SEISMIC EFFECTIVE STRESS ANALYSIS OF GRAVITY BLOCK

TYPE QUAY WALLS: APPLICATION TO PIRAEUS PORT

Panagiota TASIOPOULOU1, Nikos GEROLYMOS2 and George GAZETAS3

ABSTRACT

Experience has shown that port facilities, and especially gravity quay walls, are particularly
vulnerable to earthquake related hazards. Motivated by numerous observed cases where gravity quay
walls suffered large displacement and rotation during earthquakes, even in soils not prone to
liquefaction, an illustrative numerical analysis is presented for the response of a typical block-type
quay wall section at Piraeus port in Greece. Utilizing the Byrne's elastoplastic constitutive model, an
effective stress dynamic analysis is performed using as seismic excitation two recorded strong motions
of the seismic environment of Greece. The results emphasize the role of excess pore-water pressure
(negative or positive) build-up during shaking on the evolution of the lateral displacement and tilt of
the quay wall, shedding light on the potential pitfalls that could emerge from a total stress analysis in
which pore-water pressure generation is not directly considered. It is shown that when extensive soil
liquefaction does not take place, the negative pore water pressures develop in the backfill mitigate the
large displacement and rotation of the quay wall.

INTRODUCTION

Gravity quay wall structures have repeatedly suffered substantial outward displacement and rotation
even when subjected to moderate earthquake shaking. (e.g. Pitilakis and Moutsakis, 1989; Egan et al.,
1992; Iai et al., 1994; Suganoand Iai, 1999; Elnashai et al., 2010; Zarzouras et al., 2010). Apparently,
due to their nature, these structures are extremely vulnerable to liquefaction and lateral spreading.
Such phenomena may lead to dramatic horizontal displacements and rotations, resulting not only to
the failure of the structural component itself, but also to damage to a number of inter-connected
elements: extreme deformation or failure of piping systems and utilities. The strong rocking of quay
walls (due only to its inertial forces), when founded on a compliant and weak foundation soil in
combination with the one-sided action of the earth pressures leads to the accumulation of horizontal
displacement and rotation towards the seaside. This effect is rather amplified due to the current design
practice of quay walls: their seismic design against earth pressures unavoidably leads to large
dimensions of the walls (Okabe, 1926). The increase of their dimensions is a self-defeating and
expensive proposition, as it augments the mass of the quay wall, ultimately amplifying the inertial
forces acting on the foundation soil (Zarzouras et al., 2010): a vicious circle that may not serve either
the safety or the economy of the project.
Tilt performance criteria related to container crane operations hamper the use of gravity quay
walls for container wharf operations imposed to seismic loading. Conventional practice for evaluating

1
Phd Student, National Technical University, Athens, Greece, ptasiopoulou@gmail.com
2
Assistant Professor, National Technical University, Athens, Greece, gerolymos@gmail.com
3
Professor, National Technical University, Athens, Greece, gazetas@ath.forthnet.gr

1
their seismic stability is based on (i) pseudo-static (force-based) approaches and (ii) oversimplified
sliding block (displacement-based) methods of analysis, similar to those applied for embankments!
The deformation modes that synthesize the response of the quay wall at large displacements and near
failure conditions: sliding, overturning, and bearing capacity mobilization are evaluated separately,
totally neglecting the unavoidable interplay with one another. The state-of-practice assigns higher
factors of safety for overturning and bearing capacity than for sliding, as these are more critical modes
of failure. A key component in the above analyses is the estimation of the seismic active earth
pressure. While the Mononabe Okabe limit equilibrium method is widely used in practice to predict
active and passive earthquake pressures (Ebeling and Morrison, 1992; PIANC, 2001), the method have
limitations, and a generalized limit equilibrium approach has been recommended by the NCHRP 12-
70 Project Report 611 (Anderson et al., 2008). In both methods, however, the influence of soil
liquefaction on the failure mechanisms of the quay wall is completely ignored.
The vital role that quay walls play in the operational capacity of ports, shipyards and other
waterside facilities, increases the pressure to provide more efficient and seismically resilient new
infrastructure or improve the existing facilities through rehabilitation and seismic upgrading.
Optimizing the seismic performance of gravity quay walls requires a deep understanding of the
mechanics that govern their response, and, effective stress analysis is an essential tool that could
provide a valuable insight into this. Evidently, the whole problem is very complex. The dynamic
response of gravity quay walls is strongly affected by non-linear soil behaviour. Development of
excess pore pressures and accumulation of shear and volumetric strains both at the retained soil and
the foundation soil, produces the degradation of the shear strength of the soil which may lead to
liquefaction. The above phenomena are further complicated when accounting for soil-structure
interaction.
The goal of this paper is to investigate the seismic response of block-type gravity quay walls
emphasizing the role of pore-water pressure build-up in the soil behind and in front of the wall. Two
sub-cases are examined: In the first one (hypothetic case) only the hydrostatic conditions are
considered and the possibility of pore-water pressure build up is completely ignored. This case serves
as a reference for evaluating the influence of water flow on the system's response (second--realistic
case). The comparison is attempted at two performance levels, representing: (a) the contingency-level
earthquake (475 years return period) and (b) the ultimate-level earthquake (975 years return period)
with application to a typical gravity quay wall section at Piraeus port in Greece. To this end, the paper
utilizes the rigorous plain strain finite difference formulation of FLAC2D (Itasca, 2000), along with
Byrne's elasto-plastic constitutive model for cyclic stress-strain soil behaviour.

NUMERICAL MODELING

In the framework of the current research program which aims to the upgrade and retrofit of existing
piers in ports within Greece, pier II of Piraeus port, built in 1994-1996 was studied. A typical cross
section of pier II comprising the geometry of the block-type gravity quay wall and the idealized soil
profile is shown in Figure 1. The examined soil profile does not indicate significant liquefaction
potential; perhaps apart from the silty sand layer of medium density situated 3 m below the base of the
quay wall.
The current 2D section was simulated and analysed numerically using the finite difference
code FLAC 2D (Itasca, 2005). The distances of the boundaries from the quay wall are also shown in
Figure 1. Two types of models were dynamically analysed: i) model A where the development of
negative or positive excess pore pressures, Δu, was allowed and properly simulated, and ii) model B
where hydrostatic pressures were applied initially to establish a realistic geostatic field while further
development of Δu during the seismic stage was ignored. Both models incorporated Mohr Coulomb
plasticity model along with appropriate hysteretic damping. Especially, simulation of model A
involves the constitutive law of Byrne (1991) for pore pressure generation which is incorporated in the
standard Mohr-Coulomb plasticity model. The waterfront was simulated through constant hydrostatic
pressure on the quay wall; thus hydrodynamic effects due to sea-water waves were neglected. “Free-
field” conditions were used for the outer boundaries in order to absord wave reflections.
quay wall.

2
 P.Tasiopoulou, N.Gerolymos and G.Gazetas 3

The contact conditions between the blocks of the quay wall as well as between the quay wall
and the adjacent soil were modelled with interfaces allowing for slippage and detachement via a
Coulomb frictional law. Friction coefficients were assumed equal to 0.5 and 0.7 respectively.
The seismic input motions were chosen among the available records from earthquakes in
Greece. The goal was to examine two levels of intensity: a medium and a strong one according to
standards of Greece. To this purpose, the chosen records include Kalamata (1986) with PGA 0.26g
and Lefkada (2003) with PGA 0.42g, depicted in Figure 2, along with their acceleration spectra. The
seismic input motions were applied at the base of the models.
Vs (m/s)
9.5 m 70 m 0 250 500
0m

2.5 m

Artificial Fill 0 250 500

Sandy Gravel
17.5 m
φ = 35°
Dr = 80%

25.5 m
17.5 m

20.5 m

Silty sand, φ = 30°, Dr = 60%

25.5 m

Gravelly sand, φ = 35°, Dr = 80%

32 m
Soft marl, Su = 120 kPa
37.5 m

Figure 1. Idealized soil profile and geometry of pier 2 of II of Piraeus port.

0.5 2
0.25 1.8
α:g 0 1.6
-0.25 1.4
1.2 Lefkada
Leukada
-0.5
1
0.5 0 5 10 15 20 SA : g Kalamata
Kalamata
t (sec) 0.8
0.25
0.6
α:g 0 0.4
-0.25 0.2
-0.5 0
0 5 10 15 20 0 1 2 3 4
t (sec) T (sec)

Figure 2. Input motions and acceleration spectra.

RESULTS & CONCLUSIONS

The results are presented in Figures 3-13 through the following graphs: (i) time histories of the quay
wall horizontal displacement and rotation, (ii) contours of residual horizontal displacement, shear
strain and excess pore water pressure ratio, and (iii) displacement vectors and deformed finite
difference mesh.

Numerical results of both models are primarily shown for the higher intensity input seismic motion of
Lefkada (2003), in Figures 3 to 7. Examining initially the deformed grid after the end of shaking,
illustrated in Figure 3, it is evident that the quay wall sustained greater outward displacement and
rotation in case of model (b), where any development of Δu due to seismic loading was ignored. This
is due to negative excess water pressure ratio closely behind the quay wall (Dakoulas and Gazetas,
 2007), as indicated by the contours of excess pore pressure ratio, r u in Figure 4. The excess pore
pressure ratio, ru, is defined as the excess pore pressure Δu, over the initial vertical effective stress,
σ'vo. Figure 4 also indicates significant generation of positive excess pore pressure within the silty sand
layer of Dr =-60%. Nevertheless, it is evident that no liquefaction occurred in the backfill close to the
quay wall.
It should be noted that, for all cases considered, no slippage or detachment occurred between
the blocks of the quay wall, leading to translation of the quay wall as a rigid body. Interestingly, model
B, without Δu, sustained greater outward displacement and rotation compared to model A, with Δu,
for all earthquake motions considered. This is attributed to the extensional seaward deformation of the
backfill soil adjacent to the quay wall resulting in a geometrical imposed dilation (negative excess
pore water pressure), which overshadows the tendency of soil for volumetric contraction (positive
excess pore pressure) due to cyclic loading.
Another remarkable observation, for the strong motion record (Lefkada), is that the permanent
seaward displacement of the backfill extends all the way to the right boundary of model A, with Δu,
approximately 70 m from the quay wall. However, in case of model B, without Δu, the residual
displacements vanish rapidly after the midwidth of model, at a distance approximately 40 m from the
quay wall. On the other hand, for less strong motions (Kalamata), the distribution of backfill
displacements seems to extend to the same distance from the quay wall (Figure 9) despite the larger
outward quay wall displacement of model B, without Δu. These discrepancies in the displacement
pattern render the problem case specific and they could lead to erroneous design assumptions and
displacement-based performance requirements for deformation-sensitive inter-connected elements,
such as piping systems and container cranes, when the effect of excess pore water pressure is not
explicitly taken into account in the analysis.

(*10^1)

0)

6.000

+01
(a)
E+02 4.000
E+01

2.000

ion
+00
0

(b)
0.000

Figure 3. Deformed grid (blue) on top of undeformed grid (yellow) after the end of Lefkada 2003 seismic
motion: (a) model A with Δu and (b) model B with no Δu.

-2.000

4
 Flow Time 7.5925E+17 6.000

Dynamic Time 2.0000E+01

P.Tasiopoulou, N.Gerolymos and G.Gazetas 5 4.000
-5.639E+00 <x< 1.071E+02
-3.764E+01 <y< 7.514E+01

EX_ 1 Contours
-1.2
-1.20E+00 4.000
-0.8 2.000
-8.00E-01
ru -4.00E-01
-0.4
00.00E+00
JOB TITLE : .
4.00E-01
0.4
FLAC (Version 6.00)
0.8
8.00E-01 (*10^1)
2.000
0.000
0) Contour interval= 2.00E-01
LEGEND
Extrap. by averaging
11-Apr-14 12:51
Grid plot
step 429697 6.000
Flow Time 7.5925E+17
0 Dynamic Time
2E 1 2.0000E+01
-5.639E+00 <x< 1.071E+02 0.000 -2.000
-3.764E+01
Figure 4. Contours of excess<y<
pore7.514E+01
pressure ratio after the end of Lefkada 2003 seismic motion in case of the
model A where Δu is allowed to develop.
Max. shear strain increment
7 00.00E+00
+01 0.1
1.00E-01 4.000
+02 2.00E-01
0.2
(*10^1)
+01 Shear 3.00E-01
0.3 0.100 0.300 0.500 0.700 0.900 -2.000
stain 0.4
4.00E-01 (*10^2)
0) 0.5
5.00E-01
ent 0.6
6.00E-01
0.7
7.00E-01
0.8
8.00E-01
2.000
6.000
Contour interval= 5.00E-02
Extrap. by averaging
Grid plot
0.100 0.300 0.500 0.700 0.900
(*10^2)
0 2E 1
+01
+02 (a) 0.000
4.000
E+01 0.100 0.300 0.500 0.700 0.900
(*10^2)

02
ent

-2.000
2.000

(b)
0.100 0.300 0.500 0.700 0.900 0.000
(*10^2)
02 Figure 5. Shear strain contours after the end of Lefkada 2003 seismic motion: (a) model A with Δu and (b)
model B with no Δu.

-2.000

0.100 0.300 0.500 0.700 0.900

(*10^2)
E+17
0E+01 4.000
71E+02 (*10^1)
14E+01 Max. displacement
Vector
.00) 3.4 m
tortion
0E+00
2.000
6.000
D
+00

E+00

0E+01
1E 1
71E+02 (a) 0.000
4.000
14E+01
Max. displacement
ortion Vector
4.6 m
0E+00 JOB TITLE : .
+00 -2.000
FLAC (Version 6.00) 2.000
(*10^1)

E+00
on 6.00) LEGEND

11-Apr-14 12:51
1 step 429697 6.000
GEND Flow Time 7.5925E+17
(b)
0.100 Dynamic Time 0.300
2.0000E+01 0.500 0.700 0.900 0.000
-5.639E+00 <x< 1.071E+02 (*10^2)
51 Figure 6. Deformed -3.764E+01
grid (blue)<y<
and7.514E+01
displacement vectors after the end of Lefkada 2003 seismic motion: (a)
model A with Δu and (b) model B with no Δu.
Max. shear strain increment
5925E+17 00.00E+00
2.0000E+01 0.1
1.00E-01 4.000

. 1.071E+02
2.00E-01
0.2
(*10^1)
7.514E+01 Shear 3.00E-01
0.3 -2.000
stain 0.4
4.00E-01
on 6.00) 0.5
5.00E-01
n increment 0.6
6.00E-01
0.7
7.00E-01
0.8
8.00E-01
2.000
6.000
GEND Contour interval= 5.00E-02
Extrap. by averaging
Grid plot
50 0.100 0.300 0.500 0.700 0.900
(*10^2)
0 2E 1
2.0000E+01
1.071E+02 (a) 0.000
4.000
7.514E+01 0.100 0.300 0.500 0.700 0.900
(*10^2)

=n increment
5.00E-02
ging

-2.000
2.000

2E 1

(b)
0.100 0.300 0.500 0.700 0.900 0.000
(*10^2)
= 5.00E-02 Figure 6. Shear strain contours after the end of Lefkada 2003 seismic motion: (a) model A with Δu and (b)
ging model B with no Δu.

-2.000
2E 1
6
 6.000
11-Apr-14 12:24
step 429697P.Tasiopoulou, N.Gerolymos and G.Gazetas 7
Flow Time 7.5925E+17
Dynamic Time 2.0000E+01
-5.639E+00 <x< 1.071E+02
-3.764E+01 <y< 7.514E+01
+17
E+01 X-displacement contours
4.000
-3
-3.00E+00
1E+02 -2.50E+00 (*10^1)
Horizontal -2
-2.00E+00
4E+01 Displacement -1.50E+00
(m)
00) -1
-1.00E+00
rs -5.00E-01
00.00E+00

Contour interval= 5.00E-01 2.000

6.000
Grid plot

0 2E 1

E+01
(a)
1E+02 0.000
4.000
4E+01 0.100 0.300 0.500 0.700 0.900
E-01 (*10^2)

rs

-2.000
2.000

(b)
0.100 0.300 0.500 0.700 0.900 0.000
E-01 (*10^2)
Figure 7. Horizontal displacement contours after the end of Lefkada 2003 seismic motion: (a) model A with Δu
and (b) model B with no Δu.

-2.000

0.100 0.300 0.500 0.700 0.900

(*10^2)

(a)

(b)

Figure 8. Deformed grid (blue) on top of undeformed grid (yellow) after the end of Kalamata seismic motion: (a)
model with Δu and (b) model with no Δu.
17
E+01 4.000
E+02 (*10^1)
E+01 Max. displacement
Vector
00) 1m
rtion
E+00
2.000
6.000
1

E+00
01
E+01
(a)
E+02
0 0.000
4.000
JOB TITLE : . (*10^1)
E+01
FLAC (Version 6.00)
Max. displacement
rtion Vector
E+00 2m
(*10^1) 6.000
0 LEGEND
-2.000
2.000

) 11-Apr-14 11:42
00
step 429697
Flow Time 7.5925E+17
0 6.000
Dynamic Time 2.0000E+01 4.000
(b) -5.639E+00 <x< 1.071E+02
0.100 0.300 0.500 0.700 0.900 0.000
-3.764E+01 <y< 7.514E+01 (*10^2)
Figure 9. Deformed grid (blue) and displacement vectors after the end of Kalamata seismic motion: (a) model A
with Δu and (b) model B with no Δu.
EX_ 1 Contours
-1.2
-1.20E+00
01 4.000
2.000
02 -8.00E-01
-0.8

+01 ru -4.00E-01
-0.4 -2.000
00.00E+00

4.00E-01
0.4
0.8
8.00E-01
2.000 0.000
Contour interval= 2.00E-01
Extrap. by averaging 0.300
0.100 0.500 0.700 0.900
Grid plot (*10^2)

0 2E 1
-2.000
0.000
1
Figure 10. Contours of excess pore pressure ratio after the end of Kalamata seismic motion in case of the model
A where Δu is allowed to develop.

0.100 0.300 0.500 0.700 0.900

-2.000
(*10^2)

0.100 0.300 0.500 0.700 0.900

(*10^2)

8
 11-Apr-14 15:50
step 429697
Flow Time P.Tasiopoulou,
7.5925E+17 N.Gerolymos and G.Gazetas 9
Dynamic Time 2.0000E+01
-5.639E+00 <x< 1.071E+02
17 -3.764E+01 <y< 7.514E+01
E+01 4.000
Max. shear strain increment
E+02 00.00E+00 (*10^1)
E+01 JOB
0.04 TITLE : .
4.00E-02
Shear 0.08
8.00E-02
00) stain FLAC
0.12 (Version 6.00)
1.20E-01
ent 0.16
1.60E-01
0.2
2.00E-01
LEGEND
Contour interval= 2.00E-02 6.000
2.000
Extrap. by averaging
Grid plot
11-Apr-14 15:46
step 301622
0
Dynamic Time 2E 1
2.0000E+01
-5.639E+00 <x< 1.071E+02
E+01 -3.764E+01 <y< 7.514E+01
-02
E+02 (a) 4.000
0.000
Max. shear strain increment
E+01 00.00E+00
0.1
1.00E-01
ent Shear 0.2
2.00E-01 0.100 0.300 0.500 0.700 0.900
(*10^2)
stain 0.3
3.00E-01
0.4
4.00E-01
0.5
5.00E-01
2.000
-2.000
Contour interval= 5.00E-02
Extrap. by averaging
Grid plot

0 2E 1

-02 (b)
0.000
0.100 0.300 0.500 0.700 0.900
(*10^2)
Figure 11. Shear strain contours after the end of Kalamata seismic motion: (a) model A with Δu and (b) model B
with no Δu.
t (sec) t (sec)
0.100 0.300 0.500 0.700 0.900
0 5 10 15 20 0 2 4 6 (*10^2)
8 10
0.5 0.2
0 0 -2.000
-0.5 -0.2
Quay wall -1 Quay wall -0.4
Horizontal -1.5 Horizontal -0.6
displacement -2 displacement -0.8
(m) -2.5 (m) -1
-3 -1.2
-3.5 -1.4
Lefkada Kalamata
-40.100 0.300 0.500 -1.6 0.700 0.900
base of wall (model (a) with(*10^2)
Δu)
base of wall (model (b) without Δu)
top of wall (model (a) with Δu)
top of wall (model (b) without Δu)

Figure 12. Time histories of quay wall horizontal displacements.

 7
6
5
Quay wall
rotation 4 Lefkada (model (a) with Δu)
(°) 3 Lefkada (model (b) without Δu)
2 Kalamata (model (a) with Δυ)
1 Kalamata (model (b) without Δu)

0
-1
0 5 10 15 20
t (sec)

Figure 13. Time histories of quay wall rotation.

ACKNOWLEGMENT

This research has been co-financed by the European Union (European Social Fund – ESF) and Greek
national funds through the Operational Program "Education and Lifelong Learning" of the National
Strategic Reference Framework (NSRF) - Research Funding Program: Thales. Investing in knowledge
society through the European Social Fund, Project ID "UPGRADE".

REFERENCES

Anderson, D. G., Martin, G.A., Lam, I. (Po), and Wang, T. N., (2008), "SeismicAnalysis and Design of
Retaining Walls, Buried Structures, Slopes, and Embankments", Report 611, National
Cooperative Highway Research Program (NCHRP) Transportation Research Board,
Washington, DC.
Byrne, P. (1991). “A Cyclic Shear-Volume Coupling and Pore-Pressure Model for Sand,”
Proceedings of Second International Conference on Recent Advances in Geotechnical
Earthquake Engineering and Soil Dynamics, St. Louis, Missouri, Paper No. 1.24, pp. 47- 55.
Dakoulas P. & Gazetas G. (2007), “Insight to Seismic Earth and Water Pressures Against Caisson
Quay Walls”, Geotechnique, Vol. 57, No. 10.
Ebeling, R.M., and Morrison, E. E., (1992), "The Seismic Design of Waterfront Retaining Structures",
US Navy Technical Report NCEL TR-939, U.S. Army Engineer Waterways Experiment
Station, Vicksburg, MS.
Egan J.A., Hayden R.F., Scheibel. L.L., Otus M., Serventi G.M., (1992), “Seismic repair at Seventh
Street Marine Terminal", Grouting Soil Improvement and Geosynthetics, Geotechnical Special
Publication No.30, ASCE, p.p 867-878.
Elnashai A.S., Gencturk B., Kwon O., Al-Qadi I.L., Hashash Y., Roesler J.R., Kim S.J., Jeong S.,
Dukes J., Valdivia A., (2010), “The Maule (Chile) Earthquake of February 27, 2010:
Concequence Assessment and Case Studies”, Mid-America Earthquake Center, Report No. 10-
04, 2010-12-31.
Iai S., Matsunaga Y., Morita T., Miyata M., Sakurai H., Oishi H., Ogura H., Ando Y., Tanaka Y., Kato
M., (1994), “Effects of remedial measures against liquefaction at 1993 Kushiro – Oki
earthquake”, Proc. Fifth U.S. – Japan Workshop on Earthquake resistant design of Lifeline
Facilities and Countermeasures against Soil Liquefaction, National Center for Earthquake
Engineering Research, NCEER-94-0026, p.p. 135-152.
Itasca, (2005) "Fast Lagrangian Analysis of Continua", Itasca Consulting Group Inc., Minneapolis,
Minnesota.
Okabe S. (1926). General Theory of Earth Pressure. Jnl. Japan Soc. Civil Engrs., Tokyo
PIANC (2001), Seismic Design Guidelines for Port Structures, Int. Navigation Association

10
 P.Tasiopoulou, N.Gerolymos and G.Gazetas 11

Pitilakis K. and Moutsakis A., (1989), “Seismic analysis and behaviour of gravity retaining walls – the
case of Kalamata harbour quaywall”, Soil and Foundations, Vol. 29, No. 1, pp. 1-17.
Sugano T., and Iai S., (1999), “Damage to port facilities”, The 1999 Kocaeli Earthquake Turkey –
Investigation into Damage to Civil Engineering Structures, Earthquake Engineering Committee,
Japan Society of Civil Engineers, p.p.6-1 – 6-14.
Tasiopoulou P., Gerolymos N., Tazoh T. and Gazetas G.(2013), "Pile-Group Response to Large Soil
Displacements and Liquefaction: Centrifuge Experiments Versus A Physically Simplified
Analysis", Journal of Geotechnical and Geoenvironmental Engineering, ASCE , Vol. 139(2),
pp. 223-233.
Zarzouras Or., Gerolymos N., Gazetas G. (2010), “Seismic Response of Caisson Quay-Wall in a
Liquefied Environment. Analysis of a Case History”, Proceedings of the 6th Hellenic
Conference on Geotechnical and Geoenvironmental Engineering, Volos, Greece, September
29–October 1, 2006, Vol. 1, pp. 549–556 (in Greek).
Zhang MH and Gjorv, OE (1991) “Effect of silica fume on pore structure of cement paste,” Cement
and Concrete Research, 21(1):1006-1014