-8775/2017/v59n2a5
severe damage has occurred or accurate
measurement of vibration was done. The
Structural drawings
inverse method is at present still only being
investigated theoretically (Friswell 2008).
For the curvature/strain modes approach,
Pandey et al (1991) demonstrated that the Mass Mo Stiffness Ko
damage to beam structures can be identi-
fied using mode shape curvature. Chance et
al (1994) found the measured strain mode Reduction to 2DOF
shape to be much more feasible for damage
localisation. But recording strain over the Experimental test
whole structure is not yet feasible with cur-
rent technology; therefore, a new methodol- Mr Kr
38 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
structure. The reduced theoretical stiffness matrices Kr is used to locate the damage. & Motiershead 1995). Here the coordinates
Kr is generated by condensing the theoretical Large changes in the stiffness coefficients represent the location of sub-matrices in the
stiffness matrix Ko to a two-degrees-of- of the 2-DOF matrices indicate the location original matrix.
freedom (2-DOF) stiffness matrix. The selec- of the damage. The changes in the stiffness The reduction of the stiffness matrix
tion of these degrees of freedom corresponds matrix are shown in Equation 3. is thus accomplished by identifying those
to the vibration acceleration measurement degrees of freedom to be condensed or
sensor locations. ∆Ki–j = ∆kii ∆kij (3) reduced as slave degrees of freedom, and to
In this technique the SEREP method is ∆k ji ∆k jj express them in terms of remaining master
used to reduce the stiffness and mass matrix degrees-of-freedom.
to 2-DOFs. The SEREP condensation has the Where: The dynamic equations of equilibrium
best computational performance and leads ∆Ki–j = the change matrix of the set of nodes for an undamped n degree-of-freedom model
to smaller errors in the identification of stiff- i and j may be written as:
ness values (Koh et al 2006). ∆kii = the difference in the coefficients of
The TPC technique calculates the the Kc matrix and the Kr matrix at M ü(t) + K U(t) = F(t)(4)
identified stiffness matrix Kc by finding position (i, i)
the optimal solution of the equations of ∆kij = the difference in the coefficients of Where: ü(t) and U(t) are the acceleration and
motion, which should correspond to the the Kc matrix and the Kr matrix at displacement response vectors. The displace-
real system properties (the mode shapes and position (i, j) ment response vector U(t) in Equation 4 can
frequencies). ∆k ji = the difference in the coefficients of be expressed as shown in Equation 5 using
the Kc matrix and the Kr matrix at the mode superposition method:
Mr{ü}2×1 + [Cr]2×2{u̇}2×1 + [Kr]2×2 {u}2×1 = 0 (1) position (j, i)
∆k jj = the difference in the coefficients of U(t) = Φ q(t)(5)
Where: the Kc matrix and the Kr matrix at
Kr = The stiffness matrix of the reduced position (j, j). in which Φ is the complete eigenvector matrix
system of the full model, and q(t) is the modal coor-
Mr = The mass matrix of the reduced system When large changes are observed by com- dinate vector. It is well known that the com-
Cr = The damping matrix of the reduced paring ∆kii with ∆k jj, the location of the putation of the complete eigenvector matrix
system damage can be determined. If ∆kii is greater is not required for a large model. Therefore,
ü = The acceleration vector than ∆k jj, it means that the damage is located modal truncation is usually used in the mode
u̇ = The velocity vector near to the node i. superposition technique (Qu 2004). If p eigen-
u = The displacement vector. vectors of the full model are used in the mode
Analytical reduction of system superposition, Equatioin 5 is rewritten as:
In this technique, the input data includes the matrices using SEREP method
reduced mass matrix Mr and initial stiffness From an analytical approach standpoint, the U(t) = Φpq p(t)(6)
matrix Kr in addition to the corresponding finite element method assumes that a contin-
vectors of acceleration ü, velocity u̇, and dis- uous structure can be discretised by describ- With the arrangement of the total degrees of
placement u. The first and second measured ing it as an assembly of finite elements, each freedom, Equation 6 may be partitioned as:
modal frequencies are part of the input data with a number of boundary points that are
U(t) = Um(t) = Φmp q p(t)(7)
and are used to calculate the damping coeffi- commonly referred to as nodes. The main
cients and control the solution of finding the problem to overcome in SHM and damage U
s (t) Φsp
identified (reduced) stiffness matrix. detection is the typical mismatch of the
The technique uses the theoretical mass selected number of degrees of freedom of an This is equivalent to two equations (8 and 9):
matrix of the structure because it does not analytical and an experimental representa-
generally change (Pokharkar & Shrikhande tion of a structural dynamic system. Um(t) = Φmpq p(t)(8)
2010). The damping matrix C is calculated For damage detection, the concept of
according to Rayleigh damping. The damp- model reduction (or, alternatively, model
ing matrix is optimised due to the stiffness expansion) plays an important role. Using Us(t) = Φspq p(t)(9)
updating. The optimisation solver minimises condensation or expansion, it is possible to
the function given in Equation 2, which is compare a large analytical set of DOFs to a Equation 8 provides a description of the
based on the equation of motion. relatively small set of experimental DOFs. displacement responses at the master DOFs
Reduction and expansion also play a very in terms of the eigenvector matrix at these
q = sum [0 – [[Mr]{ü} + [α[Mr] + β[K]]{u̇} important role with regard to model updat- DOFs. The sub-matrix Φmp is generally not a
+ [K]{u}]]2(2) ing. Consequently, the set of the tested DOFs square matrix. Since the number of knowns
requires reducing the number of DOFs of a in Equation 8 are greater than the number of
Where α and β are the damping coefficients large model without losing any information unknowns, Equation 8 can be put into a nor-
(see “Damping matrix” section below). This or characteristics of the dynamic system in mal form by transforming this equation as:
technique uses the multi-objective function the modelling process.
solver to find the optimal solution of the The SEREP condensation method parti- Yp(t) = ΦTmpUm(t)(10)
function given in Equation 2. MATLAB code tions the degrees of freedom into a set of
was developed to analyse the beam based on slave DOFs and master DOFs. The DOFs are Substituting Equation 8 into Equation 10
the steps outlined in Figure 1. arranged to place the slave DOFs as the first produces:
The comparison of the identified stiffness s coordinates, while the remaining master
matrices Kc and reduced theoretical stiffness DOFs are the last m coordinates (Friswell Yp(t) = ΦTmpΦmp q̌ p (t)(11)
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 39
in which q̌p(t) is an approximate solution of
qp(t) (O’Callahan et al 1989). Although the
square coefficient matrix of q̌p(t) will in gen- 1
eral be of full rank and possess an inverse, 2ζ
the determining of the inverse of this matrix 1
using standard methods may encounter 2√2ζ
some numerical difficulty, and singular-value
Amplification
decomposition solution is usually required.
Symbolically q̌ p(t) could be solved from
Equation 11 as:
q̌ p(t) = Yp(t)[ΦTmpΦmp]–1(12) 1
40 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
6. To determine the upper limit and lower
limit of the solution, it is assumed for this
study that the stiffness degradation is
x(n) between 140% and 80%, respectively.
7. A goal function value of zero is used in
order to get the optimal value of K in the
function given in Equation 2.
8. Weighting function values of unity are
used in this study.
The technique uses the reduced theoretical
stiffness matrix as a starting point where the
solver finds the best stiffness matrix values
x(n – 1) that satisfy the equation of motion and the
associated modal frequencies.
The output of the solution process is the
identified (optimised) stiffness matrix. Kc is
x(n – 2) then used for comparison purposes to locate
the damage.
EXPERIMENTAL STUDY
The structural system used in this study is
T T a simply-supported steel beam with a span
of 2.3 m. The cross-section of the beam is a
Figure 3 The bilinear transform of an integrator rectangular section of 100 mm by 21.9 mm.
The material properties of the beam are
as follows:
g
Mass density = 7 850 kg/m3
Young’s modulus E = 205.9 GPa
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 41
θ1 θ2 v2 θ3 v3 θ4 v4 θ5 v5 θ6
A 1 2 3 4 5 6 B
1 2 3 4 5
Ra Rb
(a) The undamaged steel beam model A-B with five elements
θ1 θ2 v2 θ3 v3 θ4 v4 θ5 v5 θ6
A 1 2 3 4 5 6 B
1 2 3 4 5
Ra Rb
(b) The damaged steel beam model A-B with five elements
(c) The simply-supported steel beam model specimen with accelerometers placed on the beam
v θ
Node points Vertical DOF Rotational DOF
Figure 5 ( a) An undamaged steel beam model A-B with five elements, (b) the damaged steel beam model A-B with five elements, (c) the simply-
supported steel beam model specimen with accelerometers placed on the beam
Table 1 The reduced stiffness and mass matrices using SEREP Table 2 D
ifferences between the damaged and the theoretical
undamaged stiffness matrices for the sample steel beam
DOFs that K Theoretical reduced Theoretical reduced
is reduced to stiffness matrix (N/m) × 10 4 mass matrix (Kg) Identified changes in
DOFs that K Theoretical changes in the
stiffness matrix using
reduced to stiffness matrix (ΔK%)
6.700 –3.715 1.817 0.075 TPC technique (ΔK%)
[v 2 – v4]
–3.715 2.850 0.075 1.744
28.13 16.63 27.37 19.13
[v 2 – v4]
16.63 10.71 19.13 10.19
3.778 –2.694 2.297 1.056
[v 2 – v5]
–2.694 3.778 1.056 2.313 25.19 13.19 24.39 14.18
[v 2 – v5]
13.19 4.43 14.18 4.24
8.721 –8.479 2.366 –1.062
[v 3 – v4] 25.81 18.96 24.59 19.86
–8.472 9.021 –1.062 2.218 [v 3 – v4]
18.96 15.30 19.86 14.58
3.062 –4.137 1.861 –0.089 18.99 10.57 18.06 11.70
[v 3 – v5] [v 3 – v5]
–4.2137 7.447 –0.0897 1.831 10.57 5.39 11.70 5.17
the correct mass matrix by visual inspec- span and then cutting it after having pulled signals obtained from the four channels
tion of the model and to use the applicable it downwards. using the trapezoidal method. The accelera-
shop drawings. Figures 6(a) and 6(b) show the accelera- tion data is collected at a sampling rate of
The motion of the beam is initiated tion data captured during the test, as well 500 Hz.
by using a pullback-quick release method as the velocity and displacement signals To apply the damage detection algorithm
by tying a rope to the middle of the beam obtained by integrating the acceleration developed, as described in this paper, the
42 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
(a) The recorded acceleration signals from four channels for five-element steel beam
1 000
500
Acceleration (mm/sec2)
–500
–1 000
–1 500
5 10 15 20 25 30
Time (sec)
5
0
–5
–10
–15
5 10 15 20 25 30
Time (sec)
0.2
Displacement (mm)
0.1
–0.1
–0.2
5 10 15 20 25 30
Time (sec)
Figure 6 ( a) The recorded acceleration signals from four channels for five-element steel beam, (b) the velocity signals for five-element steel beam,
(c) the displacement signals
theoretical stiffness matrix Ko and mass Using the SEREP condensation method, frequencies for the measured values for the
matrix M of the system are reduced to the reduced stiffness and mass matrices are four channels of the five-segment beam
two DOFs using the SEREP method. The produced for each set of two DOFs, as shown model. The frequency of the first mode for
DOFs are selected to cover all the pos- in Table 1. all nodes is 9.4 Hz, while the damping ratio is
sible combinations of vertical DOFs. The By plotting the Fourier amplitude spectra 0.0185. The second mode has a frequency of
selected DOFs are v 2 – v4 , v 2 – v5, v 3 – v4 , and using Equations 23 through 26, the 38 Hz, and the damping ratio is 0.0681.
and v 3 – v5. The resultant matrices for the damping matrix, frequencies and damping The identified stiffness matrix Kc , which
condensed structural system are given in ratios are calculated. Figure 8 shows the is obtained as output from the optimisation,
Figure 7. Fourier amplitude spectra for the first three is compared with the reduced theoretical
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 43
v2 v4
Kv v Kv v
1 2 3 4 5 6 Kv = 2 2 2 4
2 –v4 Kv v Kv v
1 2 3 4 5 4 2 4 4 2x2
v2 v5
Kv v Kv v
1 2 3 4 5 6 Kv = 2 2 2 5
2 –v 5 Kv v Kv v
1 2 3 4 5 5 2 5 5 2x2
v3 v4
Kv v Kv v
1 2 3 4 5 6 Kv = 3 3 3 4
3 –v4 Kv v Kv v
1 2 3 4 5 4 3 4 4 2x2
v3 v5
Kv3v3 Kv3v5
1 2 3 4 5 6 Kv3 –v5 =
Kv5v3 Kv5v5 2x2
1 2 3 4 5
Figure 7 Reduced stiffness and mass matrices of the beam into sets of 2 DOF systems
damage. ch-4
For the condensed stiffness matrix Kv 2–v4
using DOFs (v 2 – v4), the percentage change 6 ch-1
in coefficient k v 2v 2 is larger than the change
in coefficient k v4v4 . This indicates that the
damage is located near to node 2 rather than 4
near to node 4. The same conclusion can
be established from ΔK for DOFs v 2 – v4,
v 2 – v5, v3 – v4, and v3 – v5 ( see Table 2 2
column 1).
44 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
(a) The specimen with sensors
θ1 θ2 v2 θ3 v3 θ4 v4 θ5 v5 θ6 v6 θ7
A 1 2 3 4 5 6 7
1 2 3 4 5 6
Figure 9 A
2.3 m steel beam specimen divided into six segments, where the damaged part is located between points 2 and 4: (a) the specimen with
sensors, (b) geometrical details
Table 3 P
ercentage changes in the coefficients of the reduced stiffness matrices for the measured large change that occurs in the coefficient
and the theoretical cases for the steel beam with six segments k v 2v 2 . The result for [v3 – v4] shows the large
changes concentrated in k v3v3 rather than
Identified changes in stiffness Theoretical changes in the
DOFs that K reduced to k v4v4 because node 3 is located at the dam-
matrix using TPC technique (ΔK%) stiffness matrix (ΔK%)
aged segment, while node 4 is outside the
23.81 15.69 22.61 16.74 damaged segment.
[v 2 – v4]
15.69 11.99 16.74 11.07
The result of [v4 – v5] indicates the
[v 2 – v5]
24.04 13.01 22.63 14.21 damage at the left side of the segment; both
13.01 7.20 14.21 6.14
nodes are, however, outside the damaged
34.14 25.53 33.06 26.80 segment. The result of [v4 – v5] provides a
[v 3 – v4]
25.53 21.31 26.80 19.80 good indication of the location of damage.
Table 4 P
ercentage changes in the coefficients of the reduced stiffness matrices for the measured and the theoretical cases for the steel beam with
five segments
ΔK using TPC technique (%)
DOFs that K reduced to
Plate size Plate size Plate size Plate size Plate size
100 mm × 430 mm 100 mm × 360 mm 100 mm × 290 mm 100 mm × 230 mm 100 mm × 170 mm
17.56 11.14 11.06 6.22 6.46 2.95 3.73 1.21 1.90 0.22
[v 2 – v4]
11.14 5.48 6.22 2.78 2.95 1.10 1.21 0.28 0.22 2 * 10–4
17.70 9.49 12.35 5.81 8.25 3.26 5.56 1.79 3.48 0.83
[v 2 – v5]
9.49 2.71 5.81 1.46 3.26 0.67 1.79 0.29 0.83 0.08
22.31 17.96 19.61 15.69 16.26 12.99 12.99 10.31 9.53 7.49
[v 3 – v4]
17.96 13.12 15.74 11.99 12.99 9.40 10.31 7.40 7.49 5.30
18.43 12.83 16.89 11.99 14.58 10.46 12.05 8.67 9.12 6.53
[v 3 – v5]
12.83 6.53 11.99 6.30 10.46 5.60 8.67 4.6 6.53 3.50
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 45
3 000
2 000
Acceleration (mm/sec2)
1 000
–1 000
–2 000
5 10 15 20 25 30
Time (sec)
40
30
20
Velocity (mm/sec)
10
0
–10
–20
–30
–40
0 5 10 15 20 25 30 35 40 45
Time (sec)
0.8
0.6
0.4
Displacement (mm)
0.2
0
–0.2
–0.4
–0.6
–0.8
0 5 10 15 20 25 30 35 40 45
Time (sec)
Figure 10 Recorded acceleration signals, velocity signals and displacement signals captured at points 2 to 5 for the sample steel beam
reducing the size of the damaged area. The indicate the stiffness degradation caused by two-points condensation technique (TPC)
same beam shown in Figure 5 is considered the structural damage. The bigger difference were described. The approach using TPC
here. For this case, the damage size is reduced between Kii and Kjj for any given i and j indi- represents a non-destructive test method
by cutting a piece off the plate. The width of cates the damage location. that uses vibration signal records to identify
the plate remains the same, but the length The results shown in Table 4 demonstrate the structural damage according to the mea-
of the plate representing the damage is now that the technique is successful to find the sured changes in dynamic characteristics of
reduced to 430 mm, 360 mm, 290 mm, location of the damage in the beam, even if the structure.
210 mm and 170 mm, as shown in Figure 11. the damage is small. Investigating the results The experimental results obtained from
A 500 Hz sampling rate is used for all of all tests for one set of points can indicate a steel beam model structure demonstrate
tests, and the global stiffness matrices are the ability of the technique to reflect the the usefulness of the TPC technique. This
reduced to the following DOFs: ([v 2 – v4], degradation in stiffness. method has several advantages:
[v3 – v4] and [v3 – v5]. The recorded signals ■■ The concept is simple and easy to apply
are pre-processed, and the results of five without using complex calculations.
tests using the TPC technique are shown in CONCLUSIONS ■■ Using the TPC technique, the structure
Table 4. This study investigated the changes in the is reduced to a 2-DOF structure; this
The percentage differences of the values stiffness due to the presence of structural decreases the volume of data to be
of the coefficients of the stiffness matrices damage. Experimental studies using the captured and the volume of data to be
46 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
Koh, C G, Tee, K F & Quek, S T 2006. Condensed model
identification and recovery for structural damage
assessment. Journal of Structural Engineering, ASCE,
132: 2018–2026.
Leis, J W 2011. Digital Signal Processing Using MATLAB
for Student and Researchers, 1st ed. New York: Wiley.
Liang, W & Chan, T H T 2009. Review of vibration-
based damage detection and condition assessment of
bridge structures using structural health monitoring.
(a) 430 mm Paper presented at the 2nd Infrastructure Theme
Postgraduate Conference: Rethinking Sustainable
Development: Planning, Engineering, Design and
Managing Urban Infrastructure, 26 March 2009,
Brisbane, Australia.
Monson, H H 1996. Statistical Digital Signal Processing
and Modeling, 1st ed. New York: Wiley.
O’Callahan, J, Avitabile, P & Riemer, R 1989. System
(b) 360 mm (c) 290 mm equivalent reduction expansion process (SEREP).
Proceedings, 7th International Modal Analysis
Conference, Las Vegas, NV, 30 January – 2 February
1989, 29–37.
Pandey, A K, Biswas, M & Samman, M M 1991. Damage
detection from changes in curvature mode shapes.
Journal of Sound and Vibration, 145(2): 321–332.
Pokharkar, P V & Shrikhande, M 2010. Structural
health monitoring via stiffness update. ISET Journal
(d) 210 mm (e) 170 mm of Earthquake Technology, 47(1): 47–60.
Qu, Z-Q 2004. Model Order Reduction Techniques
Figure 11 S imply-supported five-segment beam configured for each case by reducing the length of with Applications in Finite Element Analysis, 1st ed.
plate, where the lengths are: (a) 430 mm, (b) 360 mm, (c) 290 mm, (d) 210 mm, (e) 170 mm London: Springer Verlag.
Reddy, J N 2006. An Introduction to the Finite Element
dealt with at a time. Using more tests will Chopra, A K 2012. Dynamics of Structures: Theory and Method, 3rd ed. New York McGraw-Hill.
help to obtain more precise values for Applications to Earthquake Engineering, 4th ed. Salawu, O S & Williams, C 1995. Bridge assessment
the change in stiffness coefficients and Harlow, UK: Pearson Education. using forced-vibration testing. Journal of Structural
improve damage location. Czichos, H 2013. Handbook of Technical Diagnostics Engineering, 121(2): 161–173.
■■ Accuracy is achieved in detecting the Fundamentals and Application to Structures and Silva, D & Clarence, W 2000. Vibration Fundamentals
location of the damage where the TPC Systems. Berlin, Heidelberg, Germany: Springer Verlag. and Practice. Boca Raton, FL: CRC Press.
technique results have demonstrated Doebling, S W & Farrar, C R 1996. Computation of Sinou, J J 2009. A Review of Damage Detection and
good agreement with actual results. structural flexibility for bridge health monitoring Health Monitoring Mechanical Systems from
The results indicate that the TPC technique using ambient modal data. In: Lin, Y K & and Su, Changes in the Measurement of Linear and Non-
is capable of indicating the damage location T C (Eds.), Proceedings, 11th ASCE Engineering linear Vibrations. Hauppauge, NY: Nova Science
for different sizes of damage. According to the Mechanics Conference held at Fort Lauderdale, FL, Publishers, 643–702.
damage size sensitivity tests, the technique 20–22 May 1996. New York: ASCE Press, 1114–1117. Stubbs, N, Kim, J T & Topole, K 1992. An efficient and
can show stiffness degradation. Additionally, Edward, B M, Shapour, A & Balakumar, B 2011. An robust algorithm for damage location in offshore
experimental results demonstrate that the Engineer’s Guide to MATLAB, 3rd ed. Upper Saddle platforms. Proceedings, ASCE 10th Structures
SEREP reduction method is suitable for use River, NJ: Pearson Education. Congress, 13–15 April 1992, San Antonio, TX,
with the TPC technique. The damping effect Farrar, C R & Doebling, S W 1999. Damage detection 543–546.
has to be accounted for to ensure accurate and evaluation. II: Field applications to large Tadeusz, S, Tadeusz, U & Wieslaw, S 2013. Advanced
results when applying this solution technique. structures. In: Silva, I M M & Maia, N M M (Eds.), Structural Damage Detection. New York: Wiley.
Finally, the TPC technique shows the Modal Analysis and Testing, NATO Science Series Udwadia, F E 2009. A note on nonproportional
location of damage regardless of the place E, Vol. 363, London: Kluwer Academic Publishers, damping. Journal of Engineering Mechanics, 135(11):
of the acceleration sensors or the size of the 345–378. 1248–1256.
damage. The technique finds the location of Friswell, M I & Motiershead, J E 1995. Finite Element
damage when the acceleration sensors are Model Updating in Structural Dynamics. Dordrecht,
near to the damage, as well as when they are Netherlands: Springer Science & Business Media. LIST OF ACRONYMS
far from the damage. Friswell, M I 2008. Damage Identification using Inverse COMAC Coordinate Modal Assurance
Methods. Vol. 499 of the series CISM International Criterion
Centre for Mechanical Sciences. Vienna, New York: DOF Degrees Of Freedom
REFERENCES Springer Verlag, 13–66. FFT Fast Fourier Transform
Chance, J E, Worden, K & Tomlinson G R 1994. Kammer, D C 1987. Test-analysis model development FEM Finite Element Method
Processing signals for damage detection in using an exact modal reduction. International MAC Modal Assurance Criterion
structures using neural networks. Proceedings, SPIE Journal of Analytical and Experimental Modal SEREP System Equivalent Reduction
2191, Smart Structures and Materials, 187. Analysis, 2(4): 174–179. Expansion Process
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 47
SHM Structural Health Monitoring Φ+mp The generalised inverse of matrix Φmp ∆k ji The difference in the coefficients of
TPC Two-Points Condensation ωn The angular frequency the Kc matrix and the Kr matrix at
Technique C Damping matrix position (j, i)
Cr Damping matrix of the reduced system ∆k jj The difference in the coefficients of
K Globel stiffness matrix the Kc matrix and the Kr matrix at
LIST OF SYMBOLS Ko Theoretical stiffness matrix position (j, j)
α, β Damping coefficients Kc Identified stiffness matrix M Globel mass matrix
θi The rotational degrees of freedom at Kr Undamaged reduced stiffness matrix Mr Mass matrix of the reduced system
node i ∆Ki–j The change matrix of the set of nodes q(t) The modal coordinate vector
vi The degrees of freedom of vertical i and j q̌p(t) Approximate solution of δp(t)
displacements at node i ∆kii The difference in the coefficients of u Displacement vector
ξn The damping ratio for the nth mode the Kc matrix and the Kr matrix at u̇ Velocity vector
ξi Damping ration of mode ith position (i, i) ü Acceleration vector
ξj: Damping ration of mode jth ∆kij The difference in the coefficients of ü(t) Acceleration response vectors
Φ The complete eigenvector matrix of the Kc matrix and the Kr matrix at u(t) Displacement response vectors
the full model position (i, j) T The coordinate transformation matrix
48 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
The application of continuous TECHNICAL PAPER
surface wave testing for Journal of the South African
Institution of Civil Engineering
settlement analysis with ISSN 1021-2019
Vol 59 No 2, June 2017, Pages 49–58, Paper 1245
Pont Melin, Wales, UK Pretoria, and a PhD from the University of Surrey.
He has been involved with teaching and research
in geotechnical engineering for many years. His
field of interest includes the characterisation of
soil behaviour and its application in geotechnical engineering.
G Heymann, J Rigby-Jones, C A Milne
Contact details:
Department of Civil Engineering
University of Pretoria
The presence of a significant thickness of soft silts beneath the site of a proposed new Pretoria 0002
road bridge caused concerns over the suitability of shallow foundations. Despite extensive South Africa
conventional ground investigations, it was not possible to obtain reliable stiffness data, and so a T: +27 12 420 3627
E: gerhard.heymann@up.ac.za
full-scale, fully instrumented load test was undertaken to better determine ground stiffness. The
data obtained was used in conjunction with sophisticated numerical analysis to demonstrate that
JOHN RIGBY-JONES (CEng, EurIng, MICE) holds
the settlement of shallow foundations would be acceptable. Following completion of the bridge BEng and MSc degrees, and has more than 24
construction, Continuous Surface Wave testing was undertaken and a range of simple non-linear years of geotechnical engineering, geophysical
elastic settlement analyses conducted to model the behaviour of the load test. Through these and research experience gained from
analyses this paper demonstrates the benefit of using continuous surface wave ground stiffness universities, large consultancies, and lately RJM
Ground Solutions Ltd, a specialist consultancy
testing in conjunction with simple analysis methods. It underlines the importance of obtaining
set up with co-author Chris Milne. At RJM, John
high-quality stiffness data to accurately predict settlement of shallow foundations. led the development of a commercial CSW
(continuous surface wave) system, which was recognised in the 2013 Ground
Engineering Awards where RJM was a finalist in the Ground Investigation
INTRODUCTION the new bridge on spread foundations. Specialist of the Year category. In 2016 John and Chris set up Ground Stiffness
Surveys Ltd in order to focus on the delivery of a CSW testing service.
The existing 18th century masonry arch The thickness of the silt deposits limited
bridge which crosses the river Afon Alwen options for deep foundations (Solera & Contact details:
Ground Stiffness Surveys Ltd
at Pont Melin, 2 km west of Corwen in Milne 2009). Accurate settlement prediction
35 Church Hill
North Wales, was replaced in 2008 with a was required given the tight tolerance Arnside
new 45° skew bridge with 60 m single span of the bridge bearings to foundation Cumbria
between bearings (Figure 1). The presence of movements. Consequently, extensive ground LA5 0DJ
a 50 m thick sequence of saturated, soft silts investigations and a large-scale load test United Kingdom
T: +44 7879 457210
of glacio-lacustrine origin (the Rûg Silts) were undertaken at considerable cost to
E: john@groundstiffnesssurveys.com
caused concern that excessive settlement provide data for settlement analysis of
would be generated by construction of the foundations. CHRIS MILNE (CEng, FICE, EurIng, CGeol, FGS,
EurGeol) holds BSc (Hons) and MSc degrees and
a Postgraduate Diploma in Geotechnical
Engineering Design & Management. He is
currently a director of Ground Stiffness Surveys, a
specialist CSW ground stiffness profiling
developer and contractor established with fellow
director John Rigby-Jones. Chris has over 25 years
of geotechnical engineering, geophysical and commercial experience,
including for large consultancies working at senior level on major infrastructure
projects. His current role focuses on promoting the commercial application of
CSW testing in UK practice.
Contact details:
Ground Stiffness Surveys Ltd
35 Church Hill
Arnside
Cumbria
LA5 0DJ
United Kingdom
T: +44 7921 492853
E: chris@groundstiffnesssurveys.com
Keywords: continuous surface wave test, CSW, soil stiffness, non‑linear analysis,
Figure 1 S ite location foundation settlement
49
Heymann G, Rigby-Jones J, Milne CA. The application of continuous surface wave testing for settlement analysis with reference to a full-scale load test for a
bridge at Pont Melin, Wales, UK. J. S. Afr. Inst. Civ. Eng. 2017:59(2), Art. #1245, 10 pages. http://dx.doi.org/10.17159/2309-8775/2017/v59n2a6
Accurate measurement of stiffness in
SPT N-value (blows/300 mm)
the poorly understood and sensitive Rûg
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75
Silts was challenging, with a wide range 0
of values and significant scatter identified
Alluvium
from data from different sources. Solera and 5
Milne (2009) give details of the settlement Rûg Silts
analysis, which adopted a conservative 10
approach and resulted in predictions of
settlement in the range of 75–85 mm.
Depth (m)
15
This level of settlement provided
significant constraints on the construction 20
methodology, with the installation of
bearings and landing of deck beams 25
requiring programme hold periods and
careful monitoring. The actual settlements 30
Excludes refusals
suffered by the bridge foundations were on boulders
observed to be less than half the predicted 35
values. Whilst conservatism was justified SI 1986 SI 2002 SI 2003 BH201 SI 2003 BH201a
in terms of the scale and risk of the project,
the availability of accurate stiffness values Figure 2 S tandard penetration test (SPT) results
would have permitted a less conservative
and simplified analysis approach, and
provided greater confidence in the results Cone resistance (MPa)
and potentially removed the constraints 0 4 8 12 16 20 24
0
on construction associated with the
high-predicted settlements.
Seismic testing techniques have been Alluvium
used successfully for the design of many
projects around the world (Stokoe et al
2004), but its wider adoption has been
limited to date by the familiarity of design-
ers with more conventional, often intrusive
Rûg Silts
and empirically based methods. Continuous 10
Surface Wave (CSW) testing is a rapid and
cost-effective technique for measuring
seismic surface wave velocities at a range
of frequencies allowing an in situ ground
stiffness profile to be determined. Unlike
the most frequently used techniques for
assessing ground stiffness, CSW is non-
intrusive and measures the undisturbed
Depth (m)
50 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
Excavation
Concrete footing (5 m × 5 m)
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 51
of the footing is shown in Figure 6. At 200
the maximum applied load of 442 T, cor-
responding to a contact stress of 176.8 kPa, 180
the average settlement of the levelling points
was 18.3 mm. The maximum settlement 160
observed was 21.8 mm and the maximum
140
differential settlement was 5.7 mm.
Where:
E’ = drained Young’s modulus
∆σ’v = vertical effective stress increment
∆σ’r = radial effective stress increment Figure 7 G
round profile at load test
52 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
∆εv = vertical strain increment 12
v’ = Poisson’s ratio.
CONTINUOUS SURFACE
WAVE TESTING –4
The Continuous Surface Wave (CSW) test 0 50 100 150 200
is a seismic test which uses a vibratory Contact stress (kPa)
source placed on the ground surface. It Top S8 S7 S6
induces Rayleigh waves which radiate along
the ground surface and are detected by a Figure 8 E xtensometer settlement
linear array of geophones also placed on the
ground surface. The vibrator can be used as a motion is insignificant at depths below one Fundamental mode propagation is
harmonic source at a particular frequency or wavelength. dominant in normally dispersive profiles,
as a transient source which sweeps through An important part of the analysis of where stiffness increases with depth. For
a frequency range of interest. The CSW CSW results is the inversion of the disper- inversely dispersive profiles, where soft
test has the advantage over impact source sion data. Inversion is the process by which layers occur below stiff layers, higher
techniques (SASW and MASW) that the theoretical Rayleigh wave dispersion curves modes also contribute to Rayleigh wave
frequency content of the source is known of a model ground profile are matched with propagation. Correct mode identification
and can be controlled. This allows the signal the measured dispersion data. Calculation of of the experimental data is important to
quality to be assessed in the field and a poor the theoretical dispersion curves of a model ensure matching of the same modes of the
signal or interference from background noise ground profile requires the solution of the theoretical and experimental data. Leong
is easily detected. Also, a vibratory source Rayleigh eigenvalue problem. Thompson and Aung (2013) developed a weighted
provides a significantly higher signal-to-noise (1950) and Haskell (1953) used the transfer average velocity method to calculate the
ratio compared to impact sources (Jiang et matrix method and formulated an algorithm effective Rayleigh phase velocity which
al 2015). Simons et al (2002) found the CSW to model surface waves propagated by a includes the contribution of more than one
test to give more repeatable results than the layered medium known as the Thomson- mode when mode superposition occurs. The
SASW test. Haskell algorithm. Kausel and Roësset (1981) effective Rayleigh phase velocity implicitly
The geophone response is recorded for derived a formulation based on the finite accounts for the existence of higher modes,
a range of source frequencies, and at each element dynamic stiffness matrix method, and therefore eliminates the requirement
frequency the Rayleigh wave phase velocity and Kennett (1974) developed the reflec- for mode identification. This is particularly
is calculated. Due to the dispersive nature tion and transmission coefficient method. useful for inversely dispersive profiles, as
of ground profiles, the phase velocity is Numerous other techniques have been used higher modes also contribute strongly to
frequency-dependent. High frequency to model surface waves propagated by a lay- Rayleigh wave propagation. The weighted
excitation induces short wavelengths which ered medium, including the finite difference average velocity method ignores any contri-
penetrate only to shallow depth, whereas low method (Boore 1972), numerical integration bution from near field effects.
frequencies generate long waves which pen- (Takeuchi & Saito 1972) and the boundary Search algorithms are used as part of the
etrate deeper. The response at high frequency element method (Manolis & Beskos 1988). inversion analysis to find the model ground
is therefore determined by the shallow soils, These methods determine the Rayleigh dis- profile for which the dispersion curves
whereas the response at low frequency is persion curves for fundamental and higher best match the measured data. Numerous
also influenced by deeper material. Ground modes of Rayleigh wave propagation. algorithms have been proposed, including
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 53
neighbourhood algorithms (Wathelet 2008), 280
genetic algorithms (Lomax & Snieder
1994) and simulated annealing algorithms 260
(Sen & Stoffa 1991). The inversion process
is ill-posed, which implies that there may 240
be two or more theoretical profiles which
54 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
of 18.3 mm. However, it must be noted that
Go (MPa)
there is a degree of good fortune that the
0 20 40 60 80 100 120 140
0 selected softened strain level and actual
strain level correspond at this loading stage.
1
Stepwise non-linear
Depth below load test foundation (m)
2 settlement analysis
This method allows simple linear elastic
3 settlement calculations to be used to model
the non-linear behaviour of soil without
4
the need to assume an operational strain
level. The maximum stress σ v is applied
5
incrementally in a number of steps, with
the layer stiffnesses in each step being
6
progressively softened from an initial value
7
of E 0 to correspond to the calculated strain
level from the previous step. The calculated
8 strains from all previous steps are summed
Test 01 Test 02 Test 03 to determine the applicable strain for a
particular step. The methodology for the
Figure 10 S tiffness profiles from CSW tests stepwise non-linear settlement analysis is
shown in Figure 11.
observed behaviour. A number of analyses E’ 1 An analysis was undertaken using a total
= (2)
were undertaken with a range of complex- E’o 1 + 16γ(1.2 + 10 –20γ) of five steps and the GEO5 Spread Footing
ity. However, all calculation methods were software. The progressive stiffness reduc-
based on the application of a simple linear Where: γ is the current shear strain of a tion of the layer stiffnesses was undertaken
elastic soil model which subdivided the sub-layer. using the calculated vertical strain from
gravel alluvium into three layers of thick- the previous step and not the shear strain.
nesses (0.6 m, 1.0 m and 1.0 m respectively) Single-stage loading of The results of the 5-step non-linear GEO5
and the underlying Rûg Silts into layers of manually softened soil analysis are shown in Figure 6 and demon-
thicknesses 0.4 m, 0.5 m, 1.0 m, 3.0 m and In the simplest analysis method the non- strate a clear non-linear response represent-
4.5 m respectively. The same layered model linear response of the soil was modelled ing the progressively softened stiffness
was used for all analyses. A rigid bound- through a wholesale reduction of all stiff- values used in the analysis. It can be seen
ary was assumed at 12 m depth equal to ness values to an assumed operational that the analysis overestimates the settle-
2.4 times the foundation width and below strain level. Mair (1993) suggested that ments at low stress levels, although there
the depth of any significant stress increase. strain levels beneath typical shallow foun- is a significant improvement in accuracy
The small-strain shear modulus (Go), dations are of the order of 0.1% and this over the single-step analysis. Overall there
as measured by the CSW test, was used value was used as the basis for reducing the is a much improved agreement between the
to assign a small-strain drained Young’s small-strain stiffness values. Application observed data, with the calculated settle-
modulus (E’o) to each layer. E’o = 2(1 + v’)Go of the Rollins et al (1998) stiffness reduc- ment of 17.6 mm at the final loading stage
where the Poisson’s ratio (v’) was taken tion function (Equation 2) to the midrange being close to the observed settlement of
as 0.26, which is a typical value for CSW small-strain stiffness values for the 18.3 mm.
drained loading of intermediate soils. alluvium and Rûg Silts resulted in drained To explore the effects of undertaking
Axisymmetrical conditions were assumed Young’s modulus values of 83.9 MPa and stepwise analysis using a larger number of
in all analyses, and Boussinesq’s theory 22.7 MPa respectively. The settlement of steps, software (Versak v2.0) was written
used to calculate the vertical and horizontal the load test foundation was then calculated to undertake a 100-step analysis where
stress increment for each layer below the using these softened stiffness values at each the stiffness for each layer is automatically
centre of the footing. load stage with the proprietary software softened after each load increment. In
In all analyses using small-strain GEO5 Spread Footing which facilitates the contrast to the GEO5 analysis, stiffness
stiffness values, the non-linear stiffness automated linear elastic analysis of a shal- reduction was undertaken using the shear
response of the soil to straining was low foundation on a layer model. strain determined from the calculated verti-
accounted for. Many stiffness reduction The results of the GEO5 analysis are cal strains using ε s = ⅔ε v(1 + v’).
curves have been proposed to describe shown in Figure 6 and demonstrate a linear The result of the 100-step non-linear
the non-linear stress strain relationship load settlement response corresponding Versak analysis is also shown in Figure 6.
observed for soils, including Vucetic and to the softened soil stiffness values. It can As would be expected the 100-step analysis
Dobry (1991), Rollins et al (1998), Clayton be seen that at intermediate load stages, provides a smoother non-linear response
and Heymann (2001), and Archer and settlements are over-estimated due to the curve than the 5-step analysis. Whilst both
Heymann (2015). For all analyses repre- use of overly softened stiffness values corre- the 5-step and 100-step curves provide very
sentative stiffness values were calculated by sponding to a strain in excess of true values. similar overall load settlement curves, it is
applying the stiffness reduction function The error reduces with increased loading noted that the 100-step analysis provides a
suggested by Rollins et al (1998, Equation 2 and strains in the ground until, at the final slightly softer response, reflecting the more
below) to the midrange values determined load stage, the calculated settlement of frequent stiffness reduction. The calculated
from the CSW tests. 18.9 mm is close to the observed settlement settlement of 18.7 mm at the final loading
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 55
have been omitted for clarity. Clearly reliable
stiffness parameters could not be obtained
for the alluvial gravel from the standard
penetration test.
σ v3
The effect of disturbance on the SPT
results from the Rûg Silts were clear from
the significantly higher SPT values that
E2 were obtained when mechanisms were put
σ v2
in place to ensure water balance during all
σv
56 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
It is an attractive test method, because it material parameters than increasing the was undertaken by Hyder Consulting (UK)
is a non-disturbing and non-intrusive test sophistication of the analysis method. Limited (now Arcadis Consulting UK
and because the seismic wave velocity of the Limited).
material is directly related to stiffness. It is
therefore well suited to quantify the small- CONCLUSIONS
strain stiffness of the geomaterials encoun- Valuable lessons were learnt during the series REFERENCES
tered at the location of the bridge at Pont of site investigations and field testing that Archer, A & Heymann, G 2015. Using small‑strain
Melin. However, small-strain stiffness (Go were conducted before and after the con- stiffness to predict the load‑settlement behaviour of
or Eo) cannot be used directly for settlement struction of the new bridge at Pont Melin. shallow foundations on sand. Journal of the South
calculation, as the stress-strain behaviour of Quantifying stiffness parameters from African Institution of Civil Engineering, 57(2): 28–35.
soil is highly non-linear and an appropriate conventional in situ tests such as the SPT DOI: 10.17159/2309-8775/2015/v57n2a4.
stiffness reduction function is required as and CPT, as well as laboratory testing, was Boore, D M 1972. Finite difference methods for seismic
part of the settlement analysis. difficult and achieved only limited success. wave propagation in heterogeneous materials. In:
Settlement analyses to predict the load For the alluvial gravel the large scatter in the Bolt, B.A (Ed.), Methods of Computational Physics,
settlement behaviour of the full-scale SPT and CPT data made selection of the rep- Vol. 11, New York: Academic Press, 1–36.
load test conducted at Pont Melin were resentative stiffness difficult. The Rûg Silts Clayton, C R I & Heymann, G 2001. The stiffness of
undertaken using a range of simple analysis were particularly sensitive to disturbance geomaterials at very small strains. Geotechnique,
techniques in conjunction with the stiffness and the SPT data from the first site investiga- 51(3): 245–256.
parameters determined from CSW testing. tion was anomalously low, due to borehole Clayton, C R I 1995. The standard penetration test
This is in contrast to the more complex disturbance as a result of poor water balance. (SPT): Methods and use. CIRIA report R143. London:
finite element analysis undertaken as part of Only during later site investigations, when CIRIA
the bridge design. full time-site supervision was available and Clayton, C R I, Matthews, C & Simons, N E 1995. Site
The first single-stage manually softened particular care was taken to ensure good Investigation, 2nd ed. Oxford: Blackwell Science.
analysis requires the average operational water balance, were more reliable SPT data Haskell, N A 1953. The dispersion of surface waves on
strain level beneath the foundation to be obtained. Even then the uncertainty associ- multilayered media. Bulletin of the Seismological
assumed, and therefore at strain levels ated with the selection of an appropriate Society of America, 43: 17–34.
below the assumed value this method over- empirical relationship between SPT N-values Heymann, G 2013. Vibratory sources for continuous
estimates settlements and at higher strains and stiffness resulted in the adoption of con- surface wave testing. In: Coutinho, R Q & Mayne,
provides an underestimation of settlement. servative stiffness values for design. P W (Eds.), Geotechnical and Geophysical Site
The method is, however, rapid and uses the The CSW test was successfully used to Characterization (ISC’4), London: Taylor and
conventional linear-elastic analysis approach rapidly quantify the stiffness of the alluvial Francis, 1381–1386.
widely adopted by practising engineers for gravel and Rûg Silts, due to the fact that Jiang, P, Tran, K T, Hiltunen, D R & Hudyma, N 2015.
routine analyses. it is non-invasive and does not cause any An appraisal of a new generation of surface wave
The use of a stepwise analysis method disturbance during testing. In addition, the techniques at a test site in Florida. Paper presented
allows the non-linear response of the soil to CSW test excites a large volume of material at the International Foundations Congress and
be modelled. The use of proprietary software and therefore quantifies the mass stiffness Equipment Expo 2015 (IFCEE), 17–21 March, San
requires user intervention between each of the material and avoids the difficulties Antonio, TX.
step to interrogate the output strains in each which large gravels and cobbles cause for tra- Kausel, E & Roësset, J M 1981. Stiffness matrices for
layer, undertake reduction of layer stiffnesses ditional invasive tests. Clearly, the CSW test layered soils. Bulletin of the Seismological Society of
and adjust the input layer stiffnesses for the proved to be a more suitable test to quantify America, 71(6): 1743–1761.
next analysis step. Even whilst automating the stiffness of both the alluvial gravel and Kennett, B L N 1974. Reflections, rays, and
this process with a spreadsheet as far as the Rûg Silts. reverberations. Bulletin of the Seismological Society
possible, the process is time-consuming. A The use of a stepwise non-linear analysis of America, 64(6): 1685–1696.
small number of load steps are typically used method over a single-stage method was Leong, E & Aung, A 2013. Global inversion of surface
to keep analysis time within practical limits; shown to provide a significantly more accu- wave dispersion curves based on improved weighted
however, the results obtained from a 5-step rate model of the observed settlement. Little average velocity method. Journal of Geotechnical and
analysis provided a reasonable approxima- improvement was gained by increasing the Geoenvironmental Engineering, 139(12): 2156–2169.
tion to the observed load settlement curve. number of analysis steps from 5 to 100. Lomax, A J & Snieder, R 1994. Finding sets of acceptable
The automated stepwise non-linear The use of CSW stiffness data, coupled solutions with a genetic algorithm with application to
analysis undertaken using Versak software with simple analysis techniques, compared surface wave group dispersion in Europe. Geophysical
allowed a large number of load steps (100) well with observed settlements, and provided Research Letters, 21(24): 2617–2620.
to be conducted efficiently with no user a far better prediction of settlements than Lunne, T & Christoffersen, P 1985. Interpretation
intervention. However, whilst the modelled the complex finite element analysis used dur- of cone penetrometer data for offshore sands.
load settlement curve provided a smoother ing the original bridge design in conjunction Norwegian Geotechnical Institute, Publication No.
response than the 5-step analysis, the with conservative soil stiffness values. 156, Oslo, Norway, 1–12.
improvement in accuracy to the observed Mair, R J 1993. Developments in geotechnical
data was small. engineering research: Application to tunnels
Engineers have to balance resources, ACKNOWLEDGEMENTS and deep excavation. Unwin Memorial Lecture.
including money, time and effort, between The authors are grateful for the support of Proceedings of the Institution of Civil Engineers, Civil
site investigation and design. Clearly both the Welsh Government, sponsor of the new Engineering, 93: 27–41.
these processes are important, but the above bridge, who facilitated CSW testing at the Mallard, D J 1983. Testing for liquefaction potential.
case history suggests that better results were load test site. The supervision of the load Proceedings, NATO Workshop on Seismicity and
achieved by focusing on obtaining accurate test and design of the bridge foundations Seismic Risk in the Offshore North Sea Area,
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 57
1–4 June 1982, Utrecht, Netherlands. Dordrecht: Solera, S A & Milne, C A 2009. The influence of soft ground Thomson, W T 1950. Transmission of elastic waves
Reidel Publishers, 289–302. conditions on new bridge design at A5 Pont Melin Rûg, through a stratified solid medium. Journal of Applied
Manolis, G D & Beskos, D E 1988. Boundary Element Corwen. Urban Geology in Wales: 3. National Museum Physics, 21, 89–93.
Methods in Elastodynamics. London: Unwin Hyman. of Wales Geological Series No. 26, Cardiff, 23–33. Tokimatsu, K 1988. Penetration testing for dynamic
Nichol, D & Reynolds, J M 2002. Application of Stokoe, K H, Joh, S & Woods, R D 2004. Some problems. In: De Ruiter, J (Ed.), Proceedings, 1st
reflection seismology to foundation investigation at contributions of in situ geophysical measurements to International Symposium on Penetration Testing
A5 Pont Melin Rûg, North Wales. Quarterly Journal solving geotechnical engineering problems. In: Viana (ISOPT-1), 20–24 March 1988, Orlando, FL, Vol. 1,
of Engineering Geology and Hydrogeology, 35, 247–256. da Fonseca, V & Mayne, P W (Eds.), Proceedings, Rotterdam: Balkema, 117–136.
Rollins, K M, Evans, M D, Diehl, N B & Daily, W D III ISC‘2 on Geotechnical and Geophysical Site Vucetic, M & Dobry, R 1991. The effect of soil plasticity
1998. Shear modulus and damping relationships Characterization held on 19–22 September 2004, on cyclic response. ASCE Journal of the Geotechnical
for gravels. ASCE Journal of Geotechnical and Porto, Portugal. Rotterdam: Millpress, 97–132. Engineering Division, 117(1), 89–107.
Geoenvironmental Engineering, 124(5): 396–405. Stroud, M A 1989. The standard penetration test – its Wathelet, M 2008. An improved neighborhood
Sen M K & Stoffa P L 1991. Nonlinear one-dimensional application and interpretation. Proceedings, ICE algorithm: Parameter conditions and dynamic
seismic waveform inversion using simulated Conference on Penetration Testing in the UK. scaling. Geophysical Research Letters, 35(9),
annealing. Geophysics, 56, 1624–1638. London: Thomas Telford, 29–46. DOI:10.1029/2008GL033256, 2008.
Simons, N E, Menzies, B K & Matthews, M C 2002. Takeuchi, H & Saito, M 1972. Seismic surface waves. In:
A Short Course in Geotechnical Site Investigation. Bolt, B A (Ed.), Methods of Computational Physics,
London: Thomas Telford. Vol. 11, New York: Academic Press, 217–294.
58 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
Effect of the minimum TECHNICAL NOTE
void ratio on the vertical Journal of the South African
Institution of Civil Engineering
intercept of the steady state ISSN 1021-2019
Vol 59 No 2, June 2017, Pages 59–64, TN-77
1.0
Γ1
Contact details:
0.9 Department of Civil and Environmental Engineering
University of the Witwatersrand
1 Private Bag 3
Void ratio (-)
Wits
0.8 λ10
2050
South Africa
T: +27 11 326 0030
E: mackechniep@gmail.com
0.7
0.6
1 10 100 1 000
Mean effective stress (kPa)
Keywords: steady state line, critical state line, minimum void ratio,
Figure 1 Idealisation of the SSL using Equation 1 non‑plastic soils
59
Torres-Cruz LA, Geyer S, Mackechnie PR. Effect of the minimum void ratio on the vertical intercept of the steady state line of non-plastic soils.
J. S. Afr. Inst. Civ. Eng. 2017:59(2), Art. #TN-77, 6 pages. http://dx.doi.org/10.17159/2309-8775/2017/v59n2a7
of researchers and is the focus of this note. 100
The e-p’ projection is commonly modelled
90
with Equation 1:
80
e = Γ1 – λ10log10(p’) (1)
60 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
Table 1 Values of p’ and e used to calculate Γ1
Soil type p’ (kPa) e Soil type p’ (kPa) e Soil type p’ (kPa) e Soil type p’ (kPa) e
5 0.801 13 0.854 277.9 0.577 83 0.534
10 0.796 SIM (100/10) 157 0.821 GB (0/2.5) 543.2 0.566 115 0.560
23 0.774 315 0.767 840.3 0.559 HCM (20/13) 221 0.543
41 0.774 25.9 0.779 131.6 0.470 268 0.556
48 0.765 162.5 0.732 268.6 0.460 515 0.482
HSd (0/1.1/0.9)e GB (0/5)
157 0.746 328.3 0.719 360.4 0.462 0.1 0.626
SIMa (0/1.7)b
473 0.718 639.0 0.700 546.0 0.448 0.1 0.612
937 c 0.668 158.9 0.730 136.6 0.421 0.5 0.603
937 0.645 HS (0/1.1/0.35)e 315.3 0.706 144.8 0.413 0.7 0.588
GB (0/10)
1 099 0.681 632. 6 0.697 271.1 0.404 1.0 0.532
HCM (30/14)
1 217 0.595 11.5 0.780 572.6 0.392 22 0.477
1 407 0.606 158.4 0.726 75.1 0.342 108 0.544
5 0.731 159.2 0.727 131.0 0.332 209 0.536
GB (10/20)
14 0.717 160.6 0.723 274.6 0.321 272 0.530
216 0.670 224.3 0.720 579.1 0.304 528 0.448
HS (0/1.4/0.9)e
SIM (7/2.1) 236 0.665 249.5 0.720 14 0.854 0.1 0.758
434 0.628 316.4 0.713 29 0.850 0.4 0.727
547 0.595 497.5 0.700 77 0.842 1.2 0.746
1 234 0.553 544.1 0.690 HCMg (0/2.4) 111 0.828 1.8 0.667
1 0.617 629.9 0.691 215 0.825 3 0.635
HCM (50/8.9)
17 0.589 165.3 0.723 268 0.807 109 0.683
42 0.575 HS (0/1.4/0.75)e 318.1 0.710 335 0.809 139 0.567
94 0.600 627.1 0.698 9 0.789 165 0.523
SIM (15/14)
124 0.575 9.6 0.774 54 0.771 208 0.673
306 0.522 12.8 0.770 98 0.781 262 0.673
306 0.513 61.3 0.743 HCM (5/3.4) 121 0.779 1.0 0.981
HS (0/2.5)
1 289 0.421 159.7 0.714 236 0.760 1.3 1.006
1 0.477 321.3 0.695 275 0.756 2.2 0.958
11 0.461 640.5 0.675 306 0.752 3 0.815
69 0.430 162.1 0.694 51 0.715 4 0.790
HCM (70/2.2)
73 0.416 323.4 0.671 64 0.693 12 0.748
SIM (25/28) HS (0/5)
83 0.430 326.3 0.681 111 0.675 105 0.900
232 0.404 644.1 0.662 HCM (10/6.6) 115 0.709 126 0.723
690 0.357 164.7 0.618 221 0.650 194 0.877
787 0.357 HS (0/10) 324.7 0.608 286 0.665 253 0.885
4 0.410 652.0 0.606 495 0.679 0.7 1.242
7 0.401 164.5 0.602 7 0.678 2.3 1.267
SIM (40/47) 13 0.390 HS (10/20) 325.1 0.595 7 0.670 2.3 1.219
64 0.390 645.2 0.586 61 0.658 4 0.983
265 0.363 138.1 0.671 70 0.636 5 1.094
HCM (15/11) HCM (94/2)
5 0.547 275.6 0.663 109 0.633 6 1.046
GBf (0/1.1)
5 0.530 556.5 0.655 226 0.607 97 1.150
11 0.539 1 363.7 0.645 283 0.615 152 0.929
SIM (60/26) 112 0.514 119.9 0.643 314 0.584 184 1.125
138 0.514 252.6 0.633 1.4 0.650 240 1.113
GB (0/1.4)
157 0.511 498.3 0.623 2.5 0.638
HCM (20/13)
184 0.497 1 254.8 0.621 7 0.624
SIM (100/10) 10 0.877 GB (0/2.5) 138.2 0.586 32 0.579
a. SIM = Silica sand-silt mixtures tested by Thevanayagam et al (2002)
b. Values in parentheses indicate FC(%)/Cu
c. Values of p’ and e that appear in bold-italic font were not used to calculate Γ1 as they could not be adequately fitted with Equation 1
d. HS = Hostun sand tested by Li (2013)
e. Third value inside the parentheses indicates the mean grain size in mm
f. GB = Glass balls tested by Li (2013)
g. HCM = Mixtures of Hokksund sand and Chengbei silt tested by Yang et al (2006)
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 61
FC = 0%. Cu varied from 1.1 to 20. This data- 1.6
set was used herein because: (i) some of the emin method not reported
soils made up of glass balls had considerably 1.4 Γ1 = 1.52emin + 0.02
low emin values which allowed a significant R2 = 0.95
extension of the lower bound of the domain 94/2
1.2
of the Γ1-emin correlation; (ii) given that the
particle shape of the glass balls is distinctly 1.0 0/2.4
different from that of Hostun sand, this
Γ1
dataset enables a straightforward assessment 70/2.2
0.8 15/11
of whether particle shape affects the Γ1-emin 20/13 5/3.4
correlation; and (iii) this dataset also allows 50/8.9 10/6.6
0.6
assessment of the validity of the Γ1-emin 30/14
correlation at different Cu values. Figure 2 0.4
presents the PSDs of the 30 soils considered.
Values of Γ1 were calculated by fitting 0.2
Equation 1 to the (p’, e) points that defined 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
the SSL of each soil (Table 1). Some of the emin
SSLs reported by Thevanayagam et al (2002) Note: Label next to each data point indicates FC(%)/Cu
cannot be modelled with Equation 1, due
to the curvature of the SSL in e-log10p’ Figure 4 Γ1 vs emin plot corresponding to soil types reported in Yang et al (2006)
which some soils exhibit at high stress
levels (e.g. Been et al 1991; Li & Wang 1998).
1.6
Consequently, some (p’, e) points with high
emin method: ASTM D4253
p’ values were excluded from the fitting
1.4 Γ1 = 0.93emin + 0.31
process. Similarly, given the experimental
R 2 = 0.90
difficulties and uncertainties involved in
1.2
performing triaxial tests at very low values
of effective stress, two (p’, e) points with
1.0 0/1.1 0/1.1
p’ = 1 kPa were also disregarded when calcu- 10/20
0/2.5
Γ1
the independence from particle shape of the the Γ1-emin correlation observed by Cho et
RESULTS AND DISCUSSION Γ1-emin correlation for narrowly graded sands al (2006) in narrowly graded sands with low
Figures 3 to 5 suggest strong (R2 ≥ 0.90) with low values of FC. FC values is approximately the same for soils
linear Γ1-emin correlations. The data point When all 30 soils are collectively plot- with significant amounts of non-plastic fines
labels further suggest that the correla- ted (Figure 6), a single linear correlation and large Cu values, such as those repre-
tions are valid regardless of FC or Cu. The emerges (R2 = 0.85). The slight decrease sented in Figures 3 to 5. The data analysed
independence of the correlation from FC in R2 (compare to Figures 3 to 5) is likely a in this study has also been useful to expand
observed in Figures 3 and 4 is at odds with consequence of combining SSLs calculated the lower bound of the domain of the Γ1-emin
previous works (e.g. Thevanayagam et al from triaxial tests conducted in different correlation reported by Cho et al (2006)
2002; Rahman & Lo 2008; Rahman et al laboratories and following slightly different (Figure 7).
2014) which have suggested that Γ1 is fun- protocols, and emin values obtained through The authors suggest that the validity
damentally correlated to FC. Furthermore, different procedures. For example, Prochaska of the Γ1-emin correlation over such a wide
Figure 3 explicitly shows that essentially the & Drnevich (2005) showed that the maxi- range of non-plastic soil types is explained by
same Γ1-emin correlation is followed regard- mum dry unit weight, which is associated to the similarity in which both Γ1 and emin are
less of whether FC is smaller or greater than emin, can show variances of the order of ± 3% affected by a soil’s fundamental properties.
TFC. Additionally, the angular Hostun sand when estimated from different compaction For example, they are both directly corre-
and the glass balls follow the same Γ1-emin techniques. A unique linear correlation lated to particle angularity (Li 2013; Biarez &
correlation despite their significantly dif- (R2 = 0.77) continues to be apparent when Hicher 1994; Cho et al 2006), inversely cor-
ferent particle shapes (Figure 5). This result the data corresponding to Cho et al (2006) related to Cu (Li 2013; Biarez & Hicher 1994;
agrees with Cho et al (2006) who reported is included (Figure 7). This indicates that Poulos et al 1985), and respond in a similar
62 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
1.6 and valid regardless of FC, Cu, and particle
Γ1 = 1.14emin + 0.20 shape. Comparison of the results presented
1.4 R 2 = 0.85 herein (Figure 6) with the Γ1 versus emin
dataset obtained by Cho et al (2006), indi-
1.2 cates that the Γ1-emin correlation originally
observed by Cho et al (2006) for narrowly
1.0 graded sands with small amounts of fines
may be applicable to all non-plastic soils
Γ1
0.2
ACKNOWLEDGEMENTS
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 The authors gratefully acknowledge Dr Nico
emin Vermeulen (Jones & Wagener) and Dr Irvin
30 soils processed by the current authors 49 soils processed by Cho et al (2006) Luker (University of the Witwatersrand) for
Note: The emin values reported by Cho et al (2006) were obtained using a variety of methods their valuable insights and discussions.
Figure 7 Γ 1 vs emin plot including all the data points from Figures 3 to 5 and the data points from
Cho et al (2006) REFERENCES
Been, K, Jefferies, M G & Hachey, J E 1991. The critical
manner to changes in FC (Lade et al 1998; target error when calculating the SSL experi- state of sands. Geotechnique, 41(3): 365–381.
Rahman & Lo 2008). It is also important to mentally. Notwithstanding these potential Biarez, J & Hicher, P-Y 1994. Elementary Mechanics
acknowledge that the use of emin as a predic- errors, the correlation is very useful to of Soil Behaviour: Saturated Remoulded Soils.
tor of Γ1 may be limited by the differences qualitatively understand how the Γ1 values of Rotterdam: Balkema.
between the remoulding and particle crush- different non-plastic soil types will compare Carrera, A, Coop, M & Lancellota, R 2011. Influence
ing mechanisms that a soil undergoes when to one another. Additionally, when studying of grading on the mechanical behaviour of Stava
sheared to steady state and when compacted a heterogeneous soil deposit that includes tailings. Geotechnique, 61(11): 935–946.
to emin. a variety of soil types, the correlation can Castro, G 1969. Liquefaction of sands. PhD thesis,
The authors are not recommending the help reduce the number of soil types whose Cambridge, MA: Harvard University.
use of the correlations in Figures 3 to 7 to SSLs have to be experimentally determined Cho, G, Dodds, J & Santamarina, J 2006. Particle shape
replace triaxial testing to determine the to fully characterise the deposit from an SSL effects on packing density, stiffness, and strength:
SSL, as doing so can result in significant standpoint (e.g. Hemer et al 2016). Natural and crushed sands. Geotechnical and
errors. For example, in Figure 7 a deviation Geoenvironmental Engineering, 132(5): 591–602.
from the best fit line of ± 0.1 is observed in Cubrinovski, M & Ishihara, K 2000. Flow potential of
Γ1, implying a potential error of 0.2. This CONCLUSIONS sandy soils with different grain compositions. Soils
value is significantly higher than the error The Γ1-emin correlation of 30 non-plastic and Foundations, 40(4): 103–119.
of ± 0.01 in void ratio, which was suggested soils has been investigated. The results indi- Hemer, J C, Mincione, N L M & Torres-Cruz, L A
by Jefferies & Been (2006) as a reasonable cate that the correlation is linear (R2 = 0.85) 2016. Determination of steady state lines for
Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017 63
non-plastic platinum tailings. In: Jacobsz, S W. (Ed.), monotonic response of sands with different fines Yang, S L, Sandven, R & Grande, L 2006. Steady-state
Proceedings, 1st Southern African Geotechnical contents using the equivalent intergranular void lines of sand–silt mixtures. Canadian Geotechnical
Conference Held at Sun City, 4–6 May 2016. ratio. Acta Geotechnica, 10(5): 587–606. Journal, 43(11), 1213–1219.
London: CRC Press/Taylor & Francis, 275–280. Muir-Wood, D & Maeda, K 2008. Changing grading of Yang, J, Wei, L M & Dai, B B 2015. State variables for
Jefferies, M G & Been, K 2006. Soil liquefaction: A soil: Effect on critical states. Acta Geotechnica, 3(1): silty sands: Global void ratio or skeleton void ratio?
critical state approach. London: Taylor & Francis. 3–14. Soils and Foundations, 55(1): 99–111.
Lade, P V, Liggio, C & Yamamuro, J A 1998. Effects of Poulos, S J, Castro, G & France, J W 1985. Liquefaction
non-plastic fines on minimum and maximum void evaluation procedure. Geotechnical Engineering,
ratios of sand. Geotechnical Testing Journal, 21(4): 111(6): 772–792. NOTATION
336–347. Prochaska, A B & Drnevich, V P 2005. One-point e = void ratio
Li, G 2013. Étude de l’influence de l’étalement vibrating hammer compaction test for granular soils. p’ = mean effective stress (σ’1 + σ’2 + σ’3)/3
granulometrique sur le comportement mecanique Proceedings, ASCE Geo-Frontiers Conference held in q = deviator stress (σ’1 – σ’ 3)
des materiaux granulaires. PhD thesis, Nantes, Austin, TX, 24–26 January. Reston, VA: ASCE. SSL = steady state line
France: Ecole Central de Nantes [In French]. Rahman, M M & Lo, S R 2008. The prediction of Γ1 = steady state void ratio that corre-
Li, G, Ovalle, C, Dano, C & Hicher, P-Y 2013. Influence equivalent granular steady state line of loose sand sponds to a mean effective stress of
of grain size distribution on critical state of granular with fines. Geomechanics and Geoengineering, 3(3): 1 kPa
materials. In: Yang, Q, Zhang, J-M, Zheng, H & Yao, 179–190. λ10 = slope of the SSL in e-log10(p’) space
Y (Eds.), Constitutive Modeling of Geomaterials, Rahman, M M, Lo, S R & Dafalias, Y F 2014. Modelling FC = fines content (< 75 μm)
Berlin, Heidelberg: Springer Verlag. the static liquefaction of sand with low-plasticity PSD = particle size distribution
Li, X S & Wang, Y 1998. Linear representation of fines. Geotechnique, 64(11): 881–894. emin = minimum void ratio
steady state-line for sand. Geotechnical and Thevanayagam, S, Shenthan, T, Mohan, S & Liang, J TFC = threshold fines content
Geoenvironmental Engineering, 124(12): 1215–1217. 2002. Undrained fragility of clean sands, silty sands, Cu = coefficient of uniformity (D60/D10)
Mohammadi, A & Qadimi, A 2015. A simple critical and sandy silts. Geotechnical and Geoenvironmental
state approach to predicting the cyclic and Engineering, 128(10): 849–859.
64 Journal of the South African Institution of Civil Engineering • Volume 59 Number 2 June 2017
Guidelines for the preparation of papers and technical notes
Authors should comply with the following guidelines when preparing papers for publication in the journal
The Journal of the South African Institution of Civil for obtaining permission to publish experimental or should be typed using the Times New Roman
Engineering is published quarterly in March, June, data and other information that may be confidential symbol set.
September and December. Articles submitted for or sensitive. Authors are also responsible for obtain- ■■ Variables in equations (x, y, z, etc, as well as lower
publication are reviewed by a panel of referees under the ing permission from copyright owners when repro- case Greek letters) should be presented in italics.
guidance of the SAICE Journal Editorial Panel. The jour- ducing material that had been published elsewhere. Numbers (digits), upper case Greek letters, sym-
nal publishes research papers covering all the disciplines Proof of such permission must be supplied. bols of metric measurement units (m for metres,
of civil engineering (structural, geotechnical, railway, s for seconds, etc) and mathematical/trigono-
coastal/marine, water, construction, environmental, SUBMISSION PROCEDURES metrical functions (such as sin, cos and tan) are
municipal, transportation) and associated topics that AND REQUIRED FORMAT not written in italics, but in upright type (Roman).
are relevant to the civil engineering profession, and that ■■ Online submission: Manuscripts must be uploaded Variables and symbols used in the body of the text
preferably have particular relevance to civil engineering as PDF files (http://journal.saice.org.za). Individual should match the format used in the equations,
in southern Africa and the African continent. file sizes may not exceed 4 MB. Should you experi- i.e. upright or italics, whichever is applicable.
When preparing articles for publication, authors ence problems uploading your paper, please contact ■■ Metric measurement abbreviations/units should
should please take note of the following and comply with the editor (verelene@saice.org.za). conform to international usage – the SI system of
the guidelines as set out: ■■ Format: Manuscripts should be prepared in MS units should be used.
Word and presented in double line spacing, single ■■ Decimal commas may be used, but decimal points
CLASSIFICATION OF ARTICLES column layout with 25 mm wide margins. Line are preferred.
CONSIDERED FOR PUBLICATION numbers must be applied to the whole document. ■■ Symbols should preferably be defined in the text,
■■ Technical papers are well-researched, in-depth, fully All pages should bear the authors’ names and be but if this is not feasible, a list of notations may be
referenced technical articles not exceeding 6 000 numbered at the bottom of the page. With the excep- provided for inclusion at the end of the paper.
words in length (excluding tables, illustrations and tion of tables and figures (see below) the document ■■ Headings: Sections and paragraphs should not be
the list of references). Related papers that deal with should be typed in Times New Roman 12 pt font. numbered. The following hierarchy of headings
‘softer sciences’ (e.g. education, social upliftment, etc) Contributions should be accompanied by an abstract should be followed:
are accepted if they are of a technical nature and of of not more than 200 words. HEADING OF MAIN SECTION
particular interest to the civil engineering profession. ■■ First page: The first page of the manuscript should Heading of subsection
The latter type of paper will be subject not only to include the title of the paper, the number of words of Heading of sub-subsection
peer-review by civil engineers, but also to review by the main text (i.e. excluding figures, tables and the ■■ References: References should follow the Harvard
non-engineering specialists in the field covered by list of references), the initials and surnames of the system. The format of text citations should be as
the paper. authors, professional status (if applicable), SAICE follows: “Jones (1999) discovered that …” or “recent
■■ Technical notes are short, fully referenced technical affiliation (Member, Fellow, Visitor, etc), telephone results (Brown & Carter 1985; Green et al 1999)
articles that do not exceed 2 000 words. A typical numbers (landline and mobile), and e-mail and indicated that …”
technical note will have limited scope often dealing postal addresses. The name of the corresponding References cited in the text should be listed in
with a single technical issue of particular importance author should be underlined. Five keywords should alphabetical order at the end of the paper. References
to civil engineering. be suggested. by the same author should be in chronological order.
■■ Review papers are considered for publication as ■■ Figures, tables, photos and illustrations: These The following are examples of a journal article, a
either technical papers or technical notes on condi- should preferably be submitted in colour, as the jour- book and a conference paper:
tion that they are the original work of the author nal is a full-colour publication. ■■ Donald, A W 1954. Pore pressure coefficients.
and will assist the reader with the understanding, ■■ Their positions should be clearly marked in the text Géotechnique, 4(4):143–147.
interpreting or applying of the subject under review. as follows: [Insert Figure 1]. ■■ Terzaghi, K 1943. Practical soil mechanics. New
A review paper must contain criteria by which the ■■ Figures, tables, photos, illustrations and equations York: Wiley.
work under review was evaluated, and contribute should be numbered consecutively and should ■■ Smith, A W, Black, C & White, A 1999. Factors
by synthesising the information and drawing new appear in the text directly after they have been affecting the strength of soils. Proceedings, ASCE
conclusions from the dissemination of the previously referred to for the first time. Conference on Shear Strength of Soils, Colorado,
published work. ■■ Illustrations must be accompanied by appropriate pp 503–532.
■■ Discussion on published articles is welcomed up to captions. Captions for tables should appear above Papers published previously in the Journal of the
six months after publication. The length of discus- the table. All other captions should appear below South African Institution of Civil Engineering should
sion contributions is limited to 1 500 words. Where the illustration (figures, graphs, photos). be cited if applicable.
appropriate, discussion contributions will be subject ■■ Only those figures and photographs essential to ■■ Footnotes, trade names, acronyms, abbreviations:
to the normal reviewing process and will be for- the understanding of the text should be included. These should be avoided. If acronyms are used, they
warded to the authors of the original article for reply. All illustrations should be referred to in the text. should be defined when they first appear in the text.
■■ Figures should be produced using com- Do not use full stops after abbreviations or acronyms.
POLICY REGARDING LANGUAGE AND puter graphics. Hand-drafted figures will not be ■■ Return of amended papers: Papers requiring
ORIGINALITY OF SUBMITTED ARTICLES accepted. Lettering on figures should be equiva- amendments will be accepted up to six months after
■■ Language: Manuscripts should preferably be lent to a Times New Roman 9 pt font or slightly the referee reports had been sent to authors, after
presented in English, as the journal is distributed larger (up to 12 pt) if desired. Lettering smaller which the paper will be withdrawn from the system.
internationally. Articles submitted in any of the other than 9 pt is not acceptable.
official South African languages should be accompa- ■■ Tables should be typed in Times New Roman 9 pt FINAL ARTICLE
nied by an expanded abstract in English. font. They should not duplicate information already ■■ Copyright: On acceptance of the paper or technical
■■ Original work: Papers and technical notes must be given in the text, nor contain material that would note, copyright must be transferred by the author/s to
original contributions. Authors must affirm that sub- be better presented graphically. Tabular matter the South African Institution of Civil Engineering on
mitted material has not been published previously, should be as simple as possible, with brief column the form that will be provided by the Institution.
is not under consideration for publication elsewhere headings and a minimum number of columns. ■■ Photos of authors: The final corrected version of
and will not be submitted elsewhere while under ■■ Mathematical expressions and presentation the paper should be accompanied by recent, high-
consideration by the SAICE Journal Editorial Panel. It of symbols: resolution head and shoulders colour photographs
is the responsibility of the authors to ensure that pub- ■■ Equations should be presented in a clear form and a profile not exceeding 100 words for each of
lication of any paper in the journal will not constitute which can easily be read by non-mathematicians. the authors.
a breach of any agreement or the transgression of any Each equation should appear on a separate line ■■ Proofs: First proofs of papers will be sent to authors
law. The corresponding author should confirm that and should be numbered consecutively. in PDF format for verification before publication.
all co-authors have read and approved the manuscript ■■ Symbols should preferably reflect those used in No major re-writes will be allowed, only essential
and accept these conditions. Authors are responsible Microsoft Word Equation Editor or Mathtype, minor corrections.