Volume 64 Number 4

December 2022

QQ E xperimental study and classification of natural

zeolite pozzolan for cement in South Africa
QQ  pplication of a modern laser technique to evaluate
A
the performance of recycled railway ballast
QQ  proposal to facilitate BIM implementation across
A
the South African construction industry
QQ  raffic safety and the rural road environment:
T
Assessing the impact of combined roadway
conditions on crash incidence
Volume 64 No 4 December 2022
ISSN 1021-2019 (print) | ISSN 2309-8775 (online)

Publisher
South African Institution of Civil Engineering
Block 19, Thornhill Office Park,
Bekker Road, 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
day@jaws.co.za
Co-Editor-in-chief
Prof Gerhard Heymann 2 Experimental study and classification of natural
University of Pretoria
gerhard.heymann@up.ac.za zeolite pozzolan for cement in South Africa
Administrative Editor
F Sinngu, S O Ekolu, A Naghizadeh, H A Quainoo
Verelene de Koker
Tel +27 11 805 5947, Mobile +27 83 378 3996
verelene@saice.org.za
journal editorial Panel
16 Application of a modern laser technique to evaluate
Prof Y Ballim – University of the Witwatersrand
Prof W Burdzik – Retired from the University of Pretoria
the performance of recycled railway ballast
Dr P Day – Jones & Wagener (Pty) Ltd (chairman)
Prof J du Plessis – University of Stellenbosch
G M Mvelase, J K Anochie-Boateng, P J Gräbe
Prof GC Fanourakis – University of Johannesburg
Prof PJ Gräbe – University of Pretoria
Dr C Herold – Umfula Wempilo Consulting
Prof G Heymann – University of Pretoria
29 A proposal to facilitate BIM implementation
Prof A Ilemobade – University of the Witwatersrand across the South African construction industry
Prof SW Jacobsz – University of Pretoria
Prof EP Kearsley – University of Pretoria S Calitz, J A Wium
Prof CP Roth – University of Pretoria
Prof W Steyn – University of Pretoria
Mr M van Dijk – University of Pretoria
Prof C Venter – University of Pretoria 38 Traffic safety and the rural road environment:
Prof A Visser – University of Pretoria
Dr E Vorster – Zutari Assessing the impact of combined roadway
Prof J Wium – University of Stellenbosch
Prof A Zingoni – University of Cape Town conditions on crash incidence
Prof M Zuidgeest – University of Cape Town
R Ambunda, M Sinclair
PEER REVIEWING / OPEN ACCESS / COPYRIGHT
The Journal of the South African Institution of Civil
Engineering is a peer-reviewed, open-access journal
licenced under a Creative Commons Attribution Licence 49 Thanks to our Referees!
(CC BY-NC-ND) and is distributed internationally. The
content of this journal is available to users free of
charge. However, copyright remains with the South
African Institution of Civil Engineering. Permission for 50 Guidelines for the preparation of
reproduction of content, with appropriate reference to the
journal, can be obtained from the administrative editor.
papers and technical notes
Design and reproduction
Marketing Support Services, Ashlea Gardens, Pretoria
The South African Institution of Civil Engineering accepts no
responsibility for any statement made or opinion expressed
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.

Creative Commons
Attribution Licence

© South African Institution of Civil Engineering 1

TECHNICAL PAPER Experimental study and
Journal of the South African
Institution of Civil Engineering classification of natural
ISSN 1021-2019 (print) | ISSN 2309-8775 (online)
Vol 64 No 4, December 2022, Pages 2–15, Paper 1443 zeolite pozzolan for
FHATUWANI SINNGU (BTech Eng, NDip Eng)
is a civil engineering laboratory technician
cement in South Africa
at the University of Johannesburg. She is
currently completing an MPhil in concrete
research at the University of Johannesburg. F Sinngu, S O Ekolu, A Naghizadeh, H A Quainoo

Contact details:
Department of Civil Engineering Science The present study investigated the performance of South African natural zeolite for potential
University of Johannesburg
PO Box 524, Auckland Park, Johannesburg, 2006, South Africa
use as a cement extender. To the best knowledge of the authors, the natural zeolite studied
T: +27 71 686 1240 | E: fsinngu@uj.ac.za has not been employed before as pozzolan in the cement industry, and its proposed use
would be new in concrete construction. In the investigation, mortar mixtures were prepared
PROF STEPHEN EKOLU (Pr Eng, MSAICE,
QMSA, MCSSA), who holds a PhD from the using ordinary Portland cement CEM Ι 52,5N blended with 0, 10, 20 and 30% natural zeolite.
University of Toronto and an MSc (with The effectiveness of natural zeolite was also compared with performance results of mixtures
distinction) from the University of Leeds, was containing 30% fly ash. The tests conducted were workability, compressive strength, pozzolanic
formerly a lecturer at the University of the
Witwatersrand, Head of School of Civil
activity with lime, drying shrinkage, alkali-silica reaction, and sulphate resistance. Analytical
Engineering and the Built Environment at the University of Johannesburg, studies were done using X-ray diffraction and scanning electron microscopy coupled with
and is currently with the Department of Civil Engineering at the Nelson energy dispersive spectroscopy. It was found that, while incorporation of natural zeolite
Mandela University in Gqeberha (formerly Port Elizabeth). He is a rated
in cement reduces workability, its effects on mechanical properties and on durability
researcher with more than 23 years of academic and industry research
experience. His research interests include concrete materials and structures, characteristics were superior to those of fly ash, as the former effectively enhanced long-
cementitious materials, durability of concrete, service life modelling of term strength and mitigated drying shrinkage, alkali-silica reaction and sulphate attack.
concrete structures, environmental science, and engineering education. Incorporation of 20% natural zeolite in cement gave an effective overall performance, meeting
Contact details: the ASTM C618 (2015) criteria for Class N pozzolan. Based on EN 197-1 (2000) / SANS 50197‑1
Department of Civil Engineering
(2013), the blend of natural zeolite and ordinary Portland cement met the criteria for its
Nelson Mandela University
PO Box 77000, Gqeberha, 6031, South Africa classification as CEM II/A-P 32.5N,R cement type.
E: sekolu@gmail.com

DR ABDOLHOSSEIN NAGHIZADEH (Pr Eng

IRCEO) is a lecturer of structures and concrete INTRODUCTION cement, extenders also generally improve
materials in the Department of Engineering Over the past decades, the blending of the engineering properties and durability
Sciences at the University of the Free State in
cement extenders or pozzolans with ordi- performance characteristics of cementi-
South Africa. He holds BSc and MSc degrees
(with distinction) from Azad University, Iran, nary Portland cement (OPC) has become tious systems (Mardani-Aghabaglou et al
and a PhD from the University of Johannesburg. He is a researcher with more conventional practice, following numerous 2014). For example, it is well established
than 17 years of industry and research experience. His research interests early studies conducted to investigate their that cement extenders significantly reduce
include concrete technology, alternative cements, geopolymer binders, 3D
printing of concrete, and structural design and analysis.
different effects on properties of cementi- permeability levels of, and mitigate vari-
tious systems (Neville 1981; Ding & Li ous deterioration processes that occur in
Contact details:
Department of Engineering Sciences 2002). Once blended with OPC, pozzolans hardened cementitious systems (Valipour
University of the Free State react with the calcium hydroxide (CH) et al 2013). Some common extenders
5 Logeman Street, Park West, Bloemfontein, South Africa
formed from cement hydration to generate are typically blended with OPC at high
E: naghizadeha@ufs.ac.za / honair@yahoo.com
secondary calcium silicate hydrate phases proportions, such as 30 to 50% FA, GGBS,
DR HARRY A QUAINOO (AMSAICE, MAIPM, (CSH). Pozzolans or cement extenders are of etc. Consequently, such extenders tend to
MCSSA, ICIOB), who obtained his PhD and MSc
two categories comprising, (i) natural poz- reduce early-age strength and also reduce
degrees from the University of the
Witwatersrand, is a senior lecturer of zolans such as volcanic ash, natural zeolite the hydration heat of cementitious systems
transportation engineering, urban planning and (NZ), volcanic tuff, kaolin, etc, and (ii) arti- (Addis 1998).
construction management at the University of ficial pozzolans which are industrial waste In South Africa, artificial pozzolans
Johannesburg, and has more than 15 years of teaching and lecturing experience.
His areas of expertise are transportation engineering and project management,
by-products including fly ash (FA), ground comprising FA, GGBS and SF are the most
with interest in concrete materials and engineering education. granulated blast-furnace slag (GGBS) and widely used cement extenders (Addis 1998),
Contact details: silica fume (SF), amongst others (Ekolu et al as underscored by their existing standard
Department of Civil Engineering Science 2006; Naghizadeh & Ekolu 2017a; Tchadjie specifications comprising SANS 50450‑1
University of Johannesburg
& Ekolu 2018; Tchadjie et al 2021). (2014), SANS 55167-1 (2011) and
PO Box 524, Auckland Park, Johannesburg, 2006, South Africa
E: hquainoo@uj.ac.za Besides the environmental and eco- SANS 53263-1 (2011), respectively. While
nomic benefits of blending pozzolans with no similar separate standard specifications

Keywords: cement extender, pozzolan, natural zeolite, durability,

drying shrinkage, alkali-silica reaction, sulphate attack, Sinngu F, Ekolu SO, Naghizadeh A, Quainoo HA. Experimental study and classification of natural zeolite pozzolan for cement in South Africa.
J. S. Afr. Inst. Civ. Eng. 2022:64(4), Art. #1443, 14 pages. http://dx.doi.org/10.17159/2309-8775/2022/v64n4a1
supplementary cementitious materials
The Journal of the South African Institution of Civil Engineering is a peer-reviewed, open-access journal licenced under a Creative Commons Attribution Licence
(CC BY-NC-ND) and is distributed internationally. The content of this journal is available to users free of charge. However, copyright remains with the South African
2 Institution of Civil Engineering. Permission for reproduction of content, with appropriate reference to the journal, can be obtained from the administrative editor.
exist for local natural cement extenders, Table 1 C
 hemical compositions of the cementitious materials used
SANS 50197-1 (2013) makes general provi- Ordinary
sion including that for use of natural poz- Oxides Natural zeolite Fly ash
Portland cement
(%) (NZ) (FA)
zolans. Meanwhile, there is presently little (CEM Ι 52,5N OPC)
to no local research on the utilisation of Al2O3 4.49 12.9 30.27
natural pozzolans such as NZ in cement and
CaO 64.47 1.4 4.59
concrete. Prior to the present study, only one
material characterisation study on South Fe2O3 3.07 2.4 3.58
African NZ could be found, which also was MgO 1.11 1.6 1.06
not focused on the use of zeolite as pozzolan
P2O5 0.08 – 0.38
in cement (Diale et al 2011).
K 2O 0.32 – 0.77
Existing literature on the use SiO2 20.57 61.3 56.45
of zeolites as pozzolans
Natural zeolites (NZs) form when vol- Na2O 0.05 2.1 0.14

canic rocks react with alkaline ground- TiO2 0.53 – 1.57

water, leading to crystallisation in a
post-­depositional environment occurring
in shallow basins over long periods of time
(Colella et al 2001; Diale et al 2011). It
may be noted that NZs have a high water
demand. As such, a relatively higher water
content is needed to achieve adequate
workability of cementitious mixtures con-
taining NZ (Colella et al 2001; Marantos
et al 2020). The higher water demand of
NZ/OPC blended mixtures is related to the
presence of numerous small pores found
within the pozzolan’s vesicular structure.
Blends containing higher than 10% NZ may
require use of a plasticiser for the mixture
to attain similar workability as that of the
control (Marantos et al 2020).
Various literature sources have reported
contradicting findings on the effect of NZ
upon early-strength development. Sicakova
et al (2017) reported that the early-age
compressive strength of concrete increased
upon replacement of OPC with 15% NZ.
In contrast, Yilmaz et al (2007) found that
concrete mixtures containing 5, 10, 20 and
40% NZ exhibited lower early-age strength
values, relative to that of the control mix-
ture. Subsequently, however, the 28-day
compressive strength of the 10, 20 and 40%
NZ/OPC concretes increased by 15.9, 22.3
and 4.1% respectively (Yilmaz et al 2007).
Jana (2007) observed that concrete mix-
tures containing 10 to 20% NZ gave drying
shrinkage results that were similar to those
of the control OPC mixture. However,
the 30% NZ/OPC concrete mixture gave
drying shrinkage that was 20% higher than
that of the control. Interestingly, a study by
Najimi et al (2012) showed that incorpora-
tion of NZ into OPC significantly improved
the durability properties of mortars. In
their study, the 15% NZ/OPC blend was
found to effectively mitigate sulphate attack
and alkali-silica reaction (ASR).
LOI*
Relative density
Percentage retained on 45 µm sieve (%)
* LOI–loss on ignition
Study objectives
Presently, hardly any past research exists
on potential utilisation of South African
NZ as pozzolan. As such, the objectives
of the present study were to investigate
performance characteristics of NZ, and to
determine its classification along with suit-
able proportions for blending with OPC.
Proposed use of NZ in cement and con-
crete would be a new development, since
this cementitious material has not been
employed before as cement extender.
In the present study, mortar mixtures
were prepared in which CEM Ι 52,5N
OPC cement, was partially replaced with
0 to 30% NZ. The mechanical properties
measured were workability, compressive
strength development and pozzolanic
activity of NZ with lime (PAL). Also mea-
sured were durability properties compris-
ing drying shrinkage, ASR and sulphate
attack. Analytical studies were done using
X-ray diffraction (XRD) and scanning
electron microscopy (SEM) coupled with
energy dispersive spectroscopy (EDS).
EXPERIMENTAL STUDY
Materials
CEM Ι 52,5N OPC cement obtained from
Pretoria Portland Cement (PPC) (Pty) Ltd,
was used in all the mortar mixtures. In
South Africa, deposits of NZ are found in
the provinces of the Western Cape and
KwaZulu-Natal. The NZ supplied by Serina
Trading for use in the present study, was
4.04 15.8 0.42
3.14 2.33 2.24
4.03 0 2.46
obtained from the Riversdale mine in the
Western Cape. The extender was incorpo-
rated into OPC in proportions of 0, 10, 20
and 30% NZ, then used to prepare mortar
mixtures. For performance comparison of
the NZ with conventional pozzolans, mortar
mixtures containing 30% FA were also pre-
pared. The siliceous fly ash (SANS 50450-1
2014), which is a low calcium (Class F) fly
ash (ASTM 618 2015) used in the present
study, was obtained from Lethabo Power
Station as supplied by Ash Resources (Pty)
Ltd. Table 1 gives the chemical composi-
tions of the different cementitious materials
employed in the study. The PAL test was
done for both calcined and non-calcined
NZ. Calcination of NZ was done using a
laboratory furnace set at a heating rate of
2°C/min, then maintained at 500°C peak
temperature for an hour, before cooling back
to room temperature.
Local silica sand was used as the non-
reactive aggregate in mortar mixtures that
were prepared for determination of com-
pressive strength, for PAL test, sulphate
resistance and drying shrinkage tests.
The silica sand used was obtained from
Sallies Silica (Pty) Ltd. Particle sizes of
0–0.6 mm, 0.4–0.85 mm and 0.8–1.8 mm
were blended at proportions of 38, 12 and
50%, respectively, to obtain standard sand
grading (Ekolu 2014).
Greywacke (GW) aggregate, obtained
from Peninsula Quarry in the Western
Cape Province, was used to prepare mortar
samples for ASR testing. This aggregate
type is known to be highly ASR-reactive

Journal of the South African Institution of Civil Engineering Volume 64 Number 4 December 2022 3
(Naghizadeh & Ekolu 2017b). The GW
aggregate was received in 19 mm stone
size, then crushed using a laboratory jaw
crusher to obtain various fine particle
sizes. The crushed GW fine aggregate
particles were graded as per ASTM C1260
(2014) and SANS 6245 (2006). Prior to use
in mixtures, the GW fine aggregate was
washed to remove dust content, and then it
was oven-dried at 50°C for 48 hours.
Sodium hydroxide (NaOH) of technical
grade was supplied by Merck (Pty) Ltd for
use to prepare the storage solution for the
accelerated mortar bar test (AMBT), as
per ASTM C1260 (2014) and SANS 6245
(2006). To prepare one litre of 1M NaOH
solution, 40 g pellets of the chemical
reagent were dissolved in 100 g of water,
then more water was added to make up one
litre of the solution. Anhydrous sodium
sulphate (Na2SO4), also supplied by Merck
(Pty) Ltd, was used to prepare the storage
solution for the sulphate resistance test
conducted as per ASTM C1012 (2017). In
the experiment, 50 g of Na2SO4 were dis-
solved in distilled water to prepare one litre
of solution.
Mortar samples
Samples for testing of mechanical
properties
Mortar cubes of 50 mm size were prepared
for compressive strength testing as per
ASTM C109 (2019). The mortar mixtures
containing 0, 10, 20, 30% NZ and 30% FA,
were prepared at aggregate/cement ratio =
3, as given in Table 2a. Using a laboratory
mortar mixer, the dry materials were first
mixed at slow speed for one minute. With
the mixer running, water was added slowly
over a duration of one minute, then mixing
was continued for a further two minutes.
Finally, mixing was done at high speed
for one minute (ASTM C305 2016). Flow
workability of the fresh mortar mixtures
was tested as per ASTM C1437 (2015).
Fresh mortar cubes of 50 mm in size were
cast and covered with a plastic sheet, then
kept for 24 hours at room temperature in
the laboratory. Afterwards, the mortar
samples were demoulded, and then cured
in a water tank at room temperature until
testing. Compressive strength testing of the
cube samples was done at the ages of 3, 7,
28, 56, 90 and 180 days.
The PAL test is done to evaluate reac-
tivity and to classify the quality of the poz-
zolan. In the present study, both NZ and
calcined NZ (CNZ) were subjected to the
test as per ASTM C311 (2013). Mixtures
for the PAL test were prepared at poz-
zolan to lime ratio = 2, while the ratio of
standard sand to lime was = 9. The amount
of water added was adjusted for each mix,
to obtain a flow of 110 ± 5 mm. Mixing of
the mortars was done in accordance with
ASTM C305 (2016), as already described
above. After casting of the 50 mm mortar
cubes for the PAL test, they were covered
with a rigid glass and then kept at room
temperature for 24 hours. Afterwards, the
samples in their moulds were wrapped
using plastic cling film and then stored at
55°C in an oven, until the age of compres-
sive strength testing at seven days.
Prisms of 25 × 25 × 285 mm were
prepared for conducting the drying shrink-
age test, in accordance with ASTM C596
(2018). Table 2a gives the mixtures used for
the drying shrinkage test. The fresh mortar
prisms were cast, and then wrapped using
plastic cling film for 24 hours to prevent
moisture loss. The prisms were demoulded
and then cured in a water tank for six
days. After removal of the samples from
the water, the initial length readings were
recorded, following which the prisms were
air-dried. Subsequent length change mea-
surements were done after 1, 7, 14, 28, 42,
56, 70, 84, 98, 112, 126 and 140 days of dry-
ing at room temperature in laboratory air.
Samples for measurement
of durability properties
The AMBT for ASR was conducted using
mortars containing 0, 10, 20 and 30% NZ
or 30% FA. Table 2b gives the ASR mortar
mixtures made with the GW reactive
aggregate, then used to conduct the AMBT.
The 25 × 25 × 285 mm fresh mortar prisms
for AMBT were cast and wrapped using a
plastic cling film, and then kept at room
temperature for 24 hours. On the following
day, the prisms were demoulded, and then
their initial length values were measured
using a length comparator. Afterwards, the
prisms were immersed in tap water inside
an airtight sealed container, and then the
assembly was stored inside an 80°C oven
for 24 hours. Immediately after removal
of samples from the oven, their surfaces
were wiped off using a cloth, and then
initial length values were again measured.

Table 2a M
 ortar mixtures used for the workability, compressive strength and drying shrinkage tests
Mix ingredients

*OPC (CEM Ι Natural

Mix ID Sample size Fly ash (FA) Water Silica sand Tests
52,5 N) zeolite (NZ)
(mm)
(g) (g) (g) (g) (g)

OPC 50 × 50 × 50 586 – – 360 1 758 Compressive strength

10NZ 50 × 50 × 50 527 59 – 360 1 758 Compressive strength

20NZ 50 × 50 × 50 469 117 – 360 1 758 Compressive strength

30NZ 50 × 50 × 50 411 176 – 360 1 758 Compressive strength

30FA 50 × 50 × 50 411 – 176 240 1 758 Compressive strength

CEMΙ-DS 25 × 25 × 285 391 – – 215 1 172 Drying shrinkage

10NZ-DS 25 × 25 × 285 351 39 – 215 1 172 Drying shrinkage

20NZ-DS 25 × 25 × 285 313 78 – 215 1 172 Drying shrinkage

30NZ-DS 25 × 25 × 285 274 117 – 215 1 172 Drying shrinkage

30FA-DS 25 × 25 × 285 274 – 117 215 1 172 Drying shrinkage

* OPC – ordinary Portland cement CEM Ι 52,5N

4 Volume 64 Number 4 December 2022 Journal of the South African Institution of Civil Engineering
Table 2b Mortar mixtures used for the alkali-silica reaction (ASR) and sulphate resistance tests
Mix ingredients

Greywacke
*OPC (CEM Ι Natural
Mix ID Sample size Fly ash (FA) Water Silica sand aggregate Tests
52,5N) zeolite (NZ)
(mm) (GW)

(g) (g) (g) (g) (g) (g)

OPC-ASR 25 × 25 × 285 400 – – 200 – 900 ASR

10NZ-ASR 25 × 25 × 285 324 36 – 216 – 810 ASR

20NZ-ASR 25 × 25 × 285 288 72 – 216 – 810 ASR

30NZ-ASR 25 × 25 × 285 252 108 – 216 – 810 ASR

30FA-ASR 25 × 25 × 285 336 – 144 240 – 1 080 ASR

OPC-SR 25 × 25 × 285 391 – – 215 1 172 – Sulphate attack

10NZ-SR 25 × 25 × 285 351 39 – 215 1 172 – Sulphate attack

20NZ-SR 25 × 25 × 285 313 78 – 215 1 172 – Sulphate attack

30NZ-SR 25 × 25 × 285 274 117 – 215 1 172 – Sulphate attack

30FA-SR 25 × 25 × 285 274 – 117 215 1 172 – Sulphate attack

* OPC – ordinary Portland cement CEM Ι 52,5N.

Subsequently, the mortar prisms were
immersed in 1M NaOH solution inside an
airtight sealed plastic container, and then
placed back in the 80°C oven for storage.
Length change measurements were made
every 14 days until the age of 160 days.
The sulphate resistance behaviour of
mortars containing NZ was evaluated in
accordance with ASTM C1012 (2017), using
the mortar mixtures given in Table 2b.
Samples for the sulphate resistance test
comprised prisms of 25 × 25 × 285 mm.
The prisms were cast and then cured in
water at 35°C for 24 hours, along with some
cubes of 50 mm size. On the following day
the samples were demoulded and further
cured in a water bath maintained at 23°C,
until the compressive strength of the 50
mm cubes attained ≥ 20 MPa. At this
point, the initial length values of the prisms
were measured, and then the samples
were immersed in 2.8M Na2SO4 solution
170 mm for the 0.5 and 0.6 w/cm mortars,
respectively. All mixtures containing NZ
gave lower workability values relative to
those of the corresponding control mix-
tures. The 0.5 w/cm mixtures containing
10, 20 and 30% NZ gave workability reduc-
tions of 8, 13 and 17%, respectively. A simi-
lar trend was observed for the 0.6 w/‌cm
mixtures. It can be seen in Figure 1 that
the higher the NZ proportion incorporated,
the lower the workability value obtained.
These observations are in agreement with
those reported by Ding and Li (2002). As
200
180
160
140
120
Flow (mm)
expected, the mixtures containing 30% FA
exhibited higher workability results relative
to those of the corresponding control mix-
tures. The observed workability-reducing
effect of NZ is attributed to its high water
demand, as already discussed above under
the section titled Existing literature on
the use of zeolites as pozzolans on page 3
(Marantos et al 2020).
Compressive strength results
Figure 2 and Table 3 show compressive
strength results for the different mortar

inside an airtight sealed plastic container

100
stored at room temperature. Subsequent
length change measurements were done at 80
the ages of 7, 14, 42, 56, 84, 112, 140 and
160 days. 60

40

RESULTS AND DISCUSSION 20

0
Workability results OPC 10% NZ 20% NZ 30% NZ 30% FA
Figure 1 gives workability results of the Binder (%)
mortar mixtures (Table 2a) prepared at 0.5 0.6 w/cm 0.5 w/cm
and 0.6 water/cementitious (w/cm) ratios.
The control mixture prepared using plain Figure 1 F low workability results of mortar mixtures prepared at 0.5 and 0.6 water/cementitious
OPC gave flow values of 159 mm and (w/cm) ratios: OPC (ordinary Portland cement CEM Ι 52,5N), NZ (natural zeolite), FA (fly ash)

Journal of the South African Institution of Civil Engineering Volume 64 Number 4 December 2022 5
 mixtures tested at various ages of up to 180
50 days. Early strength is typically evaluated
(a) Early strength gain
for ages of up to seven days, while standard
strength is determined at 28 days (SANS
50197-1 2013). Strength tendency at ages
40
beyond 28 days is considered to be late or
long-term strength behaviour. It can be
Compressive strength (MPa)

seen in Figure 2a that, at the age of three

30 days, the control mixture showed relatively
rapid early strength gain. However, by
the age of seven days, strength results
of the 10% NZ mixture had caught up
20
with corresponding values of the control
mixture, beyond which both mixtures
exhibited similar strength behaviour at all
10 later or long-term ages of up to 180 days
(Figure 2b).
Indeed, it is evident in Figure 2b that
beyond 28 days there were significant
0
0 3 7 14 gains in compressive strength values of
Age (days) mixtures containing the extenders, while
OPC 10% NZ 20% NZ 30% NZ 30% FA the strength curve for the control OPC
mixture plateaued. By the age of 180 days,
the lagging strength curves of the 20%
50
(b) Overall strength development NZ and 30% FA mixtures had caught
up with or exceeded that of the control
mixture. For example, at 28 days, 20% NZ
40 and 30% FA samples gave values of 33.9
and 26.4 MPa, both of which were lower
Compressive strength (MPa)

than 41.1 MPa of OPC, but at 180 days all

30 three mixtures gave similar values of 40 to
43 MPa (Table 3).
It was mentioned earlier (under
Materials on page 3) that 30% FA is
20
generally recognised and conventionally
employed in structural concrete. It can be
seen in Figure 2 that 20% NZ generally gave
10 an overall strength performance closest
to that of 30% FA, considering behaviour
across all ages from early to late-age
strength development. Meanwhile, the
0
0 28 56 84 112 140 168 196 performance of 30% NZ was very weak or
Age (days) poor, and therefore not worth considering
OPC 10% NZ 20% NZ 30% FA 30% NZ for structural concrete. Drawing from
the foregoing, performance comparison
Figure 2 C
 ompressive strength results of mortar mixtures containing varied proportions between 20% NZ and 30% FA was con-
of natural zeolite (NZ) or fly ash (FA): (a) early strength gain, (b) overall strength sidered to be meaningful and was thus
development – OPC (ordinary Portland cement CEM Ι 52,5N) adopted, particularly in the analytical
studies given later in this paper (Analytical
Table 3 C
 ompressive strength results of NZ/OPC mixtures: NZ (natural zeolite), FA (fly ash), OPC studies on page 9).
(ordinary Portland cement CEM Ι 52,5N)
Compressive strength (MPa) on varied ages Drying shrinkage results
Figure 3 shows the drying shrinkage results
Mix ID 3 days 7 days 28 days 90 days 180 days
determined at different ages of up to 140
OPC 20.7 26.7 41.1 40.8 41.8 days. The control OPC mixture (OPC‑DS)
10NZ 15.8 28.9 37.6 43.4 42.4 gave the 28-day drying shrinkage of 0.04%.
20NZ 16.2 23.0 33.9 38.7 40.7 Partial replacement of OPC with 30%
NZ (30NZ-DS) resulted in a significant
30NZ 13.0 22.9 26.4 31.2 33.3
reduction of the 28-day drying shrinkage
30FA 12.5 18.9 26.4 37.9 42.6
to 0.02%. Also, the mixture containing

6 Volume 64 Number 4 December 2022 Journal of the South African Institution of Civil Engineering
 30% FA (30FA-DS) gave a considerably
0.06 low 28-day shrinkage value of 0.028%.
Throughout the 140 days of the experimen-
tal study, all mixtures containing cement
0.05 extenders showed lower drying shrink-
age levels relative to that of the control.
Mortars containing 10 to 20% NZ gave
0.04 similar results, while the mixture contain-
Drying shrinkage (%)

ing the higher 30% NZ content (30NZ-DS)

gave the lowest drying shrinkage of all the
0.03
mixtures investigated. Generally, however,
all mixtures gave shrinkage values that
were lower than the maximum limit of
0.02
0.1% specified in ASTM C596 (2018).

ASR test results

0.01
Table 4 gives the alkali content levels in
the extenders used. NZ had the highest
alkali content equivalent of 3.75%, fol-
0
0 1 7 14 28 42 56 70 84 98 112 126 140 160 lowed by FA which had an alkali level of
Age (days) 0.65%.. Each of these values exceed the
OPC-DS 30% FA-DS 10% NZ-DS 20% NZ-DS 30% NZ-DS maximum allowable alkali content of 0.6%
Na 2Oeq for OPC mixtures (Neville 1981).
Figure 3 D
 rying shrinkage results of mortar mixtures containing varied proportions of natural However, it is generally recognised that
zeolite (NZ) or fly ash (FA): OPC (ordinary Portland cement CEM Ι 52,5N) the alkalis in pozzolans do not typically
contribute to or participate in ASR (Kalina
Table 4 Alkali content levels in the extenders used et al 2021), except in some peculiar cases.
Alkalis (%) For example, Ekolu et al (2006) found that
Material volcanic tuff pozzolan promoted ASR
Na2O K 2O *Na2Oeq
expansion.
OPC 0.05 0.32 0.26 Given in Figure 4 are the ASR expan-
NZ 2.1 2.5 3.75 sion results measured over a prolonged
FA 0.14 0.77 0.65
period of up to 160 days. The control OPC
mixture (OPC-ASR) gave a high 14-day
*Na2Oeq = Na2O + 0.658K 2O
ASR expansion of 0.15%, which exceeds
the recommended maximum level of
0.10%, thereby confirming the already
0.35 established understanding that GW aggre-
gate is highly reactive. Partial replace-
0.30 ment of OPC with 10% NZ led to slight
reduction of the 14-day ASR expansion to
0.13%. Incorporation of higher zeolite pro-
0.25
portions of 20 and 30% NZ into OPC led
to corresponding significant reductions in
ASR expansion (%)

0.20 the 14-day ASR expansion, giving values

of 0.10 and 0.056% respectively, which are
0.15
incidentally better or similar to the 0.10%
value obtained for the 30% FA mixture.
Evidently NZ mitigated ASR, becoming
0.10 more effective with increase in proportion
of the pozzolan from 10 to 30% NZ. The
0.05 observed effectiveness of NZ against ASR
attack is partly attributed to the pozzolan’s
high SiO2 content (Table 1) which reacts
0
0 14 28 42 56 70 160 170 with calcium to reduce the potential for
Age (days) ASR gel formation (Naghizadeh & Ekolu
OPC-ASR 30% FA-ASR 10% NZ-ASR 20% NZ-ASR 30% NZ-ASR 2017b). But perhaps the most significant
ASR mitigating effect of pozzolans arises
Figure 4 E xpansion due to alkali-silica reaction (ASR) in mortar mixtures containing varied from the ability of the extenders to bind
proportions of natural zeolite (NZ) or fly ash (FA): OPC (ordinary Portland cement) freely available alkalis in the mixture,

Journal of the South African Institution of Civil Engineering Volume 64 Number 4 December 2022 7
 converting them into salts (Shafaatian et
0.05 al 2013).

0.04
Sulphate attack
The long-term mortar expansion results
Sulphate expansion (%)

due to sulphate attack are given in Figure 5.

0.03 The control sample (OPC-SR) gave a
28-day expansion of 0.03%, which also was
the highest value obtained among all the
0.02
mixtures investigated. Interestingly, the 20
and 30% NZ mortar mixtures exhibited
0.01 significantly lower 28-day expansion val-
ues, giving 0.019 and 0.015%, respectively.
At the age of 160 days, the 20 and 30% NZ
0
0 7 14 28 106 118 140 160 180 mortars gave long-term expansion values
Storage duration (days) of 0.022 and 0.019% respectively, which are
OPC-SR 10% NZ-SR 30% FA-SR 20% NZ-SR 30% NZ-SR much lower than the corresponding 0.038%
of OPC and 0.032% of the 30% FA mortars.
Figure 5 S ulphate expansion results of mortars containing varied proportions of natural zeolite As expected, the 30% FA mortar also
(NZ) or fly ash (FA): OPC (ordinary Portland cement) reduced sulphate expansion.

Pozzolanic reactivity and

100 thermogravimetry
It was found that CNZ gave the 7-day PAL
95 test compressive strength of 18.09 MPa,
which is much higher than the 5.67 MPa
Weight change (%)

90 of the NZ sample. Evidently, calcination

of NZ had a remarkably significant effect,
85
leading to high enhancement of pozzolanic
reactivity, as already recognised in the
80
literature (Mielenz et al 1950; Marantos et
al 2020). It is known that the crystalline
75
structures of NZs, change to amorphous
phases upon heat treatment of the zeolites
70
0 100 200 300 400 500 600 700 at high temperatures, leading to enhance-
Temperature (°C) ment in pozzolanic reactivity (Marantos et
(a) NZ lime CNZ lime Lime al 2020).
Figure 6a shows weight loss graphs
of thermogravimetric analyses for the
NZ-lime and CNZ-lime samples. Also
0
shown in Figure 6b are the weight loss
derivatives, distinctly identifying points
Derivative of weight change (%)

CSH of drastic weight loss, which indicate

–0.1 temperatures at which particular phases
underwent decomposition. The peaks seen
CC between 50 to 100oC in Figure 6b, show
–0.2 the formation of CSH in the NZ-lime and
CNZ-lime samples, owing to pozzolanic
reaction (Collier 2016). Evidently, no CSH
–0.3 peak formed in the lime sample due to
the absence of pozzolanic reaction in this
sample, as expected.
CH
It can also be seen that the sample
–0.4
0 100 200 300 400 500 600 700 of lime showed rapid conversion at a
Temperature (°C) temperature range of 340 to 439°C, as did
(b) NZ lime CNZ lime Lime both the CNZ-lime and NZ-lime samples.
Close inspection shows that the CNZ-lime
Figure 6 T hermogravimetric analysis of zeolite-lime mortar mixtures: (a) weight loss curves, graph exhibited more substantial weight
(b) derivatives: NZ (natural zeolite), CNZ (calcined natural zeolite), CH (calcium loss at around 355°C relative to that of
hydroxide), CSH (calcium silicate hydrate), CC (calcite) the NZ-lime sample (Figure 6b). This

8 Volume 64 Number 4 December 2022 Journal of the South African Institution of Civil Engineering
 Analytical studies
8 000 Based on the results of physical tests as
Q– quartz
M– mica
discussed above in the sections titled
Z – zeolite Compressive strength results, Drying shrink-
P – portlandite age results, ASR test results, and Sulphate
A – albite
attack (pages 5–8), it was observed that
6 000 C – calcite
20% NZ gave the closest overall mechanical
M Q performance to that of 30% FA, but with the
Q 30FA-ASR
A
M Q former exhibiting better durability proper-
Z A
Intensity (count/s)

CZP Q Q
P Q Q ties. Accordingly, results showed that 20%
NZ would perhaps be the maximum suitable
4 000 M
Q dosage for conventional use of the natural
Q A pozzolan in cement. On this basis, the 20%
Q 20NZ-ASR
Z M NZ and 30% FA samples were selected for
P A C Q Q
ZP Q Q the microanalytical investigations provided
in the subsequent sections.
2 000
Q X–ray diffraction
M M
Z A P OPC-ASR
P Q
C Z Q Q Q Q
A Q (a) Mortar samples subjected
to alkali-silica reaction
0
0 10 20 30 40 50 60 70 80  igure 7 shows XRD patterns of ASR
F
2 – theta mortar samples that had been immersed in
1M NaOH at 80°C for 140 days during the
Figure 7 X
 RD patterns of ASR expansion mortars containing natural zeolite (NZ) or fly ash (FA): AMBT. The strong peaks of quartz, albite
ASR (alkali-silica reaction), OPC (ordinary Portland cement) and mica minerals observed in all samples
are crystalline phases of the GW aggregate.
observation indicates that more CH had exhibited another peak at 600 to 700oC, Also seen are peaks of portlandite and
been consumed by CNZ during the lime- which is attributed to calcite (CC) forma- calcite crystals, being the crystalline prod-
pozzolan reaction, which explains this tion due to carbonation that occurred upon ucts resulting from cement hydration and
sample’s higher compressive strength of exposure of samples to atmospheric CO2 carbonation, respectively. Interestingly, zeo-
18.09 MPa relative to the lower 5.67 MPa during sample preparation (Collier 2016; lite peaks occur in all samples, as observed
of the NZ-lime mixture. All the samples Ekolu 2016, 2018). at angles of 12.5o and 32.8o (2θ). Evidently,
new zeolites formed during storage of the
samples at the elevated temperature of
8 000 80°C. The relatively higher intensity of the
Q – quartz
P – portlandite
zeolite peak seen at the 12.5° (2θ) angle for
Q C – calcite the 20% NZ sample may indicate the pres-
G – gypsum ence of some unreacted NZ particles in the
E – ettringite
mixture. By comparing the XRD patterns
6 000 of samples in Figure 7, made with the GW
G Q 30FA-SR
G aggregate versus those for the samples
Q C C
E Q Q Q prepared using silica sand as given later in
G P Q
Intensity (count/s)

E E E P Q P Q Figure 8, it is evident that the zeolite which

P
formed in the former emanated from phases
4 000
Q in the reactive GW aggregate. Clearly, no
G G
C 20NZ-SR zeolite formed in samples made with silica
E Q Q C Q sand (Figure 8).
G E P Q Q Q
E E Q P Q In all the XRD patterns of Figure 7, a
P
broad hump occurs within the range of a 26°
2 000 P Q
G to 35° (2θ) angle, which is a feature majorly
Q P Q OPC-SR attributed to the presence of an amorphous
E G C
G C Q Q Q ASR gel phase (Shi et al 2019), along with
E Q Q Q
E E Q P other non-crystalline phases such as unre-
P
acted glassy phases of NZ.
0
0 10 20 30 40 50 60 70 80
2 – theta (b) Mortar samples subjected
to sulphate attack
Figure 8 X
 RD patterns of sulphate attack in mortars containing natural zeolite (NZ) or fly ash (FA): Figure 8 gives XRD patterns of mortar
OPC (ordinary Portland cement) samples that had been subjected to

Journal of the South African Institution of Civil Engineering Volume 64 Number 4 December 2022 9
sulphate attack under storage in 2.8M
Na2SO4 solution at 23°C for 160 days.
The strong peaks of quartz seen in the
XRD patterns of all the mortar samples,
arose from the silica sand used as fine
aggregate in mortar mixtures. Similarly,
peaks of portlandite and calcite can be
seen in the XRD patterns of all mortar
mixtures, the former being a product of
cement hydration while the latter is CaCO3
formed due to carbonation of samples
under exposure to the atmosphere during
specimen preparation (Collier 2016; Ekolu
2016, 2018).
During pozzolanic reaction, the CH
found within the hardened cement paste
is consumed, leading to secondary forma-
tion of an amorphous CSH phase in the
binder matrix. In turn, a relatively denser
binder matrix develops which reduces
permeability and improves the durability
performance of the hardened cementitious
system. It can be seen in Figure 8 that
the mortar containing 20% NZ gave the
lowest peaks of portlandite, as seen at 2θ
angles 18o, 34 o and 45.7°. This observation
indicates that the 20% NZ mortar exhib-
ited much greater pozzolanic reactivity
than the other mixtures, including the
30% FA sample.
Figure 9 SEM features of alkali-silica reaction (ASR): (a) ordinary Portland cement (OPC) mortar, (b) close-up image of the cracked aggregate in (a)
10
Also in Figure 8, peaks of gypsum and
ettringite are seen in patterns of all the
mortar mixtures. It can be seen at the
11.6° (2θ) angle that the 20% NZ sample
had the lowest peak of gypsum, which
confirms the NZ’s effectiveness in mitigat-
ing sulphate attack, as discussed earlier
(Figure 5). The formation of expansive
gypsum and ettringite in the hardened
binder system is the mechanism respon-
sible for expansion due to sulphate attack,
as given in Equations 1 and 2 (Clifton &
Pommersheim 1994). However, the pres-
ence of ettringite peaks, as seen in XRD
patterns, does not strictly indicate that the
mortar mixtures were expansive, consid-
ering that the observed intensities include
the early-age ettringite phase that formed
during cement hydration. Typically, during
sulphate attack, expansive ettringite forms
in confined spaces within the hardened
cement paste. Since gypsum and ettringite
occupy more volume than their reactants,
the formation of these mineral phases
in confined spaces leads to expansion.
Indeed, the reactions given in Equations
1 and 2 result in volume increase with
volume expansion factors of 1.24 and 2.38,
respectively (Clifton & Pommersheim
1994; Mather 1978).
Volume 64 Number 4 December 2022
Na2SO4(aq) + Ca(OH)2(s) + 2H2O 
2NaOH(aq) + CaSO4 ·2H2O(s)(1)
(gypsum)
3Na2SO4(aq) + C3Al2O6(s) +
3Ca(OH)2(s) + 32H2O  6NaOH(aq) +
C6Al2(OH)12(SO4)3.26H2O(s) (2)
(ettringite)
Scanning electron microscopy
(a) Alkali-silica reaction features
Figure 9 shows SEM features of the control
OPC mortar following 160 days of the
AMBT. It can be seen in Figure 9a that
intense internal cracking of the reactive
aggregate particles occurred throughout
the mortar matrix. The observed severe
microcracking of reactive aggregate parti-
cles is a diagnostic feature typical of ASR
attack. Figure 9b gives a close-up image
of observed microcracking, showing the
ASR gel infilling characterised by the gel’s
smooth surface texture. The predominant
presence of Si/Al elements seen in the
EDS patterns, is characteristic of ASR gel.
The gel substance is a reaction product
of NaOH and Si/Al elements found in the
reactive GW aggregate. ASR gel has a high
affinity for water, which it imbibes, leading
Journal of the South African Institution of Civil Engineering
Figure 10 S EM features of alkali-silica reaction (ASR) in mortars: (a) 20% NZ microstructure, (b) 30% FA microstructure, (c) 20% NZ, EDS, (d) 30% FA,
EDS: NZ (natural zeolite), FA (fly ash)

to expansion and disruptive cracking of the
hardened cementitious system.
Figure 10 shows the SEM features of
ASR in the mortars containing 20% NZ
and 30% FA, following 160 days of AMBT
at 80°C. Some cracks can be seen in the
mortar samples of Figures 10a and 10b. The
reactive aggregate particles can be seen con-
taining a light-coloured mica mineral phase.
Evidently, the overall ASR damage was
significantly less in the mortars containing
20% NZ and 30% FA (Figures 10a and 10b),
relative to the severe damage exhibited by
the control sample (Figures 9a and 9b).
Figures 10c and 10d give the EDS analyses
of the different phases present in the 20%
NZ and 30% FA mortar samples. Upon close
inspection, it is evident in Figure 10c that the
mica phase exhibited intense cracking, while
the other phases in the aggregate remained
relatively stable. It appears that mica is the
reactive phase of the GW aggregate.
The elemental composition of the ASR
gel product that was formed in the 30% FA
mortar sample (Figure 10d) is similar to
that which occurred in the control mortar
(Figures 9a and 9b). It may, however, be
recalled that the mixtures containing 20%
NZ and 30% FA both exhibited signifi-
cantly lower ASR expansion levels relative
to that of the control, as discussed earlier
(see Figure 4).
(b) Sulphate attack features
Figure 11 shows SEM features of the control
OPC mortar that had been subjected to

Journal of the South African Institution of Civil Engineering Volume 64 Number 4 December 2022 11
Figure 11 S EM features of sulphate attack in the control ordinary Portland cement (OPC) mortar: (a) ettringite infilling in an air pore, (b) close-up
image of the ettringite in (a) showing needle-like crystals
sulphate attack for 140 days. It can be seen in
Figure 11a that the originally empty air pore
in cement paste was completely infilled by
ettringite. The close-up image of Figure 11b
shows needle-like crystals of the ettringite
infilling, also confirmed by the associated
EDS analysis. These observed SEM features
are consistent with findings from XRD
analysis (as discussed under X-ray diffrac-
tion on page 9) and with the high sulphate
expansion results of the control OPC mortar
(as discussed under Sulphate attack on page
8 (Figure 5)). The observed severe microc-
racking of mortars, as seen in Figure 11a, is
attributed to expansive formation of gypsum
and ettringite phases, which generated
intense internal stresses within the matrix,
in turn causing high physical expansion
(Figure 5) as per the mechanism explained
under Mortar samples subjected to sulphate
attack on pages 9 and 10.
12
Figure 12 shows SEM features of
sulphate attack in the mortars containing
20% NZ and 30% FA. It can be seen that
the 20% NZ mortar had a relatively dense
and uniform microstructure (Figure 12a)
with limited microcracking, which is con-
sistent with its higher sulphate resistance,
as also indicated by its observed lower
expansion level (Figure 5). In contrast,
the 30% FA mortar seemed to have been
majorly affected by sulphate attack, as the
sample exhibited extensive microcracking
(Figure 12b). However, it may be recalled
that sulphate resistance of the 30% FA
mixture was still higher than that of the
control OPC mixture, as expected.
Figures 12c and 12d give elemental
analyses of mortars containing 20% NZ
and 30% FA, respectively. The EDS analyses
results show that the 20% NZ mortar
formed Na-rich ettringite within cracks, in
Volume 64 Number 4 December 2022
contrast with the 30% FA mortar in which
normal ettringite occurred. The phase
alteration of ettringite involving alkali (Na)
enrichment may have significantly con-
tributed to the observed higher sulphate
resistance behaviour of the NZ mixture.
CLASSIFICATION OF THE
NATURAL ZEOLITE POZZOLAN
ASTM C618 (2015) specifies the criteria
that must be met for the purpose of clas-
sifying natural pozzolans as supplementary
cementitious materials or cement extenders.
In Table 5, experimental results of NZ are
compared against the standard specification
requirements. It can be seen that use of the
zeolite at 20% NZ proportion meets the
criteria for ‘Class N’ pozzolan, giving values
that are well within acceptable prescribed
limits. As seen in Table 5, the criteria
Journal of the South African Institution of Civil Engineering
Figure 12 S EM features of sulphate attack in mortars: (a) 20% NZ microstructure, (b) 30% FA microstructure, (c) 20% NZ, EDS, (d) 30% FA, EDS:
NZ (natural zeolite), FA (fly ash)

comprise chemical and physical require-
ments. Pozzolanic aluminosilicate materials
of natural origin are required to contain
SiO2 + Al2O3 + Fe2O3 content e­ xceeding
70%, of which NZ had 76.6% content,
mostly being SiO2 (61.3%) and Al2O3 (12.9%)
(Table 1). In clays and other materials, LOI
typically indicates the content of carbonates
and/or bound water, while carbon content
is the key impurity indicated by LOI of FA
(Chen et al 2019). In the present study, NZ
had somehow higher LOI of 15.8% > 10%,
most likely due to the presence of bound
water from hydroxyl-bearing minerals.
As expected, the effect of pozzolan
on physical properties depends on its
proportion in the mixture. From the
various experiments conducted and
discussed under the above two sections
titled Experimental study and Results and
discussion (pages 3–12), it was found that
20% NZ content provided an optimum
overall performance across all the crucial
characteristics and properties investigated.
Accordingly, the ASTM C618 (2015) criteria
presented in Table 5 are compared against
the performance of mixtures containing
20% NZ.
The main mechanical properties of
importance in the evaluation of pozzolans

Journal of the South African Institution of Civil Engineering Volume 64 Number 4 December 2022 13
Table 5 Classification of natural zeolite (NZ) based on the criteria specified in ASTM C618 (2015)
ASTM C618 Natural zeolite
Class N 20% NZ

SiO2 +Al2O3 +Fe2O3, min (%) 70.0 76.6

SO3, max 4.0 –

Moisture content, max (%) 3.0 0.32

Loss of ignition, max (%) 10.0 15.8

Fineness, amount retained on the 45 µm sieve, max (%) 34.0 0

Strength activity index: at 7 days, min (%) 75 86

: at 28 days, min (%) 75 83

Pozzolanic activity index with lime, min (MPa) 5.5 5.67

Increase of drying shrinkage of mortar bars at 28 days, max difference in % over control 0.03 -0.016

ASR expansion of test mixture as a percentage of low-alkali cement control at 14 days, max (%) 100 65.1

ASR mortar expansion at 14 days in alkali expansion test, max (%) 0.06 0.10

Sulphate expansion of test mixture as a % of control after 160 days exposure, max % 100 59.5

are compressive strength, PAL, and drying
shrinkage. Strength gain effects of pozzo-
lans are evaluated at the early age of seven
days and at the standard age of 28 days.
It can be seen that, at seven days and 28
days, the strength activity index values of
86% and 83% for the 20% NZ/OPC blended
mixtures are well above the 75% minimum
requirement. The foregoing is consistent
with the PAL test result of 5.67 MPa which
exceeds the 5.5 MPa minimum, showing
that the NZ has sufficient pozzolanic
reactivity. Based on ASTM C618 (2015),
pozzolans may be expected to increase
drying shrinkage up to a level of 0.03%
over that of the control. Interestingly, NZ
reduced the 28-day drying shrinkage with a
significant difference of –0.016% below that
of the control, which indicates the potential
benefit of employing the pozzolan towards
crack control.
ASR and sulphate resistance responses
of mixtures are the main durability perfor-
mance characteristics that are considered
in the criteria. The high effectiveness of NZ
in mitigating both ASR and sulphate attack,
is indicated by the expansion reductions
comprising 65.1% < 100% and 59.5% < 100%
for mixtures containing the pozzolan.
The approach of EN 197-1 (2000) and
SANS 50197-1 (2013) on quality evalua-
tion of natural pozzolans is quite different
from that of ASTM C618 (2015), with the
former specifying reactive silica content
< 25% along with strength class of cement
type, as the main criteria. For employment
of the zeolite as cement extender at 10 to
20% NZ, the blend falls under CEM II/A-P
cement type. From Figure 2 and Table 3,
the 20% NZ/OPC blend gave 3-day and
7-day strength values of 16.2 and 33.9 MPa,
which meet the criteria for Strength Class
32,5 N, R. Accordingly, the 20% NZ/OPC
blend investigated in the present study
meets the main EN/SANS criteria for clas-
sification as CEM II /A-P 32.5N,R.
CONCLUSIONS
Natural zeolite (NZ) was investigated to
determine its performance and classification
for potential use as pozzolan or cement
extender. Varying proportions of 0, 10, 20
and 30% NZ were blended with CEM I
52,5N ordinary Portland cement (OPC),
then used to prepare mortar mixtures.
Various mechanical and durability proper-
ties of NZ/OPC blended mortar mixtures
were evaluated. Performance results of NZ/
OPC blended mixtures were also compared
with those of 30% fly ash (FA) mortar. Based
on experimental observations from the
study, the following conclusions are drawn:
1. Incorporation of 20% NZ in OPC met
the ASTM C618 criteria for classifica-
tion of the zeolite as ‘Class N’ pozzolan.
The 20% NZ/OPC cement blend
fulfilled the main requirements for its
classification as CEM II / A-P 32.5N,R,
based on EN 197-1 / SANS 50197-1.
2. Blending of NZ with OPC led to
reduction in workability of mixtures.
Workability reduces with correspond-
ing increase in proportion of NZ in
mixtures.
3. Consistent with the typical effects of
common pozzolans, NZ gave lower
early-age strength but higher late-age
strength development. It was observed
that the late-age compressive strength
of 20% NZ mortar was similar to that of
the control OPC mortar.
4. The NZ pozzolan significantly reduced
the drying shrinkage of mixtures – the
higher the NZ proportion used in mix-
tures, the greater the observed reduc-
tion in drying shrinkage.
5. Mortars containing 20 to 30% NZ
exhibited significant mitigation of
expansion due to alkali-silica reaction
(ASR) and were more effective in con-
trolling ASR than the 30% FA mortar.
6. Mortars containing 20 to 30% NZ were
more effective in controlling sulphate
attack than the 30% FA mixture.
REFERENCES
Addis, B 1998. Fundamentals of Concrete. Cement and
Concrete Institute: Midrand, South Africa.
ASTM C311/C311M 2013. Standard test method for
sampling and testing fly ash or natural pozzolans
for use in Portland cement concrete. West
Conshohocken, PA: ASTM International.
www.astm.org.
ASTM C1260 2014. Standard test method for potential
alkali reactivity of aggregates (mortar bar method).
West Conshohocken, PA: ASTM International.
ASTM C1437 2015. Standard test method for flow of
hydraulic cement mortar. West Conshohocken, PA:
ASTM International.
ASTM C618 2015. Standard specification for coal fly
ash and raw or calcined natural pozzolan. West
Conshohocken, PA: ASTM International.
ASTM C1012M 2017. Standard test method for length
change of hydraulic-cement mortars exposed to a
sulfate solution. West Conshohocken, PA: ASTM
International.
ASTM C305 2016. Standard practice for mechanical
mixing of hydraulic cement pastes and mortars of

14 Volume 64 Number 4 December 2022 Journal of the South African Institution of Civil Engineering
 plastic consistency. West Conshohocken, PA: ASTM
International.
ASTM C596 2018. Standard test method for drying
shrinkage of mortar containing hydraulic cement.
West Conshohocken, PA: ASTM International.
ASTM C109 2019. Standard test method for
compressive strength of hydraulic cement mortars.
West Conshohocken, PA: ASTM International.
Chen, H-J, Shih, N-H, Wu, C-H & Lin, S-K 2019.
Effects of the loss on ignition of fly ash on the
properties of high-volume fly ash concrete.
Sustainability, 11: 2704. DOI:10.3390/su11092704.
Clifton, J R & Pommersheim, J M 1994. Sulphate attack
of cementitious materials: Volumetric relations and
expansions. Gaithersburg, MD: National Institute
of Standards and Technology, Building and Fire
Research Laboratory.
Colella C, de’Gennaro, M & Aiello, R 2001. Use of
zeolitic tuff in the building industry. Reviews in
Minerology and Geochemistry, 45(1): 551–583.
Collier, N C 2016. Transition and decomposition
temperatures of cement phases: A collection
of thermal analysis data. Ceramics Silikáty,
60(4): 338–343.
Diale, P P, Muzenda, E & Zimba, J 2011. A study of
South African natural zeolites properties and
applications. Lecture Notes in Engineering and
Computer Science, 2194(1): 698–703.
Ding, J T & Li, Z 2002. Effects of metakaolin and silica
fume on properties of concrete. ACI Materials
Journal, 99(4): 393–398.
Ekolu, S O 2016. A review on effects of curing,
sheltering, and CO2 concentration upon natural
carbonation of concrete. Construction and Building
Materials, 127: 306–320.
Ekolu, S O 2018. Model for practical prediction of
natural carbonation in reinforced concrete. Part 1.
Formulation. Cement and Concrete Composites,
86: 40–56.
Ekolu, S O, Thomas, M D A & Hooton, R D 2006.
Studies on Ugandan volcanic ash and tuff.
Proceedings, 1st International Conference on
Advances in Engineering and Technology, July 2006,
Entebbe, Uganda, 75–83.
Ekolu, S O 2014. Potential South African standard
sand for cement mortar testing and research.
Proceedings, International Confereence on
Construction Materials and Structures (ICCMATS),
24–26 November, Johannesburg, 253–260.
EN (European Standard) 2000. EN197-1 2000. Cement
Part 1: Composition, specifications and conformity
Journal of the South African Institution of Civil Engineering Volume 64 Number 4 December 2022
criteria for common cements. Brussels: European
Committee for Standardization.
Jana, D 2007. A new look to an old pozzolan,
clinoptilolite: A promising pozzolan in concrete.
Proceedings, 29th ICMA Conference on Cement
Microscopy, International Cement Microscopy
Association, Quebec City, Canada, 168–206.
Kalina, R D, Al-Shmaisani, S, Seraj, S, Cano, R, Ferron,
R D & Juenger, M C G 2021. Role of alkalis in
natural pozzolans on alkali-silica reaction. ACI
Materials Journal, 118(3): 83–89.
Marantos, I, Christidi, G E & Ulmanu, M 2020.
Zeolites formation and deposits, Chapter 3. In
Ingelazakis, V J & Zorpas, A A (Eds). Handbook of
Natural Zeolites. UAE: Bentham Science. 28–51.
Mardani-Aghabaglou, A, Inan Sezer, G & Ramyar,
K 2014. Comparison of fly ash, silica fume and
metakaolin from mechanical properties and
durability performance of mortar mixtures view
point. Construction and Building Materials,
70: 17–25.
Mather, K 1978. Tests and evaluation of Portland and
blended cements for resistance to sulfate attack. In
Mehta, P K (Ed). Cement Standards – Evolution and
Trends, ASTM STP 663. West Conshohocken, PA:
ASTM, 74–78.
Mielenz, R, Witte, L & Glants, O 1950. Effect of
calcination on natural pozzolans. In Stanton, T &
Blanks, R (Eds). Symposium on Use of Pozzolanic
Materials in Mortars and Concrete. West
Conshohocken, PA: ASTM, 43–92.
Naghizadeh, A & Ekolu S O 2017a. Pozzolanic
materials and waste products for formulation of
geopolymer cements in developing countries: A
review. Concrete/Beton, 151: 22–31.
Naghizadeh, A & Ekolu, S O 2017b. Investigation of
mixture factors influencing alkali-silica reaction
in fly ash-based geopolymer mortars. Proceedings,
International Conference on Advances in
Construction Materials and Systems (ICACMS-2017),
3–8 September, Chennai, India, 395–400.
Najimi, M, Sobhani, J, Ahmadi, B & Shekarchi, M 2012.
An experimental study on durability properties of
concrete containing zeolites as a highly reactive
natural pozzolan. Construction and Building
Materials, 35: 1023–1033.
Neville, A 1981. Properties of Concrete, 3rd ed. London:
Pitman Publishing.
SANS 6245 2006. Potential reactivity of aggregates with
alkalis (accelerated mortar prism method). Pretoria:
South African Bureau of Standards (SABS).
SANS 53263-1 2011. Silica fume for concrete. Part 1.
Definitions, requirements and conformity criteria.
Pretoria: South African Bureau of Standards (SABS).
SANS 55167-1 2011. Ground granulated blast furnace
slag for use in concrete, mortar and grout. Part 1:
Definitions, specifications and conformity criteria.
Pretoria. South African Bureau of Standards (SABS).
SANS 50197-1 2013. Cement. Part 1: Composition,
specifications and conformity criteria for common
cements. Pretoria: South African Bureau of
Standards (SABS).
SANS 50450-1 2014. Fly ash for concrete. Part 1:
Definitions, specifications and conformity criteria.
(ed. 2.00). Pretoria: South African Bureau of
Standards (SABS).
Sicakova, A, Špak, M, Kozlovská, M & Kováč, M
2017. Long-term properties of cement-based
composites incorporating natural zeolites as a
feature of progressive building material. Advances
in Materials Sciences and Engineering, Special issue,
29–31.
Shafaatian, S, Akhvan, H, Maragechi, H & Rajabipour,
F 2013. How does FA mitigate alkali-silica reaction
(ASR) in accelerated mortar bar test (ASTM
C1567)? Cement and Concrete Composites,
37: 143–153.
Shi, Z, Geng, G, Leemann, A & Lothenbach, B 2019.
Synthesis, characterization, and water uptake
property of alkali-silica reaction products.
Cement and Concrete Research, 121: 58–71.
DOI: 10.1016/j.cemconres.2019.04.009.
Tchadjie, L N & Ekolu, S O 2018. Enhancing the
reactivity of aluminosilicate materials toward
geopolymer synthesis. Journal of Material Science,
53: 4709–4733.
Tchadjie, L N, Ekolu, S O, Quainoo, H & Tematio,
P 2021. Effect of incorporation of mechanically
activated bauxite on the strength and
microstructure of volcanic ash-based geopolymer
mortars. Journal of Building Engineering, 41(7):
102384. DOI: 10.1016/j.jobe.2021.102384.
Valipour, M, Pargar, F, Sherkarchi, M & Khani, S
2013. Comparing a natural pozzolan, zeolite, to
metakaolin and silica fume in terms of their effect
on the durability characteristics of concrete:
a laboratory study. Construction and Building
Materials, 41: 879–888.
Yilmaz, B, Oteyeka, U B & Uz, V 2007. Properties of
zeolite tuff (clinoptilolite) blended portland cement.
Building and Enviroment, 42(11): 3808–3815.
15
TECHNICAL PAPER Application of a modern
Journal of the South African
Institution of Civil Engineering laser technique to evaluate
ISSN 1021-2019 (print) | ISSN 2309-8775 (online)
Vol 64 No 4, December 2022, Pages 16–28, Paper 1361 the performance of
GCULISILE MVELASE (AMSAICE) currently
works for the Council for Scientific and
recycled railway ballast
Industrial Research (CSIR) in South Africa.
She has more than 15 years of experience in
geotechnical investigation of track G M Mvelase, J K Anochie-Boateng, P J Gräbe
substructure and railway building materials.
Her research interests cover 3D mapping
techniques (GIS, LiDAR, drone and ballast 3D
scanning), prediction and performance behaviour of geomaterials, The performance of railway track structure is influenced by the ballast shape properties including
including materials evaluation and testing (i.e. use of non-potable water, roundness, flatness, elongation, sphericity, angularity and surface texture. The challenge is how to
fly ash and marginal soils in pavement construction). This article accurately measure the irregular shapes of ballast materials and directly link them to performance.
emanated from her MEng research project in the Department of Civil
In this paper, a modern three-dimensional laser technique was used to determine shapes of
Engineering in the Faculty of Engineering, the Built Environment and
Information Technology (EBIT) at the University of Pretoria. freshly produced crushed ballast and recycled ballast sampled from the heavy-haul coal line
in South Africa. The objective was to investigate the effect of ballast shapes on settlement (i.e.
Contact details:
CSIR Smart Mobility permanent deformation). All five ballast materials were scanned in the three-dimensional laser
PO Box 395, Pretoria, 0001, South Africa scanning system and the data collected was processed to reconstruct three-dimensional models
T: +27 12 841 4644 | E: gmvelase@csir.co.za of the ballast particles. The results obtained were used to develop a chart to classify ballast shapes
and link these shapes physically with settlement determined from a triaxial testing programme.
PROF JOSEPH ANOCHIE-BOATENG (Pr Eng,
Based on the triaxial test results, new empirical models were developed to determine settlement
MSAICE) is an associate professor in the
Department of Civil Engineering in the on the route corridor of the heavy-haul coal line. It is anticipated that outcomes of this study
Faculty of Engineering, the Built would assist with quality assessments and railway ballast maintenance in the field.
Environment and Information Technology
(EBIT) at the University of Pretoria. His
research focus areas are asphalt technology,
advanced imaging techniques (i.e. use of 3D INTRODUCTION depends on the effectiveness of the ballast
laser and high-resolution cameras to characterise the morphology of All over the world railway systems play a in providing drainage, stability, flexibility,
crushed stones for the building of roads and railways), forensic significant role in hauling bulk commodities uniform support to the superstructure
investigation of failures in flexible pavements, and long-term
performance of road pavements. Prof Boateng holds a PhD in Civil
to ports, and transporting freight along and distribution of the track loading to the
Engineering from the University of Illinois at Urbana-Champaign, USA. major corridors. Transnet Freight Rail (TFR) subgrade, and facilitating maintenance. An
Contact details:
is the agency that manages and operates increase in axle loads and traffic density
Department of Civil Engineering two heavy-haul lines (i.e. the coal line from increases the rate of settlement of the track.
University of Pretoria Mpumalanga to KwaZulu-Natal Province To thus keep settlement within permissible
Private Bag X20, Hatfield, 0028, South Africa
and the iron ore line from Northern Cape limits, stresses in the subgrade should be
T: +27 12 420 2978 | E: joseph.anochieboateng@up.ac.za
to Western Cape Province) in South Africa. reduced suitably to ensure stability of the
PROF HANNES GRÄBE (Pr Eng, FSAICE,
Each year, a large portion of TFR’s budget track (Li et al 2015).
FSAAE) is an associate professor in the is allocated to maintenance and renewal Indraratna et al (2006) indicated that
Department of Civil Engineering in the Faculty of track ballast materials to sustain per- the maintenance cost of track ballast can
of Engineering, the Built Environment and
formance of the railway track system. A significantly be reduced if there is a better
Information Technology (EBIT) at the
University of Pretoria. He holds the Chairs in common problem in the rail industry is the understanding of the physical and mechani-
Railway Engineering and Railway Safety, degradation of ballast under cyclic loading, cal properties of ballast. Selig and Waters
sponsored by Transnet Freight Rail and the especially on these heavy-haul lines. (1994) demonstrated that ballast contributes
Railway Safety Regulator, respectively. His research focuses on the behaviour
A typical track consists of super- most of the substructure settlement, even
of railway foundation materials subjected to cyclic loading and the
characterisation of track component performance and behaviour through structure (rails, fastenings and sleepers) though one of the functions of ballast is to
field and laboratory experimentation. He combines his experience in track and substructure (ballast, subballast and maintain good track geometry. Given that
technology, soil mechanics and advanced laboratory testing for the formation). The function of the ballast is to ballast is responsible for nearly all the track
development of novel condition monitoring techniques, maintenance
models and numerical analysis of track structures. He holds a PhD from the
transfer the load from the superstructure settlement, every effort to reduce settlement
University of Southampton (UK) in Geotechnical Engineering. to the substructure. Unlike subbase/base is focused on the ballast layer.
Contact details:
layers in pavement structures, ballast Ballast is usually composed of blasted
Department of Civil Engineering particles tend to break down, powder and rock originating from igneous, meta-
University of Pretoria deteriorate under repeated wheel load- morphic or well-cemented sedimentary
Private Bag X20, Hatfield, 0028, South Africa
ing from heavy freight trains (Moaveni rock quarries. Crushed angular hard rock
T: +27 12 420 4723 | E: hannes.grabe@up.ac.za
2015). Performance of the track system having a uniform grading and being free

Keywords: railway ballast, shape characteristics, settlement, Mvelase GM, Anochie-Boateng JK, Gräbe PJ. Application of a modern laser technique to evaluate the performance of recycled railway ballast.
J. S. Afr. Inst. Civ. Eng. 2022:64(4), Art. #1361, 13 pages. http://dx.doi.org/10.17159/2309-8775/2022/v64n4a2
modern laser, triaxial
The Journal of the South African Institution of Civil Engineering is a peer-reviewed, open-access journal licenced under a Creative Commons Attribution Licence
(CC BY-NC-ND) and is distributed internationally. The content of this journal is available to users free of charge. However, copyright remains with the South African
16 Institution of Civil Engineering. Permission for reproduction of content, with appropriate reference to the journal, can be obtained from the administrative editor.
of dust has been considered as acceptable
ballast material. The major challenge is
that ballast particles have irregular and
non-ideal shapes with variable surface tex-
tures. Improved techniques for measuring
and analysing ballast shapes are, however,
available to explore. Hayakawa and Oguchi
(2005), and Tolppanen et al (2008) reported
that digital modelling of gravel particles
based on three-dimensional (3-D) laser
scanning could be useful, reliable, repeat-
able and relatively fast to evaluate the
properties of ballast material. In addition,
Moaveni et al (2013) describe an image-
processing and analytical technique to
better quantify particle shape. Moaveni
(2015) modified and applied the Enhanced
University of Illinois Aggregate Imaging
Analyzer (E-UIAIA) to evaluate railroad
ballast field degradation. In the E-UIAIA,
aggregates are individually placed on a
conveyor belt, with three orthogonally
positioned cameras to capture the particle
projections. E-UIAIA can distinguish
flat and elongated aggregates, and auto-
matically calculates angularity and surface
texture, as well as surface area and volume
for a wide range of ballast particles. Recent
studies in South Africa have strongly
recommended the use of modern 3-D laser
scanning and numerical modelling tech-
niques to improve measurements of the
shapes of aggregates and ballast materials
(Mvelase et al 2017; Anochie-Boateng et al
2013; Mvelase et al 2012; Anochie-Boateng
et al 2012). The use of laser improves accu-
racy and reliability of results. An accurate
measurement of the shapes is important for
the development and revision of specifica-
tions for quality control and quality assur-
ance of ballast.
This paper focuses on the effect of
shapes determined from a modern 3-D
laser scanning system on settlement of
railway ballast. The selected ballast materi-
als are being investigated by TFR for their
suitability in heavy-haul railroads in South
Africa. The effect of ballast shape on settle-
ment performance was further investigated
through triaxial testing.
BALLAST SHAPE PROPERTIES
Parameters of ballast particle shape
In order to describe the particle shape in
detail, there are a number of terms, quanti-
ties and definitions used in the literature.
Shape is a first-order property that reflects
variations in the overall shape of a particle
Journal of the South African Institution of Civil Engineering Volume 64 Number 4 December 2022
(Barrett 1980). The physical dimensions,
surface area and volume have been used to
compute index parameters commonly used
to describe the shapes of aggregate and
ballast particles (Anochie-Boateng et al
2013). Although there are some differences
in their precise definitions, the long, inter-
mediate and short dimensions of a particle
are frequently used to summarise its shape.
These three dimensions are sometimes
referred to as the L, I and S diameters
respectively. L, S and I can be obtained
accurately from 3-D scanned models. It is
possible to measure these dimensions man-
ually using callipers (Krumbein & Pettijohn
1938; Krumbein 1941), although this is
time-consuming, and any set of measure-
ments may be subject to user variation.
Kuo and Freeman (2000) defined two
fundamental parameters to describe the
shape of a rock aggregate as elongation
and flatness ratios. Flatness ratio is defined
as the ratio of the particle’s shortest to
intermediate dimension (Equation 1: Sneed
& Folk 1958). Elongation ratio is defined as
the ratio of the particle’s longest dimension
in the plane perpendicular to the interme-
diate dimension (Equation 2: Sames 1966).
The shape factor of an aggregate particle
can be related to flatness and elongation
characteristics (Equation 3: Aschenbrenner
1956). The sphericity in Equation 4 was
described by Aschenbrenner (1956).
S thin dimensions that can align and form
Flatness (F) = (1)
I planes of weakness in both vertical and
I appears to increase abrasion and breakage,
Elongation (E) = (2)
L increase permanent strain accumulation
SL
Shape factor (SF) = (3)
I2
12.8(F 2 E)⅓
Ψ= (4)
{1 + F(1 + E) + 6(1 + F 2(1 + E 2))}
Where:
L = longest dimension of a particle
I = intermediate dimension of a particle
S = shortest dimension of a particle
Ψ = sphericity
Furthermore, Zingg (1935) proposed a
classification for shapes, and established
a terminology that separates flat, cubic,
ellipsoid and elongated shapes with a value
of 0.67. This chart is a graphical approach
to relate particle dimensions. Lines of equal
sphericity based on Equation 4 are added to
the Zingg diagram.
Sphericity
Lim and McDowell (2005), and Hayakawa
and Oguchie (2005) quantified sphericity
based on the surface area and volume prop-
erties of the aggregate. Sphericity is defined
as a function of the ratio of the surface area
of the particle to its volume, as shown in
Equation 5. Ballast particles have irregular
and non-ideal shapes. It is therefore diffi-
cult to obtain a direct measurement of the
surface area and volume properties using
the traditional methods for quantifying the
shape properties of aggregates.
3√36V 2
Sphericity = (5)
A
Where:
A = surface area
V = volume
Among the shape indices described above,
sphericity, flatness, elongation and shape
factor were computed directly using the data
obtained from a 3-D laser scanning device.
Current standard test to
measure ballast shape index
Raymond (1985) reported that most speci-
fications restrict the percentage of flaky
particles of which the aspect ratio exceeds
3, and exclude particles with an aspect
ratio exceeding 10. Flaky particles cannot
be used as ballast given their long and very
lateral directions. The use of flaky particles
under repeated load, and decrease stiffness
(Selig & Waters 1994).
Flakiness index test procedures are con-
tained in Technical Methods for Highways
(TMH 1) Method B3 (TMH 1 1986). Under
the new South African National Standards
(SANS), the method will be replaced by
SANS 3001-AG4 (SANS 2015). The test
provides an indication of the flatness of
ballast particles. The test procedure starts
with performing grading analysis on the
aggregate sample to be tested. Each ballast
particle retained on a specific sieve size
is then passed through a corresponding
rectangular slot of a flakiness gauge. The
particles passing the slots are regarded as
flaky, whereas particles that do not pass are
considered non-flaky. Flakiness index (%) is
calculated by dividing the mass of ballast
particles passing the slots by the total mass
of the sample.
17
Figure 1 Current test method using a flat gauge (courtesy TFR)

In the TFR specification, flakiness index

is defined as the ratio of the total mass
passing bar sieve slots to the total mass of
aggregate retained on three specific sieve
sizes. Figure 1 shows a photograph of the
flakiness gauge apparatus.
Mathematically, the flakiness index (FI)
of a ballast material can be represented by
Equation 6 as follows:

⎫ Mp ⎫
FI = ⎪ ⎪ × 100(6)
⎭ MT ⎭

Where:
Mp = total mass of aggregate passing a
bar sieve slots
MT = total mass of aggregate retained on a
specific sieve size (grading analysis)

EXPERIMENTAL PROGRAMME
AND TESTING

Material sourcing
Four recycled ballast samples from the
heavy-haul coal line in South Africa were
selected for this study. A fresh ballast
sample from a quarry was included to act Figure 2 Position of ballast sampling below the sleeper

18 Volume 64 Number 4 December 2022 Journal of the South African Institution of Civil Engineering
 Recycle ballast Km 9

Recycle
ballast Km 17 Richards Bay Coal Line

26 t Achslast 25 kV 50 Hz ~
26 t Achslast 3 kV =
20 t Achslast 3 kV =
20 t Achslast 25 kV 50 Hz ~
20 t Achslast Dieselbetrieb

Track segment
Stations Recycle ballast Km 31
Towns Recycle ballast Km 32
Km posts

Figure 3 Heavy-haul rail network between Vryheid and Richards Bay showing recycled ballast sample locations

as reference material. Recycled ballast
is referred to as reclaimed material that
is re-used in the track structure during
rehabilitation. The parent rock of recycled
ballast remains mostly the same as that of
fresh, crushed ballast. Because dolerite is
the most commonly used ballast type on
the heavy-haul line, preliminary testing for
the reproducibility of tests was conducted
using a dolerite obtained from a quarry in
South Africa. The same dolerite was used
in the main testing programme. The four
recycled ballast samples were selected from
the coal line section where repetitive bal-
last tamping and track geometry deforma-
tion (known as cant loss on curves) were
noted by line maintenance managers of
TFR. The site was chosen in consultation
with the Maintenance Depot experts and
was based on the goals of the research.
None of these selected curves were situ-
ated just before a signal, and no other rail
surface defects such as locomotive wheel
skid marks or ballast white spots were
noted. Samples were taken below the sleeper
with special focus on collecting rounded
particles for the laboratory analyses on the
sites provided by the maintenance depot as
indicated in Figure 2. The visual examination
indicated that about 90% of recycled ballast
comprised semi-angular crushed rock frag-
ments, while the remaining 10% consisted of
semi-rounded river gravels. These materials Table 1 P
 hysical characteristics of dolerite
were chosen so that a comparison of their ballast
physical characteristics could be made. South
The coal line is electrified to 25 kV AC Standard Test
African
ballast tests value
and supports 26 t/axle loading. The test sec- specification
tion forms the southern part of the coal line, Void content > 40 43
carrying mainly heavy-haul traffic with some
Relative density > 2.5 2.7
general freight. Figure 3 shows the recycled
ballast sample locations on the heavy-haul Grading pass or fail pass
coal line between Vryheid and Richards Bay, Flakiness index < 30 8
as investigated in this paper. The recycled
Los Angeles < 22 12
ballast was subjected to a traffic load of 2 738
million gross tons (MGT). This is the total Mill abrasion <7 6
tonnage carried from the inception of the Absorption <1 0.6
line in 1976, calculated based on 76 million
Weathering <5 1.2
tonnes handled by the Richards Bay Coal
Terminal annually (Kuys 2011).
Los Angeles Abrasion Test is subjected
Properties of the fresh ballast to further tests to evaluate shape, surface
For the purpose of this study, physical characteristics, grading and unit mass. Based
property tests were performed on the fresh on the results presented in Table 1, it can be
ballast, and the results were compared to the concluded that the tested sample was suitable
TFR ballast specification (TFR 2011). These for use as ballast in heavy-haul lines.
tests were conducted in the geotechnical
engineering laboratory of the TFR Track Selection of ballast particles
Testing Centre to determine whether the for laser scanning
materials were suitable for use as railway bal- To establish whether the scanning technique
last. Table 1 summarises the results of eight would lead to a reasonable prediction of
physical properties of the fresh ballast. These ballast characteristics, a quartering sam-
tests provide basic guidelines for accept- pling approach was used to select particles
ance of a given material as potential railway for scanning. Prior to quartering, grading
ballast. In general, ballast that satisfies the analyses were conducted on the samples to

Journal of the South African Institution of Civil Engineering Volume 64 Number 4 December 2022 19
 100 100
(a) Recycled ballast Km 9 (b) Recycled ballast Km 17
Initial mass (kg) = 21.3 Initial mass (kg) = 21.5
80 Scanned mass (kg) = 5.9 80 Scanned mass (kg) = 5.7

60 60
% passing

% passing
40 40

20 20

0 0
0 20 40 60 80 0 20 40 60 80
Sieve size (mm) Sieve size (mm)

100 100
(c) Recycled ballast Km 31 (d) Recycled ballast Km 32
Initial mass (kg) = 21.9 Initial mass (kg) = 21.7
80 Scanned mass (kg) = 5.7 80 Scanned mass (kg) = 6.6

60 60
% passing

% passing
40 40

20 20

0 0
0 20 40 60 80 0 20 40 60 80
Sieve size (mm) Sieve size (mm)

100 Legend (a)–(d)

(e) Freshly crushed ballast determine particle distribution according to
Spec limits the TFR specifications. These samples were
Initial mass (kg) = 44.4
80 Scanned mass (kg) = 7.3 Recycled ballast then quartered to obtain the number of par-
Recycled ballast ticles that were suitable for the shape char-
60 (scanned) acterisation. The samples were dry-sieved
% passing

using a 465 mm diameter sieve set of 63, 53,

40 Legend (e) 37.5, 26.5, 19, 13.2 and 9.5 mm. The reduced
samples were also washed and oven-dried for
Spec limits
20 scanning purposes. Figure 4 shows the grad-
Freshly crushed
ballast ing results fitted to the TFR specifications.

0 Freshly crushed
0 20 40 60 80 ballast Scanning of ballast particles
Sieve size (mm) (scanned) The 3-D laser devices at the Council for
Scientific and Industrial Research (CSIR) and
Figure 4 G
 rading analysis of samples of scanned ballast materials: (a) recycled ballast Km 9, TFR have been calibrated to determine shape
(b) recycled ballast Km 17, (c) recycled ballast Km 31, (d) recycled ballast Km 32, and surface properties of different types of
(e) freshly crushed ballast aggregate recycled ballast (Anochie-Boateng

Table 2 Representative samples of scanned materials

Freshly crushed Recycled ballast Number of particles
Sieve size
ballast Km 9 Km 17 Km 31 Km 32 per sieve size
9.5 mm 10 5 6 3 – 24
13.2 mm 7 6 10 2 3 28
19.0 mm 10 12 11 10 5 48
26.5 mm 10 30 28 14 22 104
37.5 mm 10 9 20 29 25 93
53.5 mm 10 3 1 1 1 16
63.0 mm 1 – 2 – 4 7
Scanned 58 65 78 59 60 320

20 Volume 64 Number 4 December 2022 Journal of the South African Institution of Civil Engineering
 QQ New crushed ballast Sample preparation Ballast grading Scanning
(stockpile) QQ Riffle sample to QQ Sieve initial mass QQ Place prepared
Ballast
sampling obtain approx 30 kg QQ Sieve reduced mass
ballast particle on
QQ Recycled ballast QQ Quartering to obtain QQ Separate each
the laser turntable
(track) scannable mass sample retained on QQ Select plane

QQ Wash and dry each sieve scanning mode

reduced sample QQ Process scanned

data

3D model Merge
Combine Merge
ballast particle align

Figure 5 Process used for 3D laser scanning of ballast particles

2014). All the ballast particles used for this

study were scanned in accordance with the
CSIR Guideline for Scanning of Aggregates
and Ballast Particles (Anochie-Boateng 2014).
The scanning system uses multiple scans to
produce high reliability and complete 3-D
representation of the particle being scanned.
Generally, in-between scans, the ballast parti-
cle is oriented along a different axis to ensure
that point clouds are obtained from as many
different directions as possible; this ensures
that data is obtained from all sides of the par-
ticle. After multiple scans are obtained, the
individual scans are brought into a common
(a) Actual crushed ballast (b) Modelled crushed ballast reference system through a process that is
usually referred to as alignment or registra-
tion. After the scans are registered, they are
subsequently merged to create a complete 3-D
ballast model of the ballast particle. The total
number of 320 particles, with their corre-
sponding sieve sizes, are presented in Table 2.
Figure 5 shows a schematic representa-
tion of the process used for the scanning
of the 3-D ballast models. Figure 6 shows
photographs of actual crushed ballast and
models of the scanned ballast. From these
images it is clear that the laser-scanning
technique is capable of producing models
that are similar to the actual particle. It
(c) Actual recycled ballast (d) Modelled recycled ballast
is therefore reliable to analyse the ballast
Figure 6 A
 ctual particles versus 3-D laser modelled recycled ballast: (a) actual crushed ballast, particles to obtain an indication of their
(b) modelled crushed ballast, (c) actual recycled ballast, (d) modelled recycled ballast shape properties.

Journal of the South African Institution of Civil Engineering Volume 64 Number 4 December 2022 21
Discussion of flatness and
elongation results
As presented earlier in Table 2, scanned
particles involved about 320 particles
between 9.5 mm and 63 mm sieve sizes.
This data was used in the development of a
classification system. Dimensions (longest,
intermediate and shortest) of the individual
ballast particles were obtained directly

Shorted dimension (S)

from the three-dimensional bounding box
shown in Figure 7.
The following steps were followed to
compute flatness and elongation ratios
from the scanned results:
QQ The flatness ratio of an individual bal-
(L)
last particle was computed by dividing io n
Inte ens
rme d im
the shortest dimension by the interme- diat g es t
e di
m en Lon
diate dimension (Equation 1). sion
QQ The elongation ratio of an individual
(I)

ballast particle was computed by divid-

ing the longest dimension in the plane
perpendicular to the intermediate
dimension (Equation 2).
QQ Furthermore, Zingg (1935) proposed a
classification for shapes, and established
a quadrant that separates the terms
with a value of 0.67 to each other.
QQ The shape factor is computed by the
ratio of flatness and elongation in
Equation 3.
QQ The sphericity using elongation and
flatness is shown in Equation 4.
Discussion of shape
classification results
The classification methodology is based
on measuring the shape characteristics of
recycled ballast from four different positions
and one freshly crushed ballast sample using
a 3-D laser scanner. Scanned particles were
visually classified as four shapes (cubic,
flaky, elongated or flat and elongated). The
chart shown in Figure 8 was used to dis-
Figure 7 Properties of a 3-D laser-based modelled ballast particle (courtesy CSIR)
be scattered within the shaded area. Any
particle scattered below the shaded area is
too flaky or flat and elongated, while any
particle scattered above the shaded area is
too spherical.
Several rock types may fracture to flat
and/or elongated to cubic particles when
crushed. It is easy to determine the main
shape of the particles as cubic, flaky, elon-
gated and flat and elongated. A desirable
ballast particle should have as many sides
as possible (i.e. cubic). This will ensure
1.0
0.9
0.8
0.7
Elongation ratio
that different ballast particles can interlock
and remain in that stable position. Flaky
particles, on the other hand, have large
surfaces which reduce the void spaces and
can break under load.
One can more easily evaluate the
shapes of particles by plotting the ratios of
dimensions, elongation ratio and flatness
ratio. These do not present the size of the
particles but purely the shape relation-
ship of each individual particle. Figure 9
shows the total percentage of particles
SF = 0.33 SF = 0.67
SF = 1.00
Ψ = 0.95
Flaky = 122 Cubic = 98
SF = 1.50

tinguish these shapes. The sphericity value 0.6

gives a very good indication on the propor- Ψ = 0.90
0.5
tions of particle dimensions. However, one
cannot determine whether the sample has 0.4
flat, elongated or flat and elongated particles SF = 3.00
using sphericity alone. 0.3 Ψ = 0.80
When all scanned particles from the
0.2
four shapes were combined, grouped Ψ = 0.70
samples had the highest occurrence of 0.1 Ψ = 0.60
flaky particles, namely 122 in number, Ψ = 0.50
Flat & Elongated = 43 Elongated = 57
0
followed by the cubic particles (98), elon- 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
gated (57) and flat and elongated (43), as Flatness ratio
shown in Figure 8. The shaded area in the Recycled ballast Km 9 (mean value) Recycled ballast Km 17 (mean value)
chart represents ballast shape with a limit Recycled ballast Km 31 (mean value) Recycled ballast Km 32 (mean value)
of SF ranging between 0.67 and 1.5 and Recycled ballast Km 32 (mean value)
sphericity between 0.8 and 0.9. Therefore,
the recommended average shape should Figure 8 A shape classification chart for ballast material

22 Volume 64 Number 4 December 2022 Journal of the South African Institution of Civil Engineering
 that fall into the four different quadrants,
Percentage of particles (%) namely cubic, flaky, elongated or flat and
0 10 20 30 40 50 elongated particles, at the sampling points.
The sample with the highest percentage
of cubic particles was the recycled ballast

Cubic
Km 31, followed by the recycled ballast
Km 32, recycled ballast Km 17 and recycled
ballast Km 9. This differed from the results
of freshly crushed ballast where most of the
Flaky

ballast particles were found to be slightly

flat and elongated, possibly due to the dif-
ferent techniques that the quarry uses to
crush stone. This could be a reason why
Elongated

cubic particles are mostly preferred over

Flat &

flaky and elongated particles, which tend to

break under load. Li et al (2015) preferred
cubic particles for stable and strong ballast,
whereas elongated, flaky, smooth shapes
Elongated

should be avoided.

Flat and elongated particles

Recycled ballast Km 9 Recycled ballast Km 32 The longest, intermediate and shortest
Recycled ballast Km 17 Freshly crushed ballast dimensions of the ballast particles were
Recycled ballast Km 31 used to compute the flat and elongation
ratio using Equations 1 and 2. This means
Figure 9 Percentage of particles at sampling points that the recycled ballast samples were
expected to have higher flatness ratio val-
Table 3 Statistical parameters for flatness ratio and elongation ratio ues, greater than 0.66, compared to freshly
Average Standard deviation crushed ballast. For equal dimensional
Sample description ballast particles, the flat and elongation
Flatness ratio Elongation ratio Flatness ratio Elongation ratio
ratio approaches a minimum value of 1.
Recycled ballast Km 9 0.70 0.72 0.18 0.14 The flat and elongation ratio statistical
Recycled ballast Km 32 0.66 0.78 0.16 0.12 analysis for the five types of ballast studied
Recycled ballast Km 31 0.69 0.80 0.13 0.11 are presented in Table 3. The average
flatness ratio of recycled ballast Km 9 was
Recycled ballast Km 17 0.66 0.77 0.17 0.13
the highest, followed by recycled ballast
Freshly crushed ballast 0.65 0.66 0.21 0.16
Km 32. The average flatness ratio of freshly
crushed ballast was the lowest, followed by
recycled ballast Km 17, Km 31 and Km 32.
100 It can be seen that the flatness ratios of all
90
four recycled ballast samples were compa-
rable to the fresh ballast, although a little
80 bit more. However, elongation ratios of the
70 four recycled materials were higher than
% of ballast particles

the fresh ballast. This implies that broken

60 particles become more elongated, which
50 could affect the performance of the ballast
layer. The distributions of the flatness ratio
40
and elongation ratio of all the studied bal-
30 last samples are plotted in Figures 10 and
11, respectively. The results indicate that
20
the particles were slightly elongated and
10 moderately flat.

0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Sphericity results
Flatness ratio (S/I) The surface area and volume of individual
Recycled ballast Km 9 Recycled ballast Km 31 Freshly crushed ballast stone particles were used to compute the
Recycled ballast Km 17 Recycled ballast Km 32 sphericity of the ballast materials using
Equation 5. The sphericity approaches
Figure 10 Distributions of flatness ratio a maximum value of one (1) for round

Journal of the South African Institution of Civil Engineering Volume 64 Number 4 December 2022 23
 sphericity values, followed by recycled
100 ballast Km 31, recycled ballast Km 9 and
90
recycled ballast Km 17.
It was also observed that the distribution
80 of sphericity values for different samples
70 may differ significantly, despite their average
% of ballast particles

values being closer to each other. Therefore,

60 a single property such as flakiness index
50 cannot accurately differentiate shape
properties of the ballast samples. The use
40
of the laser scanning technique provided
30 more information required to distinguish
the shapes of the different samples used in
20
this study. The plots of sphericity, elonga-
10 tion ratio and flatness ratio presented in
Figures 10, 11 and 12 indicate that the
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 recycled ballast differs from freshly crushed
Elongation ratio (I/L) ballast in terms of its shape.
Recycled ballast Km 9 Recycled ballast Km 31 Freshly crushed ballast
Recycled ballast Km 17 Recycled ballast Km 32 Correlation of flakiness index
with ballast shape indices
Figure 11 Distributions of elongation ratio The flakiness index values were correlated
with the average sphericity and flatness
shaped aggregate particles. Figure 12 objects. These spheres had different sizes ratio (see Figure 13). A reasonably good
shows the distributions of the sphericity and were manufactured using different correlation was found between the sphe-
computed using surface area and volume. types of material (steel, ceramic, rubber ricity and flakiness index (R2 = 0.604).
The sphericity distributions are plotted for and plastic). All the spherical objects plot- However, a poor correlation was found
the 320 particles scanned. Included in the ted closer to a value of one (1). Recycled between flakiness ratio and flakiness index
figure are sphericity values of 14 spherical ballast Km 32 had relatively higher (R2 = 0.360). The recycled ballast samples

100

90

80

70
% of ballast particles

60

50

40

30

90% of the fresh ballast has

20
sphericity greater than 0.6

3 36πV 2
10 Sphericity =
A

0
0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1
Sphericity
Recycled ballast Km 9 Recycled ballast Km 17
Recycled ballast Km 31 Recycled ballast Km 32
Freshly crushed ballast Spheres

Figure 12 Distributions of sphericity

24 Volume 64 Number 4 December 2022 Journal of the South African Institution of Civil Engineering
 14 14
y = –115.398x + 91.324 y = –48.372x + 38.706
12 R2 = 0.604 12 R2 = 0.360
Flakiness index (%)

Flakiness index (%)

10 10

8 8

6 6

4 4

2 2

0 0
0.6 0.7 0.8 0.9 1.0 0.6 0.7 0.8 0.9 1.0
Sphericity Flatness ratio

Recycled ballast Km 9 Recycled ballast Km 17 Recycled ballast Km 31 Recycled ballast Km 32 Freshly crushed ballast

Figure 13 Correlation of flakiness index with sphericity and flatness ratio

at Km 31 and KM 9 showed a higher flat-
ness ratio with a lower flakiness index
and hence appear to contribute to the
poor correlation.
TRIAXIAL TESTING PROGRAMME
A proper evaluation of the behaviour of
railway ballast requires that the specimens
must be prepared to the particle size repre-
sentative of the field conditions (Tutumler
et al 2011). Repeated load triaxial testing
was conducted in accordance with the TFR
in-house triaxial testing protocol developed
for granular material (Arangie 1995). The
specimen dimensions, i.e. a height (H) of
385 mm and a diameter (D) of 210 mm,
corresponding to an acceptable H/D ratio
0 10 000
0
0.5
of 1.85, were used for the triaxial tests. A
210 mm diameter and 380 mm tall triaxial
specimen yields a volume of 0.013 m3.
The chamber was pressurised, carefully
avoiding applying the full confined pres-
sure before removal of the vacuum. The
vacuum pipe was ventilated to prevent any
back-pressure.
In this study, a deviator stress of
101 kPa and a confining pressure of 90 kPa
were applied to the ballast samples to
determine settlement. The cell pressure of
90 kPa is representative of field conditions
on the coal line in KwaZulu-Natal. The
axial loading values were obtained based
on Geotrack software (Arangie 1995) to
represent axle loads of 26 ton (~ 94 kPa)
and 34 ton (~121 kPa) for the current and
Number of cycles, N
20 000 30 000 40 000
the predicted axle load due to an increasing
coal export market (Arangie 1995). The
30-ton (~107 kPa) loading represents the
locomotive axle load.
Repeated load triaxial testing was
performed on the five ballast samples. This
test closely resembles the load environ-
ment and failure mechanisms of ballast in
track. All tests were conducted at the CSIR
pavement materials testing laboratory.
During testing, a repeated haversine load
of 0.1 seconds followed by a 0.9-second rest
period is applied to the ballast sample.
For the first stage, the samples were
loaded at 90 kPa confining pressure and
100 kPa axial loading with 0.1-second load
duration followed by a 0.9-second rest
period. This was immediately followed
by increasing the repeated axial stress
level to 140 kPa under the same confining
stress condition of 90 kPa for an additional
50 000 50 000 load application. The samples were
loaded at 90 kPa confining pressure and
100 kPa, then at 140 kPa axial loading.

1.0
DISCUSSION OF TRIAXIAL
TEST RESULTS
Settlement, SN1 (mm)

1.5

2.0
First-stage settlement
Figure 14 shows that the highest settlement
2.5 was found in recycled ballast Km 32, fol-
lowed by recycled ballast Km 17, recycled
3.0 ballast Km 9, recycled ballast Km 31 and
finally the freshly crushed ballast. A number
3.5 of mathematical functions have been pro-
posed to predict the settlement behaviour
4.0 of ballast (Selig & Waters 1994; Indrarantna
Recycled ballast Km 9 Recycled ballast Km 17 Recycled ballast Km 31 et al 2006). Considering the trends of per-
Recycled ballast Km 32 Freshly crushed ballast manent strains with respect to the number
of load cycles in the triaxial simulation of
Figure 14 First-stage settlement results ballast materials in the laboratory, these

Journal of the South African Institution of Civil Engineering Volume 64 Number 4 December 2022 25
 Number of cycles, N
50 000 55 000 60 000 65 000 70 000 75 000 80 000 85 000 90 000 95 000 100 000
0

0.1

0.2

0.3
Settlement, SN1 (mm)

0.4

0.5

0.6

0.7

0.8

0.9

1.0
Recycled ballast Km 9 Recycled ballast Km 17 Recycled ballast Km 31 Recycled ballast Km 32 Freshly crushed ballast

Figure 15 Second-stage settlement results

researchers found that semi-log and power In this study a power function of settle- the accuracy of the model in estimating the
functions better describe the behaviour ment and number of load repetitions was settlement of recycled ballast during tri-
of ballast material. The researchers also used to analyse the triaxial test results. axial testing. The sample with the highest
found the power function as presented settlement was recycled ballast Km 32, fol-
in Equation 7 to be more accurate than Second-stage settlement lowed by recycled ballast Km 17, recycled
the semi-logarithmic function regarding As expected, a significant amount of settle- ballast Km 9, recycled ballast Km 31 and
­modelling the behaviour ballast settlement. ment occurred in the first stage of testing, freshly crushed ballast. From Table 4 it can
due to the ballast particles being free to be observed that the recycled ballast expe-
SN = aN b(7) settle and undergo reorientation during the rienced the greatest amount of settlement
first-stage loading. Figure 15 shows that the before the second-stage settlement.
SN = a(k log N + 1) (8) highest settlement was found in recycled
ballast Km 9, followed by recycled ballast Combined recycled ballast model
Where: Km 17, recycled ballast Km 32, recycled A close examination of all four recycled
SN = settlement of ballast ballast Km 31 and freshly crushed ballast. ballast settlement variables suggests that
N = load cycles Table 4 presents total settlement of the the individual databases could be combined
a = settlement at first cycle tested samples computed using Equation 7. for analysis. A simplified model to combine
b&k = empirical constant The power function R2 was used to assess all data of the recycled ballast was therefore

Table 4 Settlement of tested materials using power function

First-stage settlement (SN1) Second-stage settlement (SN2) Total settlement
Sample description
a b SN1 R2 a b SN2 R2 SN (SN1 + SN2)

Recycled ballast Km 9 0.92 0.08 2.37 0.98 0.01 0.04 0.74 0.96 3.11

Recycled ballast Km 17 2.22 0.04 3.69 0.98 0.02 0.35 0.64 0.96 4.33

Recycled ballast Km 31 0.62 0.10 1.88 0.99 0.01 0.35 0.46 0.93 2.34

Recycled ballast Km 32 2.72 0.03 3.86 0.94 0.01 0.38 0.85 0.93 4.44

Freshly crushed ballast 0.11 0.08 0.26 0.98 0.01 0.46 0.92 0.98 1.18

26 Volume 64 Number 4 December 2022 Journal of the South African Institution of Civil Engineering
 5.0 5.0

4.5 4.5

4.0 4.0

3.5 3.5
Settlement (mm)

Settlement (mm)
3.0 3.0

2.5 2.5

2.0 2.0

1.5 1.5

1.0 1.0

0.5 0.5

0 0
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
Flatness ratio Sphericity

Recycled ballast Km 9 Recycled ballast Km 17 Recycled ballast Km 31 Recycled ballast Km 32

Figure 16 Effect of flatness ratio and sphericity on settlement

developed to simulate ballast values that
had been subjected to repeated loading and
had undergone degradation. When ballast
has been in service for a significant period
of time, it becomes damaged and contami-
nated, resulting in a change in shape and
size, which in turn leads to a reduction in
performance. The developed first-stage
model is shown in Equation 9 and the
second-stage model for recycled ballast is
shown in Equation 10.
Recycled Ballast Models
First stage SN1 = 1.10 N0.09(9)
Second stage SN2 = 0.01 N0.36(10)
of ±3.5 mm, maintenance is necessary to settlement, and therefore the performance
return the railway track to its original posi- of the rail track structure can be linked to
tion, since differential settlement can result ballast shape.
in many problems, including derailment
(Zaayman 2013).
FINDINGS AND CONCLUSIONS
Flakiness index A modern 3-D laser scanning technique
Figure 17 shows the effect of the flakiness was used to determine shape properties of
index with settlement. An overall good railway ballast materials, which were linked
correlation was observed, with a good cor- to the settlement performance of the mate-
relation between the flakiness index and rials. The 3-D laser technique was used to
the settlement (R2 = 0.904) of the recycled directly obtain the flakiness index, flatness
materials. This shows that laser scanning ratio, elongation ratio and sphericity of
is a versatile technique to determine the ballast particles. Settlement of the ballast
shape of railway ballast. The flakiness was determined in the laboratory from
index increases linearly with increasing triaxial testing with conditions derived

5.0
EFFECT OF BALLAST SHAPE 4.5 y = 0.2675x + 2.5057
ON SETTLEMENT R2 = 0.9041
4.0

Flatness ratio and sphericity 3.5

Settlement (mm)

The threshold value of sphericity will 3.0

be 1.0, as indicated by use of spheres in
2.5
Figure 12. The flatness ratio and spheric-
ity determined from the laser were used 2.0
to evaluate the effect of ballast shape
1.5
on settlement. Figure 16 shows a plot of
settlement against flatness ratio and sphe- 1.0

ricity respectively. There was no proper 0.5

relationship observed. Ballast deteriorates
0
due to many reasons, including break- 0 2 4 6 8 10 12
age of angular corners and sharp edges Flakiness index (%)
resulting in finer materials with minimum Recycled ballast Km 9 Recycled ballast Km 31
dimensions, as shown in Figure 4. Once Recycled ballast Km 17 Recycled ballast Km 32
the vertical alignment (track settlement)
in the track reaches the allowable limit Figure 17 Effect of flakiness index on settlement

Journal of the South African Institution of Civil Engineering Volume 64 Number 4 December 2022 27
from railway track field loading conditions.
Based on the findings and results discussed
in this paper, the following conclusions
are drawn:
QQ Much-needed improvement in the
measurement of the ballast shape has
been successfully demonstrated in
this study using the 3-D laser-based
scanning technique. This implies that
a more accurate measurement of the
highly irregular shapes of ballast mate-
rials can be achieved through automa-
tion instead of the traditional methods
that are susceptible to error.
QQ The findings presented in this study
showed that railway ballast shapes
can be adequately captured by a
modern 3-D laser scanning system.
Furthermore, these shapes can be linked
with settlement results from triaxial
testing to mimic field loading of the bal-
last materials.
QQ The desired particle shape should be
distributed within the shaded area of
the shape classification chart. Flaky
or elongated particles have a specific
dimension that is significantly greater
than other dimensions and tend to
produce preferential planes of weakness
when present in large proportions and
clustered together in the material.
QQ The results of cyclic loading of the
tested materials confirmed that settle-
ment of good quality freshly crushed
ballast was low when compared with
recycled ballast material under the same
field loading conditions.
QQ The power mathematical function bet-
ters the model settlement behaviour of
recycled (degraded) ballast materials.
This validates the existing classical
power function that is commonly used
to model permanent deformation (set-
tlement) of the behaviour of ballast
materials.
ACKNOWLEDGEMENTS
This project was funded by the Council for
Scientific and Industrial Research (CSIR)
and Transnet Freight Rail. The project was
done at the CSIR towards a postgraduate
Master’s degree from the University of
Pretoria.
28
REFERENCES
Anochie-Boateng, J K, Komba, J J & Tutumluer, E 2012.
Aggregate surface areas quantified through laser
measurements for South African asphalt mixtures.
Transportation Engineering, 138(8): 1–11.
Anochie-Boateng, J K, Komba, J J & Mvelase, G M 2013.
Three-dimensional laser scanning technique to
quantify aggregate and ballast shape properties.
Journal of Construction and Building Materials,
43: 389–398.
Anochie-Boateng, J K 2014. Test protocol for
determining particle shape and surface properties
of aggregate and ballast using laser scanning
technique. CSIR Technical Manual CSIR/BE/IE/
IR/2014/0038/C. Pretoria: CSIR Built Environment.
Arangie, P B D 1995. The influence of ballast fouling
on the functional behaviour of the ballast pavement
layer. MEng Dissertation. University of Pretoria.
Asschenbrener, B C 1956. A new method of expressing
particle sphericity. Journal of Sedimentary Petrology,
26: 15–31.
Barrett, P J 1980. The shape of rock particles: A critical
review. Sedimentology, 27: 291–303.
Hayakawa, Y & Oguchi, T 2005. Evaluation of gravel
sphericity and roundness based on surface area
measurement with a laser scanner. Journal of
Geology, 31(6): 735–741.
Indraratna, B, Shahin, M A & Rujikiatkamjorn, C 2006.
Stabilization of rail tracks and underlying soft soil
formations. Australia: University of Wollongong.
Krumbein, W C & Pettijohn, F J 1938. Manual of
Sedimentary Petrography. New York: Appleton-
Century, Inc.
Krumbein, W C 1941. Measurement and geological
significance of shape and roundness of sedimentary
particles. Sedimentary Petrology, 11: 64.
Kuo, C Y & Freeman, R B 2000. Imaging indices for
quantification of shape, angularity and surface
texture of aggregates. Proceedings, Transportation
Research Board 79th Annual Meeting, Washington
DC, Paper No: 000686.
Kuys, W 2011. Heavy haul operations in South Africa.
Proceedings, 2011 International Heavy Haul
Association Conference, 19–22 June 2011, Calgary,
Canada.
Li, D, Hyslip, J, Sussmann, T & Chrismer, S 2015.
Railway Geotechnics. New York: Taylor and Francis.
Lim, W L & McDowell, G R 2005. Discrete element
modelling of railway ballast. Granular Matter,
7(1): 19–29.
Moaveni, M 2015. Advanced image analysis and
techniques for degradation characterization of
aggregates. PhD Thesis. University of Illinois at
Urbana-Champaign.
Volume 64 Number 4 December 2022
Moaveni, M, Wang, S, Hart, J M & Ahuja, N 2013.
Aggregate size and shape evaluation using
segmentation techniques and aggregate image
processing algorithms. Transportation Research
Record 2335, Paper No. 13-4167: 50–59.
Mvelase, G M, Anochie-Boateng, J K & Gräbe, P J 2012.
Application of laser-based technology to quantify
shape properties of railway ballast. Proceedings, 31st
Southern African Transportation Conference, 9–12
July 2012, Pretoria.
Mvelase, G M, Gräbe, P J & Anochie-Boateng, J 2017.
The use of laser technology to investigate the effect
of railway ballast roundness on shear strength.
Transportation Geotechnics, 11: 97–106.
Raymond, G P 1985. Research on railroad ballast
specification and evaluation. Transportation
Research Record 1006: 1–8.
Sames, C W 1966. Morphometric data of some recent
pebble associations and their sedimentary petrology.
Journal of Sedimentary Research, 26: 15–31.
SANS (South African National Standard) 2015. SANS
3001-AG4 Civil engineering test methods. Part AG4:
Determination of the flakiness index. Pretoria: SABS
Standards Division.
Selig, E T & Waters, J M 1994. Track Geotechnology
and Substructure Management, 1st ed. London:
Thomas Telford.
Sneed, E D & Folk, R L 1958. Pebbles in the lower
Colorado River, Texas: A study in particle
morphogenesis. Journal of Geology, 66(2): 114–150.
TMH 1 (Technical Methods for Highways 1) 1986.
Standard Methods of Testing Road Construction
Materials. Pretoria: CSIR National Institute for
Transport and Road Research.
Tolppanen, P, Illerstrom, A & Stephansson, O 2008.
3D Laser analysis of size, shape and roughness of
railway ballast. Research in Progress. Sweden: Royal
Institute of Technology.
TFR (Transnet Freight Rail) 2011. Specification for the
supply of stone. Transnet Freight Rail Technical
Document S406. Johannesburg: Transnet.
Tutumluer, E, Qian, Y, Hashash, Y M A, Ghaboussi, J
& Davis, D 2011. Field validation discrete model
for railroad ballast. Proceedings, AREMA Annual
Conference, 18–21 September 2011, Minneapolis,
MN.
Zaayman, L 2013. Mechanisation of Track Work
in Developing Countries. MTech Dissertation.
University of Johannesburg.
Zingg, T 1935. Contribution to the gravel analysis
(Beitrag zur Schotteranalyse). Schweizerische
Mineralogische und Petrographische Mitteilungen,
15: 38–140. DOI: doi.org/10.3929/ethz-a-000103455.
Journal of the South African Institution of Civil Engineering
A proposal to facilitate TECHNICAL PAPER
BIM implementation Journal of the South African
Institution of Civil Engineering

across the South African ISSN 1021-2019 (print) | ISSN 2309-8775 (online)
Vol 64 No 4, December 2022, Pages 29–37, Paper 1450

construction industry STEPHAN CALITZ, a Candidate Civil Engineer,

obtained his BEng and MEng (cum laude) in
Civil Engineering at the University of
Stellenbosch. He is currently employed as a
S Calitz, J A Wium civil engineer in the Civil and Transport
Department of Lithon Project Consultants in
Namibia.

The value of Building Information Modelling (BIM) has been noted by the global construction
industry, and several countries are facilitating BIM through national strategies and initiatives. Contact details:
Lithon Project Consultants Department of Civil Engineering
However, the South African Construction Industry (SACI) lacks widespread and effective BIM Windhoek Stellenbosch University
implementation. Therefore, this study presents a proposal to facilitate BIM implementation Namibia Private Bag X1
across the industry. T: +264 81 207 1300 Matieland 7602
E: stephan.calitz@lithon.com South Africa
A qualitative, exploratory methodology was followed involving semi-structured interviews
with nine BIM experts. The interviews were analysed using thematic analysis. An understanding
PROF JAN WIUM (Pr Eng, FSAICE) is
was gained of the challenges associated with BIM implementation across the construction professor in the Department of Civil
industry. Possible solutions were found in terms of initiatives and strategies to facilitate Engineering at Stellenbosch University. He
and promote BIM implementation locally. Key role-players responsible for facilitating BIM specialises in Construction Engineering and
Management, and in Structural Engineering.
implementation in the SACI were identified.
Finally, a proposal is made to facilitate BIM implementation across the local industry, which
comprises three key concepts, namely leadership, strategy, and roles and responsibilities. This
research contributes to BIM adoption and implementation in the SACI and can be used for Contact details:
future research on national BIM implementation strategies. Department of Civil Engineering
Stellenbosch University
Private Bag X1
Matieland 7602
INTRODUCTION environment (Georgiadou 2019). Therefore, South Africa
BIM has become the centre of digital T: +27 21 808 4348
Background information management processes in the E: janw@sun.ac.za

The fourth industrial revolution (4IR) Architecture Engineering Construction

introduced trends in the construction and Operations (AECO) industry.
industry such as automation, big data, The rate of BIM adoption and digitali-
cloud computing, smart cities, and the sation still varies rapidly amongst countries
Internet of Things (IoT). A pivotal moment and organisations because of different
in the construction industry was the intro- cultures, government initiatives and a
duction of Building Information Modelling variation in the rate of the development
(BIM). According to the International of national BIM standards and guidelines
Organisation of Standardisation (ISO) (Akintola et al 2017; Edirisinghe & London
19650 standard, BIM is “… a shared digital 2015). Some countries, such as the United
representation of a built asset to facilitate Kingdom (UK), have recognised the
design, construction and operation pro- economic value of BIM and put in place
cesses to form a reliable basis for decisions.” strategies to facilitate BIM implementation
(BS 2019). In simpler terms, a BIM model (HM Government 2015; Royal Institution
contains all information about a built asset of Chartered Surveyors 2020). However,
(graphical data, non-graphical data, and South Africa has yet to reach a high level of
all project-related documents), allowing BIM maturity across the industry (Akintola
for better decision-making during the et al 2020; Govender 2018; Mtya 2019).
asset’s design, construction and operational Although a handful of larger AECO firms
phases. BIM processes increase efficiency have gained some BIM capabilities, BIM
and productivity since it allows multidis- implementation in South Africa is neither
ciplinary project teams to work collabora- widespread nor optimal (Akintola et al
tively and exchange information in a digital 2017; Kiprotich 2014; Mtya 2019).

Keywords: Building Information Modelling (BIM),

Calitz S, Wium JA. A proposal to facilitate BIM implementation across the South African construction industry. BIM implementation strategies, BIM adoption,
J. S. Afr. Inst. Civ. Eng. 2022:64(4), Art. #1450, 9 pages. http://dx.doi.org/10.17159/2309-8775/2022/v64n4a3
digital construction, construction management
The Journal of the South African Institution of Civil Engineering is a peer-reviewed, open-access journal licenced under a Creative Commons Attribution Licence
(CC BY-NC-ND) and is distributed internationally. The content of this journal is available to users free of charge. However, copyright remains with the South African
Institution of Civil Engineering. Permission for reproduction of content, with appropriate reference to the journal, can be obtained from the administrative editor. 29
 In South Africa BIM implementa-
tion is driven amongst others by the
BIM Institute, the BIM Academy Africa Clients Project managers
and software developers (Natspec 2019). Facilities management Planning and phasing
However, the BIM Institute operates sepa- Brief Analysis Reporting
rately from the government and software
developers (Natspec 2019) and there is o n M a nag
a ti e
still a lack of collaborative efforts between m
Contractors

m
Info

e nt
the influential industry stakeholders to Architects Manufacturing/
BIM
promote BIM. Design and presentation
Model
prototyping
Documentation Clash detection
Work scheduling
Problem statement
Akintola et al (2017) report that BIM
has been present in the industry since Engineers
2010; however, its implementation in Civil and structural Quantity surveyors
South Africa has not been effective nor Mechanical, electrical Material take-offs
and plumbing Cost analysis
widespread because there is no structured
Performance analysis
approach or strategy at a national level to
promote and facilitate BIM implementa-
tion. Therefore, several studies have Figure 1 BIM application per project stakeholder (adapted from Ismail (2019))
highlighted the need for more research
on BIM implementation in South Africa
(Chimhundu 2015; Froise 2014; Mtya 2019).
Several more recent studies also found that
BIM uptake in South Africa is low com- 1
pared to other countries (Govender 2018;
Design effort

1 Ability to control cost

Meno 2020; Mtya & Windapo 2019). As
a result, the competitiveness of the South
African construction industry (SACI) is
potentially under threat, and international
companies are in an advantageous posi-
tion to take work from local companies
(Froise 2014).
Research design
This paper presents a proposal to facilitate
BIM implementation across the local con-
struction industry. First, existing research
is explored with a specific focus on BIM
implementation challenges and possible
solutions to mitigate the challenges. Then,
a qualitative methodology is followed
which entails conducting semi-structured
interviews with nine BIM experts. Finally,
the interview findings are compared with
the literature to provide a holistic, updated
account of the status and prospects regard-
ing BIM implementation in the SACI. The
research contributes to BIM research and
development and provides a proposal to
guide the industry on the way forward.
LITERATURE REVIEW
Value of BIM
The benefit of BIM is experienced
throughout an asset’s life cycle, from
initiation through to operation phases.
Furthermore, BIM according to the
2 4 Cost of design changes
3
4
Pre-construction
Time
Figure 2 Traditional versus IPD design process (adapted from Smith & Tardif (2009))
ISO 19650 standard is a collaborative pro-
cess that requires input from all project
stakeholders in each phase. Therefore,
project stakeholders might contribute to
the BIM process differently. Figure 1 illus-
trates how project stakeholders contribute
and benefit differently to and from the
BIM process.
Effective BIM implementation contrib-
utes to an integrated project delivery (IPD)
approach (Froise 2014). The focus of an
integrated project delivery approach is on
project collaboration. Effective collabora-
tion promotes quick and reliable exchange
of project information, which leads to
better project performance (Georgiadou
2019). Another significant advantage of
BIM applied to a project is that a greater
design effort is applied earlier on in the
project through integrated project delivery
than with traditional construction proj-
ects, as illustrated in Figure 2 (Smith &
Tardif 2009).
2 IPD design process
3 Traditional design process
Construction
In addition, BIM allows for the integra-
tion of multidisciplinary design fields early
in the project, which reduces errors, design
clashes and constructability constraints,
which in turn reduces the amount of
rework on a project (Ghaffarianhoseini et
al 2017; Hwang et al 2019). Although the
IPD process involves a similar design effort,
this effort is shifted to an earlier stage
(see the design effort shift in Figure 2).
Putting in more effort at the start of the
project enables important decisions to be
made before construction commences and
therefore reduces change orders later in
the project, which will have a much more
significant impact on the project cost
(McGraw Hill Construction 2014b).
Past research on BIM
implementation
BIM implementation should be considered
according to the different levels or dimen-
sions of analysis, namely organisation,

30 Volume 64 Number 4 December 2022 Journal of the South African Institution of Civil Engineering
Table 1 Identified challenges to BIM implementation in the SACI
Challenges
Lack of BIM competency in industry and
education
Lack of BIM research and awareness
Lack of industry guidance and leadership for BIM
implementation
High implementation costs and uncertainty in
returns
Lack of a legal framework
Cultural and social barriers
Lack of government support
Lack of support and drive from all project
stakeholders and industry institutions
Ineffective traditional procurement system
Table 2 Identified solutions to implement BIM in the SACI
Solutions (initiatives)
Develop national standards and guidelines for
local BIM implementation
Develop and promote education and training
programmes
Promote research and development in BIM
Increase government support
Increased support from industry stakeholders
Explore alternative procurement routes
project, and industry (Brito et al 2021).
While considering all levels of analysis,
this study focused mainly on an indus-
try perspective, specifically the SACI.
Vidalakis et al (2020) proposed that BIM
adoption should be facilitated in the
industry through effective leadership,
industry initiatives and peer education.
However, Georgiadou (2019) noted that
there is no single solution to implement
BIM that works for everyone. Countries,
projects and organisations all have differ-
ent demographics and challenges that need
to be overcome. Therefore, it is important
to first gain a thorough understanding of
BIM implementation challenges before the
solutions are considered. Table 1 provides
Journal of the South African Institution of Civil Engineering Volume 64 Number 4 December 2022
Source from literature
Chimhundu 2015; Kekana et al 2015; Tabesh 2015;
Moodley et al 2016; Pillay et al 2018; Mtya 2019;
Meno 2020
Froise 2014; Kiprotich 2014; Chimhundu 2015;
Kekana et al 2015; Govender 2018; Mtya 2019;
Meno 2020
Chimhundu 2015; Wortmann et al 2016; Akintola
et al 2020
Kiprotich 2014; Chimhundu 2015; Akintola et al
2017; Meno 2020
Froise 2014; Chimhundu 2015; Akintola et al 2017;
Meno 2020
Akintola et al 2017; Chimhundu 2015; Froise 2014;
Meno 2020
Kiprotich 2014; Chimhundu 2015; Akintola et al
2017; Meno 2020
Akintola et al 2017; Chimhundu 2015; Froise 2014;
Govender 2018; Kekana et al 2015
Froise 2014; Chimhundu 2015; Akintola et al 2017;
Ogwueleka & Ikediashi 2017; Govender 2018
Source from literature
Akintola et al 2017; Chimhundu 2015; Kiprotich
2014; Meno 2020; Mtya 2019; Wortmann et al 2016
Chimhundu 2015; Kekana et al 2015; Froise 2014;
Kiprotich 2014; Meno 2020; Pillay et al 2018;
Tabesh 2015
Chimhundu 2015; Froise 2014; Kiprotich 2014;
Meno 2020; Mtya 2019
Akintola et al 2017; Chimhundu 2015; Froise 2014;
Kiprotich 2014; Mtya 2019; Pillay et al 2018
Chimhundu 2015; Froise 2014; Govender 2018;
Kiprotich 2014; Meno 2020; Mtya 2019
Froise 2014; Chimhundu 2015; Akintola et al
2017; Meno 2020; Govender 2018; Ogwueleka &
Ikediashi 2017
a summary of the main challenges to BIM
implementation in the local industry, as
identified by past research.
Although South African research on the
solutions or initiatives to BIM implementa-
tion is limited, some studies have identified
solutions for the limited implementation, as
illustrated under various themes in Table 2.
International BIM
implementation efforts
Since South Africa does not have a national
BIM implementation plan or strategy, South
Africa now has the advantage of develop-
ing a strategy based on experience from
other countries. The BIM implementation
strategies of three BIM-mature countries are
roughly described in this section, namely
Ireland, Singapore and the United Kingdom.
Although many countries have similar
strategies, these three were regarded as
insightful and some aspects of these strate-
gies could be applied to the SACI.
Ireland
The National BIM Council (NBC) of
Ireland released a Digital Roadmap in 2017.
The NBC consists of a committee of clients
and representatives across the construc-
tion industry supply chain, and their key
deliverable was to develop the National
Roadmap to guide the digital transition in
the built environment of Ireland (National
BIM Council Ireland 2017).
The roadmap is divided into four
categories: leadership, standards, educa-
tion and training, and procurement. Each
category contains several actions, events,
or initiatives that will help improve digital
uptake in the industry and improve the
industry’s performance. The Ireland road-
map also contains a second section where
the key role-players in the industry respon-
sible for events and activities are identified
(National BIM Council Ireland 2017).
Singapore
Singapore’s latest strategy involving BIM
implementation is the Integrated Digital
Delivery (IDD) implementation plan, which
is led by the IDD steering committee
(Building and Construction Authority 2018).
Four million dollars were set aside as part
of the IDD implementation plan to fund
digital platforms that should assist construc-
tion firms to accelerate their digitalisation
(Building and Construction Authority 2018).
The IDD implementation plan consists of
three main action plans, namely:
1. Promote and develop IDD by demon-
strating it through actual projects.
2. Develop IDD platforms, solutions, and
standards.
3. Increase IDD competency.
United Kingdom
The Digital Built Britain Level 3 BIM is the
UK’s latest strategic plan that builds on the
BIS (Department of Business, Innovation and
Strategy) BIM strategy and Construction
2025 (HM Government 2015). The Digital
Built Britain vision does not only involve
stakeholders in the construction sector, but
it involves a national movement covering
other sectors such as transport, education,
health and finance, and includes smart cities
and the digital economy (HM Government
31
2015). The Digital Built Britain strategy
provides opportunities and innovations such
as IoT, AI, smart cities and data analytics, to
name a few (Royal Institution of Chartered
Surveyors 2020). However, this strategy
is perhaps somewhat too advanced to be
adopted by South Africa, which is still at an
infant BIM adoption stage.
Literature review conclusion
The literature review provided insight
into the challenges and solutions of BIM
implementation. In addition, the methods
used by past authors were considered for
the current study. Finally, the national BIM
implementation efforts of Ireland, Singapore
and the United Kingdom have lessons for
the implementation of BIM in the South
African context. The information from the
literature was used as a basis to formulate a
research methodology and to develop new
findings in terms of BIM implementation
challenges, solutions to the challenges, and
national BIM implementation strategies.
METHODOLOGY
Research approach
From the past literature on BIM implemen-
tation, it is observed that quantitative stud-
ies used surveys as the main form of data
collection, and qualitative studies used focus
group interviews. For example, Mtya (2019)
conducted a quantitative study by making
use of surveys to assess South Africa’s BIM
maturity and capability. However, the aim of
the current study is more complex, requir-
ing an in-depth understanding. Therefore, a
qualitative approach was adopted with the
main form of data being semi-structured
interviews with a focus group of BIM
experts from the construction industry.
This focus group could provide the authors
with informed answers to complex ques-
tions that might have several answers or
interpretations. This approach is similar
to the approach taken by Sahil (2016) and
Akintola et al (2017). Hence, these studies
were used as guidance in developing the
research methodology.
The chosen qualitative approach is
phenomenological, similar to Sahil (2016),
which is “… a study that attempts to under-
stand people’s perceptions and perspectives
relative to a particular situation”, according
to Leedy and Ormrod (2016). Focus group
interviews allowed the authors to study the
perspectives and perceptions of BIM experts
to gain insight into BIM implementation in
South Africa. The study is also exploratory,
similar to that of Chimhundu (2015).
Focus group demographics
The focus group participants were
required to have a mature understanding
of BIM and the SACI. Furthermore, the
focus group needed to represent a diverse
range of industry stakeholders, including
architects, contractors, engineers, clients
and academics, to prevent the results from
being skewed towards a particular group’s
perspective. An adequate representation
from the public and private sectors was
maintained for a similar reason. Seven
participants were from South Africa and
two were from Ireland and New Zealand
respectively but have lived in South Africa
and could therefore provide meaningful
insight to the discussions. The collective
BIM experience of all participants adds up
to a total of 68 years. Their collective expe-
rience in the construction industry adds up
to 163 years. The participants’ demograph-
ics, background and experience are shown
in Table 3. The participants’ names are not
disclosed due to ethical considerations.
Data collection and analysis
Due to the rapid transformation of BIM in
the construction industry, the literature
alone did not contain the latest information
about BIM. Therefore, there was a need
to conduct interviews with BIM experts.
The data collection process involved (1) the
focus group selection process, (2) interview
guide development, (3) conducting of the
interviews, and (4) interpreting the inter-
view results. Ethical approval for the study
was obtained from Stellenbosch University
under project number ING-2019-17095.
The data analysis process commenced
during the interviews while the interviewer
interpreted the participants’ perceptions
and made notes accordingly. The interview
notes were sent to the participants for their
review as a means of validation. Then the
notes were analysed using the thematic
analysis process as described by Braun and
Clarke (2006). The interview responses were
described using thick description, which
describes the participants’ perceptions in
detail and captures the phenomenon’s com-
plexity to provide readers with an in-depth
and complete picture of the phenomenon,
allowing them to make informed interpreta-
tions (Leedy & Ormrod 2016). After describ-
ing the interview responses per question,
the findings were obtained per research
objective through a further step of thematic

Table 3 Participant background and expertise

Code Background Profession Company type

Civil engineering background from three continents. Working on government

P1 civil engineer client (public)
strategies to increase BIM implementation in local government.

Civil engineer. Experience on BIM projects internationally. Current business owner

P2 civil engineer consulting (private)
providing consulting for construction technology and digital transformation.

P3 Owner of project management consultancy providing BIM services. project manager project management (private)

Architectural background. BIM experience internationally. Current work involves education (public) and
P4 researcher and consultant
digital twin cities, city development strategies and lean construction practices. consulting (private)

P5 Property developer working on an integrated BIM project in South Africa. property developer client (private)

P6 BIM software consultant and vendor. software specialist software supplier (private)

P7 Civil engineer and BIM champion, working on government projects. civil engineer client (public)

Architectural background. Many years of BIM experience. Contributed to architect, BIM specialist/ BIM consulting and support
P8
developing Ireland’s digital roadmap. Providing BIM services and consulting. consultant services (private)

BIM manager, BIM specialist/

P9 Digital engineer at a large consulting firm and running a BIM education business. consulting (private)
consultant

32 Volume 64 Number 4 December 2022 Journal of the South African Institution of Civil Engineering
analysis. Then the findings were compared
to the literature. To conclude the process,
the findings and discussion were used to
develop the final proposal to facilitate BIM
implementation across the SACI.
FINDINGS AND DISCUSSION
Challenges to BIM implementation
The challenges faced in South Africa were
grouped according to five main themes,
namely educational, cultural, legal, finan-
cial, and governmental. Although these
challenges are grouped in specific themes,
they should be viewed as part of an exten-
sive system with interlinking relationships.
Educational
One of the primary barriers to BIM imple-
mentation was found to be the shortage of
BIM skills in the industry. Moodley et al
(2016), Pillay et al (2018) and Tabesh (2015)
confirmed that the education and training
curricula in South Africa did not produce
graduates with the required BIM skills needed
to deliver mature BIM projects. Furthermore,
the participants emphasised the general lack
of BIM awareness in the industry, especially
among contractors and asset owners. P6
(from Table 3) claimed that industry profes-
sionals have a misinformed view of BIM,
which prevents them from adopting it.
Govender (2018) explained that this misin-
formed view is a result of the lack of efforts
from industry institutions publishing infor-
mation about BIM since they are responsible
for informing and guiding the industry.
Cultural
Another common challenge was found
to be the resistance to change. Industry
professionals are used to a certain way of
carrying out tasks and delivering projects,
and are reluctant to explore modern pro-
cesses. This was reflected in the literature
by Akintola et al (2017), Chimhundu
(2015) and Froise (2014). Another signifi-
cant theme that was identified from the
interview responses was the lack of col-
laboration between project stakeholders.
Due to the competitive culture of the South
African construction industry, project
stakeholders are used to working in silos.
However, working in silos is detrimental to
the success of a BIM project, since a BIM
project requires collaboration amongst
stakeholders. Froise (2014) reinforces that
the lack of collaboration in the industry
creates a fragmented culture.
Journal of the South African Institution of Civil Engineering Volume 64 Number 4 December 2022
Legal
The two major themes regarding the legal
challenges to BIM implementation were
found to be the lack of formal guidance on
projects regarding BIM processes and deliv-
erables, as well as the outdated procurement
system. Similarly, Akintola et al (2017) high-
lighted the lack of guidance in the form of
BIM standards. Govender (2018) explained
that, due to the traditional procurement sys-
tem and the way contracts are setup, project
stakeholders are hesitant to share informa-
tion with one another, creating information
silos. This lack of collaboration inhibits
effective BIM implementation.
Financial
It was found that organisations in the SACI
perceive BIM implementation as expensive
and risky. Implementing BIM requires initial
investments in upskilling, training, upgrading
hardware and network infrastructure, and
software licences. Although there is some
debate about whether BIM is profitable or
not, some participants strongly suggested that
the expensive software licences do not meet
the industry and market demands of South
Africa. Furthermore, participants highlighted
the risk of theft and vandalism, especially on
rural and low-income sites in South Africa.
Governmental
The research found that there is a lack of
government drive in promoting and facilitat-
ing BIM across the construction industry.
Since 57% of construction projects in South
Africa are publicly funded (CIDB 2021) and
the industry is heavily regulated by public
bodies (Anthony 2017), government leader-
ship has a significant impact on BIM imple-
mentation. The importance of government
policies and regulations was confirmed as a
critical driver of uniform BIM implementa-
tion (Gu & London 2010; Porwal & Hewage
2013). The concern that evolved from the
interview discussions was it would seem that
the majority of industry experts are from
the private sector and that the public sector
may lack the expertise and competency to
drive BIM. Furthermore, several participants
mentioned that there is an industry percep-
tion that technology adoption contradicts
government job creation initiatives.
Possible solutions to facilitate and
promote BIM implementation
Raising awareness
From the interviews, it seemed that the first
step towards increasing BIM uptake was
to raise awareness across the industry. P1
(from Table 3) mentioned the need to initi-
ate discussions between the private sector,
the public sector, and educational institu-
tions. Similarly, Froise (2014) recommended
that contractors, as well as all organisations
across the supply chain, should be made
aware of BIM. P7 noted that BIM awareness
could also be aimed at the public, which
could promote public participation and
excite the public about future projects.
Education and training
A major point of discussion was the need
for a national upskilling programme aimed
at developing the digital skills of exist-
ing employees and students. Similarly,
Moodley et al (2016) also recommended
the concept of a national education and
training programme to upskill the industry.
P6 and P8 mentioned that people need
to learn how to produce, manage and
exchange digital information. P1, P2 and P4
recommended allocating a percentage of a
project’s profit margin towards implement-
ing BIM and upskilling the workforce’s
digital capabilities.
In addition to education and training,
it was found that more research is needed
on the impact of BIM in South Africa,
which is supported by earlier studies by
Chimhundu (2015) and Mtya (2019). Since
BIM processes are rapidly evolving, it is
essential to carry out continuous research to
establish the current best standards for BIM
implementation. This will change in time as
the technology develops and the country’s
BIM maturity increases. Furthermore, there
is a need to research the required skill sets
in the industry to establish how university
curricula need to be updated to produce
competent graduates. It was found that built
environment faculties should collaborate
with the industry to develop a strategy to
upskill the industry.
Promoting pilot projects
Promoting pilot projects and case studies
was a commonly mentioned initiative to
increase BIM uptake. Well documented case
studies could provide proof of the concept
to show organisations that BIM imple-
mentation is profitable in South Africa.
Furthermore, case studies could be used to
establish BIM guidelines and develop BIM
standards and policies. The concept of pilot
projects formed a key part of Singapore’s
integrated digital delivery strategy (Building
and Construction Authority 2018), which
motivates the need for such an initiative.
33
Developing standards
Similar to the concept of case studies, it
was perceived that the development of a
national BIM standard for South Africa
could accelerate BIM uptake across South
Africa, since standards dictate a best
practice to implement BIM and guarantee
success. Therefore, implementing BIM
according to a BIM standard should be less
risky, as explained by P4. Furthermore, a
national BIM standard would eliminate
confusion and fragmented BIM implemen-
tation (Akintola et al 2017; Chimhundu
2015; Mtya 2019). At the time of the
interviews several participants suggested
that the ISO 19650 standard be adapted
to South Africa since it is regarded as the
current international BIM standard (part
of this standard has now been published
in 2022 as SANS 19650-1). Akintola et
al (2017) noted the importance of such a
national standard being a jointly agreed
upon document from a selected group of
industry professionals and BIM experts
representing the private and public sectors.
The process of adopting the various parts
of EN 19650 and development of related
national annexes should therefore proceed.
Updating procurement systems
The participants commented that tra-
ditional procurement systems inhibit
collaborative practices such as BIM and
IPD. The need to update procurement
systems in South Africa was confirmed by
Windapo (2017). Participants suggested
that contracts should accommodate the
application of BIM systems and processes,
addressing intellectual property and copy-
right concerns.
Developing software that meets
South African needs
The study found that available software
packages might not meet the SACI and
market demand. P4 explained that the
commonly used software packages are
aimed at large AECO firms while the bulk
of the industry consists of smaller firms.
Similarly, Chimhundu (2015), Kiprotich
(2014) and Mtya (2019) confirmed that
smaller organisations struggle to imple-
ment BIM and that the available software
does not meet their needs.
Government initiatives
Another solution identified was that the
government should mandate BIM on public
projects. P1 recommended the implemen-
tation of a framework with “hard” and
34
“soft” mandates. “Soft” mandates mean
that BIM is encouraged, promoted and
preferred, and “hard” mandates mean BIM
is formally enforced.
The participants reasoned that govern-
ment organisations should realise the value
of BIM and conform to the global trends of
digital transformation to start developing
digital and smart cities. The UK has been
mandating BIM on public projects for sev-
eral years, and it forms part of their Digital
Built Britain agenda and development of
smart cities. P5 recommended that the gov-
ernment could require building plan sub-
missions in the form of electronic submis-
sions rather than paper hard copies, which
is what Singapore started to do many years
ago (McGraw Hill Construction 2014a).
Even though the digital competency might
not yet lie within the governmental bodies
to effectively construct and operate smart
cities, the digital data could still be col-
lected to be used in the future. Data is a
valuable asset in today’s digital era.
Key role-players responsible for
facilitating BIM implementation
Many past studies have dealt with the
challenges to BIM implementation in South
Africa and identified some solutions to
mitigate these challenges. However, limited
research covers the important role-players
needed to facilitate, accelerate and promote
BIM implementation in South Africa. P8
emphasised the importance of leadership
and accountability when it comes to BIM
implementation. Therefore, this section
covers the main role-players needed to
increase BIM uptake in South Africa.
The SACI is a complex environment
consisting of many different organisation
types from the public and private sectors,
each contributing to the delivery of con-
struction projects. This complex environ-
ment makes it impossible to identify single
organisations that could be responsible for
facilitating BIM implementation across the
industry. Therefore, this study considered
the interview discussions and divided the
industry up into four main functional
groups that play a significant role in BIM
implementation. The four groups are
government, the education sector, private
organisations, and software developers. It
should be noted that some organisations
may fall into multiple categories or func-
tional groups and these main groups are
developed purely for simplification reasons.
The role and description of each of these
groups are discussed below.
Volume 64 Number 4 December 2022
Government
To simplify, the government group consists
of all government entities, municipalities,
industry bodies, professional institutions,
and voluntary associations (including
the Engineering Council of South Africa
(ECSA), the South African Institution of
Civil Engineering (SAICE), Consulting
Engineers South Africa (CESA), South
African Institute of Architects (SAIA)
…). The reasoning behind this grouping
was that these organisations all play a role
regarding governance, regulation, legisla-
tion and professionalism. Furthermore,
government plays the role of being an
initiator and driver, regulator, educa-
tor, funding agency, demonstrator and
researcher (Ethiopian Construction Project
Management Institute 2019).
Education sector
The education sector consists of organisa-
tions that provide training, education and
research, such as universities, colleges,
schools, training centres and research
centres. These organisations are publicly
or privately funded. The education sector
plays an important role in research and
development and in producing competent
professionals. Furthermore, research is
needed on the impact of BIM and the
newly required skill sets and professions
in the SACI. Furthermore, the education
sector is responsible for teaching profes-
sionals the requisite technical BIM skills
of how to produce, manage and exchange
digital information.
Private organisations
Private organisations refer to all construc-
tion project organisations that are privately
funded, such as architects, engineers and
other consultants, contractors, clients, and
all other organisations across the supply
chain. Private organisations have the most
difficult task among the four identified
role-players, since BIM implementation
within these firms requires private invest-
ment, which puts the business at risk.
Furthermore, BIM adoption requires new
skill sets and processes to be implemented
in a firm. Therefore, effective change
management is essential to increase a firm’s
digital or BIM competency. However, P8
explained that organisations such as those of
architects, engineers and contractors, prac-
tise a professional service. Hence, it is their
responsibility to remain updated on industry
trends and to carry out their service using
the best available tools and processes.
Journal of the South African Institution of Civil Engineering
Software developers
Software developers play an important role
regarding BIM implementation since they
develop the platforms used to carry out
BIM processes. According to Akintola et al
(2017), software vendors were at the time
the current drivers of BIM in South Africa
due to the lack of government intervention.
However, P4 mentioned that BIM software
is aimed at the international market and
does not suit the SACI and market needs.
Therefore, software developers could find
the opportunity in the market and produce
software that meets the South African
demographics. For example, they could pro-
duce simpler products with limited functions
at a better price, aimed at smaller contrac-
tors who make up the bulk of the industry.
Furthermore, software developers have the
responsibility to train professionals on soft-
ware use and provide technical support.
PROPOSAL
While some groups and organisations in
South Africa are promoting the imple-
mentation of BIM in their sectors, the
local industry lacks a coordinated effort to
drive the process. From this investigation
the authors were able to obtain a holistic
perspective on the phenomenon, which
allowed for the development of a proposal
to facilitate BIM in the SACI. The proposal
is based on the findings from the inter-
views and the literature. The discussion
below presents the proposal under three
headings, namely leadership, strategy, and
roles and responsibilities.
Leadership
Throughout the literature and the inter-
views, it was found that leadership is vital
for the success of BIM implementation
in the industry. Other countries formed
organisations to lead the industry. For
example, Ireland established a National BIM
Council, and the UK a BIM Task Group.
It seemed that most of these committees,
councils or boards consist of experts and
professionals representing various fields of
the industry. A similar approach is proposed
for South Africa – such a local group could
be called the South African National BIM
Council (SANBC).
The SANBC could be separate from the
existing industry bodies, professional insti-
tutions and voluntary associations such as
the Construction Industry Development
Board, the Council for the Built
Environment, or SAICE. However, these
industry bodies could have representatives
in the SANBC, and the initiatives proposed
by the SANBC could be reflected and
executed by these industry organisations.
Since the key role-players responsible for
facilitating BIM across the construction
industry were identified as the government,
the education sector, private organisations
and software developers, the representation
of these groups in the SANBC is critical to
the success of BIM leadership in the SACI.
Strategy
Once the SANBC is established, a clear
BIM strategy or action plan with specific
initiatives should be developed. For exam-
ple, Ireland’s NBC developed a “digital
roadmap” to guide the industry. This strat-
egy could include clear goals to develop the
construction industry, such as those set by
the UK in their Government 2025 strategy.
However, the UK’s latest goals in their
Digital Built Britain agenda include goals
reaching broader than just the construc-
tion industry and involve a digital agenda
towards smart cities and IoT. South Africa
could consider adopting some aspects of
these global strategies towards a South
African BIM strategy, while considering
the nature and demographic makeup of the
construction industry.
Pilot projects could form a central part
of a national BIM strategy since it was a
common theme in the interview discus-
sions, and it was part of Singapore’s IDD
strategy. Well documented pilot projects
could provide proof of concept while
upskilling the industry. Pilot projects
could be monitored through research. The
research could update the BIM strategy,
develop new guidelines and standards, and
further guide the industry.
Roles and responsibilities
Leadership and strategy could give the
industry direction. However, the strat-
egy needs to be executed and monitored.
Therefore, the main industry role-players
could be assigned specific roles and respon-
sibilities aligned with the strategy. The
proposed initiatives falling under the main
industry role-players are shown in Figure 3.
These initiatives are adopted from the
findings of the interview discussions. The
layout of Figure 3 suggests that the SANBC

South African National BIM Council (SANBC)

QQ Lead BIM implementation while guiding industry stakeholders in their role.
QQ Act as the sole source of information released about BIM in the South African construction industry.
QQ Develop a national BIM implementation strategy and action plan.
QQ Develop BIM standards.
QQ Raise awareness through discussions, seminars and workshops.
QQ Initiate and monitor BIM incentives.
QQ Initiate pilot projects and document them as case studies.

Government Education sector Private organisations Software developers

QQ Develop a BIM QQ Develop a national upskilling QQ Determine the business QQ Provide software that
implementation framework and education programme. value of BIM and develop matches the South African
with hard and soft mandates. QQ Conduct research on BIM in an organisational strategy to industry and market needs.
QQ Update contracts and South Africa, including the implement BIM. QQ Provide technical BIM
procurement systems. industry’s skill demand. QQ Implement BIM and upskill support.
QQ Collect data to develop digital QQ Update university curricula the organisation.
and smart cities. to release BIM competent QQ Partner with other
QQ Adapt regulations and graduates. organisations.
policies. QQ Upskill the industry.
QQ Develop incentives towards
BIM education such as
subsidies and tax reliefs.

Figure 3 Roles and responsibilities of main industry role-players

Journal of the South African Institution of Civil Engineering Volume 64 Number 4 December 2022 35
represents the leadership or drivers of the
four main industry groups or role-players.
CONCLUSION
It was identified that there is not wide-
spread and effective BIM implementa-
tion in the SACI, which threatens the
industry’s competitiveness. Therefore, the
study aimed at developing a proposal to
facilitate BIM implementation across the
SACI.
The research provides an updated
report on the challenges to BIM implemen-
tation faced by the construction industry
and to the solutions, in terms of initiatives
and strategies, to mitigate these challenges.
The key role-players responsible for BIM
implementation in the SACI were identi-
fied, namely, the government, education
sector, private organisations and software
developers.
Finally, a proposal is made to facilitate
BIM implementation across the SACI,
which entails three key concepts,
namely leadership, strategy, and roles
and responsibilities. This research could
assist the industry to become aware of the
challenges to BIM implementation and it
could guide industry stakeholders on what
can be done to increase BIM uptake in the
industry.
Valuable findings
This section presents new and significant
findings to BIM research in South Africa:
QQ BIM implementation is a significant
challenge in the South African context,
due to the costs associated with network
infrastructure upgrades. Furthermore,
the risk of theft and vandalism poses a
specific challenge.
QQ A key finding is that the government
should start collecting digital data
to develop digital and smart cities,
to benefit from the operational and
maintenance benefits. Clients, public
or private, could require building plan
submittals in the form of BIM models.
QQ The available BIM software does not
meet the specific demands of the SACI
and market needs. The bulk of the
industry consists of smaller organisa-
tions that do not find much benefit in
complex, expensive software aimed at
large international firms.
QQ There is a growing demand for digi-
tally skilled professionals, especially
regarding BIM modelling and software
development.
36
Recommendations
The proposal to facilitate BIM implementa-
tion mentioned the need for investigating
the impact of BIM, the skills demand and
the software demand. In addition, the rec-
ommendations for future research include:
QQ The change in required digital skills to
prepare for future changes and required
skill sets.
QQ Determining the best way to measure
and compare BIM maturity amongst
organisations and countries.
QQ Measuring BIM maturity in South
Africa and continuously assessing the
maturity increase amongst different
organisations and project types.
QQ Continuously researching whether the
available software and technology in the
market meet the needs of South African
firms, especially smaller firms with
limited budgets.
QQ Best practices in South Africa to imple-
ment BIM in an organisation, in a project
and in the industry (case studies could
form an essential part of this research).
REFERENCES
Akintola, A, Root, D & Senthilkumar, V 2017.
Key constraints to optimal and widespread
implementation of BIM in the South African
construction industry. Proceedings, 33rd ARCOM
Annual Conference, 4–6 September 2017,
Cambridge, UK, 25–34.
Akintola, A, Venkatachalam, S & Root, D 2020.
Understanding BIM’s impact on professional
work practices using activity theory. Construction
Management and Economics, 38(5): 447–467.
Anthony, A 2017. Best practice in South African
construction procurement law. Chapter 16. In Thai,
K V (Ed), Global Public Procurement Theories and
Practices. Cham, Switzerland: Springer, 291–310.
Braun, V & Clarke, V 2006. Using thematic analysis
in psychology. Qualitative Research in Psychology,
3(2): 77–101.
Brito, D M, Ferreira, E A M & Costa, D B 2021.
Framework for Building Information Modelling
adoption based on critical success factors
from Brazilian public organizations. Journal of
Construction Engineering and Management, 147(7).
Building and Construction Authority 2018. Media
release on IDD implementation plan. Singapore:
Building and Construction Authority.
Chimhundu, S 2015. A study on the BIM adoption
readiness and possible mandatory initiatives for
successful implementation in South Africa. MSc
Dissertation. University of the Witwatersrand.
CIDB (Construction Industry Development Board)
2021. Construction monitor – supply and demand.
Pretoria: CIDB.
Volume 64 Number 4 December 2022
Edirisinghe, R & London, K 2015. Comparative
analysis of international and national level
BIM standardization efforts and BIM adoption.
Proceedings, 32nd CIB W78 Conference,
(Applications of IT in the Architecture,
Engineering and Construction Industry),
26–29 October 2015, Eindhoven, Netherlands,
149–158.
Ethiopian Construction Project Management Institute
2019. Roadmap for adoption and implementation
of BIM technology in the Ethiopian architectural,
engineering and construction (AEC) industry. Addis
Ababa.
Froise, T 2014. Building information modelling as
a catalyst for an integrated construction project
delivery culture in South Africa. PhD Thesis.
Port Elizabeth: Nelson Mandela Metropolitan
University.
Georgiadou, M C 2019. An overview of benefits and
challenges of building information modelling (BIM)
adoption in UK residential projects. Construction
Innovation, 19(3): 298–320.
Ghaffarianhoseini, A, Tookey, J, Ghaffarianhoseini, A,
Naismith, N, Azhar, S, Efimova, O & Raahemifar, K
2017. Building Information Modelling (BIM) uptake:
Clear benefits, understanding its implementation,
risks and challenges. Renewable and Sustainable
Energy Reviews, 75: 1046–1053.
Govender, K 2018. The awareness of integrated project
delivery and building information modelling:
Facilitating construction projects. International
Journal of Sustainable Development and Planning,
13(1): 121–129.
Gu, N & London, K 2010. Understanding and
facilitating BIM adoption in the AEC industry.
Automation in Construction, 19(8): 988–999.
HM Government (UK) 2015. Digital Built Britain
level 3: Building Information Modelling – strategic
plan. London.
Hwang, B G, Zhao, X & Yang, K W 2019. Effect
of BIM on rework in construction projects in
Singapore: Status quo, magnitude, impact, and
strategies. Journal of Construction Engineering and
Management, 145(2): 1–16.
Ismail, M 2019. BIM implementation: Theory and
practices. BIM Talk. BIM Africa.
Kekana, G, Aigbavboa, C & Thwala, W 2015.
Understanding Building Information Modelling
in the South African construction industry.
Proceedings, Organization, Technology and
Management in Construction (OTMC) Conference,
Primošten, Croatia, 2–6.
Kiprotich, C 2014. An investigation on Building
Information Modelling in project management:
Challenges, strategies and prospects in the
Gauteng construction industry, South Africa. MSc
Dissertation. University of the Witwatersrand.
Leedy, P D & Ormrod, J E 2016. Practical Research:
Planning and Design, 11th ed. Vol. 12. Harlow, UK:
Pearson.
Journal of the South African Institution of Civil Engineering
McGraw Hill Construction 2014a. The business value
of BIM for construction in major global markets.
SmartMarket Report.
McGraw Hill Construction 2014b. The business value
of BIM for owners. SmartMarket Report.
Meno, T 2020. An assessment of risk associated with
digitalisation in the South African construction
industry. University of Johannesburg.
Moodley, V, Mathye, K & Radebe, S 2016. Teaching
BIM in schools of architecture of South African
universities. University of Witwatersrand.
Mtya, A 2019. Evaluation of Building Information
Modelling (BIM) adoption, capability and maturity
within South African consulting and construction
firms. MPhil Thesis. University of Cape Town.
Mtya, A & Windapo, A 2019. Chapter 13: Drivers and
barriers to the adoption of Building Information
Modelling (BIM) by construction firms in South
Africa. In Wu, P & Wang, X (Eds), Innovative
Production and Construction. Singapore: World
Scientific Publishers, 215–223.
National BIM Council Ireland 2017. Roadmap to
digital transition for Ireland’s construction industry
2018–2021. Dublin.
Natspec 2019. Global 2019 update report. Sydney,
Australia: International Construction Information
Society.
Ogwueleka, A C & Ikediashi, D I 2017. The future
of BIM technologies in Africa: Prospects and
Journal of the South African Institution of Civil Engineering Volume 64 Number 4 December 2022
challenges. In Wu, P & Wang, X & Li, H (Eds),
Integrated Building Information Modelling. UAE:
Bentham Science, 307–314.
Pillay, N, Musonda, I & Makabate, C 2018. Use of BIM
at higher learning institutions: Evaluating the level
of implementation and development of BIM at built
environment schools in South Africa. University of
Johannesburg.
Porwal, A & Hewage, K N 2013. Building Information
Modelling (BIM) partnering framework for public
construction projects. Automation in Construction,
31: 204–214.
Royal Institution of Chartered Surveyors (UK)
2017. Building Information Modelling for project
managers. London.
Royal Institution of Chartered Surveyors (UK) 2020.
The future of BIM: Digital transformation in the UK
construction and infrastructure sector. London.
Sahil, A 2016. Adoption of Building Information
Modelling in developing countries: A
phenomenological perspective. Master’s Dissertation.
Colorado State University.
SANS 19650-1 2022. Organization and digitization of
information about buildings and civil engineering
works, including building information modelling
(BIM) – Information management using building
information modelling - Part 1: Concepts and
principles. Pretoria: South African Bureau of
Standards (SABS).
Smith, D K & Tardif, M 2009. Building Information
Modelling: A strategic implementation guide for
architects, engineers, constructors, and real estate
asset managers. New York: Wiley.
Tabesh, M 2015. An investigation to introduce BIM
in undergraduate civil engineering teaching to
improve construction processes. MEng Dissertation.
Stellenbosch University.
BS (British Standard) 2019. ISO 19650-1. Organization
and digitization of information about buildings
and civil engineering works, including building
information modelling – Information management
using building information modelling. Part 1:
Concepts and principles. London: British Standards
Institution.
Vidalakis, C, Abanda, F H & Oti, A H 2020. BIM
adoption and implementation: Focusing on SMEs.
Construction Innovation, 20(1): 128–147.
Windapo, A 2017. Determinants of the sustainability
and growth of construction contractors. University of
Cape Town. DOI: 10.20944/preprints201707.0091.v1.
Wortmann, A E, Root, P D S & Venkatachalam, D S
2016. Building Information Modelling (BIM)
standards and specifications around the world
and its local applicability to the South African
AEC sector: A critical review. Proceedings,
1st International BIM Academic Forum (BAF)
Conference, 13–15 September, Glasgow, Scotland.
37
TECHNICAL PAPER Traffic safety and the
Journal of the South African
Institution of Civil Engineering
rural road environment:
ISSN 1021-2019 (print) | ISSN 2309-8775 (online)
Vol 64 No 4, December 2022, Pages 38–48, Paper 1383 Assessing the impact
DR ROBERT AMBUNDA is a lecturer and the of combined roadway
conditions on crash incidence
lead researcher in the Mobility Lab at the
School of Engineering and the Built
Environment, University of Namibia. He
serves as an active member of the
SLOCAT-VREF Young Leaders in Sustainable
Transport programme. His research interests
R Ambunda, M Sinclair
include areas around transport policy;
sustainable, resilient and equitable transport systems; transport
modelling; climate change and adaptation; and road safety. In his current
post he is responsible for teaching (undergraduate and postgraduate), This study was designed to explore the combinational effect of the roadway and traffic
leading and coordinating research, and commercial projects in
characteristics of national rural roads on fatal and serious-injury crashes in Namibia. Namibia,
transportation. He holds an MSc and PhD in Transport Engineering from
Stellenbosch University. like many other countries in Southern Africa, experiences high numbers of high-severity
crashes, particularly on its rural road network. The study applied Generalised Poisson regression
Contact details:
Department of Civil and Department of Civil Engineering models to analyse the study dataset. The results indicate that high-severity crashes are almost
Mining Engineering Stellenbosch University always influenced by a combination of factors, not simply the presence of a single factor. For
University of Namibia Private Bag X1 example, on higher-order roads, a higher proportion of heavy vehicles in the traffic stream had
PO Box 3624 Matieland 7602
Ongwediva 9000 South Africa
the highest absolute impact on high-severity crash levels, but this effect was heightened when
Namibia hilliness, wider lanes, surfaced shoulders and higher operating speeds were also present at
T: +264 81 219 0246 the same time. On lower-order roads a higher proportion of light vehicles in the traffic stream
E: otumuna@yahoo.com / rambunda@unam.na
had the highest absolute impact on high-severity crashes, and again this was compounded in
the presence of wider unpaved shoulder widths and higher operating speeds. Overall, wider
PROF MARION SINCLAIR is an Associate
lane widths and wider unpaved shoulder widths, especially when found in combination with
Professor in Transportation within the
Department of Civil Engineering at each other, were linked to a subsequent increase in higher-severity crash levels on higher-order
Stellenbosch University. Her specialist field roads. The study shows conclusively that the development of safer roads relies not on simply
is road safety, focusing on human factors implementing minimum safety standards for individual design elements, but on understanding
contributing to road risk, as well as on
societal problems emerging from traffic
the relationship between features of road design, traffic parameters and road safety, to
crashes and injuries. She has spent the last recognise which coincidences of factors affect crash risks on different classes of road. This
19 years working in this field, shared between international road safety knowledge can result in more careful road design so that crash likelihood is reduced.
planning and implementation, and research and teaching of road safety
problems and solutions in South Africa. In her current post she is
responsible for teaching road safety to undergraduate and postgraduate
civil engineering students, and for carrying out research and INTRODUCTION interrelated factors including the geometric
commercial projects. The demand for more transport and mobil- design of the road, vehicle and environ-
Contact details: ity, which has surged in recent years, has mental factors, and driver behaviour and
Department of Civil Engineering precipitated an increase in the levels of perceptions (Fitzpatrick 2014; Turner et al
Stellenbosch University
motorisation in the modern world (Starkey 2015; Ambunda & Sinclair 2019; Chatterjee
Private Bag X1
Matieland 7602 et al 2017). One of the major issues arising & Mitra 2019).
South Africa from and being attributed to high levels of Notwithstanding the general recogni-
T: +27 21 808 3838 tion that road-user behaviour on the
motorisation is road traffic crashes, which
E: msinclair@sun.ac.za
are recognised now to be one of the leading roadways is the primary cause of road traf-
causes of premature mortality globally. The fic crashes, the road environment and its
World Health Organisation (WHO) reports properties (geometric, pavement and traffic
that about two-thirds of premature deaths characteristics) play a significant role in
caused by injuries stem from road traffic influencing crash risk, due to their impact
crashes (WHO 2018), and that on average on how a driver perceives and responds to
around 1.3 million people are killed in traf- the roadway (Taylor et al 2000; Deller 2013;
fic crashes every year. Ambunda & Sinclair 2019).
Road traffic crashes are complex Roadway components are designed
events that are difficult to predict due around minimum design standards which in
to the fact that their constituent causes turn are based on consideration of expected
generally include a combination of several driver characteristics, reactions and

Keywords: crash prediction models, road crashes,

severe road crashes, road classifications, Namibia Ambunda R, Sinclair M. Traffic safety and the rural road environment: Assessing the impact of combined roadway conditions on crash incidence.
J. S. Afr. Inst. Civ. Eng. 2022:64(4), Art. #1383, 11 pages. http://dx.doi.org/10.17159/2309-8775/2022/v64n4a4

The Journal of the South African Institution of Civil Engineering is a peer-reviewed, open-access journal licenced under a Creative Commons Attribution Licence
(CC BY-NC-ND) and is distributed internationally. The content of this journal is available to users free of charge. However, copyright remains with the South African
38 Institution of Civil Engineering. Permission for reproduction of content, with appropriate reference to the journal, can be obtained from the administrative editor.
traffic conditions (Bester & Makunje 1998). to capacitate transport engineers to detect crashes on a given national rural road is
Examples include average driver height (for and even predict high-risk road and traffic important in evaluating the different road
vertical crest standards), average reaction characteristics that have the potential to design variable covariates and alternatives
time (to determine intersection sight dis- influence the frequency and severity of road (Wang et al 2013; Glavić et al 2016), as well
tances) and numerous others. Driver behav- crashes, by using information that is readily as to understand which factors may be more
iour, however, is seldom predictably based available, dependable and cost-effective. challenging to drivers than are convention-
on simple characteristics or averages. Most Reducing the frequency and severity of ally assumed. Various methods have been
often a driver responds to a road environ- road traffic crashes continues to be a prior- used to carry out road safety analysis, with
ment based on the messages that he or she ity for transportation and traffic engineers. several statistical methods commonly
‘reads’ from the roadway conditions (Bester The UN Decade(s) of Action for Road used by road traffic engineers (Hauer 2014;
& Makunje 1998; Ambunda & Sinclair Safety (2010–2030) has highlighted the Murthy & Rao 2015; Ihueze & Onwurah
2019). These may include – but are not lim- need for road designers and authorities to 2018). These methods have, however, tended
ited to – components of road geometry, road better understand the complexity of crash to approach crash risk from the perspective
surface and traffic characteristics relating to cause and injury severity, and to work more of single or limited numbers of covariates
traffic composition and speed (Fitzpatrick proactively to reduce the potential for high- (Ben-Bassat & Shinar 2011; Liu et al 2016;
2014; Ambunda & Sinclair 2019; Chatterjee severity crashes. The UN has defined a Gitelman et al 2019).
& Mitra 2019). The response of drivers to number of road safety performance targets To this end, the goal of the study was to
individual elements is shaped by a number that countries across the world, includ- develop a method to quantitatively inves-
of things, including (among others) their age ing Namibia, have signed up to achieve tigate the extent of the combined effect of
and driving experience, the cognitive load (Ambunda 2018). These include better various national rural road geometric, pave-
placed on them at the time, their stimulus- road management strategies, which in turn ment and traffic covariables on crash occur-
response compatibility (largely determined include improved systems to understand rences for Namibia’s rural roads, providing a
by previous exposure to the particular road crash causation. straightforward, mathematically sound way
road element); and their prior expectations Road safety analyses can be useful in of predicting road crashes based on combina-
(Vaa 2014; Bongiorno et al 2017). How identifying national road sections prone to tions of these covariates. The intention was
individuals respond to a complex road envi- high road crash incidences and high injury to develop a method that could be used by
ronment is really only partially understood. severity, while determining the factors traffic engineers in similar operating condi-
Some studies (for example, looking at driver significantly contributing to the high road tions to predict crash likelihood where there
behaviour in curves, or in the presence crash occurrences (Noland & Oh 2004). may not be perfect crash data, but where
of wide shoulders) have started to build a Estimating the factors influencing road roadway and traffic data is accessible.
better understanding of how individual ele-
ments of the road environment may influ-
ence driver behaviour, but this knowledge is NAMIBIAN NATIONAL RURAL ROAD NETWORK
only partial and is particularly inadequate
when two or more characteristics exist in
parallel (Deller 2013; Ambunda & Sinclair
2019). In a­ nalysing crash locations where
high-severity crashes have been experienced
repeatedly, it becomes possible to investigate
how the combination of roadway conditions
at the time of the crash may have influenced
crash risk, based on the coincidence of fac-
tors in each location.
Most developing countries, including
Namibia, are faced with a lack of locally
derived or locally tested tools to predict
and mitigate crash likelihood. The context-
sensitivity of even the most common
road safety standards has seldom been Legend
Municipalities and towns
evaluated in countries of the developing
Trunk roads
world, where road user behaviour and
Main roads
training (in particular) may differ from National and regional borders
what is expected in other parts of the world.
Furthermore, the costs associated with
achieving or maintaining international
safety standards can be extremely high. As
such, road safety authorities in countries
such as Namibia tend to be reactive instead
of proactive to road safety issues. For this
reason it becomes particularly important Figure 1 Namibian national rural road network

Journal of the South African Institution of Civil Engineering Volume 64 Number 4 December 2022 39
Table 1 Functional classes of rural and urban roads TRH26 (CSRA 1988)
Main class
High-order rural roads (HORR)
Low-order rural roads (LORR)
DATA
Study area
Fatal- and serious-injury crash data
was sourced for the Namibian national
rural road network from the Namibian
National Road Safety Council (NRSC),
the Motor Vehicle Accident Fund of
Namibia (MVA) and the Namibian police
for the aforementioned road classes (see
Figure 1 on page 39). The national road
network was divided into several classes
(see Table 1) according to the functions of
the roads and the traffic volumes experi-
enced on these roads using the Technical
Recommendations for Highways on Road
Classifications and Access Management
(TRH 26) (CSRA 1988). The national
rural road network spans all fourteen
regions in Namibia and is maintained by
the Namibian Roads Authority through
subsidies provided by the Namibian
Government, road user taxes and other fees
Acronym Rural classes
R1 Rural principal arterial
R2 Rural major arterial
R3 Rural minor arterial
R4 Rural collector road
R5 Rural local road
R6 Rural walkway
collected by the Road Fund Administration
(RFA) (Eggleston et al 2016).
Data on roadway design and conditions
was obtained from the Roads Authority
(RA) of Namibia. This focused on traf-
fic volumes, speeds (operational, design
and posted), road lane characteristics,
road shoulder characteristics, road align-
ment, sight distances, access density, and
pavement conditions. The collection of
roadway data also involved on-site data
collection on selected rural roads to
supplement or corroborate data obtained
from the relevant authorities. A summary
of the covariates used in the model devel-
opment procedure is presented in Table 2.
Sixteen covariates were included in the
model development process, of which 14
were tested in the models. The covariates
were divided into two groups, namely
numerical and categorical covariates. Of
these covariates, nine related to the geo-
metric characteristics of the rural roadway
system and seven related to the character-
istics of the rural roadway, describing the
traffic modal split, terrain and roadway
surface types and conditions.
Crash data
The study focused on higher-severity
crashes (fatal- and serious-injury crashes)
on trunk and main roads on the national
rural road network. Crash data was
obtained from the three main authorities
responsible for crash reporting in Namibia
(NRSC, MVA and the Namibian police).
Using the crash data, the crash rates were
determined as crashes per million vehicle
kilometres per kilometre of road section
over the study period of five years (2012–
2016) (see Figure 2). The study analysis
required continuous, reliable and accurate
historical crash data, with information
on traffic characteristics, traffic exposure
covariates and the road environment,
all of which are vital for an appropriate
statistical analysis. The historical crash
data collected from the Namibian road
safety authorities was not geo-coded (this
is a common problem in crash reporting
in many developing countries), with the
majority of the site-specific crash informa-
tion missing. As a result, it was necessary
for the study to address the deficiencies in
both the crash and roadway data by devel-
oping an approach to remove incomplete
crash record data and gather additional

Table 2 Descriptive statistics of crash model covariates

Descriptive statistic summary of covariates

Covariate Minimum Maximum Mean Standard deviation Variance

AADT_Light (AADTL) 85 14005 2328.44 2921.117 8532926.924

AADT_Heavy (AADTH) 2 1400 345.29 376.970 142106.055

AADT_Total 91 15362 2673.73 3231.762 10444282.947

Operating_Speed (Ops) 0 120 44.02 53.010 2810.020

Lane_Width (LW) 2.940 12.450 5.156 2.552 6.513

Numerical

No_Lanes (NL) 1 6 1.79 .683 0.466

Surface_SW (SSW) 0.000 3.175 0.255 0.562 0.316

Ground_SW (GSW) 0.000 8.9900 1.713 0.652 0.425

Horizontal_(Curves/Length) (Hor) 0.000 0.709 0.176 0.143 0.020

Access_Density (AD) 0.000 0.409 0.121 0.086 0.007

Section_Length (SL) 12.230 22.967 15.462 1.486 2.207

SSD 15 2215 179.76 63.002 3969.274

Surface_type (ST) 0 1 – – –
Categorical

Shoulder_type (ShoT) 0 1 – – –

Terrain_Vertical (TV) 0 1 – – –

Pavement_Condition (PC) 0 1 – – –

40 Volume 64 Number 4 December 2022 Journal of the South African Institution of Civil Engineering
site-specific information to enable a com-
prehensive statistical analysis of crash
locations on national rural roads.
METHOD
Model development
The study developed Generalised Linear
Models (GLMs) to predict national rural
road crash incidences and investigate
the combinational effects of geometric
and traffic characteristics on road safety.
The approach involved the aggregation
of design and traffic factors, and fatal
and serious injury (FSI) to satisfy the
mathematical form assumptions of the
prediction models – namely (i) to gener-
ate logical results that do not cause the
prediction of negative crash incidences and
should ensure a prediction of zero crashes
for zero values of exposure and length vari-
ables, and (ii) there must exist a known link
function that can linearise the model form
for the purpose of coefficient estimation
(Ghanbari 2017). The modelling approach
was tested and validated using data from
two datasets, representing FSI crashes
on higher- and lower-order rural roads.
Through the use of data manipulation it
was possible to satisfy the assumptions of
the GLMs and thus develop robust crash
prediction models (CPMs). The study pro-
duced and compared the GLM CPMs using
the Generalised Poisson (GP) models and
Negative Binomial models to determine the
best-performing model.
The GLMs that were developed share
a number of unique properties, such as
linearity and a common method for param-
eter estimation that allow the development
of effective CPMs for highways. The GLMs
consist of the following three components
(Oppong 2012):
1. A random component, which speci-
fies the conditional distribution of the
response variable Yi given the explora-
tory variables xij
2. A linear function of the regression
variables, called the linear predictor,
on which the expected value μi of Yi
depends:
Yi = α + β1 Xi1 + … + βk Xi1 = x'i β(1)
3. An invertible link function:
g(μi) = ηi(2)
 The invertible link function transforms
the expectation of the response to the
linear predictor. The inverse of the link
function is sometimes called the mean
function:
g–1(ηi) = μi (3)
The GLMs are an extension of the linear
models to include response variables that
follow any probability distribution in the
exponential family of distributions (Eenink
et al 2005; Field 2013; Denis 2021). Several
measures were used to examine the validity
of the models and how they fit the data
in the study. These measures, given in
Section 4.5, include the Scaled Deviance
(SD), Pearson’s Chi-Square (PCS), Akaike
Information Criterion (AIC), Corrected
Akaike Information Criterion (AICc) and
Bayes Information Criterion (BIC).
Outlier analysis
The study applied two-dimensional (2D)
box plots as a diagnostic tool for detecting
outliers and data-influential points, and
ultimately used the winsorisation technique
to address detected outliers in the dataset
(see Figure 2). Winsorising is the process of

2.8 0.20

2.6
0.18
2.4
0.16
2.2

2.0 0.14

1.8
0.12
1.6
Crash_Rate (W)
Crash_Rate

1.4 0.10

1.2 0.08

1.0
0.06
0.8

0.6 0.04
Median = 0.0261
0.4 25%–75% = Median = 0.0261
(0.0113, 0.0678) 0..02
25%–75% =
0.2 Non-outlier range = (0.0113, 0.0678)
(0, 0.1977) Non-outlier range =
0
0 Outliers (0, 0.1524)
Extremes Outliers
–0.2 –0.02
(a) (b)

Figure 2 2D box plots of crash rate distribution before and after winsorisation

Journal of the South African Institution of Civil Engineering Volume 64 Number 4 December 2022 41
replacing a specified set of extreme values of
a given variable in a set of sample data with
specified values computed from the data
(Chambers 2000; Reifman & Keyton 2010).
The outliers were not removed from the
dataset, even though they were extreme data
points not typical of the rest of the crash
data used in the study. The study recognised
that the presence of these data points could
have a significant impact on the estimates
of the safety performance parameters of the
models, thus a more robust outlier detecting
and inclusion method was adopted in the
study for outlier resistance modelling to
allow for robust safety inferences by down-
weighting (winsorising) the outlying crash
observations rather than rejecting them.
The winsorised models provide a superior
fit and standard deviations for the param-
eter estimates.
A predefined rule was used to adjust an
outlying (positive) value Yi of the dataset
variable Y downwards, leaving the remain-
ing values unchanged (Hicks & Fetter 1991;
Reifman & Keyton 2010). The value of the
adjusted variable is denoted Yi * and the cor-
responding winsorised estimator adjusted to
a fixed cut-off is represented by Equation 4.
n
Ŷt = ∑j=1 adj w tj yj(4)
Where:
t = truncation level
yj = reported crash rate for the jth unit
∑n w j
adj =
∑nw tj
t On the biplot, the eigen vectors ê1 and
wtj = wj, if wj yj ≤ t , if wj yj > t
yj
The weights of the observations of which
the expanded weighted value was larger
Table 3 Goodness-of-fit measures for all CPMs
Parameter CPM 1 (high-order
rural roads)
Scaled Deviance (SD)
Pearson Chi-Square (PCS)
Scaled Pearson Chi-Square (SPCS)
Akaike’s Information Criterion (AIC)
Finite Sample Corrected AIC (AICc)
Bayesian Information Criterion (BIC)
1 Information criteria are in smaller-is-better form.
2 The full log likelihood function is displayed and used in computing information criteria.
42
than t (where t = 0.75) were truncated so
that the expanded values were equal to t.
The t value was determined after observ-
ing the distribution of the data points and
where the extreme data points were posi-
tioned. The truncated portions were then
smoothed over for all observations.
Crash model biplots
Biplots are a graphical representation of
information in an n × p data matrix, with
information in rows representing samples
and information in columns representing
covariates. In the principal component (PC)
analysis, plots were obtained by graphing
the first two principal components of the
units (Gower et al 2011). In biplots the idea
is to add information about the covariates
to the PC graph.
Construction of biplots
The best two-dimensional approximation
of data in an n × p matrix was determined
by approximating the jth observation vector
xj in terms of the sample values of the first
two PCs. The approximation is given by
Equation 5:
xj = x + ŷj1ê1 + ŷj2 ê2(5)
Where: ê1 and ê2 are the first two eigen
vectors of (n – 1)S = x'c xc . Where S denotes
the contribution of the first principal
component to the total variance, x'c is the
original observed data matrix, and xc is
equal to the mean corrected data with row
vectors (xj – x)’.
ê2 define the plane. The coordinates ŷj1
and ŷj2 for j = 1, …, n, define the n units in
that plane – principal component scores.
The variables x1 … xp are positioned
Goodness-of-fit1, 2
Negative Binomial Models
CPM 2 (low-order
rural roads)
1 499.000 957.000
14.148 46.288
1 499.000 957.000
–2 715.819 –172.822
–2 715.671 –172.759
–2 662.694 –148.503
Volume 64 Number 4 December 2022
on the graph by the row vectors of
Ê = [ê1, ê2]:p × 2, since:
Y = [Y1 Y2] = [ê'1 ê'2][x1 : xp](6)
The lengths of the vectors from x1 to xp
were adjusted to ensure that all the vari-
ables were plotted on the same graph as the
points (ŷj1, ŷj2); j = 1, … n.
Assessment of goodness-of-fit
The study used five goodness-of-fit s­ tatistic
tests to evaluate the fit-of-the-crash
­prediction models developed to the national
rural road network – the Scaled Deviance
(SD), Pearson Chi-Square (PCS), Scaled
PCS (SPCS), Akaike’s Information Criterion
(AIC), Finite Sample Corrected AIC (AICc)
and Bayesian Information Criterion (BIC)
values. A summary of the goodness‑of‑fit
measures of the road crash models d ­ eveloped
is p
­ resented in Table 3.
The Generalised Poisson (GP) regression
crash prediction models were found to be
the best fit for the datasets compared to
Negative Binomial (NB) models, due to their
significantly lower AIC, AICc and BIC values
generated. CPMs with negative AIC, AICc
and BIC values have better performance
compared to positive test values due to mini-
mal information loss. Further, comparing the
performance of the GP and NB models, the
GP models generated higher values of the
different residuals (SD, PCS and SPCS) than
those generated by the NB models.
RESULTS
Crash model biplots
The model biplots developed by the study
indicate the variance structure of the study
Generalised Poisson Models
CPM 1 (high-order CPM 2 (low-order
rural roads) rural roads)
2 452.567 1 033.016
2 951.391 813.087
4 618.773 1 669.331
9 985.664 –1 417.109
–9 985.499 –1 417.021
–9 911.472 –1 387.927
Journal of the South African Institution of Civil Engineering
 Horizontal_(Curves/Length)

Shoulder_type
AADT_Heavy

Ground_SW
AADT_Light

Ops
SSD
Terrain_Vertical
PC 2 (16%)

No_Lanes
Pavement _Condition
Lane_Width
Surface_Type

Access_Density

HORR

Surface_SW LORR
PC 1 (26%) 0.75 alpha ellipses

Figure 3 Principal component biplot

covariates for the entire rural road network Table 4 Principal component summary
dataset. The biplot generated in Figure 3 Eigenvalues (all rural roads)
shows the projected observations (points) Extraction: principal components
Principal
and the projected covariates (vectors) component Cumulative
approximated by the first two principal Eigenvalue % Total variance Cumulative (%)
eigenvalue
components (PCs) shown in Table 4.
1 3.676 26.257 3.676 26.257
The PC biplot, represented graphi-
cally in Figure 3, describes the distribution 2 2.240 15.999 5.916 42.256
and possible influence of the principal
components on higher-severity crash levels 3 1.337 9.549 7.253 51.805

on both road classifications – high-order 4 1.221 8.721 8.474 60.525

rural roads (HORRs) and low-order rural
roads (LORRs). The Principal Component 5 1.104 7.885 9.577 68.411
Analysis (PCA) does not discard any covari- 6 0.865 6.180 10.443 74.590
ates. Instead, it reduces the large number
of dimensions by constructing PC plots 7 0.859 6.138 11.302 80.728
(vectors and clusters). The PCs describe
8 0.714 5.102 12.016 85.830
the variation and account for the varied

Journal of the South African Institution of Civil Engineering Volume 64 Number 4 December 2022 43
influences of the covariates (Greenacre
2010). The PC plots thus indicate graphi-
cally how covariates correlate with each
other. Specifically, the angles between the
vectors inform how the covariates correlate
with each other: when two vectors are
close, forming a narrow angle (less than
90 degrees) this implies positive correla-
tion (belonging to the same cluster group),
while a wide angle (equal to or greater than
90 degrees) points to negative correlations
(belonging to different cluster groups)
(Greenacre 2010). Such influences can then
be traced back from the PC plots using the
developed crash prediction models to dis-
cern what produces the differences among
the covariates and the covariate clusters
(Acar & Yener 2009; Ginanjar et al 2017).
The first two PCs, represented by the
cluster of dots, explain 26 percent (PC1) and
16 percent (PC2) of the variance contributed
by the covariates on the different road clas-
sifications at an alpha ellipses level of 0.75
(see Figure 3). Without factoring in autocor-
relation, the biplot gives an indication of
which covariates are likely to explain the
correlation with crash incidences. For the
HORRs in the dataset, the model variance in
PC 1 is potentially explained by:
QQ The widths of the surfaced shoulders
(SSW).
The model variance for HORRs in PC 2 is
potentially explained by:
QQ The access density (AD)
QQ The number of lanes on the HORR sec-
tions (NL)
QQ The hilliness of the vertical alignment
(TV)
QQ The stopping sight distance (SSD) avail-
able to drivers on HORRs (SSD)
QQ The heavy and light vehicle annual aver-
age daily traffic (AADTH and AADTL)
QQ The number of horizontal curves per
km on the road sections (Hor).
For the LORRs in the dataset, the biplot
indicates that the model variance in PC 1 is
potentially explained by:
QQ The width of the unpaved shoulder
(GSW) on the LORRs
QQ The type of hard shoulder available on
the LORR sections (ShoT)
QQ The 85th percentile operating speed
(Ops).
The variance in the model for LORR sec-
tions in PC 2 is potentially explained by:
QQ The condition of the pavement surface
on the LORRs (PC)
QQ The width of the available lanes (LW)
QQ The type of surface (paved or unpaved)
on the LORRs (ST).
44
A detailed crash prediction model analysis
(coefficient b* estimates) for the HORRs
(CPM 1) and LORRs (CPM 2) is presented
in the following section.
Crash prediction models estimates
The estimates (standardised regression
coefficient b*) for all the Generalised
Poisson Crash Prediction Models
(GP-CPMs) developed and applied in this
study to fit the datasets used – high-order
rural roads (HORR) and low-order rural
roads (LORR) – are summarised in Table 5.
The crash prediction models tested all the
study covariates, with the HORR GP-CPM
eventually exhibiting the largest number
(five covariates) of covariates showing
statistically significant (p < 0.05) relation-
ships with crash incidences. The CPM on
LORR had four covariates that presented
statistically significant associations with
crash incidences.
Of the fourteen covariates tested in the
models, seven covariates were found to
exhibit statistically significant (p < 0.05)
effects in the CPMs. The following seven
covariates tested in all the crash prediction
models that did not show any statistically
significant affiliation with the crash inci-
dences on the rural roads are:
1. The number of lanes available to traffic
(NL)
2. The type of surface on the different
road classifications (ST)
Table 5 Best-fit Road Crash Prediction Models (GP-CPMs) performance
Tested parameters
CPM 1
High-order rural roads
AADT_Heavy (AADTH)
85th Percentile Speed (Ops)
Lane Width (LW)
Surface_SW (SSW)
Terrain_Vertical (TV)
AADT_Light (AADTL)
No_Lanes (NL)
Surface_Type (ST)
Shoulder_Type (ShoT)
Ground_SW (GSW)
Horizontal_(Curves/ length) (Hor)
Access_Density (AD)
Pavement_Condition (PC)
SSD
Volume 64 Number 4 December 2022
3. The proportion of hard shoulder sur-
faces (ShoT)
4. The number of horizontal curves per
km length of rural road section (Hor)
5. The number of access points per km
road section length (AD)
6. The condition of the pavement surface
(PC)
7. The stopping sight distance available on
the rural road sections (SSD).
Three covariates were shown to be influen-
tial (p < 0.05) in both CPMs, with varying
effects (standardised coefficient b* esti-
mates) on crash incidences. These covari-
ates are: (1) the operating speed (Ops) on
the road sections, (2) the vertical terrain –
hilliness (TV), and (3) the ground shoulder
width (GSW) on the road sections (all
indicated in red in Table 4). The 85th per-
centile operating speed exhibited positive
association with crash incidences in both
CPM 1 (0.032) and CPM 2 (0.049). Similar
to the operating speed, the hilliness of
the vertical alignment exhibited a positive
association to crash incidences on all rural
road classifications (CPM 1 b* = 0.112;
CPM 2 b* = 0.066). The ground shoulder
width (GSW) demonstrated statistically
significant relations to crash incidences
in both CPM 1 and CPM 2. In CPM 1, the
GSW covariate indicated a positive relation
to crash incidences, with a 0.108 coefficient
b* estimate. In contrast, the GSW covariate
in CPM 2 indicated a negative relation to
Standardised regression
coefficient (coefficient b*)
CPM 2
Low-order rural roads
0.682 –
0.032 0.049
0.137 –
– –0.138
0.112 0.066
– 0.315
– –
– –
– –
0.108 –0.205
– –
– –
– –
– –
Journal of the South African Institution of Civil Engineering
crash incidences, with a –0.205 coefficient
b* estimate.
Two covariates exhibited statistically
significant (p < 0.05) parameter estimates
in CPM 1 only. These covariates are:
(1) the proportion of heavy vehicles in the
annual average daily traffic (AADTH), and
(2) the width of the rural road lanes (LW).
The proportion of heavy vehicles in the
traffic stream (AADTH) covariate demon-
strated positive associations with the crash
incidences. The standardised regression
coefficient b* for the AADTH covariate is
0.682. The model results also indicate that
the lane widths (LW) are positively related
(b* = 0.0.137) to the crash incidences on
higher-order rural roads. Two other covari-
ates emerged with statistically significant
coefficient b* estimates on lower-order
rural roads (CPM 2). These covariates are:
(1) the surfaced shoulder width (SSW) and
(2) the proportion of light vehicles in the
annual average daily traffic (AADTL) on
the road sections. The SSW was identified
to have a negative relation (b* = –0.138)
with crash incidences in CPM 2 while the
AADTL exhibited a positive relation (b* =
0.315) to crash incidences.
DISCUSSION
The study developed novel crash prediction
models for both classes of national rural
road classifications in Namibia – high-
order rural roads (HORR) and low-order
rural roads (LORR). The results of the
crash prediction models developed provide
a platform to further link and examine the
impact of road and traffic characteristics
on driver behavioural traits and their
distribution across the national rural road
network. The study found several positive
associations between road characteristics
and fatal- and serious-injury crashes. In
particular, hilliness, wide travel lanes,
higher proportions of heavy vehicles (HVs)
on higher-order roads and of light vehicles
(LVs) on lower-order roads, operating speed
and shoulder widths. Possibly more inter-
esting is the fact that combinations of these
factors resulted in a higher likelihood of
fatal and serious injury crashes. Thus, for
example, on higher-order roads, while the
proportion of HVs had the highest individ-
ual impact, its impact was increased when
the road was hilly, also where there were
wider shoulders or higher operating speeds.
Of course, the combined effect of multiple
factors makes sense, as each of these ele-
ments individually has the potential to
influence the level of attention that a driver
needs to be paying to make sense of the
road environment, and to ‘read’ the road
cues safely. Where multiple factors need to
be interpreted, cognitive overload is more
likely to occur, and driver misperceptions
are likely, which can in turn lead to driver
errors and misjudgements.
The positive impact of hilliness, or the
effect of vertical alignment on crash occur-
rence, is interesting, as it is a factor that
appears to be largely innocuous unless it
exists in combination with other factors, as
found in the study. While several previous
studies found no significant correlation
between hilliness as a single variable with
the occurrence of road crashes (Bester &
Makunje 1998; Taylor et al 2002; Gitelman
et al 2016), sites demonstrating a combina-
tion of hilliness and bendiness have been
found to experience an elevated frequency
of road crashes (Wang et al 2013; Ambunda
2018). This has been generally attributed to
poor coordination between the horizontal
and vertical alignment, leading to poor
driver perceptions and driving errors
(Bester & Makunje 1998; Walmsley et al
1998; Hanno 2004; Laird et al 2010). Such
previous research thus supports the study
finding in this research, that hilliness can
be a significant contributor to crash risk
when it occurs in combination with other
road design and traffic parameters.
The results from CPM 1 found that an
increase in the width of the travel lanes was
associated with a higher occurrence of FSI
crashes, when it occurs in combination with
other road design and traffic parameters.
On Namibia’s high-order rural roads, lane
widths are mostly wide (LW > 3.5 m) with
extremely narrow (SSW < 1.5 m) or zero
surfaced shoulder widths (Ambunda 2018).
Previous research tells us that higher lane
widths and longer forward visibility are both
associated with increased speed, as drivers
feel more comfortable with the available
space ahead of them (Warrd et al 2004; Ma
et al 2009; Liu et al 2016). Typically, the
higher widths are also associated with high
levels of lateral lane deviations. In the local
Namibian context, the Namibian National
Road Safety Council has already identified
higher levels of same-direction road crash
frequency where there are wide lanes pres-
ent (NRSC 2012; Nghishihange 2018). While
higher speeds may directly affect crash
injury levels, it is likely the higher number of
crashes themselves are precipitated through
merging manoeuvres, brought about by
greater speed differentials.
Another important finding of the study
is the influence of the higher proportion of
heavy vehicles (HV) and light vehicles (LV)
in the traffic stream on the safety of road
users on the different selected national
rural roads. The novel models demonstrat-
ed that an increase in the volume propor-
tion of HVs and LVs, in combination with
other factors, increased the occurrence of
FSI crashes on HORRs and LORRs respec-
tively. The effect contributed by this modal
split may partly be attributed to the speed
variance between vehicles on high-speed
highways (Gargoum & El-Basyouny 2016),
as well as the fact that delays caused by
slow-moving vehicles have been shown to
increase the potential for risky overtaking
behaviour by drivers of faster vehicles (Li et
al 2020; Huang et al 2018; Emo et al 2016).
Both of these elements are likely to be
more pronounced on HORRs, where speed
is a motorist priority, than on LORRs.
Another possible contributing factor in the
Namibian context is that HV drivers often
drive long distances. Lack of quality sleep
and severe fatigue are significantly associ-
ated with more frequent human errors
(Aworemi et al 2010; Gastaldi et al 2014).
Fatigue itself is difficult to measure and to
prove as a contributory cause in crashes,
but even so, should not be overlooked as a
potentially important element in explain-
ing the relationship between HV presence
and high severity crashes on HORRs.
The operating speed of a road is of
course fundamental to the incidence of
crashes, with research showing conclusively
that more crashes occur with increasing
operating speeds (Bamdad & Mirbaha
2018; Wang et al 2018). This study found
that operating speed demonstrated a statis-
tically significant positive association with
the occurrence of FSI crashes in all the
CPMs, an effect that was magnified when
high operating speeds co-existed with
several road design and traffic factors. The
study findings corroborated previous stud-
ies on the impact of operating speeds on
Namibian national roads (Ambunda 2018;
Ambunda & Sinclair 2019).
The study found that the ground
shoulder width (GSW) had dissimilar but
statistically significant influences on high-
order and low-order national rural roads
when combined with several factors. On
high-order rural roads, the model demon-
strated that ground shoulder widths in this
study had a positive association with FSI
crash frequencies. International research
has produced mixed findings on the impact

Journal of the South African Institution of Civil Engineering Volume 64 Number 4 December 2022 45
of wide hard shoulders on traffic safety
(Stamatiadis et al 2009; Summala 1996;
Knuiman et al 1993), but there is consensus
regarding the role of the hard shoulder type
itself. Paved shoulders are overwhelmingly
shown to be safer than unpaved shoulders on
higher-order roads (Ogden 1997; Hallmark et
al 2009). As such, road managers in Namibia
are more likely to see improvements in safety
through the paving of gravel shoulders,
rather than by increasing shoulder widths.
On low-order rural roads – mostly low-­
volume paved roads or one-lane gravel roads
– the ground shoulder width, in combination
with other factors, demonstrated a negative
correlation to the occurrence of FSI road
crashes. This finding in the study suggests
that an increase in the width of the ground
shoulder on LORRs may have the potential
to reduce crash frequencies. This confirms
the findings of a study by Ambunda (2018),
that ground shoulder widths on low-order
rural roads are mostly non-compliant with
design guidelines (existing GSW < Technical
Recommendations for Highways 17 (TRH
17) GSW), so increasing the GSW to bring
them in line with design standards could
reduce crash frequencies. Driver speed selec-
tions tend to be lower on roads with gravel
shoulders due to visual cues (colour differ-
ence between the paved roadway surface
and the gravel-surfaced shoulder) that give
a perception of a narrower driving lane. The
finding on the impact of GSW on low-order
rural roads also corroborates results from
several international researchers (Zegeer et
al 1987; Gitelman et al 2019).
The surfaced shoulder width (SSW)
demonstrated a negative association with
the frequency of FSI road crashes on lower-
order rural roads (CPM 2) when combined
with other design and traffic factors. This
means that increasing the width of paved
shoulders on road sections could result in
a reduction in road crashes. This finding
goes hand in hand with the design compli-
ance finding, where existing SSWs were
found to be significantly non-compliant
(existing SSW < recommended TRH 17
SSW) with design guidelines. International
research has confirmed that drivers tend to
select lower speeds on narrow travel lanes
(LW < 3.2 m) with narrow surfaced shoul-
ders (SSW < 1.5 m) due to the perception
of lower safety (Karlaftis & Golias 2009;
Martens et al 1997; Godley et al 2004).
Arguably, in the local Namibian con-
text, the combination of narrow shoulders
and wider travel lanes (LW > 3.5 m) may
be encouraging drivers to select high
46
speeds due to a false sense of security and
perceived space to correct driving errors.
Despite the wider lanes on the Namibian
national road network (Ambunda 2018),
the narrow shoulder could also inadver-
tently lead drivers to steer away from the
left shoulder and drive more closely to the
centre of the rural road – this is a phenom-
enon found in previous research (Liu et al
2016; Ambunda & Sinclair 2019). In such
a case, the likelihood of head-on crashes
increases significantly. High levels of head-
on crashes have been confirmed by crash
statistics from the Namibian National Road
Safety Council (NRSC 2012).
International standardisation of geo-
metric design elements tends to give us
confidence that roadway design features
have been designed as safely as possible.
However, given that there are over 1.3
million deaths each year on roads that
largely conform to such standards, it would
appear that there may well be room for
improvement. Moreover, whether a road
is used safely by drivers often comes down
to the behaviour that the driver believes to
be expected, based on his or her reading
of the road design. In looking at possible
reasons for the relationships between the
covariables most significantly associated
with increased crash risk in this study, it is
apparent that those elements that appear
to have the highest direct impact on safety
levels are those that potentially mislead the
driver (e.g. into thinking they are safer than
they are: wider lanes and higher speed lim-
its may be examples here) or those that cre-
ate an element of surprise (e.g. combined
hilliness and bendiness) or a degree of
ambiguity about what is expected or safe.
CONCLUSION AND
RECOMMENDATIONS
The study developed novel crash prediction
models (CPMs) for the various road classifi-
cations on the Namibian national rural road
network. The method developed predicts
crash likelihood in the future for Namibian
roads, by recognising which factors, and
importantly, which combinations of fac-
tors, have historically led to higher crash
occurrence. The study results showed that
individual safety standards are potentially
less effective when multiple factors co-exist,
therefore leading to more research being
required to develop improved standards
that recognise that factors seldom exist
in isolation, and account for the impact of
high-cognitive loading on road users.
Volume 64 Number 4 December 2022
The insights from the study will also
contribute to the United Nations global
road safety performance targets, specifically
those targets which focus on safer roads and
road environments for all road users, based
on high technical/design standards with
reference to road safety and incorporating
safe-system principles. Using the existing
rural road design data as the key cog in
the CPMs, the models developed are also
intended to supplement and potentially
replace traditional road safety tools which
are typically dependent on high-quality
and timely crash reporting. The models’
application and insights could expand the
road safety stakeholder’s ability to determine
road sections with potential crash risk and
eliminate the risk for road users. These
CPMs are novel in that they quantify the
operational characteristics of the roads at
existing high-crash locations, and identify
elements that apparently contribute to
elevated crash risk, either because they
mislead the driver or because they create an
element of surprise and ambiguity on the
road. The CPMs also present an opportu-
nity to examine, through before and after
studies, the impact of the road environment
on human-centred crash risk factors by
comparing the changes that may result from
road characteristic improvements.
The replication and applicability of the
models at an aggregated level in countries
with similar rural road environments
will need to be investigated further for
comparison purposes, as the models are
predicated on accurate road design and
traffic data, which may not always be pres-
ent. However, such data is arguably easier
to collect and corroborate than detailed
crash data, which is notoriously unreliable
in many parts of the developing world. The
models present an opportunity to identify
(and treat) potential high-risk locations for
high-severity crashes, based primarily on
geometric and traffic data, using reliable
and accurate crash histories. This is a prac-
tical and more proactive alternative to the
traditional process of waiting for a serious
crash problem to develop and addressing it
after the fact.
DECLARATION OF INTEREST
None
ACKNOWLEDGEMENTs
The authors would like to thank the
Namibian National Road Safety Council
Journal of the South African Institution of Civil Engineering
and the Roads Authority of Namibia for
providing the crash data and design data
respectively.
FUNDING
This research did not receive any specific
grant from funding agencies in the public,
commercial or not-for-profit sectors.
REFERENCES
Acar, E & Yener, B 2009. Unsupervised multiway data
analysis: A literature survey. IEEE Transactions
on Knowledge and Data Engineering, 21(1): 6–20.
DOI: 10.1109/TKDE.2008.112.
Ambunda, R 2018. Analysis of the relationship
between fatal rural road crashes and road
design elements: Namibian case study. Master’s
Dissertation. Stellenbosch University.
Ambunda, R & Sinclair, M 2019. Effect of two-lane two-
way rural roadway design elements on road safety.
International Journal of Innovative Technology and
Exploring Engineering, 8(4): 632–637.
Aworemi, R J, Abdul-Azeez, A I, Oyedokun, J A &
Adewoye O J 2010. Efficacy of drivers’ fatigue on
road accidents in selected southwestern states
of Nigeria. International Business Research,
3(3): 225–233.
Bamdad, M B & Mirbaha, B 2018. Evaluating the
relationship between operating speed and collision
frequency of rural multilane highways based on
geometric and roadside features. Civil Engineering
Journal, 4(3): 609. DOI: 10.28991/cej-0309120
Ben-Bassat, T & Shinar, D 2011. Effect of shoulder
width, guardrail and roadway geometry on driver
perception and behaviour. Accident Analysis &
Prevention, 43(6): 2142–2152.
Bester, C J & Makunje, J A 1998. The effect of rural
road geometry on safety in southern Africa.
Proceedings, International Symposium on Highway
Geometric Design Practices, 30 August – 1
September 1995, Boston, MA, pp 1–10.
Bongiorno, N, Bosurgi, G, Pellegrino, O & Sollazzo G
2017. How is the driver’s workload influenced by the
road environment? Procedia Engineering, 187: 5–13.
DOI: 10.1016/j.proeng.2017.04.343.
Chambers, R L 2000. Winsorisation for identifying and
treating outliers in business surveys. Proceedings,
2nd International Conference on Establishment
Surveys, 17–21 June 2000, Buffalo, NY, pp 717–726.
Chatterjee, S & Mitra, S 2019. Safety assessment of
two-lane highway using a combined proactive and
reactive approach: Case study from Indian national
highways. Transportation Research Record, 2673(7).
CSRA (Committee of State Road Authorities) 1988.
TRH 26: Road Classification and Access
Management. Pretoria: CSRA.
Deller, J 2013. The influence of road design speed,
posted speed limits and lane widths on speed
selection: a literature synthesis. Proceedings,
Australasian Transport Research Forum, 2–4
October 2013, Brisbane, Australia.
Denis, D 2021. Multiple linear regression. In Denis, D
(Ed). Applied Univariate, Bivariate, and Multivariate
Statistics using Python. New York: Wiley, 232–285.
Eenink, R, Reurings, M, Elvik, R, Cardoso, J, Wichert,
S & Stefan, C 2005. Accident prediction models and
road safety impact assessment: Recommendations
for using these tools. Sustainable Development,
Global Change & Ecosystem, 506184: 1–20.
Eggleston, Z, Hansen, P & Carrera, F 2016.
Improvement of road safety: Route-based assessment
on major roads in Namibia. Worcester, MA:
Worcester Polytechnic Institute.
Emo, K, Matthews, G & Funke, G J 2016. The slow
and the furious: Anger, stress and risky passing in
simulated traffic congestion. Transportation Research
Part F: Traffic Psychology and Behaviour, 42: 1–14.
Field, A 2013. Discovering Statistics using IBM SPSS
Statistics, 4th ed. London: SAGE.
Fitzpatrick, C D 2014. The effect of roadside elements
on driver behaviour and run-off-the-road crash
severity. Master’s Dissertation. Boston, MA:
University of Massachusetts.
Gargoum, S A & El-Basyouny, K 2016. Exploring
the association between speed and safety: A path
analysis approach. Accident Analysis & Prevention,
93: 32–40. DOI: 10.1016/j.aap.2016.04.029.
Gastaldi, M, Rossi, R & Gecchele, G 2014. Effects of
driver task-related fatigue on driving performance.
Procedia – Social and Behavioural Sciences,
111: 955–964. DOI: 10.1016/j.sbspro.2014.01.130.
Ghanbari, A 2017. Impact of pavement condition
indicators on road safety in New Brunswick. Master’s
Dissertation. Fredericton, NB: University of New
Brunswick.
Ginanjar, I, Pasaribu, U S & Indratno, S W 2017.
A measure for objects clustering in principal
component analysis biplot: A cause study in
inter-buses maintenance cost data. Proceedings,
American Institute of Physics Conference, 1827(3).
DOI: 10.1063/1.4979432.
Gitelman, V, Pesahov, F, Carmel, R & Bechor, S 2016.
The identification of infrastructure characteristics
influencing travel speeds on single-carriageway
roads to promote self-explaining roads.
Transportation Research Procedia, 14(6): 4160–4169.
DOI: 10.1016/j.trpro.2016.05.387.
Gitelman, V, Doveh, E, Carmel, R & Hakkert, S 2019.
The influence of shoulder characteristics on
the safety level of two-lane roads: A case study.
Accident Analysis & Prevention, 122(2): 108–118.
DOI: 10.1016/j.aap.2018.10.003.
Glavić, D, Mladenovic, M N, Stevanovic, A, Tubic, V,
Milenkovic, M & Vidas, M 2016. Contribution
to accident prediction models development
for rural two-lane roads in Serbia Mark. Promet-
Traffic & Transportation, 28(4): 415–424.
DOI: 10.7307/ptt.v28i4.1908.
Godley, S T, Triggs, T J & Fildes, B N 2004. Perceptual
lane width, wide perceptual road centre marking
and driving speeds. Ergonomics, 47: 237–256.
DOI: 10.1080/00140130310001629711.
Gower, J, Lubbe, S & Le Roux, N 2011.
Understanding Biplots. New York: Wiley.
DOI: 10.1002/9780470973196.ch10.
Greenacre, M 2010. Biplots in practice. Journal
of the Royal Statistical Society: Series
A (Statistics in Society), 51(4): 375–392.
DOI: 10.1111/j.1467-985x.2012.01069_5.x.
Hallmark, S L, McDonald, T J, Tian, Y & Andersen, D J
2009. Safety benefits of paved shoulders. InTrans
Project 05–239. Ames, IA: Iowa State University.
Hanno, D 2004. Effects of the combination of horizontal
and vertical alignments on road safety. MSc
Dissertation. Vancouver: University of British
Columbia.
Hauer, E 2014. Statistical Road Safety Modelling.
Transportation Research Record, 1897(1): 81–87.
DOI: 10.3141/1897-11.
Hicks, S & Fetter, M 1991. An evaluation of robust
estimation techniques for improving estimates of total
hogs. Fairfax, VA: US Department of Agriculture.
Huang, Y, Sun, J & Zhang, L H 2018. Effects of
congestion on drivers’ speed choice: Assessing the
mediating role of state aggressiveness based on taxi
floating car data. Accident Analysis & Prevention,
117: 318–327.
Ihueze, C & Onwurah, U O 2018. Road traffic accidents
prediction modelling: An analysis of Anambra State,
Nigeria. Accident Analysis & Prevention, 112(12):
21–29. DOI: 10.1016/j.aap.2017.12.016.
Karlaftis, M G & Golias, I 2009. Impacts of roadway
condition, traffic and manmade features on
road safety. International Journal of Civil
Engineering & Technology, 4(7): 19–25.
DOI: 10.1061/(ASCE)HY.1943-7900.0000746.
Knuiman, M W, Council, F M & Reinfurt, D W 1993.
Association of median width and highway accident
rates. Transportation Research Record, 1401: 70–82.
Laird, J, Harris, R & Shen, S 2010. Accidents on rural
roads – For better or worse. University of Leeds, UK.
Li, G, Lai, W, Sui, X, Li, X, Qu, X, Zhang, T & Li, Y
2020. Influence of traffic congestion on driver
behaviour in post-congestion driving. Accident
Analysis & Prevention, 141: 105508.
Liu, S, Wang, J & Fu, T 2016. Effects of lane width,
lane position and edge shoulder width on driving
behaviour in underground urban expressways:
A driving simulator study. International Journal
of Environmental Research and Public Health,
13(10): 1–14. DOI: 10.3390/ijerph13101010.
Ma, Y, Yang, X & Zeng, Y 2009. Association analysis of
urban road free-flow speed and lane width. Tongji
University Journal, 37: 1621–1626.
Martens, M, Comte, S & Kaptein, N 1997. The effects of
road design on speed behaviour: A literature review.
Crowthorne, UK: Transport Research Laboratory.
https://trid.trb.org/view/634567.

Journal of the South African Institution of Civil Engineering Volume 64 Number 4 December 2022 47
Murthy, N & Rao, R S 2015. Development of a model
for road accidents based on intersection parameters
using regression models. International Journal of
Scientific & Research Publications, 5(1): 1–8.
Nghishihange, S 2018. Factors contributing to
motor vehicle accidents and severity of related
injuries in Oshana region, Namibia. Master’s
Dissertation. Windhoek: University of Namibia.
DOI: 10.22201/fq.18708404e.2004.3.66178.
Noland, R B & Oh, L 2004. The effect of infrastructure
and demographic change on traffic-related fatalities
and crashes: A case study of Illinois county-level
data. Accident Analysis & Prevention, 36(4): 525–
532. DOI: 10.1016/S0001-4575(03)00058-7.
NRSC (Namibia National Road Safety
Council) 2012. Road accidents in Namibia –
Statistical report 2012. Windhoek, Namibia.
DOI: 10.1017/CBO9781107415324.004.
Ogden, K W 1997. The effects of paved shoulders on
accidents on rural highways. Accident Analysis &
Prevention, 29(3): 353–362.
Oppong, R A 2012. Statistical analysis of road
accidents fatality in Ghana using Poisson regression.
Master’s Dissertation. Kumasi, Ghana: Kwame
Nkrumah University of Science and Technology.
Reifman, A & Keyton, K 2010. Winsorise.
Encyclopaedia of Research Design, 1636–1638.
Stamatiadis, N, Pigman, J, Sacksteder, J, Ruff,
W & Lord, D 2009. Impact of shoulder width
48
and median width on safety. Washington,
DC: Transportation Research Board, National
Cooperative Highway Research Program (NCHRP),
Report No 633.
Starkey, P, Riehle, E B & Ambunda, R 2017.
Promoting rural access and mobility in northern
Namibia: An integrated approach. Proceedings,
8th Africa Transportation Technology Transfer
Conference, 8–10 May 2017, Livingstone, Zambia.
Available at: https://www.researchgate.net/
publication/318680016.
Summala, H 1996. Accident risk and driver
behaviour. Safety Science, 22: 103–117.
DOI: 10.1016/0925-7535(96)00009-4.
Taylor, M C, Lynam, D A & Baruya, A 2000. The
effects of drivers’ speed on the frequency of road
accidents. TRL Report 421. Crowthorne, UK:
Transport Research Laboratory. https://trid.trb.
org/view/651648.
Taylor, M C, Baruya, A & Kennedy, J V 2002. The
relationship between speed and accidents on
rural single-carriageway roads. TRL Report 511.
Crowthorne, UK: Transport Research Laboratory.
http://www.safespeed.org.uk/TRL511.pdf.
Turner, B, Woolley, J & Cairney, P 2015. An analysis of
driver behaviour through rural curves: Exploratory
results on driver speed. Proceedings, Australasian
Road Safety Conference, 14–16 October 2015,
Queensland, Australia, Vol 10, pp 1–10.
Volume 64 Number 4 December 2022
Vaa, T 2014. From Gibson and Crooks to Damasio:
The role of psychology in the development of driver
behaviour models. Transportation Research Part F:
Traffic Psychology and Behaviour, 25(Part B): 112–
119. DOI 10.1016/j.trf.2014.02.004.
Walmsley, D A, Summersgill, I & Binch, C 1998.
Accidents on modern rural single-carriageway trunk
roads. TRL Report 336. Crowthorne, UK: Transport
Research Laboratory.
Wang, C, Quddus, M A & Ison, S G 2013. The effect
of traffic and road characteristics on road safety: A
review and future research direction. Safety Science,
57: 264–275. DOI: 10.1016/j.ssci.2013.02.012.
Wang, B, Hallmark, S, Savolainen, P & Dong, J 2018.
Examining vehicle operating speeds on rural
two-lane curves using naturalistic driving data.
Accident Analysis & Prevention, 118(4): 236–243.
DOI:10.1016/j.aap.2018.03.017.
Warrd, D, Steyvers, F J & Brookhuis, K A 2004. How
much visual road information is needed to drive
safely and comfortably. Safety Science, 42: 639–655.
DOI: 10.1016/j.ssci.2003.09.002.
WHO (World Health Organization) 2018. Global
status report on road safety 2018. Geneva,
Switzerland.
Zegeer, V C, Reinfurt, D W, Hummer, J, Herf, L &
Hunter, W 1987. Safety effects of cross-section design
for two-lane roads. Chapel Hill: University of North
Carolina, Highway Safety Research Centre.
Journal of the South African Institution of Civil Engineering
 Thanks to our Referees!
The SAICE Journal Editorial Panel would like to thank the persons listed below, all of whom served as referees from June 2021 up to the end
of October 2022. The quality of our journal is not only a reflection of the level of expertise of participating authors, but certainly also of the
high standard set by our referees. If we have overlooked anyone by mistake, please accept our apologies, and know that your assistance is
equally appreciated!

Dr André Archer Dr Stephan Jooste Prof Chris Roth

Dr Ryan Bradley Prof Johan Joubert Dr Martin Rust
Mr Johan Brink Prof Elsabé Kearsley Mr Ken Schwartz
Ms Megan Brink Prof Nthatisi Khatleli Dr Sarah Skorpen
Mr Keith Brown Mr Hans King Prof Jeff Smithers
Mr Gys Burger Dr Hendrik Kirsten Mr Hendrik Steenkamp
Mr Arthur Chapman Dr Abisai Konstantinus Prof Miroslav Sýkora
Dr Renier Cloete Dr Maxim Kovtun Mr Frans Van der Merwe
Dr Louis Coetzee Dr Christoff Krogscheepers Mr Danie Van der Spuy
Mr Sean Cooke Dr Maggi Loubser Dr Pierre Van der Spuy
Dr Talia Da Silva Burke Mr Andrew MacKellar Dr André Van Tonder
Mr Gary Davis Dr Charles MacRobert Prof Gideon Van Zijl
Prof Nick Dekker Mr Isak Malherbe Prof Marianne Vanderschuren
Mr Johan De Koker Mr Stephen Mallory Mr Isak Venter
Dr Wibke De Villiers Mr Mike McDonald Dr Karien Venter
Dr Louw Du Plessis Dr Terence Milne Mr Robert Verger
Dr Grizelda Du Toit Mr Derek Mostert Dr Nico Vermeulen
Mr Johann Enslin Prof Pilate Moyo Dr Eduard Vorster
Prof Phil Everitt Mr Johannes Neethling Dr Fritz Wagener
Dr André Görgens Prof Sunday Nwaubani Mr Gavin Wardle
Mr Siegfried Grosskopf Mr Johan O'Connell Dr Andrew Watson
Mr Gary Hayes Mr Pieter Onderwater Dr Pieter Wessels
Dr Francois Heyns Prof Mike Otieno Dr Gabi Wojtowitz
Dr Ramon Hingorani Mr Greg Palmer Dr Piotr Wolski
Mr Gerhard Hitge Mr Michael Palmer Mr Shuaib Yunos
Prof Florian Hoertkorn Mr Trevor Pape Mr Eduardo Zannoni
Prof Adeshola Ilemobade Prof Geoff Pegram
Prof Heinz Jacobs Prof Ferdinand Pospischil

Journal of the South African Institution of Civil Engineering Volume 64 Number 4 December 2022 49
 Guidelines for the preparation of
papers and technical notes
INTRODUCTION
The Journal of the South African Institution of Civil Engineering,
which is a peer-reviewed, open-access journal, is published quar-
terly in March, June, September and December. Articles submitted
for publication are reviewed by a panel of referees under the guid-
ance of the SAICE Journal Editorial Panel. The journal publishes
research papers covering all the disciplines of civil engineering
(structural, geotechnical, railway, coastal/marine, water, construc-
tion, environmental, municipal, transportation) and associated
topics that are relevant to the civil engineering profession, and
that preferably have particular relevance to civil engineering in
Southern Africa and the African continent. When preparing arti-
cles for publication, authors should please take note of, and comply
with, the guidelines as set out below.
CLASSIFICATION OF ARTICLES CONSIDERED
FOR PUBLICATION
QQ Technical papers are well-researched, in-depth, fully referenced
technical articles not exceeding 6 000 words in length (exclud-
ing tables, illustrations and the list of references). Related papers
that deal with ‘softer sciences’ (e.g. education, social upliftment,
etc) are accepted if they are of a technical nature and of particu-
lar interest to the civil engineering profession. The latter type of
paper will be subject not only to peer-review by civil engineers,
but also to review by non-engineering specialists in the field
covered by the paper.
QQ Technical notes are short, fully referenced technical articles
that do not exceed 2 000 words. A typical technical note will
have limited scope often dealing with a single technical issue of
particular importance to civil engineering.
QQ Review papers are considered for publication as either technical
papers or technical notes on condition that they are the original
work of the author and will assist the reader with the under-
standing, interpreting, or applying of the subject under review.
A review paper must contain criteria by which the work under
review was evaluated and must contribute by synthesising the
information and drawing new conclusions from the dissemina-
tion of the previously published work.
QQ Discussions on published articles are welcomed up to six
months after publication. The length of discussion contributions
is limited to 1 500 words. Where appropriate, discussion contri-
butions will be subject to the normal reviewing process and will
be forwarded to the authors of the original article for reply.
POLICY REGARDING LANGUAGE AND ORIGINALITY
OF SUBMITTED ARTICLES
QQ Language: Manuscripts should be presented in English, as the
journal is distributed internationally.
QQ Original work: Papers and technical notes must be original
contributions. Authors must affirm that submitted material has
not been published previously, is not under consideration for
publication elsewhere and will not be submitted elsewhere while
50
under consideration by the SAICE Journal Editorial Panel. It is
the responsibility of the authors to ensure that publication of any
paper in the journal will not constitute a breach of any agree-
ment or the transgression of any law. The corresponding author
should confirm that all co-authors have read and approved the
manuscript and accept these conditions. Authors are responsible
for obtaining permission to publish experimental data and other
information that may be confidential or sensitive. Authors are
also responsible for obtaining permission from copyright owners
(usually the publishers) when reproducing material that had been
published elsewhere. Proof of such permission must be supplied.
SUBMISSION PROCEDURES AND REQUIRED FORMAT
QQ Online submission: Manuscripts must be uploaded as PDF
files (http://journal.saice.org.za). Individual file sizes may not
exceed 10 MB. Should you experience problems when uploading
your paper, please contact the administrative editor (verelene@
saice.org.za).
QQ Format: Manuscripts should be prepared in MS Word and
presented in double-line spacing, single-column layout with
25 mm wide margins. Line numbers must be applied to the
whole document. All pages should bear the authors’ names and
be numbered at the bottom of the page. With the exception of
tables and figures (see below) the document should be typed in
Times New Roman 12 pt font. Contributions should be accom-
panied by an abstract of not more than 200 words.
QQ First page: The first page of the manuscript should include
the title of the paper, the number of words of the main text
(i.e. excluding figures, tables and the list of references), the
initials, titles and surnames of the authors, professional status
(if applicable), SAICE affiliation (Member, Fellow, Visitor, etc),
telephone numbers (landline and mobile), and e-mail and postal
addresses. The name of the corresponding author should be
underlined. Five keywords should be suggested.
QQ Figures, tables, photos and illustrations:
QQ These should preferably be submitted in colour, as the jour-
nal is a full-colour publication.
QQ Their positions should be clearly marked in the text as fol-
lows: [Insert Figure 1].
QQ Figures, tables, photos, illustrations and equations should
be numbered consecutively and should appear in the text
directly after they have been referred to for the first time.
QQ Illustrations must be accompanied by appropriate captions.
Captions for tables should appear above the table. All other
captions should appear below the illustration (figures,
graphs, photos).
QQ Only those figures and photographs essential to the under-
standing of the text should be included. All illustrations
should be referred to in the text.
QQ Figures should be produced using computer graphics. Hand-
drafted figures will not be accepted. Lettering on figures
should be equivalent to a Times New Roman 9 pt font, or
Volume 64 Number 4 December 2022 Journal of the South African Institution of Civil Engineering
 slightly larger (up to 12 pt) if desired. Lettering smaller than
9 pt is not acceptable.
QQ Tables should be typed in Times New Roman 9 pt font. They
should not duplicate information already given in the text,
nor contain material that would be better presented graphi-
cally. Tabular matter should be as simple as possible, with
brief column headings and a minimum number of columns.
QQ Mathematical expressions and presentation of symbols:
QQ Equations should be presented in a clear form which can
easily be read by non-mathematicians. Each equation
should appear on a separate line and should be numbered
consecutively.
QQ Symbols should preferably reflect those used in Microsoft
Word Equation Editor or Mathtype, or should be typed using
the Times New Roman symbol set.
QQ Variables in equations (x, y, z, etc, as well as lower-case Greek
letters) should be presented in italics. Numbers (digits),
upper-case Greek letters, symbols of metric measurement
units (m for metres, s for seconds, etc) and mathematical/
trigonometrical functions (such as sin, cos and tan) are not
written in italics, but in upright type (Roman). Variables
and symbols used in the body of the text should match the
format used in the equations, i.e. upright or italics, whichever
is applicable.
QQ Metric measurement abbreviations/units should conform to
international usage – the SI system of units should be used.
QQ Decimal commas may be used, but decimal points are pre-
ferred in accordance with international usage.
QQ Symbols should preferably be defined in the text, but if this is
not feasible, a list of notations may be provided for inclusion
at the end of the paper.
QQ Headings: Sections and paragraphs should not be numbered.
The following hierarchy of headings should be followed:
HEADING OF MAIN SECTION
Heading of subsection
Heading of sub-subsection
QQ References: References should follow the Harvard system. The
format of text citations should be as follows: “Jones (1999) dis-
covered that …” or “Recent results (Brown & Carter 1985; Green
et al 1999) indicated that …”
 References cited in the text should be listed in alphabetical
order at the end of the paper. References by the same author
should be in chronological order. The following are examples of
a journal article, a book and a conference paper:
QQ Donald, A W 1954. Pore pressure coefficients.
Géotechnique, 4(4): 143–147.
QQ Terzaghi, K 1943. Practical soil mechanics. New York: Wiley.
QQ Smith, A W, Black, C & White, A 1999. Factors affecting the
strength of soils. Proceedings, ASCE Conference on Shear
Strength of Soils, Colorado, pp 503–532.
Papers published previously in the Journal of the South African
Institution of Civil Engineering should be cited if applicable.
Journal of the South African Institution of Civil Engineering Volume 64 Number 4 December 2022
QQ Footnotes, trade names, acronyms, abbreviations: These
should be avoided. If acronyms are used, they should be defined
when they first appear in the text. Do not use full stops after
abbreviations or acronyms.
QQ Return of amended papers: Papers requiring amendments will
be accepted up to six months after the referee reports had been
sent to authors, after which time the paper will be withdrawn
from the system.
FINAL ARTICLE
QQ Proofs: First proofs of accepted papers will be sent to authors
in PDF format for verification before publication. No major
re-writes will be allowed at this stage, only essential minor
corrections.
QQ Photos of authors: The final corrected version of the paper
should be accompanied by recent, high-resolution head and
shoulders colour photographs and a profile not exceeding 100
words for each of the authors.
QQ Copyright: As a peer-reviewed, open-access journal
licenced under a Creative Commons Attribution Licence
(CC BY‑NC‑ND), the content of this journal is available to users
free of charge. However, copyright remains with the South
African Institution of Civil Engineering. On acceptance of a
paper or technical note, copyright must therefore be transferred
by the author/s to the South African Institution of Civil
Engineering (as the publisher) on the form that will be provided
by the Institution. Authors retain the following rights:
QQ All propriety rights other than copyright, such as patent
rights.
QQ The right to grant or refuse permission to third parties to
re-publish all or part of the paper or translations thereof.
Such parties must obtain SAICE’s written permission as
well. However, such permission will not be refused by SAICE
except at the discretion of the author. SAICE may grant
rights with respect to entire volumes.
QQ The right to use all or part of this paper in future works of
the author(s), such as reports, lectures, lecture notes, press
releases, reviews, textbooks or reprint books.
QQ In the case of “work made for hire”, the author allows the
employer the right to make copies of this paper for his own
use, but not for re-sale.
Permission for reproduction of content, with appropriate refer-
ence to the journal, can be obtained from the administrative
editor.
QQ Repository policy (depositing of articles by authors after
publication): After publication, a PDF of their published article
is sent to authors for their records. Authors may then upload
the published article to their work/institutional website, or any
other repository of their choice, such as ResearchGate and/or
similar. Only the published version of the paper may be used/
uploaded in this manner, i.e. not the originally submitted or the
accepted/pre-published version.
51