See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/269873174
Statistics of Flow and the Scaling of Ceramic Water Filters
Article  in  Journal of Environmental Engineering · July 2014
DOI: 10.4028/www.scientific.net/AMR.1132.267
CITATIONS READS
0 165
5 authors, including:
Ebenezer Annan Kabiru Mustapha
University of Ghana Kwara State University
20 PUBLICATIONS   43 CITATIONS    11 PUBLICATIONS   20 CITATIONS   
SEE PROFILE SEE PROFILE
Karen Malatesta W. O. Soboyejo
Princeton University Worcester Polytechnic Institute
29 PUBLICATIONS   376 CITATIONS    494 PUBLICATIONS   6,233 CITATIONS   
SEE PROFILE SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Development of metals/metal oxide nanocrystals conducting polymer blends nanocomposites for photovoltaic and optoelctronic applications. View project
Statistical Analysis of Strength and Fracture toughness of Electro-porcelians View project
All content following this page was uploaded by Ebenezer Annan on 15 July 2015.
The user has requested enhancement of the downloaded file.
Statistics of Flow and the Scaling of Ceramic Water Filters
Ebenezer Annan1; Kabiru Mustapha2; Olushola S. Odusanya3; Karen Malatesta4; and Winston O. Soboyejo5
Abstract: According to the World Health Organization (WHO), there was an increase in the number of people that have access to safe
drinking water between 2006 and 2010. Such trends can be accounted for partly by the increasing usage of ceramic water filters that can
remove microbial pathogens from water. However, the initial flow rates in such filters are often limited to ranges between 1 and 3 L=h. In this
paper, six frustum-shaped ceramic water filters of the same clay:sawdust composition were tested. Each ceramic water filter was filled with
water and allowed to filter 20 times. Each time, the flow rate and water level were measured for a consecutive 12 h. Permeability values were
estimated for each run of the ceramic water filters. Statistical analysis was performed on flow rates (in the first hour), mean flow rates, and
estimated permeability values. The flow rate values (in the first hour) for the six ceramic water filters were found to be between 1.4 and
3.0 L=h. An effective permeability was obtained for ceramic water filters with a range of microscale and nanoscale pore sizes. The statistical
variations in the flow rates and effective permeabilities were elucidated along with the potency of a multiple ceramic water filter system for
scale-up studies in serving communities that need portable water. DOI: 10.1061/(ASCE)EE.1943-7870.0000862. © 2014 American Society
of Civil Engineers.
Author keywords: Ceramic water filter; Permeability; Flow rate statistics; Scale-up filters; Multiple filter system.
Introduction Sobsey 2002) that can improve the health and well-being of people
in rural/urban areas in the developing world.
It is estimated that 1.8 million deaths per year and 61.9 million Several point-of-use treatment systems (Murcott 2006; Sobsey
disability-adjusted life-years are attributed to the drinking of unsafe 2002) can be used to provide microbiological, chemical, or
water and poor sanitation (WHO/UNICEF 2000). The WHO also physical water treatment. They include (1) disinfection [chlorina-
suggests that over 99.8% of the deaths are caused by unsafe water, tion, solar disinfection (SODIS), solar pasteurization, ultraviolet
sanitation, and hygiene in developing countries (WHO/UNICEF irradiation with lamps, and boiling] (Liu and Fitzpatrick 2010),
2000). Furthermore, in developing countries, most of the childhood (2) particle filtration (cloth fiber, ceramic filter, biosand, and other
mortality occurs in children who are less than 5 years old (WHO/ slow sand filter technologies) (Heather et al. 2010; Yakub et al.
UNICEF 2004). 2013), (3) adsorption media (granular activated carbon and
Those who have access to improved water may also experience activated alumina or clay) (Liu and Fitzpatrick 2010; Yakub and
contamination between the source and the point of use. Hence, the Soboyejo 2013), (4) combined system (combined flocculation
key is to ensure that people have access to improved water at the and disinfection or filtration) (Liu and Fitzpatrick 2010; Preston
point of use. This is crucial for the achievement of the Millennium et al. 2010), and (5) other approaches (plain sedimentation settling,
Development Goals (MGDs) of halving the number of people safe storage, flocculation with iron or alum salts, and membrane
without access to improved water supply and sanitation by the year processes) (Preston et al. 2010; Davies et al. 2010). Within the
2015 (WHO/UNICEF 2004). To achieve this target, it is important range of possible water treatment methods, water filters are one
to explore point-of-use water purification methods (Murcott 2006; of the most highly rated systems for the removal of turbidity
(Van Halem et al. 2009; Brown et al. 2009; Brown and Sobsey
1Ph.D. Candidate, African Univ. of Science and Technology, P.M.B. 2010), fluoride removal (Yakub and Soboyejo 2013), and bacteria
681, Garki - Abuja, Federal Capital Territory, Nigeria; and Assistant inactivation (Van Halem et al. 2009; Brown et al. 2009; Brown and
Lecturer, Univ. of Ghana, Accra, Ghana. E-mail: eannan@ug.edu.gh Sobsey 2010; Oyanedel-Craver and Smith 2008; Yakub et al.
2Ph.D. Candidate, African Univ. of Science and Technology, P.M.B. 2013). Filters are advantageous because they are relatively inexpen-
681, Garki - Abuja, Federal Capital Territory, Nigeria. E-mail: sive (Van Halem et al. 2009; Brown et al. 2009).
mustkabir@googlemail.com
3Deputy Director, Biotechnology Laboratory, Sheda Science and Tech- Ceramic water filters are produced from a mixture of clays and
nology Complex (SHESTCO), P.M.B. 186, Garki - Abuja, Federal Capital sawdust. The mixtures are then molded into pot shapes using steel
Territory, Nigeria. E-mail: shola2@hotmail.com molds that are subjected to hydraulic pressure. After drying, the
4Lecturer and Research Specialist II, Dept. of Mechanical and Aero- green bodies are treated at approximately 850°C (Yakub and
space Engineering, Princeton Univ., 1 Olden St., Princeton, NJ 08544. Soboyejo 2013; Yakub et al. 2013; Van Halem 2006). During
E-mail: kmalates@princeton.edu the heating process, the carbonaceous materials are burnt at temper-
5Professor, Dept. of Mechanical and Aerospace Engineering, Princeton atures between 400 and 500°C and sintered between 800 and 900°C
Univ., 1 Olden St., Princeton, NJ 08544; Professor, Dept. of Materials (Yakub and Soboyejo 2013; Yakub et al. 2013; Van Halem
Science and Engineering, African Univ. of Science and Technology, et al. 2009; Brown et al. 2009; Brown and Sobsey 2010;
P.M.B. 681, Garki - Abuja, Federal Capital Territory, Nigeria (correspond- Oyanedel-Craver and Smith 2008; Van Halem 2006). They are then
ing author). E-mail: soboyejo@princeton.edu
Note. This manuscript was submitted on August 21, 2013; approved on furnace-cooled (to minimize thermal shock) and dipped in water for
May 8, 2014; published online on July 10, 2014. Discussion period open 24 h. After dipping to saturation, the filters are soaked in colloidal
until December 10, 2014; separate discussions must be submitted for in- silver or silver nitrate (Yakub and Soboyejo 2013; Yakub et al.
dividual papers. This paper is part of the Journal of Environmental En- 2013; Van Halem et al. 2009; Brown et al. 2009; Brown and Sobsey
gineering, © ASCE, ISSN 0733-9372/04014039(11)/$25.00. 2010; Oyanedel-Craver and Smith 2008; Van Halem 2006) to
© ASCE 04014039-1 J. Environ. Eng.
J. Environ. Eng. 
Downloaded from ascelibrary.org by RUTGERS UNIVERSITY on 08/29/14. Copyright ASCE. For personal use only; all rights reserved.
y
introduce a coating layer that disinfects microbial pathogens (Dies
2003). Although the use of ceramic water filters has been promoted
by nongovernmental organizations such as Potters for Peace (Liu
and Fitzpatrick 2010), the current understanding of the flow and ts
permeability in ceramic water filters is limited (Liu and Fitzpatrick
2010; Dies 2003; Lantagne 2002; Swanton 2008). r(y)
Prior work on flow through porous ceramic water filters has been h(t)
carried out by Van Halem (2006), Yakub (2012), Yakub et al.
(2013), Schweitzer et al. (2013), and Bear (1972). The Darcy equa-
tion (Bear 1972) is very instrumental in flow studies. Van Halem
et al. (2009), Yakub (2012), and Yakub et al. (2013) used Darcy
flow concepts to model the flow through porous ceramic water fil-
ters. Yakub (2012) also considered the flow through the sides and
base of frustum-shaped ceramic water filters. Flow characteristics of tb θ
frustum-shaped ceramic water filters with varying sawdust:clay
content were also studied by Yakub et al. (2013). However, the ear- Fig. 1. Schematic of frustum-shaped ceramic water filter
lier models did not fully explore the statistical variations in the
Darcy flow parameters and the applicability of the Darcy flow
model under scenarios in which the filters are used multiple times. where ρ = density of fluid (in this case, water); πr2o = area; g = the
The most recent publication on flow modeling—Schweitzer acceleration due to gravity; hðtÞ = the height of the waterhead; tb
et al. 2013—deduced equations that can be used to compute the and ts = the respective thicknesses of the bottom and sides of the
flow rate of frustum-shaped and parabolic-shaped ceramic water ceramic water filter; Qs and Qb = the respective flow rates through
filters (with uniform thickness of the sides and bottom). They per- the sides and bottom of the ceramic water filter; and θ = angle of
formed tests on ceramic water filters for three runs (every 8 h=day). inclination at the corners. The flow through an annular element
This indeed suggested that, barring clogging, increasing the along the sides of the filter is given by
frequency of loading of the filter or runs may increase the volume
of water produced. The hydraulic conductivities obtained in Z
hðtÞ
their work (k¼ 1.2 × 10−7 m=s for parabolic-shaped ceramic water ¼ 2πKρg− Qs ðh − yÞðro þ y tan θÞdy ð2Þfilters and k¼ 0.78 × 10 7 m=s for frustum-shaped ceramic water μts 0
filters) were found to be comparable to previous estimates in the
literature (Oyanedel-Craver and Smith 2008; Van Halem 2006). Integration of Eq. (2) gives
However, the dependence of flow rate on permeability or the
hydraulic pressure head has not been statistically fully explored 
 πKρg 3
under enough multiple loadings or runs. Therefore, the objectives Q ¼ r 2s oh ðtÞ þ
tan θh ðtÞ ð3Þ
of this paper are to (1) explore and expand variations in flow rate (in tsμ 3
the first hour) values and permeability (in the first hour) values for
20 runs of six frustum-shaped ceramic water filters, (2) give insight The total flow, Q, through the sides and bottom of the filter is,
into the spread of flow rate (in the first hour) values and permeabil- therefore, given by the sum of Eqs. (1) and (3)
ity (in the first hour) values using statistical distribution tools,
(3) understand the behavior or spread of the mean flow rate values 
 ð Þ ð Þ K r2 2o roh t h t
and effective permeability values using statistical tools within the Q ¼ πρghðtÞ þ þ tan θ ð4Þμ t t 3t
context of Darcy’s equation, and (4) assess the scale-up potential of b s s
ceramic water filters in serving a community. The implications of
the results are then discussed. Thus, a more general expression for the flow rate through the
frustum-shaped ceramic water filter following the Schweitzer et al.
(2013) derivation procedure (Yakub 2012; Yakub et al. 2013) is
Analytical Modeling given in Eq. (4). This is expressed as a function of the height of
the waterhead in the ceramic water filter. The height of the water
Schweitzer et al. (2013) modeled the flow through ceramic water level in the ceramic water filter is a function of time, which depends
filters. In their analysis, they assumed uniform thickness for the on the volumetric flow rate and the filter geometry. The volume of
walls and bottom of the frustum-shaped ceramic water filter. water, VðtÞ, contained in the frustum-shaped ceramic water filter at
Schweitzer and co-workers initiated their modeling by the any given time is given as
Darcy equation (Bear 1972). The modified Schweitzer et al. 
 
(2013) modeled equation is presented for the flow rate of 3ð Þ ¼ ð Þ þ ð Þ þ h ðtÞtan
2θ
V t π R2h t Rh2frustum-shaped ceramic water filters. The equation is made general t tan θ ð5Þ3
by allowing for differences in the thicknesses of the walls and
bottom of the ceramic water filter. Fig. 1 shows the parameters
of the flow through the ceramic water filter. The expressions obtained for hðtÞ were generally found to be of
The flow rates for the bottom and sides of the ceramic water the form
filter following the similar procedure of Schweitzer et al. (2013)
are given in Eqs. (1) and (3) h3ðtÞfα 23t − β3g þ h ðtÞfα2t − β2g þ hðtÞfα1t − β1g ¼ 0 ð6Þ
¼ K πr
2
Q o
ρghðtÞ ð1Þ where α1, α2, α3, β1, β2, and β3 are all constants defined in theb μ tb following equation, and t is defined as time.
© ASCE 04014039-2 J. Environ. Eng.
J. Environ. Eng. 
Downloaded from ascelibrary.org by RUTGERS UNIVERSITY on 08/29/14. Copyright ASCE. For personal use only; all rights reserved.
¼ KπρgR
2
¼ ¼ KπρgR phenomena during the early stages of the flow experiments (Yakubα1 ; β1 πR2;αμt 2 ;
  μt et al. 2013). The saturated ceramic water filters were then mountedb s
2 in the receptacles and filled with 9 L of tap water at the start of the
¼ ¼ tan θ Kπρg tan θβ R tan θ; α ; β ¼ ð7Þ experiments. The flow through the filters was then measured by2 3
3 μt 3s 3 recording the volume of water discharged from the ceramic water
Eq. (6) shows that hðtÞ can converge. The measured flow rate filter after each hour. The waterhead, h, was also measured. The
plot versus the time-dependent values of hðtÞwas used to determine waterhead, h, is a function of time and is usually represented as
permeability values and was then consequently used in the fitting hðtÞ in equations. Six ceramic water filters were used in this section
process. of the research. Each ceramic water filter was loaded with water 20
times, and flow experiments were performed at each loading or run
to determine the variability in the flow parameters. In this way, the
Materials and Methods means and standard deviations of the permeabilities and flow rates
were obtained along with the statistical distributions that provided
Ceramic Water Filter Preparation the reasonable characterization of the measured flow rates.
Two types of clays were used in the fabrication of the filters. They are
Ewuya and Iro clays, which were mined from Abeokuta, Ogun State, Multiple Filter Studies
Southern Nigeria. The sawdust was obtained from a saw mill in Five ceramic water filters were connected such that both the fillings
Sapon Market, Abeokuta-Nigeria. The clays and the sawdust were of the filters and outflows were connected in parallel. The parallel
sieved through 1-mm pores. They were then mixed in the outflows were then connected to a common pipe, thus creating a
volume ratios of 50% clay and 50% sawdust. Two liters of water common discharge outlet. The system had a main supply that was
were added before molding into a frustum shape (Oyanedel-Craver filled with water and was connected to the filtration system. Each
and Smith 2008; Yakub 2012). The water was added in a small ceramic water filter was initially ensured to be filled with 9 L of
incremental amount until the clay mixture coalesced into a clay ball water before starting the timer. The volume of water that flowed out
whose surface is not cracked and also does not stick on the board of the common outlet at hourly time intervals was recorded and
or sides of the mixing machine. The molded shapes were then continued for 12 h.
air-dried for 2 weeks before firing. The firing involved heating
from room temperature (approximately 25–30°C) to 850°C,
followed by furnace cooling in air. The firing was done in a kiln (with Results and Discussion
thermocouple measurement of temperature). During heating, the
sawdust was burnt off at a temperature of approximately 500°C X-Ray Diffraction and X-Ray Fluorescence Analyses
(Yakub 2012; Yakub et al. 2013; Plappally et al. 2011). The initial
heating rate of 45°C=h was increased to 100°C=h for the firing Figs. 2(a–c) show the XRD diffraction patterns obtained for the
process. clays. These show the minerals in the Ewuya and Iro clays. The
Ewuya clay contained mostly kaolinite and silica, while the Iro clay
had silica, mica, and montmorillonite prior to firing. However, after
Porous Material Characterization firing, the fired clays were found to also contain allophane (alumi-
X-ray diffraction (XRD) analysis was carried out on the as-received num silicate hydrate and potassium silicate hydroxide) [Fig. 2(c)].
clays and the sintered product. X-ray diffraction was performed Table 1 provides the chemical compositions of the Ewuya and
using an XPERT-PRO diffractometer (PANalytical BV, Almelo, Iro clays. The major differences between the Ewuya and Iro clays
Netherlands) with theta/theta geometry. The system was operated were in the Al2O3 and SiO2 contents. The Ewuya clay contains
in a cobalt tube at 35 kVand 50 mA. The goniometer was equipped approximately 74.43% by weight SiO2 and 11.46% by weight
with an automatic divergence slit and a PW3064 spinner stage. The Al2O3, and the Iro clay has 61.88% by weight SiO2 and
XRD patterns of all specimens were recorded in the 10-degree to 15.14% by weight Al2O3. The more plastic Iro clay had a higher
50-degree, 2θ range, with a step size of 0.017 degrees and a Al2O3 content of approximately 15.14% by weight compared to the
counting time of 14.1565 s per step. Qualitative phase analysis Ewuya clay, which contained approximately 11.46% by weight
was conducted using X’pert HighScore Plus search match software Al2O3. Conversely, the Iro clay contained approximately
(PANalytical B.V., Almelo, Netherlands). 61.88% by weight SiO2 compared to the Ewuya clay, which con-
X-ray fluorescence (XRF) spectroscopy of the clay samples tained approximately 74.43% by weight SiO2. Hence, the mixing
was also conducted to determine the chemical compositions. of the Ewuya and Iro clays produced a composite clay with inter-
The XRF data were collected using a Thermo Fisher ARL9400 mediate silica and alumina contents that optimized the plasticity
XP+ Sequential XRF spectrometer equipped with a WinXRF soft- and the thermal shock resistance of the clay mixtures during hy-
ware package. The samples were milled in a tungsten-carbide draulic pressing and sintering, respectively.
milling pot to achieve particle sizes of less than 75 microns.
The samples were then dried at 100°C and roasted at 1,000°C to Flow through Ceramic Filters
determine loss on ignition (LOI) values. A gram of the sample
was mixed with 6 g of lithium teraborate flux and fused at Figs. 3(a–f) show the mean flow rates obtained for the 20 runs for
1,050°C to make a stable fused glass bead. For trace element each ceramic water filter of the six ceramic water filters that were
analyses, the sample was mixed with a polyvinyl acetate (PVA) tested. The individual flow rates were found to be between approx-
binder and pressed into a pellet using a 10-t press. imately 1.4 and 3.0 L=h during the first hour of the flow. However,
the flow rates decreased with increasing time. This was due to the
decrease in the hydraulic pressure and possibly clogging or unclog-
Flow Experiments ging for increasing flow durations (Van Halem 2006; Schweitzer
The ceramic water filters were soaked in water for 12 h prior to et al. 2013; Yakub 2012). Fig. 3 shows the measured flow rate
the flow experiments. This was done to avoid transient flow variations for different durations with their analytical modeled
© ASCE 04014039-3 J. Environ. Eng.
J. Environ. Eng. 
Downloaded from ascelibrary.org by RUTGERS UNIVERSITY on 08/29/14. Copyright ASCE. For personal use only; all rights reserved.
α α
(a) (b) Note: R – Silica, S – Mica, T – Montmorillonite 
(c) Note: U – Aluminum silicate hydrate, V – Potassium aluminum silicate hydroxide
Fig. 2. X-ray diffraction patterns of clays showing peak minerals identified: (a) Ewuya clay; (b) Iro clay; (c) fired clay
plots. These clearly show that the flow rates vary significantly as a Davies et al. 2010; Brown et al. 2009; Brown and Sobsey 2010;
function of time and variations in the individual filters. Also, for a Oyanedel-Craver and Smith 2008; Van Halem 2006).
single ceramic water filter, no clear trend was observed for mea- The mean flow rate of water discharge for the six filters was
sured flow rate from one run to the next run over the entire 20 runs. 2.3 L=h during the first hour of discharge. The effective permeabil-
Furthermore, the variations in the flow rates would be expected ity of each ceramic filter was also obtained by fitting the measured
to result in variations in the permeabilities extracted from the over- flow rate data to Eq. (4). The permeability values for the six filters
all flow data (Yakub and Soboyejo 2013; Yakub et al. 2013; were found to be between approximately 0.44 × 10−14 and
2.54 × 10−14 m2, with an average value of 1.19 × 10−14 m2 (hy-
draulic conductivity; k¼ 1.46 × 10−7 m=s). The average effective
Table 1. X-Ray Fluorescence Results of Clays permeability value is found to be very comparable to permeability
% composition Ewuya clay Iro clay values found for ceramic water filters from Cambodia
SiO 74.43 61.88 (1.37 × 10
−7 m=s) and Ghana (1.3 × 10−7 m=s) in Van Halem’s
2
TiO 1.00 0.80 (2006) work. Also, the average effective permeability computed2
Al2O3 11.46 15.14 is well within the hydraulic conductivity range (1.15 × 10
−7 to
Fe O 4.51 8.84 5.01 × 10−72 3 m=s) obtained by Oyanedel-Craver and Smith
MnO 0.09 0.08 (2008) for three ceramic water filters. Furthermore, Yakub et al.
MgO 0.24 1.19 (2013) found permeability values between approximately
CaO 0.22 0.53 0.1 × 10−14 and 5.0 × 10−14 m2. A linear dependence of the flow
Na2O 0.14 0.33
K O 1.13 0.84 on the permeability exists, which therefore suggests that the effec-2
P O 0.02 0.01 tive permeability approach captures the trends in the flow rate data.2 5
Cr O 0.01 0.02 Figs. 4(a–f) and 5(a–f) show the variability in the permeabilities2 3
NiO <0.01 <0.01 and flow rates, respectively, for the six filters that were each tested
V2O5 0.01 0.02 over 20 runs. In each case, the filters exhibited statistical variations
ZrO2 0.09 0.05 attributable to pore clogging and unclogging and changes in
CuO <0.01 <0.01 crack sizes within the porous structures (Schweitzer et al. 2013;
LOI 6.12 10.20 Van Halem 2006; Dies 2003; Yakub 2012; Plappally et al.
Total 99.47 99.93 2011). The variations in the flow rates were found to be reasonably
© ASCE 04014039-4 J. Environ. Eng.
J. Environ. Eng. 
Downloaded from ascelibrary.org by RUTGERS UNIVERSITY on 08/29/14. Copyright ASCE. For personal use only; all rights reserved.
(a) (b)
(c) (d)
(e) (f)
Fig. 3. Comparison of experimental flow rates to model simulation for ceramic water filters: (a) Filter F1; (b) Filter F2; (c) Filter F3; (d) Filter F4;
(e) Filter F5; (f) Filter F6
characterized by normal distributions that fitted the measured ex- water Filters F4 and F6 were found to have effective permeability
perimental data. The mean flow rate and the effective permeability P-values fairly close to the chosen 0.05 alpha value. This, therefore,
values spreads for the 20 runs of each filter are plotted (with fitted depicts a weak dependence in terms of the distribution being nor-
normal distribution) in Figs. 6 and 7 respectively. The distribution mally distributed. The mean flow rate P-values of ceramic water
of the flow rate (in the first hour and mean) and permeabilities (in Filter F6 was found to be 0.03 and thus failed the normality test.
the first hour and mean) obtained for 20 runs are fairly normally Tables 2 and 3 give the P-values computed using the Kolmogorov-
distributed, and this is confirmed using the Kolmogorov-Smirnov Smirnov normality test and the OriginPro software statistical tool.
normality test and the OriginPro software statistical tool.
Furthermore, the average flow rates obtained for the six filters
were found to increase with increasing permeability values. At the Scale-up Potential
0.05 level, all data drawn for the flow rate (in the first four) and The scale-up filtration system experiments conducted suggest that
the permeabilities were significantly normally distributed. Ceramic it is possible to scale up the number of water filtration systems to
© ASCE 04014039-5 J. Environ. Eng.
J. Environ. Eng. 
Downloaded from ascelibrary.org by RUTGERS UNIVERSITY on 08/29/14. Copyright ASCE. For personal use only; all rights reserved.
(a) (b)
(c) (d)
(e) (f)
Fig. 4.Histogram of flow rates (in the first hour) with normal distribution fit for ceramic water filters: (a) Filter F1; (b) Filter F2; (c) Filter F3; (d) Filter
F4; (e) Filter F5; (f) Filter F6
serve rural communities. Fig. 8 shows the arrangement of the a system will prove capable of producing safe drinking water for
ceramic water filters in the multiple filter studies. The cumulative eight different households (four members) for a day. Even with de-
water discharge over time obtained is plotted in Fig. 9. An average creasing waterhead for a continuous 12 h, the average flow rate was
flow rate of 7 L=h (in the first hour) was found for five ceramic 4.5 L=h, producing 37 L of water (extrapolation from Fig. 9).
water filters connected in parallel. The system was able to produce Based on the parameters obtained from this study, at maximum
a total volume of 23 L of water, with decreasing waterhead, within waterhead for 10 h, a community of 525 people could be served
the first 5 h (refer to Fig. 9). Therefore, with an average flow rate of by 15 multiple-filter systems. These 15 systems should be located
7 L=h and considering the WHO guidance for daily safe drinking at accessible positions in the community. It is important to charac-
water of 2 L per day, the volume of water that flowed out of the terize the individual ceramic water filters of the multiple-filter sys-
common outlet (at maximum waterhead for 10 continuous h, tem in terms of flow well. Moreover, the number of filters used in
7 L=h × 10 h ¼ 70 L) can serve a community of 35 people. Such the multiple-filter system can be increased to serve even larger
© ASCE 04014039-6 J. Environ. Eng.
J. Environ. Eng. 
Downloaded from ascelibrary.org by RUTGERS UNIVERSITY on 08/29/14. Copyright ASCE. For personal use only; all rights reserved.
(a) (b)
(c) (d)
(e) (f)
Fig. 5. Histogram of permeability values (in the first hour) with normal distribution fit for ceramic water filters: (a) Filter F1; (b) Filter F2; (c) Filter
F3; (d) Filter F4; (e) Filter F5; (f) Filter F6
communities. The results obtained from this study, therefore, sug- strength, and Darcy’s flow equation. The key condition for oper-
gest the possibility of having a combined multi-filter system that is ation without fracture would be to ensure that the stress intensity
connected to water obtained from a borehole or a polluted source. factor, K, applied to the cracks in the ceramic water filters is less
However, one problem with such multi-filter systems is that if one than the fracture toughness (KIC).
ceramic water filter unit is cracked or not functioning well, the en-
tire water discharged is contaminated. Furthermore, as a way of
reducing the total number of ceramic water filters in multiple-filter Implications
systems, larger ceramic water filters can be produced. These can be The implications of the aforementioned results are quite significant.
designed by considering the fracture toughness, pressure-induced First, they show that the variations in the flow rates across the water
© ASCE 04014039-7 J. Environ. Eng.
J. Environ. Eng. 
Downloaded from ascelibrary.org by RUTGERS UNIVERSITY on 08/29/14. Copyright ASCE. For personal use only; all rights reserved.
(a) (b)
(c) (d)
(e) (f)
Fig. 6. Histogram (with normal distribution fit) of mean flow rates for ceramic water filters: (a) Filter F1; (b) Filter F2; (c) Filter F3; (d) Filter F4;
(e) Filter F5; (f) Filter F6
filters are well-characterized by normal distributions during the first needed to determine a material parameter that captures the effective
hour of flow. Furthermore, the effective permeability values asso- flow through the porous structures of the filters. However, these
ciated with the 20 consecutive flows through the filters under require more detailed analyses of flow data that are probably well
investigation were well-characterized by normal distributions. This beyond the capabilities of most ceramic water filter factory engi-
suggests that 20 consecutive filter tests can be used to establish the neers. Hence, simple software is needed to enable engineers to
variability in the initial flow rates and permeabilities of the filters. establish the variability in the effective permeabilities for
The initial flow rates are easier to estimate, especially within rural efficient applications in quality control.
village settings in which ceramic water filters were fabricated and In any case, the correlations established between the effective
used. However, they do not capture the overall trends in the flow permeabilities and the flow rate after the first hour suggest that
over the period of discharge. Hence, the effective permeabilities are either the permeability or flow rate approach may be used to
© ASCE 04014039-8 J. Environ. Eng.
J. Environ. Eng. 
Downloaded from ascelibrary.org by RUTGERS UNIVERSITY on 08/29/14. Copyright ASCE. For personal use only; all rights reserved.
(a) (b)
(c) (d)
(e) (f)
Fig. 7. Histogram (with normal distribution fit) of effective permeabilities for ceramic water filters: (a) Filter F1; (b) Filter F2; (c) Filter F3; (d) Filter
F4; (e) Filter F5; (f) Filter F6
Table 2. Normality Test Analysis of Ceramic Water Filters for Flow Rate and Permeability
Filter F1a Filter F2a Filter F3a Filter F4a Filter F5a Filter F6a
Filter parameter K1b Q1c K2b Q2c K3b Q3c K4b Q4c K5b Q5c K6b Q6c
P-value 0.87 1.00 0.26 0.40 1.00 1.00 0.91 0.94 0.94 0.63 1.00 1.00
Normality test PNTd PNTd PNTd PNTd PNTd PNTd PNTd PNTd PNTd PNTd PNTd PNTd
Note: The normality test was verified at an alpha α-value of 0.05.
aFilters F1–F6 are porous clay ceramic water filters.
bKn (where n is an integer from 1 to 6) are permeability histogram data in the first hour of flow.
cQn (where n is an integer from 1 to 6) are flow rate histogram data in the first hour of the flow experiment.
dPNT = passed the Kolmogorov-Smirnov normality test.
© ASCE 04014039-9 J. Environ. Eng.
J. Environ. Eng. 
Downloaded from ascelibrary.org by RUTGERS UNIVERSITY on 08/29/14. Copyright ASCE. For personal use only; all rights reserved.
Table 3. Normality Test Analysis of Ceramic Water Filters for Mean Flow Rates and Effective Permeabilities
Filter F1a Filter F2a Filter F3a Filter F4a Filter F5a Filter F6a
Filter parameter K1b Q1c K2b Q2c K3b Q3c K4b Q4c K5b Q5c K6b Q6c
P-value 0.52 0.72 0.85 0.18 0.61 0.35 0.07 0.20 0.32 0.45 0.07 0.03
Normality test PNTd PNTd PNTd PNTd PNTd PNTd PNTd PNTd PNTd PNTd PNTd FNTe
Note: The normality test was verified at an alpha α-value of 0.05.
aFilters F1–F6 are porous clay ceramic water filters.
bKn (where n is an integer from 1 to 6) are permeability histogram data in the first hour of flow.
cQn (where n is an integer from 1 to 6) are flow rate histogram data in the first hour of the flow experiment.
dPNT = passed the Kolmogorov-Smirnov normality test.
eFNT = failed the Kolmogorov-Smirnov normality test.
Fig. 8. Multi-filter scale-up system
over periods of 2–3 years in which most ceramic water filters are
expected to provide filtered water at rates between approximately
1 and 3 L=h. These are clearly the challenges for future work.
Conclusions
This paper presents the results of combined analytical and exper-
imental study of flow through frustum-shaped ceramic water filters.
The statistical variations in flow rates were found to be well-
described by normal distributions. The average flow rate was
2.3 L=h for first-hour use at maximum waterhead, with the effec-
tive permeabilities for the six filters ranging between 0.44 × 10−14
and 2.54 × 10−14 m2. Permeabilities for the first hour of use can be
deduced from the Darcy equation. The flow through frustum-
shaped ceramic water filters was well-described by Darcy’s equa-
Fig. 9. Volume of water discharged for scale-up system tion. The results suggest that reasonable filtered water can be
obtained by adopting filter testing methods that involve the use
of 20 tests in the establishment of the statistical variations in flow
characterize and test the filters. Further work is clearly needed to rates and effective permeabilities for effective filter quality control.
characterize the statistical variations in the flow rate parameters The average flow rate and effective permeabilities were found to be
at different stages of the filter life. There is also a need to establish well-characterized by the normal distribution. The linear depend-
the acceptable variances in filter flow rates and permeabilities ence of flow rates (in the first hour) on the measured permeabilities
© ASCE 04014039-10 J. Environ. Eng.
J. Environ. Eng. 
Downloaded from ascelibrary.org by RUTGERS UNIVERSITY on 08/29/14. Copyright ASCE. For personal use only; all rights reserved.
suggests that filter quality may be assessed using either flow rate or Liu, X., and Fitzpatrick, C. S. B. (2010). “Removal of humic substances
permeability measurements. using solar irradiation followed by granular activated carbon adsorp-
The multiple-filter study also shows that a combination of filters tion.” Water Sci. Technol. Water Supply, 10(1), 15–22.
may be used to provide drinking water for communities of different Murcott, S. (2006). “Implementation, critical factors and challenges to
sizes. Since the overall flow rates from multiple-filter systems scale scale-up of household drinking water treatment and safe storage
systems.” Proc., Household Drinking Water Treatment and Safe
with the number of filters, the filter sizes and the number of filters Storage (HWTS) Electronic Conf., USAID/Hygiene Improvement
can be scaled to provide adequate drinking water for communities. Project (HIP).
Furthermore, the multiple filters can be placed in strategic locations ORIGIN 7.0 PRO [Computer software]. OriginLab, Northampton, MA.
within rural communities to provide easy access to safe drinking Oyanedel-Craver, V. A., and Smith, J. A. (2008). “Sustainable colloidal-
water in rural/urban communities in developing countries. silver-impregnated ceramic filter for point-of-use water treatment.”
J. Environ. Sci. Technol., 42(3), 927–933.
Plappally, A. K., Yakub, I., Brown, L. C., Soboyejo, W. O., and Soboyejo,
Acknowledgments A. B. O. (2011). “Physical properties of porous clay ceramic-ware.”
J. Eng. Mater. Technol., 133(3), 1–9.
Ebenezer Annan’s Ph.D. research at the African University of Preston, K., Lantagne, D., Kotlarz, N., and Jellison, K. (2010). “Turbidity
Science and Technology (AUST), Abuja, Federal Capital Territory, and chlorine demand reduction using alum and moringa flocculation
Nigeria, was sponsored by the Next Generation of Academics in before household chlorination in developing countries.” J. Water
Africa (NGAA) project at the University of Ghana, Accra-Ghana. Health, 8(1), 60–70.
The NGAA project is a Carnegie cooperation of the New York– Schweitzer, R. W., Cunningham, J. A., and Mihelcic, J. R. (2013).
U.S.A. funding project at the University of Ghana, Ghana. The “Hydraulic modeling of ceramic water filters for point-of-use water
treatment.” Environ. Sci. Technol., 47(1), 429–435.
authors are also grateful to the World Bank STEP-B Program, Sobsey, M. D. (2002). Managing water in the home: Accelerated health
the African Centers of Excellence Program, the African Develop- gains from improved water supply, World Health Organization, Geneva.
ment Bank, and the Princeton Grand Challenge Program for finan- Swanton, A. A. (2008). “Evaluation of the complimentary use of the
cial support. The authors would like to thank the staff at Mateng, ceramic (kosim) filter and aqua-tabs in Northern Region, Ghana.”
Nigeria Limited, Abeokuta-Nigeria, for their support in the fabri- M. E. thesis, Massachusetts Institute of Technology, Cambridge, MA.
cation of ceramic water filters used in this research. Van Halem, D. (2006). “Ceramic silver impregnated pot filters for house-
hold drinking water treatment in developing countries.” M.Sc. thesis,
Delft Univ. of Technology, Delft, The Netherlands.
References Van Halem, D., Lann, H. V. D., Heijman, S. G. J., Dijk, J. C. V., and Amy,
G. L. (2009). “Assessing the sustainability of the silver-impregnated
Bear, J. (1972). Dynamics of fluids in porous media, American Elsevier ceramic pot filter for low-cost household drinking water treatment.”
Publication Company, New York. Phys. Chem. Earth, 34(1–2), 36–42.
Brown, J., Proum, S., and Sobsey, M. D. (2009). “Sustained use of a house- WinXRF version 3.0 [Computer software]. Thermo Fisher Scientific,
hold-scale water filtration device in rural Cambodia.” J. Water Health, Ecublens, Switzerland.
7(3), 404–412. World Health Organization (WHO)/United Nations Children’s
Brown, J., and Sobsey, M. J. (2010). “Microbiological effectiveness of Fund (UNICEF). (2000). Assessment Rep., Water Supply and Sanitation
locally produced ceramic filters for drinking water treatment in Council, New York, 〈http://www.who.int/docstore/water_sanitation
Cambodia.” J. Water Health, 8(1), 1–8. _health/Globassessment/Globaltoc.htm〉 (Jun. 6, 2013).
Davies, S. H., Baumann, M. J., Byun, S., Corneal, L. M., Tarabara, V. V., World Health Organization (WHO)/United Nations Children’s Fund
and Masten, S. J. (2010). “Fabrication of catalytic ceramic membranes (UNICEF). (2004). Meeting the MDG drinking water and sanitation
for water filtration.” Water Sci. Technol. Water Supply, 10(1), 81–85. target: A mid-term assessment of progress, Joint Monitoring
Dies, W. R. (2003). “Development of a ceramic water filter for Nepal.” M. Programme for Water Supply and Sanitation, New York.
E. thesis, Dept. of Civil and Environmental Engineering, Massachusetts X’pert HighscorePlus version 2.1 [Computer software]. PANalytical B.V.,
Institute of Technology, Cambridge, MA. EA Almelo, The Netherlands.
Heather, M. M., Edward, A. M., and Khosrow, F. (2010). “A critical Yakub, I. (2012). “Micro- and nano-porous adsorptive materials for
evaluation of two point-of-use water treatment technologies: Can they removal of contaminants from water at point-of-use.” Ph.D. thesis,
provide water that meets WHO drinking water guidelines?” J. Water Princeton Univ., Princeton, NJ.
Health, 8(4), 611–630. Yakub, I., et al. (2013). “Porosity, flow, and filtration characteristics of
Lantagne, D. (2002). “Investigation of the Potters for Peace colloidal silver- frustum-shaped ceramic water filters.” J. Environ. Eng., 10.1061/
impregnated ceramic filter: Intrinsic effectiveness and field performance (ASCE)EE.1943-7870.0000669, 986–994.
in rural Nicaragua.” Proc., 5th Specialized Conf. on Small Water and Yakub, I., and Soboyejo, W. O. (2013). “Adsorption of fluoride from water
Wastewaster Treatment Systems, International Water Association, using sintered clay-hydroxyapatite composites.” J. Environ. Eng.,
London. 10.1061/(ASCE)EE.1943-7870.0000692, 995–1003.
© ASCE 04014039-11 J. Environ. Eng.
View publication stats J. Environ. Eng. 
Downloaded from ascelibrary.org by RUTGERS UNIVERSITY on 08/29/14. Copyright ASCE. For personal use only; all rights reserved.