Professional Documents
Culture Documents
Journal Vol 63 2021 March
Journal Vol 63 2021 March
Journal Vol 63 2021 March
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 11 805 5947, Fax +27 11 805 5971
http://www.saice.org.za
civilinfo@saice.org.za
Contents
Editor-in-chief
Dr Peter Day
Consultant, Jones & Wagener (Pty) Ltd
Adjunct Professor of Geotechnical Engineering
University of Stellenbosch
Tel +27 11 519 0200
day@jaws.co.za
2 Reliability basis for assessment of existing building
joint EditorS-in-chief
Prof Chris Clayton structures with reference to SANS 10160
University of Southampton
c.clayton@soton.ac.uk M Holický, J V Retief, C Viljoen
Prof Gerhard Heymann
University of Pretoria
gerhard.heymann@up.ac.za 11 Measurement of probabilistic ballast
Editor
Verelene de Koker particle dynamics using Kli-Pi
Tel +27 11 805 5947, Mobile +27 83 378 3996
verelene@saice.org.za A Broekman, P J Gräbe
journal editorial Panel
Prof Y Ballim – University of the Witwatersrand
Prof W Burdzik – University of Pretoria 24 Reliability performance of bridges designed
Prof CRI Clayton – University of Southampton
Dr P Day – Jones & Wagener (Pty) Ltd (chairman)
according to TMH7 NA load model
Prof J du Plessis – University of Stellenbosch
Prof GC Fanourakis – University of Johannesburg
R Lenner, S E Basson, M Sýkora, P F Van der Spuy
Prof PJ Gräbe – University of Pretoria
Dr C Herold – Umfula Wempilo Consulting
Prof G Heymann – University of Pretoria 37 The variability in commercial laboratory aggregate
Prof A Ilemobade – University of the Witwatersrand
Prof SW Jacobsz – University of Pretoria testing for road construction in South Africa
Prof EP Kearsley – University of Pretoria
Prof C Roth – University of Pretoria I J Breytenbach, H G Fourie
Prof W Steyn – University of Pretoria
Mr M van Dijk – University of Pretoria
Prof C Venter – University of Pretoria
Prof A Visser – University of Pretoria
45 Review of compatibility between
Dr E Vorster – Zutari
Prof J Wium – University of Stellenbosch
SANS 10400 deemed-to-satisfy masonry
Prof A Zingoni – University of Cape Town wall provisions and loading code
Prof M Zuidgeest – University of Cape Town
Peer reviewing W I de Villiers, G P A G van Zijl, W P Boshoff
The Journal of the South African Institution of
Civil Engineering is a peer-reviewed journal
that is distributed internationally 61 Design and construction of laboratory-scale
Design and reproduction
Marketing Support Services, Ashlea Gardens, Pretoria
activated carbon, gravel and rice husk filter
Papers for consideration should be submitted columns for the treatment of stormwater
online at: http://journal.saice.org.za
runoff from automobile workshops
The South African Institution of Civil Engineering accepts no
responsibility for any statement made or opinion expressed C O Ataguba, I Brink
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.
Contact details:
Department of Civil Engineering
Stellenbosch University
INTRODUCTION as such, deviate from the design basis. Whilst
Private Bag X1, Matieland, Stellenbosch 7602, South Africa Acceptance of the principles of structural the same principles of structural reliability on
T: +27 82 446 5084, E: jvr@sun.ac.za reliability and risk is manifested by the which the basis of design rests would apply
introduction of the semi-probabilistic limit to the reliability of the existing structure
PROF CELESTE VILJOEN (Pr Eng, FSAICE), states approach, such as the partial factor throughout its service life, the specific condi-
Vice-Dean Teaching in the Faculty of
Engineering at Stellenbosch University, was
or load and resistance factor methods, not tions requiring assessment of the structure
until recently Head of the Division of only for South Africa, but also interna- provide a first indication that operational
Structural Engineering and Civil Engineering tionally by many countries and interna- procedures for design and assessment are
Informatics at the same university. Her
tional organisations, such as ISO TC98, bound to differ. At the minimum level, the
research considers structural risk and
reliability, including the assessment of various the Joint Committee for Structural Safety need for assessment of existing structures
structural standard provisions in reinforced concrete design, steel design and (JCSS) and the European Committee for arises from the basic requirement for validat-
loading. She has supervised many postgraduate students to completion and Standardisation CEN TC250. Key concepts ing reliability compliance after decades of
has authored more than 50 published papers. She is widely involved in
of the reliability-based approach are that service. For structures designed to present
structural standardisation, being a member of various SABS Technical
Committees. She is also convenor of the SANS 10100-3 Working Group, a structural performance is set in terms of standards, strategic upper bounds for assess-
member of the Working Group for the Revision of ISO 13824, and a member target levels of reliability, with operational ment are required from needs such as life
of the International Joint Committee on Structural Safety. design procedures that are calibrated to extension for economic reasons or sustain-
Contact details: exceed the target levels, whilst accounting ability objectives and future imperatives to
Department of Civil Engineering for the diverse variabilities and uncertainties account for climate change during the service
Stellenbosch University
Private Bag X1, Matieland, Stellenbosch 7602, South Africa
of all design or basic variables. life. Important classes of differences between
T: +27 +27 21 808 4444, E: celesteviljoen@sun.ac.za It is evidently not feasible to account, design and assessment are the options for
during the design stage, for all possible condi- optimal decision-making under the respec-
tions to which the structure will be exposed tive conditions, including the state of infor-
during a service life of multiple decades. An mation availability and gathering at the stage
extensive set of scenarios can be identified of decision-making.
Keywords: structural reliability, existing structures, assessment, where conditions to which the structure is Whilst the emphasis in structural
standard procedures, applications exposed, or even conditions of the structure engineering practice is rightfully placed on
Holický M, Retief JV, Viljoen C. Reliability basis for assessment of existing building structures with reference to SANS 10160.
2 J. S. Afr. Inst. Civ. Eng. 2021:63(1), Art. #1041, 9 pages. http://dx.doi.org/10.17159/2309-8775/2021/v63n1a1
the design of new structures, the bulk of methodologies to South Africa by Retief design of new structures and assessment of
structures is in use, requiring some degree et al (2019). existing structures.
of structural management. Based on a The International Standard ISO 13822 As a general guideline the conversion
growth rate of a few percent of the stock of (ISO 2016a), first edition published in 2001, from the wide-ranging principles and
structures over a 50-year service life, the served as point of departure for the CEN procedures to the specific assessment
ratio of existing-to-new structures is around Eurocode initiative. This standard, adopted procedures proposed for incorporation
30, with 3 to 6 structures within their last 10 as South African National Standard 13822 into SANS 10160 is intended to provide for
to 20 years of service life for each new struc- (ISO 2016a), represents the most direct standardised general practice; serving as
ture being designed. Additional complica- provision for existing standards in this a first round of assessment to identify the
tions are early obsolescence or extended country. The specific case of the assessment need for it and specifics of more refined
service life that will both have implications of existing concrete structures is extensively investigations, assessment, decision-
for the relative importance of the reliability developed in the fib Bulletin 80 (2016), with making and intervention. As a first step,
performance of existing structures. the proposed methodologies intended to be procedures are directed towards comply-
Professional duties regarding both public compatible with the Eurocode standardised ing with all the requirements that can be
and economic interests should include the approach, and with a view of extending the regarded as being equivalent to that for
reliability of the collection of existing struc- scope of a future edition of the fib Model design, such as the relevant limit states and
tures. A standardised approach towards reli- Code for structural concrete. An extensive design situations, yet fully adapted to the
ability assessment has the potential for the review of theoretical concepts for the proba- existing structure under investigation.
most effective manner of dispensing with bilistic assessment of existing structures The intention is that assessment proce-
professional responsibilities, even if only the that could serve as basis for operational dures are limited to the semi-probabilistic
requirements for the basis of assessment partial factor assessment methods is com- approach, whilst requiring an appreciation
are provided in a standardised format. The piled by the Joint Committee on Structural of risk and reliability concepts to be able
primary objective of this paper is to dem- Safety (JCSS 2001a), complementing the to adjust design parameters in accordance
onstrate that the principles of reliability and Probabilistic Model Code (JCSS 2001b); with information on the existing structure.
its conversion into operational requirements the latter arguably being biased towards Where the situation may arise that more
and procedures are sufficiently mature in the design process. These standardisation advanced methods are justified, such as the
standardised design to be converted also advances are based on an extensive body application of full reliability-based assess-
to the assessment of existing structures, of literature and research, represented for ment or risk-based and risk-optimised
in spite of the lag between development of example by the seminal papers by Allen methods, the standardised approach should
these two modes of managing and decision- (1991) and Ellingwood (1996), and more serve as initial investigation leading to
making on structural reliability. recent extensive investigations by Holický specialist assessment using Eurocode pro-
(2005), Vrouwenvelder and Scholten (2010), cedures, whilst retaining consistency with
Eurocode development of Tanner et al (2011), Steenbergen et al (2015), South African requirements.
provisions for existing structures Sýkora et al (2017), and Holický (2017). Since the condition of the existing
This paper provides a review of recent structure related to its resistance generally
developments for the introduction of the South African applications plays a primary role in its load-bearing
reliability basis for the assessment of exist- The focus for the review and appraisal of assessment, the scope of this review is
ing structures in the format of standardised this extensive body of information is provid- simplified by mainly considering concrete
requirements, and alternative approaches ed by relevance to the Head Standard for the structures, as opposed to the comprehen-
and procedures. The survey includes back- South African Loading Code SANS 10160-1 sive Eurocode approach where all structur-
ground information that served as input to (SANS 2018) Basis of structural design al materials are included (in addition to all
the development process, including historic which provides the requirements for the the structural classes of buildings, bridges
progression of provisions for the closely reliability performance of buildings and and industrial and civil engineering struc-
related respective fields of design and similar structural systems. The background tures). Even within the limited scope of this
assessment. The focus is primarily on the to SANS 10160 is provided by Retief and review, a substantial fraction of existing
extension of the Eurocode Head Standard Dunaiski (Eds) (2009), including the reli- building structures will be covered. The
EN 1990 (EN 2002) to include assessment, ability principles (Retief & Dunaiski 2009), investigation on existing concrete struc-
as envisaged in prEN 1990 (EN 2020). The the relationship to Eurocode (Retief et al tures by the International Federation for
formulation of Assessment and retrofit- 2009) and an outline of the implementation Structural Concrete ( fib Bulletin 80 2016)
ting of existing structures (CEN/TS 17440 of reliability principles in operational partial provides additional background informa-
2020) and the accompanying Background factor design in Eurocode by Holický et al tion to complement this survey. Some indi-
Documents represent a key step in the (2009). The most significant differences of cation is provided for extending assessment
development process. The CEN/TS is based the South African standard compared to procedures to other structural materials,
on a number of national standards related Eurocode are the limited scope of applica- mainly considering steel.
to the assessment of existing structures, tion, target reliability levels and resulting
compiled in a Joint Research Centre report formats for action combinations, and the
(JRC 2015). The paper provides an exten- relationship between respective standards ASSESSMENT PROCESSES
sion of the overview of recent develop- for actions and materials-based resistance. AND PROCEDURES
ments provided by Holický et al (2019) and These differences need to be reflected Initial initiatives to formalise the assess-
consideration of the applicability of these when establishing the relationship between ment of existing structures emphasised
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 3
the process of information gathering and QQ The general framework for assessment assessment to verify acceptable reliability
decision-making, whilst quantitative meth- to determine their actual conditions and optimal performance. Such conditions
odologies were presented in general terms QQ Methodologies for gathering information may include exceedance of the design ser-
of the principles of reliability and risk. Flow and updating of basic variables for the vice life, outdated standards, deteriorating
charts of iterative processes of applying representation of actions and material and obsolete materials, damage to the struc-
more detailed investigations and decision properties ture, the need for upgrading, refurbishing or
branches summarise the assessment pro- QQ Structural analysis for quantitative verifi- life extension, and enhancing sustainability.
cess for both the JCSS (2001a) report and cation of structural elements and systems In principle, possible critical situations
ISO 13822 (ISO 2016a). The requirements QQ Alternative verification methods for the structure should be identified in
for a code for the assessment and format ranging from partial factors to risk advance in the specification of the future
of its contents are provided by the JCSS assessment performance of the structure in terms of
report, whilst the ISO Standard provides a QQ Construction and operational utilisation and safety planning.
pro forma layout and specification. These interventions Various lists of circumstances leading
documents represent key contributions to QQ Consideration of the special case of the to the initiation of assessment of the struc-
the advancement of standardised assess- assessment of heritage buildings. ture are provided as an indication of the
ment. However, essential quantitative Since quantitative reliability verification scope of the document. These conditions
procedures are provided in theoretical and plays a central role in the assessment process should serve to determine the objectives,
conceptual terms, lacking standardised and represents the most significant recent planning and methodology to be followed.
procedures, such as approach and method- advancement, this paper focuses mainly on The following list is based on the CEN-TS
ologies required to bring assessment on par verification processes. Other topics are con- document, as elaborated on by Holický et
with operational design standards. sidered mainly from the perspective of their al (2019) and extended by the JCSS (2001a)
The development of CEN/TS 17440 relationship to reliability verification. list, differentiating between circumstances
(abbreviated as CEN/TS) for the assess- that may have a bearing on how the assess-
ment of existing structures represents a Scope of situations requiring ment is to be done:
significant advancement in closing the gap assessment QQ Deviations from the original project
between standardised design and assess- The scope of conditions and situations description are observed
ment. The presentation of a set of method- where the need for the reliability assessment QQ Adverse results of a recent investigation
ologies for the verification of the reliability of an existing structure arises can be related or trends from periodic investigations
of an existing structure forms an important to the reliance of modern societies and QQ Concern about the structural safety
part of this advancement. The verification economies on the built environment, where caused by evidence of damage
methodologies are consistent with the the structural or load-bearing behaviour QQ Deterioration due to time-dependent envi-
relationship between the risk-based and of the facility forms an integral function. ronmental actions (e.g. corrosion, fatigue)
reliability-based approaches for semi-prob- Trends in the continued use of structures, QQ Unusual incidents during use which
abilistic partial-factor methods specified particularly for public buildings, but often could have damaged the structure (such
by ISO 2394 (ISO 2016b). Since ISO 13822 in conjunction with economic constraints, as impact of vehicles, avalanches, fire,
(ISO 2016a) has served as starting point require the application of advanced methods earthquake)
for the CEN/TS, the general format is also that may exceed the level of sophistication QQ Suspicion of possible impairment of the
consistent with that International Standard. of design standards. The socio-economic structural safety related to structural
The implication is that the adaptation of the differences between the design of new struc- materials, construction methods or the
CEN/TS to South African standards will be tures and the assessment of existing struc- structural system
consistent with the adopted International tures, as summarised in Table 1, provide an QQ Discovery of design or construction errors
Standards SANS/ISO 2394 (ISO 2016b) and indication of the context to be applied to the QQ Planned change of the use of the
SANS/ISO 13822 (ISO 2016a). respective requirements (Holický 2005). structure
Integration of the CEN/TS into prEN More specifically, the nature and level QQ Changes in the structural system (e.g.
1990 (EN 2020) should result in the final of assessment of an existing structure are retrofitting, modifications, extensions
step of the advancement of standardised closely related to the need and motivation to the structure)
assessment. In accordance with the com- for such an assessment. Any situation where QQ A change of loads and loading conditions
prehensive nature of Eurocode, all classes the condition of the structure deviates from QQ Extension of the design working life or
and types of buildings and civil engineering the requirements and assumptions used by expiry of residual service life granted
infrastructure, bridges, industrial and geo- present structural standards may require an from earlier assessment
technical structures exposed to the inclusive
range of actions covered by the standard are Table 1 D
ifferences between the design of new structures and the assessment of existing structures
included in the scope of the CEN/TS. Aspect Design of new structures Assessment of existing structures
The range of CEN/TS topics provides for
Marginal costs of reliability Marginal costs of reliability improvements are
the basis of assessment that is equivalent to Economic
improvements are usually low usually high
the basis for design. The main topics include:
QQ The scope of conditions that may lead to Restrictions are usually less Restriction of the use and damage of
Social
significant than in existing structures economic and cultural assets are significant
the need for assessment of the structure
QQ Requirements and prerequisites in New materials are often used, Allowance for sustainability is enhanced
Sustainability
sustainability is difficult substantially by using original materials
accordance with current standards
4 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
QQ Requirement of authorities, insurance Table 2 Definition of consequences classes
companies or owners, or demanded by a The more severe consequences of
maintenance plan Consequences class
loss of human life economic
QQ Simply because of doubts about the
safety of the structure. CC4 Highest consequences Extreme Huge
The proper management of a structure
CC3 Higher consequences High Very great
throughout its service life could be added
to the list, with the process starting as early CC2 Normal consequences Medium Considerable
as the commissioning of a new building. CC1 Lower consequences Low Small
More detailed specification of the
CC0 Lowest consequences Very low Insignificant
assessment process can be formulated
for classes of circumstances, such as a
general class where the integral reliability Reliability targets and classes Table 3 I ndicative reliability levels related
needs consideration, concern about vari- According to the Eurocode basic require- to one year and ultimate limit states
ous classes of resistance impairment, and ments, the selection of reliability levels given in prEN 1990 (EN 2020)
circumstances related to loading. by member states should be based on the
CC1 CC2 CC3
following considerations:
Reliability requirements QQ The possible consequences of failure Pf = 10 –5 Pf = 10 –6 Pf = 10 –7
The central role of risk-based and reliabil- in terms of risk to life, injury, potential
β1 = 4,2 β1 = 4,7 β1 = 5,2
ity-based procedures for the design of new economic losses
structures clearly also applies to the assess- QQ The possible cause and/or mode of
ment of existing structures with the objec- attaining a limit state uncertainty comes from actions that have
tive to verify acceptable levels of structural QQ Public aversion to failure statistically independent maxima in each
performance. Reliability requirements for QQ The expense and procedures necessary year. Then the reliability (complementary to
assessment should therefore be consistent to reduce the risk of failure. failure probabilities) Φ(βn) related to the ref-
with those embedded in design procedures, Reliability levels which are related to the erence period of n years is determined from
both in terms of target levels of reliability, probability of structural failure are related to annual reliability Φ(β1), where the index β1 is
and in reliability differentiation between the consequences of failure, which are clas- related to one year, as the product of n annual
limit states and consequence classes for sified into five consequence classes (CC0 – reliabilities, thus on average as [Φ(β1)]n.
structures. Differences in reliability targets CC4) that depend on societal and economic Consequently, the reliability index βn can be
and classes in EN 1990 and SANS 10160-1 effects, as listed in Table 2. Eurocode pro- assessed from β1 using the expression indi-
are therefore an important issue in adapt- vides for CC1 – CC3, considering the two cated in the Eurocode EN 1990 (EN 2002):
ing CEN/TS procedures for inclusion in the extreme classes (CC0 and CC4) to be outside
South African standard. its scope. Indicative reference target reli- Φ(βn) = [Φ(β1)]n(1)
The comparison between design ability levels for the ultimate limit state are
and assessment summarised in Table 1 given in prEN 1990 (EN 2020), as listed in Equation 1 indicates that, for mutually
provides a clear indication of the need Table 3 for the one-year reliability index β1 independent occurrences of failure in subse-
for adapted reliability requirements for and failure probability Pf to be determined quent years, the commonly used reliability
assessment. The derivation of appropriate by member states. Procedures are based on index βn = 3,8 for n = 50 corresponds to
reliability levels of existing structures is CC2, typically with adjustments indicated β1 = 4,7. However, the statistical maxima of
actively considered in the literature – see, for other consequence classes. actions (and other time-dependent variables)
for example, Vrouwenvelder and Scholten The values in Table 3 are based on pre- in subsequent years are usually correlated.
(2010), Steenbergen et al (2015), Holický et vious studies and Annex C of prEN 1990 Consequently, the occurrences of failures
al (2015), Sýkora et al (2017), and Holický (EN 2020); seismic situations are excluded. in subsequent years are inter-dependent.
(2017). More elaborate reviews of the lit- It is not specified whether the values in Then the relationship (Equation 1) should be
erature are captured by background inves- Table 3 are applicable to accidental and fire generalised to take the correlation of failure
tigations, such as JCSS (2001a), Lüchinger design situations, and recommendations for events in subsequent years into account. A
et al (2015), fib Bulletin 80 (2016). The Serviceability Limit States are absent. procedure for adjustment of reliability to
relevant information from the literature is The final draft of prEN 1990 (EN 2020) different reference periods for correlated
effectively embedded in the CEN/TS docu- does not include possible transformation events is provided by Holický et al (2018).
ment but could serve as background input of the reliability level related to one year
towards the extension of SANS 10160-1 to levels related to other reference periods, Observations on target reliability
to account explicitly for the assessment of even though this is needed to specify reli- The following conclusions are related to
existing structures. ability elements for design and assessment the target reliability level:
In addition to the general motivation of of common structures. QQ The target reliability levels recommend-
refining the rational basis for determining ed in various national and international
reliability levels, the main issues related to Adjustment of reliability level documents are inconsistent in terms of
assessment are safety costs, varying uncer- Transformation of target reliability index the values and reference periods.
tainties, time-related reference periods and β related to different reference periods QQ In the latest draft prEN 1990 (EN 2020)
remaining service life. of n years is well known when the main of the revision of Eurocode EN 1990 the
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 5
target reliability level is indicated only for if they can be justified on the basis of eco- Here Pf,t denotes the target probability of
one-year and 50-year reference periods. nomic, social and sustainable consideration failure that is to be specified taking into
QQ Transformation formulae for adjustment (see Annex F to ISO 13822 (ISO 2016a)). account economic and societal conse-
of the reliability level to different refer- An adequate value of the reliability quences of failure, and the costs of improv-
ence periods, taking mutual dependence index β should generally be determined ing structural reliability.
of failure probabilities in subsequent considering the appropriate reference peri-
years into account, are missing. od. For serviceability and fatigue the refer- Risk assessment approach
QQ The proposed transformation formula for ence period equals the remaining working The reliability is examined by acceptable
reliability index βnk depends on the refer- life, while for the ultimate limit states the risk expressed in a symbolic form as:
ence period n and independence interval k. reference period is in principle the same as
QQ Reliability index βnk decreases with the the design working life specified for new Risk = Pf C = P{g(Xi) < 0}C < Risk t(6)
reference period n and increases with structures (50 years for buildings).
the independence interval k. Reliability of a structure is given by the Here C generally represents any type of
QQ When determining the target reliability condition g(Xi) > 0, where g(Xi) denotes economic and societal consequences,
index, the assumption of annual inde- the limit state function and Xi represents and Risk t the relevant target risk level.
pendence of failures (k = 1) may be unsafe. the basic variables. Commonly the limit Appropriate target risk level Risk t is to be
state function can be considered in a specified individually, whilst accounting for
simplified form: the specific condition of an assessed struc-
RELIABILITY VERIFICATION ture. This may be a complicated task, par-
Reliability verification of an existing struc- g(Xi) = R − E > 0 (2) ticularly in the case of heritage buildings,
ture shall be made using valid codes of where historical and artistic aspects are
practice, as a rule based on the limit state Here R denotes resistance while E denotes usually also involved. A general flowchart
concept. Attention should be paid to both load effect. The reliability condition of the risk assessment procedure is shown
the ultimate and serviceability limit states. (Equation 2) can be verified by various in the CEN/TS Annex A. Principles of risk
Verification may be carried out using partial methods. assessment may be used for cost optimisa-
safety factor or structural reliability methods The following procedures are included tion procedures.
with consideration of the structural system in the CEN/TS.
and ductility of components. The reliability
assessment shall take the remaining working Partial factor method APPLICATION EXAMPLE
life of the structure into account, the refer- When using the partial factor method, An existing building constructed in 1970
ence period, and anticipated changes in the the reliability requirement (Equation 2) is to be renovated and a new assessment is
environment of the structure. (g(Xi) > 0) is substituted by the condition: required. The following example is limited
The conclusion from the assessment to a simply supported reinforced concrete
shall withstand a plausibility check. g(xdi) = g(xd1, xd2, xd3, ...) > 0, panel (a prefabricated hollow core floor
Discrepancies between the results of xdi = xki or xdi = xki γ(β) or panel) of a span L = 6 m. The panel should
structural analysis (e.g. insufficient safety) xdi = xki ⁄ γ(β)(3) be exposed to additional permanent load
and the real structural condition (e.g. no due to a newly designed floor surface. The
sign of distress or failure, satisfactory Here xdi denotes the design values of basic following requirements are specified: the
structural performance) must be explained. variables Xi determined using their charac- remaining working life is 50 years, and the
It should be kept in mind that many teristic values xki and relevant partial factors target reliability index related to the refer-
engineering models are conservative and γ(β). The partial factors γ(β) may be adjusted ence period of 50 years is βt = 3,8. The tar-
cannot always be used directly to explain taking the specified reliability index β and get reliability can be adjusted to the actual
an actual situation. actual characteristics of the basic variable economic and societal conditions. Table 4
The target reliability level used for into account (see fib Bulletin 80 (2016)). indicates expected characteristic values of
verification can be taken as the level of actions specified in accordance with valid
reliability implied by the acceptance criteria Assessment value method standards, the load effect (the mid-span
provided in valid design codes. The target The condition (Equation 2) is modified by point bending moment) and the corre-
reliability level shall be stated together with the requirement: sponding current capacity of the panel.
clearly defined limit state functions and The load effect E and the resistance R of
specific models of the basic variables. The g(xdi) = g(xd1, xd2, xd3, …) > 0, the panel are expressed as bending moments
target reliability level can also be established ΦXi(xdi) = Φ(–αiβ)(4) at the mid-span point of the panel. The
taking the required performance level for characteristic values of the actions Ek and
the structure into account, together with Here αi denotes the FORM sensitivity fac- resistance Rk alone cannot be used to assess
the reference period and possible failure tors and Φ the normal distribution function. structural reliability; however, when Ek < Rk
consequences. In accordance with ISO this may indicate some safety margin.
2394 (ISO 2016b), the performance require- Probabilistic method The structural reliability of the consid-
ments applied in the assessment of existing The requirement g(Xi) > 0 (Equation 2) is ered concrete panel can be verified consid-
structures are the same as those used in the examined by the failure probability: ering the limit state function (Equation 2)
design of new structures. Lower reliability in a common form used for bending capac-
targets for existing structures may be used P f = P{g(Xi)<0} < P f,t(5) ity of reinforced concrete cross section:
6 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
Table 4 Characteristic values of actions and resistance possible structural or operational interven-
tions are considered. Then the procedure
Permanent load G Imposed load Q Panel resistance R
for the cost optimisation indicated below
Characteristic values 6.26 kN/m2 1.50 kN/m2 43.2 kN/m2 may be useful. In such a case the total cost
Characteristic total actions 7.76 kN/m2 – Ctot(x) including possible structural or
operational interventions may be consid-
Characteristic Ek and Rk 34.9 kNm 43.2 kNm
ered in the fundamental form (indicated in
COMREL):
Table 5 Results of the applied assessment methods
Ctot(x) = C0 + C1x + Cf P f (x)(8)
Load effect Resistance Reliability
Assessment method Results
(kNm) (kNm) index β
Here x = R/R0 denotes the intervention
48.3 parameter (material consumption) to
Partial factors of Eurocodes 37.6 – Negative
45.1 increase the present resistance R0 to a
desired level R, C0 denotes necessary costs
Adjusted partial factors 41.9 39.3 Assumed 3.8 Negative
of intervention independent of the param-
Assessment values 40.3 41.0 Assumed 3.8 Positive eter x, C1 denotes marginal costs per unit of
the parameter x, Cf denotes costs due to fail-
Probabilistic method – – Assessed 4.2 Positive
ure and malfunctioning, and Pf (x) denotes
the probability of failure and malfunction-
g(Xi) = R
– E = As f y(d − Asf y /(2bfc)) remaining working life of 50 years the ing. Figure 1 (obtained using the software
− (g+q) L2/8(7) considered panel seems to be sufficiently product COMREL) indicates variation of
reliable. the total costs Ctot(x) with the parameter x,
Here As denotes reinforcement, f y strength Cost optimisation procedures may be for selected relative costs C0 = 0, C1 = 10,
of reinforcement, fc strength of concrete, particularly effective when the reliability of Cf = 1 000 000 and structural failure Pf (x)
d the effective depth and b the width of the a minor structure is acceptable (as in the of the panel having the resistance x R0.
panel cross section. case of the assessed floor panel); however, Figure 1 indicates that the optimum
The application of the partial factors for to increase the current reliability level increase of the resistance R seems to be
design employing EN 1990 Equations (6.10) for the required remaining working life, 10 percent (R/R0 ≈ 1,10). However, the
and (6.10a) respectively serves as reference for
the assessment. A set of partial factors, which
are adjusted for assessment, is applied to all 30
the design variables for verification. Next,
the probability models for the basic variables
are adjusted and applied in accordance with
the assessment value approach. Finally, a 25 Ctot = C1 R/R0 + Cf Pf
reliability assessment compares the reliability
achieved to the target. The results of the panel
assessment, including recommended deci-
sions based on the applied verification meth- 20
ods, are indicated in Table 5. The consecutive
steps demonstrate the effects respectively
of reduced standardised assessment partial
Cost
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 7
convincing result should be based on appro- Table 6 Basic reliability elements for the ultimate limit state
priate failure cost Cf and marginal costs C1. Reliability class
This simple illustrative example of a Reliability element
floor panel may provide guidance to the RC2 RC3
8 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
Table 8 Partial factors for design adjusted to assessment level for the assessment of existing structures
should explore the implications for all
γG γQ γs γc
Application level these and other issues on the conversion
RC2 RC3 RC2 RC3 RC2 RC3 RC2 RC3 from design to assessment. The extensive
Joint Research Centre (JRC 2015) survey
Design 1.2 1.25 1.6 1.8 1.15 1.2 1.5 1.6
on international practice should provide
Assessment I 1.15 1.2 1.4 1.6 1.1 1.15 1.4 1.5 useful guidance on converting the CEN/TS
procedures into operational rules for South
Assessment II 1.15 1.15 1.25 1.4 1.05 1.1 1.35 1.4
African conditions.
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 9
SANS 10160 to normal design practice and Holický, M 2005. Chapter 9: Assessment of existing Kemp, A R, Milford, R V & Laurie, J A P 1987.
allowing for the use of specialist Eurocode structures. In Ellingwood, B R & Kanda, J (Eds). Proposals for a comprehensive limit states
procedures where needed, rather than to Structural Safety and its Quality Assurance. Reston, formulation for South African structural codes.
adopt Eurocode standards comprehensively VA: American Society of Civil Engineers (ASCE). Civil Engineer in South Africa, 29(9): 351–360.
(Retief & Dunaiski 2009). Holický, M, Retief, J V & Dunaiski, P E 2009. The Lüchinger, P, Fischer, J, Chrysostomou, C et al 2015.
Two complementary perspectives are reliability basis of Eurocode. In Retief, J V & New European technical rules for the assessment
therefore presented in the paper, namely Dunaiski P E (Eds). Background to SANS 10160. and retrofitting of existing structures. JRC Science
alternative methodologies for progressively Stellenbosch: SunMedia. and Policy Report. Luxembourg: Publications Office
advanced assessment analysis and verifica- Holický, M, Retief, J V & Wium, J A 2010. Partial of the European Union.
tion for Eurocode, and a more operational factors for selected reinforced concrete members: Milford, R V 1985. Calculation of a load index for
semi-probabilistic approach that could Background to the revision of SANS 10100-1. ultimate limit state codes. Report 85/6. Pretoria:
be applied in standardised procedures. Journal of the South African Institution of Civil CSIR National Building Research Institute.
The mutual advantages would be the Engineering, 52(1): 36–44. Milford, R V 1986. A guide for the calibrating SABS
presentation of procedures that could be Holický, M, Diamantidis, D & Sýkora, M 2015. materials codes. Report 86/16. Pretoria: CSIR
used extensively in normal practice, based Determination of target safety for structures. National Building Research Institute.
on the advanced procedures that could be Proceedings, 12th International Conference on Retief, J V & Dunaiski, P E (Eds) 2009. Background to
used by specialists where the stakes are Applications of Statistics and Probability in Civil SANS 10160. Stellenbosch: SunMedia.
high for assessment verification. Engineering (ICASP12), 12–15 July 2015, Vancouver, Retief, J V, Dunaiski, P E 2009. The Limit States Basis of
Canada. Design for SANS 10160-1. Stellenbosch: SunMedia.
Holický, M 2017. Operational approach to assessment Retief, J V, Dunaiski, P E & Holický, M 2009. Review
REFEFRENCES of existing structures. Proceedings, 12th of Eurocode from the Perspective of the Revision of
Allen, D E 1991. Limit states criteria for structural International Conference on Structural Safety & SANS 10160. Stellenbosch: SunMedia.
evaluation of existing buildings. Canadian Journal Reliability (ICOSSAR), 6–10 August 2017, Vienna, Retief, J V, Viljoen, C & Holický, M 2019. Standardized
of Civil Engineering, 18(6): 995–1004. Austria. basis for assessment of existing structures. In
Botha, J, Retief, JV, Viljoen, C 2018. Reliability Holický, M, Diamantidis, D & Sýkora, M 2018. Zingoni, A (Ed) Advances in Engineering Materials,
assessment of the South African wind load Reliability levels related to different reference Structures and Systems: Innovations, Mechanics
formulation. Journal of the South African Institution periods and consequence classes. Beton- und and Applications. Abingdon, UK: Taylor & Francis,
of Civil Engineering, 60(3): 30–40. Stahlbetonbau, 113(1): 22–26. pp 2120–2125.
CEN/TS 17440 2020. Technical Specification. Holický, M, Viljoen, C & Retief, J V 2019. Assessment SANS (South African National Standard) 2018. SANS
Assessment and retrofitting of existing structures. of existing structures. In Zingoni, A (Ed). Advances 10160-1:2010/2018. Basis of Structural Design and
Brussels: European Committee for Standardisation. in Engineering Materials, Structures and Systems: Actions for Buildings and Industrial Structures.
Diamantidis, D & Sýkora, M 2019. Reliability Innovations, Mechanics and Applications. Part 1: Basis of Structural Design. Pretoria: SABS
differentiation and uniform risk in standards: a Abingdon, UK: Taylor & Francis, pp 2109–2114. Standards Division.
critical review and a practical appraisal. Paper ISO (International Organisation for Standardisation/ SANS (South African National Standard) 2000. SANS
presented at the Symposium on Future Trends in South African National Standard). 2016a. ISO 10100-1: 2000. The Structural Use of Concrete,
Civil Engineering, 17 October 2019, Zagreb, Croatia. 13822:2010/SANS 13822:2016. Bases for Design Part 1: Design. Pretoria: SABS Standards Division.
EN (European Standard) 2002. EN 1990:2002. of Structures. Assessment of Existing Structures. Steenbergen, R, Sýkora, M, Diamantidis, D, Holický, M
Eurocode – Basis of Structural Design. Brussels: Geneva, Switzerland: ISO. & Vrouwenvelder, T 2015. Economic and human
European Committee for Standardisation: ISO (International Organization for Standardisation/ safety reliability levels for existing structures.
EN (European Standard) 2020. prEN 1990:2002 South African National Standard). 2016b. ISO Structural Concrete, 16(3): 323–332.
(supersedes 1990:2002). Eurocode – Basis of 2394:2015/SANS 2394:2016. General Principles on Sýkora, M, Diamantidis, D, Holický, M & Jung, K 2017.
Structural and Geotechnical Design. Brussels: Reliability for Structures. Geneva, Switzerland: ISO. Target reliability for existing structures considering
European Committee for Standardisation. JCSS (Joint Committee on Structural Safety) 2001a. economic and societal aspects. Structure and
Ellingwood, B R 1996. Reliability-based condition Diamantidis D (Ed.) Probabilistic assessment of Infrastructure Engineering, 13(1): 181–194.
assessment and LRFD for existing structures. existing structures. Joint Committee on Structural Tanner, P, Lara, C & Prieto, M 2011. Semi-probabilistic
Structural Safety, 18(2–3): 67–80. Safety. SARL, France: RILEM Publications. models for the assessment of existing concrete
fib Bulletin 80 2016. Partial factor methods for existing JCSS (Joint Committee on Structural Safety) 2001b. structures. Proceedings, 11th International
concrete structures. Lausanne, Switzerland: Joint Committee on Structural Safety. Probabilistic Conference on Applications of Statistics and
International Federation for Structural Concrete Model Code. http://www.jcss.byg.dtk.dk. Probability in Civil Engineering, 1–4 August 2011,
(fib). JRC (Joint Research Centre) 2015. Scientific and Policy Zurich, Switzerland.
Fischer, K, Viljoen, C, Köhler, J & Faber, M H 2019. Report, No 1247. New European Technical Rules Vrouwenvelder, T & Scholten, N 2010. Assessment
Optimal and acceptable reliabilities for structural for the Assessment and Retrofitting of Existing criteria for existing structures. Structural
design. Structural Safety, 76: 149–161. Structures. Luxembourg: JRC. Engineering International, 20(1): 62–65.
10 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
TECHNICAL PAPER
Measurement of
probabilistic ballast particle
Journal of the South African
Institution of Civil Engineering
Broekman A, Gräbe PJ. Measurement of probabilistic ballast particle dynamics using Kli-Pi.
J. S. Afr. Inst. Civ. Eng. 2021:63(1), Art. #966,13 pages. http://dx.doi.org/10.17159/2309-8775/2021/v63n1a2 11
LITERATURE (2017) investigated the potential replace- Strack (1979) explored the effect of particle
The background literature discusses the rel- ment of t raditional LVDTs (linear variable rearrangement on the distribution of princi-
evance of ballast in the track structure and differential transformers) altogether using pal stresses using numerical models. Typical
emerging sensor technologies that enable accelerometers (Bowness et al 2006; Graizer ballast maintenance activities further disturb
quantitative measurements of the ballast 2010; Kempe 2011). The peak-to-peak the particle arrangement and interlocking
particles. Different modelling approaches for displacement accuracy was reported as 20% where the lateral and longitudinal resistance
describing inter-particle contact and behav- over a wide range of train velocities. are reduced by up to 60% and 40% respec-
iour – both deterministic and probabilistic The introduction of inertial navigation tively (Tutumluer et al 2006). Furthermore,
perspectives – are reviewed, and conclude systems and units (IMU), together with the the long-term rate of permanent deformation
with recent examples of “smart ballast” to advancement of micro-electro-mechanical (Lekarp et al 2000) and cumulative perma-
directly quantify the dynamic ballast prop- system (MEMS) manufacturing techniques, nent deformation (Gräbe & Clayton 2009) is
erties subject to quasi-static loading. have led to the development of high- highly dependent on the development of PSR.
performance sensors in miniaturised form The most unfavourable stress combination
Track structure factors. Woodman (2007) defines inertial occurs approximately at the subballast inter-
Most of track maintenance pertains to the navigation as: “… a self-contained navigation face depending on the Young’s modulus of
correction of deteriorating track geometry. technique in which measurements provided the subgrade and ballast. Gräbe and Clayton
The action of repeated cyclic loading by accelerometers and gyroscopes are used to (2009) describe the PSR as a combination of
degrades the ballast through inter-particle track the position and orientation of an object three distinct mechanisms:
attrition, fouling and weathering of the par- relative to a known starting point, orientation 1. Longitudinal PSR is positioned between
ticles. Although some of the settlement can and velocity.” A rigid IMU system provides two adjacent sleepers for shallow depths
be attributed to the underlying earthwork the ability to describe the body frame (of and single-axle combinations.
(subgrade and subballast), most of the track the sensor), relative to the global reference 2. Transverse PSR is present at the discon-
settlement is caused by the rearrangement frame, through the measurement of three tinuous sleeper ends.
and movement of the unbound ballast parti- orthogonal accelerometers and three ortho 3. The depth of influence of PSR correlates
cles. Low-frequency components associated gonal gyroscopes. The linear accelerations with the axle count per bogie.
with the moving load are characterised by and angular velocities are measured by the
geometry-dependent features of the train, accelerometers and gyroscopes respectively. Micromechanics of granular
such as inter-axle, inter-bogie and inter- MEMS manufacturing technologies refer to materials
wagon dimensions. Intermediate frequencies the construction of miniaturised IMUs using The deterministic or continuum approach
are characteristic features of the granular traditional semiconductor and integrated cir- to analysing soils and granular media is
layers, with higher frequencies associated cuit (IC) fabrication techniques (Nathanson & generally employed to model the response
with track components and rail defects. For Wickstrom 1965). The typical construction to loading using theories of elasticity and
the investigation of ballast, the frequencies and operation of such a device are discussed plasticity. For the investigation of discrete
of interest are below 250 Hz (Kouroussis in detail by Lemkin and Boser (1999). particle media, however, analytical solutions
et al 2015). and new instrumentation technologies are
Deterministic mechanics required if the micromechanics of mesoscale
Emerging sensor technologies of granular materials behaviour are to be better understood. The
Different instrumentation technologies have Traditionally, investigations of granular two most popular numerical methods are
been developed, adapted and improved over materials, particularly numerical models, finite element modelling (FEM) and discrete
time to measure a wide assortment of track laboratory and field investigations, have element modelling (DEM). The complexity
components and associated phenomena. been from a deterministic point of view. of obtaining realistic results stem from the
Although the functional importance of The combined probabilistic response of the significant number of particles present and
ballast is well understood and extensively particle matrix manifests as macroscopic or the small time-step required for convergence.
documented, few studies have attempted to deterministic effects and properties such as To complement the theoretical investigations,
directly instrument ballast particles. Some stress, permanent strain and confinement. the use of miniaturised, MEMS-based sensor
notable examples are that of Bennett et al The transfer of load from the sleeper technologies has recently been explored. Few
(1999) with the development of a cylindri- to the ballast produces the same effects as comprehensive, constitutive relationships
cally shaped device to measure strain of with triaxial compression (Li et al 2015). exist that accurately describe the complexi-
asphalt surfacings, and Nejikovsky and The middle of the ballast layer experiences ties involved with a discrete medium, thus
Keller (2000) with the on-board measure- a reduction in vertical stress compared to the wide implementation of FEM and DEM.
ment of lateral acceleration. Straser and the ballast directly beneath the sleeper, but Cundall and Strack (1979) pioneered the
Kiremidjian (1998) developed wireless with higher levels of confinement. The lowest theoretical establishment of DEM modelling.
monitoring systems for civil structures. Any layer of ballast, near the subballast interface, Discrete modelling has the advantage of
attempts to instrument a ballast particle develops tensile strain incrementally as the simulating the relatively large displacements,
directly are met with difficulty, owing to number of load cycles increases. These par- rotations and contacts occurring within the
the irregular morphology and freedom of ticles tend to spread and rearrange, resulting matrix. A comprehensive, two-dimensional
movement about multiple degrees of free- in the development of voids. The substantial DEM analysis by Zhang et al (2016) inves-
dom (DoF). Only the particles located at the influence of particle breakage on the tigated the dynamic ballast behaviour for
surface, which is not of interest, are suited permanent strain was highlighted by Lobo- high-speed railway applications. The largest
for visual study. Recently, Lamas-Lopez et al Guerrero and Vallejo (2006). Cundall and stresses and accelerations were concentrated
12 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
within the first 200 mm of ballast where of instrumenting field ballast with a uniaxial successfully developed (Figure 1) and cali-
peak frequencies were observed in the region MEMS accelerometer, in combination with a brated for this research project (Broekman
below 100 Hz and in the 150–300 Hz range. theoretical model. Good agreement between & Gräbe 2018). The calibration procedure
A dimensionless scalar value of β relates the the results was found. The resonant fre- provided comparable instrumentation
ratio of the principal stress of a particle to quency of the ballast accelerations was found performance compared to that of Milne et
the average of the three principal stresses. to be between 80 Hz and 100 Hz. Aikawa al (2016b). The name Kli-Pi is derived from
When the displacements of these particles (2009) supplemented this method with an the Afrikaans word for small rock to convey
are graphed for a moving load, a circular additional, secondary accelerometer to derive the miniaturised rock or ballast-shaped ana-
pattern of movement is produced. The angular velocities. Rotations exhibited identi- logue. The key performance characteristics
spectral frequency densities with the greatest cal frequencies to that of the accelerations of Kli-Pi can be summarised as follows:
amplitude were observed to be a function with more rapid attenuation. Only vertical QQ High-performance, Linux-based
of the inter-axle and inter-bogie spacing of accelerations were measured. Li et al (2015) microprocessor and MEMS IMU
the train. Other studies improved upon this reported a similar range of resonance fre- (H3LIS221DL Motion Sensor).
research with the implementation of more quencies in-between 50 Hz and 210 Hz. Milne QQ Sampling rate of 3 000 measurements
representative non-convex particle geometry et al (2016a) utilised several wireless, minia- per second, divided evenly among
(Wilke et al 2016) generated from photo- turised smart ballast particles to investigate the 6 DoF (tri-axis accelerometer and
grammetric laser scanning (Indraratna et al aerodynamic effects on the granular media tri-gyroscope). The data for each axis of
2011; Mvelase et al 2012) of the particles. for train velocities exceeding 400 km/h. each sensor is thus recorded at 500 Hz.
Some authors have suggested that the The most recent and noteworthy example QQ Programmable full scale of ±16 G and
driving force behind the rearrangement, of smart ballast is the development of ±2 000 degrees per second (DPS) for the
orientation and dynamic response of “SmartRock” by Liu et al (2016a). The wire- accelerometer and gyroscope respectively.
granular media is the manifestation of the less internal unit containing the electronics QQ High-strength, non-convex, 3D printed
maximisation of the system’s entropy (Cowin is surrounded by a 3D printed exoskeleton, PLA (Polylactic Acid) exoskeleton for
& Satake 1979). This statistical foundation measuring only 60 × 60 × 60 mm in size. ample environmental protection.
suggests a possible amalgamation between The unit transmits the IMU measurements QQ Sufficient battery capacity to provide a
the measured deterministic response and wirelessly using a Bluetooth interface. The two-month standby time or ten hours of
proven probabilistic characteristics exhibited. maximum sampling frequency is reported continuous data acquisition.
Rothenburg and Kruyt (2009) concluded as 64 Hz which can resolve the peak-to-peak QQ High-bandwidth, in-situ wireless com-
that the best descriptor of the inter-particle translational and rotational accelerations. munication using a Wi-Fi connection
contact forces, angles and displacements Short-term testing of 500 loading cycles illus- for the transfer of data (post-recording)
is with the use of statistics, not simplified trates the complexity in the rearrangement of over a local area network (LAN).
laws of mechanics. The observed particle the SmartRock in the surrounding granular The 500 Hz sampling rate, or 2 ms between
displacements, for identical contact angles matrix. A decrease in lateral stiffness was successive readings, provides the abil-
(the corresponding conditional probability), observed at the edge of the sleeper resulting ity to accurately integrate the data using
are merely the outcome or realisation of from the decrease in resistance to movement. numerical methods. A second-order
some independent random variable. The Additional research was carried out to move Lagrange interpolation polynomial was
shape of the probability distribution func- the SmartRock to different positions beneath used to approximate the discrete data
tion (PDF) was derived using techniques to the sleeper (Liu et al 2016b) together with the points and is referred to as super-sampling.
maximise the available entropy (Kruyt & influence of geogrids installed within the bal- Following the super-sampling operation,
Rothenburg 2002). Depending on the average last (Liu et al 2016c). The SmartRock proved Simpson’s method is employed to numeri-
coordination number (CN) of the particles, vital to measure the mesoscale effects associ- cally integrate the array of data. The linear
the PDF is either Gaussian-shaped for dense ated with different geometric and material acceleration and rotational velocity were
systems (5.5 < CN < 6) or exponentially variables. The main conclusions reached integrated using this method to obtain the
shaped (4 < CN < 5). Makse et al (2000) were the increased amplitude of rotation by displacements and rotations for the three
observed similar results to that of Kruyt and positioning the SmartRock at the edge of orthogonal axes of measurement. Due to the
Rothenburg (2002). Kitamura (1980) sum- the sleeper, the significant reduction in the influence of electromagnetic interference,
marises his findings as follows: “Furthermore, movement with an increase in depth, and the only the gravitational vector was employed
it is shown that the discontinuous motions complex interaction between the measured in the alignment of the body vector’s z-axis
largely contribute to the deformation process translation and rotation. with that of the global frame. The installa-
of particulate material, and the method to tion method ensured that the lateral axes of
quantitatively estimate the discontinuous the body frame were aligned with that of the
motions is proposed based on the probabilistic KLI-PI DEVELOPMENT global frame. For both the installation and
considerations for the motions of particles.” A detailed design methodology outlining experimentation procedure, the following
the desired functionality and performance coordinate system was used:
Smart ballast characteristics of an improved smart ballast QQ Longitudinal | parallel to the direction
The fusion of recent research applications prototype, to meet the required objectives of of the rails | IMU x-axis or roll (φ)
involving smart sensor technologies and the study, was established. The methodology QQ Lateral | parallel to the long axis of the
ballast particles is collectively referred to as incorporated the design, construction and sleepers | IMU y-axis or pitch (θ)
“smart ballast” instrumentation. Zhai et al calibration of the instrument. The smart QQ Vertical | parallel to the gravitational
(2004) provides one of the earliest examples ballast prototype, named “Kli-Pi”, was vector | IMU z-axis or yaw (ψ)
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 13
Figure 1 Kli-Pi 3D printed shell / exoskeleton and the internal IMU box containing all the electronics
EXPERIMENTAL PROCEDURE coincided with an increase in the gradient surrounding the instruments. During sub-
A full-scale field trial was conducted to and a decrease in train velocity as the train sequent retrieval of the instruments, it was
investigate the representative dynamic passed the static reference measurement noted that the orientation remained unal-
response of the ballast. The aim of the point. The granular layers were saturated tered. The Wi-Fi access point (denoted AP)
experiment was to replicate the typical following short bouts of rain during the used for communicating with all the Kli-Pis,
characteristics and operating behaviour of preceding week. Additionally, during was positioned directly above the vertically
field conditions for heavy-haul conditions. retrieval of the instruments, the water table aligned column of instruments to ensure
Figure 2 illustrates the instrumentation was positioned approximately 100 mm a reliable connection. The user interface
configuration together with the depth of above the subballast interface. was situated a safe distance away from the
the instrumentation. The experiment was The section was instrumented with a track using an Ethernet interface. The data
carried out during December 2017 near the total of four Kli-Pis. For ease of installation presented was obtained from the passage
Komati Power Station / Broodsnyersplaas and retrieval, the Kli-Pis were installed of six class 10E locomotives (21 ton/axle)
(26°04’32.9”S 29°30’17.8”E). This primary directly beneath one of the sleepers in the and 100 loaded wagons (estimated 20 ton/
rail artery is used for exporting coal from shoulder ballast section (Figure 3). This axle), representing a period of approximately
the surrounding mines in the Mpumalanga position also minimises the disturbance 100 seconds.
Province of South Africa. Gräbe and to the ballast matrix. Furthermore, this The software language of choice was
Clayton (2009) summarised the track position is synonymous with the expected Python for its versatile set of available
specifications and dimensions. The track location where significant rotation (Liu et al libraries and wide implementation in the
consists of equidistantly spaced (650 mm) 2016b) and PSR (Gräbe and Clayton 2009) research community. Python was used for
concrete (PY) sleepers using a standard would be measured in addition to the influ- both the calibration of the IMU and the
Cape gauge of 1 067 mm. The ballast ence of overlapping stress frusta from the analysis of the large data sets to extract
depth is 300 mm with the subballast and dual-wheel axle configuration. The ballast meaningful data as efficiently as possible.
subgrade material extending to a depth stones were carefully placed back in position The analysis workflow consisted of filtering
of 900 mm. This rail corridor started by hand after the installation process to methods (Butterworth filtering), numerical
service in 1976. The direction of travel ensure adequate interlocking of the particles approximation and integration methods,
Z
AP Y
K4
X
Z
Y
250
PY sleeper
50
K3
Ballast X
150
K2
150
Subballast
interface K1
14 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
frequency domain transformations and peak
identification. Similar to Lamas-Lopez et al
(2017) and Milne et al (2016b) a Butterworth
filter of order 3 with a bandpass of 4-200 Hz
0.4
was selected for the acceleration data. To
remove the bias present in the rotational
velocity data obtained from the gyroscope,
0.2
Displacement (mm)
an average measurement over a 3-second
period (with no trains passing) is subtracted
(per axis) from the entire dataset. Any other
calculations derived from these datasets, 0
0.03
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 15
insights into the geometry of the train,
sleeper-ballast resonance and horizontal
20.0 1.20 × 10 –8 forces. For the vertical deflection (z-axis) of
Kli-Pi 3, positioned beneath the sleeper, the
17.5
16 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
displacements. The negative gradient of the
spectra with the progression of time for all
the FFT-based figures is a direct result of the 20.0 5.00 × 10 –4
decreasing train velocity over time, owing to
Frequency (Hz)
12.5
energy methods 3.00 × 10 –4
Considering that the motion of the Kli-Pi 10.0
was quantified about all three spatial
7.5 2.00 × 10 –4
dimensions, application of kinetic and
potential energy principles is a cogent 5.0
analysis method. The assumption can be 1.00 × 10 –4
postulated whereby the position of the 2.5
Kli-Pi, over a relatively long period of time,
0
remains unchanged. Even though the 20 40 60 80 100
filtered data represents a relative position Time (sec)
(zero-centred data), the assumption is that
a constant position before and after the Figure 8 Frequency domain rotation: x-axis of Kli-Pi 3
passage of the train represents an absolute
datum. The total work associated with the
Kli-Pi consists of both kinetic (UK) and 1.00
potential energy (UP) components. The
angular moment of the Kli-Pi is ignored 0.75
owing to the difficulty associated with the
calculation of the moment of inertia and the 0.50
small rotation angles. The work generated
by the potential or gravitational field is 0.25
0
energy of a rigid body (Stewart 2012). Also,
following from Newton’s second law of
–0.25
motion, the mechanical work exerted on a
rigid body can be shown to equal the change
–0.50
in the kinetic energy. If the total energy is
to be conserved for a Kli-Pi moving along
–0.75
a path C, given by a function r (t), through
a conservative force field F , Equation 1 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3
–1.00
must hold: 64.0 64.5 65.0 65.5 66.0 66.5 67.0 67.5 68.0
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 17
the sleeper (Kli-Pi 4). A large component
2.00 of the displacement is associated with the
longitudinal direction (x-axis), caused
1.75
predominantly by tractive wheel-rail forces.
1.50 This longitudinal component, together
with a slightly smaller lateral component,
Kinetic energy (mJ)
2
likely due to the dissipation of energy and
the increasing number of potential pathways
0
in the ballast matrix through which the load
can transfer. The magnitude of the CoV
–2 is similar among all three axes, with the
largest measurements associated with the
–4 longitudinal (x-axis) displacement instead of
the expected vertical (z-axis) displacement
associated with the loading vector. The CoV
–6
0 20 40 60 80 100 associated with the longitudinal direction
Time (sec) (x-axis) is pronounced only with the sleeper
K4 – 0 mm K3 – 250 mm K2 – 400 mm K1 – 550 mm and the sleeper-ballast interface.
Figure 14 illustrates positive skewness
Figure 11 Potential energy of all Kli-Pis for all the displacement statistics. The data
is significantly skewed in the longitudinal
to 0.219 mJ for Kli-Pi 1. The significance of to the stochastic nature of the measure- (x-axis) and lateral (y-axis) directions
the statistical characteristics of this metric ments, especially that of the non-uniform compared to the vertical (z-axis) direction,
follows in the next section. displacements, producing a relatively large primarily caused by the locomotive tractive
number of peak-to-peak amplitudes that forces. These forces are largely dissipated
Statistical descriptors need to be accounted for. For each DoF, through the sleeper-ballast interface near
Reviewing both the measurements and the minimum, mean, maximum, standard the position of Kli-Pi 3, where after the
literature, the underlying probabilistic deviation, coefficient of variance (CoV) skewness decreases noticeably for the
nature of the particle response is clear. and skewness statistics were calculated. Of instruments positioned at greater depth in
Descriptive statistics is employed to all the statistics analysed, the mechanical the ballast.
quantitatively compare these results. For work provided the most representative and
each dataset, a total of 2 000 of the largest significant results compared to that of the Rotation
peak-to-peak amplitudes was extracted displacement and rotation statistics. The relationship between the rotation
to obtain a representative sample for each statistics and depth of installation is
DoF. This number was obtained through Displacement less clear compared to the displacement
an iterative process until a representative Figure 12 illustrates the difference among statistics. From the time-series data it was
distribution of peak-to-peak amplitudes the three spatial dimensions. For any given concluded that the preferred axis of rotation
had been extracted. This number is larger spatial direction, the largest mean peak- is primarily associated with the behaviour
than the number of passing axles due to-peak displacement is associated with of the sleeper. The largest mean rotation
18 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
(Figure 15) is associated with Kli-Pi 3 which
is prone to transverse PSR resulting from 0.20
the sleeper above, followed by Kli-Pi 1. 0.18
The mean peak-to-peak amplitude for any
0.16
instrument or DoF remains relatively small,
with only a fraction of a degree rotation 0.14
Displacement (mm)
being measured. This finding was verified 0.12
during the analysis stage, whereby the 0.10
rotational velocity (prior to the integration
0.08
to calculate the attitude or orientations)
was rarely measured to be greater than one 0.06
degree per second. 0.04
The CoV statistics (Figure 16) illustrate
0.02
the rigidity of the sleeper where the CoV is
0
relatively small in the longitudinal (x-axis) Mean (X) Mean (Y) Mean (Z)
and lateral (y-axis) directions. Comparing Spatial direction
the two metrics, the difference in magni- Kli-Pi 4 (0 mm) Kli-Pi 3 (250 mm) Kli-Pi 2 (400 mm) Kli-Pi 1 (550 mm)
tude of the rotational CoV is significantly
smaller than that of the displacement CoV Figure 12 Displacement: mean for all axes
(Figure 13). A small reduction in the CoV
with depth is present when comparing Kli-
Pi 3 and Kli-Pi 4. 2.0
The skewness metric of the rotational
1.8
measurements is much smaller in magni-
tude compared to that of the displacements 1.6
Coefficient of Variance (CoV)
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 19
from both the dissipation of energy and
1.50 the increasing number of load paths with
1.35 depth as the load is distributed. From the
results presented by Kruyt and Rothenburg
1.20
(2002) it is known that with an increase in
1.05 the confinement stress, the shape of the
Rotation (arcmin)
1.0 CONCLUSIONS
The importance of studying and quantifying
0.9
the mechanical behaviour of granular media
0.8 was noted by Cowin and Satake (1979).
Coefficient of Variance (CoV)
20 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
the overlapping stress frusta of the wheel
loads, non-uniform strain distribution was 2.0
measured. The variation in the loading, 1.8
material and support conditions in the field
1.6
test significantly amplified the standard
1.4
a strong correlation with the depth of instal-
1.2
lation. The skewness statistic of the Kli-Pi
installed at the greatest depth converged 1.0
to a value of 1.1. This corresponds with a 0.8
confined granular state where the probability 0.6
distribution function tends towards an
0.4
exponential distribution, corresponding to
existing literature. The increase of the skew- 0.2
ness statistic corresponds with an increase in 0
both the standard deviation and CoV. Standard Deviation Coefficient of Variance (CoV) Skewness
Test regimes over a longer period should
be perused with the inclusion of typical Kli-Pi 4 (0 mm) Kli-Pi 3 (250 mm) Kli-Pi 2 (400 mm) Kli-Pi 1 (550 mm)
Table 1 Peak-to-peak displacement statistics for the field test (all Kli-Pis)
Kli-Pi 1 Kli-Pi 2 Kli-Pi 3 Kli-Pi 4
Parameter
X Y Z X Y Z X Y Z X Y` Z
Mean (mm) 0.036 0.030 0.074 0.031 0.028 0.104 0.041 0.053 0.109 0.106 0.074 0.161
Minimum (mm) 0.008 0.006 0.009 0.007 0.004 0.007 0.007 0.004 0.009 0.013 0.008 0.005
Maximum (mm) 0.162 0.137 0.229 0.155 0.150 0.363 0.292 0.355 0.419 1.027 0.563 0.685
Standard deviation (mm) 0.032 0.028 0.056 0.291 0.298 0.925 0.064 0.073 0.100 0.177 0.117 0.187
Coefficient of Variance 0.881 0.916 0.762 0.925 1.050 0.887 1.574 1.385 0.918 1.666 1.591 1.165
Skewness 4.390 4.028 1.119 4.238 3.230 1.124 9.049 4.972 1.440 6.042 6.656 4.498
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 21
Table 2 Peak-to-peak rotation statistics for the field test (all Kli-Pis)
Kli-Pi 1 Kli-Pi 2 Kli-Pi 3 Kli-Pi 4
Parameter
X Y Z X Y Z X Y Z X Y Z
Mean (arcmin) 0.879 0.768 0.620 0.625 0.506 0.460 1.330 0.724 0.508 0.548 0.733 0.647
Minimum (arcmin) 0.261 0.332 0.283 0.265 0.185 0.203 0.180 0.239 0.192 0.285 0.390 0.345
Maximum (arcmin) 2.277 1.652 1.358 1.635 1.262 1.100 3.732 2.651 1.583 1.097 1.425 1.344
Standard deviation (arcmin) 0.542 0.322 0.260 0.314 0.264 0.204 1.145 0.481 0.293 0.184 0.241 0.218
Coefficient of Variance 0.617 0.419 0.419 0.503 0.522 0.444 0.861 0.664 0.577 0.336 0.329 0.337
Skewness 0.909 0.864 1.027 1.387 1.062 1.383 0.760 2.691 2.049 0.921 0.955 1.112
Table 3 Work energy statistics (all Kli-Pis) theory. PhD Dissertation, Japan: Kyoto University,
p 17.
Field test
Parameter Kouroussis, G, Caucheteur, C, Kinet, D, Alexandrou, G,
Kli-Pi 1 Kli-Pi 2 Kli-Pi 3 Kli-Pi 4 Verlinden, O & Moeyaert, V 2015. Review of
Mean (mJ) 0.287 0.415 0.424 0.639 trackside monitoring solutions: From strain gauges
to optical fibre sensors. Sensors, 15(8): 20115–20139.
Minimum (mJ) 0.037 0.028 0.034 0.032 Kruyt, N P & Rothenburg, L 2002. Probability density
Maximum (mJ) 0.885 1.445 1.641 2.699 functions of contact forces for cohesionless
frictional granular materials. International Journal
Standard deviation (mJ) 0.219 0.363 0.390 0.633
of Solids and Structures, 39: 571–583.
Coefficient of Variance 0.761 0.874 0.921 0.990 Lamas-Lopez, F, Cui, Y J, Costa Aguiar, S & Calon, N
2017. Assessment of integration methods
Skewness 1.756 1.442 1.122 1.113
for displacement determination using field
accelerometer and geophone data. Journal of
ACKNOWLEDGEMENT Broekman, A & Gräbe, P J 2018. Development and Zhejiang University, 18(7): 553–566.
Transnet Freight Rail is gratefully acknowl- calibration of a wireless, inertial measurement Lekarp, F, Isacsson, U & Dawson A 2000. State of the
edged for sponsoring the Chair in Railway unit (Kli-Pi) for railway and transportation Art. II: Permanent strain response of unbound
Engineering in the Department of Civil applications. Proceedings, 37th Annual South African aggregates. Journal of Transportation Engineering,
Engineering at the University of Pretoria. Transportation Conference (SATC 2018), pp 521–531. 126(1): 76–83.
Cowin, S C & Satake, M 1979. Continuum mechanical Lemkin, M & Boser, B E 1999. A three-axis
and statistical approaches in the mechanics of micromachined accelerometer with a CMOS
DECLARATION granular materials. Journal of Rheology, 23(2): position-sense interface and digital offset-trim
The authors declare no potential conflicts of 243–256. electronics. IEEE Journal of Solid-State Circuits,
interest with respect to the research, author- Cundall, P A & Strack, O D L 1979. A discrete 34(4): 456–468.
ship, and/or publication of this article. numerical model for granular assemblies. Li, D, Hyslip, J, Sussmann, T & Chrismer, S 2015.
Géotechnique, 29(1): 47–65. Railway Geotechnics. Boca Raton, FL: CRC Press,
Gräbe, P J & Clayton C R I 2009. Effects of principal pp 93–100.
REFERENCES stress rotation on permanent deformation in rail Liu, S, Huang, H, Qiu, T & Gao, L 2016a. Comparison
Aikawa, A 2009. Techniques to measure effects of track foundations. Journal of Geotechnical and of laboratory testing using SmartRock and discrete
passing trains on dynamic pressure applied to Geoenvironmental Engineering, 135(4): 555–565. element modelling of ballast particle movement.
sleeper bottoms and dynamic behaviour of ballast Graizer, V M 2010. Strong motion recordings and Journal of Materials in Civil Engineering, 29(3): 1–7.
stones. Quarterly Report of Railway Technical residual displacements: What are we actually Liu, S, Huang, H, Qiu, T & Gao, Y 2016b. Study on
Research Institute, 50(2): 102–109. recording in strong motion seismology? the ballast particle movement at different locations
Audley, M & Andrews, J D 2013. The effects of tamping Seismological Research Letters, 81(4): 635–639. beneath crosstie using “SmartRock”. Proceedings,
on railway track geometry degradation. Proceedings Indraratna, B, Ngo, N T & Rujikiatkamjorn, C 2016 Joint Rail Conference, Columbia, SC.
of the Institution of Mechanical Engineers, Part F: 2011. Behaviour of geogrid-reinforced ballast Liu, S, Haung, H, Qiu, T & Kwon, J 2016c. Effect of
Journal of Rail and Rapid Transit, 227(4): 376–391. under various levels of fouling. Geotextiles and geogrid on railroad ballast particle movement.
Bennett, R, Hayes-Gill, B, Crowe, J A, Armitage, R, Geomembranes, 29(3): 313–322. Transportation Geotechnics, 9: 110–122.
Rodgers, D & Hendroff, A 1999. Wireless Ju, S H, Lin H T & Huang J Y 2009. Dominant Lobo-Guerrero, S & Vallejo, L E 2006. Discrete element
monitoring of highways. Smart Structures and frequencies of train-induced vibrations. Journal of method analysis of railroad ballast degradation during
Materials 1999: Smart Systems for Bridges, Sound and Vibration, 319: 247–259. cyclic loading. Granular Matter, 8(3): 195–204.
Structures and Highways, 3671: 173–182. Kempe, V 2011. Inertial MEMS: Principles and Makse, H A, Johnson, D L & Schwartz, L M 2000.
Bowness, D, Lock, A C, Powrie, W, Priest, J A & Practice. 1st ed. Cambridge, UK: Cambridge Packing of compressible granular materials. Physical
Richards, D J 2006. Monitoring the dynamic University Press. Review Letters, 84(18): 4160–4163.
displacements of railway track. Journal of Rail and Kitamura, R 1980. Analysis of deformation mechanism Milne, D, Le Pen, L, Watson, G, Thompson, D et al
Rapid Transit, 221: 3–22. of particulate material based on the probability 2016a. Measuring ballast acceleration at track
22 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
level. Proceedings, 3rd International Conference on ASME Joint Railroad Conference, Newark, NJ, Wilke, D N, Govender, N, Pizette, P & Abriak, N E
Railway Technology: Research, Development and pp 111–124. 2016. Computing with non-convex polyhedra on
Maintenance. 5–8 April 2016, Cagliari, Sardinia, Italy. Rothenburg, L & Kruyt, N P 2009. Micromechanical the GPU. Proceedings, 7th International Conference
Milne, D, Le Pen L, Watson, G, Thompson, D et al definition of an entropy for quasi-static deformation on Discrete Element Methods, 1–4 August 2016,
2016b. Proving MEMS technologies for smarter of granular materials. Journal of the Mechanics and Dalian, China, pp 1371–1377.
railway infrastructure. Procedia Engineering, 143: Physics of Solids, 57: 634–655. Woodman, O J 2007. An introduction to inertial
1077–1084. Rucker, W 1982. Dynamic interaction of a railroad-bed navigation. Technical Report No 696. Cambridge,
Milne, D R M. Le Pen L M, Thompson D J, Powrie, W with the subsoil. Proceedings, Soil Dynamics and UK: University of Cambridge Computer Laboratory.
2017. Properties of train loading frequencies and Earthquake Engineering Conference, Southampton, Zhai, W M, Wang, K Y & Lin, J H 2004. Modelling and
their applications. Journal of Sound and Vibration, UK, pp 435–448. experiment of railway ballast vibrations. Journal of
397: 123–140. Straser, E G & Kiremidjian, A S 1998. A modular, Sound and Vibration, 270: 673–683.
Mvelase, G M, Anochie-Boateng, J K & Gräbe, P J 2012. wireless damage monitoring system for structures. Zhang, X, Zhao, C & Zhai, W 2016. Dynamic
Application of laser-based technology to quantify Technical Report No 128. Stanford, CA: John A. behaviour analysis of high-speed railway ballast
shape properties of railway ballast. Proceedings, 31st Blume Earthquake Engineering Center, Stanford under moving vehicle loads using discrete element
Annual Southern African Transport Conference University. method. International Journal of Geomechanics,
(SATC), Pretoria, pp 243–254. Stewart, J 2012. Calculus, Early Transcendentals. 17(7).
Nathanson, H C & Wickstrom, R A 1965. A resonant- 7th ed. Boston, MA: Cengage Learning.
gate silicon surface transistor with high-Q bandpass Tutumluer, E, Huang, H, Hashash, Y & Ghaboussi, J
properties. Applied Physics Letters, 7(4): 84. 2006. Aggregate shape effects on ballast tamping BIBLIOGRAPHY
Nejikovsky, B & Keller, E 2000. Wireless and railroad track lateral stability. Proceedings, Selig, E T & Waters, J M 1994. Track Geo-technology
communications-based system to monitor AREMA Annual Conference, 23 May 2006, Salt and Substructure Management. 1st ed. London:
performance of rail vehicles. Proceedings, IEEE/ Lake City, UT, pp 17–20. Thomas Telford Ltd.
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 23
TECHNICAL PAPER
Reliability performance of
bridges designed according
Journal of the South African
Institution of Civil Engineering
Lenner R, Basson SE, Sýkora M, Van der Spuy PF. Reliability performance of bridges designed according to TMH-7 NA load model.
24 J. S. Afr. Inst. Civ. Eng. 2021:63(1), Art. #1102, 13 pages. http://dx.doi.org/10.17159/2309-8775/2021/v63n1a3
Van der Spuy & Lenner 2018; Pérez &
Lenner 2019). The second case study is 40
based on WIM data from National Route 1
(N1) near Kilner Park, as used in the study
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 25
TMH7), this amplification is removed
from the NA load model specifically for 2.0
this study to enable a comparison with the
load effects derived on the basis of static
Fréchet Two-parameter
WIM data. α > 1.14 lognormal
The investigation of uncertainty in 1.5 α = 3V + V 3
dynamic effects of heavy traffic at the ulti-
mate limit states is beyond the scope of this Gumbel α = 1.14
Skewness (α)
contribution; for further discussion see for
1.0 Three-parameter
instance work by Lenner & Sýkora (2016) Gamma
lognormal for any α α = 2V
and Ivanković et al (2019).
Weibull
α < 1.14
0.5
Roosboom case study (RCS) Beta with zero
The first case study consists of a reliability lower bound α < 2V
analysis considering only the traffic load
effect based on seven years of WIM data 0
for a single lane. It is first necessary to Coefficient of Variation (V)
obtain the distribution function describing
the load effect and then formulate the limit Figure 2 R
elation between skewness and coefficient of variation for monthly maxima for each
state function. This enables the calculation span length of RCS; the points indicate V and α for various span lengths (Basson 2020)
of the reliability index by means of the First
Order Reliability Method (FORM). Spuy 2020). However, larger blocks tend to as close to zero, which indicates a distribu-
remove variations in observations and yield tion that approaches a normal distribution
Overview of traffic load data that better adheres to the iid require- (refer to Basson 2020 for full details).
effect calculation ment. Considering seven years of available Most of the points in Figure 2 are
The work presented in this study is based recordings, monthly maxima of the bend- located above the two parameter lognor-
on the load effects obtained previously ing moment are further used in this study. mal line, which suggests that the three
by means of moving-load analysis which The block is large enough to reduce the parameter lognormal (LN3) distribution
accounts for single-truck events and mul- variations, and there are enough blocks for or Weibull distribution (for maxima) is
tiple-truck events (Van der Spuy & Lenner the statistical description. Detailed analysis appropriate for modelling the monthly
2018). Only the sagging bending moment in of statistical inference methods and proba- maxima. The GEV distribution can be
increments of 5 m for 5–50 m span lengths bilistic modelling of load effect maxima are used conveniently, as it approaches the
is presented in this work to limit the scope. offered by O’Brien et al (2015). Weibull distribution when the skewness
A simplified approach exploring a is less than 1.14. Both the LN3 and GEV
Probabilistic modelling of wide range of extreme value distribution distribution are asymmetrical and have
traffic load effects as developed by Holický (2013) is adopted three model parameters. Both models are
Consideration of single-loading events here. It is particularly useful in finding flexible in allowing for both positive and
leads to a distribution of the load effects applicable models by plotting the skew- negative skewness and can thus cater for
which need to be described by a probability ness and the coefficient of variation of different sample characteristics obtained
distribution. There are different approaches the data set against plotted distributions. from the different span lengths. As a
that vary from utilising basic fitting of a Considering the monthly blocks, the result, the LN3 and GEV distributions are
normal distribution to the population or to sample skewness and coefficient of varia- further investigated with an aim to decide
the tail (Nowak 1993; Nowak & Rakoczy tion, denoted by α and V respectively, are on an appropriate model.
2013; Soriano et al 2016), to the more com- used to plot data points on the developed
mon extreme value theory and tail fitting diagram for each span length. Assessment of selected
(Coles 2001; Caprani et al 2008; Enright Figure 2 clearly indicates the fit of the probabilistic distributions
2010; O’Brien et al 2015). data to various distributions including Goodness-of-fit tests performed in R
A block maxima approach is adopted the GEV family, i.e. the Gumbel, Fréchet (R Core Team 2018) assess the suitability
in this study as it is a proven and effective and Weibull distribution. The V ranges of both the LN3 and GEV distribution in
technique in studying extreme traffic between 0.08 and 0.10, which indicates representing the monthly maxima. The
events (Crespo-Minguillón & Casas 1997; a relatively narrow distribution shape in tests consist of diagnostic plots and the
Caprani et al 2003; Getachew & O’Brien comparison to the characteristics of daily modified Anderson-Darling (AD) (Ang
2007; Hajializadeh et al 2012). It essentially and weekly maxima where V ranges 0.12 to & Tang 2006) hypothesis testing. For
isolates a maximum event in each block. 1.16, and 0.10 to 0.14 respectively. All span each considered distribution, the model
To utilise the technique, the underly- lengths have a positive skewness (right tail) parameters are inferred by the maximum
ing assumption is that the data must be equal to or less than 0.8, except for a 15 m likelihood estimation (MLE).
independent and identically distributed span that has a small negative skewness Visual inspection of the diagnostic
(iid) (Coles 2001). The minimum size of (left tail). Furthermore, the skewness for plots for all span lengths shows that both
a block is one day (Caprani 2012; Van der the 10 m and 15 m spans can be observed models fit the data well. The points on
26 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
Emperical and theoretical density Q–Q plot
0.020
180
160
0.015
Empirical quantiles
140
Density
0.010
120
100
0.005
80
0 60
60 80 100 120 140 160 180 200 80 100 120 140 160 180
Data Theoretical quantiles
0.8 0.8
Empirical probabilities
0.6
0.6
CDF
0.4
0.4
0.2
0.2
0 0
60 80 100 120 140 160 180 200 0 0.2 0.4 0.6 0.8 1.0
Data Theoretical probabilities
Figure 3 Diagnostic plots for the monthly maxima of sagging bending moment for a 5 m span RCS represented by the LN3 distribution (Basson 2020)
the probability plots adhere to a straight The AD test is further used as a numeri- Table 1 M
odified Anderson-Darling p-values for
line and a good fit is seen for the density cal measure to substantiate the results LN3 distribution and GEV distribution
plots. For illustrative purposes, the Q-Q obtained from the diagnostic plots. When fitted to monthly maxima of RCS
plots obtained for a 5 m span are shown the p-value obtained from the AD test Span p-value p-value
in Figure 3 for LN3 and in Figure 4 for exceeds the significance level, the data length (m) (LN3) (GEV)
GEV. In the figures, the following plots are can be represented by the selected model.
5 0.73 0.75
included (clockwise from top left): A significance level of 0.05 is typically
1. Histogram with a fitted probability accepted (Fisher 2006). The LN3 distribu- 10 0.27 0.45
density function tion is converted to a normal distribution
15 0.15 0.33
2. Plot with empirical and theoretical (Holický 2013). The AD test applied to the
quantiles – probability paper of a theo- transformed data obtains p-values as pro- 20 0.06 0.09
retical distribution vided in Table 1. The p-values exceed 0.05,
3. Empirical and theoretical cumulative which indicates that the monthly maxima 25 0.58 0.58
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 27
Probability plot Quantile plot
1.0
360
0.8
340
0.6
Empirical
320
Model
0.4 300
280
0.2
260
0
0 0.2 0.4 0.6 0.8 1.0 260 280 300 320 340 360
Empirical Model
360
0.015
340
Return level
f(z)
320 0.010
300
0.005
280
260
0
1e-01 1e+00 1e+01 1e+02 1e+03 240 260 280 300 320 340 360 380
Return period z
Figure 4 Diagnostic plots for the monthly maxima of sagging bending moment for a 5 m span RCS represented by the GEV (Basson 2020)
The Q-Q plots in Figures 3 and 4 suggest is deemed appropriate for shorter spans, estimate the overall β, where the probabil-
that upper tail behaviour is described approx- while it may be reduced for longer spans ity distribution function of a standardised
imately only. Detailed analysis that might with the dominant dead load. normal distribution FU(x) relates β to pf .
improve the model for traffic load extremes When only traffic load effects are ana-
will be the scope of further research. lysed, a reliability criterion can be defined pf = P(E > ed ) = 1 – F50(ed )(1)
as the instance when the actual traffic load
Design load effect effects exceed the design load effect. A ref- pf = FU(+αEβ) = FU(–0.7β)(2)
The overall reliability level, described by erence period of 50 years is chosen in order
the reliability index β, can be split into the to provide for comparison with the target where FU (·) denotes the cumulative distri-
resistance part R and load effect part E reliability commonly listed in codes such bution function of the standardised normal
(EN 2002; ISO 2015). For this case study, as EN 1990 (EN 2002) or SANS 10160‑1 variable and F50 (·) is the 50-year maximum
the load effect part is of interest, where (SANS 2019). Therefore, the actual load load effect distribution.
the reliability level is expressed with the effects derived from the WIM data should
load effect index βE equal to αE β. The correspond to a 50-year maximum. The 50-Year maximum load effect distribution
variable αE is a sensitivity factor of traffic design load effect can then be described To obtain the maximum load effect
load effects obtained from FORM. This as a fractile of the 50-year maximum load distribution for a 50-year period, the
describes the relative importance of the effect distribution F50 (x). The probability monthly maxima distribution is statistically
load effects in obtaining β and is accepted of failure pf is then approximated by the projected by raising the probability
in this work as –0.7 for dominant action probability that the design load effect ed distribution function F(x) to an appropriate
as recommended by EN 1990 (EN 2002) is exceeded by the traffic load effects E, as power n (per Equation 3). In Equation 3,
and ISO 2394 (ISO 2015). This αE-value shown in Equation 1. Equation 2 is used to n is the number of monthly maxima in
28 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
Notional lane Notional lane Notional lane Notional lane
Notional lane Notional lane Notional lane Notional lane Notional lane
Carriageway width = Above 7.4 m up to and including 11.1 m Carriageway width = 4.8 m up to and including 7.4 m
Design scenario 2: 1 traffic lane; 3 notional lanes Design scenario 4: 2 traffic lanes; 2 notional lanes
a reference period, i.e. n = 600, and Fn(x) and 4), and its unbounded right tail (for is often narrower than a traffic lane. As a
denotes the probability distribution function positive skewness) allows for the thorough result, a number of notional lanes can be
of the 50-year maxima, F50 in Equation 1. investigation of the design load effects. As a fitted to a bridge deck which in reality only
result, an LN3 distribution is adopted here carries a single traffic lane (bridges are built
Fn(x) = [F(x)]n(3) for the monthly maxima. to accommodate not only a traffic lane,
but also shoulders). Four design scenarios
For each span length, the 50-year maxi- Reliability analysis for are defined where initially an assumption
mum load effect distribution was obtained traffic load effects of one traffic lane corresponding to one
from both the LN3 distribution and the notional lane is investigated, followed by the
GEV distribution of monthly maxima. Formulation of limit state function opposite notion of one traffic lane occupy-
Depending on the selected model, the char- The limit state function Z defines the ing a deck which is in reality designed for
acteristics of the projected maximum load desired failure mode for the investigation three notional lanes. The same is carried
effect distribution will differ. According to as per Equation 4. Model uncertainty θE is out for two possible traffic lanes and three
Castillo et al (2005), a lognormal distribu- introduced in this contribution according to and two notional lanes respectively. A sum-
tion fitted to maximum values converges to a lognormal distribution with a mean of 1.0, mary of the scenarios under consideration is
a Gumbel distribution as the power n tends and a COV of 0.1 is commonly used in lit- depicted in Figure 5 and provided in Table 2.
to infinity. erature (JCSS 2001; fib Bulletin 80 2016; Von
The Weibull distribution for maxima, Scholten 2004). The design load effect ed is Table 2 Overview of design scenarios for RCS
represented by the GEV distribution, has a deterministic value provided by TMH7 for Design
a finite upper bound and its maximum NA loading, while E50 is described by the Description
scenario
domain of attraction remains a Weibull 50-year maximum load effect distribution
1 1 traffic lane, 1 notional lane
distribution (Castillo 1988). The concern with an LN3 distribution.
with using the maximum Weibull distri- 2 1 traffic lane, 3 notional lanes
bution is that the upper bound remains E = θE E50 3 2 traffic lanes, 3 notional lanes
unchanged irrespective of the reference Z = ed – θE E50(4)
4 2 traffic lanes, 2 notional lanes
period projected to. This means that the
50-year maximum load effect distribution FORM is implemented to estimate failure
has the same upper bound as the monthly probability for each span length from The design values of traffic load effects, ed,
maxima distribution. This could potential- Equation 4; the overall reliability index β is for each scenario given in Table 3 represent
ly be problematic, as the design load effects then derived from Equation 2. the rounded maximum obtained bending
may exceed the bound determined by the moments according to TMH7 NA loading
distribution function parameters. Overview of reliability analysis for the midspan of a beam when designing
The physical argument that might jus- for different design scenarios for one, two and three notional lanes. It is
tify the upper bound could be formulated The actual traffic load effects are derived important to note that dynamic effects are
on the basis of legal limits for axle and from the WIM data for a single traffic lane. excluded from the NA load model based
vehicle loads. However, the concern about To make a meaningful comparison with on the Swiss formula, and the ULS partial
frequent overloading of trucks makes this TMH7 it is necessary to investigate different factor γf = 1.65 is obtained as a product
argument doubtful. Further, the LN3 dis- scenarios. The concept of a notional lane as of γfL = 1.5 and γf3 = 1.1. The calculated
tribution shows a similar fit to the monthly defined in TMH7 is different to the actual reliability indices β for each span length are
maxima as the GEV distribution (Figures 3 traffic lane, meaning that a notional lane given in Table 4.
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 29
11 950 Overall bridge width
1 750
250
300
200
1 800
4 450 2 000
Figure 6 Cross-section of bridge deck with traffic lane arrangement (units in mm) of KPCS (Basson 2020)
Table 3 D
esign traffic load effects (in kNm)
Traffic lane 1 Traffic lane 2 for ULS determined from NA loading
3.7 m 3.7 m in TMH7
–0.18 Span One Two Three
length notional notional notional
(m) lane lanes lanes
5 360 650 920
10 1 050 1 960 2 810
LLDF2 = 0.49 15 2 080 3 950 5 740
20 3 470 6 900 8 900
25 5 200 9 720 12 500
LLDF1 = 0.95 30 7 280 12 900 16 600
35 9 720 16 400 21 200
1.18
40 13 100 20 200 26 200
45 15 700 24 300 31 600
Figure 7 L ateral load distribution factors corresponding to the position of traffic load effects in
50 18 500 28 700 37 300
two lanes (Basson 2020)
In order to make a meaningful com- show high reliability, mostly in excess of concrete bridge with a twin spine deck is
parison of the obtained results for a 50-year β = 10, leading to very low probability of considered as a representative example for
reference period, EN 1990 (EN 2002) failure. Care must be exercised in this case, bridges in South Africa. The effect of lane
recommends βt = 3.8 while the South as this is pointing only to the global action loading and load distribution between two
African National Standard, SANS 10160-1 effects and it might not necessarily be true main girders is investigated.
(SANS 2019), uses a βt of 3.0 for the same to a critical element of a bridge, for instance
reference period. It is readily observed that a single girder. Yet, it does provide an insight Bridge deck analysis and design
the way each scenario is defined is funda- into the performance of NA loading and the The cross-sectional properties of the
mental to the obtained reliability index. The inherent conservatism in the definition of designed bridge deck, together with the
extreme case of one notional lane to one notional lane width. Design scenarios 3 and traffic lane arrangement, are illustrated in
traffic lane shows very low β (therefore high 4 exhibit similar results, where two traffic Figure 6. Two actual traffic lanes of 3.7 m
probability of failure) for short spans, while lanes are considered and again, when the width each are considered with necessary
for spans ≥ 25 m the performance is deemed number of notional lanes is matched with surfaced shoulders on both sides. The
satisfactory. The poor performance in this the number of reflected traffic lanes, a poor deterministic analysis reveals that the spine
scenario can be contributed to the way a reliability performance is observed. At the beam is the critical member. Note that this
notional lane is defined, and that it is physi- same time, for all considered scenarios, the bridge is not skew. As the maximum of
cally impossible to have one notional lane longer span lengths prove to show satisfac- design load effects to load effects per meas-
equal to one traffic lane. It is neglecting the tory performance when a target β of 3.0 or ured vehicles is considered in this inves-
mandatory shoulders on the roads – mean- even 3.8 is considered. tigation, it is expected the results of the
ing the deck width is always wider than a analysis would be similar for skew bridges.
traffic lane and therefore necessitates more The permanent loads for the design
notional lanes during the design. Kilner Park Case Study (KPCS) include the dead load of the bridge
On the other hand, the case of wide The second case study utilises three deck, typical F-shape Type A parapets
shoulders results in a scenario where three consecutive years of WIM data for the two (SANRAL 2012) and asphalt road surfacing
notional lanes are compared to a single outer traffic lanes in both directions (Van with a thickness of 40 mm (SANRAL 2010).
traffic lane. In this case the obtained values der Spuy et al 2019c). A 20 m reinforced NA loading (excluding dynamic effects)
30 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
is applied to the grillage model using Table 4 O
verall reliability indices β provided for each design scenario and corresponding to a
three notional lanes that correspond to 50-year period of RCS
the considered deck width as per TMH7 Design Design Design Design
requirements. The ULS mid-span moment Span length (m)
scenario 1 scenario 2 scenario 3 scenario 4
for the considered beam according to the
5 –1.2 9.7 3.8 –0.7
analysis equals to 11 600 kNm, which
10 –1.6 10.8 4.2 –0.7
results in 29 Y40 (450 MPa yield strength)
steel bars required for flexural resistance. 15 1.2 15.0 7.5 2.3
The thickness of the deck here is the main 20 1.7 10.6 6.1 3.2
factor influencing the amount of moment
25 2.8 10.3 6.2 3.6
distribution per spine. The next section
30 3.8 10.6 6.6 3.9
explains lateral load distribution factors
used in this case. 35 4.3 10.4 6.4 3.9
40 5.1 10.1 6.3 3.8
Influence of traffic load effects 45 5.2 10.2 6.5 4.0
on critical spine beam
50 5.6 10.7 6.9 4.4
The transverse stiffness of the considered
superstructure is a key factor for resulting
load effects due to WIM-based loading on
the spine beam. By using the transverse Traffic lane 1 Traffic lane 2 Traffic lane 3
3.5 m 3.5 m 3.5 m
influence line for the critical element
(Figures 7 and 8) it is possible to capture the –0.18
load sharing effect and determine how much
of the total load is resisted by the spine.
The lateral load distribution factor (LLDF) LLDF3 = 0.09
numerically describes the contribution of
each WIM lane to the bending moment in
the investigated girder and is located at the LLDF2 = 0.52
centre of the traffic lane. Uncertainty in the
lateral load distribution is accounted in the
probabilistic analysis by load effect model
LLDF1 = 0.96
uncertainty, θE, similarly to Equation 4.
Apart from the scenario depicted in 1.18
Figure 6, a further loading arrangement is
investigated for the case of a bridge utilised Figure 8 L ateral load distribution factors corresponding to the position of traffic load effects in
as a three-lane deck providing a reduced three lanes (Basson 2020)
lane width. The critical arrangement of
WIM traffic lanes, along with the transverse
influence line in this case corresponds to 2.0
Figure 6. This loading scenario is provided
in order to capture the possibility that a
Fréchet Two-parameter
bridge initially designed for two traffic lanes α > 1.14 lognormal
with three notional lanes may actually be, 1.5 α = 3V + V 3
under special circumstances, utilised as a
three-lane bridge with reduced shoulders. Gumbel α = 1.14
Skewness (α)
Probabilistic modelling of
0
traffic load effects 0 0.1 0.2 0.3 0.4
The monthly maxima combined for two Coefficient of Variation (V)
opposite travel directions, representing the
extreme traffic load effects for the critical Figure 9 A
diagram where the relation between skewness and coefficient of variation is used to
spine beam, are modelled and assessed find applicable models to represent the combined monthly maxima for each load case
similarly to the RCS. Figure 9 illustrates the of KPCS; the points indicate V and α for various span lengths (Basson 2020)
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 31
Emperical and theoretical density Q–Q plot
0.0035
0.0030 600
0.0025 500
Empirical quantiles
0.0020
Density
400
0.0015
300
0.0010
0.0005 200
0
100 200 300 400 500 600 700 100 200 300 400 500 600 700
Data Theoretical quantiles
0.8 0.8
Empirical probabilities
0.6 0.6
CDF
0.4 0.4
0.2 0.2
0 0
100 200 300 400 500 600 700 0 0.2 0.4 0.6 0.8 1.0
Data Theoretical probabilities
Figure 10 Diagnostic plots of sagging moments provided in R for two southbound lanes combined of KPCS fitted with the LN3 distribution (Basson 2020)
characteristics of each load case and relates per Table 5) are considered. As in the first Table 5 M
odified Anderson-Darling p-values
the sample skewness to the coefficient of case study, the adopted probabilistic model for LN3 distribution fitted to
variation. It is observed that the sample provides only an approximation of the upper combined monthly maxima of KPCS
skewness varies considerably, which can be tail behaviour, and detailed analysis might Load
conveniently captured by the LN3 distribu- improve predictions of traffic load extremes. Permutations p-value
Case
tion. Consistently with the RCS, the use of The modified AD test results are sum-
1 Nb1, Nb2 0.11
the Weibull distribution is avoided, as the marised in Table 5. Nb1 and Nb2 respective-
fixed upper bound is deemed to be unjusti- ly refer to the slow lane and fast lane in the 2 Sb1, Sb2 0.98
fied. As only extracted load effects were northbound direction. The same reference 3 Nb2, Nb1 0.88
available for this project, it is impossible to is made for Sb1 and Sb2 in the southbound 4 Sb2, Sb1 0.63
determine the cause of the high skewness direction. The traffic lane combinations
5 Nb1, Sb1 0.14
for the single-load effect situated in the for each load case are described in Table
Frechet domain. 5, where the traffic lanes are ordered from 6 Sb1, Nb1 0.89
The moment parameters for the LN3 left to right in accordance to the closest and 7 Nb1, Nb2, Sb2 0.11
distribution are estimated using MLE. Visual furthest traffic lane from the critical spine
8 Sb1, Sb2, Nb2 0.71
inspection of the diagnostic plots shows beam. It shows the recorded traffic used in
that the LN3 distribution fits the combined each of the considered lanes. The p-value for
monthly maxima well. For illustrative pur- each load case exceeds a significance level of The results show relatively high p-values
poses, Figure 10 shows the diagnostic plots 0.05, which confirms that the LN3 distribu- for all the load cases, except for Load Cases
for the load case when the two traffic lanes tion can be used to represent the combined 1, 5 and 7 that have much lower p-values.
in the southbound direction (Load Case 2 monthly maxima. This observation concurs with the diagnostic
32 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
Table 6 C
onventional probabilistic models describing the basic variables in the limit state function
Symbol Standard
Distribution Mean
Name of basic variable of basic Unit deviation Reference
type (μ X)
variable (X) (σX)
Traffic load (50 years) Q kNm LN3 Based on moment parameters of 50-yr maximum load effect distribution
plots. A shared characteristic between these various researchers, two different models different load cases to find the governing
load cases is that the concurrent traffic of uncertainty for the resistance are further load case for the critical member under
load effects in the northbound slow lane investigated in order to assess the sensitiv- consideration. For future studies, it is rec-
(Nb1) have the largest contribution to the ity of their parameters. ommended to identify and separate different
combined monthly maxima. The results for loading event types. By fitting distributions
these load cases suggest that the combined Reliability analysis results and discussion to the individual loading event types, the
monthly maxima imposed on the critical Reliability analysis of the critical element accuracy of the probabilistic models should
spine beam may originate from a mixture of is again performed using FORM. Overall improve. This will also lend further sup-
loading events. For example, the combined reliability indices β obtained for each load port to the assumption of independent
monthly maxima may include side-by-side case depend on the selected model of the identically distributed extremes of traffic
truck events together with single or following resistance model uncertainty θR. The result- load effects. The results also show that the
truck events. The 50-year maximum load ing values for β for each load case are pre- coefficient of variation of θR influences β,
effect distribution can again be obtained sented in Table 7. For Model 1, the obtained which agrees with literature (Allaix 2007;
by Equation 3, where F(x) is the probability β-values range between 6.2 and 7.8, whereas Holický et al 2010).
distribution function of the LN3 distribution for Model 2, the values range between When comparing the results to target
describing the combined monthly maxima. 6.2 and 7.0. The lowest β is obtained for reliability indices βt of 3.8 and 3.0, the
Load Case 8, when three traffic lanes are
Formulation of limit state function for positioned on the bridge and the slow lane Table 7 R
eliability indices β, corresponding to a
the critical element reliability analysis in the southbound direction has the largest 50-year reference period, provided for
The limit state function in Equation 5 is loading contribution on the critical spine each load case depending on the model
used to determine the reliability perfor- beam. It is seen that the addition of another of the resistance model uncertainty θR
mance of the bending moment capacity of traffic lane is not as influential as in the RCS
β
the critical spine beam at ULS. The func- due to its reduced influence according to the Load Case
tion consists of independent basic variables transverse stiffness. Model 1 Model 2
that describe the resistance, permanent The range of β-values is caused by con- 1 7.1 6.6
load, traffic load effects and model uncer- siderably varying statistical moments for Q
tainties relevant to the critical spine beam. amongst the load cases under consideration. 2 6.6 6.5
The higher the mean, coefficient of variation 3 6.3 6.4
f y As and skewness of the traffic load effects, the
Z=θ
R fy As 1 – d – θG(G + Gw) 4 7.6 6.8
1.34fcubd lower the obtained β-values are. Depending
– θQQ(5) on whether the traffic moves in the north- 5 7.8 7.0
bound or southbound direction and whether
6 7.8 6.9
Table 6 defines the basic variables pre- a slow lane or a fast lane is located above the
sented in Equation 5 with respective critical spine beam, the moment parameters 7 7.3 6.7
probabilistic models. Due to ambiguities describing the traffic load effects vary. The
8 6.2 6.2
in model uncertainty models presented by results show the importance of investigating
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 33
β-levels would mostly increase, as the typi-
90 cally conservative nominal value of DAF
is considered in design (Van der Spuy et
80
al 2019c), while measurements typically
70 suggest very low dynamic amplification
due to heavy traffic (besides effects on
Relative importance α2 (%)
60
local or short-span members). DAF com-
50 monly exhibits an inverse proportionality
between the dynamic amplification and
40 vehicle weight and a number of lanes, and
a reduction in its scatter with increasing
30
weight and number of lanes (Lenner &
20 Sýkora 2016; Ivanković et al 2019; O’Brien
et al 2009; O’Connor & Enevoldsen 2007).
10
As the results of this study are presented
0 for sagging moments, they need to be
Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 interpreted with caution for other types of
Load case 2 Load case 3 Load case 8 load effects:
QQ Similar results are expected for hogging
θR 10.24 43.56 6.25 19.36 8.41 27.04
moments.
fy 9.61 14.44 5.76 6.25 7.84 9.00 QQ For shear, particularly at shorter spans
Q 65.61 20.25 77.44 64.00 70.56 49.00 and local verifications, a detailed analy-
sis of axle loads is needed. This would
θQ 10.24 9.61 7.29 6.76 9.00 9.00
likely result in a lower reliability index
(Van der Spuy 2020).
elative importance α2 of the resistance model uncertainty θR, yield strength f y , traffic
Figure 11 R Reliability of steel bridges is commonly
load (50 years) Q and traffic load model uncertainty θQ for Load Cases 2, 3 and 8 dominated by fatigue, and modelling of
traffic load extremes becomes less impor-
obtained β-values are significantly higher. all the basic variables in the limit state tant; an example of fatigue verification of a
This observation shows that TMH7 NA function are considered in the summation. steel bridge using WIM was provided in a
load model exhibits satisfactory reliability Thus, α 2 describes the relative importance recent study (Ma et al 2018).
performance for the considered bending of each basic variable. The primary focus
capacity of the critical elements of a 20 m of the sensitivity analysis is Load Cases
twin spine deck but might be deemed 2, 3 and 8, where the lowest β-values are Conclusions and
uneconomic. It is, however, important to obtained. Variables Q, θR, f y and θQ show recommendations
consider shear forces as well. Yet, according the largest significance, with Q and θR The Roosboom Case Study, considering
to Tabsh and Nowak (1991), and Matos et having the highest α-values. Considering sagging moments as dominating traffic
al (2019), uneconomically high β-values the two models for θR, Figure 8 provides load effects with an assumed sensitivity
(larger than 5–6) could be expected for the relative importance α 2 of the most factor αE = –0.7, indicates that NA loading
structures in a good condition. The high significant variables for Load Cases 2, 3 generally performs well for spans ranging
β-levels are also in agreement with the and 8. From Model 1 to Model 2, the coef- from 15 to 50 metres in length. However,
results of a previous study by Teichgräber ficient of variation of θR increases and, in a poor reliability performance is seen for
et al (2019) who found many sources of response, the relative importance of θR short-span narrow bridges, especially for
“hidden safety” in the traffic-load model also increases, whereas Q becomes less 5 m and 10 m spans where the number
LM1 in EN 1991-2, leading to ~25% over significant (Figure 11). This would justify of traffic lanes is equal to the number of
design. Future research should investigate considering a reduced value of the sensitiv- notional lanes for NA loading. The find-
other bridge deck types for a 20 m span to ity factor for traffic load effect, well below ings agree with Oosthuizen et al (1991),
determine whether the reliability results |0.7|. However, as a next step of future Anderson (2006), and Van der Spuy and
concur with the results of the twin spine research it is recommended to verify and Lenner (2019a), who also found deficien-
deck. In addition, it is also recommended improve the theoretical probabilistic mod- cies in the NA loading for short-span and
to investigate shear as a failure mode (see els used for the basic variables, especially narrow bridges. The generally acceptable
the Discussion further down). θR. Actual data would improve the uncer- performance of the rest of the bridges
tainty quantification of a variable, which in can be contributed to the relatively high
Sensitivity analysis results and discussion turn would improve the accuracy of α 2 and partial factor, but mostly to the nature of
The FORM analysis also provides the the obtained β. the notional lane definition of the NA load
sensitivity factors α for all basic variables. model. Owing to the geometry of a typical
The factor indicates the influence of a vari- highway bridge, more notional lanes of the
able on the obtained β. As |α| numerically Discussion NA model are used for design than the
increases, so does the significance of the This study is limited to the investigation bridge can physically carry. The Roosboom
basic variable. Note that ∑α 2 = 1, when of static effects only. If DAF were included, case study identifies deficiencies in TMH7
34 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
for normal traffic conditions, and it also 7th International Conference on Structural Hajializadeh, D, O’Brien, E J, Enright, B, Caprani,
identifies span lengths that achieve a high Engineering, Mechanics and Computation, Cape C C, Sheils, E & Wilson, S P 2012. Probabilistic
reliability performance when considering Town, pp 1865–1870. study of lifetime load effect distribution of bridges.
sagging moments. This implies that the Bruls, A, Croce, P & Sanpaolesi, L 1996. ENV1991. Proceedings, 6th International ASRANet Conference
design load model could be optimised to be Part 3: Traffic loads on bridges: Calibration of for Integrating Structural Analysis, Risk and
more cost-efficient for a range of bridges. load models for road bridges. Proceedings, IABSE Reliability, 2–4 July 2012, London.
To study a reliability performance of a Colloquium on the Basis of Design and Actions Holický, M 2009. Reliability Analysis for Structural
critical element, the Kilner Park case study on Structures. Delft, The Netherlands: IABSE, Design. Stellenbosch: SunMedia.
entailed investigation of a single spine of pp 439–454. Holický, M 2013. Introduction to Probability and
a 20 m twin-spine deck. The results agree Caprani, C C 2012. Calibration of a congestion Statistics for Engineers. Berlin, Heidelberg: Springer.
with the Roosboom case study and satisfac- load model for highway bridges using traffic Holický, M, Retief, J & Wium, J 2010. Partial factors for
tory performance is observed. In fact, the microsimulation. Structural Engineering selected reinforced concrete members: Background
obtained β-values range between 6.2 and International, 22(3): 342–348. to a revision of SANS10100-1. Journal of the South
7.8, and far exceed the target levels βt of Caprani, C C, Belay, A & O’Connor, A J 2003. African Institution of Civil Engineering, 52(1): 37–44.
3.8 and 3.0 as required by the Eurocode Site-specific probabilistic load modelling for Holický, M & Sýkora, M 2012. Conventional
and the SANS codes, respectively. A revi- bridge reliability. Proceedings, 3rd International probabilistic models for calibration of codes.
sion of the load model could therefore lead Conference on Current and Future Trends in In Faber, M, Koehler, J & Nishijima, K (Eds).
to potential cost savings. To extend the Bridge Design, 28–30 April 2003, Shanghai, China, Applications of Statistics and Probability in
background information for such revision, pp 341–348. Civil Engineering, Boca Raton, FL: CRC Press,
it is recommended to verify the findings Caprani, C C, O’Brien, E J & McLachlan, G J 2008. pp 969–976.
of this study by studies based on WIM Characteristic traffic load effects from a mixture ISO (International Organisation for Standardisation)
data from other heavy-freight routes in of loading events on short to medium span bridges. 2015. ISO 2394:2015. General Principles on
South Africa. It should be further verified Structural Safety, 30(5): 394–404. Reliability for Structures. Geneva, Switzerland: ISO.
whether similar conclusions can be drawn Castillo, E 1988. Extreme Value Theory in Engineering. Ivanković, A, Skokandić, D, Žnidarič, A & Kreslin, M
for hogging moments as expected. For Boston, MA: Academic Press. 2019. Bridge performance indicators based on traffic
shear, particularly at shorter spans and Castillo, E, Hadi, A S, Balakrishnan, N & Sarabia, J M load monitoring. Structure and Infrastructure
local verifications, the detailed analysis of 2005. Extreme Value and Related Models with Engineering, 15(7): 899–911.
axle loads is needed and lower reliability Applications in Engineering and Science. New York: Jacob, B & Feypell-de La Beaumelle, V 2010. Improving
indices may be obtained. Wiley. truck safety: Potential of weigh-in-motion
Coles, S 2001. An Introduction to Statistical Modelling technology. IATSS Research, 34(1): 9–15.
of Extreme Values. London: Springer. Jacob, B, O’Brien, E J & Jehaes, S 2002. COS 323.
ACKNOWLEDGEMENTS Crespo-Minguillón, C & Casas, J R 1997. A Weigh-in-motion of road vehicles. Final report. Paris:
This study has been supported by the comprehensive traffic load model for bridge safety Laboratoire Central des Ponts et Chaussées.
Wilhelm Frank Trust in South Africa, and checking. Structural Safety, 19(4): 339–359. JCSS (Joint Committee on Structural Safety)
partly by the Ministry of Education, Youth CSRA (Committee of State Road Authorities) 1998. 2001. Probabilistic Model Codes. JCSS Working
and Sports of the Czech Republic under TMH7. Code of Practice for the Design of Highway Document. http://www.jcss.ethz.ch
Grant LTT18003, and by the Czech Science Bridges and Culverts in South Africa. Parts 1 and 2. Lenner, R & Sýkora, M 2016. Partial factors for loads
Foundation under Grant 20-01781S. The Pretoria: Department of Transport. due to special vehicles on road bridges. Engineering
authors thank the N3TC for access to the De Wet, D P G 2010. Post-calibration and quality Structures, 106: 137–146.
Roosboom WIM data. and Bakwena conces- management of weigh-in-motion traffic data. MEng Lenner, R & Sýkora, M 2017. Partial factors for
sion for use of the Kilner Park WIM data. Dissertation. Stellenbosch University. imposed loads in areas for storage and industrial
DOT (Department of Transport) 2013. National use. Structure and Infrastructure Engineering,
Transport Master Plan (NATMAP) 2050. Chapter 7: 13(11): 1425–1436.
REFERENCES Freight Transport. Pretoria: DOT. Lenner, R, De Wet, D P G &Viljoen, C 2017. Traffic
Allaix, D L 2007. Bridge reliability analysis with an EN (European Standard) 2002. EN 1990:2002. characteristics and bridge loading in South Africa.
up-to-date traffic load model. PhD Thesis. Turin, Eurocode – Basis of Structural Design. Brussels: Journal of the South African Institution of Civil
Italy: Politecnico di Torino. European Committee for Standardisation. Engineering, 59(4): 34–46.
Anderson, J R 2006. Review of South African live load Enright, B 2010. Simulation of traffic loading on Ma, R, Xu, S, Wang, D & Chen, A 2018. Vehicle models
models for traffic loading on bridge and culvert highway bridges. PhD Thesis. Dublin, Ireland: for fatigue loading on steel box-girder bridges
structures using Weigh-in-Motion (WIM) data. University College Dublin. based on weigh-in-motion data. Structure and
MEng Dissertation. University of Cape Town. fib Bulletin 80 2016. Partial factor methods for Infrastructure Engineering, 14(6): 701–713.
Ang A & Tang, W 2006. Probability Concepts in existing concrete structures. Lausanne, Switzerland: Matos, J C, Moreira, V N, Valente, I B, Cruz, P J,
Engineering: Emphasis on Applications to Civil International Federation for Structural Concrete Neves, L C & Galvão, N 2019. Probabilistic-based
and Environmental Engineering. 2nd ed. New York: (fib). assessment of existing steel-concrete composite
Wiley. Fisher, R A 2006. Statistical Methods for Research bridges: Application to Sousa River Bridge.
Basson, S E 2020. An investigation into the reliability Workers. Edinburgh: Genesis Publishing. Engineering Structures, 181: 95–110.
performance of bridges designed according to TMH7. Getachew, A & O’Brien, E J 2007. Simplified site- Morales-Nápoles O & Steenbergen R 2014. Analysis of
MEng Dissertation. Stellenbosch University. specific traffic load models for bridge assessment. axle and vehicle load properties through Bayesian
Basson, S E & Lenner, R 2019. Reliability verification of Structure and Infrastructure Engineering, networks based on weigh-in-motion data. Reliability
bridges designed according to TMH7. Proceedings, 3(4): 303–311. Engineering and System Safety, 125: 153-164.
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 35
Nowak, A S 1993. Load model for highway bridges. SANRAL (South African National Roads Agency Towards an Integrated Vision. Proceedings of the
Structural Safety, 13(1–2): 53–66. Limited) 2010. Table 6000/1 Project document: 6th International Symposium on Life-Cycle Civil
Nowak, A S & Rakoczy, P 2013. WIM-based live load Project specifications. Pretoria: SANRAL. Engineering (2018), Ghent, Belgium: IALCCE,
for bridges. KSCE Journal of Civil Engineering, SANRAL (South African National Roads Agency pp 2837–2845.
17(3): 568–574. Limited) 2012. SANRAL Structural Drawings. Van der Spuy, P 2020. Derivation of a traffic load
O’Brien, E J, Rattigan, P, González, A, Dowling, J & Pretoria: SANRAL. https://www.nra.co.za/live/ model for the structural design of highway bridges in
Žnidarič, A 2009. Characteristic dynamic traffic content.php?Category_ID=206. South Africa. PhD Thesis. Stellenbosch University.
load effects in bridges. Engineering Structures, SANS (South African National Standard) 2019. SANS Van der Spuy, P & Lenner, R 2018. Developing a
31(7): 1607–1612. 10160-1:2019. Basis of Structural Design and Actions new bridge live load model for South Africa.
O’Brien, E J, Schmidt, F, Hajializadeh, D et al 2015. A for Buildings and Industrial Structures. Pretoria: Proceedings, 9th International Conference on Bridge
review of probabilistic methods of assessment of SABS Standards Division. Maintenance, Safety and Management (IABMAS),
load effects in bridges. Structural Safety, 53: 44–56. Soriano, M, Casas, J R & Ghosn, M 2016. Simplified 9–13 July 2018, Melbourne, pp 1405–1410.
O’Connor, A & Enevoldsen, I 2007. Probability-based probabilistic model for maximum traffic load from Van der Spuy, P & Lenner, R 2019a. Towards a new
bridge assessment. Proceedings of the Institution of weigh-in-motion data. Structure and Infrastructure bridge live load model for South Africa. Structural
Civil Engineers – Bridge Engineering, 160: 129–137. Engineering, 13(4): 454–467. Engineering International, 29(2): 292–298.
O’Connor, A J & O’Brien, E J 2005. Traffic load Steenbergen, R D & Vrouwenvelder, A C 2010. Van der Spuy, P, Lenner, R, De Wet, T & Caprani, C C
modelling and factors influencing the accuracy Safety philosophy for existing structures and 2019b. Multiple lane reduction factors based on
of predicted extremes. Canadian Journal of Civil partial factors for traffic loads on bridges. Heron, multiple lane weigh in motion data. Structures,
Engineering, 32(1): 270–278. 55(2): 123–140. 20: 543–549.
Oosthuizen, A P, Meintjies, C J, Trumpelmann, V, Sýkora, M, Holický, M, Prieto, M & Tanner, P 2015. Van der Spuy, P, Lenner, R & Meyer, M 2019c.
Peters, D, Ullmann, K K & Oppermann, G H 1991. Uncertainties in resistance models for sound and Dynamic amplification factor for South African
TMH7 Part 2: Traffic Loading (1991). Proposed corrosion-damaged RC structures according to EN bridges. Proceedings, 7th International Conference
Substitution of Section 2.6. Pretoria: Committee of 1992-1-1. Materials and Structures/Materiaux et on Structural Engineering, Mechanics and
State Road Authorities, Department of Transport. Constructions, 48(10): 3415–3430. Computation, 2–4 September, Cape Town,
Pérez, S & Lenner, R 2019. Bridge reduction factors Tabsh, S W & Nowak, A S 1991. Reliability of highway pp 1859–1864.
based on Monte Carlo routine with copulas. girder bridges. Journal of Structural Engineering, Von Scholten, C, Enevoldsen, I, Arnbjerg-Nielsen, T et
Engineering Structures, 198. doi.org/10.1016/j. 117(8): 2372–2388. al 2004. Reliability-based classification of the load
engstruct.2019.109530 Teichgräber, M, Nowak, M, Köhler, J & Straub, D carrying capacity of existing bridges. Copenhagen:
R Core Team 2018. R: A language and environment for 2019. The effect of traffic load model assumptions Danish Ministry of Transport, Road Directorate.
statistical computing, version 3.5.0. Vienna, Austria: on the reliability of road bridges. In Life-Cycle
R Foundation for Statistical Computing. Analysis and Assessment in Civil Engineering:
36 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
The variability in commercial TECHNICAL PAPER
laboratory aggregate Journal of the South African
Institution of Civil Engineering
Breytenbach IJ, Fourie HG. The variability in commercial laboratory aggregate testing for road construction in South Africa.
J. S. Afr. Inst. Civ. Eng. 2021:63(1), Art. #1038, 8 pages. http://dx.doi.org/10.17159/2309-8775/2021/v63n1a4 37
aggregate material which prematurely fails same sample to multiple laboratories for or SANRAL (2013). After reviewing the
in service. comparison, this study considers multiple available project data, which included
Properties of soil – and in this case samples tested in duplicate only, and as duplicate test values, the following tests
aggregate – materials are affected by a num- such a detailed statistical analysis on vari- were selected for comparison:
ber of aspects, including the inherent nature ance between different laboratories’ results QQ Polished Stone Value (SANS 5848)
of the material itself, sampling disturbances, is not possible due to limited data. Instead, QQ Aggregate Crushing Value (SANS 3001:
a measurement error and statistical uncer- the discussion intends to highlight the high AG10) – dry test
tainty associated with the number of tests degree of variation in test results on indi- QQ Aggregate Crushing Value (SANS 3001:
performed (Juang et al 2019). Jacobsz and vidual samples tested in duplicate, despite AG10) – wet test
Day (2008) highlighted the variability in the participating laboratories being SANAS- QQ 10% Fines Aggregate Crushing Value
results of Atterberg Limits reported on the accredited for the tests performed, with the (SANS 3001: AG10) – dry test
same sample by different commercial geo- exception of the polished stone value (PSV) QQ 10% Fines Aggregate Crushing Value
technical laboratories in South Africa and for which none of the laboratories had (SANS 3001: AG10) – wet test
the implications of using erroneous results SANAS accreditation. QQ 10% Fines Aggregate Crushing Value
in geotechnical design. A similar exercise (SANS 3001: AG15) – ethylene glycol-
was undertaken by Theron et al (2015) who soaked test (basic crystalline materials
considered the results of hydrometer tests METHODOLOGY only)
performed by multiple laboratories on the Data was captured from existing projects QQ Water Absorption of the +5.00 mm
same sample materials – the outcome of the commissioned by SANRAL tasked with sample fraction (SANS 3001: AG20)
comparison was described as troublesome. identifying and proving sources of suitable QQ Water Absorption of the –5.00 mm
Another example is given by Takács and hard rock aggregate materials to be quar- sample fraction (SANS 3001: AG21)
Mahler (2011) who compared results pro- ried for various purposes (e.g. base course,
duced by 63 laboratories in Hungary, testing surfacing stone, etc). During the course of Material Sample Selection
more than 500 samples on eight separate the projects, duplicate tests were performed Following the identification of preferred test
sample groups. on selected samples, and only certain methods, suitable project result data had
However, while literature on the reli- critical tests or samples were duplicated. to be selected. The selected test data was
ability or reproducibility of soil materials is Samples were produced from drill core sourced from fourteen project sites on vari-
fairly readily available, similar discussions retrieved during intrusive drilling, which ous geological bedrock types. The tests were
on aggregate material tests were found was grouped into batched samples and performed by a total of seven different com-
to be extremely limited, especially in the crushed using a laboratory crusher. Test mercial laboratories, all of which are SANAS-
South African environment, and more duplication was to be done by splitting the accredited for the tests under consideration,
specifically for South African aggregate nominated sample after sample prepara- except the PSV test. The geological composi-
tests. Moreover, literature related to tion. One half of the sample was tested by tion of the data set is summarised in Table 1.
aggregate tests mainly focus on concrete the primary laboratory (i.e. primary tests)
production – for example Grieve (2002) – while the second half was sent to a second, Statistical Analyses
and not on road construction aggregate, independent laboratory (i.e. duplicate tests) Data sets were entered into a spreadsheet
despite some overlapping tests between the to perform the same test, using the same and each material test was used for a sepa-
two applications. Dumas (2000) compiled standardised test method. The reasoning rate analysis. The test result data for each
a proposed appendix to the Technical behind this approach was that by splitting test method was used to perform a general
Methods for Highways (TMH1) series the sample, the test material should be statistical analysis (i.e. descriptive statistics)
widely used in South Africa. The document practically identical and therefore compar- which was supplemented by a simple cor-
comprehensively considers the reliability ing test results directly should be possible. relation analysis between the primary and
of test results and, amongst other aspects, Preference for duplicate testing was duplicate test results in order to determine
takes into account the statistical signifi- given to tests which have lower or upper the correlation coefficient. The results of
cance of test results and their variability. limit specifications in COLTO (1998) and/ the statistical analyses will not be discussed
The TMH1 series of tests was replaced
by SANS 3001, which includes a series Table 1 Number of data sets
on the determination of uncertainty of Geological material
measurement, repeatability, reproducibility
Test Granite/ Total
and bias, i.e. SANS 3001-PR1 (SANS 2010), Dolerite Tillite Quartzite
Gneiss
as well as an assessment of repeat, check
PSV 15 17 2 6 40
or duplicate tests, i.e. SANS 3001-PR2
(SANS 2011). The latter standard provides ACV (dry) 15 18 3 6 42
guidelines to assess duplicate tests by cal- ACV (wet) 15 18 3 6 42
culating the permissible range between the 10% FACT (dry) 15 18 3 6 42
primary and duplicate test results. 10% FACT (wet) 15 18 3 6 42
The approach followed in this compari-
10% FACT (ethylene glycol) 15 15
son differs from that followed by Jacobsz
Water Absorption (+5.00 mm) 15 17 3 6 41
and Day (2008) and Theron et al (2015)
Water Absorption (–5.00 mm) 15 17 3 6 41
in that where these authors submitted the
38 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
in depth, but relevant aspects will be high-
lighted in the following sections. 80
Lab 1
75 Lab 2
Direct Simple Comparison Lab 3
In order to directly compare results of 70 Lab 4
primary and duplicate tests, it was decided Lab 5
that the simplest and most direct method 65
would be to illustrate the results graphically.
PSV
Scatter plots were prepared for each type of 60
test used, plotting the sample number on the
55
x-axis and the two test results (i.e. primary
and duplicate) on the y-axis. In this manner 50
the primary and duplicate test results can
be compared easily. The scatter plots were 45
further refined to distinguish between the
seven participating laboratories that supplied 40
0 5 10 15 20 25 30 35 40
results. Results were then further assessed Sample number
according to SANS 3001 PR2 (SANS 2011) in
order to assess the variability in the results. Figure 1 Polished Stone Value comparison
RESULTS 25
Lab 1
23 Lab 3
Polished Stone Value (PSV)
Lab 5
The comparative plot of results for the PSV 21
Lab 6
tests is illustrated in Figure 1. From the
19 Lab 7
graph it is clear that the correlation between
the primary and duplicate test results is 17
ACV (%)
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 39
Aggregate Crushing Value (ACV)
500 The comparative scatter plots for the dry
and wet ACV tests are shown in Figures 2
450
and 3, respectively.
400
The results of the dry ACV test had a
correlation coefficient of 0.50, which is a
350 notable improvement compared to the PSV
10% FACT (kN)
40 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
data sets, as only basic crystalline materials
were subjected to this test and hence sam- 1.4
Lab 1
ple numbers for Figure 6 do not correspond
Lab 3
to those of Figures 4 and 5. 1.2
Lab 5
The data for the dry tests shows very Lab 6
poor correlation, with discrepancies 1.0 Lab 7
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 41
PSV ACV Dry (%)
80 24
75 22
Secondary test results
22 450
Secondary test results
18 350
16 300
14 250
12 200
10 150
9 14 19 24 250 300 350 400 450 500
Primary test results Primary test results
350
400
Secondary test results
300 R2 = 0.3411
350
250
300
200
250 R2 = 0.6698
150
100 200
150 200 250 300 350 400 450 200 250 300 350 400 450
Primary test results Primary test results
1.4
0.6
1.2
0.5 1.0 R2 = 0.5919
R2 = 0.0651
0.8
0.4
0.6
0.3
0.4
0.2 0.2
0.2 0.4 0.6 0.8 1.0 1.2 1.4 0.2 0.7 1.2 1.7
Primary test results Primary test results
Figure 9 P
rimary (x-axis) versus secondary (y-axis) test results
42 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
Lab 6 was ultimately identified to be a devia- results is not only of theoretical signifi- problems with test results which could be
tion from the test method by an inexperi- cance. One should consider the practical addressed timeously and without incurring
enced technician. This in itself highlights and financial implications of using only a potentially severe financial repercussions
the importance of technician training and single set of unverified and possibly errone- and project delays. Coincidentally, dupli-
certification to perform specific tests. ous test results. To illustrate the point, the cate testing identified at least three serious
Despite the much-improved correlation following two scenarios can be considered: discrepancies in project data used for the
coefficient of the water absorption for the QQ Scenario 1: A site is investigated as analyses reported here and allowed the
fine fraction data (i.e. 0.79), results still potential hard rock quarry; however, source of the discrepancies to be identified
compare relatively poorly. Primary and unverified test results inaccurately and addressed timeously. The cost of dupli-
duplicate test results were identical on only report lower than true values which cate testing amounts to only a fraction of
three occasions, while the largest single happen to fall below the minimum the cost of repairing or rebuilding a failed
difference in results was reported as 0.9% material specification. The site is dis- road section or repeating the investigations
(i.e. sample 10). Applying the principles of qualified from use, and the funds and for an alternative material source.
SANS 3001 PR2 (SANS 2011) once more, time utilised investigating the site are The cost of periodic maintenance (e.g.
30 of 41 test results (i.e. 73%) exceeded the wasted. Even worse, additional funds resealing) for a single carriageway road
permissible range, rendering those results and time are required to investigate with an assumed width of ten metres
“unacceptable”. and prove a new site, while the original typically ranges from R2.0 million to R3.0
In this data set seven samples conflicted site was in fact suitable, and the pro- million per kilometre, as summarised by
with regard to passing or failing the maxi- ject is delayed with contractual and SANRAL’s Annual Integrated Report for
mum water absorption of 1.5% allowed cost implications. 2019 (SANRAL 2019) and indicated in
by the COLTO (1998) specification, while QQ Scenario 2: A site is investigated as Table 2. Similarly, the cost of rehabilitation
one sample was confirmed to fail by both potential hard rock quarry. The unveri- or partial reconstruction would be at least
primary and duplicate tests. As before, the fied laboratory analyses inaccurately ten times the cost of resurfacing. The total
degree of variation for such a relatively sim- overestimate test results and report cost of all duplicate tests considered in this
ple test is a serious concern. For example, them to be within specification when, research comparison amounted to less than
sample 10 reported a primary result of 1.6% in fact, they fall below specification. R1.0 million, which is negligible compared
water absorption while the duplicate result The source is subsequently approved to the high cost of road construction and
reported only 0.7%. Though this is an for material production and is used maintenance. Therefore, insisting on dupli-
extreme example, it illustrates the problem in construction, but ultimately the cate tests to ensure reliable test results,
well in that one result is less than half of aggregate material fails prematurely in will recoup the costs of the duplicate tests
the other. service because of its inferior quality. within the first kilometre of a typical reseal
Costs are incurred to identify a new project if scenarios 1 or 2, as mentioned
suitable source of aggregate and the earlier, can be avoided.
DISCUSSION road must be repaired, or worse, rebuilt In cases where large variations occur
In order to further illustrate the difference at great expense. between primary and duplicate test results,
between primary and duplicate test results The scenarios above hold substantial prac- additional tests can be performed to deter-
concisely, Figure 9 shows direct compari- tical and financial risk for virtually all par- mine – with confidence – which set of test
sons between the test results. From the ties involved; however, the second scenario results is reliable and reproducible. In this
regression coefficients (R2) it is clear that clearly holds the largest risk as the implica- manner, a laboratory can also make sure
there are large shortcomings in reliability. tion of premature pavement failure is dire. that equipment and reporting techniques
The significance of the poor relationship Regardless of the scenario, duplicate testing are as they should be. This, however, has
between the primary and duplicate test could be used to identify inaccuracies or practical problems, too, as funds have
Table 2 Typical costs for road maintenance and construction (SANRAL 2019)
Type of road project Frequency Cost
Capital projects
Strengthening: Increases structural capacity of pavement through recycling of existing layers or addition
Every 20 – 25 years R2m – R3m per kilometre
of new granular layers or structural asphalt layers
Improvement: Improves road with unacceptable quality of service. Examples: addition of passing lanes,
Every 20 – 25 years R12m – R18m per kilometre
addition of paved shoulders, improved intersections. Is often combined with rehabilitation of pavement
New facilities: Improve capacity to accommodate traffic and expand road network. Examples: greenfields
roads, upgrades from single to dual carriageway, new bridges, replacement of intersections with Every 20 – 25 years R20m – R40m per kilometre
interchanges
Road maintenance
Routine maintenance: Cleaning drains and culverts, cutting vegetation, repainting road markings,
Ongoing R0.1m a kilometre per year
repairing guard rails and signs, patching and sealing cracks
Periodic maintenance: Scheduled waterproofing of roads by application of surface seals and thin
8 – 12 years R2m – R3m per kilometre
functional asphalt layers
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 43
finite limits and verification testing cannot respective test. Large discrepancies in test document as an appendix to the TMH series.
be conducted on an indefinite number results must be identified, and where neces- Unpublished document.
of samples until satisfactory (i.e. statisti- sary additional tests of the same sample Grieve, G R H 2002. Inter-laboratory proficiency
cally acceptable) results are produced. It material must be undertaken to identify evaluations between several Gauteng-based
is further time-consuming and project which test result set is in fact correct, and concrete and aggregate testing laboratories. Journal
timelines do not make provision for also to back-trace the origin of the errone- of the South African Institution of Civil Engineering,
additional testing. ous results to establish the root cause. As 44(1): 31–34.
specified in SANS 17025 (SANS 17025 2018), Jacobsz, S W & Day, P 2008. Are we getting what
the responsibility rests with the laboratory we pay for from geotechnical laboratories? Civil
CONCLUSION AND to undertake a root cause analysis. Engineering, 16(4): 8–11.
RECOMMENDATIONS Estimation of uncertainty of measure- Juang, C H, Zhang, J, Shen, M & Hu, J 2019.
Actual project data was presented as ment for soils and gravels is better defined, Probabilistic methods for unified treatment of
proof that confirmation test analyses are probably due to the high frequency and geotechnical and geological uncertainties in
required in order to verify material test volume of testing concerned with these a geotechnical analysis. Engineering Geology,
results. While minor discrepancies are materials. However, in the case of aggregate 249: 148–161.
inevitable and test results are practically testing, further research is required to SANAS (South African National Accreditation System)
unlikely to match precisely under real- determine the maximum systematic error 2016a. Regulatory document R-48-04: Proficiency
world conditions, variations were often not to enable laboratories to identify erroneous Testing and other Comparison Programme
acceptable. In this case, data from fourteen test results. This is not possible without Requirements for Calibration Laboratories. Pretoria:
high-quality hard rock quarry projects was verification testing. Uncertainty of meas- SANAS.
used where duplicate tests were performed. urement is often underestimated or even SANAS (South African National Accreditation System)
A simple comparison between specifica- ignored by design engineers, technologists 2016b. Regulatory document R-80-03: Proficiency
tion test results clearly illustrated that the and technicians when assessing material Testing and other Comparison Programme
relationship between primary and duplicate test results. Requirements for Testing and Medical Laboratories
test results is often inconsistent and that Compared to the cost of repairing or and Blood Transfusion Services. Pretoria: SANAS.
the test results are therefore treated with rebuilding a road section due to premature SANRAL (South African National Roads Agency
scepticism, despite all participating labora- failure ascribed to the use of inferior Limited) 2013. SAPEM 2013. South African
tories being SANAS-accredited and using aggregate material, the additional cost of Pavement Engineering Manual. Chapter 4:
standardised test methods. In this case duplicate tests is negligible. Seen in a bigger Standards. Pretoria: SANRAL.
study, material sources were of a high qual- context, the additional cost of duplicate SANRAL (South African National Roads Agency
ity and hence the majority of test results test analyses can massively reduce the Limited) 2019. SANRAL Integrated Report
fall well within material specification, financial and practical risk to any road 2018/2019. Volume 1. Pretoria: SANRAL.
despite frequent large variations. However, construction project. SANS (South African National Standard) 2010.
the implications of unreliable test results SANS 3001 – PR1 (2010). Part PR1: Determination
will become more severe where materials of Uncertainty of Measurement, Repeatability,
are within closer bounds (i.e. borderline ACKNOWLEDGEMENT Reproducibility and Bias. Pretoria: SABS Standards
cases) to material specification limits. The authors acknowledge the South Division.
It is also disconcerting that technical African Roads Agency SOC Limited for SANS (South African National Standard) 2011. SANS
signatories of the commercial laboratories permission to use selective project data. 3001 – PR2 (2011). Part PR2: Use and Assessment of
involved with the projects used in this Repeat, Check or Duplicate Tests. Pretoria: SABS.
comparison issued material test results – SANS (South African National Standard) 2018.
which reflected both the primary and DISCLAIMER SANS 17025:2018. General Requirements for the
duplicate test sets on one document – The views expressed by authors are in Competence of Testing and Calibration Laboratories.
without revisiting results that were clearly their personal capacity and do not reflect Pretoria: SABS Standards Division.
highly variable or questionable. SANRAL’s viewpoint or policy. Takács, A & Mahler, A 2011. Statistical evaluation
Based on the findings of this comparison of geotechnical laboratory round robin tests in
it is recommended that material test sched- Hungary. Proceedings, 15th European Conference
ules for hard rock quarry material investiga- REFERENCES on Soil Mechanics and Geotechnical Engineering,
tions also make provision for (at least limit- COLTO (Committee of Land Transport Officials) Athens, Greece, Vol 1, pp 293–297.
ed) duplicate testing as a measure of quality 1998. Standard Specifications for Road and Bridge Theron, E, Badenhorst, W & Stott, P 2015. Duplicate
control. The principles of SANS 3001 PR1 Works for State Road Authorities. Halfway House: testing conducted on the input parameters for
(SANS 2010) and SANS PR2 (SANS 2011) South African Institution of Civil Engineering the estimation of potential expansiveness of clay.
must be applied to determine if the range (SAICE). Proceedings, 16th African Regional Conference on
between primary and duplicate test Dumas, B (Provincial Administration, Western Cape) Soil Mechanics and Geotechnical Engineering, 27
results are within acceptable limits for the 2000. General technical information: Proposed April 2015, Hammamet, Tunisia.
44 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
Review of compatibility TECHNICAL PAPER
between SANS 10400 Journal of the South African
Institution of Civil Engineering
deemed-to-satisfy ISSN 1021-2019
Vol 63 No 1, March 2021, Pages 45–60, Paper 1062
in South Africa is regulated by the Design of masonry structures (performance- Contact details:
Application of the National Building based). SANS 10400 is performance-based Address at the time of preparing this paper:
Department of Civil Engineering, Stellenbosch University
Regulations, based on the National in nature but contains extensive deemed-to-
Private Bag X1, Matieland 7602, Stellenbosch, South Africa
Building Regulations and Standards Act of satisfy solutions, the typical mixed approach Current address:
1977 (RSA 1977). The standard was first taken in transitioning from prescriptive to Faculty of Engineering, Built Environment and Information Technology
published in 1985 but has since been updat- performance-based regulation. University of Pretoria
Private Bag X20, Hatfield 0028, South Africa
ed several times to the current edition A number of developments over the T: +27 12 420 2746, E: billy.boshoff@up.ac.za
SANS 10400 (SANS 2010a). On a practical past decade or two warrant a reconsidera-
level, all housing construction in South tion of these deemed-to-satisfy solutions
Africa is regulated by the National Home in SANS 10400, specifically with regard to
Builders Registration Council (NHBRC), masonry walling solutions:
the establishment of which is enshrined QQ Loading: The South African loading
Keywords: low-income housing, National Building Regulations,
in The Housing Consumers Protection code SANS 10160 2011 (SANS 2011) has SANS 10400, South African loading code, concrete
Measures Act (RSA 1995; NHBRC 2015). been revised, in the form of an adaption masonry, simplified micro-model
De Villiers WI, Van Zijl GPAG, Boshoff WP. Review of compatibility between SANS 10400 deemed-to-satisfy masonry wall provisions and loading code.
J. S. Afr. Inst. Civ. Eng. 2021:63(1), Art. #1062, 16 pages. http://dx.doi.org/10.17159/2309-8775/2021/v63n1a5 45
(b)
Modelling approach
Significant advances in numerical methods
and computational capabilities in recent
decades have altered the way in which
masonry is analysed. For masonry finite
element modelling, two main approaches
have been established, namely macro- and
micro-modelling, with the level of abstrac-
tion directly related to the complexity
(a) and size of the problem to be analysed,
(Giambanco et al 2001; Reyes et al 2008;
Figure 1 T ypical 40 m2 government subsidised concrete masonry house: (a) plan (CMA 2011, Roca et al 2010; Abdulla et al 2017).
reproduced with the kind permission of the CMA) and (b) under construction Macro-modelling assumes a smeared
continuum approach, where the unit, mortar
of EN 1990 (EN 1990) and EN 1991 (EN be taken into consideration in the speci- and unit-mortar interface behaviours are
1991), with notable changes and addi- fications for masonry. combined in a representative continuous
tions. The design of single-storey mason- QQ Outdated mechanical limits: Current material. In contrast, micro-modelling repre-
ry structures must take seismic loading mechanical limits set in the South sents a high degree of detail where the unit,
into account more comprehensively in African prescriptive standards of mortar and unit-mortar interface are mod-
certain areas of the country, and signifi- conventional masonry units are largely elled distinctly. Simplified micro-modelling
cant improvements have been made to based on yield line theory analysis (JSD (SMM) is a subset of micro-modelling as its
the South African wind data map. 1995), taken from the withdrawn British name implies, wherein the units are modelled
QQ Category 1 buildings: The Application Standard BS 5628-1 (BS 1978). as expanded elements, with solely elastic
of the National Building Regulations This paper therefore investigates the material properties, to encompass the volume
(SANS 2010b) has been revised with response of conventional concrete masonry of the unit and the mortar in order to main-
significant changes, including the intro- walls in the context of South African low- tain the overall geometry. The relatively weak
duction of Category 1 buildings, specifi- income housing (LIH) by means of finite mortar joint and unit-mortar interface are
cally aimed at drawing more low-income element (FE) analysis. The analyses are combined into a single zero-thickness inter-
structures into a regulatory framework. performed on single-storey, unreinforced, face element in which the nonlinear material
QQ Adoption of Eurocode 6: The South single-leaf, external masonry walls, which behaviour is concentrated (Lourenço 1996).
African masonry industry is in the conform to the deemed-to-satisfy solutions The SMM approach is considered the most
process of updating its suite of masonry of SANS 10400. Two critical wall layouts appropriate for this study in terms of the
standards to the EN approach, Eurocode are identified (W1 and W2), modelled in wall scale to be analysed and the computing
6 Design of Masonry Structures, mark- DIANA FE analysis software and subjected requirements (Figure 2 refers).
ing a transition from prescriptive to to three load conditions as required by SANS
performance-based standards. 10160: the serviceability limit state (SLS) and Constituent material model
QQ Advances in numerical analyses: A the ultimate limit state for wind (ULS-W) The constitutive material model chosen is
significant amount of research has been and seismic (ULS-S) actions. The results of the Combined Cracking-Shearing-Crushing
conducted internationally, using finite the analyses provide insight into the in-plane (CCSC) model, implemented in DIANA. The
element modelling, to better understand and out-of-plane structural behaviour of plasticity-based model is defined by a multi-
the discontinuous behaviour of masonry conventional concrete masonry walls of LIH surface yield function, shown in Figure 3,
structures, and these advances need to housing, relative to the expected loading. consisting of a tension cut-off, a Coulomb
46 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
hhmm
hhu
u 16-noded
hhm 16-noded
m interface
interface
element
elementfor for
mortar
mortarjoints
joints
Interface
interface Potentialcrack
elements potential crack
elements (joints)
(joints) ininthe
theunit
unit
Zerothickness
zero thickness
+h
h u+h
u mm
20-noded solid
20-noded solid
element for
element for bricks
bricks
Continuum elements (units)
continuum
(a) elements (units) (b)
Figure 2 S MM approach: (a) in 2D with expanded unit elements (adapted from Lourenço & Rots 1997) and (b) in 3D with solid brick elements and
2D interface elements (adapted from Macorini & Izzuddin 2011)
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 47
6 000 mm
Secondary: 0 5.6 mm × 2
756 mm
Primary: 0 5.6 mm × 3
(equivalent of)
1 008 mm
2 646 mm
Brickforce: 0 2.8 mm × 2
Brickforce: 0 2.8 mm × 2
882 mm
Brickforce: 0 2.8 mm × 2
750 mm 900 mm 600 mm 3 000 mm 750 mm
630 mm 630 mm
Secondary: 0 5.6 mm × 2
Primary: 0 5.6 mm × 2
1 010 mm
(equivalent of)
2 650 mm
Brickforce: 0 2.8 mm × 2
1 010 mm
Brickforce: 0 2.8 mm × 2
48 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
Figure 6 Boundary conditions for wall W1 (left) and W2 (right), inner perspective
walls provide lateral support to the modelled Table 2 Input parameters for wall numerical analyses
walls, whilst allowing for some rotation of Parameter Symbol DIANA Method Value Unit
these joints. It is assumed that the roof truss
system does not provide substantial lateral Unit
load transfer, based on the type of connec- Density ρu – EXP 2 090 kg/m3
tion and the typical poor quality of the con- E-modulus Eu YOUNG EXP 17 700 N/mm2
nection between the roof and walling in LIH.
Therefore the roof line is modelled as unsup- Poisson’s ratio νu POISSON LIT 0.16 –
MATERIAL INPUT PARAMETERS Mode I fracture energy GIf,c GF EXP 0.047 N/mm
together with 10 mm mortar joints. The Mode II fracture energy GIIf,c MO2VAL LIT 0.47 N/mm
blocks have a length of 290 mm, width of
Compressive strength fc,c COMSTR EXP 12.1 N/mm2
140 mm and height of 116 mm. Extensive
material parameter characterisation is Shear traction contrib Css,c CS LIT 1.0x10 -3 –
required for the selected constituent mate- Compr fracture energy Gc,c GC LIT 10.0 N/mm
rial model, CCSC, that was selected in
Eq plastic relative displ κp,c DUPEAK LIT 0.030 mm/mm
DIANA. Table 2 contains the material input
parameters that were used to define the Tangential stiffness ks,c DSSX/Y LIT 763x10 -3 N/mm3
concrete masonry, as well as the method by Normal stiffness kn,c DSNZ LIT 1770x10 -3 N/mm3
which they were determined. Experiments
Joint interface
were conducted by Fourie (2017) to deter-
mine several of the unit, crack and joint Tensile strength f t,j TENSTR LIT 0.12 N/mm2
interface parameters, indicated by EXP Mode I fracture energy GIf,j GF LIT 0.005 N/mm
in Table 2 under Method. Finite element
Cohesion cj COHESI EXP 0.17 N/mm2
analysis (FEA) was used to determine the
tensile strength of the crack interface and Friction angle φj PHI EXP 49.5 °
the compressive fracture energy and equiva- Dilatancy coefficient ψj PSI LIT 0 °
lent plastic relative displacement of the
Mode II fracture energy GII f,j MO2VAL LIT 0.05-0.08σ N/mm
joint interface by numerically fitting data to
experimental data. Suitable literary sources Compressive strength fc,j COMSTR EXP 5.5 N/mm2
(LIT) were used to determine the remain- Shear traction contrib Css,j CS LIT 0.7 –
ing parameters. For further details on the
Compr fracture energy Gc,j GC FEA 18.0 N/mm
material input parameters or the process to
obtain them see De Villiers et al (2018). Eq plastic relative displ κp,j DUPEAK FEA 0.030 mm/mm
Characteristic values of the current Tangential stiffness ks,j DSSX/Y LIT 214 N/mm3
material input data cannot be established
Normal stiffness kn,j DSNZ LIT 520 N/mm3
since the data is statistically insufficient.
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 49
The use of nominal values is allowed in Table 3 Critical design loads for SLS, ULS-W and ULS-S to SANS 10160
such instances according to SANS 10160-1 Roof OP
(SANS 2018a), and it requires the use of (N/mm2) IP
Self-Weight Wind Zone A Zone B
mean values for the parameters govern-
ing the structural stiffness. The choice of SLS W1–44 –10.1 × 10 –3 43.9 × 10 –3 1.4 × 10 –3 1.1 × 10 –3 24.4 × 10 –3
input parameters, as well as the modelling
SLS W1–40 –10.1 × 10 –3 36.3 × 10 –3 1.2 × 10 –3 0.9 × 10 –3 20.2 × 10 –3
approach, was validated experimentally
in a separate process for both the in-plane SLS W1–36 –10.1 × 10 –3 29.4 × 10 –3 0.9 × 10 –3 0.7 × 10 –3 16.3 × 10 –3
Eurocode 6 (Sýkora & Holický 2010), and the ULS-W 0.9 0.0 1.6 –
disparity in classifications, a direct compari-
ULS-S 1.0 0.0 – 1.0
son of the material partial safety factors is
not reasonable. However, in broad terms, the
SANS 51996-1-1 (least conservative) ranges combinations of permanent, imposed, wind as detailed in Table 2. The roof assembly
from 1.5 to 3.0, the UK National Annex to and seismic actions for the two wall con- consists of six bay Howe type trusses,
Eurocode 6 of 2005 (BSI 2005) from 2.3 to figurations under consideration. The three assuming a timber density of 5 000 N/m 3
3.0, and SANS 10164-1 (most conservative) load cases considered are the serviceability according to Table A.4 of SANS 10160-2
from 2.9 to 3.5. limit state (SLS) and the ultimate limit (SANS 2011a) for the structural pine,
Despite their importance, material par- states under wind (ULS-W), based on basic and 0.5 mm metal sheeting with a self-
tial factors are not included in this study. fundamental wind speeds of 44, 40 and weight of 39.5 N/m 2 according to Table
Admittedly, this may result in a less con- 36 m/s respectively, and seismic (ULS-S) A.5 of SANS 10160-2 (SANS 2011a). The
clusive evaluation of the concrete masonry actions. The latter are determined by means roof is classified as an inaccessible roof
walls and the design loads applied to them, of the equivalent lateral static force method. according to Table 5 of SANS 10160-2
which are adjusted with partial factors. Table 3 summarises the factored loads (SANS 2011a) and loads for normal main-
However, including the material partial fac- applied in the numerical analyses for the tenance and repair of 400 N/m 2 would be
tors renders the finite element output less vertical loads on the roof and the horizontal included. However, since an additional
clear, impacting in particular the fracture out-of-plane (OP) and in-plane (IP) loads on compressive load on the walls is favour-
behaviour of the concrete masonry as well the walls, whereas Table 4 details the partial able, the load combination nullifies the
as the investigation into the relative impor- load factors for the three load cases, accord- roof-imposed load.
tance of the parameters. ing to SANS 10160-1 (SANS 2018a).
Wind load
Load assumptions The loads due to wind actions are deter-
DESIGN LOADS The assumptions made in determining mined according to SANS 10160-3 (SANS
SANS 10400-B (SANS 2012) requires these critical load cases are detailed in the 2018b). The pertinent parameters are
the structural strength and stability of following three sub-sections according to summarised in Table 5 and assumptions
the structure to be assessed by loading it self-weight and imposed load, wind load discussed thereafter.
with the relevant actions as determined and seismic load. In most instances, the parameter result-
according to the South African load- ing in the most critical load is selected.
ing code, SANS 10160 (SANS 2011). All Self-weight and imposed load The basic fundamental wind speed is taken
relevant design situations are taken into The self-weight of the walls is based on as the highest value for any area in South
consideration to arrive at the most critical the density determined experimentally, Africa of 44 m/s; however, loads based on
50 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
Table 5 Wind load parameters to SANS 10160-3 (SANS 2018b) Building Code (ICBO 1997) and introduced
Parameter Symbol Value Clause
in the South African loading code to com-
pensate for a lower nominal peak ground
Fundamental value of basic wind speed νb 44 m/s 7.2.2 acceleration of 0.1 g (Wium 2010). However,
Terrain category - C Table 2 the UBC permits a reliability redundancy
factor range of 1.0 to 1.5, compared to
Terrain roughness factor cr(z) 0.73 7.3.2, Table 3
a range of 1.2 to 1.5 in SANS 10160-4
Topography factor c0(z) 1 7.3.3 (SANS 2017). The lower limit of 1.2 was
Air density ρ 1.2 kg/m3 Table 4 set to compensate for the higher behaviour
factors for reinforced concrete shear walls
Peak wind pressure qp(z) 1213 N/m2 7.4, Equation 6
used in the UBC (ICBO 1997) compared to
Eurocode 8 (EN 1998; Wium 2010).
Table 6 Seismic load parameters to SANS 10160-4 (SANS 2017) For determining the seismic design
Parameter Symbol Value Clause
load, the reliability redundancy factor is
chosen as the lower limit of the allowable
Peak ground acceleration ag 0.15 g 5.2, Figure 1 range (1.2 to 1.5), hence less conservative,
Ground type – 4 5.1.2, Figure 2, Table 2 for two reasons. First, a higher peak ground
acceleration of 0.15 g was selected for the
Building importance factor γ1 1.0 7.3, Table 3
analyses, not 0.1 g. Second, the lower limit
Reliability redundancy factor ρ 1.2 7.3, Equation 6 of 1.2 in SANS 10160-4 (SANS 2017) was
Behaviour factor q 1.5 8.2, Table 4 introduced to compensate for the higher
behaviour factors of reinforced concrete
Fundamental period of vibration factor CT 0.05 8.5.2.1
shear walls. This discrepancy in behaviour
factors is less relevant for this study, given
basic fundamental wind speeds of 40 and Seismic load that a consistent behaviour factor for unre-
36 m/s are also included in the results for The loads due to seismic actions are inforced masonry of 1.5 is used. It would
comparative purposes. The terrain cat- determined according to SANS 10160-4 hence be justifiable to use a reliability
egory is chosen as the most likely scenario (SANS 2017). The pertinent parameters are redundancy factor of 1.0. However, compli-
for single-storey residential structures summarised in Table 6 and assumptions ance with SANS 10160-4 (SANS 2017)
in a suburban or peri-urban setting. The discussed thereafter. is considered salient and a factor of 1.2
default topography factor is chosen, on The highest peak ground acceleration is used.
the assumption that it is unlikely that low- for natural seismicity in South Africa is
income housing is developed on extreme selected. The most unfavourable of ground Load application
terrain, which is costly to construct on. types is chosen and the selected build- The critical design loads, detailed in
The highest air density value is chosen, to ing importance factor is commensurate Table 3, are applied to the two wall con-
result in the highest critical load. with a typical residential structure. The figurations, as shown in Figure 7 for the
Additionally, SANS 10400-B (SANS 2012) behaviour factor for unreinforced masonry SLS and ULS-W, and in Figure 8 for the
specifies minimum wind pressures to be is used, given that minimum detailing and ULS-S. Loads that act out-of-plane (OP) of
applied to housing structural systems of reinforcement requirements are adhered the wall, are applied as a uniform distrib-
370 N/m2 and to housing structural elements to. The fundamental period of vibration is uted load over the entire wall, including
of 450 N/m2. The peak wind pressures chosen based on structural system type. the wind or seismic load, as applicable.
determined according to SANS 10160-3 The reliability redundancy factor is not Under wind action, the most critical
(1 213 N/m2) (2018) exceed these minimum present in the parent standard, Eurocode 8 load case is the modelled wall acting as
load requirements. (EN 1998), but was taken from the Uniform side wall in the context of a 40 m 2 house
IP Roof wind IP
OP Zone B
OP Zone B
OP Zone A
OP Zone A
Figure 7 Load applications for wall W1 (left) and W2 (right) for SLS and ULS-W (N/mm2)
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 51
IP IP
OP OP
Figure 8 Load applications for wall W1 (left) and W2 (right) for ULS-S (N/mm2)
structure. The OP wind loads are there- a gable wall configuration, which does not deflections provided in Figures 12 and 17
fore differentiated into Zones A and B, support trusses, the roof load is only applied are measured at the top midspan position,
according to SANS 10160-3 (SANS 2018b). to the W1 configuration. as indicated in Figure 9. The IP displace-
The total OP force (N) is applied uniform- In the past, numerical and experimental ments provided in Figures 13 and 18 are
ly over the masonry portions of the model studies on unreinforced masonry walls have measured at the top left corner of each
walls to account for the lack of surface been focussed on the effect of either OP or respective wall. Typical crack positions
area over the wall openings. IP loading. In more recent research, the sig- found in the numerical analyses are also
The horizontal in-plane (IP) load is dis- nificance of the interaction of these two load identified in Figure 9, for the later discus-
tributed over the full height and thickness conditions has gained prominence, but still sion on crack damage classification and
of the wall. The load arises from the lateral focused on masonry infill walls. Few numer- crack width in Figure 14.
loads on the walls adjacent to the wall ical investigations have taken into account Typical failure modes are presented in
modelled, and includes either the wind or the combined effect of IP and OP actions Figure 10 for the SLS and ULS-W and in
seismic load, as applicable. The IP load var- on load-bearing unreinforced masonry, and Figure 11 for the ULS-S. Compressive fail-
ies linearly for the wind load case, with the even fewer experimental studies (Milani ure or crushing is not identified in either of
maximum (presented in Table 3) applied 2008; Agnihotri et al 2013; Najafgholipour the wall configurations for both OP and IP
at the top of the wall. Using the principle et al 2013; Dolatshahi et al 2015). Typically failures. This is not remarkable given the
that lateral loads are applied at the location the findings are that the IP load may have a small structure size and low vertical loads.
of the mass, the IP load for the seismic crucial effect on the OP capacity of the wall. The contours in Figure 10 indicate the OP
load case is distributed uniformly over the The wall slenderness and aspect ratios large- deflections and OP failure dominates for
height of the wall. ly determine the interaction level. Therefore, both the SLS and ULS-W, since most of the
The vertical loads transferred from the this study applies the simultaneous action of total load applied is in the lateral direction.
roof to the wall are applied at each truss OP and IP loading. The ‘Base 1’ crack, located in the lowest
support point and distributed over the width bed joint of the wall, indicates tensile fail-
of masonry block to prevent stress concen- ure, together with shear failure in columns
trations at these points. The roof self-weight RESULTS adjacent to the door openings (‘Door’) for
and the wind load, if applicable, are included wall configurations W1 and W2 for both
in this load. Vertical uplift is indicated by a Results overview load cases. Subsequent tensile cracks also
positive value, whereas a compressive force Figure 9 provides an interpretation form in the pier adjacent to the window
is indicated by a negative value. Since W2 is key for results discussed later. The OP opening (‘Pier 1’ and ‘Pier 2’).
OP deflection
IP displacement
IP displacement OP deflection
Pier 3
Pier 1 Lintel
Door Pier 3
Pier 1 Lintel
Door
Pier 4 Pier 4
Pier 2 Pier 2
Figure 9 Crack position and deflection/displacement measurement legend (W1 left, W2 right)
52 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
Figure 10 Typical failure for SLS and ULS-W for walls W1 (left) and W2 (right)
The contours in Figure 11 indicate the Table 7 Summary of OP SLS and ULS-W loads and IP ULS-S loads
IP displacements. IP failure dominates for SANS 10160 Design load /
the ULS-S load case, due to shear action Numerical resistance
design load numerical resistance
in the walls bearing most of the seismic OP/
SLS/ULS
IP [N] [N[ [N] [N] [-] [-]
load. Tensile/flexural cracks dominate,
starting with ‘Base 2’ next to the door W1 W2 W1 W2 W1 W2
and ‘Pier 1’ to ‘Pier 3’ surrounding the SLS-44 OP 12 482 11 530 16 086 21 430 0.8 0.5
window opening. Crack onset at ‘Base 2’
SLS-40 OP 10 315 9 529 16 086 21 430 0.6 0.4
is tensile, and progresses to sliding shear,
whereas ‘Pier 4’ is a combination of shear SLS-36 OP 8 356 7 718 16 086 21 430 0.5 0.4
and tensile stepped cracks. The W2 wall
ULS-W-44 OP 33 285 30 746 15 597 16 543 2.1 1.9
configuration is made particularly vulner-
able by the slender column to the left of ULS-W-40 OP 27 508 25 410 15 597 16 543 1.8 1.5
the door opening. ULS-W-36 OP 22 281 20 582 15 597 16 543 1.4 1.2
As an overview, the design loads and
ULS-S IP 19 613 20 539 25 964 11 295 0.8 1.8
numerical wall resistance determined
through the numerical analyses are pre-
sented in Table 7 for the three load cases, three load cases. The South African load- load-carrying capacity of the walls, namely
namely SLS, ULS-W (for three basic funda- ing code (SANS 10160 2011) OP design in W1 ULS-W, W2 ULS-W (both for all
mental wind speeds of 44, 40 and 36 m/s) loads for each of the three load cases are three basic fundamental wind speeds) and
and ULS-S. For each instance of the critical also included to contextualise the results. W2 ULS-S. It is important to note that,
load direction, the ratio of the design In Table 3 the design loads were provided in line with recent findings in literature
load to the numerical wall resistance is in the form of pressures (N/mm2) but are (Vaculik 2012; Derakhshan et al 2018), the
included. Failure is therefore indicated by a converted to forces (N) to facilitate the OP response for the ULS-S is inadequate
ratio of greater than 1.0. comparison of the design loads and the and that OP behaviour of unreinforced
resistance capacities of the walls. masonry cannot be disregarded under
Out-of-plane response Of the six analyses presented in seismic action. SANS 10400-B (SANS 2012)
The OP load/deflection responses for W1 Figure 12, in half of them the OP design imposes a 1:175 deflection limit on such
and W2 are depicted in Figure 12 for the load significantly exceeds the OP building walls, which is well above the
Figure 11 Typical failure for ULS-S for walls W1 (left) and W2 (right)
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 53
35 000 35 000
NUM SLS
SANS 10160 ULS-W 36 m/s
W2 OP load (N)
W1 OP load (N)
NUM SLS
NUM ULS-W
15 000 NUM ULS-W 15 000
0 0
0 1 2 3 0 1 2 3 4 5 6 7
W1 OP deflection (mm) W2 OP deflection (mm)
Figure 12 Out-of-plane response of walls W1 (left) and W2 (right) for SLS, ULS-W and ULS-S
OP deflection range encountered in these by the particularly weak geometry derived as serviceability performance criteria for
analyses of between 2.5 mm and 7 mm. for W2. masonry walls and were subsequently
The gable of the W2 configuration was The IP displacement of both of the included in SANS 10400.
not buttressed, contrary to the specifica- wall configurations is not significant (less Several typical crack positions were
tions of SANS 10160-4 (SANS 2017). It was than 1 mm). Arguably, the IP displace- identified in Figures 9, 10 and 11 for the
presumed that this lack of lateral support ment would be greater if the load/dis- SLS and ULS-W and ULS-S load cases
to the gable would cause instabilities in the placement path is continued numerically, respectively. The most dominant crack
analyses, but this element was noncriti- but the laborious work of overcoming is identified for each of the analyses per-
cal in the OP loading conditions under the post-peak divergence prevents this formed and plotted against the OP loads for
consideration. This is most probably due pursuit. There are no limitations specified W1 and W2 in Figure 14. Cracks occurred
to the presence of more vulnerable, slender for IP displacement in SANS 10400 as in a number of typical positions for the
elements in the wall in other locations. A they are for IP deflections. However, the IP and OP-dominant loading conditions,
wall with better-proportioned openings pronounced reduction in load-carrying as illustrated in Figure 10, as well as in
may well cause the gable to become critical, capacity, together with the negligible IP Figure 11 for the ULS-W and in Figure 12
requiring buttressing. displacement, is typical of the extremely for the ULS-S. For each analysis performed,
brittle behaviour of masonry. the most dominant crack is identified and
In-plane response plotted against the OP load in Figure 14
The IP load/displacement responses for W1 Crack damage classification (left) for W1 and (right) for W2. For both
and W2 are depicted in Figure 13 for the The crack damage is classified and pre- W1 and W2, the ‘Base’ cracks as well as
three load cases. The South African load- sented in Figure 14 for W1 and W2. Only cracks around the window openings in the
ing code (SANS 10160 2011) IP design loads the dominant crack for each load case ‘Piers’ are prolific. Most cracks measured
for each of the three load cases are also combination is included for clarity. The in the numerical analyses of the concrete
included to contextualise the results. frame of reference for the crack widths masonry walls fall below the ‘negligible’
Reflective of the OP response, in three is taken from the damage categories and (less than 0.25 mm) classification. With
of the six analyses presented in Figure 13, maximum crack widths in SANS 10400-B further development of the walls’ post-peak
the IP design load significantly exceeds (SANS 2012) and the South African Home responses, the cracks would undoubtedly
the IP load-carrying capacity of the walls, Building Manual (NHBRC 2015). The widen. However, in the numerical analyses
namely in W1 ULS-W, W2 ULS-W (both damage categories vary from less than performed, these initial cracks suffice to
for all three basic fundamental wind 0.25 mm, classified as negligible, to greater demonstrate crack development and to
speeds) and W2 ULS-S. Notably, the than 25 mm, classified as very severe. produce a significant reduction in the load-
seismic IP capacity of W1 is 2.3 times These classifications were developed by carrying capacity, which is typical of the
greater than W2. This is explained in part Watermeyer and Tromp (1992) to serve brittle nature of masonry.
54 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
30 000 25 000
NUM ULS-S
25 000 SANS 10160 ULS-S
20 000
15 000
W2 IP load (N)
W1 IP load (N)
15 000
SANS 10160 ULS-W 36 m/s NUM SLS SANS 10160 ULS-W 40 m/s
NUM SLS SANS 10160 ULS-W 36 m/s
5 000
5 000 NUM ULS-W SANS 10160 SLS 44 m/s NUM ULS-W
SANS 10160 SLS 40 m/s SANS 10160 SLS 44 m/s
SANS 10160 SLS 40 m/s
SANS 10160 SLS 36 m/s
SANS 10160 SLS 36 m/s
0 0
0 0.05 0.10 0.15 0.20 0 0.2 0.4 0.6 0.8
W1 IP displacement (mm) W2 IP displacement (mm)
Figure 13 In-plane response of walls W1 (left) and W2 (right) for SLS, ULS-W and ULS-S
Summary points of diversion, such as adjusting the method was used throughout. However,
The progression of most of the analyses step size, increasing the tolerance of the convergence was rarely achieved. It is clear,
was limited by divergence. Several steps convergence criteria, employing other though, from the load-displacement or
were taken to move the analyses past these iterative procedures, etc, and the arc length load-deflection trace, that the linear-elastic
35 000 35 000
20 000 20 000
SANS 10160 ULS-S
Negligible
0.25 mm
0 0
0 0.1 0.2 0.3 0 0.1 0.2 0.3
W1 crack width (mm) W2 crack width (mm)
Figure 14 Crack damage classification for walls W1 (left) and W2 (right) for SLS, ULS-W and ULS-S
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 55
region has been surpassed and that post- Table 8 CON adjusted joint parameters
peak global stiffening is improbable. Joint interface parameters Symbol Baseline Adjusted Unit
Given the conventional strength and
stiffness range of the concrete masonry Tensile strength f t,j TENSTR 0.12 0.84 N/mm2
studied, as well as the typically conservative Mode I fracture energy GIf,j GF 0.005 0.011 N/mm
nature of deemed-to-satisfy solutions, it was
Cohesion cj COHESI 0.17 1.17 N/mm2
anticipated prior to the study that the resis-
tance of these conventional masonry walls
would exceed the design loads in both ULS Material parameters, as provided in Table 2, are
load cases. Contrary to this expectation, To investigate the potential increase in OP kept constant.
both wall configurations failed to resist the load-carrying capacity due to improved The outcome of these analyses (Table 9
ULS-W design loads for basic fundamental material properties, reasonable maximum and Figure 17) shows an increase in the
wind speeds of 44, 40 and 36 m/s, as did W2 values for the three most influential OP load-carrying capacity of 53% for W1 and
for the ULS-S design loads, and by a large parameters (joint tensile strength, joint 11% for W2, due to the improved material
margin. Plausible origins of these failure cohesion and joint mode I fracture energy) properties. This reduces the discrepancy
are the applied design load, the material were sought in literature. Thereafter, to the most critical design load by 47%
input parameters, the derived geometry of ULS-W load analyses were performed on and 13% for W1 and W2, respectively.
the wall configurations and the assumed W1 and W2 with these three adjusted Increasing the critical joint parameters
boundary conditions. These aspects are joint parameters. Experimental data on to reasonable maximum values does not
discussed in the following section. the joint tensile properties is scarce, but increase W2’s resistance sufficiently to
reasonable maximum values for mode I withstand any of the wind design loads.
fracture energy and cohesion were found In the case of W1, the wall’s resistance is
DISCUSSION in literature for normal density concrete increased such that the ULS-W load case
blocks with general purpose mortar joints, based on the 36 m/s basic fundamental
Design load conducted by Van Der Pluijm (1999), wind speed can be resisted, but the 40 and
Most of the assumptions or selections and are detailed in Table 8. All other 44 m/s not.
detailed for determining the wind load
were made to achieve the most critical
wind loading, not the most likely. However, 6 000 mm
even design wind loads based on 40 and
756 mm
36 m/s basic fundamental wind speeds,
which account for the majority of areas in
South Africa, exceed the walls’ capacities in
1 008 mm
most instances.
2 646 mm
882 mm
wind pressure specified in SANS 10400-B
(SANS 2012) (370 N/m2). The substantially
higher design load for the ULS-W case is in
part due to the recent revision of the wind 1 200 mm 900 mm 1 200 mm 1 500 mm 1 200 mm
loading code, SANS 10160-3 (SANS 2018b).
The wind load partial factor has
4 950 mm
increased from 1.3 to 1.6 and the highest
fundamental basic wind speed from 36 m/s
630 mm 630 mm
56 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
Figure 16 Baseline (left) and adjusted (right) boundary conditions for return walls
Table 9 P
eak OP resistance to ULS-W and IP to ULS-S for adjusted material parameters, geometry Boundary conditions
and boundary conditions The conservative assumption was made
that the timber truss system provides
Peak resistance [N] OP W1 OP W2 IP W2
negligible lateral support to the top of the
Baseline 15 597 16 543 11 295 walls. The effect of this assumption could
be meaningful, but its validity is sustained
Adjusted material parameters 23 835 18 373 –
given the similarly weak OP resistance
Adjusted geometry 18 687 18 946 29 534 of the opposite wall, which is meant to
provide the additional lateral resistance, as
Adjusted boundary conditions 18 011 18 197 –
well as the typically poor quality of con-
nection between truss and wall in LIH.
It is important to recall that in all the the seismic design principles of the loading A potential source of error could be
analyses of W1 and W2, mean material code. excessive rotation of the short return walls,
parameter values are used, and the material To investigate the potential improve- which provide lateral restraint to the walls.
resistance has not been reduced by means of ment in OP resistance of both walls under The pinned modelling of the walls could
material partial safety factors. Applying this ULS-W loading and IP resistance for W2 underestimate the rotational restraint
necessary reduction for ULS-based design under ULS-S loading due to more robust that a full-length return wall would offer,
would further widen the discrepancy. geometry, the original window opening thereby allowing greater OP deflection.
length is halved to 1 500 mm, and the door To investigate the effect of this, the trans-
Geometry and window openings are positioned in less lational restraint on the return walls is
The limitations on wall panel sizes and extreme positions in the wall, as illustrated applied to all nodes in the boundary plane,
openings set out in the SANS 10400-K in Figure 15 for W1 and W2. All other origi- as opposed to just the central row of nodes,
deemed-to-satisfy solutions are taken from nal dimensions of the walls are maintained. as shown in Figure 16.
the JSD Code of Practice: Foundations and The outcome of the OP ULS-W The outcome of the OP ULS-W
Superstructures for Single Storey Residential analyses (Table 9 and Figure 17) shows analyses (Table 9 and Figure 17) shows
Buildings of Masonry Construction (JSD an increase in the load-carrying capacity an increase in the load-carrying capacity
1995). Different wall panel configurations of 20% for W1 and 15% for W2, due to of 15% for W1 and 10% for W2, due to
were analysed using the yield line approach the reduced window opening and the less the adjusted boundary conditions and
to derive the panel sizes, and the then extreme positions of the openings. This increased rotational restraint. This reduces
current South African masonry structural reduces the discrepancy to the most critical the discrepancy to the most critical design
design code SABS 0164-1 (SABS 1980) was design load by 17% for both W1 and W2. load by 14% for W1 and 12% for W2.
applied to the respective elements to derive The outcome of the IP ULS-S analyses Increasing the rotational restraint provided
the opening limitations (Watermeyer 1996). on W2 (Table 9 and Figure 18) shows an by the return walls does not increase the
The total area of openings for both increase in the load-carrying capacity of walls’ resistances sufficiently to withstand
W1 and W2 falls within the specifications 160% for W2, due to the reduced window any of the wind design loads.
of seismic design principles set out in the opening and the less extreme positions
loading code, SANS 10160-4 (SANS 2017) of the openings. The IP resistance of W2
Clause 6.2.2, of being less than one third now exceeds the seismic design load by CONCLUSIONS
of the overall wall area. The openings 40%. Reducing the window opening by half The FE analyses were executed success-
are positioned “as uniformly as possible”, does not increase the walls’ OP resistances fully for the two wall configurations under
but given the large opening length of 3 m sufficiently to withstand any of the wind three load cases. The failure modes can
permitted in the deemed-to-satisfy solu- design loads. However, this mitigation be broadly classified as tensile for the
tions of SANS 10400-K (SANS 2011b), it strategy significantly increases W2’s IP OP-dominant cases and a combination of
does result in large openings at both wall load-carrying capacity to successfully resist tensile/shear failure for the IP-dominant
ends, which is undesirable according to the full seismic design load. cases. The analyses revealed the wall
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 57
35 000 35 000
W2 OP load (N)
W1 OP load (N)
10 000 10 000
5 000 5 000
0 0
0 1 2 3 4 0 1 2 3
W1 OP deflection (mm) W2 OP deflection (mm)
Figure 17 OP ULS-W response for adjusted material parameters, geometry and boundary conditions for W1 (left) and W2 (right)
58 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
panel and opening sizes and the current formulation. Journal of the South African Institution Residential Buildings of Masonry Construction.
EC-based loading code, both for wind, i.e. of Civil Engineering, 60(3): 30–40. Halfway House: SAICE and IStructE Joint
SANS 10160-3 (SANS 2018b) and seismic, BS (British Standard) 1978. BS 5628-1:1978. Code of Structural Division.
i.e. SANS 10160-4 (SANS 2017) loads. Practice for the Use of Masonry. Structural Use of Kruger, A, Retief, J & Goliger, A 2017. Development
Based on a case study of extreme wind Unreinforced Masonry. London: British Standards of an updated fundamental basic wind speed map
loads on an inland housing development in Institution. for SANS 10160-3. Journal of the South African
South Africa, Mahachi et al (2018) came to BSI (British Standards Institution) 2005. NA to BS EN Institution of Civil Engineering, 59(4): 12–25.
the conclusion that a review of the techni- 1996-1-1. UK National Annex to Eurocode 6. Design Laubscher, J 2014. Reviewing challenges between the
cal standards in housing development is of Masonry Structures. General Rules for Reinforced need for government-subsidised housing in South
necessary, specifically the NHBRC Home and Unreinforced Masonry Structures. London: BSI. Africa and the sustainability requirements of the
Building Manual. Griffith (2000) reports CMA (Concrete Manufacturers Association) 2011. National Building Regulations. Proceedings, World
on a similar case of discrepancies between Initiatives. http://www.cma.org.za/Initiatives/ SB14, 28–30 October 2014, Barcelona, Spain,
the capacity of the ‘deemed-to-comply’ wall The-CMA-House. pp 18–24.
provisions of the South Australian Housing De Villiers, W I 2019. Computational and experimental Lourenço, P B 1996. Computational Strategies for
Code and the Australian masonry stan- modelling of masonry walling towards performance- Masonry Structures. Delft, The Netherlands: Delft
dard’s design load. It has long been found based standardisation of alternative masonry units University Press.
that, within the field of masonry buildings, for low-income housing. PhD Thesis. Stellenbosch Lourenço, P B & Rots, J 1997. Multisurface interface
low-rise, unreinforced ones with light University. model for analysis of masonry structures. Journal of
roofs (such as LIH), experience the most De Villiers, W I, Fourie, J & Boshoff, W P 2018. Engineering Mechanics, 123: 660–669.
wind damage (Sparks et al 1989), especially Numerical modelling of alternative masonry units. Macorini, L & Izzuddin, B 2011. A non-linear interface
non-engineered ones, relying on empirical Proceedings, International RILEM Symposium on element for 3D mesoscale analysis of brick-masonry
design procedures. This, coupled with the Concrete Modelling (CONMOD18), 26–29 August structures. International Journal for Numerical
significant changes in the required wind 2018, Delft, The Netherlands. Methods in Engineering, 85: 1584–1608.
and seismic design loads with the revision Department of Human Settlements (DHS) 2017. Mahachi, J, Bradley, R & Goliger, A 2018. Wind
of the South African loading code, war- Delivery of Serviced Sites and Houses/Units from storm damage to houses: Planning and design
rants a review of the SANS 10400 deemed- HSDG. Pretoria: DHS. consideration. Proceedings, Out-Of-The-Box
to-satisfy wall provisions. Derakhshan, H, Lucas, W, Visintin, P & Griffith, M Conference, 24–25 October 2018, Pretoria,
It is recommended that this is done with 2018. Out-of-plane strength of existing two-way pp 82–92.
the preferential housing solution in mind, by spanning solid and cavity unreinforced masonry Milani, G 2008. 3D upper bound limit analysis of
first specifying desired wall configurations, walls. Structures, 13: 88–101. multi-leaf masonry walls. International Journal of
based on constructability, typical South DIANA 2017. Finite Element Analysis User’s Manual Mechanical Sciences, 50: 817–836.
African building practice and skill level, fen- 10.2. Delft: DIANA FEA BV. Najafgholipour, M, Maheri, M & Lourenço, P B 2013.
estration requirements for building energy Dolatshahi, K, Aref, A & Whittaker, A 2015. Capacity interaction in brick masonry under
usage, natural lighting and ventilation, Interaction curves for in-plane and out-of-plane simultaneous in-plane and out-of-plane loads.
and fire safety. Numerous such specifica- behaviours of unreinforced masonry walls. Journal Construction and Building Materials, 38: 619–626.
tions are well documented in the relevant of Earthquake Engineering, 19: 60–84. NHBRC (National Home Builders Registration
parts of SANS 10400 and can be used as a EN (European Standard) 1990. EN 1990 (2002). Council) 2015. Home Building Manual and Guide.
basis to determine the desired geometry. Eurocode: Basis of Structural Design. Brussels: Johannesburg: NHBRC.
Subsequently, the derived wall configura- European Committee for Standardization (CEN). Reyes, E, Casati, M J & Gàlvez, J C 2008. Cohesive
tions can be structurally analysed using the EN (European Standard) 1991. EN 1991 (2002+). crack model for mixed mode fracture of brick
simplified micro-modelling approach. Eurocode 1: Actions on Structures. Brussels: CEN. masonry. International Journal of Fracture,
EN (European Standard) 1998. EN 1998-1 (2004). 151: 29–55.
Eurocode 8: Design of Structures for Earthquake Roca, P, Cervera, M, Gariup, G & Pela, L 2010. Structural
ACKNOWLEDGEMENTS Resistance. Part 1: General Seismic Actions and analysis of masonry historical constructions:
Funding: This work was supported by the Rules for Buildings. Brussels: CEN. Classical and advanced approaches. Archives of
National Research Foundation of South Fourie, J 2017. Characterisation and evaluation of the Computer Methods in Engineering, 17: 299–325.
Africa (grant numbers 87961, 106965). mechanical properties of alternative masonry units. RSA (Republic of South Africa) 1977. National Building
MEng Dissertation. Stellenbosch University. Regulations and Building Standards Act, 103 1977,
Giambanco, G, Rizzo, S & Spallino, R 2001. Numerical Government Gazette, No 145 (5640), 6 July 1977,
REFERENCES analysis of masonry structures via interface models. Pretoria: Government Printer.
Abdulla, K, Cunningham, L & Gillie, M 2017. Computer Methods in Applied Mechanics and RSA (Republic of South Africa) 1995. Housing
Simulating masonry wall behaviour using a Engineering, 190: 6493–6511. Consumers Protection Measures Act, 9 of 1995.
simplified micro-model approach. Engineering Griffith, M 2000. Experimental Study of the Flexural Government Gazette, 401(19418), 2 November 1995,
Structures, 151: 349–365. Strength of URM (Brick) Walls. Adelaide, Australia: Pretoria: Government Printer.
Agnihotri, P, Singhal, V & Rai, D 2013. Effect of University of Adelaide. SABS 1980. SABS 0164-1:1980. Code of Practice. The
in-plane damage on out-of-plane strength of ICBO (International Conference of Building Officials) Structural Use of Masonry, Part 1: Unreinforced
unreinforced masonry walls. Engineering Structures, 1997. Uniform Building Code (UBC)™. Whittier, CA: Masonry Walling. Pretoria: South African Bureau of
57: 1–11. ICBO. Standards.
Botha, J, Retief, J & Viljoen, C 2018. Reliability JSD (Joint Structural Division) 1995. Code of Practice: SANS (South African National Standard) 1989.
assessment of the South African wind load design Foundations and Superstructures for Single Storey SANS 10164-1 1989. The Structural Use of Masonry,
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 59
Part 1: Unreinforced Masonry Walling. Pretoria: SANS 2017. SANS 10160-4 2017. Basis of Structural Van Der Pluijm, R 1999. Out-of-plane bending of
SABS Standards Division. Design and Actions for Buildings and Industrial masonry: Behaviour of bed joints. PhD Thesis.
SANS 2010a. SANS 10400 2010. The Application of Structures. Part 4: Seismic Actions and General Eindhoven, The Netherlands, Technische
the National Building Regulations, 3rd ed. Pretoria: Requirements for Buildings. Pretoria: SABS Universiteit Eindhoven, pp 1849–1862. doi.
SABS Standards Division. Standards Division. org/10.6100/IR528212.
SANS 2010b. SANS 10400-A 2010. The Application of SANS 2018a. SANS 10160-1 2018. Basis of Structural Van Zijl, G P A G 2000. Computational modelling of
the National Building Regulations. Part A: General Design and Actions for Buildings and Industrial masonry creep and shrinkage. PhD Thesis. Delft,
Principles and Requirements, 3rd ed. Pretoria: SABS Structures. Part 1: Basis of Structural Design. The Netherlands: Delft University of Technology.
Standards Division. Pretoria: SABS Standards Division. Watermeyer, R 1996. Recent developments in providing
SANS 2011. SANS 10160 2011. Basis of Structural Design SANS 2018b. SANS 10160-3 2018. Basis of Structural houses of masonry construction in South Africa.
and Actions for Buildings and Industrial Structures, Design and actions for Buildings and Industrial The Structural Engineer, 74(19): 325–331.
1.1 ed. Pretoria: SABS Standards Division. Structures. Part 3: Wind. Pretoria: SABS Standards Watermeyer, R 2004. The impact of structural
SANS 2011a. SANS 10160-2 2011. Basis of Structural Division. engineering on the sustainability of human
Design and Actions for Buildings and Industrial SANS 2018c. SANS 51996-1-1 2018. Eurocode 6: Design settlements in developing countries. Proceedings, 2nd
Structures. Part 2: Self-weight and Imposed Loads. of Masonry Structures. Part 1-1: General Rules for International Conference on Structural Engineering,
Pretoria: SABS Standards Division. Reinforced and Unreinforced Masonry Structures, Mechanics and Computation, Cape Town.
SANS 2011b. SANS 10400-K 2011. The Application of 1st ed. Pretoria: SABS Standards Division. Watermeyer, R & Tromp, B 1992. A systematic
the National Building Regulations. Part K: Walls, Sisulu, L 2016. Housing Backlog at 2.1 million, says approach to the design and construction of single-
3rd ed. Pretoria: SABS Standards Division. Minister Sisulu. Engineering News, 22 April 2016. storey residential masonry structures on problem
SANS 2011c. SANS 10400-L 2011. The Application of Sparks, P, Liu, H & Saffir, H 1989. Wind damage soils. The Civil Engineer in South Africa, March:
the National Building Regulations. Part L: Roofs, to masonry buildings. Journal of Aerospace 83–96.
3rd ed. Pretoria: SABS Standards Division. Engineering, 2(4): 186–198. Watermeyer, R & Milford, R 2003. The use of
SANS 2011d. SANS 10400-XA 2011. The Application Sýkora, M & Holický, M 2010. Probabilistic model for performance-based building codes to attain
of the National Building Regulations. Part X: masonry strength of existing structures. Engineering sustainable housing objective: The South African
Environmental Sustainability; Part XA: Energy Mechanics, 174(1): 61–70. approach. Presented at the Global Policy Summit on
Usage in Buildings. Pretoria: SABS Standards Talocchino, G 2005. Design and construction criteria the Role of Performance-Based Building Regulations
Division. for domes in low-cost housing. MEng Dissertation, in Addressing Societal Expectations, International
SANS 2012. SANS 10400-B 2012. The Application Johannesburg: University of the Witwatersrand. Policy, and Local Needs. Washington, DC, US.
of the National Building Regulations. Part B: Vaculik, J 2012. Unreinforced masonry walls subjected Wium, J A 2010. Background to draft SANS 10160
Structural Design, 3rd ed. Pretoria: SABS Standards to out-of-plane seismic actions. PhD Thesis. (2009) Part 4. Seismic loading. Journal of the South
Division. Adelaide, Australia: University of Adelaide. African Institution of Civil Engineering, 52(1): 20–27.
60 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
Design and construction TECHNICAL NOTE
of laboratory-scale Journal of the South African
Institution of Civil Engineering
automobile workshops
management, water sanitation and hygiene
(WASH), and the application of computer technology in water resources
and sanitation facilities management. He is currently a PhD student at
Stellenbosch University.
Contact details:
C O Ataguba, I Brink Department of Civil Engineering
Stellenbosch University
Private Bag X1, Matieland,
Stellenbosch 7602
The design and construction of low-cost laboratory-scale filter columns using locally available South Africa
Nigerian filter materials – granular activated carbon (GAC), gravel (GR) and rice husk (RH) – were T: +234 803 567 1452
E: clematrix2008@gmail.com
carried out and reported. The filter materials and columns were designed, constructed and used
for the treatment of stormwater runoff from selected automobile workshops in Nigeria over
DR ISOBEL BRINK is a senior lecturer in water
a period of three rainy months. The combined granular activated carbon and rice husk filter quality and environmental engineering at
systems performed best with pollutant removal efficiency of 58%. It was shown that the materials, the Department of Civil Engineering,
considered as waste, could be recycled and used as filter materials in the treatment of stormwater Stellenbosch University. She is interested in
new innovations in water quality
from automobile workshops. This low-cost technology for stormwater runoff treatment, especially
improvement methods and undertakes
for automobile workshops at large scale and in-situ, can be further explored. projects with wide application. Recent
research includes simple point-of-use
systems for potable water treatment, river water quality modelling, and
the use of green infrastructure and LID technologies for surface runoff
INTRODUCTION and non-availability of appropriate labour water quality improvement.
Stormwater runoff from automobile to operate the technologies (Bahgat et al
Contact details:
workshops washes debris, waste, oil com- 1999). These challenges have created Department of Civil Engineering
binations, grease, etc, through the drainage opportunities for the exploration of locally Stellenbosch University
infrastructure within the urban catchment available materials/technologies. Private Bag X1, Matieland,
Stellenbosch 7602
and discharges into receiving water bodies. In this research, the potential of com-
South Africa
This has resulted in a high degree of water bined (a) gravel – activated carbon, (b) acti- T: +27 21 808 4195
pollution in Nigeria (Idu 2015; Ekiye & vated carbon – rice husk, and (c) single rice E: icbrink@sun.ac.za
Zejiao 2010). husk as filter materials in the treatment of
Studies have shown that land use plays a selected automobile workshop stormwater
critical role in the concentration and com- runoff was investigated. These materials
position of pollutants that are discharged are readily available and affordable in
into receiving water bodies (Khatun et al Nigeria. Five automobile workshops were
2014; Shrestha 2017; Wang et al 2013). selected from the two study towns of Idah
In a study carried out by Pitt et al (1995), and Lokoja in Nigeria for this stormwater
it was reported that stormwater runoff sampling and treatment. These selected
from vehicle service and parking lots was workshops were named Automobile
found to have the highest levels of metals, Workshops 1–5.
petroleum hydrocarbon compounds and This technical note reports the design,
organics pollutants when compared with construction and preliminary use of the
other urban land uses. combined (a) gravel – granular activated
Conventional wastewater/stormwater carbon, (b) granular activated carbon – rice
treatment methods, such as reverse osmo- husk, and (c) single rice husk filter materi-
sis, chemical precipitation, electrodialysis, als as vertical columns for the treatment of
ion exchange, ultrafiltration, etc, have automobile workshop stormwater runoff.
been reported to be unsuitable for adop- Results of the laboratory investigation
tion in developing countries due to high of the pre-treatment and post-treatment Keywords: automobile workshop, filter media, pollution,
operating costs, high energy requirement quality of the stormwater samples collected stormwater runoff, treatment
Ataguba CO, Brink I. Design and construction of laboratory-scale activated carbon, gravel and rice husk filter columns for the treatment of stormwater
runoff from automobile workshops. J. S. Afr. Inst. Civ. Eng. 2021:63(1), Art. #1158, 6 pages. http://dx.doi.org/10.17159/2309-8775/2021/v63n1a6 61
(a) (b) (c)
Figure 1 F ilter materials: (a) granular activated carbon, (b) gravel, (c) rice husk
62 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
Table 1 Details of design equations, data and references for the filter designs
Filter Design equation Design data Reference
Vf ⎫⎪ ⎫⎪ = = Q ÷
Depth of column H = 0.40 m
πD 2
⎭h⎭ A
m Q Height of filter L = 0.3 m
Freeboard t = 0.10 m
4
Diameter of column D = 0.10 m
Wegelin (1996);
GR filter design Volume of gravel VG = 0.0024 m3
Density-Mass-Volume Relation
Density of gravel ρG = 1 400 kg/m3
Mass (m) Mass of gravel MG = 3.4 kg
Density ρ =
Volume (v) Filtration rate Vf = 1.0 m/h
Flow rate Q = 0.008 m3/h
L(m)
Design and construction of Filter material
granular activated carbon, gravel
and rice husk filter columns
The vertical downflow bed filter system PVC filter mesh
has been adopted from Wegelin (1996).
This design concept is based on its ease of Araldyte (sealant)
use of gravity flow in underdeveloped areas
(Diaper 1965; Pratap et al 2007). Flow regulator
The details of the design equations,
data and the relevant references used in Discharge hose
the design of the filters are presented in
Table 1 with the design equations from the
Filter stand
references, while the dimensions are the
authors’ work. The design data in Table 1 Effluent collector
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 63
Table 2 Pollutant removal efficiency of the filters for Automobile Workshop 4 at week 1
Te Te Te
Untreated
Test parameters GAC-RH (GAC-RH) GR-GAC (GR-GAC) RH (RH)
sample
% % %
Dissolved oxygen, DO (mg/L) 4.62 36.712 5.66 22.466 6.13 16.027 7.3
Total dissolved solids, TDS (mg/L) 604.08 68.776 1 181.01 38.956 646.56 66.581 1 934.69
Total solids, TS (mg/L) 786.08 82.162 1 297.01 70.567 787.56 82.128 4 406.69
Total suspended solids, TSS (mg/L) 182 92.638 116 95.307 141 94.296 2472
Oil and grease, O&G (mg/L) 3.43 46.238 4.88 23.511 4.24 33.542 6.38
Filter system setups, stormwater over regeneration for this study, as it is eco- filter system was computed, as presented in
sampling, treatment and nomical, and materials are locally available Tables 2, 3 and 4.
laboratory analyses at little or no cost. The parameters analysed The pollutant removal efficiency Te of a
The different filter materials designed in included: pH, conductivity, turbidity, oil particular filter system with respect to any
Table 1 were placed in the different filter and grease (O&G), dissolved oxygen (DO), particular parameter is given as:
columns and fastened to the fabricated total dissolved solids (TDS), total solids
stand as shown in Figure 4(b). (TS), total suspended solids (TSS), cadmium Cd – C x
Te(x)% = × 100
Stormwater sampling from each (Cd), copper (Cu), lead (Pb) and iron (Fe). Cd
automobile workshop spanning a period These parameters were selected based on
of nine weeks during the rainy season was the characteristic pollutants associated with Where:
carried out according to Lowe et al (2018). this land use (Pitt et al 1995). The param- Cd = concentration of a particular param-
The influents and effluents from the differ- eters were analysed in accordance with eter from its untreated sample
ent filter setups were analysed for quality APHA (2017) at the Water Quality Control C x = concentration of the same parameter
before (raw stormwater) and after treatment Laboratory at the National Geosciences from the filtered sample x
with the filter media (filtered effluents). Research Laboratories in Kaduna, Nigeria. (for x = filter systems GAC-RH, GR-GAC
Replacement of filter material was chosen The efficiency of pollutant removal of each or RH).
Table 3 Pollutant removal efficiency of the filters for Automobile Workshop 4 at week 2
Te Te Te
Untreated
Test parameters GAC-RH (GAC-RH) GR-GAC (GR-GAC) RH (RH)
sample
% % %
Dissolved oxygen, DO (mg/L) 4.28 45.408 5.1 34.949 5.88 25.000 7.84
Total dissolved solids, TDS (mg/L) 641.46 68.547 1 164.78 42.887 770.05 62.242 2 039.44
Total solids, TS (mg/L) 854.46 80.200 1 357.78 68.537 971.05 77.498 4 315.44
Total suspended solids, TSS (mg/L) 213 90.641 193 91.520 201 91.169 2 276
Oil and grease, O&G (mg/L) 3.73 43.910 5.12 23.008 4.81 27.669 6.65
64 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
Table 4 Pollutant removal efficiency of the filters for Automobile Workshop 4 at week 3
Te Te Te
Untreated
Test parameters GAC-RH (GAC-RH) GR-GAC (GR-GAC) RH (RH)
sample
% % %
Dissolved oxygen, DO (mg/L) 4.11 47.975 5.65 28.481 6.4 18.987 7.9
Total dissolved solids, TDS (mg/L) 642.26 66.919 1 302.18 32.927 768.36 60.423 1 941.45
Total solids, TS (mg/L) 849.26 79.602 1 477.18 64.520 949.36 77.198 4 163.45
Total suspended solids, TSS (mg/L) 207 90.684 175 92.124 181 91.854 2 222
Oil and grease, O&G (mg/L) 3.3 48.031 4.92 22.520 4.76 25.039 6.35
70
(a) (b) 60
Average pollutants removal (%)
50
40
30
20
10
0
GAC-RH RH GR-GAC
(c) GAC (c) RH (c) Gravel
Filter media
Figure 4 C
onstructed / fabricated filter columns: (a) constructed filter column, (b) filter column
setup with stands, (c) filter column with different filter materials, from left to right: GAC, Figure 5 A
verage pollutant removal efficiencies
RH and gravel of the different filter media
Journal of the South African Institution of Civil Engineering Volume 63 Number 1 March 2021 65
husk and GAC can be used as source of APHA (American Public Health Association) 2017. Maebh, A G 2016. Development of filtration
heat energy for local industries after sun- Standard Methods for the Examination of Water technologies for effective, cost-efficient and robust
drying, while the gravel can be used for and Wastewater. 22nd ed. Washington, DC: APHA. water treatment. PhD Thesis. Galway, Ireland:
concreting. The impact of the filters dis- Bahgat, M, Dewedar, M A & Zayed, A 1999. Sand- National University of Ireland.
posal on the environment will be reduced. filters used for wastewater treatment: Buildup and McDougall, G J 1991. The physical nature and
Tables 2, 3, and 4 show the computed distribution of microorganisms. Water Research, manufacture of activated carbon. Journal of
average performance of these filter systems 33(8): 1949–1955. the Southern African Institute of Mining and
in the treatment of the stormwater runoff Baker, F S, Miller, C E, Repik, A J & Tolles, E D 1992. Metallurgy, 91(4): 109–120.
in terms of pollutant removal efficiency. Activated carbon. In Kirk Othmer Encyclopaedia of Nurul-Amin, M D, Kaneco, S, Kitagawa, T et al
Figure 5 shows that the combined GAC– Chemical Technology, Wiley Online Library. 2006. Removal of arsenic in aqueous solutions by
RH filter system performed best with Cecen, F & Aktas, Ö 2011. Activated Carbon for adsorption onto waste rice husk. Industrial and
average Te ≈ 58%. The RH filter system and Water and Wastewater Treatment: Integration of Engineering Chemistry Research, 45: 8105–8110.
the combined Gravel–GAC filter systems Adsorption and Biological Treatment. New York: Pitt, R, Field, R, Lalor, M & Brown, M 1995. Urban
followed the combined GAC–RH filter Wiley. stormwater toxic pollutants: Assessment, sources
system with average Te ≈ 49% and average Chukwudebelu, J A, Igwe, C C & Madukasi, E I and treatability. Water Environment Research,
Te ≈ 40% respectively. 2015. Prospects of using whole rice husk for the 67(3): 260–275.
production of dense and hollow bricks. African Pratap, M R, Khambhammettu, U, Clark, S E & Pitt, R
Journal of Environment, Science and Technology, 2007. Stormwater polishing: Upflow vs downflow
CONCLUSION AND 9(5): 493–501. filters. Proceedings, World Environmental and
RECOMMENDATION Diaper, E W J 1965. Upflow and downflow filtration Water Resources Congress, American Society of
The results obtained from the use of this through graded media. Proceedings of the Institution Civil Engineers, 15–19 May, Tampa, FL.
technology showed that the combined of Civil Engineers, 30(2): 437. Shrestha, D 2017. Characterization and modelling
GAC and RH filter system performed best Ekiye, E & Zejiao, L 2010. Water quality monitoring of stormwater for the City of Calgary. MSc
in the pollutant removal efficiency (58%). in Nigeria. Case study of Nigeria’s industrial cities. Dissertation. Canada: University of Calgary.
The RH filter (49%) and the combined GR Journal of American Science, 6, 22–28. Sukia, S K 2020. Rice husk-derived adsorbents for
and GAC (40%) filter systems respectively Idu, A J 2015. Threats to water resources development in water purification, Chapter 6. In Naushad, M &
followed in pollutant removal efficiency. Nigeria. Journal of Geology and Geophysics, 4: 205. Lichtfouse, E (Eds.), Green Materials for Wastewater
This technology can be upscaled for use Karthik, R G H, Abhishek, N P, Kishor, V R, Nikhil, E, Treatment, Cham, Switzerland: Springer.
in other environments, with modifications Poornima, K B & Shivakumara, B 2018. Comparative USACE (United States Army Corps of Engineers)
in the available filter materials and the analysis of locally available adsorbents for 2001. Adsorption Design Guide. Guide 1110-1-2. US
need for sustainable use of the technology. purification of water. International Research Journal Department of the Army.
Other cheap locally available agricultural of Engineering and Technology, 5(5): 1722–1726. Wang, S, He, Q, Hainan, A, Wang, Z & Zhang, Q
waste materials, such as sugarcane bagasse, Khatun, A, Bhattacharyya, K G & Sarma, H P 2014. 2013. Pollutant concentrations and pollution loads
groundnut/melon shells, maize cobs, etc, Levels of pollutants in runoff water from different in stormwater runoff from different land uses in
can be explored for use as filter materi- land uses in Guwahati City, Assam, India. Archives Chongqing. Journal of Environmental Sciences,
als. It is highly recommended that other of Applied Science Research, 5: 96–100. 25(3): 502–510.
low-cost technology for stormwater runoff Lakhote, A, Ahire, P, Dabholkar, P, Dhadambe, V & Wegelin, M 1996. Surface Water Treatment by
treatment, especially for automobile work- Gharat, S 2016. Comparative analysis of design of Roughing Filters. A Design, Construction and
shops at large scale and in-situ, be explored, water filter for rural areas. International Journal Operation Manual. St. Gallen, Switzerland: SKAT
researched and prototypes rolled out. of Innovative Research in Advanced Engineering, Consulting Ltd.
3(12): 23–27. Xu, M, Liu, X, Tang, Q, Qu, R & Xu, Q 2013.
Lowe, J, de Leon, D, Collins, J, Hoover, R & Book, S Utilization of rice husks modified by
REFERENCES 2018. Standard Operating Procedure for Collecting organomultiphosphonic acids as low-cost
Andersen, J E H 2016. Hydraulic performance of Grab Samples from Stormwater Discharges, biosorbents for enhanced adsorption of heavy metal
advanced treatment media to improve quality Vol 1.1. Publication No 18-10-023. Olympia, ions. Bioresource Technology, 149: 420–424.
of stormwater from airports exposed to de-icing WA: Washington State Department of Ecology. Zunaira, A & Zhi, C 2017. Removal of arsenic from
chemicals. MSc Dissertation. Trondheim, Norway: https://fortress.wa.gov/ecy/publications/ drinking water using rice husk. Applied Water
Norwegian University of Science and Technology. summarypages/1810023.html. Science, 7: 1449–1458.
66 Volume 63 Number 1 March 2021 Journal of the South African Institution of Civil Engineering
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 permission to publish experimental data and other or should be typed using the Times New Roman
Engineering is published quarterly in March, June, information that may be confidential or sensitive. symbol set.
September and December. Articles submitted for Authors are also responsible for obtaining permis- QQ Variables in equations (x, y, z, etc, as well as
publication are reviewed by a panel of referees under sion from copyright owners when reproducing mate- lower case Greek letters) should be presented
the guidance of the SAICE Journal Editorial Panel. rial that had been published elsewhere. Proof of such in italics. Numbers (digits), upper case Greek
The journal publishes research papers covering all the permission must be supplied. letters, symbols of metric measurement units
disciplines of civil engineering (structural, geotechnical, (m for metres, s for seconds, etc) and mathemati-
railway, coastal/marine, water, construction, environ- SUBMISSION PROCEDURES cal/trigonometrical functions (such as sin, cos
mental, municipal, transportation) and associated topics AND REQUIRED FORMAT and tan) are not written in italics, but in upright
that are relevant to the civil engineering profession, and QQ Online submission: Manuscripts must be uploaded type (Roman). Variables and symbols used in the
that preferably have particular relevance to civil engi- as PDF files (http://journal.saice.org.za). Individual body of the text should match the format used
neering in southern Africa and the African continent. file sizes may not exceed 10 MB. Should you experi- in the equations, i.e. upright or italics, whichever
When preparing articles for publication, authors ence problems uploading your paper, please contact is applicable.
should please take note of the following and comply the editor (verelene@saice.org.za). QQ Metric measurement abbreviations/units should
with the guidelines as set out: QQ Format: Manuscripts should be prepared in MS conform to international usage – the SI system of
Word and presented in double line spacing, single units should be used.
CLASSIFICATION OF ARTICLES column layout with 25 mm wide margins. Line QQ Decimal commas may be used, but decimal
CONSIDERED FOR PUBLICATION numbers must be applied to the whole document. points are preferred.
QQ Technical papers are well-researched, in-depth, All pages should bear the authors’ names and be QQ Symbols should preferably be defined in the text,
fully referenced technical articles not exceeding numbered at the bottom of the page. With the excep- but if this is not feasible, a list of notations may be
6 000 words in length (excluding tables, illustra- tion of tables and figures (see below) the document provided for inclusion at the end of the paper.
tions and the list of references). Related papers that should be typed in Times New Roman 12 pt font. QQ Headings: Sections and paragraphs should not be
deal with ‘softer sciences’ (e.g. education, social Contributions should be accompanied by an abstract numbered. The following hierarchy of headings
upliftment, etc) are accepted if they are of a techni- of not more than 200 words. should be followed:
cal nature and of particular interest to the civil QQ First page: The first page of the manuscript should HEADING OF MAIN SECTION
engineering profession. The latter type of paper will include the title of the paper, the number of words of Heading of subsection
be subject not only to peer-review by civil engineers, the main text (i.e. excluding figures, tables and the Heading of sub-subsection
but also to review by non-engineering specialists in list of references), the initials and surnames of the QQ References: References should follow the Harvard
the field covered by the paper. authors, professional status (if applicable), SAICE system. The format of text citations should be as
QQ Technical notes are short, fully referenced technical affiliation (Member, Fellow, Visitor, etc), telephone follows: “Jones (1999) discovered that …” or “recent
articles that do not exceed 2 000 words. A typical numbers (landline and mobile), and e-mail and results (Brown & Carter 1985; Green et al 1999)
technical note will have limited scope often dealing postal addresses. The name of the corresponding indicated that …”
with a single technical issue of particular importance author should be underlined. Five keywords should References cited in the text should be listed in
to civil engineering. be suggested. alphabetical order at the end of the paper. References
QQ Review papers are considered for publication as QQ Figures, tables, photos and illustrations: These by the same author should be in chronological order.
either technical papers or technical notes on condi- should preferably be submitted in colour, as the The following are examples of a journal article, a
tion that they are the original work of the author journal is a full-colour publication. book and a conference paper:
and will assist the reader with the understanding, QQ Their positions should be clearly marked in the QQ Donald, A W 1954. Pore pressure coefficients.
interpreting or applying of the subject under review. text as follows: [Insert Figure 1]. Géotechnique, 4(4):143–147.
A review paper must contain criteria by which the QQ Figures, tables, photos, illustrations and equa- QQ Terzaghi, K 1943. Practical soil mechanics. New
work under review was evaluated, and contribute tions should be numbered consecutively and York: Wiley.
by synthesising the information and drawing new should appear in the text directly after they have QQ Smith, A W, Black, C & White, A 1999. Factors
conclusions from the dissemination of the previously been referred to for the first time. affecting the strength of soils. Proceedings, ASCE
published work. QQ Illustrations must be accompanied by appropriate Conference on Shear Strength of Soils, Colorado,
QQ Discussion on published articles is welcomed up to captions. Captions for tables should appear above pp 503–532.
six months after publication. The length of discus- the table. All other captions should appear below Papers published previously in the Journal of the
sion contributions is limited to 1 500 words. Where the illustration (figures, graphs, photos). South African Institution of Civil Engineering should
appropriate, discussion contributions will be subject QQ Only those figures and photographs essential to be cited if applicable.
to the normal reviewing process and will be for- the understanding of the text should be included. QQ Footnotes, trade names, acronyms, abbreviations:
warded to the authors of the original article for reply. All illustrations should be referred to in the text. These should be avoided. If acronyms are used, they
QQ Figures should be produced using com- should be defined when they first appear in the text.
POLICY REGARDING LANGUAGE AND puter graphics. Hand-drafted figures will not be Do not use full stops after abbreviations or acronyms.
ORIGINALITY OF SUBMITTED ARTICLES accepted. Lettering on figures should be equiva- QQ Return of amended papers: Papers requiring
QQ Language: Manuscripts should preferably be lent to a Times New Roman 9 pt font or slightly amendments will be accepted up to six months after
presented in English, as the journal is distributed larger (up to 12 pt) if desired. Lettering smaller the referee reports had been sent to authors, after
internationally. Articles submitted in any of the than 9 pt is not acceptable. which the paper will be withdrawn from the system.
other official South African languages should be QQ Tables should be typed in Times New Roman
accompanied by an expanded abstract in English. 9 pt font. They should not duplicate information FINAL ARTICLE
QQ Original work: Papers and technical notes must already given in the text, nor contain material QQ Copyright: On acceptance of the paper or technical
be original contributions. Authors must affirm that would be better presented graphically. note, copyright must be transferred by the author/s
that submitted material has not been published Tabular matter should be as simple as possible, to the South African Institution of Civil Engineering
previously, is not under consideration for publica- with brief column headings and a minimum on the form that will be provided by the Institution.
tion elsewhere and will not be submitted elsewhere number of columns. QQ Photos of authors: The final corrected version of
while under consideration by the SAICE Journal QQ Mathematical expressions and presentation the paper should be accompanied by recent, high-
Editorial Panel. It is the responsibility of the authors of symbols: resolution head and shoulders colour photographs
to ensure that publication of any paper in the journal QQ Equations should be presented in a clear form and a profile not exceeding 100 words for each of
will not constitute a breach of any agreement or which can easily be read by non-mathematicians. the authors.
the transgression of any law. The corresponding Each equation should appear on a separate line QQ Proofs: First proofs of papers will be sent to authors
author should confirm that all co-authors have and should be numbered consecutively. in PDF format for verification before publication.
read and approved the manuscript and accept these QQ Symbols should preferably reflect those used in No major re-writes will be allowed, only essential
conditions. Authors are responsible for obtaining Microsoft Word Equation Editor or Mathtype, minor corrections.