Construction and Building Materials 170 (2018) 366–377

Contents lists available at ScienceDirect

Construction and Building Materials

journal homepage: www.elsevier.com/locate/conbuildmat

Virtual Sliding Pipe Rheometer for estimating pumpability of concrete

Venkatesh Naidu Nerella, Viktor Mechtcherine ⇑
Technische Universität Dresden, Institute of Construction Materials, Dresden, Germany

h i g h l i g h t s

A single-phase CFD model was developed for simulating pumpability tests with Sliper.

Models without a lubricating layer (LL) are not generally applicable to concrete.

The lubricating layer was implemented in the pipe flow CFD model.

Influences of rheological parameters and thickness of LL were demonstrated.

Particle-concentration-based models were suggested to estimate LL properties.

a r t i c l e i n f o a b s t r a c t

Article history: Predicting the pumping pressure needed to ensure a consistent flow rate of concrete is crucial to the
Received 1 September 2017 success of current construction processes and newly introduced, but fast-growing 3D-concrete print-
Received in revised form 26 February 2018 ing techniques. The Sliding Pipe Rheometer (Sliper) has recently proved to be a reliable experimental
Accepted 1 March 2018
tool in predicting pumping pressure. Building on the experimental results of an earlier investigation
Available online 23 March 2018
by means of Sliper, a single-fluid numerical model for simulating Sliper tests (virtual Sliper) was
developed using Computational Fluid Dynamics (CFD). Various observations as well as numerical lim-
Keywords:
itations of the model and their physical origins were analysed. It was demonstrated that lubricating
Pumpability
Computational Fluid Dynamics
layer has vital influence on concrete pumping and that single phase numerical models, not consider-
Lubricating layer ing the lubricating layer are only applicable to some specific concrete compositions. Hence, the initial
Sliding Pipe Rheometer single-phase model was improved by implementing a separate lubricating layer; its properties were
calculated using Chateau-Ovarlez-Trung and Krieger-Dougherty models. Experimental and numerical
comparative analyses confirm the validity of the above-mentioned approach in calculating lubricating
layer properties. Parameter sensitivity analysis showed that the plastic viscosity and thickness of the
lubricating exert the dominant influences on pumping pressure. The virtual pumpability testing tool
as developed should enable a more purposeful material design of pumpable concrete and
pre-estimation of pumping processes.
Ó 2018 Elsevier Ltd. All rights reserved.

1. Introduction blockages or circuit breakdowns over longer time durations

Concrete pumping has emerged as one of the most important
processing technologies in the construction industry. It often
makes possible the reduction of construction costs and consider-
ably speeds up the construction process. Concrete pumping is also
crucial to many advancements in construction-related processes
such as Digital Construction (DC), often called 3D-printing with
concrete. In the case of onsite DC techniques, concrete must be
transported or pumped over longer distances, periodically
stopping after having pumped each layer, but without any
⇑ Corresponding author.
E-mail address: mechtcherine@tu-dresden.de (V. Mechtcherine).
(see Fig. 1).
1.1. Testing pumpability of concrete
Concrete pumping occurs by pushing concrete using high pres-
sure into pipelines made of either flexible, abrasive resistant mate-
rial or steel. In other words, the force of the applied pressure causes
the concrete material to deform in the direction of the force
applied and so to transmit the force further. The pressure required
to pump concrete depends on its composition as well as on the
pumping specifications such as distance, height to pump, pipe
diameter, and discharge rate. Any changes in the composition of
concrete such as water-to-binder ratio, aggregate size distributions
(grading), and admixtures can exert a pronounced influence on its

https://doi.org/10.1016/j.conbuildmat.2018.03.003
0950-0618/Ó 2018 Elsevier Ltd. All rights reserved.
 V.N. Nerella, V. Mechtcherine / Construction and Building Materials 170 (2018) 366–377 367

Fig. 1. Examples of concrete printing technologies: a) a CONPrint3D: concrete boom pump is depicted pumping and precision-placing fresh concrete (source: TU Dresden), b)
Contour Crafting: a concrete pump is depicted pumping concrete to a rail-mounted robotic printhead (source: University of Southern California).

rheological properties, indeed its behaviour during pumping. ifications of pumping circuit, one can estimate the discharge pres-
Hence, predicting the discharge pressure needed to pump a partic- sure P required for a pumping circuit under field conditions using
ular concrete mixture over a given distance or height at a specific Eq. (1) (considering plug/slip flow with no deformation in the plug
rate is not trivial. Much research has been conducted on this topic, [10,11]):
especially during the last two decades, which has led in turn to dif-
4L 16  L  Q
ferent prediction models and test approaches [1–7]. Convention- P¼ aþ bþqgH ð1Þ
ally, this crucial task has been accomplished using methods
D p  D3
based on empirical test methods [6] or experience [8]. However, where L is the length, D is the diameter of the pipeline, q is the den-
such empirical approaches have proven unreliable because of dif- sity of concrete, H is the pumping height, and Q is the desired flow
ferences in laboratory testing conditions and the corresponding rate.
actual situation during pumping [4,6,9]. Not so in the case of the Assuming a linear relationship between P and Q, see [4,10] and
Sliding Pipe Rheometer (Sliper, see Fig. 2) which is a relatively Fig. 2c, according to Eq. (2),
new device developed to overcome the conventional problems in
P ¼ A þ B  Q þ PH ð2Þ
testing the pumpability of concrete and estimating discharge pres-
sures for various concrete types [4]. where PH = pressure induced by deadweight of concrete in the Sli-
As the first part of this research, extensive laboratory investiga- per pipe, parameters a and b can be calculated from the P-Q plot
tions [6] were conducted using Sliper. Using the rheological of Sliper experiments using Eq. (3) which is derived by combining
parameters a and b obtained from Sliper experiments and the spec- Eqs. (1) and (2):

Pressure
sensor
Upper pipe
(on the top of
the piston)
Handles

(b)
Lower pipe
Pressure P

(inside the pipe)

Piston B
(inside the
lower pipe)
A
PH

Flow rate Q
(a) (c)

Fig. 2. a) Sliding Pipe Rheometer, courtesy of Schleibinger Geräte T. u. G. GmbH; b) pressure sensor (in dotted circle); c) schematic view of P-Q plot (A = parameter related to
yield stress, B = parameter related to plastic viscosity, PH = deadweight pressure of concrete).
368 V.N. Nerella, V. Mechtcherine / Construction and Building Materials 170 (2018) 366–377
Bpd
3
dA
a¼ ; b¼
4l 16  l
where A is the P-intercept of the P-Q curve, B is the slope of P-Q
curve, and l and d are the length and diameter of Sliper pipe, 0.5
m and 0.126 m, respectively.
1.2. Research significance
Today’s very broad spectrum of concrete compositions, which
certainly become even broader in the future, makes it necessary
to perform numerous experiments to characterise their rheological
properties and subsequently determine their pumpability. More-
over, considering on the one hand that concrete properties are
time- and thixotropy- (resting time) dependent and that on the
other hand pumping in the case of modern processes such as DC
demands pumpability for longer durations, extensive experimental
studies are needed to assure the pumpability of concrete, even if its
composition is unaltered. Therefore, a purely experimental
approach for assessing pumpability does not seem feasible in all
the cases, since extensive experimental tests demand a great deal
of time and economic investment.
As in many other fields, numerical simulations present a
promising tool in complementing experimental studies [12] since
they may not only save time and costs but also increase the accu-
racy of the predictions, enable better insight into material beha-
viour, and support material design as well. With this in mind the
experimental research on the pumpability of concrete [6,11] has
been extended by developing a numerical model which simulates
the Sliper experiment and in this way enables the estimation of
the discharge pressure numerically. These numerically determined
pressures from the virtual Sliper can thereafter be utilized to pre-
dict the pumping pressure for pumping circuits of various lengths
and configurations. The virtual Sliper also makes it possible to
investigate numerous concrete types with wide variations in their
rheological properties without conducting physical experiments.
Thus, virtual Sliper can play an important role in establishing intel-
ligent material design approaches for pumpable concretes. In addi-
tion, virtual Sliper becomes an inherent part of DC approaches,
where minimal involvement of human labour is envisioned with
enhanced machine-material integration (autonomous in the lines
of Industry 4.0). Fresh concrete properties can be determined
online or ‘in the process’ during transporting and the printing pro-
cess, these properties could then be sent to virtual Sliper. Virtual
Sliper predicts the pumping pressures needed to ensure consistent
pumping flow-rates considering the ‘change’ in rheological proper-
ties as well as in target structure geometry. Thus, a numerical
model to simulate Sliper experiments or another adequate tech-
nique to predict pumping behaviour is indispensable.
1.3. Numerical modelling of fresh concrete
The common numerical approaches in simulating the behaviour
of fresh concrete are Computational Fluid Dynamics (CFD)
[9,13,14] and the Discrete Element Method (DEM) [15,16]. In the
DEM concrete is modelled by a collection of interacting distinct
particles, which can establish and break contacts dynamically.
The method enables modelling of large strain and stress phenom-
ena of heterogeneous materials. However, its applicability to large
scale simulations (concrete volume of more than 10’s of litres) is
strongly hindered by the requirement that millions of particles
be simulated and subsequently demanding substantial computa-
tional resources.
In CFD, concrete is modelled as a single continuum phase or as
multiple continuum phases, while the modelled domain is dis-
cretized as interconnected mesh elements. This means that the
rheological properties of concrete in a phase are constant through-
ð3Þ out the modelled geometry. Numerical solution for fresh concrete’s
rheological behaviour is then made possible by adding apparent
viscosity, commonly following the Bingham or Herschel-Bulkley
models, to the Navier-Stokes equations [13,17,18]. This method
excludes the interactions among solid particles from the simula-
tion and, hence, is not appropriate to simulate heterogeneous phe-
nomena such as segregation and blocking directly [12].
Nevertheless, CFD has proven to be a useful tool for simulating
fresh concrete flows where macroscopic phenomena such as pres-
sure vs. flowrate relationships during pumping are the focus [9]. It
is worthy of note that it is possible to couple CFD calculations, rep-
resenting the fluid phase, with DEM, which represents the particle
phase, but such calculation procedures are very complex, very
time-consuming, and still in the early stages of their development
[14].
It is possible to simulate large volume concrete flows such as
castings [13] and full scale pumping [19] using CFD. Thrane et al.
[20] simulated mould-filling process of SCC as well as slump flow
simulations with a phenomenological micro-mechanical CFD
model. Tichko et al. [21] developed a CFD model for computing
formwork pressures by simulating form filling through bottom-
up concrete pumping. Further literature presents various CFD
modelling using single-phase flow and multi-phase approaches;
see [12,13,22–27]. Jo et al. [9] developed a computational approach
to estimate lubricating layer (hereinafter LL) in concrete pumping.
They emphasized the inadequacy of conventional pumping pres-
sure prediction methods with the help of field measurements
and a mock full-scale pumping test. Tan et al. [28] developed a
multi-phase numerical model to investigate wear mechanisms of
concrete piping wall. They combined DEM and CFD approaches:
DEM to model concrete aggregates as discrete particles and CFD
to model the continuous fluid phase. Choi et al. [19] simulated
fresh concrete flow in a 170 m long pumping circuit using a
single-fluid CFD approach. The best fitting thickness for LL was
determined by comparing calculated pressure with the experimen-
tally measured values at a flow rate of 50 m3/h. Furthermore, they
analysed pressure profiles comparatively along the circuit geome-
try with the pressure profiles measured experimentally and
observed that the deviations between the CFD-calculated results
and experimental data were below 7% [19]. Choi et al. [5] also
investigated the formation of the lubrication layer and simulated
the mechanism of shear-induced particle migration (SIPM) during
concrete pumping. Influences of particle shapes were indirectly
modelled by solving SIPM equations, implementing a User-
Defined Scalar into ANSYS Fluent [5]. It was concluded that the
LL was the dominant aspect in defining the flow of concrete during
pumping and found numerical simulations in considering particle
shape to correlate well with experimentally measured velocity
profiles using an Ultrasonic Velocity Profiler. Chen et al. [29] per-
formed CFD simulations of wet shotcrete flow in a pipe, including
modelling of LL. In addition to finding the best fitting thickness of
the LL, the authors also compared numerically determined LL
thicknesses with experimentally measured LL thicknesses and
found that LL for tested shotcrete compositions varied from 1.5
mm to 3.5 mm.
The approach presented in the article at hand is significantly
different to those previous to it. In all the above listed literature,
numerical models were primarily developed to study concrete flow
characteristics, and that of LL, in a long pumping circuit. So far no
numerical model is available to link the resulting rheological prop-
erties as observed in the Sliper tests to any other more commonly
used experimental procedure such as the use of a rheometer or vis-
cometer. This is a rather significant knowledge deficiency that
should be overcome with the utmost urgency, considering the pro-
ven applicability of Sliper for pumping pressure prediction and its
 V.N. Nerella, V. Mechtcherine / Construction and Building Materials 170 (2018) 366–377 369

newness and unavailability in most concrete laboratories and 3. Single-phase numerical model
ready-mix concrete practitioners. This paper addresses this chal-
lenge by presenting virtual Sliper – a CFD model able to predict Sli- In the first step of the Sliper modelling experiments, concrete
per experimental results using rheological input parameters was assumed to be a continuous Bingham fluid, and the approach
according to the Bingham model as obtained from viscometer tests. is termed as ‘Initial single-phase model (ISPM). The numerical
From this point on, this paper is organised into the following model and the corresponding simulation results are presented in
segments: Experimental background (Section 2), Single-phase the following sections.
numerical model (Section 3), and User-defined, single-phase model
(Section 4), followed by the summary and conclusions (Section 5). 3.1. Modelling schema
In both numerical model sections, the modelling schema is pre-
sented in detail, followed by results and discussion. In the initial Concrete flow was modelled by the conservation equations for
single-phase model, Sliper simulations were carried out without mass and momentum of steady-state, incompressible flow. The
considering the LL. It was demonstrated that single-phase numer- continuity equation for the conservation of mass is [30]:
ical models which do not consider the LL are applicable only to
some specific concrete compositions. With that in mind, an initial r ~
u ¼ 0: ð4Þ
single-phase model was improved by implementing a separate LL !
where u is the velocity, r is the divergence in a tensor notation,
using a user-defined function; its properties were calculated using
and r  ~
u is the time rate of change of the volume of a moving fluid
Chateau-Ovarlez-Trung and Krieger-Dougherty models. Finally, the
element in an infinitesimally small cell [17]. The divergence of the
applicability of user-defined single-phase model for predicting
velocity characterizes the volume density of the velocity flux arriv-
pressured measured with Sliper is demonstrated for all concrete
ing and leaving a finite volume.
compositions under investigation.
The momentum in the numerical model is conserved according
to Eq. (5):
2. Experimental background
r  ðq~uÞ ¼ r  r þ q~
u~ g ð5Þ
As the first part of this research, extensive laboratory investiga- where ~ g is the gravitational acceleration, q is the density and r is
tions of the pumpability of fresh concrete were conducted using the stress tensor.
Sliper [6]. Measurement of the pipe velocity representing concrete r can be subdivided into normal stresses rii and the shear stres-
discharge rate (Q) and the pressure acting at the piston head (P) ses sij as:
yielded Sliper parameter a, which is related to the yield stress of 0 1 0 1 0 1
the LL and parameter b related to the plastic viscosity of the LL.
rxx sxy sxz p 0 0 rxx þ p sxy sxz
B C B C B
Twelve different concrete mixtures were systematically investi- r ¼ @ syx ryy syz A ¼ @ 0 p 0 A þ @ syx ryy þ p syz C A
gated, with the major focus on the influence of various constituents szx szy rzz 0 0 p szx szy rzz þ p
of concrete on its rheological properties and pumpability. The r ¼ pI þ s
results obtained clearly demonstrated the links between such mix- ð6Þ
ture parameters as water-to-binder ratio, aggregate shape, use of
admixtures, and consistency class on the one hand and the pumpa- where p is an isotropic pressure term, rp is the pressure gradient,
bility of concrete on the other. Fig. 3 gives an overview of the mix- I is the unit dyadic, and s is the shear stress tensor [31] as reported
tures under investigation. All these concretes were also tested in [12] .
using a ConTec Viscometer 5 and the flow table. The evaluation From Eqs. (5) and (6) we obtain Eq. (7):
of the results emphasised the advantages of using Sliper to charac- r  ðq~uÞ ¼ r  s  rp þ q~
u~ g ð7Þ
terise the pumping behaviour of concrete, while a positive correla-
tion was found between the Sliper viscosity parameter b and the For incompressible flow, shear stress can be expressed as:
plastic viscosity m as obtained from the viscometer tests. Finally, r  s ¼ gD~
u ð8Þ
the predicting capability of Sliper was validated under field condi-
tions by measuring full-scale pumping pressure. The details of the
q~ uÞ ¼ rp þ gD~
u  ðr  ~ u þ q~
g ð9Þ
mixtures tested, testing procedures, and the results obtained are
presented and discussed in the previous article [6]. Here g is the dynamic viscosity; see also Section 3.2.

Fig. 3. Overview of concretes investigated experimentally by Mechtcherine et al. [6].

370 V.N. Nerella, V. Mechtcherine / Construction and Building Materials 170 (2018) 366–377

The concrete flow was considered isothermal in accordance To complete the implementation, the critical shear rate c_ c was
with [12,21,24]. assumed to be close to zero, thus prompting the solver to examine
a second case in Eq. (13) immediately after flow initiates, i.e., even
3.2. Implementation of Bingham model at very low shear rates. Low c_ c implies a very high dynamic viscos-
ity g0 as long as c_ 6 c_ c .
Fresh concrete is a heterogeneous material with constituents of
various sizes and shapes. Its rheological behaviour is that of a yield
stress fluid, i.e., it yields non-zero shear stress at a shear rate of 3.3. Geometry, boundary conditions and model parameters
zero. With increasing shear rates, the shear stress increases often
non-linearly, thus accounting for shear thickening or thinning The numerical discretization of the computational domain plays
effects. Hence, concrete can be classified as a non-Newtonian fluid a vital role in the overall performance of the model. The accuracy of
with yield stress. The Hershel-Bulkley model represents this shear the model significantly increases with more detailed, finer dis-
stress-shear rate (s  c_ ) relationship (Eq. (10), Fig. 4a) using three cretization but leads also to drastic increases in computational
parameters: yield stress s0 , plastic viscosity lpl , and power law time. Hence, an optimum is to be strived for, which would repre-
index n. For the sake of simplicity, the description of concrete as sent the physical phenomena sufficiently while being cost effec-
a Newtonian fluid with yield stress is often applied alternatively, tive. The Sliper pipe is axi-symmetric; see Fig. 2. The flow of the
the so-called Bingham plastic model (Eq. (11), Fig. 4b). This model concrete can be assumed to be axi-symmetric in a macroscopic
was proven to represent concrete’s rheological behaviour with perspective. Therefore, instead of the whole pipe geometry, a quar-
adequate accuracy [16,32,33]. ter of it was modelled using the meshing tool ICEM CFD; see Fig. 5.
The modelled geometry is created as an assembly of four surfaces:
s ¼ s0 þ lpl c_ n ð10Þ pipe wall (Sliper pipe), top (top surface of filled concrete), bottom
(piston head of Sliper, where the pressure sensor is located) and
s ¼ s0 þ lpl c_ ð11Þ internal walls, the inner boundaries of the Sliper quadrant under
consideration.
In the present research, the Bingham model was implemented
The boundary conditions applied are a no-slip moving wall for
by adopting the built-in Herschel-Bulkley model, also called the
the pipe wall, symmetry for the internal walls, and a fixed wall for
bi-viscous model [24] of commercial CFD solver ANSYS Fluent
the top and bottom walls. Because of this relatively simple geom-
(Eq. (12)). The bi-viscous model emulates two viscous fluids: one
etry, the use of the single-phase material model, and a structured
with a very high, constant dynamic viscosity g0 and one with grad-
grid, the computational time is relatively low. The boundary layer
ually decreasing dynamic viscosity gi . The change from one beha-
close to the pipe wall is important in investigating the lubricating
viour to another occurs coincidently with the passing of the
layer phenomenon. The fineness of discretization (dimensions of
critical shear rate c_ c : finite elements) plays significant role in numerical simulations.
8   h i
> s c_ c_ Considering this, a numerical refinement near the wall surface
< g0 ¼ c_ 0c 2  c_ c þ k ð2  nÞ þ ðn  1Þ c_ c c_ 6 c_ c
g¼ ð12Þ was employed; see Fig. 5. Instead of creating fine mesh throughout
h in1
>
: g ¼ s0 þ k c_
i c_ c_ c c_ > c_ c the entire modelled domain, which would increase the computa-
tional time drastically, the region close to the pipe wall were mod-
Assigning the power law index n equal to 1, the above equation elled using a finer mesh. Since the observed thickness of
result in Bingham model as long as shear rate is higher than critical lubrication layer Varies between 1 mm and 5 mm (see details in
shear rate, i.e. c_ > c_ c . Section 4.1), the elements size close to the pipe boundary was cho-
8   sen as 0.84 mm. This resulted in the total number of 58,460 ele-
< g ¼ s0 2  c_ þ k c_ 6 c_ c ments in the model.
c_ c c_ c
g¼ 0
ð13Þ
: g ¼ s0 þ k c_ > c_ c Table 1 gives a conceptual physical description of the numerical
i c_ model and experiments. Input and output parameters of the

Fig. 4. Shear stress vs. shear rate curves f, or the a) bi-viscous and b) Bingham models.
 V.N. Nerella, V. Mechtcherine / Construction and Building Materials 170 (2018) 366–377
Fig. 5. a) Discretization of a Sliper quadrant with b) a detailed view of the optimized mesh.
Table 1
Experimental and numerical model parameters.
Parameters Experiments
Input 1) Concrete composition
2) Velocity of pipe (varies with weights added)
Output 1) Pressure measured with the sensor
2) Pumpability curves (Excel)
numerical model were chosen to reflect the corresponding experi-
mental parameters closely.
Concrete composition and pipe velocity V are the varying
parameters in the Sliper experiments. In other words, different
rheological properties were attained by mix design, and different
pipe velocities were obtained by changing the weights attached
to the pipe [6]. Similarly, the rheological parameters of the con-
cretes under investigation, yield stress s0 and plastic viscosity l;
as measured using viscometer and pipe velocity from Sliper tests,
were used as input parameters for the numerical model. The out-
put parameter from the experiments was pressure P measured
by a sensor at the bottom of the Sliper pipe for different discharge
rates Q, so that pumpability P-Q curves could be plotted [6]. In the
numerical model, pressure acting at the bottom wall was ‘‘mea-
sured” by means of a virtual sensor, which calculates it as area-
weighted averages.
3.4. Results and discussion
Seven concrete mixtures, namely 1, 2, 3, 5, 7, 10 and 11, were
chosen from experimental study [6] based on availability of the
viscometer results, but for a better overview the notations are used
unaltered in this paper. Fig. 6 shows measured pressures from Sli-
per experiments and calculated pressures from ISPM simulations
over corresponding flow rates.
Both experimental and numerical results confirm the consensus
that with an increase in flow rate the discharge pressure becomes
higher as well, showing a nearly linear P-Q relationship. The pres-
sure values calculated with ISPM showed good agreement with the
Sliper measurements, both qualitatively and quantitatively, for
371
Numerical Model
1) Rheological properties (from viscometer tests)
2) Velocity of pipe (wall velocity parameter)
1) Pressure as ‘‘measured” by virtual pressure sensor
2) Pumpability curves (Excel)
mixtures with low water-binder ratio of 0.3, i.e., Mixtures 5, 7,
10 and 11. The representative results are shown in Fig. 6a on the
example of Mixture 10. In contrast, for mixtures with high
water-binder ratios of 0.45 and 0.6, especially for Mixtures 2 and
3 (see Fig. 6b), the calculated results showed little agreement with
experimental findings. Correspondingly, the P-Q diagram can be
roughly subdivided in two areas as showed in Fig. 7.
The primary reason for the poor agreement between the calcu-
lated and measured P-Q curves for concretes with relatively high
w/b appears to be the effect of the lubricating layer, which forms
at the pipe wall and has immense influence on the discharge pres-
sures according to [4,6,10,11,19,34,35]. Since ISPM does not con-
sider the LL, i.e., constant material properties are assigned
throughout the model domain (Sliper pipe), the pumping beha-
viour of mixtures with relatively high water-to-binder ratios can-
not be predicted well because the rheological properties of the LL
in such mixtures differ very pronouncedly from those of bulk con-
crete and the thickness of the LL is relatively high in comparison to
concretes with lower w/b. This conclusion is supported by the
results reported by Jo et al. [9], who reported a decrease of LL thick-
ness from 5 mm to 1 mm with decreasing water-to-binder ratio.
Furthermore, Choi et al. [35] measured increases in LL thickness
with increasing w/c. Thus, neglecting LL in the ISPM makes the
model only applicable to concrete with a low w/b.
Still further, from the fundamentals of the numerical model
developed, it can be deduced that the larger the discrepancies in
Sliper and viscometer measurements for a concrete mixture,
the poorer the experimental and numerical correlation (ENC) is.
Table 2 shows that mixtures with high w/b, i.e., Mixtures 1, 2
and 3, yielded very low values of Sliper viscosity parameter b,
372 V.N. Nerella, V. Mechtcherine / Construction and Building Materials 170 (2018) 366–377

Fig. 6. a) Comparison of experimental results and corresponding numerical simulations for Mixture 10 (w/b = 0.3) and b) Mixtures 1 and 3 (both with w/b = 0.45).

Fig. 7. Correlation of the results of the simulation comparing ISPM with experimental results; the base figure is adopted from [6]. Note: correlation for Mixture 1 is better
than that for Mixture 2 or Mixture 3; see Discussion.

Table 2
Experimental results from Sliper and viscometer tests, adopted from [6] showing relative viscosities and discrepancy.

Mixture-Nr Sliper Viscometer Sliper Viscometer Discrepancy

b l Rb Rl D
Pas/mm Pas Pas/mm Pas %
Mixture 1 0.59 76 0.51 0.60 17.57
Mixture 2 0.14 38 0.12 0.30 147.73
Mixture 3 0.81 230 0.70 1.83 159.16
Mixture 5 2.3 236 2.00 1.87 6.35
Mixture 7 1.74 164 1.51 1.30 13.98
Mixture 10 1.15 126 1.00 1.00 0.00
Mixture 11 3.1 291 2.70 2.31 14.32

0.59 Pas/mm, 0.14 Pas/mm and 0.81 Pas/mm, respectively, while
the corresponding plastic viscosities measured by viscometer were
76 Pas, 38 Pas and 230 Pas, respectively. To understand the rela-
tive sensitivity of both Sliper and viscometer to mixture variation,
a value denoted as ‘discrepancy’ D (Eq. (14)) was calculated. When
comparing measurements from different instruments, to avoid the
common variables such as geometrical factors and test conditions,
a ‘relative parameter’ can be calculated. Ferraris et al. calculated
relative plastic viscosity to compare measurements from various
rheometers [36]. By dividing Sliper and viscometer measurements
of viscosity viz. bn and ln of any composition n with the same of a
reference composition r, relative viscosities were calculated. In this
research, Mixture 10 was chosen as reference considering in the
measured spectrum of viscosities its closeness to the medium vis-
cosity measurements as well as the low influence of the lubricating
layer; see also Section 4.2. In the next step, the discrepancy is
 V.N. Nerella, V. Mechtcherine / Construction and Building Materials 170 (2018) 366–377
quantified by dividing the difference in relative viscosities as
obtained from Sliper and viscometer measurements with relative
viscosity obtained by using Sliper Rb.
D ¼ ½ðbn =br Þ  ðln =lr Þ=Rb: ð14Þ
The calculated discrepancy values are presented in Table 2
as percentages. Very high discrepancies in cases of Mixtures 2
and 3 indicate very poor ENC. The negative sign in the discrep-
ancy indicates that Sliper relative viscosity parameter showed
higher reduction than that of relative plastic viscosity from
viscometer.
It is noteworthy here that Sliper measurements catch the
influence of the LL much better than does the viscometer, which
originates from the devices’ respective geometries and function-
ing principles in both experiments. In the viscometer, the whole
concrete is sheared with the help of special ribs designed to pre-
vent slippage. Although in Sliper varying from mixture to mix-
ture, primarily the material in the vicinity of the pipe wall is
deformed. The discrepancy of mixtures with poor ENC, i.e., 2, 3
and 1 all are negative and high, thus showing higher reductions
in relative viscosities measured with Sliper than those with vis-
cometer. Thus, it is evident that among the concretes under con-
sideration, the behaviours of Mixtures 2 and 3 followed by
Mixture 1 were significantly influenced by a prominent LL. Mix-
ture 1 showed a somewhat better ENC than Mixture 3 even
though it had the same w/b. This can be explained by the fact
that the yield stress of M1 was with 245 Pa among the lowest
of all tested concretes, and much lower than 450 Pa and 533 Pa
measured for M2 and M3, respectively. The low yield stress of
M1 appears to have resulted in the partial shearing of the plug
during Sliper testing; see [10,37]. In other words, here the shear
of the plug affects the measured pressure in addition to the defor-
mation of the LL. Thus, since ISPM considers only the shear of
concrete bulk, M1 yielded better ENC in comparison to M2 and
M3. This view is also supported by the best-fit results presented
in Section 4.2.
Obviously, the applicability of ISPM is strongly limited due to its
not considering the LL. To overcome this limitation, ISPM was
improved by implementing a user-defined function as described
in the next section.
4. User-defined single-phase model
A single-phase model with user-defined functions, hereinafter
UDFM, enables the incorporation of the LL’s influence into the ini-
tial single-phase model. Even though it is possible to model the LL
and plug as separate phases using a multi-phase model, such an
approach demands relatively large amounts of computational
resources, and so it increases model complexity and, accordingly,
computational time. In contrast, the single-phase model calcula-
tion with UDF took approximately the same time as the simula-
tions using the initial single-phase model.
4.1. Modelling schema
The LL forming during the flow of concrete in a pipe is assumed
to be composed of cement, water, admixtures and fine aggregates;
therefore, it can be broadly considered to be fine mortar. The
absence of coarse aggregates in the fine mortar layer reduces its
yield stress s0 and plastic viscosity l dramatically. From the per-
spective of numerical modelling, this means that the fluid elements
(cells) near the wall have lower yield stress s0 and plastic viscosity
l than more distant elements. To assign varying rheological prop-
erties appropriately across the cross-section of Sliper, the numeri-
cal model must identify the positions of all elements in a specified
373
region near the pipe wall and assign the corresponding properties.
This task was accomplished by implementing ‘user defined func-
tions’ (UDF) in ANSYS Fluent controlling s0 and l values of the
‘fluid elements’ based on their distance from the pipe wall.
The UDF approach improves model performance by adding four
parameters to the initial model: layer thickness TLL, yield stress fac-
tor sf , viscosity factor lf and gradient factor; see Fig. 8. The yield
stress and viscosity factors are the ratios of yield stress and plastic
viscosities of mortar to the same values for concrete. Note that in
the literature the terms ‘dimensionless viscosity’ and ‘dimension-
less yield stresses’ or ‘relative viscosity’ and ‘relative yield stress’
have been used previously [36,38]. However, these terms were
introduced as the ratio of higher viscous material to the lower vis-
cous material, while the present case is the inverse, hence the
choice of the new terms. TLL defines the region in which the prop-
erties of fine mortar should be applied. Defining two different sets
of material properties in the LL domain on one hand and plug
domain on the other implies an abrupt local variation of the model
parameters, causing numerical instability. Hence, the rheological
properties over the TLL are varied with a gradient, having a lower
limit at the pipe wall and an upper limit (equal to plug properties)
at the inner boundary of the LL. Feys et al. [39] observed that the
flow of SCC during pumping can be described as plug flow with a
uniform velocity profile and a LL with large velocity gradient. In
addition, the results from Jo et al. [9] and Choi et al. [40] affirmed
the existence of a large velocity gradient in the vicinity of the pipe
wall, thus substantiating the approach chosen for UDFM. It possi-
ble to vary properties within the LL using linear, exponential or
any other mathematical function. For the current model, a linear
function was chosen.
Determining the thickness and rheological properties of the LL
is not trivial. Experimental investigation of LL requires special
equipment such as Ultrasonic Velocity Profiler [19] and/or tri-
bometer [37]. Such an examination has not been a part of the pre-
ceding experimental work [6]. Therefore, a theoretical approach for
estimation of UDFM model parameters sf and lf was followed.
There are numerous investigations on predicting yield stress and
plastic viscosity of a suspension based on the corresponding values
of the suspending fluid [41–47]. Mahaut et al. [38,45] reported that
the yield stress of suspensions depends only on coarse particle con-
centration (suspending phase) and yield stress of the carrying fluid
and is independent of physicochemical interactions. They observed
that elastic modulus-to-concentration relationship followed the
Krieger-Dougherty Law and was in agreement with Chateau et al.
[44], deriving the following model:
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
sc ð/Þ u 1/
¼u ð15Þ
sc ð0Þ t1  / 2:5/m
/m
According to Krieger-Dougherty [41], the viscosity of suspen-
sions can be related to the coarse particle concentration and vis-
cosity of fluid matrix as:
lc ð/Þ 1
¼ ð16Þ
lc ð0Þ 1  / ½g/m
/m
In the above models, sc ð/Þ, lc ð/Þ are the yield stress and viscosity of
the suspension with particle concentration (packing fraction) of /.
sc ð0Þ and lc ð0Þ are the yield stress and plastic viscosity of the sus-
pension with particle concentration of zero; thus, the plastic viscos-
ity of the fluid matrix, /m is the maximum possible packing fraction
and ½g is the intrinsic viscosity [40,48].
Eqs. (15) and (16) are valid for monodisperse suspensions. For
multidisperse suspensions such as concretes, as reported in [40],
the Farris [42] model should be employed:
374 V.N. Nerella, V. Mechtcherine / Construction and Building Materials 170 (2018) 366–377
Fig. 8. Sketch of modelled plug and lubricating layers in UDFM including model parameters.
 ½ga /a;m  ½gb /b;m
lc ð/Þ / /
¼ 1 a 1 b
lc ð0Þ /a;m /b;m
where indices a and b indicate two phases in the suspension, e.g.,
fine sand and coarse sand in concrete.
Following the above models, Jo et al. [2] deduced for concrete
Eqs. (18) and (19):
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffivffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
sco ð/Þ u 1  /cement u 1  /sand
¼u u independent of this work ascertained this phenomenon very
si ð0Þ t1  /cement 2:5/cement;m t1  /sand 2:5/sand;m
/cement;m
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u
u 1  /gravel
u
t 2:5/gravel;m
/
1  / gravel
gravel;m
 2:5/cement;m  2:5/sand;m
lco ð/Þ / /sand
¼ 1  cement 1
li ð0Þ /cement;m /sand;m
where /cement , /sand and /grav el are the concentrations of cement, sand
and gravel in cement paste, mortar, and fresh concrete, respec-
tively; /cement;m , /sand;m and /grav el;m are the corresponding maximum
concentrations; sco ð0Þ and lco ð0Þ are the plastic viscosity and yield
stress of concrete with solid particle concentration /; si ð0Þ and
li ð0Þ are the plastic viscosity and yield stress of concrete with zero
solid particle concentration.
Within the constraints of the work at hand, upon implementa-
tion of LL through UDF, a parametric study was carried out by vary-
ing one model parameter while the other two were kept constant.
The thickness of the LL (TLL) was varied from 1 mm to 5 mm based
on literature and reported expert opinions [9,40]. The viscosity fac-
tor lf was varied from 0.001 to 0.8, while yield stress factor sf was
varied for a few mixtures from 0.08 to 0.8 and kept constant at
sf ¼ 0:2 for the other mixtures; see Section 4.2. Finally, the lf val-
ues at which numerical simulations fitted the experimental results
best were compared with lf values calculated using Eq. (16) and
considering the LL to be equivalent to the representative mortar.
It is noteworthy that the rheological properties of the lubrication
layer in pipe flow may not be necessarily equal to those of consti-
tuting mortar belonging to the corresponding concrete. The valid-
ity of this condition depends upon the concrete composition and
flow characteristics. For flows with a very thin LL, the constituting
of fine mortar consisting of cement paste and very fine sand is
apparently realistic. However, in this first study, for sake of sim-
plicity, the properties of the LL were assumed to be equivalent to
those of constituting mortars.
4.2. Results and discussion
Though ISPM yielded good correlations for mixtures with low
w/b, it does not imply the absence of LL during the pumping of
these mixtures. In reality, the existence of LL for all mixtures is
very probable. However, the crucial point to notice is the thickness
and properties of this layer as well as flow characteristics, i.e. plug
ð17Þ
flow or plug-plus-shear flow. Kaplan et al. [10] suggested that con-
crete flows as slip-plus-shear flow after the interface shear stress si
exceeds the yield stress s0 of concrete. Even in case of low w/b,
common for SCC, presence of LL with partial shearing of plug
region is apparent. Feys et al. [39] suggested that due to very low
yield stress part of bulk concrete is partially sheared during the
pumping of SCC. Research carried out at the TU Dresden [11,37]
clearly with SCC compositions having w/b = 0.29 (w/(c + 0.4fa))
/sand;m
or 0.39 (w/(c + fa)) and resulting in the complete shear of bulk con-
crete. Nonetheless, if the thickness of the layer is small for the
ð18Þ tested mixture or the difference in plug and LL properties is rela-
tively low, then the simulation results obtained without imple-
menting LL (ISPM) correlate reasonably well with the measured
!2:5/grav el;m results.
/grav el
Disregarding the good correlation of a few mixtures in cases of
1
/grav el;m
ISPM, all the mixtures under investigation were simulated using
ð19Þ UDFM. In the first step, a sensitivity analysis for the model param-
eters was conducted, documenting the high influence of lf on the
calculated pressure. Fig. 9a illustrates the comparative results for
the calculated pressure with viscosity factors varying from 0.001
to 0.2, while layer thickness and yield stress factor were kept con-
stant at 5 mm and 0.2, respectively. It was also observed that the
increment in viscosity factor can cause drastic drops of the pres-
sure at higher discharge rates Q and lower pressure drops at lower
discharge rates Q. Similarly, layer thickness TLL also showed consid-
erable influence on the calculated results. Fig. 9b gives the compar-
ative results of measured and calculated pressure values for layer
thicknesses of 3, 4 and 5 mm.
In contrast, sf variation yielded minimal influence on calculated
pressures. Fig. 10 shows the comparative results for yield stress
factors of 0.08, 0.2 and 0.8 with layer thickness of 5 mm and viscos-
ity factors of 0.001 (Fig. 10a) and 0.02 (Fig. 10b). Similar low influ-
ences of yield stress on rheological behaviour of concrete have
been generally reported, especially in cases of high shear flows
where the final shape of the fluid body is not of primary concern
[9,14,24]. Here it may probably originate from the flow character-
istics of concrete in the pipe. Prior to a Sliper experiment the for-
mation of the LL is ensured with help of pre-strokes [6]. During
the progress of a Sliper experiment, due to the presence of LL
and ever-increasing pipe velocity, shear rates in the vicinity of
the pipe wall are relatively high. From the fundamentals of the
Bingham model, it is clear that at higher shear rates the influence
of yield stress is relatively low since the plastic viscosity tends
towards dynamic viscosity; see also Fig. 4. Considering the low
influence of sf on calculated pressures, its value was set constant
at 0.2 for all the subsequent simulations.
Table 3 presents the final best-fit parameters from the numeri-
cal simulations as well as lf values calculated using Eq. (16). It is
noteworthy that Mixtures 1 and 5 contained round natural quartz
aggregates while all the other mixtures contained crushed basalt
aggregates. The intrinsic viscosity ½g for the round and crushed
aggregates was assumed, broadly based on [40], to be 2.5 and 5,
 V.N. Nerella, V. Mechtcherine / Construction and Building Materials 170 (2018) 366–377 375

Fig. 9. Results of simulations using UDFM; comparison of results calculated a) with different viscosity factors (layer thickness 5 mm, yield stress factor 0.2), b) with different
layer thicknesses of 3, 4 and 5 mm (yield stress factor 0.2, viscosity factor 0.001).

Fig. 10. Results of UDFM simulations for different yield stress factors carried out with layer thickness 5 mm and viscosity factor a) 0.001 and b) 0.02.

Table 3
Viscosity factors lf calculated using Eq. (16) and best-fit values from simulations. LL remains constant throughout the pumping period [5,10,35,37]
and the dependency of TLL on concrete composition is becoming
Mixture Eq. (16) Best fit (simulation with experiments)
more and more clear [9,11,37,39,49]. Furthermore, the gradient
1 0.10 0.50/0.40/0.12 for TLL = 5/4/1 mm, respectively of LL properties and, as a consequence, velocity profiles in pipe
2 0.07 0.07 for TLL = 5 mm
flow could also be experimentally deduced or assumed based on
3 0.05 0.02 for TLL = 5 mm
5 0.11 0.07/0.10 for TLL = 1/2 mm, respectively the literature [9,19].
7 0.07 0.80 for TLL = 1 mm Nonetheless, the dependence of any suspension’s rheological
10 0.07 0.07 for TLL = 1 mm properties on their suspending fluid is influenced by the
11 0.07 0.07 for TLL = 1 mm

a) Maximum packing fraction of the particles;

respectively. For Mixtures 2, 5, 10 and 11, best-fit values matched
positively with viscosity factors calculated/predicted using Eq.
(16); for Mixture 3 the match was very close. The only divergent
case in these results was Mixture 7 for reasons yet to be deter-
mined. Mixture 1 in a peculiar case showed a positive ENC for three
different TLL, viz. 5 mm, 4 mm and 1 mm, each with different lf ,
viz. 0.50, 0.40 and 0.12. The ambiguous results are rather another
upshot of slip-plus-shear flow that occurs in case of low yield
stress concrete; see Section 3.4. Nevertheless, acceptable agree-
ment of best-fit values and predicted viscosity factors make it pos-
sible to estimate lf parameters roughly for further studies, largely
independent of empirical knowledge. In addition, by assigning cal-
culated sf and lf values, one can determine TLL for a particular
composition if experimental pressure and flow rate results are
available. Also, input parameters of UDFM, that Eq. (16) does not
indicate, are the thickness of LL and the gradient of LL properties.
However, as repeatedly proven, once formed, the thickness of the
b) Packing fraction of particles or paste volume in the concrete;
c) Particles, shape, size distribution and size range: minimum
and maximum diameter;
d) Intrinsic viscosity of particles.
In addition, in the case of applying particle concentration mod-
els for Sliper model, TLL has also to be considered. Considering the
complexity of this phenomenon, the findings presented in this
study should be taken as preliminary observations on the applica-
bility of particle concentration models for determining properties
of the LL. Further studies are essential to explain how the parame-
ters influence the applicability of such models in determining the
properties of the LL.
Fig. 11 brings together the results of experimental measure-
ments and of ISPM and UDFM simulations. The results of numerical
simulation using the calibrated UDFM model showed very good
agreement with experimental results for both mixtures. For other
mixtures similar results were obtained as well.
376 V.N. Nerella, V. Mechtcherine / Construction and Building Materials 170 (2018) 366–377
Fig. 11. Experimental and numerical results for a) Mixture 10 having a low water content and b) Mixture 3 with high water content.
The layer thickness in the Sliper experiments appears to vary
from 1 mm for mixtures with low water content (low w/b) to 5
mm for mixtures with high water content, in agreement with ear-
lier research findings [4,9,10]. The yield stress of LL is 5 times lower
than that of concrete based on assumed sf . The plastic viscosity
factor, which is the ratio of plastic viscosities of LL to the same of
concrete, varied from 1/1.25 to 1/20. For Mixtures 1, 2, 10, and
11 the plastic viscosity of the LL is approximately 15 times lower
than that of concrete; note the different TLL.
It is commonly understood that if a concrete mixture can
develop a thick LL, of which the yield stress and plastic viscosity
are always lower than those of concrete, then the pumping dis-
charge pressures are lower in comparison to concrete with a thin
lubrication layer. It is also well known that paste-rich concretes
and concretes with low yield stresses are prone to develop a
thicker LL. The numerical results and experimental measurements
clearly support the earlier knowledge. The mixtures with low
water content (low w/b) which are also exhibit higher yield stress
and plastic viscosity [6] had a best fit TLL of 1 mm indicating very
moderate LL formation.
When looked more specifically, Mixture 2 with a w/b of 0.6 is
very flowable even without superplasticizer [6], and its plastic vis-
cosity measured by means of viscometer is the lowest among all
the compositions tested. Consequently, the pressures P measured
as a function of flowrate Q for Mixture 2 are significantly lower
than all other tested compositions. The w/b of 0.6 and higher paste
content of Mixture 2 ensured a thicker LL with low plastic viscosity
in the Sliper experiments, as demonstrated by viscosity parameter
b, which was equal to 0.14 Pas/mm; see Table 3 in [6]. Thus, irre-
spective of the lower l for the plug, high TLL and lower plastic vis-
cosity are to be assigned to the LL for simulating high w/b
mixtures.
In contrast, mixtures with the lowest w/b of 0.3 seem to
develop a very thin LL. The mixtures 5, 7, 10 and 11, all with lower
w/b ratio (0.3) exhibited best-fit TLL values of 1 mm. For Mixture 5
best-fit lf is 1/10 to 1/14, depending on layer thickness. Mixture 5
has a w/b of 0.3 and is without any secondary cementitious mate-
rials; it has the lowest water content among all mixtures having
W/B of 0.3. For Mixtures 1 and 5 the plastic viscosity of the LL is
approximately 10 times lower than plastic viscosity of concrete.
For the best fit of Mixture 3, a 50 times lower plastic viscosity of
the LL than that of the plug was determined. At first glance, this
may lead to false conclusions about Mixture 3’s having a 5 mm
thick LL of very low lf , lower than that of Mixture 2. However,
in the experimental investigations it was observed that in the case
of Mixture 3, due to a potential experimental error in the viscome-
ter tests, the torque-rotational speed curve T-N did not fit well with
the P-Q curves of the Sliper tests. While for all other mixtures T-N
curves correlated well with P-Q curves, the T-N curve measured for
Mixture 3 was very steep, leading to a much higher plastic viscos-
ity measurement from the viscometer [6]. Such a high plastic vis-
cosity, when used as input for numerical model, leads to the
prediction of overly high Sliper pressures. Consequently, best fit
values for TLL and lf have to be high and very low, respectively.
5. Summary and conclusions
A numerical tool to simulate Sliper tests is highly relevant in
predicting the pumping behaviour of concrete. Such a tool, a virtual
Sliper, was developed using the single fluid CFD method, short
ISPM. ISPM predicted the general trend of pressure variation with
respect to the discharge rate correctly and provided acceptable
agreement with experimental results in the case of concretes with
low water-to-binder ratios. However, since the lubrication layer LL
was not considered, the applicability of ISPM was limited to the
scenarios in which formation of LL was not pronounced.
Thus, an improved single-phase model with User Defined Func-
tion (UFD) was developed, which can model the lubricating layer.
This approach also makes possible the estimation of the thickness,
yield stress and plastic viscosity of the LL for different concrete
compositions. In the framework of a parameter study performed
using the UDF model it was observed that both the thickness and
plastic viscosity of the LL have pronounced influences on concrete
pumpability, while the yield stress parameter showed a negligible
effect. Additionally, the best-fit results suggest that the assumption
of a yield stress factor parameter equal to 1/5 is appropriate, while
plastic viscosity factor and thickness parameters varied within the
ranges from 1/1.25 to 1/20 and from 1 to 5 mm, respectively. The
parameter study and the subsequent empirical fitting of the calcu-
lated results indicate the validity of calculating the viscosity of the
lubricating layer as a function of concrete viscosity using the
Krieger-Dougherty model.
The virtual pumpability testing tool developed should enable
both a more purposeful material design of pumpable concrete
and estimation of pumping processes. In addition, virtual Sliper
might become an inherent part of DC approaches where minimal
involvement of human labour is envisioned with enhanced
machine-material integration. Virtual Sliper can predict the
required pumping pressures with help of supporting on-line
 V.N. Nerella, V. Mechtcherine / Construction and Building Materials 170 (2018) 366–377
sensors to measure transient rheological properties of concrete
during pumping and printing and has the potential to play vital
role in autonomous concrete construction.
Acknowledgements
This research work is part of the doctoral work of the first author
and was partly accomplished at the Weimar Institute of Applied
Construction Research (IAB gGmbH), Weimar, Germany. The
authors gratefully acknowledge the support of Dr.-Ing. Justus
Lipowsky working at IAB gGmbH.
References
[1] N.R. Guptill, D.J. Akers, R.A. Kelsey, J.S. Pierce, C. Bognacki, J.C. King, P.E.
Reinhart, J.L. Cope, W.C. Krell, R.J. Rhoads, M. Garner, G.R. Mass, K.L. Saucier, D.
J. Green, P. McDowell, P.R. Stodola, T.C. Holland, D.T. Parekh, W.X. Sypher, T.A.
Johnson, R.J. Phares, R.E. Tobin, S.A. Kalat, K. Wolf, Placing Concrete by
Pumping Methods, Report by ACI Committee 304, 1996.
[2] D. Kaplan, Pumping of Concrete PhD thesis, Laboratoire Central des Ponts et
Chaussées, Paris, 2001.
[3] F. Chapdelaine, Fundamental and Practical Study on the Pumping of Concrete
PhD thesis, Laval University, 2007.
[4] K. Kasten, Sliding Pipe Rheometer: A Method to Establish the Flow Properties
of High Viscous Media in Pipelines PhD thesis, TU Dresden, 2010.
[5] M.S. Choi, Y.J. Kim, S.H. Kwon, Prediction on pipe flow of pumped concrete
based on shear-induced particle migration, Cem. Concr. Res. 52 (2013) 216–
224, https://doi.org/10.1016/j.cemconres.2013.07.004.
[6] V. Mechtcherine, V.N. Nerella, K. Kasten, Testing pumpability of concrete using
Sliding Pipe Rheometer, Constr. Build. Mater. 53 (2014) 312–323, https://doi.
org/10.1016/j.conbuildmat.2013.11.037.
[7] D. Feys, K.H. Khayat, A. Perez-Schell, R. Khatib, Development of a tribometer to
characterize lubrication layer properties of self-consolidating concrete, Cem.
Concr. Compos. 54 (2014) 40–52, https://doi.org/10.1016/j.
cemconcomp.2014.05.008.
[8] D. Feys, K.H. Khayat, R. Khatib, How do concrete rheology, tribology, flow rate
and pipe radius influence pumping pressure?, Cem Concr. Compos. 66 (2016)
38–46, https://doi.org/10.1016/j.cemconcomp.2015.11.002.
[9] S.D. Jo, C.K. Park, J.H. Jeong, S.H. Lee, S.H. Kwon, A computational approach to
estimating a lubricating layer in concrete pumping, Comput. Mater. Contin. 27
(2012) 189–210, https://doi.org/10.3970/cmc.2011.027.189.
[10] D. Kaplan, F. De Larrard, T. Sedran, Design of Concrete Pumping Circuit, ACI
Mater. J. 102 (2005) 110–117.
[11] V.N. Nerella, E. Secrieru, V. Mechtcherine, Experimental study on influence of
lubricating layer on pumping pressure using Sliding Pipe Rheometer (SLIPER),
Proc.: 19h Int. Conf. Build. Mater. – Ibausil, Weimar, 2015: pp. 1-1247–1254.
[12] N. Roussel, A. Gram, Simulation of Fresh Concrete Flow. State-of-the Art Report
of the RILEM Technical Committee 222-SCF, Springer, Netherlands, 2014.
[13] N. Roussel, M.R. Geiker, F. Dufour, L.N. Thrane, P. Szabo, Computational
modeling of concrete flow: general overview, Cem. Concr. Res. 37 (2007)
1298–1307, https://doi.org/10.1016/j.cemconres.2007.06.007.
[14] N. Roussel, A. Gram, M. Cremonesi, L. Ferrara, K. Krenzer, V. Mechtcherine, S.
Shyshko, J. Skocec, J. Spangenberg, O. Svec, L.N. Thrane, K. Vasilic, Numerical
simulations of concrete flow: a benchmark comparison, Cem. Concr. Res. 79
(2016) 265–271, https://doi.org/10.1016/j.cemconres.2015.09.022.
[15] V. Mechtcherine, A. Gram, K. Krenzer, J.-H. Schwabe, S. Shyshko, N. Roussel,
Simulation of fresh concrete flow using Discrete Element Method (DEM):
theory and applications, Mater. Struct. 47 (2014) 615–630, https://doi.org/
10.1617/s11527-013-0084-7.
[16] S. Remond, P. Pizette, A DEM hard-core soft-shell model for the simulation of
concrete flow, Cem. Concr. Res. 58 (2014) 169–178, https://doi.org/10.1016/j.
cemconres.2014.01.022.
[17] J.D. Anderson, The Governing Equations of Fluid Dynamics, in: Comput. Fluid
Dyn. J. Wendt, Ed., Springer-Verlag, Berlin/Heidelberg, 2009: pp. 15–51. doi:
10.1007/978-3-540-85056-4_1.
[18] K. Vasilic, B. Meng, H.C. Kühne, N. Roussel, Flow of fresh concrete through steel
bars: a porous medium analogy, Cem. Concr. Res. 41 (2011) 496–503, https://
doi.org/10.1016/j.cemconres.2011.01.013.
[19] M. Choi, N. Roussel, Y. Kim, J. Kim, Lubrication layer properties during concrete
pumping, Cem. Concr. Res. 45 (2013) 69–78, https://doi.org/10.1016/j.
cemconres.2012.11.001.
[20] L.N. Thrane, Form Filling with Self-Compacting Concrete, Danish Technological
Institute – Technical University of Denmark, Lyngby, 2007.
[21] S. Tichko, J. Van De Maele, N. Vanmassenhove, G. De Schutter, J. Vierendeels, R.
Verhoeven, P. Troch, Numerical simulation of formwork pressure while
pumping self-compacting concrete bottom-up, Eng. Struct. 70 (2014) 218–
233, https://doi.org/10.1016/j.engstruct.2014.04.008.
[22] J. Gram, Silfwerbrand, Numerical simulation of fresh SCC flow: applications,
Mater. Struct. 44 (2010) 805–813, https://doi.org/10.1617/s11527-010-
9666-9.
377
[23] O. Svec, J. Skocek, H. Stang, J.F. Olesen, P.N. Poulsen, Flow simulation of fiber
reinforced self compacting concrete using Lattice Boltzmann method, Proc.:
8th Int. Congr. Chem. Cem., Madrid, 2011.
[24] K. Vasilic, T. Universit, K. Vasili, N. Roussel, A Numerical Model for Self-
Compacting Concrete Flow through Reinforced Sections : a Porous Medium
Analogy, PhD Thesis, TU Dresden, 2015.
[25] J.E. Wallevik, Thixotropic investigation on cement paste: experimental and
numerical approach, J. Nonnewton. Fluid Mech. 132 (2005) 86–99, https://doi.
org/10.1016/j.jnnfm.2005.10.007.
[26] M.R. Geiker, M. Brandl, L.N. Thrane, D.H. Bager, O. Wallevik, The effect of
measuring procedure on the apparent rheological properties of self-
compacting concrete, Cem. Concr. Res. 32 (2002) 1791–1795, https://doi.org/
10.1016/S0008-8846(02)00869-4.
[27] J.E. Wallevik, O.H. Wallevik, Analysis of shear rate inside a concrete truck
mixer, Cem. Concr. Res. 95 (2017) 9–17, https://doi.org/10.1016/j.
cemconres.2017.02.007.
[28] Y. Tan, H. Zhang, D. Yang, S. Jiang, J. Song, Y. Sheng, Numerical simulation of
concrete pumping process and investigation of wear mechanism of the piping
wall, Tribol. Int. 46 (2011) 137–144, https://doi.org/10.1016/j.
triboint.2011.06.005.
[29] L. Chen, G. Liu, W. Cheng, G. Pan, Pipe flow of pumping wet shotcrete based on
lubrication layer, Springerplus. 5 (2016) 945, https://doi.org/10.1186/s40064-
016-2633-3.
[30] R. Schwarze, CFD-Modellierung: Grundlagen und Anwendungen bei
Strömungsprozessen, Springer-Verlag, Berlin/Heidelberg, 2013. doi: 10.1007/
978-3-642-24378-3.
[31] N.S. Martys, A classical kinetic theory approach to lattice Boltzmann
simulation, Int. J. Mod. Phys. C. 12 (2001) 1169–1177, https://doi.org/
10.1142/S0129183101002474.
[32] Gram, Modelling Bingham Suspensional Flow PhD Thesis, KTH Royal Institute
of Technology, 2015.
[33] W.R. Schowalter, G. Christensen, Toward a rationalization of the slump test for
fresh concrete: comparisons of calculations and experiments, J. Rheol. 42
(1998) 865–870, https://doi.org/10.1122/1.550905.
[34] V.N. Nerella, Experimental and CFD Study on Pumpability of Concrete using
Sliding Pipe Rheometer and ANSYS Fluent Master Thesis, TU Dresden, 2012.
[35] S.H. Kwon, P.J. Kyong, J.H. Kim, P.S. Surendra, State of the Art on Prediction of
Concrete Pumping, Int. J. Concr. Struct. Mater. 10 (2016) 75–85, https://doi.
org/10.1007/s40069-016-0150-y.
[36] C.F. Ferraris, L.E. Brower, D. Beaupré, F. Chapdelaine, P. Domone, E. Koehler, L.
Shen, M. Sonebi, L. Struble, D. Tepke, O. Wallevik, J.E. Wallevik, Comparison of
concrete rheometers : international tests at MB (Cleveland OH, USA), 2004.
[37] E. Secrieru, S. Fataei, C. Schröfl, V. Mechtcherine, Study on concrete
pumpability combining different laboratory tools and linkage to rheology,
Constr. Build. Mater. 144 (2016) 451–461, https://doi.org/10.1016/
j.conbuildmat.2017.03.199.
[38] F. Mahaut, S. Mokéddem, X. Chateau, N. Roussel, G. Ovarlez, Effect of coarse
particle volume fraction on the yield stress and thixotropy of cementitious
materials, Cem. Concr. Res. 38 (2008) 1276–1285, https://doi.org/10.1016/j.
cemconres.2008.06.001.
[39] D. Feys, R. Verhoeven, G. De Schutter, Pipe flow velocity profiles of complex
suspensions, like concrete, 8th Natl. Congr. Theor. Appl. Mech. (NCTAM 2009).
(2009) 66–73. https://biblio.ugent.be/publication/967634.
[40] M.S. Choi, J.K. Young, K.K. Jin, Prediction of Concrete Pumping Using Various
Rheological Models, Int. J. Concr. Struct. Mater. 8 (2014) 269–278, https://doi.
org/10.1007/s40069-014-0084-1.
[41] I.M. Krieger, T.J. Dougherty, A mechanism for non-newtonian flow in
suspensions of rigid spheres, Trans. Soc. Rheol. 3 (1959) 137–152, https://
doi.org/10.1122/1.548848.
[42] R.J. Farris, Prediction of the viscosity of multimodal suspensions from
unimodal viscosity data, J. Rheol. 12 (1968) 281, https://doi.org/10.1122/
1.549109.
[43] R.J. Flatt, P. Bowen, Yodel: a yield stress model for suspensions, J. Am. Ceram.
Soc. 89 (2006) 1244–1256, https://doi.org/10.1111/j.1551-2916.2005.00888.x.
[44] X. Chateau, G. Ovarlez, K.L. Trung, Homogenization approach to the behavior of
suspensions of noncolloidal particles in yield stress fluids, J. Rheol. 52 (2008)
489–506, https://doi.org/10.1122/1.2838254.
[45] F. Mahaut, X. Chateau, P. Coussot, G. Ovarlez, F. Mahaut, X. Chateau, P. Coussot,
G. Ovarlez, Yield stress and elastic modulus of suspensions of noncolloidal
particles in yield stress fluids, J. Rheol. 52 (2008) 287–313, https://doi.org/
10.1122/1.2798234.ions of noncolloidal particles in yield stress fluids, (2008).
doi:10.1122/1.2798234.
[46] C. Hu, F. de Larrard, The rheology of fresh high-performance concrete, Cem.
Concr. Res. 26 (1996) 283–294, https://doi.org/10.1016/0008-8846(95)00213-
8.
[47] H. Hafid, G. Ovarlez, F. Toussaint, P.H. Jezequel, N. Roussel, Effect of particle
morphological parameters on sand grains packing properties and rheology of
model mortars, Cem. Concr. Res. 80 (2016) 44–51, https://doi.org/10.1016/j.
cemconres.2015.11.002.
[48] K. Krenzer, Development of a State-Dependent DEM-Contact-Model to
Simulate Mixing Processes of Fresh Concrete PhD Thesis, TU Dresden, 2016.
[49] D. Feys, G. De Schutter, R. Verhoeven, K.H. Khayat, Similarities and differences
of pumping conventional and self-compacting concrete, in: Des. Prod. Place.
Self-Consolidating Concr., Springer Netherlands, Dordrecht, 2010, pp. 153–
162, https://doi.org/10.1007/978-90-481-9664-7_13.