Water Research 197 (2021) 117085

Contents lists available at ScienceDirect

Water Research
journal homepage: www.elsevier.com/locate/watres

New hydraulic insights into rapid sand filter bed backwashing using
the Carman–Kozeny model
Onno J.I. Kramer a,b,c,d,f,∗, Peter J. de Moel a,c,e, Johan T. Padding b, Eric T. Baars c,
Sam B. Rutten d,g, Awad H.E. Elarbab d, Jos F.M. Hooft c, Edo S. Boek f, Jan Peter van der
Hoek a,c
a
Delft University of Technology, Faculty of Civil Engineering and Geosciences, Department of Water Management, PO Box 5048, 2600 GA, Delft, the
Netherlands
b
Delft University of Technology, Faculty of Mechanical, Maritime and Materials Engineering, Department of Process and Energy, Leeghwaterstraat 39, 2628
CB, Delft, the Netherlands
c
Waternet, PO Box 94370, 1090 GJ, Amsterdam, the Netherlands
d
HU University of Applied Sciences Utrecht, Institute for Life Science and Chemistry, PO Box 12011, 3501 AA Utrecht, the Netherlands
e
Omnisys, Eiberlaan 23, 3871 TG, Hoevelaken, the Netherlands
f
Queen Mary University of London, Division of Chemical Engineering, School of Engineering and Materials Science, Mile End Road, London E1 4NS, United
Kingdom
g
Wetsus, European Centre of Excellence for Sustainable Water Technology, P.O. Box 1113, 8900 CC, Leeuwarden, the Netherlands

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

Article history: Fluid flow through a bed of solid particles is an important process that occurs in full-scale water treat-
Received 30 December 2020 ment operations. The Carman–Kozeny model remains highly popular for estimating the resistance across
Revised 26 February 2021
the bed. It is common practice to use particle shape factors in fixed bed state to match the predicted
Accepted 22 March 2021
drag coefficient with experimentally obtained drag coefficients. In fluidised state, however, where the
Available online 23 March 2021
same particles are considered, this particle shape factor is usually simply omitted from the model with-
Keywords: out providing appropriate reasoning. In this research, it is shown that a shape factor is not a constant
Drinking water treatment particle property but is dependent on the fluid properties as well. This dynamic shape factor for irregu-
Multiphase flows larly shaped grains increases from approximately 0.6 to 1.0 in fluidised state.
Filter-backwash
Hydraulics drag relations We found that unstable packed beds in moderate up-flow conditions are pseudo-fixed and in a setting
Particle orientation state. This results in a decreasing bed voidage and simultaneously in a decreasing drag coefficient, which
Dynamic particle shape factors seems quite contradictory. This can be explained by the collapse of local channels in the bed, leading to
a more uniform flow distribution through the bed and improving the available surface for flow-through.
Our experimental measurements show that the drag coefficient decreases considerably in the laminar
and transition regions. This is most likely caused by particle orientation, realignment and rearrangement
in particles’ packing position.
A thorough hydraulic analysis shows that up-flow filtration in rapid sand filters under backwash condi-
tions causes the particle bed to collapse almost imperceptibly. In addition, an improved expression of
the drag coefficient demonstrated that the Carman–Kozeny model constant, however often assumed to
be constant, is in fact not constant for increasing flow rates. Furthermore, we propose a new pseudo-3D
image analysis for particles with an irregular shape. In this way, we can explain the successful method
using optimisation of the extended terminal sub-fluidisation wash (ETSW) filter backwashing procedure,
in which turbidity and peaks in the number of particles are reduced with a positive effect on water
quality.
© 2021 The Author(s). Published by Elsevier Ltd.
This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/)

1. Introduction

The downward flow of fluid through a bed of solid particles

∗
Corresponding author. is an important process that occurs in full-scale water treatment
E-mail address: [email protected] (O.J.I. Kramer). operations (Crittenden et al., 2012). In the field of drinking wa-

https://doi.org/10.1016/j.watres.2021.117085
0043-1354/© 2021 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/)
O.J.I. Kramer, P.J. de Moel, J.T. Padding et al. Water Research 197 (2021) 117085

water flow expands the filter bed and flushes out the collected
Nomenclature fine particles. After a filtration and backwashing sequence, the cy-
cle starts again (Edzwald, 2011). To reduce the passage of parti-
Ar Archimedes number [-] cles through restarted filters immediately after a backwash proce-
ci Coefficients [-] dure, Amburgey proposed the extended terminal sub-fluidisation
D Inner column or cylinder vessel diameter [m] wash method (ETSW) to remove significantly higher amounts of
d Measured diameter [m] backwash remnant particles (Amburgey, 2005). ETSW is a proven
dp Effective or average or particle equivalent diameter method of terminating the backwash cycle with a sub-fluidisation
[m] wash for a time interval adequate to pass one theoretical filter vol-
ds,i Sieve mesh diameter [m] ume of water upward through the filter (Amburgey and Amirthara-
fL Dimensionless drag coefficient (laminar representa- jah, 2005).
tion) [-] ETSW is a relatively straightforward process comprising an ap-
fL,CK Dimensionless drag coefficient (laminar representa- propriate additional sub fluidisation wash step at the end of a nor-
tion for Carman–Kozeny) [-] mal backwash procedure. The ETSW washing phase is started at
fT Dimensionless drag coefficient (turbulent represen- the end of the backwash sequence with the purpose of rinsing out
tation) [-] the remainder of the particles, sheared from the filter during the
fT,CK Dimensionless drag coefficient (turbulent represen- fluidisation stage of the backwash. The ESTW is set to a wash wa-
tation for Carman–Kozeny) [-] ter flow rate below the minimum fluidisation velocity for a time
g Local gravitational field of earth equivalent to the sufficient to displace one entire filter volume of particle-laden wa-
free-fall acceleration [m/s²] ter. The results presented by Amburgey (2005) show a significant
L Relative total fluid bed height [m] improvement in water quality by reducing, or eliminating, the ini-
L Fluid bed height [m] tial particle passage during the restart of a packed filter after back-
Lm f Bed height at minimum fluidisation [m] washing. ETSW reduces the turbidity during filter maturation and
L0 Fixed bed height [m] particle number spikes in practice, but no hydraulic explanation for
N Total number of particles / total number of experi- its effectiveness was provided.
ments [#] In this paper, we report laboratory measurements in which up-
P/L Pressure drop head loss [kPa/m] flow filtration procedures were simulated. During the experiments,
Pmax Total maximum differential pressure over the bed an unexpected reduction in both bed voidage and bed resistance
[kPa] was detected which could not be explained by traditional drag co-
Qw Water flow [m³/h] efficient correlations as a function of the particle Reynolds number.
Re p Reynolds particle number [-] We propose this can be explained by differences in preferred ori-
Re Modified Reynolds particle number [-] entation of particles under different hydraulic conditions.
Re ,m f Modified Reynolds particle number at minimum flu-
idisation [-]
vs Linear superficial velocity or empty tube fluidisation 2. Hydraulic drag
velocity [m/s]
T Temperature [°C] 2.1. Hydraulic models for fluid flow in particle beds
Vp Volume of an individual particle [m³]
Well-known hydraulic models for flow through particle
Greek symbols
beds have been introduced for the laminar flow regime by
ε Voidage of the system [m³/m³]
Blake (1922) and Kozeny (1927) based on Darcy (1856), for the
ε0 Fixed bed voidage [-]
turbulent flow regime by Burke and Plummer (1928), and for the
εm f Voidage at minimum fluidisation [-]
transitional flow regime by Ergun (1952) and Carman (1937) based
η Dynamic fluid viscosity [kg/m/s]
on the Forchheimer flow conditions (Forchheimer, 1930). In the
υT Kinematic fluid viscosity [m²/s]
laminar regime, the fluid flows in smooth parallel layers without
φs Shape of diameter correction factor [-]
any disturbance between the layers, while in the turbulent flow
ρf Fluid density [kg/m³]
regime, the fluid undergoes irregular fluctuations and changes in
ρp Particle density [kg/m³]
both magnitude and direction. Transitional flow is a mixture of
Abbreviations laminar and turbulent flow, with turbulence in the centre of pore
CFD Computational Fluid Dynamics spaces and laminar flow near the particle surfaces. Each of these
DPI Dots per inch flows behaves in different ways in terms of their frictional energy
ETSW Extended terminal sub-fluidisation wash loss.
FTU Formazin turbidity units De facto, the most commonly used and most accurate
HSV Colour model (hue, saturation, value) in graphic de- (Říha et al., 2018) equation for expressing the relationship be-
sign tween permeability and voidage, particle size and tortuosity is the
RSF Rapid sand filtration Carman–Kozeny equation. This equation is applied in various fields,
WHO World Health Organisation such as groundwater flow, water treatment and many other wa-
ter processes (Camp, 1946). The Carman–Kozeny equation is im-
portant not only for the determination of permeability in porous
ter treatment, rapid sand filtration (RSF) is the most common media (Yang, 2003), but also for the estimation of the voidage in
unit operation applied to capture fine particles from the water a fluidised bed (Kramer et al., 2020a). In groundwater flow, the
in the filter bed. Accumulated particles cause a gradual increase flow regime is generally considered to be laminar (Crittenden et al.,
in head loss and reduce overall effectiveness of the filter bed 2012). In fluidised beds, however, not only laminar, but also tran-
(Howe et al., 2012). When the head loss exceeds a certain thresh- sitional and, to a certain degree, turbulent flow regimes occur. The
old after a period of time, the filter is backwashed, usually as- resistance of laminar flow through a granular bed can be calcu-
sisted by some auxiliary scouring. During backwashing, an upward lated using the well-known equation proposed by Blake (1922),

2
O.J.I. Kramer, P.J. de Moel, J.T. Padding et al. Water Research 197 (2021) 117085

Kozeny (1927) and Carman (1937): In the literature (Wadell, 1933; Cleasby et al., 1981;
Michaelide et al., 2017) particle shape is often characterised
P v s η ( 1 − ε )2
= 180 2 ( Reε < 2 ) (1) by sphericity, which is generally defined as the ratio of the surface
L dp ε3 area of an equal volume sphere to the surface area of the granule.
That being said, sphericity has limited value in actual practice
Where: P/Lis the pressure drop head loss, vs is the superfi-
for several reasons. First, filter media are currently still routinely
cial velocity, η the dynamic fluid viscosity, d p the average or equiv-
measured and specified using a laborious sieve analysis proce-
alent particle diameter, ε the bed voidage and Re the modified
dure, and less by means of advanced image analysis techniques.
particle Reynolds number (Eq. (3)). This equation shows that the
Second, in the literature there is no general consensus on how
resistance of laminar flow is linear with the viscosity and with the
to apply sphericity for commercially available granular media. If
superficial velocity, as is also known to be the case for laminar
the particle has a known geometric shape, sphericity can be cal-
flow through pipes according to Hagen-Poiseuille (Munson et al.,
culated mathematically. However, for irregularly shaped particles,
2020) and laminar flow through aquifers (Darcy, 1856). The exper-
indirect methods are required. To determine shape factors, several
imentally obtained value of 180 is often indicated as the Kozeny
empirical fitting models are available (Dharmarajah, 1982). Often,
drag coefficient (Kozeny, 1927). Further background information
these models are based on head loss or voidage experiments.
about the Carman equation is elucidated in the Supplementary
Carman (1937) proposed that the K-value was independent of
Material (Section 3).
media properties and introduced a correction factor to account for
Since in water treatment the operational field lies in the vicin-
the non-spherical nature of filter grains, using sphericity and the
ity of incipient fluidisation and since turbulent flow regimes are
shape factor (Fair et al., 1933). According to Yang (2003), shape
exceptional, there is a preference for using the Carman–Kozeny
factors are open to criticism, because a range of granules with dif-
drag relation. Kozeny proposed a fixed pore shape factor K = 180
ferent shapes may have the same shape factor, which is inevitable
in Eq. (1) to fit the model results to experimental data. In the lit-
if complex shapes are to be described by only a single parameter.
erature, many other values are proposed (Ozgumus et al., 2014;
Wen and Yu (1966) proposed shape factor corrections, used in the
Erdim et al., 2015; Hoyland, 2017; Schulz et al., 2019), of which
Ergun equation, based on experimental data for the minimum flu-
K = 150 is also often used (Burke and Plummer, 1928; Ergun, 1952;
idisation voidage. According to the literature (Dharmarajah, 1982;
Kramer et al., 2020a). The factor 180 does not appear to be con-
Crittenden et al., 2012), a commonly used particle shape factor in
stant at higher velocities when the flow is no longer laminar (Reε
fixed bed state is necessary in order to match the Carman–Kozeny
> 2), as Carman (1937) showed. He introduced a drag coefficient,
predicted drag coefficient with the experimentally obtained drag
which can be written in the laminar form ( fL ) or in the turbulent
coefficient. The main drawback of this approach is that the model
form ( fT ) as a function of the Reynolds number Reε :
parameters often have no direct connection to the underlying
fL = fT Reε = 180 + 2.9 Reε 0.9 (Reε < 600 ) (2) porous medium structure (Schulz et al., 2019), do not have a phys-
ical interpretation, or are difficult or even impossible to measure,
Where fL is the dimensionless drag coefficient (laminar rep- especially by means of sieve analysis (di Felice, 1995).
resentation) and fT the dimensionless drag coefficient (turbulent In contrast, in the fluidised state, particle shape factors are usu-
representation) and where the modified particle Reynolds number ally omitted to match the Carman–Kozeny predicted drag coeffi-
Reε is defined as: cient with the experimentally obtained drag coefficient, without
ρ f d p vs 1 providing any explanation. Most likely this is done because these
Reε = (3) parameters are difficult to measure directly (de Jong and Nom-
η 1−ε
den, 1974; Hoyland, 2017; Hunce et al., 2018). This is remarkable at
Where ρ f is the fluid density. The experimentally obtained re- the very least since the same particles are involved, in fixed as well
lation for fL or fT will be indicated below as the Carman drag coef- as fluidised states. In addition, particle shape factors as a function
ficient (Carman, 1937). For further background information on the of fluid velocity and temperature in liquid-solid fluidised systems
many forms of this equation, reference is made to the literature are rarely evaluated in the literature. In the practice of filtration
(Erdim et al., 2015) and model derivations given in the Supplemen- (Crittenden et al., 2012) and other forms of flow through granular
tary Material (Section 3). beds, the measured resistance does not appear to correspond with
the calculated resistance. To obtain a fit, an experimentally deter-
2.2. Particle correction shape factors mined shape factor φs is used to modify the measured diameter
(d) into an effective diameter (d p ):
Within hydraulic modelling, particles are often idealised to
d p = φs d (4)
spheres for mathematical convenience. Water treatment-related
particles generally have more complex geometric features, sum- Where d is the measured diameter and φs the diameter shape
marised under the term shape, including the geometric shape, correction factor. For further background information on the man-
the presence of surface disparities such as protuberances and re- ifold empirical adjustments for flow resistance, reference is made
entrant features as well as particle property irregularities. Shape to the literature (Říha et al., 2018).
is therefore difficult to define. Although the literature on parti-
cle shape is extensive (Gauvin and Katta, 1973; Holdich, 2002; 3. Research aims
Seville and Yu, 2016) and a number of shape factors and descrip-
tors have been proposed (Clift et al., 1978; Allen, 1990), there is Based on theoretical factors and the specific ETSW procedure
no universal agreement on how to define particle shape. Therefore, as applied after filter backwashing, it is clear that in order to ob-
there is no agreement on how to correctly measure it, nor does tain a thorough understanding of the hydraulics of up-flow filtra-
the literature provide consistent methods to cope with naturally tion systems applied in full-scale water treatment unit operation,
irregularly shaped particles that can be used in voidage prediction several goals must be met. Initially, a more effective drag coeffi-
models for liquid-solid fluidisation systems. The size and shape of cient must be introduced to be able to visualise changes in drag
the particles has important implications for, for instance, filter de- for various velocities and temperatures. The magnitudes of parti-
sign (Crittenden et al., 2012), but there is no easy way to account cle correction shape factors must be shown in fixed and fluidised
for this. state. To demonstrate the degree of irregularity of considered

3
O.J.I. Kramer, P.J. de Moel, J.T. Padding et al. Water Research 197 (2021) 117085

Table 1
ImageJ morphological properties (Ferreira and Rasband, 2012).

Property Meaning

Aspect ratio The width to height ratio of an elongated particle

Ferret diameter Distance between two parallel tangents on opposite sides of the image of a randomly oriented particle
Circularity Perimeter of sphere of the same projected area divided by the actual projected perimeter of the particle
Solidity The area of a particle divided by its convex hull area; the imaginary convex hull around it

granular material as well as to determine preferred particle orien- rapid sand and multimedia filters. By increasing the flow rate, the
tation, an applicable image analysis method must be developed. In bed height was carefully observed and measured to determine any
addition, a hydraulic explanation must be offered for the effective change in bed height and consequently average voidage. By mea-
ETSW method for future optimisation purposes (water throughput suring the differential pressure, the flow rate of the incipient flu-
and operational time) and to improve water quality. idisation point was detected. By further increasing the flow rate,
a sequence of bed voidages and differential pressures was mea-
4. Materials and methods sured to be able to compose an expansion curve. Based on acquired
bed voidage, differential pressure, superficial fluid velocity, particle
4.1. Particle selection and physical properties size and kinematic fluid viscosity, the drag coefficients were deter-
mined as presented in Section 2.1.
For up-flow filtration and fluidisation experiments, rapid fil- Photographs of particles, technical information about experi-
tration sand grains were selected. The particle density was mental set-up devices and operational procedures can be found in
determined using measured differential pressures and a labo- the Supplementary Material (Sections 1 and 2).
ratory pycnometer. The hydraulic equivalent particle diameter
(Dallavalle, 1948; Crittenden et al., 2012; Davis, 2010) was calcu- 4.3. Dynamic shape factor determination
lated using the sieve diameters, expressed in Eq. (5):
 To investigate the variability of the particle shape factor, the
dp = ds,1 ds,2 (5) ratio of the measured drag coefficient fL and the drag coefficient
Where ds,i is the sieve mesh diameter. The dimensions of the for Carman–Kozeny fL,CK , using Eq. (2), was calculated. We hy-
particles were also determined by using ImageJ as a particle anal- pothesised that this ratio would depend on the hydraulic state, i.e.
yser (Ferreira and Rasband, 2012), for which a 4.188 g sample of on the ratio Reε /Reε,m f of the actual modified particle Reynolds
filter sand was scanned on an A4 flatbed scanner, at a resolution of number (Eq. (3)) and that for incipient fluidisation. Inverting the
2400 dpi. Irregularly shaped particles, in particular elongated parts, Carman–Kozeny equation, for the measured voidage, velocity and
tend to lie horizontally oriented. ImageJ determines for each indi- viscosity, it was possible to estimate the average particle size. Ac-
vidual particle which pixel in the 2D-image belongs to this particle. cordingly, a dynamic shape factor could be determined for various
For each particle, the software determines some 34 size parame- flow rates using a numerical solving method based on Bolzano’s
ters, such as equivalent circle diameter, and equivalent ellipse di- numerical intermediate value theorem. This is a straightforward
mensions with related morphological particle properties (Table 1), solving method described in: (Apostol, 1967). A simplified equation
such as aspect ratio, Ferret diameter, circularity and solidity. Sta- states:
tistical information such as mean, range and frequency distribution

fL,CK
can be determined from all these measurements. φs = (6)
fL
The total mass and measured density were used to calculate
the total volume of the particles. Assuming an ellipsoid based on Where fL,CK is the dimensionless drag coefficient (laminar rep-
length and width of the average of the 2D dimensions gives a resentation for Carman–Kozeny).
pseudo-3D height as average value for all particles.
5. Results and discussion
4.2. Hydraulic experimental set-up
5.1. Particle properties
Filtration and expansion experiments were carried out at three
locations: in Waternet’s Weesperkarspel drinking water pilot plant Samples of RSF granules were obtained from a full-scale RSF
located in Amsterdam, the Netherlands; at the University of Ap- unit operation located at the Waternet facility in Loenderveen, the
plied Sciences Utrecht, the Netherlands; and at Queen Mary Uni- Netherlands. The applied filter sand range (see Fig. 2) was ob-
versity of London, United Kingdom. In all experiments, locally pro- tained by using sieves of 0.80 and 1.25 mm. This was confirmed
duced drinking water was used. The set-up (Fig. 1) consisted of a by sieving with 10 sieves between 0.6–2.8 mm (NEN-EN 933–2,
4 m high transparent PVC pipe with an inner diameter of 57 mm. n.d.), showing that 10 wt-% of the particles were smaller and 1 wt-
Water temperature was regulated with a boiler, a cooler and a % were larger than this range, which corresponds to a respective
thermostat by recirculating water through a buffer vessel con- 20% and 1% in terms of the number of particles. We measured
nected to a water reservoir. An overflow at the top of the reac- an average particle diameter of d50 = 0.90 mm based on particle
tor returned water to the buffer vessel. From the buffer vessel, wa- counting.
ter was pumped through the reservoir connected to the thermostat The full-colour scan obtained from the A4 flatbed scanner was
which was set to a programmed water temperature. initially converted into a single-colour picture for maximum parti-
Through filtration and expansion experiments, bed voidage and cle distinction from which a binary (black and white) picture was
pressure drop were measured as a function of various linear fluid created with a pixel size of around 0.01 mm (2400 dpi). The size
flow rates and temperatures. Hydraulic experiments were started of each (black) particle was measured through pixel counting. Ob-
in fixed bed state, obtained after an initial backwash with full served particles with their particle numbers can be shown as out-
fluidisation, which resembled the typical practical conditions of lines on top of the original scan (see Fig. 3). The 4.188 g sam-

4
O.J.I. Kramer, P.J. de Moel, J.T. Padding et al. Water Research 197 (2021) 117085

Fig. 1. Schematic overview of experimental set-up in all experimental locations.

Fig. 2. Rapid sand filter granules 0.8 < dp < 1.25 mm.

ple contained 2040 particles, each analysed by its size, shape and culation of an equivalent ellipse with the same area for each parti-
other morphological properties in ImageJ. cle, giving two independent diameters d50 of 1.57 and 1.21 mm on
Figs. 4 and 5 show the measured distribution and cumulative average, with an aspect ratio of 1.30. Assuming that an ellipsoid
frequency, respectively, of observed equivalent particle diameter. with these two axes has the same volume as the average parti-
The green areas indicate the specification of the filter material cle gives the third diameter of 0.78 mm. This dimension explains
based on information from the supplier of the two sieves (bot- the sieve passage as particles are passing sieves in an upright ori-
tom sieve and top sieve). The projected area per particle, deter- entation, contrary to the scanned particles lying in a horizontal
mined with ImageJ, was used to obtain the equivalent diameter of orientation.
a circle resembling the basics of square sieve grids, as illustrated Fig. 5 illustrates how smaller particles are obtained which did
in Fig. 4. Giving the exact contours per particle allows for the cal- not pass through the lowest sieve with the smallest mesh size,

5
O.J.I. Kramer, P.J. de Moel, J.T. Padding et al. Water Research 197 (2021) 117085

Fig. 3. Scanned sample with particle analysis with ImageJ, with particle outlines and particle numbers, with 2 × 2 mm gridlines.

Table 2
Particle properties rapid filtration sand grains.

Variable RSF grains Unit

Particle sieve sizes 0.80–1.25 [mm]

Particle density 2638 ± 11 [kg/m3 ]
2)
Equivalent circle diameter 2D (average) 1.38 [mm]
Uniformity 2D (d60 /d10 ) 2) 1.26 [-]
Circularity 2D (average) 2) 0.73 [mm]
Ellipsoid 3D-d50 2) 1.57 • 1.21 • 0.78 [mm]
Ellipsoid aspect ratios 3D - l/b-b/h 2) 1.30 - 1.55 [-]
Geldarts type 3) D
1
Measured particle size distributions are given in the Supplementary Material (Sec-
tion 4).
2
ImageJ (Ferreira and Rasband, 2012).
3
Geldarts particle classification (Geldart, 1973) type D: spoutable particles.

probably as a result of short sieving time. In contrast, hardly any

Fig. 4. Equivalent diameter (assumed to be a circle): histogram of circle diameter larger particles were found since they did not pass through the top
from particle areas of ImageJ (in blue), compared to filter sand specifications (green sieve with the largest mesh size. Fig. 5 also shows that the parti-
area). The average diameters based on particle counting (d50 ) are presented as ver-
cles are not perfect spheres, as a sieve curve would suggest, but
tical lines in green (sieve experiment) and blue (ImageJ), respectively. (For interpre-
tation of the references to colour in this figure legend, the reader is referred to the
that they are much better described as an ellipsoid with three dis-
web version of this article.) tinct dimensions: width, height and length.
In sum, it can be concluded that the analysed RSF grains are
moderately irregularly shaped, and far from spherical. In the next
section, it will become clear why this information is important in
relation to decreasing bed resistance for increasing flow rates. Par-
ticle analysis using flatbed scans with pseudo-3D image processing
is superior to 1D sieve analyses. The determined particle properties
of RSF grains are presented in Table 2.
In industrial practice, samples of granules are still commonly
sieved to obtain information about the particle size distribution for
process monitoring. With this method, the morphological proper-
ties of the particles are lost. However, as irregularly shaped par-
ticles, and rod-like particles in particular, can realign and pass to
lower sieves, this gives a distorted view of the actual particle sizes.
The easily acquired particle information, i.e. average particle size,
is entered accordingly into the drag coefficient or voidage predic-
tion model. In practice, this classical method works well for spher-
ical particles, but it is less accurate for irregularly shaped parti-
cles such as RSF grains (Fig. 2). Irregularly shaped particles ex-
perience anisotropic drag in fixed and fluidised state, due to the
surrounding fluid and interactions with adjacent particles and the
Fig. 5. Cumulative frequency grain sizes.
wall (Mahajan et al., 2018).

6
O.J.I. Kramer, P.J. de Moel, J.T. Padding et al. Water Research 197 (2021) 117085

Fig. 6. Up-flow filtration and expansion curve for rapid filter sand granules 0.8 < Fig. 8. Drag (turbulent representation) fT against modified Reynolds number for
dp < 1.25 mm. Experimental voidage (circles), Carman–Kozeny predicted voidage rapid filter sand granules 0.8 < dp < 1.25 mm.
(•••) and differential pressure (triangles) against superficial fluid velocity, (- - -)
maximum pressure drop, (—) voidage at minimum fluidisation.

Fig. 9. Drag (laminar representation) fL against modified Reynolds number for rapid
filter sand granules 0.8 < dp < 1.25 mm.
Fig. 7. Fluidisation characteristics of rapid filter sand granules 0.8 < dp < 1.25 mm.
Experimental voidage (circles) and Carman–Kozeny predicted voidage (•••) against
superficial fluid velocity for six different water temperatures.
Experimental data tabulation (Section 9), minimum fluidisation
points (Section 5) and model predictions (Section 3) are given in
5.2. Hydraulic experiments and expansion curves the Supplementary Material.

The hydraulic experiments we conducted provided crucial in- 5.3. Measured and predicted drag coefficients
formation about differential pressure and bed height in fixed, in-
cipient and fluidised states, as a function of linear flow rates and The drag coefficient against the modified Reynolds number
water temperatures. In total, eight up-flow experiments were con- Reε for RSF grains is presented in turbulent representation f T in
ducted for RSF grains at six different water temperatures (6–29 °C). Fig. 8 and in laminar representation fL in Fig. 9. In both fig-
Additional expansion experiments were performed at seven differ- ures, the Kozeny and Carman (Eq. (2)) drag relations are plotted.
ent water temperatures (5–36 °C) to determine the minimum or Fig. 8 shows a clear deviation between the measured and pre-
incipient fluidisation points, causing the first particles to expand dicted values for low Reynolds numbers (Reε < 10). In most other
as well as allowing the precise expansion behaviour of RSF grains literature references (Burke and Plummer, 1928; Carman, 1937;
to be determined. For up-flow filtration, the fluid flow was grad- Ergun, 1952; Gibilaro, 2001; Bird et al., 2007; Hoyland, 2017), such
ually increased until the incipient fluidisation point was reached, deviations between measurements and drag models are artificially
where the differential pressure levelled at a maximum value. For hidden by the use of logarithmic scales over several orders of mag-
moderate fluidised state, the fluid flow was slightly further in- nitude.
creased under constant maximum differential pressure. The ex- For relatively high Reynolds numbers (Reε > 10), the predicted
pansion curve for RSF grains is presented in Fig. 6 The measured Carman values coincide relatively well with the measured values.
voidage in fixed bed state varies between ε0 = 0.44 ± 0.03. The For low Reynolds numbers (especially Reε < 5), the measured fL
voidage at the minimum fluidisation velocity is εm f = 0.45 ± 0.01. values are significantly larger than the theoretically expected f L
Fig. 6 shows the expansion behaviour of RSF grains in the fluidised =180, corresponding to the Kozeny pore shape factor K = 180. The
state. The measured bed voidages at various water temperatures laminar representation of fL makes changes in drag much more
agree reasonably well with the predicted values according to the concise compared to the commonly used turbulent representation
Carman–Kozeny model. The influence of the water temperature in of fT . Fig. 9 clearly shows that the drag decreases considerably in
the fluidised state can be seen in Fig. 7. The temperature effect in case the fluid flow increases slightly, under laminar or transitional
the fixed bed state was less clearly observable (Fig. 6). A moder- conditions. This is happening while the particles are still assumed
ate linear relationship (R2 = 0.99) between the degree of subsi- to be packed in the fixed bed state.
dence% of the packed bed and water temperature (5–35 °C) was More information about the Carman model can be found in the
found: ε /ε = −0.0 0 01T + 0.0 08. The temperature dependence of Supplementary Material (Sections 3).
the bed setting is caused by the changes in viscosity of water, To show explicitly that the drag decreases mainly in the (sup-
which affects the frictional forces acting on the filter bed. posedly) fixed bed, the ratio of measured to predicted drag f L / fL,CK

7
O.J.I. Kramer, P.J. de Moel, J.T. Padding et al. Water Research 197 (2021) 117085

Fig. 10. Drag (laminar representation) relative to the Carman–Kozeny drag against
modified Reynolds numbers relative to modified Reynolds numbers at minimum
fluidisation for rapid filter sand granules 0.8 < dp < 1.25 mm.

Fig. 12. Initial fixed bed.

Fig. 11. Shape factor against modified Reynolds numbers relative to modified
Reynolds numbers at minimum fluidisation for rapid filter sand granules 0.8 < dp
< 1.25 mm.

is plotted against the Reynolds number ratio Reε /Reε,m f in Fig. 10.
For increasing flow rates, where Reε /Reε,m f is still well below 1,
the ratio fL / fL,CK decreases rapidly. For every percent that the bed
voidage drops, the drag coefficient fL decreases by a factor 4. The
large ratio of measured to predicted drag occurs at Reε /Reε,m f <
0.5, which indicates that it cannot be explained by effects occur-
ring around vm f . This will be investigated further in the next sub-
section.
A practical approach is to use a particle shape factor to com-
pensate for observed deviations in fL . The use of a shape factor,
however, becomes redundant in the fluidised state, since the pre-
diction of the voidage is increasingly consistent with the models
for (Reε > 10) in Figs. 9 and 10. We reiterate that the common Fig. 13. Fixed bed setting after t=5 [min] ETSW.
use of two different shape factors in the literature for fixed beds
(Ergun, 1952; Yang, 2003; Li et al., 2018) and fluidised beds (di Fe-
lice, 1995; Gibilaro, 2001; Rhodes, 2008) is remarkable, because it rangement in their packing position. During the pilot plant exper-
concerns the same particles. Using Eq. (6), shape factors were cal- iments, the observed bed height decreased slightly (approximately
culated for RSF grains. Fig. 11 shows that in the fixed bed state, the 1%), which can be seen in Fig. 12, Figs. 13 and 14. Videos of RSF
particle shape factor varies between approximately 0.6 < φs < 0.9. grains (Kramer and van Schaik, 2020) as well as calcite pellets
In the fluidised state, however, the shape factor slowly approaches (Kramer et al., 2020b), at flow rates well below the minimum flu-
φs = 1. idisation points, show clearly that particles already start to move
locally while overall they remain in a fixed bed state. The collapse
5.4. Hydraulic explanation for up-flow filtration and ETSW of the packed bed caused by ETSW is also demonstrated in four
videos.
When a fluid flows in upward direction through a particle bed The collapse of the particles in the fixed beds below the incip-
in fixed state, the fluid exerts a certain drag on the particles. ient fluidisation point is most likely caused by particle orientation,
The transition from a fixed bed to a fluidised bed occurs when realignment and rearrangement in their packing position. An un-
the pressure drop across a packed bed reaches a constant max- steady packed bed in moderate up-flow conditions is pseudo-fixed
imum value, after the fluid flow exceeds the minimum fluidisa- and in a setting state. This results in a decreasing bed voidage and
tion velocity. The particles start to elevate, and irregularly shaped simultaneously in a decreasing drag coefficient, which seems quite
particles also show changing orientation, realignment and rear- contradictory. However, this can be explained by the collapse of

8
O.J.I. Kramer, P.J. de Moel, J.T. Padding et al. Water Research 197 (2021) 117085

tude of turbidity and particle count spikes decrease. Therefore, the

successful optimisation ETSW filter backwashing procedure pro-
posed by Amburgey (2005) can be explained hydraulically.

5.5. Preliminary considerations of hydraulics-based ETSW

According to Amburgey (2005), adjustments in fluid velocity

should be limited to the minimal fluidisation velocity. Since this
velocity depends on both the filter bed material and on water
temperature, optimising the up-flow velocity could become com-
plicated due to different packing materials and seasonal tempera-
ture fluctuations. Smaller RSF grains fluidise more easily compared
to the larger RSF grains, resulting in a gradual transition. In other
words, because of differences in particle size, the minimum fluidis-
ation velocity cannot be determined exactly. In the event that dur-
ing ETSW velocities occur that are too high, the benefits of bed set-
ting, a more uniform flow distribution throughout the bed and bet-
ter available surface for through-flow are nullified. Therefore, when
optimising the ETSW time, it may be more important to focus on
effective bed subsidence instead of focussing on incipient fluidi-
sation. Amburgey (2005) assumed that at least one bed volume
would be required to wash out remaining particles. As we provide
a hydraulic explanation for improved water quality, this assump-
Fig. 14. Fixed bed setting after t=30 [min] ETSW.
tion does not need to apply. Therefore, considering this hydraulic
explanation, the ETSW time could potentially be shortened to less
local channels (Figs. 15b and c) in the bed, leading to a more uni- than one bed volume. This would decrease down-time due to back-
form flow distribution through the bed and improving the avail- washing and increase up-time (filtration time) as breakthrough at
able surface for flow-through (Hassett, 1961). This is the reason restart will be significantly reduced, or even eliminated. With re-
why ETSW, as an additional method to a backwash filter recipe, gard to water quality and costs, the elimination of breakthrough at
works effectively (Fig. 15), since it levels the voidage in the par- restart will improve the overall bulk quality of the produced water,
ticle bed. In this way, the successful ETSW method proposed by increase the total volume produced in between backwashes and
Amburgey (2005) can be explained. subsequently save overall production costs. In practice, the first
During full-scale operational ETSW procedures, small (1– half hour of filtration, after backwashing, is discarded as this con-
10 mm) spouting channels were observed: this is shown in Fig. tains the highest concentrations of remaining particulates and con-
15(b) and Fig. 15(c). The amount and intensity of spouts decreased tamination.
during the procedure. A plausible explanation is that the up-flow
filtration caused the channels to collapse in the bed, leading to a 5.6. Full-scale consequences of utilising the ETSW procedure
more uniform flow distribution through the bed and improving the
available surface for flow-through. We hypothesise that this phe- Turbidity is a measure of the amount of particulate matter in
nomenon has a positive effect on water quality, since the magni- water and is one of the most widely used parameters for mea-

Fig. 15. (a) Full-scale RSF during collapse pulsing 40 m/h backwashing scour; (b) Full-scale RSF during ETSW 5 m/h up-flow; (c) Full-scale RSF during ETSW bed, spouting
channels.

9
O.J.I. Kramer, P.J. de Moel, J.T. Padding et al. Water Research 197 (2021) 117085

6. Conclusions

Our experiments showed that irregularly shaped particles such

as RSF grains in fixed beds under moderate up-flow conditions
show non-monotonous hydraulic behaviour related to changing
orientation, realignment and rearrangement in particle packing po-
sition. This happens in such a manner that the bed offers the
maximum flow passage, i.e. voidage to the fluid. Initially, the bed
voidage, the flow-through interface available in the bed, as well
the drag decrease and accordingly increase for larger fluid flows.
These phenomena also affect the volume-based voidage prediction,
something that to the best of our knowledge has not been eluci-
dated in the literature.
In the field of engineering, a shape factor is applied to compen-
sate for the above aspect in the fixed and incipient fluidisation but
Fig. 16. Turbidity measurements of four full-scale RSF first filtrates in Loenderveen is omitted when the fluidised state is described. In this research,
before the ETSW procedure was established. it was shown that such a shape factor is not a constant particle
property but is dependent on the fluid properties as well. This dy-
namic shape factor for irregularly shaped RSF grains increases from
approximately 0.6 until 1.0 in fairly fluidised state. The use of con-
stant shape factors regarding natural, irregularly shaped granules is
therefore not recommended. Our new straightforward image anal-
ysis method, based on a weighted particle sample to estimate the
mean measures of spheroids, shows that sieve measures are not
representative of natural, irregularly shaped granules. Based on ex-
perimental data obtained in this work, it is shown that, for in-
creasing fluid flow rates, the fixed bed voidage decreases slightly
(approximately 1%) while the drag coefficients decline to a con-
siderable degree. The drag coefficient reached values up to 450
(in laminar representation), which is considerably larger than the
well-known Kozeny value of 180.
The reduction in bed voidage as well as bed resistance is due
to the realignment of the particles in the packed bed and the col-
lapse of flow channels created during a normal backwash proce-
dure. Particle orientation, realignment and rearrangement implies
Fig. 17. Turbidity measurements of four full-scale RSF first filtrates in Loenderveen
after the ETSW procedure was established.
unstable fixed beds that become denser in laminar up-flow condi-
tions. This results in a decreasing bed voidage and simultaneously
in a decreasing drag coefficient. A complete fluidised state, i.e. ex-
ceeding the minimum fluidisation velocity, is not necessary. In ad-
dition, besides the particle size, also the particle shape is a deci-
suring water quality prior to disinfection (Edzwald, 2011). High sive factor in the phenomena occurring in the particle beds. The
turbidity levels can reduce disinfection capacity. Adequate turbid- collapse of local channels in the bed leads to a more uniform flow
ity removal implies the partial removal of pathogens in water distribution through the bed and improves the available surface for
(Lechevallier et al., 1991), since most pathogens tend to aggregate flow-through. In this way, the successful ETSW method proposed
with particles. Before 2016, at the full-scale RSF in Loenderveen by Amburgey (2005) can be explained.
(the Netherlands), considerable turbidity spikes were detected in
the first filtrate after backwash cycles. After a backwash cycle had 7. Recommendations
been completed and the filter returned to operational mode, the
turbidity of the first filtrate was measured in four filters (Fig. 16). With the knowledge acquired during these experiments, the
After approximately 30 min, the turbidity increased from 0.2 FTU ETSW procedure can, in future research, be further optimised by
up to 1–3 FTU. A similar elevated pattern could be seen for bacte- adjustments to the up-flow velocity and time.
rial levels Enterococci (N =4/10 L), E.coli (N =14/10 L) and traces of During our experiments, it was possible to capture the almost
sulphite-reducing Clostridia (N=103 /10 L). The WHO (World Health unnoticeable changes in the particle bed on video. In a full-scale
Organization, 2017) promotes a turbidity threshold of below 0.2 installation, real-time monitoring of the particle collapse in the
NTU. rapid sand filter bed is practically impossible. Thorough pilot plant
Accordingly, an optimisation process has been started at the research is required to fully understand the specific processes of a
full-scale facility in Loenderveen to improve water quality by re- particular water treatment. Every water treatment plant has a dif-
ducing bacterial breakthroughs. The goal was to decrease the con- ferent context i.e. water quality (load of particulate material), bed
centrations of bacteria to improve the oxidation and disinfection composition, different process strategies and different water tem-
efficiency in the ozonation unit operation down-stream (van der peratures.
Helm, 2007). The ETSW procedure was initialised on the full-scale Since the configuration of the particles in the packed filter bed
facilities additional to the standard backwash procedure. Further- is generated by frictional forces, it is recommended to investigate
more, we investigated whether it was desirable to remove the first alternative methods that include frictional forces, so that subsi-
filtrate for approximately three hours. After the ETSW procedure dence can take place. Although filter resting could also be a so-
was utilised, turbidity spikes in the first filtrates were excluded lution to acquire bed setting and prevent the first filtrate turbidity
(Fig. 17) in favour of water quality. spike after the backwash expansion, it reduces the operational up-

10
O.J.I. Kramer, P.J. de Moel, J.T. Padding et al. Water Research 197 (2021) 117085

time. In order to extend the hydraulic impact discussed here to an Davis, M., 2010. Water and Wastewater Engineering–Design principles and Practice,
even more effective ETSW, a model to predict the permeability and 1st ed. McGraw-Hill, New-York.
de Jong, J.A.H, Nomden, J.F., 1974. Homogeneous gas-solid fluidization. Powder Tech-
bed setting of the filter bed needs to be further substantiated and nol. 9, 91–97.
developed to be able to optimise the ETSW procedure. The newly Dharmarajah, A.H., 1982. Effect of Particle Shape on Prediction of Velocity-Voidage
proposed pseudo-3D ImageJ analysis method can be used for fu- Relationship in Fluidized Solid-Liquid Systems. Ames.
di Felice, R., 1995. Review article number 47: of hydrodynamics of liquid fluidisa-
ture CFD modelling research, taking into account the influence of tion. Chem. Eng. Sci. 50, 1213–1245. doi:10.1016/0 0 09- 2509(95)98838- 6.
non-spherical particles, in which the obtained measures of the el- Edzwald, J.K., 2011. Water quality and treatment: a handbook on drinking water.
lipsoids can be put to good use. American Water Works Association, 6th ed. American Society of Civil Engineers,
McGraw-Hill, New York.
Erdim, E., Akgiray, Ö., Demir, I., 2015. A revisit of pressure drop-flow rate correla-
Declaration of Competing Interest tions for packed beds of spheres. Powder Technol. 283, 488–504. doi:10.1016/j.
powtec.2015.06.017.
Ergun, S., 1952. Fluid flow through packed columns. Chem. Eng. Sci. 48, 89–94.
The authors declare that they have no known conflicts of in- Fair, G.M., Hatch, L.P., Hudson, H.E., 1933. Fundamental factors governing the
terests or personal relationships that could have appeared to influ- streamline flow of water through sand. Jo (Amo Water Works Assoco) 25, 1551–
ence the work reported in this article. 1565. doi:10.1002/j.1551-8833.1933.tb18342.x.
Ferreira, T., Rasband, W., 2012. ImageJ user guide, ImageJ user guide IJ 1.46r. doi:10.
1038/nmeth.2019.
Acknowledgements Forchheimer, P., 1930. Hydraulik, 3rd ed. B.G. Teubner, Leipzig.
Gauvin, W.H., Katta, S., 1973. Momentum transfer through packed beds of various
particles in the turbulent flow regime. AIChE J. 19, 775–783. doi:10.1002/aic.
This research is part of the project “Hydraulic modelling of 690190415.
liquid-solid fluidisation in drinking water treatment processes” car- Geldart, D., 1973. Types of gas fluidization. Powder Technol. 7, 285–292. doi:10.1016/
ried out by Waternet (the water utility of Amsterdam and sur- 0 032-5910(73)80 037-3.
Gibilaro, L.G., 2001. Fluidization-Dynamics, the Formulation and Applications of a
roundings), Delft University of Technology and HU University of Predictive Theory for the Fluidized State. Butterworth-Heinemann, Oxford.
Applied Sciences Utrecht. Financial support came from Waternet’s Hassett, N.J., 1961. The mechanics of fluidization. Br. Chem. Eng. 19, 777–
Drinking Water Production Department. 780.
Holdich, R.G., 2002. Fundamentals of Particle technology, Fundamentals of Parti-
We acknowledge and thank our students from Delft Univer- cle Technology. Midland Information Technology and Publishing, Leicestershire
sity of Technology, HU University of Applied Sciences Utrecht and doi:10.4236/ns.2010.210132.
Queen Mary University of London and in particular Victor Shao and Howe, K.J., Hand, D.W., Crittenden, J.C., Rhodes Trussell, R., Tchobanoglous, G., 2012.
Principles of Water Treatment, 1st ed. John Wiley & Sons, Inc., New Jersey.
Cas van Schaik for the precise execution of many laboratory and Hoyland, G., 2017. General combined model for the hydrodynamic behaviour of
pilot plant experiments. fixed and fluidised granular beds. Water Res. 111, 163–176. doi:10.1016/j.watres.
This research project did not receive any specific grant from 2017.01.008.
Hunce, S.Y., Soyer, E., Akgiray, Ö., 2018. On the backwash expansion of graded filter
funding agencies in the public, commercial or not-for-profit sec-
media. Powder Technol. 333, 262–268. doi:10.1016/j.powtec.2018.04.032.
tors. Kozeny, J., 1927. Über kapillare leitung des wassers im boden. Akad. Wiss.Wien 136,
271–306.
Kramer, O.J.I., Padding, J.T., van Vugt, W.H., de Moel, P.J., Baars, E.T., Boek, E.S., van
Supplementary materials der Hoek, J.P., 2020a. Improvement of voidage prediction in liquid-solid flu-
idized beds by inclusion of the Froude number in effective drag relations. Int. J.
Supplementary material associated with this article can be Multiphase Flow 127. doi:10.1016/j.ijmultiphaseflow.2020.103261.
Kramer, O.J.I., van Schaik, C., 2020. Videos of Up-Flow Filtration Velocities for Rapid
found, in the online version, at doi:10.1016/j.watres.2021.117085. Sand Filter Grains 0.8-1.25Mm in Water [Data set] 4TU.Centre for Research Data.
4TU.Centre for Research Data, The Netherlands doi:10.4121/13337057.v1.
References Kramer, O.J.I., van Schaik, C., Nijssen, T.M.J., 2020b. Videos of Fluidisation of Calcite-
Pellets 0.8-0.9mm and 1.4-1.7Mm in Water for Various Flow Rates [Data set]
Allen, T., 1990. Particle Size Measurement–Powder Technology Series, 4th ed. Chap- 4TU.Centre for Research Data. 4TU.ResearchData, The Netherlands doi:10.4121/
man and Hall, London doi:10.1007/978- 94- 009- 0417- 0. 13277246.v1.
Amburgey, J.E., 2005. Optimization of the extended terminal subfluidization wash Lechevallier, M.W., Norton, W.D., Lee, R.G., 1991. Occurrence of giardia and cryp-
(ETSW) filter backwashing procedure. Water Res. 39, 314–330. doi:10.1016/j. tosporidium spp. Surf. Water Suppl.. Appl. Environ. Microbiol. 57, 2610–2616.
watres.2004.09.020. doi:10.1128/AEM.57.9.2610-2616.1991.
Amburgey, J.E., Amirtharajah, A., 2005. Practical and theoretical guidelines for im- Li, X., Liu, M., Li, Y., 2018. Hydrodynamic behavior of liquid–solid micro-fluidized
plementing the extended terminal subfluidization wash (ETSW) backwashing beds determined from bed expansion. Particuology 38, 103–112. doi:10.1016/j.
procedure. J. Water Supp.: Res. Technol.-Aqua 54, 329–337. doi:10.2166/aqua. partic.2017.08.002.
20 05.0 031. Mahajan, V.v., Nijssen, T.M.J., Kuipers, J.A.M., Padding, J.T., 2018. Non-spherical parti-
Apostol, T.M., 1967. One-Variable Calculus, with an Introduction to Linear Algebra, cles in a pseudo-2D fluidised bed: modelling study. Chem. Eng. Sci. 192, 1105–
2nd ed. John Wiley & Sons, Inc., New York. 1123. doi:10.1016/j.ces.2018.08.041.
Bird, R.B., Stewart, W.E., Lightfoot, E.N., 2007. Transport Phenomena, Revised, 2nd Michaelide, E., Crowe, C.T., Schwarzkopf, J.D., 2017. Multiphase Flow Handbook, 2nd
ed. John Wiley & Sons, Inc. International, New-York. ed. Taylor & Francis Inc, London.
Blake, F.C., 1922. The resistance of packing to fluid flow. Trans.Am. Inst. Chem. Eng. Munson, B.R., Rothmayer, A.P., Okiishi, T.H., Huebsch, W.W., 2020. Fundamentals of
14, 415–421. Fluid Mechanics, 7th ed. John Wiley & Sons, New York.
Burke, S.P., Plummer, W.B., 1928. Suspension of macroscopic particles in a turbulent NEN-EN 933-2, 2020. Tests for Geometrical Properties of Aggregates - Part 2: Deter-
gas stream. Ind. Eng. Chem. 20, 1200–1204. doi:10.1021/ie50227a026. mination of Particle Size Distribution - Test sieves, Nominal Sizes of Apertures,
Camp, T.R., 1946. Sedimentation and the design of settling tanks. Trans. Am. Soc. n.d.. Dutch norm ICS-code 91.100.15, 93.080.20.
Civil Eng. 111, 895–936. Ozgumus, T., Mobedi, M., Ozkol, U., 2014. Determination of Kozeny constant based
Carman, P.C., 1937. Fluid flow through granular beds. Trans., Inst. Chem. Eng. 15, on porosity and pore to throat size ratio in porous medium with rectangular
32–48. doi:10.1016/S0263-8762(97)80 0 03-2. rods. Eng. Appl. Comput. Fluid Mech. 8, 308–318. doi:10.1080/19942060.2014.
Cleasby, J.L., Asce, M., Fan, K.S., 1981. Predicting fluidization and expansion of filter 11015516.
media. J. Environ. Eng. Division 107, 455–471. Rhodes, M., 2008. Introduction to particle technology. Chemical Engineering
Clift, R., Grace, J.R., Weber, M.E., 1978. Bubbles, Drops, and Particles. Academic and Processing, 2nd ed John Wiley & Sons Ltd., Chichester doi:10.1002/
Press., San Diego, San Diego, C.A. 9780470727102.
Crittenden, J.C., Trussell, R.R., Hand, D.W., Howe, K.J., Tchobanoglous, G., 2012. Říha, J., Petrula, L., Hala, M., Alhasan, Z., 2018. Assessment of empirical formulae for
MWH’s Water Treatment: Principles and Design, 3rd ed. John Wiley & Sons, determining the hydraulic conductivity of glass beads. J. Hydrol. Hydromech. 66,
New York. 337–347. doi:10.2478/johh- 2018- 0021.
Dallavalle, J.M., 1948. Micromeritics–the Technology of Fine Particles, 2nd ed. Pit- Schulz, R., Ray, N., Zech, S., Rupp, A., Knabner, P., 2019. Beyond Kozeny–Carman:
man Publishing Ltd, London. predicting the permeability in porous media. Transp. Porous Media 130, 487–
Darcy, H., 1856. Les fontaines publiques de la ville de Dijon: Exposition et appli- 512. doi:10.1007/s11242- 019- 01321- y.
cation des principes a suivre et des formules a employer dans les questions Seville, J.P.K., Yu, C.Y., 2016. Particle Technology and engineering, an Engineer’s
de distribution d’eau; ouvrage terminé par un appendice relatif aux fournitures Guide to Particles and powders: Fundamentals and Computational Approaches,
d’eau de plusieurs villes au filtr. Victor Dalmont. Libraire Des Corps imperiaux 1st ed. Butterworth-Heinemann.
Des Ponts Et Chaussées Et Des Mines. Dijon.

11
O.J.I. Kramer, P.J. de Moel, J.T. Padding et al. Water Research 197 (2021) 117085

van der Helm, A.W.C., 2007. Integrated Modeling of Ozonation for Optimization of World Health Organization, 2017. Guidelines for Drinking-Water Quality, 4th ed.
Drinking Water Treatment. Water Management Academic Press, Delft. World Health Organization.
Wadell, H., 1933. Sphericity and roundness of rock particles. J. Geol. 41, 310–331. Yang, W.C., 2003. Handbook of Fluidization and Fluid-Particle Systems, 1st
doi:10.1086/624040. ed. Chemical Engineering. CRC Press, New-York doi:10.1016/S1672-2515(07)
Wen, C.Y., Yu, Y.H., 1966. A generalized method for predicting the minimum 60126-2.
fluidization velocity. Am. Institu. Chem. Eng. J. 12, 610–612. doi:10.1002/aic.
690120343.

12