Chemical Engineering and Processing: Process Intensi Fication

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Chemical Engineering and Processing 105 (2016) 21–29

Contents lists available at ScienceDirect

Chemical Engineering and Processing:


Process Intensification
journal homepage: www.elsevier.com/locate/cep

Comparative study of in-line coagulation and/or ozonization


pre-treatment for drinking-water production with spiral-wound
ultrafiltration membranes
Fátima Rojas-Serrano* , Jorge Ignacio Pérez, Miguel Ángel Gómez
Technologies for Water Management and Treatment Research Group, Department of Civil Engineering, University of Granada, Campus de Fuentenueva s/n
18071, Granada, Spain

A R T I C L E I N F O A B S T R A C T

Article history:
Received 10 June 2015 The application of pre-treatments is a common approach for improving ultrafiltration membrane
Received in revised form 20 March 2016 limitations. However, most studies have been performed for lab-scale hollow-fibre or flat-sheet
Accepted 12 April 2016 membrane modules. The present study was conducted to test the performance of fullscale spiral-wound
Available online 14 April 2016 ultrafiltration (UF) membranes for potabilization of influents rich in humic and fulvic acids under
different scenarios: (i) direct UF, (ii) with in-line pre-coagulation, (iii) with pre-ozonation, and (iv) with
Keywords: both pre-treatments combined. The membranes were polyvinylidene fluoride with 0.03 mm effective
Ultrafiltration pore size. Permeate flow rate was fixed, 1.0 m3/h, equivalent to 48 LMH flux. Membrane performance was
Spiral-wound
compared for the aforementioned cases with the following results: feed water pretreatment was
Pre-coagulation
required since the effluent quality achieved with direct ultrafiltration was poor and fouling increased.
Pre-ozonation
Humic-fulvic-rich influents Between pre-ozonation and pre-coagulation, the former was disregarded given the negligible
Membrane fouling improvement achieved with respect to single ultrafiltration. By contrast, coagulation-flocculation
results were remarkably better for both organic matter removal and fouling control. The application of
the two pretreatments in series resulted in effluent qualities comparable to those achieved with
coagulation-flocculation but fouling rates were much lower and transmembrane pressure growth was
minimal, this constituting a promising option for fouling control.
ã 2016 Elsevier B.V. All rights reserved.

1. Introduction transmembrane pressure, TMP, vacuum vs. pressure driven, etc.),


increasing the versatility of the systems.
The use of ultrafiltration membranes for water treatment has Nevertheless, the main drawbacks of the membranes remain
spread worldwide because of its multiple advantages. The unsolved. Firstly, the poor screening of small-sized natural
separation mode is size exclusion, by molecular sieving through organic matter (NOM) constitutes a serious handicap, especially
fine pores [1]. These membranes are effective in screening out for surface water. Secondly, the membranes are porous materials
bacteria and suspended solids [2] while avoiding the disadvan- subject to fouling, i.e. permeability loss from the accumulation of
tages of chemical disinfection. Moreover, the minor footprint and aquatic substances on or inside the membrane matrices [5].
the simple operation of the systems [3], together with the rising Fouling obligates the operation to stop periodically for back-
competitiveness of the membrane prices [4], make this technology washing, cleaning or relaxation periods, thus shortening mem-
an attractive alternative for drinking-water production. In fact, brane life [6] and increasing operational costs. Depending on their
vigorous research is currently being focused on membrane nature, foulants can be classified as colloidal, organic, scaling,
materials (ceramic, polymeric, metallic, etc.), configurations biofouling, etc. [7] although NOM is considered the worst
(hollow fibre, flat-sheet, tubular, spiral-wound, etc.) and operation constraint for drinking-water applications [8].Thus, membrane
modes (dead-end vs. cross-flow, constant flux vs. constant fouling and low retention of NOM are related and require
integrated solutions.
Two different approaches to this problem can be found in the
literature: (i) the optimisation of the operating conditions and (ii)
the application of pre-treatments, [4,5]. In fact, in a previous work
* Corresponding author.
[9], our research group optimised the operating conditions for a
E-mail address: rojas_f@ugr.es (F. Rojas-Serrano).

http://dx.doi.org/10.1016/j.cep.2016.04.004
0255-2701/ ã 2016 Elsevier B.V. All rights reserved.
22 F. Rojas-Serrano et al. / Chemical Engineering and Processing 105 (2016) 21–29

drinking-water production unit by means of direct ultrafiltration. 2. Materials and methods


We found that operating conditions certainly influenced mem-
brane performance since the right combination of variables could 2.1. Influent water
lower the TMP increase and the resulting energy consumptions.
Nonetheless, even when the best combination was applied, the The influent was synthetic but the matrix water came from
feasibility of the process was questioned as the permeate recovery Canales Reservoir in Granada (Spain), with the following physico-
was reduced as well. As a result, the pre-treatment option seems chemical parameters: pH 8.00; turbidity 5.0 NTU and UV254
more realistic, especially when the influents are of low quality. 0.4 m1. The dissolved organic carbon (DOC) was 0.5 mg/L and the
Several studies have combined ultrafiltration and pre-treat- suspended solids concentration (SS) was negligible. As a result, the
ments, offering widely different possibilities: coagulation [10– organic load was increased in order to simulate medium- to low-
12], adsorption [13], pre-filtration [14], pre-oxidation [15,16], etc. quality surface water. A mixture of humic and fulvic acids
However, most studies have used laboratory-scale membranes (Carbotecnia Húmico 10-10, Carbotecnia S.L., Spain) was dissolved
with only a few exceptions [12,17]. On the other hand, despite the in the source water so that the resulting influent DOC was
potential of spiral-wound (SW) modules for drinking water 7.5  1 mg/L, although this concentration could vary slightly
production, SW membranes having higher filtration surface and according to the changes in the matrix water. The influent
packing density [18], hollow fibres [10,12–16,19,20] or to a lesser characterization is included in Table 1.
extent flat sheets [11,14,21,22] have traditionally been preferred
to the detriment of the former. Thus, a comparative study was 2.2. Ultrafiltration modules
conducted to evaluate the performance of full-scale spiral-wound
membranes in different scenarios: direct filtration and ultrafil- The membrane modules used were commercially available
tration preceded by other processes. spiral-wound SpiraSep 960 by TriSep Corporation (Goleta,
Of the possible pre-treatments to be applied, coagulation- California, USA); made of polyvinylidene fluoride (PVDF), with
flocculation (CF) was chosen for being the most promising, as 20.9 m2 filtration area and 0.03 mm effective pore size. The
shown by a great number of CF-UF works. Indeed, it has been modules operate submerged and driven with a vacuum, with
successfully applied to increase organic matter removal as well as 0.7 bar being the threshold transmembrane pressure (TMP).
to improve membrane performance [22–25]. On the other hand,
ozonation (OZ) was selected as well, since few examples of the 2.3. Description of the experimental plant
combination OZ-UF were found in the literature, possibly because
of the incompatibility of many polymeric membranes with ozone The matrix water was drawn from the inlet channel to be
[4]. In fact, to date, most studies have used ceramic membranes in discharged into a PVC intermediate tank (2000 L). A centrifuge feed
which permeability after ozone application increased with respect pump sucked the water from the intermediate tank and fed the
to single ultrafiltration [26,27]. Both pre-treatments involve the experimental plant, with the humic/fulvic acid solution being
addition of chemicals although their mechanisms sharply differ. injected upstream of the feed pump for proper mixing inside the
On the one hand, CF consists of adding an agent (coagulant) that pump body. The feed flow rate was adjusted to 1.1 m3/h.
destabilises the colloidal particles in the influent, making their The facility was designed to work in different scenarios, so that
aggregation possible. The resulting aggregates are presumably the pipe downstream of the feed pump branched off (Fig. 1).
filterable through the membrane. By contrast, OZ involves Depending on the position of the manual valves, the influent water
oxidising the organic matter present in the water by adding could follow three possible paths:
ozone. The size and nature of the fragments generated depend on
the extent of the oxidation reaction, affecting the success of the (a) Direct ultrafiltration: The first branch entered the membrane
process. tank, where the water was filtered from outside to inside
Finally, the main purpose of this study was to determine which driven by the permeate pump. The operation was flux constant,
combination of the following was more effective in terms of 48 LMH, and part of the permeate was reversed through the
fouling reduction and NOM removal: direct ultrafiltration, CF-UF membrane during backwashing every 10 min with 2 m3/h flow
and OZ-UF. In addition, the combination of the three was also rate and 30 s duration. Aeration was continuous (18 Nm3/h).
tested in different sequences, to verify whether the order had any These operating conditions were fixed according to a previous
influence. In this way, CF-OZ-UF and OZ-CF-UF were tested as study [9]. The operation was dead-end and without a reject.
well. (b) Coagulation-flocculation pre-treatment: The second branch
was connected to a static mixer through which the coagulant

Table 1
Average physicochemical characteristics of the influent and the different effluents.

Description Influent UF effluent CF-UF effluent OZ-UF effluent CF-OZ-UF effluent OZ-CF-UF effluent

Value Value % rem.


Turbidity, NTU 5.39  0.16 1.83  0.29 65.56 0.95  0.06 81.45 1.17  0.41 80.02 1.00  0.00 81.11 1.00  0.00 80.9
DOC, mg/L 7.55  0.16 6.34  0.10 13.66 4.41  0.58 42.86 5.35  0.28 29.81 4.30  0.06 40.32 4.71  0.24 39.75
UVA254, m1 18.49  0.14 11.07  0.40 40.07 3.42  0.37 81.3 10.85  0.71 41.89 3.60  0.10 80.74 3.67  0.06 80.15
SUVA, m1/mg/L 2.49  0.10 1.75  0.04 30.6 0.78  0.11 68.32 2.03  0.18 17.06 0.84  0.01 66.27 0.78  0.05 68.67
Colour436, m1 3.01  0.11 1.33  0.15 54.62 0.20  0.05 93.04 1.03  0.08 66.96 0.22  0.05 92.54 0.23  0.06 92.43
Fe3+, mg/L 0.00  0.00 0.00  0.00 – 0.06  0.02 – 0.00  0.00 – 0.01  0.00 – 0.01  0.00 –
pH 8.21  0.04 8.23  0.04 – 7.85  0.08 – 8.18  0.08 – 7.91  0.04 – 7.86  0.09 –
F. Rojas-Serrano et al. / Chemical Engineering and Processing 105 (2016) 21–29 23

Fig. 1. Flow diagram of the experimental set-up.

was injected. A ferric chloride matrix solution was prepared by was maintained constant at 25 g O3/Nm3 in order to achieve a O3:
dilution of 40% (m/m) FeCl3 solution, (Brenntag, Germany) in DOC ratio of 0.6 (mg/mg), assuming full ozone transfer.
pure water, enabling a rate of 50 mg FeCl3/L in the influent, the The experimental facility was enclosed and insulated so that the
coagulant dose being fixed according to previous experiments temperature variations were insignificant. However, temperature
[28] .The coagulant-influent mixture entered the in-line together with TMP and O3 concentrations were recorded by a data
flocculator, which consisted of a PVC pipe, with 50 mm register (RSG30, Endress Hauser, Switzerland).
external diameter (2.4 mm thickness) and 100 m long. The Furthermore, the facility was provided with by-passes allowing
flocculator outlet was connected to the membrane tank. Then, the simultaneous operation of all the processes described above.
ultrafiltration was carried out as described under (a). Sample points were available for influent, permeate, and also for
(c) Ozonation pre-treatment: A third by-pass connected the feed the mixture influent-coagulant, i.e. flocculated water, and for the
pump discharge to a venturi for ozone injection. Ozone was mixture influent-ozone, i.e. ozonated water.
produced from pure oxygen (Air Liquide S.A.), by means of a
generator (C-L010DTI, AirTree Ozone Technology Co., Ltd., 2.4. Experimental procedure
Taiwan) operating at 1.2 bar. The oxygen flow rate applied was
0.2 Nm3/h. The whole study was conducted over approximately one year
divided into different stages:
An in-line ozone analyser (Mini-HiCon, In USA Corp., USA)
allowed measuring the ozone concentration in the gas line. After 1. Direct ultrafiltration (UF).
ozone injection, the water-ozone mixture entered a 12-min 2. Coagulation-flocculation followed by ultrafiltration (CF-UF).
retention-time contact column to favour the ozone transfer, and 3. Ozonation followed by ultrafiltration (OZ-UF).
then left the column through an overflow, separating the gas and 4. Coagulation-flocculation in series with ozonation followed by
water. The ozonated water was conducted to a separate tank, from ultrafiltration (CF-OZ-UF) and vice versa (OZ-CF-UF).
which it was pumped to the ultrafiltration circuit following the
same route as mentioned above. The non-reacting gas was released Each stage lasted eight weeks, after which the membrane
in the upper part of the overflow, ending at a destruction cell, module in operation was removed and replaced by a new one. The
where the ozone reacted with saturated solution of NaHSO3 only exception was stage 4, for which one single membrane
producing O2. Nonetheless, before ozone destruction, the off-gas module was used for the two options tested. Each combination was
ozone concentration was measured by a second in-line ozone tested for eight weeks also, but with a four-week relaxation period
analyser (Mini-HiCon built in SC010, In USA Corp., USA). Thus, the in between.
transferred ozone dose, TOD, could be calculated with Eq. (1),[29]. The operation consisted of 6 working-days/1 relaxation-day
with 240-min-long assays. The influent and effluent were sampled
Q gas 
TOD ¼  ½O3 0  ½O3 res ð1Þ daily to determine which configuration was more effective in
Q liq
terms of permeate quality. In addition, ozonated and flocculated
where TOD is the transferred ozone dose (g O3/L), Qgas and Qliq are water samples were also taken during stage 4 in order to determine
the flow rates for the inlet gas (Nm3/h) and the feed water (L/h), the water quality in the intermediate phases of the combined
respectively, and [O3]0 and [O3] res are the ozone concentrations process. On the other hand, the transmembrane pressure (TMP)
(gO3/Nm3) in the inlet gas and the off-gas streams, respectively. was registered every 20 s for fouling evaluation. Ozone concen-
Mozia et al. [30] have successfully tested pre-ozonation with doses trations in the inlet and off-gas lines were also registered with the
up to 0.6 (mg/mg) O3:TOC. Therefore the inlet ozone concentration same frequency.
24 F. Rojas-Serrano et al. / Chemical Engineering and Processing 105 (2016) 21–29

0.25
0.20
0.15
0.10
TMP, bar 0.05
0.00
-0.05
-0.10
-0.15
-0.20 y = -4.10E-06x-0.10
R2 = 0.99
-0.25
0 2000 4000 6000 8000 10000 12000 14000 16000
Time, s
Fig. 2. TMP time course for a normal operating period within direct UF stage.

After each assay, the membrane was soaked in clean water. were called fouling rates. The comparison of the average fouling
Furthermore, cleaning in place (CIP) was carried out the last rates for each stage allowed an assessment of which scenario was
working day of each operating period, during which the membrane more favourable in terms of fouling avoidance.
tank was emptied and filled with permeate while NaClO (50 mg/L)
was dosed. After the membrane tank was filled, recirculation 2.7. Statistical analyses
between that and the permeate tank was conducted for 20 min.
Then, the membrane tank was emptied again and filled with tap Regression models were built for fouling-rate calculation. In
water until the next operation period. addition, analysis of the variance tests (ANOVA) were made in
order to compare the quality of the samples and to determine
2.5. Analytical methods whether there were statistically significant differences among
them at a significance level of 5%, p < 0.05.
Physicochemical analyses were performed daily for the influent
and effluent as well as for intermediate samples in the last stage. 3. Results
Samples were integrated, 100 mL being collected every 30 min for
the whole operation time. 3.1. Direct ultrafiltration
Turbidity, DOC, UVA254, colour436 nm, and pH were determined.
In addition, the iron content was analysed for the effluents The physicochemical parameters of the effluents resulting
resulting from the application of pre-coagulation. Turbidity was from the ultrafiltration as a single treatment are shown in Table 1.
determined as a measurement of the scattered light (DINKO D-112, The yields for turbidity were acceptable for the entire operating
Spain) while Fe3+ concentration was measured by means of a period. However, the colour removal was poor (55%) and, in fact,
spectrophotometric test kit (Merck KGaA, Germany). UV–vis the colour of the effluent samples was visible to the naked eye. On
spectrophotometer Helios g (Spectronic Unicam, USA) was used the other hand, even when the efficiency for UVA254 was medium
for the analysis of colour436nm, UVA254nm, and Fe565nm, with 1 cm (40% average), DOC removal was very low, with yields of less than
quartz cell. Glass-fibre filters (Millipore AP4004705) were used to 15%.
obtain filtrated samples for DOC measurement, made by combus- Fig. 2 shows the TMP time course for a normal operating period
tion TOC Analyser (Formacs HT, SKALAR, The Netherlands). Specific within stage 1. The positive values correspond to backwashing
ultraviolet absorbance (SUVA) index was calculated as UV254 nm whereas the negative correspond to permeate production.
 100/DOC (mg/L). Table 2 summarizes membrane fouling rates for the different
stages. According to the high coefficients of determination found
2.6. Fouling evaluation for direct ultrafiltration (average R2 = 0.99), linear regression
closely fit the data. The fouling rate reached for the first working
Given that the operation was flux constant, membrane fouling period was 2.6  106 bar/s, with 0.05 bar TMP increment after
was indirectly quantified by means of the time course of TMP. 240 operating min. The fouling rate continued to increase up to
Linear regression was applied to TMP values immediately prior to 4.1 106 bar/s during the first week. Then, although TMP build-
backwashing vs. operation time for each assay. The resulting slopes up continued throughout the entire experimental period, the slope

Table 2
Summary of fouling rates, Vf, results for the whole study.

Process Starting Vf, bar/s Average Vf, bar/s Maximum Vf, bar/s Minimum Vf, bar/s R2 Average TMP, bar
UF 2.6E-06 5.0E-06 5.5E-06 2.6E-06 0.99 0.07
CF + UF 1.1E-06 5.0E-07 2.7E-06 2.2E-07 0.75 0.01
OZ + UF 2.5E-06 1.7E-06 2.8E-06 1.4E-06 0.87 0.02
CF + OZ + UF 3.0E-07 3.5E-07 3.0E-07 4.3E-07 0.33 0.00
OZ + CF + UF 4.1E-07 2.9E-07 4.1E-07 8.3E-07 0.37 0.00
F. Rojas-Serrano et al. / Chemical Engineering and Processing 105 (2016) 21–29 25

0.30
0.25
0.20
0.15
0.10
TMP, bar 0.05
0.00
-0.05
-0.10
-0.15
-0.20
y = -1.64E-06x-0.15
-0.25 R2 = 0.82
-0.30
0 2000 4000 6000 8000 10000 12000 14000 16000
Time, s

Fig. 3. TMP time course for a normal operating period within CF-UF stage.

growth began to slow, the maximum value being 5.5  106 bar/s. performance of the respective membrane modules with the same
The average value for the whole stage was 5.0  106 bar/s criterion. The best fouling slope was positive, 2.2  107 bar/s,
whereas the average TMP increment for operating periods was whereas the maximum fouling resulted in 2.7  106 bar/s rates.
0.07 bar. However, the CIP and relaxations applied were very The average TMP increment for each working period was less than
effective and the TMP threshold value was never reached. 0.01 bar (Table 2).

3.2. Coagulation-flocculation prior to ultrafiltration 3.3. Ozonation prior to ultrafiltration

The application of coagulation-flocculation before ultrafiltra- Despite the high ozone concentration applied, 25 g O3/Nm3,
tion considerably improved the effluent quality, as shown in TOD was lower than expected, as there was always residual ozone
Table 1.The yields registered for organic matter removal were in the off-gas line, 4.2 g O3/Nm3 on average. The average TOD was
medium in terms of DOC, i.e. 43% average, but high in terms of 3.7 mg/L, the ozone dose being equivalent to the O3: DOC ratio of
UVA254, i.e. 81% average. Colour removal was significant also, with 0.5 (mg/mg).
very high yields, averaging 93%. The turbidity of the effluent also The effluent characteristics resulting from the application of
improved, the average value being below 1 UNT. In addition, the ozonation as a pre-treatment are shown in Table 1. Ozonation
maximum Fe effluent concentration according to Directive 98/83/ proved adequate for the removal of turbidity but less effective for
EC for water intended for human consumption, 0.2 mg Fe/L, was organic matter reduction, according to the DOC and UVA254 yields
never reached, although the pH slightly fell. (30 and 42%, respectively). On the other hand, the average colour-
The fouling rate was initially 1.1 106 bar/s but gradually reducing efficiency was 67%, while the pH barely changed.
decreased, 5.0  107 bar/s being the average value for the stage. The initial fouling rate was 2.5  106 bar/s but after several
Fig. 3 shows the TMP time course for a normal operating period days of operation, values decreased almost one point, averaging
within CF-UF stage. The figures fluctuated frequently and the around 1.7  106 bar/s (See Table 2). Fig. 4 shows the TMP time
correlation factors were weaker, with average R2 values of around course for a random operating period within OZ-UF stage. The
0.75. In any case, linear regression was still the best fit and it was coefficients of determination, R2, were strong, between 0.85 and
necessary to maintain the same model in order to compare the 0.9, although no positive fouling slopes were found in this case.

0.30
0.25
0.20
0.15
0.10
TMP, bar

0.05
0.00
-0.05
-0.10
-0.15
-0.20
y = -2.71E-06x-0.16
-0.25 R2 = 0.90
-0.30
0 2000 4000 6000 8000 10000 12000 14000 16000
Time, s

Fig. 4. TMP time course for a normal operating period within OZ-UF stage.
26 F. Rojas-Serrano et al. / Chemical Engineering and Processing 105 (2016) 21–29

0.30
0.25
0.20
0.15
0.10
TMP, bar
0.05
0.00
-0.05
-0.10
-0.15
-0.20 y = 0.32E-06x-0.11
-0.25 R2 = 0.37
-0.30
0 2000 4000 6000 8000 10000 12000 14000 16000
Time, s

Fig. 5. TMP time course for a normal operating period within CF-OZ-UF experiments.

However, the rates were very similar from one day to another. The pre-treatment applied. By contrast, ozonated-water DOC was
average TMP increment for working periods was 0.02 bar but 7.03 mg/L when ozonation was applied first, whereas the same
TMP growth was controlled by means of the weekly CIPs and parameter was 4.97 mg/L when coagulation-flocculation was
relaxation periods. Additionally, the application of ozone did not performed first.
compromise the integrity of the PVDF membrane throughout the Concerning the membrane performance, the initial fouling
operation period. rates were 4.1 107 bar/s and 3.0  107 bar/s for OZ-CF-UF
and CF-OZ-UF, respectively (Table 2). However, after the first
3.4. Combined pre-treatments before ultrafiltration operation hours, most of the fouling slopes turned positive for the
two alternatives tested, as reflected in Fig. 5 and Fig. 6. In fact, the
Ozonation followed by coagulation and subsequent ultrafiltra- average values for the overall operating period were positive for
tion resulted in very similar effluent characteristics to those found both cases, 2.9  107 bar/s and 3.5  107 bar/s, respectively. The
with the opposite combination. In fact, no statistically significant pressure time course for a normal working period was almost
differences (p-value > 0.05) were found. The average parameters constant, varying by only a few thousandths of a bar. Nevertheless,
for the corresponding effluents are listed in Table 1. The yields for the correlations were very weak, with average R2 coefficients
the removal of DOC and UVA254 were around 40 and 80%, around 0.4.
respectively. Nonetheless, the best results were achieved again for
turbidity and colour, with 81 and 92% removal percentages. The Fe 4. Discussion
content of the effluents was consistently below 0.2 mg/L.
The analyses of the intermediate samples showed no remark- 4.1. Effluent quality
able differences for the flocculated water regardless of whether
coagulation-flocculation was performed before or after ozonation. The application of direct ultrafiltration with spiral-wound
On the contrary, the ozonated water was influenced by the order membranes proved insufficient for the production of drinking
applied. In general the values for turbidity, colour or UVA254 were water from influents rich in humic/fulvic content. The permeate
considerably higher when coagulation-flocculation was the first was heavily coloured and DOC and UVA254 removal percentages

0.30
0.25
0.20
0.15
0.10
TMP, bar

0.05
0.00
-0.05
-0.10
-0.15
y = 0.76E-06x-0.10
-0.20
R2 = 0.40
-0.25
-0.30
0 2000 4000 6000 8000 10000 12000 14000 16000
Time, s
Fig. 6. TMP time course for a normal operating period within OZ-CF-UF experiments.
F. Rojas-Serrano et al. / Chemical Engineering and Processing 105 (2016) 21–29 27

were very low. Such high organic content in the effluent fully applied to any water source. Ozonation results could be
compromises the organoleptic properties of water, making it potentially improved if higher ozone transfer was achieved. This
unsuitable for human consumption. Moreover, the addition of would involve increasing the contact ozone-water by enlarging the
chlorine would be necessary, which in turns would relocate the column-retention time or providing a packing, together with the
risks of the disinfection by-products (DBPs) in the drinking-water application of higher inlet ozone concentrations. Both measures
distribution networks [31]. entail higher investment in equipment and energy and, although a
However, the combination coagulation-flocculation-ultrafiltra- cost analysis lies beyond the scope of the present study, according
tion proved quite effective in lowering the organic matter content, to our results and those of other authors [39,40], the resulting
especially the aromatic part, since UVA254 elimination averaged permeate quality does not appear to justify the cost rise, making
81%. The yields for DOC removal were the highest for all the pre-ozonization unadvisable for ultrafiltration of influents with
scenarios tested, the results being consistent with those reported high organiccontent. By contrast, although there is a margin for
by other authors for lab-scale hollow-fibre membranes [32–34]. On improving coagulation-flocculation, e.g. testing new coagulants or
the other hand, the slight drop in pH was due to the acidification of optimising the coagulant injection, the process is much cheaper
the water caused by iron hydroxide formation, but the average and easier to control, making this option preferable among all the
value was in all cases within the requirements of the Council scenarios tested.
Directive 98/83/EC. Furthermore, the efficiency of the CF pre-
treatment for the removal of colour and turbidity was remarkable 4.2. Membrane fouling
and the maximum iron content met EU standards.
By contrast, the removal of UVA254 with ozonation pre- The application of pre-treatments was found to help control
treatment was only slightly better than for single ultrafiltration, fouling, given that the resulting average rates were considerably
and the DOC removal achieved indicated that complete minerali- lower for any of the combined processes. In fact, ultrafiltration
zation of the organic matter was far from being reached. In fact, the without pre-treatment was the only case where the starting
SUVA index changed slightly with respect to the influent. In fouling rate continued to increase over the entire stage.
addition, instead of the strong colour degradation expected Between coagulation-flocculation and ozonation, the former
according to other authors [35,36], the maximum yield recorded was demonstrably better. Fig. 4 shows that the slope for OZ-UF
was only 67%, this implying a serious shortcoming, since humic descended more sharply than that for CF-UF, Fig. 3, indicating that
and fulvic acids are highly responsible for colouring natural waters. the membrane became fouled more quickly. Compared to single
These data indicate that the ozone transfer could have been ultrafiltration, the average fouling rates were one order of
insufficient for the nature and relatively high concentration of magnitude lower, this difference not being as significant for
organic matter in the influent water. According to Eq. (1) [29], for a ozonation (Table 2). The starting fouling rates were also lower for
given gas-to-liquid ratio, TOD can be increased by minimizing the CF-UF. On the other hand, although the average rate for the CF stage
residual ozone concentration and/or raising the one at the inlet. was negative, some positive values resulted within this stage,
Throughout the experimental period the ozone-water contact was indicating a certain degree of pressure recovery during operation.
lower than expected, since residual ozone was constantly found in This fact could not be attributed to any particular condition but to
the off-gas line. Thus, the resulting 0.5 (mg/mg) O3:DOC ratio the coagulation-flocculation process itself, as the same coagulation
would have been inadequate. Byun et al. [35] reported that DOC parameters were applied throughout the stage. Nonetheless,
rejection increased only from 11% to 26% when raising the O3:DOC correlation factors were better for OZ-UF, a result which can be
ratio from 0.2 to 0.9 (mg/mg) for a PVDF-UF membrane. Similarly, explained by considering that linear regression perfectly fit the
for ceramic membranes, Szymanska et al. [37] found that the TMP time course for direct ultrafiltration; and given that ozonation
quality of the permeate could not be further improved even when barely decreased the organic content of the water going through
the ozone dosage was increased up to 3.1 (mg/mg) O3:TOC ratios the membrane, it seems logical that OZ-UF correlation factors were
for a flat-sheet microfiltration membrane. The results of these similar to those for UF and stronger than for CF-UF.
authors, for 11.7 mg DOC/L and 8 mg TOC/L influents, respectively, The results reached for full-scale CF-UF with spiral-wound
suggest that even when the ozone transfer had been greater, DOC membranes were consistent with works with other types of
removal would have barely increased. membranes, where the coagulant was found to be responsible for
Concerning the application of the two pre-treatments in series, the formation of a porous cake that could ease the filtration process
physicochemical parameters for the final effluents were very [41,42]. The role of ozone in fouling is less discussed in the
similar regardless of the order applied, and the soaking period in literature. Lee et al. [40] found the fouling layer to be thinner for
between did not appear to influence the effluent quality. However, hybrid ozonation-microfiltration than for single microfiltration.
ozonated-water analyses revealing much lower DOC, when Likewise, according to Szymanka et al. [37], the higher the ozone
coagulation-flocculation was performed first, confirmed that the dosage and flow rate, the lower the membrane fouling for ceramic
success with the combined process was attributable to coagula- membranes. Both studies used flat-sheet membranes. In our case,
tion. Jeong et al. [38], recently testing combined coagulation- the comparison of the average TMP increments reached for single
ozonation-ultrafiltration for lab-scale hollow-fibre membranes, ultrafiltration and for hybrid ozonation-ultrafiltration indicated
also concluded that coagulation was mainly responsible for that the application of ozone also helped delay the fouling of spiral-
decreasing DOC in the effluent. In any case, given that the removal wound membranes. From the rates being similar from one day to
efficiency was slightly lower than that achieved with pre- another it could be deduced that fouling generation was more or
coagulation-flocculation, double pre-treatment was disregarded less stabilized. However, unlike the CF-UF process, no positive
as a means of improving the final water quality. values were found, implying that the fouling layer that formed was
From the results for permeate quality, coagulation-flocculation less reversible. Nonetheless, fouling mechanisms occurring as a
was deemed the most recommendable pre-treatment for spiral- result of the application of in-line CF-UF or OZ-UF should be further
wound ultrafiltration membranes, given that the organic matter studied by means of respective membrane-module autopsies.
removal was considerably higher than for ozonation and nearly the Finally, the application of the two pre-treatments in series
same as for double pre-treatment. These findings confirm the provided the best membrane performance. The starting fouling
potential of this type of membrane for drinking-water production, rates were the lowest for the entire study and average fouling rates
although further research is needed before this technology can be were positive regardless of the order applied, signifying that there
28 F. Rojas-Serrano et al. / Chemical Engineering and Processing 105 (2016) 21–29

was pressure recovery throughout the respective operational fouling rates demonstrate. These findings make this combination a
periods, as reflected by the ascending slopes in Figs. 5 and 6. The promising option for fouling control.
CF-OZ-UF figures proved slightly better than OZ-CF-UF, which
could be the result of having used the same membrane module for Acknowledgements
both experiments. Despite the relaxation period in between, CF-
OZ-UF being tested first could have affected the results but the This study, CTM2010-18899-TECNO, was supported by the
differences were not relevant and the average TMP increments Spanish Ministry of Science and Innovation. It was held at the
were on the order of thousandths of a bar in both cases. Institute of Water Research (University of Granada) and Parque de
Nonetheless, the coefficients of determination for both alternatives las Ciencias, Granada (Spain) with the collaboration of CADAGUA S.
were the worst reached throughout the study, indicating very A. Thanks are also extended to Mr. José Manuel Quesada Gallego,
weak data correlations, probably as a consequence of the stronger from Ozono Electronica Iberica S.L., for his support during
changes in the influent water with the double pre-treatment. ozonation start-up and to Mr. David Nesbitt, professional transla-
These results indicate that the combination of the two pre- tor, for reviewing the English of the manuscript.
treatments can successfully reduce depositions on membrane
surface apparently because of the changes in the chemistry of the References
influent water, making the membrane less prone to fouling. These
findings were corroborated by Jeong et al. [38], who recently [1] R.W. Baker, Membrane Technology and Applications, John Wiley & Sons, Ltd.,
Chichester, UK, 2004, doi:http://dx.doi.org/10.1002/0470020393.
reported that the flux decline was remarkably better after [2] J.G. Jacangelo, R.Rhodes Trussell, M. Watson, Role of membrane technology in
ozonation together with coagulation than only with coagulation. drinking water treatment in the United States, Desalination 113 (1997) 119–
However, the results of the present study should be completed 127, doi:http://dx.doi.org/10.1016/S0011-9164(97)00120-3.
[3] A.W. Zularisam, A.F. Ismail, R. Salim, Behaviours of natural organic matter in
with more exhaustive analyses of the effluent, flocculated and membrane filtration for surface water treatment—a review, Desalination 194
ozonated water samples as well as membrane autopsies, in order (2006) 211–231, doi:http://dx.doi.org/10.1016/j.desal.2005.10.030.
to identify which changes in molecular size and functional groups [4] W. Gao, H. Liang, J. Ma, M. Han, Z. Chen, Z. Han, et al., Membrane fouling control
in ultrafiltration technology for drinking water production: a review,
in the water entering the membrane explain the lower TMP Desalination 272 (2011) 1–8, doi:http://dx.doi.org/10.1016/j.desal.2011.01.051.
increase and fouling. [5] H. Huang, K. Schwab, J.G. Jacangelo, Pretreatment for low pressure membranes
In summary, the hybrid process was a suitable option to control in water treatment: a review, Environ. Sci. Technol. 43 (2009) 3011–3019, doi:
http://dx.doi.org/10.1021/es802473r.
membrane fouling. Nevertheless, although cost analysis is beyond
[6] I. Sutzkover-Gutman, D. Hasson, R. Semiat, Humic substances fouling in
the scope of this study, the energy consumption of the combined ultrafiltration processes, Desalination 261 (2010) 218–231, doi:http://dx.doi.
pre-treatments is higher than for a single one. Therefore, the org/10.1016/j.desal.2010.05.008.
economic feasibility of combined processes compared to single [7] N. Li, A. Fane, W. Ho, T. Matsuura, Advanced Membrane Technology and
Applications, first ed., John Wiley & Sons, Ltd., Hoboken, NJ, USA, 2008.
coagulation-flocculation pre-treatment should be assessed as well. [8] G. Amy, Fundamental understanding of organic matter fouling of membranes,
Desalination 231 (2008) 44–51, doi:http://dx.doi.org/10.1016/j.
5. Conclusion desal.2007.11.037.
[9] F. Rojas-Serrano, R. Álvarez-Arroyo, J.I. Pérez, F. Plaza, G. Garralón, M.A. Gómez,
Ultrafiltration membranes for drinking-water production from low-quality
The present study on the performance of full-scale spiral- surface water: a case study in Spain, Membr. Water Treat. J. 6 (2015) 77–94,
wound ultrafiltration membranes confirms the potential of this doi:http://dx.doi.org/10.12989/mwt.2015.6.1.077.
[10] J.-D. Lee, S.-H. Lee, M.-H. Jo, P.-K. Park, C.-H. Lee, J.-W. Kwak, Effect of
configuration for drinking-water production. However, the mem- coagulation conditions on membrane filtration characteristics in coagulation—
branes must be used together with appropriate pre-treatment microfiltration process for water treatment, Environ. Sci. Technol. 34 (2000)
since the humic/fulvic acid removal achieved with direct filtration 3780–3788, doi:http://dx.doi.org/10.1021/es9907461.
[11] K.J. Howe, M.M. Clark, Effect of coagulation pretreatment on membrane
was very poor, DOC and UVA254 rejection being, respectively, 14% filtration performance, Am. Water Works Assoc. 98 (2006) 133–146.
and 40%. Furthermore, membrane fouling increased without any [12] Q. Xiangli, Z. Zhenjia, W. Nongcun, V. Wee, M. Low, C.S. Loh, et al., Coagulation
pressure recovery over time, with the risk of quickly reaching pretreatment for a large-scale ultrafiltration process treating water from the
Taihu River, Desalination 230 (2008) 305–313, doi:http://dx.doi.org/10.1016/j.
threshold transmembrane pressure.
desal.2007.11.032.
Pre-treatments such as ozonation or coagulation-flocculation [13] M. Campinas, M.J. Rosa, Assessing PAC contribution to the NOM fouling control
offer better effluent quality and improve membrane performance. in PAC/UF systems, Water Res. 44 (2010) 1636–1644, doi:http://dx.doi.org/
However, the effect of ozonation is less remarkable than that of 10.1016/j.watres.2009.11.012.
[14] M. Peter-Varbanets, J. Margot, J. Traber, W. Pronk, Mechanisms of membrane
coagulation-flocculation. The first option is not recommended fouling during ultra-low pressure ultrafiltration, J. Membr. Sci. 377 (2011) 42–
given the scant upgrade achieved with 0.5 mg/mg O3:DOC ratios 53, doi:http://dx.doi.org/10.1016/j.memsci.2011.03.029.
(3.7 mg O3/L) and the higher investment needed to potentially [15] T. Lin, S. Pan, W. Chen, S. Bin, Role of pre-oxidation, using potassium
permanganate, for mitigating membrane fouling by natural organic matter in
improve the effluent quality. By contrast, coagulation-flocculation an ultrafiltration system, Chem. Eng. J. 223 (2013) 487–496, doi:http://dx.doi.
prior to ultrafiltration has proved to be the most effective pre- org/10.1016/j.cej.2013.03.024.
treatment to reduce the organic matter content in the permeate [16] H. Xu, W. Chen, H. Xiao, X. Hu, Stability of an ultrafiltration system for drinking
water treatment, using chlorine for fouling control, Desalination 336 (2014)
among the possibilities tested, with 43% and 81%, DOC and UVA254 187–195, doi:http://dx.doi.org/10.1016/j.desal.2013.12.002.
removal, respectively. In addition, average fouling rates, being [17] S. Panglisch, G. Kraus, A. Tatzel, J.-P. Lickes, Membrane performance in
three times lower than for ozonation-ultrafiltration and one order combined processes including ozonation or advanced oxidation, powdered
activated carbon and coagulation—investigations in pilot scale, Desalination
of magnitude lower than for direct ultrafiltration, indicate that pre-
250 (2010) 819–823, doi:http://dx.doi.org/10.1016/j.desal.2008.11.049.
coagulation helps substantially to delay fouling. [18] V. Singh, P.K. Jain, C. Das, Performance of spiral wound ultrafiltration
Both pre-treatments applied in series led to effluent quality membrane module for with and without permeate recycle: experimental and
theoretical consideration, Desalination 322 (2013) 94–103, doi:http://dx.doi.
similar to that achieved with single coagulation-flocculation,
org/10.1016/j.desal.2013.05.012.
meaning that double pre-treatment is useless from the standpoint [19] M. Kabsch-Korbutowicz, A. Biłyk, M. Mołczan, The effect of feed water
of permeate quality. However, the combination of ozonation and pretreatment on ultrafiltration membrane performance, Polish J. Environ.
coagulation-flocculation minimizes fouling rates, in such a way Stud. 15 (2006) 719–725.
[20] X. Guo, Z. Zhang, L. Fang, L. Su, Study on ultrafiltration for surface water by a
that transmembrane pressure barely changes during operation (on polyvinylchloride hollow fiber membrane, Desalination 238 (2009) 183–191,
the order of magnitude of thousandths of a bar) and pressure is doi:http://dx.doi.org/10.1016/j.desal.2007.11.064.
recovered throughout the operating period, as positive average [21] C. Guigui, J.C. Rouch, L. Durand-Bourlier, V. Bonnelye, P. Aptel, Impact of
coagulation conditions on the in-line coagulation/UF process for drinking
F. Rojas-Serrano et al. / Chemical Engineering and Processing 105 (2016) 21–29 29

water production, Desalination 147 (2002) 95–100, doi:http://dx.doi.org/ for drinking water production, Desalination 145 (2002) 237–245, doi:http://
10.1016/S0011-9164(02)00582-9. dx.doi.org/10.1016/S0011-9164(02)00418-6.
[22] Y. Chen, B.Z. Dong, N.Y. Gao, J.C. Fan, Effect of coagulation pretreatment on [33] J. Wang, X.C. Wang, Ultrafiltration with in-line coagulation for the removal of
fouling of an ultrafiltration membrane, Desalination 204 (2007) 181–188, doi: natural humic acid and membrane fouling mechanism, J. Environ. Sci. (China)
http://dx.doi.org/10.1016/j.desal.2006.04.029. 18 (2006) 880–884, doi:http://dx.doi.org/10.1016/S1001-0742(06)60008-9.
[23] K. Konieczny, D. Sa˛kol, J. Płonka, M. Rajca, M. Bodzek, Coagulation— [34] S. Delgado Diaz, L. Vera Peña, E. González Cabrera, M. Martínez Soto, L.M. Vera
ultrafiltration system for river water treatment, Desalination 240 (2009) 151– Cabezas, L.R. Bravo Sánchez, Effect of previous coagulation in direct
159, doi:http://dx.doi.org/10.1016/j.desal.2007.11.072. ultrafiltration of primary settled municipal wastewater, Desalination 304
[24] J.C. Rojas, J. Pérez, G. Garralón, F. Plaza, B. Moreno, M.a. Gómez, Humic acids (2012) 41–48, doi:http://dx.doi.org/10.1016/j.desal.2012.08.005.
removal by aerated spiral-wound ultrafiltration membrane combined with [35] S. Byun, J.S. Taurozzi, A.L. Alpatova, F. Wang, V.V. Tarabara, Performance of
coagulation–hydraulic flocculation, Desalination 266 (2011) 128–133, doi: polymeric membranes treating ozonated surface water: effect of ozone
http://dx.doi.org/10.1016/j.desal.2010.08.013. dosage, Sep. Purif. Technol. 81 (2011) 270–278, doi:http://dx.doi.org/10.1016/j.
[25] H.E. Wray, R.C. Andrews, Optimization of coagulant dose for biopolymer seppur.2011.07.021.
removal: impact on ultrafiltration fouling and retention of organic [36] A. Kerc, M. Bekbolet, A.M. Saatci, Effect of partial oxidation by ozonation on the
micropollutants, J. Water Process Eng. 1 (2014) 74–83, doi:http://dx.doi.org/ photocatalytic degradation of humic acids, Int. J. Photoenergy 5 (2003) 75–80,
10.1016/j.jwpe.2014.03.007. doi:http://dx.doi.org/10.1155/S1110662X03000163.
[26] B. Schlichter, V. Mavrov, H. Chmiel, Study of a hybrid process combining [37] K. Szymanska, A.I. Zouboulis, D. Zamboulis, Hybrid ozonation–microfiltration
ozonation and microfiltration/ultrafiltration for drinking water production system for the treatment of surface water using ceramic membrane, J. Membr.
from surface water, Desalination 168 (2004) 307–317, doi:http://dx.doi.org/ Sci. 468 (2014) 163–171, doi:http://dx.doi.org/10.1016/j.memsci.2014.05.056.
10.1016/j.desal.2004.07.014. [38] K. Jeong, D.-S. Lee, D.-G. Kim, S.-O. Ko, Effects of ozonation and coagulation on
[27] J. Kim, S.H.R. Davies, M.J. Baumann, V.V. Tarabara, S.J. Masten, Effect of ozone effluent organic matter characteristics and ultrafiltration membrane fouling, J.
dosage and hydrodynamic conditions on the permeate flux in a hybrid Environ. Sci. (China) 26 (2014) 1325–1331, doi:http://dx.doi.org/10.1016/
ozonation–ceramic ultrafiltration system treating natural waters, J. Membr. S1001-0742(13)60607-5.
Sci. 311 (2008) 165–172, doi:http://dx.doi.org/10.1016/j.memsci.2007.12.010. [39] M.S. Siddiqui, G.L. Amy, B.D. Murphy, Ozone enhanced removal of natural
[28] F. Rojas-Serrano, J.I. Pérez, M.Á. Gómez, Integrated in-line coagulation-aerated organic matter from drinking water sources, Water Res. 31 (1997) 3098–3106.
ultrafiltration for drinking-water production: a case study from laboratory to [40] S. Lee, K. Lee, W.M. Wan, Y. Choi, Comparison of membrane permeability and a
pilot plant, J. Environ. Sci. Health A (2015), doi:http://dx.doi.org/10.1080/ fouling mechanism by pre-ozonation followed by membrane filtration and
10934529.2015.1064284. residual ozone in membrane cells, Desalination 178 (2005) 287–294, doi:
[29] P. Xu, M.L. Janex, P. Savoye, A. Cockx, V. Lazarova, Wastewater disinfection by http://dx.doi.org/10.1016/j.desal.2004.11.040.
ozone: main parameters for process design, Water Res. 36 (2002) 1043–1055, [41] A.W. Zularisam, A.F. Ismail, M.R. Salim, M. Sakinah, T. Matsuura, Application of
doi:http://dx.doi.org/10.1016/S0043-1354(01)00298-6. coagulation–ultrafiltration hybrid process for drinking water treatment:
[30] Z. Cui, T. Taha, Enhancement of ultrafiltration using gas sparging: a comparison optimization of operating conditions using experimental design, Sep. Purif.
of different membrane modules, J. Chem. Technol. Biotechnol. 78 (2003) 249– Technol. 65 (2009) 193–210, doi:http://dx.doi.org/10.1016/j.
253, doi:http://dx.doi.org/10.1002/jctb.763. seppur.2008.10.018.
[31] J.C. Rojas, B. Moreno, G. Garralón, F. Plaza, J. Pérez, M.a. Gómez, Influence of [42] W.Z. Yu, H.J. Liu, L. Xu, J.H. Qu, N. Graham, The pre-treatment of submerged
velocity gradient in a hydraulic flocculator on NOM removal by aerated spiral- ultrafiltration membrane by coagulation—effect of polyacrylamide as a
wound ultrafiltration membranes (ASWUF), J. Hazard. Mater. 178 (2010) 535– coagulant aid, J. Membr. Sci. 446 (2013) 50–58, doi:http://dx.doi.org/10.1016/j.
540, doi:http://dx.doi.org/10.1016/j.jhazmat.2010.01.116. memsci.2013.06.012.
[32] P. kyu Park, C. hak Lee, S.J. Choi, K.H. Choo, S.H. Kim, C.H. Yoon, Effect of the
removal of DOMs on the performance of a coagulation-UF membrane system

You might also like