Journal of Environmental Management 214 (2018) 362e369

Contents lists available at ScienceDirect

Journal of Environmental Management

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

Research article

Evaluation of adsorbent and ion exchange resins for removal of

organic matter from petroleum refinery wastewaters aiming to
increase water reuse

ria da Fonseca b, *
Rodrigo de Abreu Domingos a, Fabiana Vale
a
PETROBRAS, School of Chemistry, Federal University of Rio de Janeiro, Av. Henrique Valadares, 28, Centro, Rio de Janeiro, RJ, 20231-030, Brazil
b
ria, Ilha do Funda
School of Chemistry, UFRJ, Federal University of Rio de Janeiro, Av. Athos da Silveira Ramos 149, Cidade Universita ~o, Rio de Janeiro, RJ,
21941-909, Brazil

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

Article history: The oil refinery industry seeks solutions to reduce its water uptake and consumption by encouraging the
Received 3 July 2017 reuse of internal streams and wastewater from treatment systems. After conventional treatment the
Received in revised form petroleum refinery wastewater still contains a considerable quantity of recalcitrant organics and the
11 February 2018
adsorption on activated carbon is currently used in Brazilian refineries, although it is still expensive due
Accepted 5 March 2018
to the difficulty of its regeneration. This study evaluated the use of adsorbent and ion exchange resins for
the removal of organic matter from refinery wastewater after conventional treatment in order to verify
its feasibility, applying successive resin regenerations and comparing the results with those obtained for
Keywords:
Adsorbent resins
activated carbon process. Adsorption isotherms experiments were used to evaluate commercial resins,
Ion exchange resins and the most efficient was subjected to column experiments, where absorbance (ABS) and total organic
Organic matter carbon (TOC) removal were measured. The adsorption isotherm of the best resin showed an adsorptive
Petroleum refinery capacity that was 55% lower than that of activated carbon. On the other hand, the column experiments
Water reuse indicated good removal efficiency, and the amount of TOC in the treated wastewater was as good as has
been reported in the literature for activated carbon. The regeneration efficiency of the retained organics
ranged from 57 to 94%, while regenerant consumption ranged from 12 to 79% above the amount rec-
ommended by the resin supplier for the removal of organic material from natural sources, showing the
great resistance of these recalcitrant compounds to desorption. Finally, an estimate of the service life of
the resin using intermediate regeneration conditions found it to be seven times higher than that of
activated carbon when the latter is not regenerated.
© 2018 Elsevier Ltd. All rights reserved.

1. Introduction
Water availability per capita in Brazil is higher than in many
other countries in the world. However, this water is unevenly
distributed throughout the territory, with about 80% of the water
resources being in the Amazon basin, far from most industrial and
more densely populated regions. These regions face critical water
availability levels according to UN classification indexes (ANA,
2013). The recent water crisis faced by some Brazilian states
worsens the situation and directly affects the industrial sector. The
industries face increasing costs when purchasing water and spend
* Corresponding author.
E-mail addresses: rodrigo.domingos@petrobras.com.br (R. de Abreu Domingos),
fabiana@eq.ufrj.br (F.V. da Fonseca).
more resources to treat it. Due to these factors, process optimiza-
tion procedures and reuse actions are being increasingly applied in
order to improve process efficiencies and water recovery.
The oil industry is included in this context since it is a major water
consumer and, in many cases, is situated in water-scarce regions both
in Brazil and worldwide. According to PETROBRAS (2015), a Brazilian
petroleum refinery consumes an average of 0.9 m3 water per m3 of
processed oil, with water-based cooling systems and steam gener-
ation being the largest consumers (Amorim, 2005; Pombo, 2011). As
a result, a large amount of wastewater is also generated, which must
be treated before disposal since tit contains high levels of pollutants
such as oil, ammonia, phenols, sulfides, heavy metals, and organic
compounds (Nacheva, 2011).
Wastewater treatment in refineries primarily consists of two
steps: removal of contaminants (oil and suspended solids) by

https://doi.org/10.1016/j.jenvman.2018.03.022
0301-4797/© 2018 Elsevier Ltd. All rights reserved.
 R. de Abreu Domingos, F.V. da Fonseca / Journal of Environmental Management 214 (2018) 362e369
physical and physical-chemical processes, followed by a biological
step to remove organics and nitrogen compounds. As presented by
IPIECA (2010) and Tchobanoglous et al. (2003), a third stage, usually
called tertiary treatment, may be required to make the effluent
suitable for disposal or reuse as cooling tower makeup water,
process water or boiler feed water.
Tertiary treatment processes vary depending on the contaminants
to be removed. In refineries, clarification, filtration, or membrane
separation are commonly applied to remove suspended solids, which
is then followed by one or more dissolved salt removal processes,
such as electrodialysis reversal (EDR), reverse osmosis (RO), or ion
exchange (IEX) resins. However, most dissolved salt removal pro-
cesses require the reduction or elimination of residual organic com-
pounds present in the wastewater, since these may foul membranes
and IEX resins irreversibly. These recalcitrant compounds are difficult
to remove using conventional wastewater treatment because they are
poorly biodegradable, as indicated by Touma (2013a).
The most widely used process for removing these compounds in
industry is adsorption on granular activated carbon (GAC)
(Tchobanoglous et al., 2003; Yang, 2013), which is usually employed
after a biological treatment step as the cost of GAC consumption
without prior treatment would make the process economically
unfeasible. Although GAC adsorption has been used for many years
for water and wastewater treatment, it has some disadvantages
when used in refineries. Due to the high levels of organic matter
(OM) present after biological treatment, the GAC is quickly satu-
rated and should be replaced with new inventory, or regenerated/
reactivated. Thermal reactivation of GAC is not usually applied since
it is not performed in situ, requiring GAC to be transported to
reactivation sites that are not always close to the refineries. In
addition, it requires a frequent labor force for handling, high capital
expenditure, and has high GAC losses of 5e10% according to
Sufnarski (1999). GAC regeneration also follows this trend and thus
is not applied commonly either.
As an alternative to using GAC in refineries, polymeric IEX and
adsorbent resins are based on adsorption and IEX mechanisms.
They are more easily regenerated than GAC (Yang, 2013), as they
can be regenerated in situ by pH adjustment, solvents, or water
vapor, and they also result in lower losses than seen for GAC (Xu
et al., 2003). This makes them good options for the capture of OM
from oil refinery wastewaters even though their initial costs are
higher than GAC, as noted by Yang (2013). The use of resins to
capture some organic compounds, such as fulvic acids and tannins,
is currently applied using upstream demineralization systems,
usually to protect RO membranes and IEX resins.
The main objective of this work is to verify the applicability and
feasibility of a residual OM removal process using adsorbent and
IEX resins at the laboratory scale, in order to find an alternative to
the use of GAC for tertiary water treatment in oil refineries aiming
to reduce water consumption by reusing water. Initially, the
adsorption capacities of resins were measured using isotherms and
the best performing material was chosen for column tests. Column
experiments were then performed to verify the performance of the
resin when removing organics from wastewater and its ability to be
regenerated by removing the OMs using regenerant chemicals.
After the experiments, the service life of the resin when performing
intermediate regenerations was estimated and compared with that
of GAC under similar conditions as obtained from literature data
(Touma, 2013a), with the latter not being regenerated.
2. Materials and methods
2.1. Effluent source
Effluents from the wastewater treatment plants of Brazilian
363
refineries were used in this study. The samples were all collected
after membrane bioreactors (MBR) in 20 L bottles after passing
through ultrafiltration membranes, making them almost free of
undissolved OM and microorganisms, which also ensured
improved preservation of their original characteristics.
The effluents were sampled from two different refineries where
MBR are installed, herein called A and B. Refinery A is an industrial-
scale plant and provided the first samples for the initial isotherms.
Refinery B, which contained a pilot-scale MBR, provided the rest of
the samples for the adsorption/IEX isotherms and column experi-
ments. The OM content varied between samples due to the use of
different petroleum sources in the refinery feed and operational
variations in the wastewater treatment steps. Because of the large
effluent volume required, samples with the same characteristics
could not be used in all the experiments. The initial TOC value for the
refineries A and B was 14,2 ± 1,1 and 18,8 ± 0,7 mg/L, respectively.
2.2. Adsorbent and IEX resins
The macroporous adsorbent and IEX resins used in this study are
listed in Table 1, along with their main characteristics as listed by
their manufactures: matrices, functional groups, operational flow
required (in bed volumes per hour), and possible regenerant
chemicals. The resins were conditioned according to the manu-
facturer's recommendations by soaking them in water for half an
hour in a mixer and then removing the air inside their pores with a
vacuum pump before each experiment. The commercial names of
the resins are not included in order to maintain manufacturer's
confidentiality.
Since resin R8 was used throughout all experiments, its true and
apparent densities were also obtained according to DOW (2014)
and ASTM D2187-94 (2004), respectively. The true density was
1042 kg m
3, whereas the apparent density was 621 kg m
3. The
calculated porosity based on these values was 0.40. The average
particle diameter of these resins was 0.75 mm.
2.3. Granular activated carbon
NORIT 1240 W GAC was used in this study. GAC samples were
prepared by grinding and sieving the commercial samples (Mesh
12  40). After this, GAC was washed several times to remove fine
particles, dried at 100  C in an incubator and had the air inside the
pores removed using a vacuum pump.
2.4. Analytical methods
The source water and all samples were analyzed for total organic
carbon (TOC) and UV254 absorbance. The concentration of TOC was
utilized as the primary marker for performance. TOC testing was
performed according to APHA (2005) 5310C method using a
HiPerTOC TOC Analyzer. The UV absorbance was monitored
following the APHA (2005) 5910 method using an UVmini 1240
Shimadzu UVevisible spectrophotometer.
2.5. Equilibrium isotherm experiments
Equilibrium isotherm experiments were conducted to investi-
gate the capacities of the resins to remove the residual OM from
refinery wastewater through adsorption and IEX mechanisms,
allowing the most efficient resins to be identified and tested in
columns. Data collected in these tests were also compared to GAC
isotherm performances. The Langmuir and Freundlich models were
also verified.
Sixteen experiments were conducted, as each resin had two
isotherms, one for each refinery wastewater. They were performed
364 R. de Abreu Domingos, F.V. da Fonseca / Journal of Environmental Management 214 (2018) 362e369
Table 1
Characteristics of evaluated resins.
Resin Type Polymer structurea
R1 e R5 Adsorbent S-DVB
R6 Adsorbent Chloromethyl S-DVB
R7 Anionic S-DVB
R8 Anionic S-acrylate-DVB
a
S-DVB: styrene - divinylbenzene.
by dosing the effluent with 0.05, 0.1, 0.2, 0.4, 0.6, 0.8, 1.2 and 2.0 g of


resin per liter at 170 rpm in a Nova Etica shaker at 25  C until
equilibrium was achieved. GAC isotherms for each refinery waste-
water were obtained followed similar procedures to allow a com-
parison of the isotherms of the resins, using ASTM D3860-98
(2008) as a reference.
2.6. Adsorption and regeneration column experiments
Packed-bed column experiments were performed to evaluate
the adsorption and IEX performances of the resins by extracting
their breakthrough curves in order to verify their regeneration
capacity, and to test them through several treatment cycles to
confirm their potential as regenerable resins. Resin regeneration
should be understood to be the process of removing the retained
OM by desorption and IEX through the passage of a regenerant
solution.
During the adsorption/IEX step, herein called the “operation
step” or “operation cycle,” the effluent was continuously fed from
the top of the column using a dosing pump. The samples were
collected from the bottom of the column and an automatic sam-
pling valve was used when the column was not directly monitored.
The regenerant solution was also fed from the top of the column
during the regeneration/desorption process at flowrates defined
according to the resin manufacturer's lab-scale procedures. The
regeneration step was preceded and followed by deionized water
rinses.
The glass column used in the laboratory experiments was
designed and optimized in order to avoid radial and axial disper-
sion effects, liquid maldistribution, and inadequate flow through
the bed, while simultaneously seeking to reduce sample con-
sumption, which was achieved by using smaller diameters. A
schematic draw of the column set-up used in the experiments is
present as supplementary material. As a result, an internal column
diameter of 12.6 mm and a bed height of 500 mm were chosen,
corresponding to a bed volume (BV) of 64 mL. Correlations based on
bed parameters and flow operational conditions were extracted
from Delgado (2006), Inglezakis (2010), Inglezakis and Poulopoulos
(2006), and Thomas and Crittenden (1998) to achieve the desired
glass column dimensions.
It should be noted that all experiments to evaluate resin effi-
ciencies were performed with a single column, giving a lower
empty bed contact time (EBCT) compared to industrial scale sys-
tems, where it is common to find adsorption beds in series to allow
higher EBCT and lower outlet OM.
2.7. Column experiment 1
This experiment focused on acquiring the breakthrough curve
for the selected resin, R8, in a 72 h experiment and on determining
the efficiency of the regeneration process using a 10% NaCl solution
as a regenerant after the end of the operation. Experiment duration
of 72 h was defined according to the resin limitation of a maximum
of 3 days of operation without regeneration, which aims to avoid
irreversible adsorption.
Functional Group Operation flow (BV/h) Regenerant chemical
None 5e20 NaOH or HCl or ethanol
Amine 2e5 NaOH or HCl
Amine 5e40 NaCl or NaOH
Amine 7e20 NaCl and NaOH
The high ionic strength caused by excess NaCl in water during
the regeneration procedure causes the resin to contract and release
the retained OM particles.
According to the manufacturer recommendation, the flowrate
through resin R8 should be from 8 BV h
1 to 20 BV h
1. For this
experiment, a mean value of 13e14 BV h
1 was chosen.
2.8. Column experiment 2
This experiment was similar to the previous one, differing only
in the type and flow of the regenerant chemicals used. The 10% NaCl
solution was supplemented with 2% NaOH solution in an attempt to
improve the regeneration efficiency. NaOH increases the pH and
assists in hydrolysis of organics and ion exchange, facilitating the
breaking of bonds formed between the resin and the OM molecules.
Furthermore, a lower regenerant flow rate was applied to in-
crease the EBCT between the resin and the chemical solution dur-
ing the regeneration step.
2.9. Column experiment 3
The third column experiment evaluated the performance of the
resins during consecutive 24 h of test, aiming to check if interme-
diate regenerations allowed the recovery of the initial adsorption
capacity and if the efficiency improved due to the reduced opera-
tional cycle duration being less stressful to the resins. The objective
of these smaller tests was to verify if irreversible adsorption was
occurring in the first two column experiments due to the longer
tests lengths.
In this case, the influent sample TOC was smaller than in column
experiments 1 and 2. Another change was the effluent flow rate
during operation, which was decreased to the lowest level sug-
gested by the manufacturer, 8 BV h
1, in order to improve the EBCT
and favor adsorption.
2.10. Theoretical models
There are many solutions for determining the adsorbate mass
balance in a fluid flowing through a packed bed, which depend on
the process conditions and the adsorbate-adsorbent groups. Two
different solutions were used from Richardson and Harker (2002).
Equation (1) considers that equilibrium is achieved between the
adsorbate in the fluid and in the solid, and that the bed operates
isothermally. On the other hand, equation (2) treats the bed as
isothermal without reaching equilibrium. According to Ruthven
(1984), small-scale columns approximate isothermal conditions.
In most environment-related applications of adsorption treatment,
such as IEX and catalysis, species are in such low concentrations
that the effect of the heat of adsorption on the temperature can be
neglected (Inglezakis and Poulopoulos, 2006).
ut
¼ z
z0 (1)
1 þ m1i f 0 ðCÞ
 R. de Abreu Domingos, F.V. da Fonseca / Journal of Environmental Management 214 (2018) 362e369 365

8 2  39 these results for GAC, it was determined that performing adsorp-

> 3t
1 >
1< = tion/IEX column experiments with resins was of the utmost
C 6 2q 7
¼ 1 þ erf 4 rffiffiffiffiffiffiffiffiffiffiffiffiffiffi
h i 5 (2) importance, since their application would only be feasible
C0 2 >
: 1þ5j >
;
2 compared to GAC if regenerations were efficient enough to allow a
5q
longer service lifetime in order to compensate for their higher in-
In these equations, t is the time; u is the velocity of the particles; vestment and the extra operating costs spent on regenerative
C and C0 are the concentrations at position z and z0 (initial) along chemicals.
the bed, respectively; mi is a bed porosity function, f(C) is the The isotherm data was fitted to the Langmuir and Freundlich
representative function of the considered isotherm; and t, q, and j models. For both refineries, a much higher R-squared value was
are functions of effective diffusivity, time, position, velocity, bed achieved for the Freundlich model compared to the Langmuir
porosity, isotherm, mass transfer coefficient, and particle model. The improved fit of the Freundlich model was expected as it
dimensions. considers the heterogeneity of the active sites of the adsorbent and
the formation of many layers of adsorbate on the adsorbent surface,
whereas the Langmuir model does not. The Freundlich parameters,
3. Results and discussion
Kn, 1/n (n
1), and the adsorption capacity at equilibrium at the
initial sample concentration, (qe)C0, that were obtained for the ef-
3.1. Equilibrium isotherms experiments
fluents of both refineries and the data was presented in
supplementary materials. A better result for the Freundlich model
Adsorption isotherms of the evaluated resins and GAC for the
was also obtained by Li et al. (2004) for polymeric resins in non-
effluents from refineries A and B are shown in Fig. 1, where qe is the
aqueous solutions.
adsorption capacity and Ce is the equilibrium concentration.
The Freundlich parameters obtained for resin R6 indicate that it
Resin R6 showed a slight advantage over the other resins at
possessed a good adsorption capacity. However, R6 had a low
lower equilibrium concentrations; however, it had a reduced ca-
(qe)C0, which is an important parameter for column performance.
pacity at higher concentrations. Resins R5 and R8 also had good
For this parameter, the best resin was R8. Therefore, resin R8 was
results compared to the other five resins. The highest adsorptive
chosen for column experiments since it had one of the best
capacities for the initial sample conditions, Ce ¼ C0, were obtained
isotherm profiles as well as the best performance in terms of (qe)C0.
for resins R5 and R8.
Additionally, resin R8 changed its color as it was contaminated
GAC was favorable and it had a much higher adsorption capacity
(resin R6 maintained its color), allowing observation of the prog-
across the entire concentration range that was monitored. Based on
ress of contamination through the bed.

3.2. Column experiment 1

The adsorption/IEX curve obtained from column experiment 1,

Fig. 1. Resins and GAC isotherm and comparison with resins e Refinery A (a) and
Refinery B (b). Experimental conditions: Resin concentration (g/L) 0.05, 0.1, 0.2, 0.4, 0.6,
0.8, 1.2 and 2.0, T ¼ 25  C, 170 rpm.
also called a breakthrough curve, is shown in Fig. 3a. It displays the
ratio of the treated effluent TOC to the initial value (C/C0) as a
function of bed volumes treated. Operational data for the first two
column experiments are displayed in Table 2. All column experi-
ments tested resin R8 and used a bed volume of 64 mL (diam-
eter ¼ 12.6 mm and height ¼ 500 mm).
A rapid increase in the treated effluent TOC occurs up to about
200 BV, when the curve reaches an almost constant baseline. This
result does not necessarily indicate a rapid saturation of the bed,
since it keeps adsorbing part of the OM. During this initial increase,
certain recalcitrant organic species that are less favorably retained
through adsorption and IEX rapidly reach the bottom of the bed,
whereas other organic species are more easily retained.
Touma (2013a) presented pilot scale and prototype GAC col-
umns with similar EBCT and effluent TOC levels, resulting in a C/C0
ratio of 0.80 for 900 BV of consumed effluent for a single vessel. The
consumption of the same volume of effluent in present work
experiment resulted in a maximum of C/C0 ¼ 0.62. This shows that
resin R8 had a similar or even better TOC removal efficiency than
GAC.
Fig. 2 shows the regeneration curve (outlet TOC) and the
retained and removed OM from the resins bed during the regen-
eration process. The curve reached the expected profile and TOC
values were almost 4000 mg L
1 at the desorption peak. The
regeneration data for the first two column experiments are shown
in Table 3.
At the end of the experiment, after almost 12 BV of regenerant
consumption, the desorption/regeneration process had removed
around 68% of the initially retained OM on the resin, as determined
by graphical integration, and 1250 g of NaCl per liter of resin had
been consumed. However, the regeneration could have ended after
4 BV of regenerant consumption, when the outlet TOC had already
366 R. de Abreu Domingos, F.V. da Fonseca / Journal of Environmental Management 214 (2018) 362e369

Table 2
Column experiments 1 and 2 e Operational data.

Column Experiment 1 2
Operation cycle Single Single
Fluid Refinery B effluent Refinery B effluent
Average initial TOC (mg/L) 18.0 17.1
Average flow (BV/h) 12.7 13.0
EBCT (min) 4.7 4.6
Operation cycle duration (h) 68.6 65.7

3.3. Column experiment 2

As expected, the shape of the column experiment 2 break-

through curve was very similar to that seen for experiment 1, since
both tests used the same sample and operational parameters.
Fig. 3a shows that C/C0 did not exceed 0.65 (outlet
TOC ¼ 11.0 mg L
1). As described for experiment 1, the TOC level
stabilized as the treated effluent reached 200 BV. Integrating the
TOC curve showed that about 46% of the total effluent OM was
retained in the resin bed, similar to the amount seen in experiment
1, where 45% of the OM was adsorbed or ion-exchanged.
OM removal in the second regeneration experiment (Fig. 3b)
was better than seen previously, with 80% of the total OM retained
in the resin bed at the beginning of the regeneration procedure
being removed, 12% more than seen for regeneration with only 10%
NaCl, as performed in experiment 1. Even though the total OM
passed through the column was approximately 8% lower than for
experiment 1, which could have contributed to the reduced irre-
versible adsorption in this experiment, the better results could also
have been favored by the greater EBCT due to the decreased
regenerant flow rates, as well as the additional use of NaOH. The
regeneration consumption was the same as seen in column
experiment 1.

3.4. Column experiment 3 e consecutive operation cycles

Fig. 2. Operational breakthrough curve (a) and regeneration curve (b) e Column
experiment 1. Experimental conditions: Average flow ¼ 12.7 BV/h; EBCT ¼ 4.7 min;
Operation cycle duration ¼ 68.6 h.
stabilized at low levels. At this point, the regeneration level reached
430 g of NaCl per liter of resin. This level of regenerant consumption
was 79% above the 240 g NaCl L
1 recommended by the manufac-
turer in their procedure for the regeneration of resin R8. This
discrepancy may be possible due to the presence of recalcitrant
organic compounds in this effluent, which are more difficult to
desorb than the natural OMs (NOMs) commonly found in surface
water and used by the manufacturer as a reference for resin R8 data.
Fig. 4 shows the outlet TOC values for all operation cycles during
column experiment 3. These values did not exceed 9.3 mg L
1 in all
eight operation cycles, which is equivalent to a maximum C/C0 of
0.61. However, the curves show a tendency for TOC to increase
through the tests, especially in the last three, indicating gradual bed
saturation. The operational and regeneration data obtained from
this experiment are displayed in Table 4 and Table 5, respectively.
Further important evidence is that the breakthrough curves
always began at TOC values above zero, indicating that some OM
compounds had a low tendency towards adsorption or IEX, even
when the bed was not yet contaminated. This portion corre-
sponded to approximately 24e31% of the refinery wastewater OM.
The presence of non-adsorbable compounds was also shown for
GAC by Touma (2013b), where values from 10 to 20% were found.
OM removal efficiency ranged from 57 to 94% (76.1% ± 13.8%) for
the eight regenerations carried out in experiment 3. This does not
allow one to conclude if reducing the operation cycle duration from
72 h to 24 h decreased irreversible adsorption.

Table 3
Column experiments 1 and 2 e Regeneration step data.

Column Experiment 1 2
Operation cycle Single Single
Fluid/Solution NaCl 10% NaCl 10% and NaOH 2%
Average flow (BV/h) 3.6 2.4
EBCT (min) 16.7 25.2
Regeneration duration (min) 211 340
Regenerant volume consumed (mL) 806 850
 R. de Abreu Domingos, F.V. da Fonseca / Journal of Environmental Management 214 (2018) 362e369 367

These values are smaller than those obtained in the two previous
experiments, but are up to 66% higher than the value provided by
the manufacturer for NOM removal, 240 g NaCl L
1.

3.5. Application of theoretical models

As shown in Fig. 5, the theoretical breakthrough curves calcu-

lated from equations (1) and (2) were compared to the data from
the first operation cycle of experiment 3, since only the first oper-
ation cycle started with a fresh non-contaminated resin.
As can be seen, neither curve agreed with the experimental data.
However, the curve shape under isothermal and equilibrium con-
ditions (obtained using equation (1)) is similar to the experimental
data, since it includes a breakthrough from the starting time. The
disagreement between equation (1) and the experimental data is
primarily caused by the theoretical model not taking into account
non-adsorbable species, which correspond to about 25% of the total
TOC.
In order to adapt the model results to the data obtained in
experiment 3, it was necessary to add the non-adsorbable OM to
the model results. The result of this adjustment for the first oper-
ation cycle of experiment 3 is shown in Fig. 6. According to the
figure, the data for the first operation cycle were in good agreement
with the model when the estimated non-adsorbable OM portion
was added, allowing the results to be used to predict the total
operation cycles for resin as described in the next section.

3.6. Estimate of GAC and resin operation cycles

Fig. 3. Adsorption breakthrough curve (a) and regeneration curve (b) e Column
experiment 2 and comparison with column experiment 1. Experimental conditions:
Average flow ¼ 13 BV/h; EBCT ¼ 4.6 min; Operation cycle duration ¼ 65.7 h.
The average regeneration level obtained in experiment 3 was
338 ± 50 g NaCl L
1 resin, ranging from 270 to 400 g NaCl L
1 resin.
In order to estimate the maximum GAC and resin operation
cycles before saturation, it is necessary to define a breakpoint of
interest. This definition depends on the maximum TOC accepted by
the input specifications of downstream processes, usually for
removal of dissolved solids in tertiary treatment.

Fig. 4. Operation breakthrough curves e column experiment 3. Experimental conditions: Average flow ¼ 7.1e8.3 BV/h; EBCT ¼ 7.2e8.5 min; Operation cycle duration ¼ 21.6e26.5 h.

Table 4
Column experiment 3 e Operational data.

Operation cycle 1 2 3 4 5 6 7 8
Fluid Refinery B effluent
Average initial TOC (mg/L) 14.3 14.3 14.2 13.9 13.9 14.4 14.5 15.3
Average flow (BV/h) 7.8 7.6 8.3 7.1 7.2 7.5 7.4 7.4
EBCT (min) 7.7 7.9 7.2 8.5 8.3 8.0 8.1 8.1
Operation cycle duration (h) 23.6 21.6 24.5 23.7 26.5 24.4 23.5 23.8
Final TOC outlet value (mg/L) 7.3 7.1 6.9 7.2 8.1 8.6 8.6 9.3
368 R. de Abreu Domingos, F.V. da Fonseca / Journal of Environmental Management 214 (2018) 362e369

Table 5
Column experiment 3 e Regeneration data.

Operation cycle 1 2 3 4 5 6 7 8
Fluid/Solution NaCl 10% and NaOH 2%
Average flow (BV/h) 2.0 2.3 2.2 2.1 2,0 2.0 2.2 2.1
EBCT (min) 29.4 26.7 27.3 28.9 29.6 29,3 27.5 28.7
Regeneration duration (min) 139 110 131 108 128 100 99 127
Regenerant volume consumed (mL) 304 283 308 240 278 219 231 284

Fig. 5. Theoretical breakthrough curves e column experiment 3. Experimental conditions: Average flow ¼ 7.1e8.3 BV/h; EBCT ¼ 7.2e8.5 min; Operation cycle
duration ¼ 21.6e26.5 h.

Considering that the average outlet TOC for all column experi-
ments performed after achieving a constant baseline was around
7e11 mg L
1, RO systems could not be used in downstream resins
beds as most RO membranes do not tolerate TOC values above
3e5 mg L
1 without an extensive study of the impact of the TOC on
the membranes. Moreover, normal IEX systems would also expe-
rience high contamination on anionic resins. On the other hand,
EDR membranes can cope with TOC of up to 15 mg L
1 (GE, 2013),
and could be used as a dissolved salt removal process after TOC
removal using resin R8 beds.
In this work, only experiment 3 ran consecutive operation cycles
that could be used to estimate the maximum number of operation
cycles the resin could withstand until the breakpoint is reached. A
breakpoint of about C/C0 ¼ 0.70 will be used as a safe estimate,
which represents a treated effluent with a TOC of 10 mg L
1. This
value is similar to the target TOC used by Touma (2013b) for TOC
removal by adsorption using GAC beds.
Since GAC performance was not tested through column
experiments during this study, results were gathered from the
literature for comparison with resin R8. Touma (2013a) tested GAC
with Brazilian refinery wastewater in rapid small-scale column
tests (RSSCT) with the same TOC level as used in experiment 3 of
this study and a similar EBCT. To reach C/C0 ¼ 0.70 with a single
column, Touma (2013a) consumed about 328 BV of effluent.
An estimate of the total number of resin operation cycles could
not be made using the data from experiment 3 as it had not ach-
ieved the breakpoint of C/C0 ¼ 0.70. Therefore, the total number of
operation cycles was estimated using the theoretical model from
equation (1), adjusted to consider the non-adsorbable OM as
described in Section 3.6.
As shown in Fig. 6, there was a discrepancy between the
experimental data for the eight operation cycles and the adjusted
model curve for the first operation cycle due to bed saturation, with
this discrepancy being especially strong after the fourth adsorption
curve. The discrepancy in the retained OM for each operation cycle
allowed a correlation between the retained OM per operation cycle

Fig. 6. Adjusted equilibrium and isothermal curve e column experiment 3. Experimental conditions: Average flow ¼ 7.1e8.3 BV/h; EBCT ¼ 7.2e8.5 min; Operation cycle
duration ¼ 21.6e26.5 h.
 R. de Abreu Domingos, F.V. da Fonseca / Journal of Environmental Management 214 (2018) 362e369
and the total processed effluent to be determined. This correlation
was used to predict the total effluent consumption if the breakpoint
of C/C0 ¼ 0.70 had been reached, giving a total operation cycle of
2410 BV for the resin. Thus, when running 24 h operation cycles, it
would be possible to operate the bed thirteen times before reaching
the breakpoint and resin saturation, which is seven times longer
than the GAC operation cycle according to Touma (2013a).
This number may be even higher if more effective regenerations
could be conducted to extend the life of the resin, for example by
using higher temperatures or higher regenerant concentrations.
Considering that the initial investment cost for resins is higher than
for GAC and that the operational costs for regenerations are also not
negligible, resins should provide a high number of operation cycles
before reaching their breakpoint in order to be more attractive than
GAC.
4. Conclusions
The removal of organic matter from wastewater from two pe-
troleum refineries using adsorbent and IEX resins was studied. In
isotherm experiments, Freundlich model was fitted to determine
the adsorptive capacity of the resins IEX and GAC for both re-
fineries. However, the selected resin R8 had good performance in
column experiments, with literature comparison showing that the
TOC removal efficiency of the resin when operated in a continuous
modus was equal to or even better than that of GAC. Isotherms and
column experiments also showed that some recalcitrant organic
compounds have low or no tendency to be retained on resins. The
amount of such materials was found to vary from 24 to 31%. These
compounds also have high resistance to desorption, requiring
higher regeneration EBCT and regenerant consumption.
The regeneration efficiencies for all conditions used in this study
were lower than would be expected to prevent rapid saturation of
resins, ranging from 57 to 94%, showing the great resistance of the
recalcitrant compounds to desorption. Different conditions, such as
varying the temperature and regenerant concentrations, should be
considered in future studies. The service life of the resin when using
intermediate regeneration conditions was estimated to be seven
times higher than that of GAC when the latter is not regenerated.
Resins must provide a high number of operation cycles before
reaching their breakpoint to be more attractive than GAC, since
they have higher investment and operational costs than GAC.
Furthermore, the only dissolved solids removal process capable
of handling the TOC outlet from the columns is EDR. Most RO
systems do not tolerate TOC values above 3e5 mg L
1 without an
extensive study of their impacts on the membranes, and normal IEX
systems would expect high contamination levels on anionic resins.
Appendix A. Supplementary data
Supplementary data related to this article can be found at
https://doi.org/10.1016/j.jenvman.2018.03.022.
369
References
Amorim, R.S., 2005. Abastecimento de a
gua de uma refinaria de petro
leo: Caso
REPLAN. Masters Thesis. Federal Fluminense University, Rio de Janeiro.
ANA, 2013. Age

^ncia Nacional de Aguas. Conjuntura dos Recursos Hídricos no Brasil,
Brasilia.
APHA, 2005. Standard Methods for the Examination of Water and Wastewater.
American Public Health Association/American Water Works. Association/Water
Environment Federation, Washington DC.
ASTM, 2004. Standard Test Methods for Physical and Chemical Properties of Par-
ticulate Ion-exchange Resins (West Conshohocken, Pensilv^ ania).
ASTM, 2008. Standard Practice for Determination of Adsorptive Capacity of Acti-
vated Carbon by Aqueous Phase Isotherm Technique (West Conshohocken,
Pensilva^nia).
Delgado, J.M.P.Q., 2006. A critical review of dispersion in packed beds. Heat. Mass
Transf. 42, 279e310. https://doi.org/10.1007/s00231-005-0019-0.
General Eletric Power & Water, 2013. Electrodialysis Reversal Fact Sheets - GE 2020
EDR Systems. http://www.gewater.com/products/electrodialysis-reversal-edr.
html (Accessed 05.06.17).
Inglezakis, V.J., 2010. Ion Exchange and adsorption fixed bed operations for
wastewater treatment e Part II: scale-up and approximate design methods.
J Eng Stud Res 16, 42e50.
Inglezakis, V.J., Poulopoulos, S.G., 2006. Adsorption, Ion Exchange and Catalysis e
Design of Operations and Environmental Applications, first ed. Elsevier B.V,
Amsterdan.
IPIECA, 2010. Petroleum Refining Water/wastewater Use and Management - Op-
erations Best Practice Series. http://www.ipieca.org/publication/petroleum-
refining-water-wastewater-use-and-management/ (Accessed 05.07.17).
Li, H., Xu, M., Shi, Z., He, B., 2004. Isotherm analysis of phenol adsorption on
polymeric adsorbents from nonaqueous solution. J. Colloid Interface Sci. 271,
47e54.
Nacheva, P.M., 2011. Water Management in the Petroleum Refining Industry, Water
Conservation. http://www.intechopen.com/books/water-conservation/water-
management-in-the-petroleum-refining-industry/ (Accessed 15.02.17).
PETROBRAS, 2015. Elevamos reúso de a
gua em 3 bilho ~ es de litros em nossas refi-
narias no Brasil. http://www.petrobras.com.br/fatos-e-dados/elevamos-reuso-
de-agua-em-3-bilhoes-de-litros-em-nossas-refinarias-no-brasil.htm (Accessed
05.06.17).
Pombo, F.R., 2011. Gest~ao da demanda de a
gua na indústria de refino de petro
leo:
desafios e oportunidades de racionalizaç~ ao. Ph.D. Thesis. COPPE, Federal Uni-
versity of Rio de Janeiro, Rio de Janeiro.
Richardson, J.F., Harker, J.H., 2002. Chemical Engineering e Particle Technology and
Separation Processes, fifth ed. Butterworth-Heinemann, Oxford.
Ruthven, D.M., 1984. Principles of Adsorption and Adsorption Processes, first ed.
John Wiley & Sons, Inc., New Jersey.
Sufnarski, M.D., 1999. The Regeneration of Granular Activated Carbon Using Hy-
drothermal Technology. Masters thesis. University of Texas, Austin.
Tchobanoglous, G., Burton, F.L., Stensel, H.D., 2003. Wastewater Engineering:
Treatment and Reuse, fifth ed. McGraw-Hill Companies, New York.
The Dow Chemical Company, 2014. Dow Ion Exchange Resins. Measuring the
Density. http://dowac.custhelp.com/app/answers/detail/a_id/147/ (Accessed
05.06.17).
Thomas, W.J., Crittenden, B., 1998. Adsorption Technology and Design, first ed.
Elsevier Science & Technology Books, Oxford.
Touma, S.L., 2013a. Testes acelerados em colunas de pequena escala: o uso de carva ~o
ativado granular em projetos de reúso de efluentes de refinarias de petro
leo.
Masters Thesis. Federal University of Rio de Janeiro, Rio de Janeiro.
Touma, S.L., 2013b. Cost Analysis of A Granular activated carbon plant design based
on RSSCT: a case study for reuse of wastewater in a Brazilian petroleum re-
finery. In: Annual World Conference on Carbon. Rio de Janeiro.
Xu, Z., Zhang, Q., Fang, H.H.F., 2003. Applications of porous resin sorbents in in-
dustrial wastewater treatment and resource recovery. Crit. Rev. Environ. Sci.
Technol. 33, 363e389. https://doi.org/10.1080/10643380390249512.
Yang, R.T., 2013. Adsorbents: Fundamentals and Applications, first ed. John Wiley &
Sons, Inc., New Jersey.