Accepted Manuscript

Cleaning of ceramic membranes for produced water filtration

T. Zsirai, H. Qiblawey, P. Buzatu, M. Al-Marri, S.J. Judd

PII: S0920-4105(18)30222-5
DOI: 10.1016/j.petrol.2018.03.036
Reference: PETROL 4777

To appear in: Journal of Petroleum Science and Engineering

Received Date: 28 May 2017

Revised Date: 5 March 2018
Accepted Date: 6 March 2018

Please cite this article as: Zsirai, T., Qiblawey, H., Buzatu, P., Al-Marri, M., Judd, S.J., Cleaning of
ceramic membranes for produced water filtration, Journal of Petroleum Science and Engineering (2018),
doi: 10.1016/j.petrol.2018.03.036.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
Zsirai et al, Cleaning of ceramic membranes for produced water filtration
ACCEPTED MANUSCRIPT
Cleaning of ceramic membranes for produced water filtration
Zsirai, T.1, Qiblawey, H.2, Buzatu, P.1, Al-Marri, M.1,2, and Judd, S.J.1,3
1
Gas Processing Center, Qatar University; 2Department of Chemical Engineering, Qatar
University. 3Cranfield Water Science Institute, Cranfield University.

Abstract
The application of ceramic microfiltration membranes to the tertiary treatment of produced

PT
water from an Arabian Gulf oilfield has been studied using a dedicated pilot plant. Studies
were based on a previously published protocol in which the retentate stream was recycled so
as to successively increase the feed concentration throughout the experimental run. Chemical
cleaning in place (CIP) was applied between each run and the flux and permeability recovery

RI
recorded for various cleaning protocols studied, the CIP being based on the combination of
caustic soda (NaOH) and citric acid. Surface analysis of the membrane, and specifically its
hydrophilicity, was also conducted.

SC
Results indicated the main influencing factor on permeability recovery from the CIP to be the
employment of backflushing during the CIP itself. A final flux of 700 L m-2 h-1 was sustained

U
through the application of 6 wt% NaOH with 6 wt% citric acid combined with backflushing
at approximately twice the rate of the filtration cycle flux. A consideration of the impact of
AN
this flux value on the viability of two commercially-available ceramic membrane
technologies indicated the footprint incurred to be slightly lower than that of the upstream
induced gas flotation technology and corroborated a previously published estimate. The flux
was sustained despite surface analysis indicating a loss of the innate hydrophilicity of the
M

ceramic membrane.

Keywords Produced water; ceramic membranes; chemical cleaning; footprint; flux.

D
TE

1 Introduction
Produced water (PW) generated from oil exploration requires rigorous removal of suspended
matter (free oil and particulate solids) as tertiary treatment (downstream of hydrocyclone and
EP

gas flotation) if it is to be desalinated for reuse (Alzahrani et al, 2014) or re-injected into low-
permeability reservoirs (Judd et al, 2014; Xu et al, 2016). The application of membrane
technology for this duty has been recently reviewed (Munirasu, 2016; Dickhout et al, 2017),
C

and the option of ceramic membrane filtration widely explored (Ebrahimi et al, 2010; Guirgis
et al, 2015; Weschenfelder, 2015, 2016).
AC

The viability of membrane processes generally is largely dependent on sustaining a high

membrane flux to minimise the process footprint, a particularly important attribute on
offshore oil platforms where available space is at a premium. The use of silicon carbide (SiC)
ceramic membranes for this duty has been demonstrated to provide a reliably high treated
water quality (6.3 - 7.6 mg/L oil and grease (O&G), 4 - 8 NTU turbidity) for microfiltration
(MF, pore size 2 µm) and ultrafiltration (UF, pore size 0.04 µm) membranes based on recent
pilot-scale studies (Zsirai et al, 2016) using real PW from an offshore oil platform. The
larger-pore MF membrane was shown to provide a significantly higher flux than the UF
membrane, but was also subject to greater flux (and permeability) decline. Moreover, it was
noted that there was a marked deterioration in both permeate water quality and permeability
with successive experimental runs on the chemically-cleaned membrane. The work

1
Zsirai et al, Cleaning of ceramic membranes for produced water filtration
ACCEPTED MANUSCRIPT
emphasised the need to improve the efficacy of the chemical clean-in-place (CIP) applied
between runs to recover both the permeability and selectivity of the membrane.

The current paper reports a series of trials conducted to identify the optimal protocol for
applying a CIP so as to sustain a viably high permeate flux. The impact on the technology
footprint was then determined, using a previously published protocol (Judd et al, 2014), based
on the specifications of two commercially-available candidate ceramic membrane filtration
systems for tertiary PW treatment.

2 Materials and methods

PT
The pilot plant operation methodology has been described previously (Zsirai et al, 2016). The
pilot plant (Fig. 1) and membrane elements were supplied by Liqtech International (Ballerup,
Denmark), the membranes having been pre-conditioned through their used for 15-17 trials in

RI
a previous campaign prior to the commencement of the current study. The plant comprised
two streams fitted with tubular membrane modules operated in crossflow mode. The plant,
allowed operation at a fixed conversion with automated physical cleaning provided by

SC
backflushing and manually-applied chemical cleaning in place (CIP). Experimental runs were
conducted with recirculation of the retentate (the concentrate stream from the membrane),
intended to increase the feedwater suspended oil and solids concentration over the course of

U
the run and so increase the challenge to the membrane. All runs were operated at a
transmembrane pressure of 0.55-0.6 bar with backflushing for 5 s every 600 s at a back-
AN
pressure of 3 bar and a filtration cycle conversion of 20%, these conditions having been
identified as being optimal in previous studies (Zsirai et al, 2016). The selection of the
cleaning reagents was similarly based on this study.
M

cooling b'pulse
D

Feed Membranes
PW tanks
TE

storage
Feed/permeate*
tank
Transfer
EP

pump
C

Permeate
Media-filtered/
collection Membrane
permeate*
tank pump
AC

Figure 1: Pilot plant

Experiments were conducted on single MF and UF SiC membrane elements (25 mm

diameter, 305 mm long, 37 channels of 3 mm square and with a total membrane area of 0.09
m2) which were chemically cleaned between runs. The MF and UF membranes were rated as
2 μm and 0.04 μm pore size respectively. The elements were challenged with PW shipped in
5-tonne batches from an oil platform operating in the Arabian Gulf, two different batches
being employed throughout the trials with the PW nitrogen-blanketed to suppress oxidation.

2
Zsirai et al, Cleaning of ceramic membranes for produced water filtration
ACCEPTED MANUSCRIPT
The PW was sampled downstream of the induced gas flotation (IGF) step of the PW
treatment train of the oil platform, the train being based on a classical two-stage
hydrocyclone-IGF process. The work was limited to eight runs conducted on the two
preconditioned membrane elements.

Experiments were conducted on 200 L feed volumes of PW with retentate recycling

throughout which served to agitate the feed by jet mixing during the experimental run.
Further jet mixing of the storage tanks was conducted prior to dispensing of the 200L
volumes to ensure an even dispersion of the oil and solids in the stock solution. Chemical
clean in place (CIP) studies were then carried out between each filtration run according to a

PT
multiple-stage protocol (Table 1), applying supplementary backflushing during the CIP for
the final three runs (c – e). The backflushing was designed to help ensure that any dislodged
solids were not immediately re-deposited onto the membrane surface by the flow of cleaning

RI
reagent. Reagents employed comprised an organic acid (citric acid), supplemented with
sulphuric acid to lower the pH to 2, and caustic soda at pH 13 (Table 1). Whilst both base-
acid and acid-base sequential cleaning were initially studied, no apparent consistent change

SC
was noted between the two sequences. The feedwater quality and CIP protocols employed for
the trials conducted based on two different batches of PW, denoted by trials 1-3 (Batch 1) and
a-e (Batch 2), are summarised in Table 2, the increase feed concentration arising from the
recirculation of the retentate.

Table 1: Cleaning-in-place procedure

U
AN
Reagent, ~50L aliquots Recirculation time and temperature
1. Mains water 30 min at ~25oC
2. 6% NaOH 30 min at pH 13
3. Mains water 30 min at ~25oC
M

4. 6% citric acid 30 min at pH 2

5. Mains water 30 min at ~25oC
D

Table 2: Feed water quality and CIP conditions

Feed O&G Feed turbidity Cleaning reagents Cleaning
TE

Run Start End Start End NaOH Citric acid sequence BF

1-3 24-25 32-84 21-34 39-74 2% 2% B-A N
a 52 148 123 202 2% 6% A-B N
b 40 76 99 158 6% 2% A-B N
EP

c 47 52 72 461 2% 2% A-B Y
d 50 95 408 955 2% 6% A-B Y
e 42 93 663 998 6% 2% A-B Y
C

Runs 1-3 refer to Batch 1, Runs a-e refer to Batch 2; B-A base followed by acid; A-B acid followed by base; BF backflush
during CIP.
AC

Surface analysis of the virgin and used membrane material was by contact angle
measurement, conducted according to standard methods (APHA, 2001), to assess the surface
hydrophobicity. SEM-EDAX measurements were also made to assess the extent of formation
of calcium, barium and iron alkaline scales. The autopsies were intended to provide an
indication of the degree of permanent fouling from oil and inorganic reagents. Analysis for
turbidity and oil and grease (O&G) was according to standard methods (SMWW 2130 B and
5520 C).

3 Results
The flux and permeability profiles followed the same trends as in previous studies (Zsirai et
al, 2016), with the flux decline over the filtration cycle and recovery during the backflush

3
Zsirai et al, Cleaning of ceramic membranes for produced water filtration
ACCEPTED MANUSCRIPT
cycle both decreasing with time (Fig. 2). As outlined in previous work, the end permeability
refers to the permeability resulting from the same hydraulic load and, if based on the same
PW batch, roughly the same pollutant load between tests. The start and end permeabilities
between tests (respectively the green and red bars in Figure 3) are therefore directly
comparable.

PT
RI
SC
(a)

U
AN
M
D

(b)
Figure 2: Typical filtration transient: flux and pressure transient from the first (a) and the last hours (b) of
the concentration test.
TE

Results indicate a significant difference in fouling propensity between the two batches. There
are a number of possible explanations for this, including an approximate doubling of the
EP

O&G content and more than order of magnitude increase in turbidity (Table 2). It is also
possible that Batch 2 had a higher concentration of organic matter of higher fouling
propensity. This would include production chemicals based on flocculent polymers, which
are known to foul polymeric membranes though are less onerous to ceramic materials.
C

However, the very significantly higher colloidal content, manifested as the turbidity, is the
most likely cause of the decreased end permeability from Run 3 to Run a for the MF
AC

membrane (Fig. 3b). This is corroborated by the far less significant corresponding decrease
for the UF membrane (Fig. 3a), which would be expected to be more resistant to colloidal
fouling.

Notwithstanding differences in fouling propensity of the two batches the greatest impact on
sustaining permeability is from the application of backflushing during the CIP (Runs c-e) for
both the UF and the MF membranes, although for the final UF run the CIP failed to recover
the permeability (Fig. 3a). In the case of the MF a final permeability of between 1190 to 1280
LMH/bar was sustained for the three runs (Fig. 3b), with the CIP recovering the permeability
to 1480-1630 LMH/bar for the start of the following run. This compares with final
permeabilities of 210-310 LMH/bar for Runs a and b, prior to the introduction of
backflushing during the CIP. In the case of the UF membranes the corresponding final

4
Zsirai et al, Cleaning of ceramic membranes for produced water filtration
ACCEPTED MANUSCRIPT
permeability values were 367-384 LMH/bar for the backflushed CIP tests compared with
202-232 LMH/bar prior to the introduction backflushing during the CIP. The increased
efficacy of the backflushed CIP was thus almost four times greater for the MF membrane
compared to the UF.

1600
Start End
1400
Batch 2

PT
1200
Permeability, LMH/bar

1000 Backflushing w. CIP

RI
800

SC
600

400

U
200
AN
0
1 2 3 a b c d e
Run
(a)
M

7000
Start End
D

6000
Batch 2
TE

5000
Permeability, LMH/bar

Backflushing w. CIP
EP

4000

3000
C

2000
AC

1000

0
1 2 3 a b c d e
Run

(b)
Figure 3: Permeability trend with successive experimental runs (a) UF, and (b) MF

A mean treated water quality of <7.1 ± 1.6 mg/L O&G (oil and grease) in for the UF
membrane and 8.5 ± 1.6 mg/L for the MF was maintained throughout. This arose despite the
increase in feed water O&G concentration and the impact of recycling of the retentate, and

5
Zsirai et al, Cleaning of ceramic membranes for produced water filtration
ACCEPTED MANUSCRIPT
was in keeping with the outcomes of the previous study employing the same protocol (Zsirai
et al, 2016). The use of the backflush during the CIP also appeared to stabilise the permeate
turbidity to <25 NTU, corresponding to >98% removal which is assumed to take place at the
membrane surface. The rejection provided by the MF membrane following the backflushed
CIP was slightly higher than that of the UF for both turbidity and O&G, which is counter-
intuitive given the significantly higher selectivity expected for the UF membrane given the
rated pore size.

The membrane autopsy indicated only negligible quantities of inorganic scalants and metal
hydroxides (calcium and iron) on the membrane surface for both the MF and UF membranes,

PT
as might be expected given the strength and rigour of the acid cleaning sequence of the CIP.
This was despite the PW being heavily supersaturated with hardness (between 9 and 15 g L-1
as CaCO3). However, the surface was evidently affected by contact with the oil, with the

RI
contact angle measurements (θ = 106-116°) indicating significant hydrophobicity for the used
membranes, compared with complete hydrophilicity (θ = 0°) for the virgin membrane
material. It can therefore be surmised that the decline in permeability is primarily attributable

SC
to the oil content of the PW, rather than the inorganic scales. Since the membranes had
already been used for a number of tests prior to the current study (Section 2) it is likely that
they in a hydrophobic state throughout the campaign.

U
Examination of the filtration transients for Runs c-d, where backflushing during the CIP was
applied, indicated a negligible permeability decline between backflush events for the final 5-7
AN
backflushes of each run. A final flux of around 700 LMH was correspondingly sustained for
the MF membrane, notwithstanding the loss of membrane surface hydrophilicity.
M

4 Discussion
There have a number of studies of the application of ceramic membrane technology to the
D

treatment of real petroleum industry wastewaters reported since 2010 (Table 3). Few of these
have been conducted at pilot scale, challenged with real oilfield produced water (OFPW),
TE

and/or operated for a significant period of time. An early site-based trial (Lee et al, 2005),
apparently employing a spiral-wound hydrophilic polymeric membrane, achieved a flux of
10-20 LMH and maximum permeabilities of 6-8 LMH/bar after an extended operating period
of 650 hours. Since then somewhat higher fluxes and permeabilities from similar site based
EP

trials using ceramic membranes have been reported, ranging from 60 (Pedenaud et al, 2011),
to more than 1000 LMH/bar (Prado‐Rubio et al, 2011) depending on the feedwater quality
and the state of the membrane (Table 3).
C

However, it is evident that the sustainable flux attainable is greatly dependent on the CIP
AC

efficacy. Permeability recoveries varying by factors of 2-3 have been reported from pilot
plant studies based on real PW (Prado‐Rubio et al, 2011), with significant permeability
reduction ensuing from insufficiently cleaned membranes. A conclusion common to all
studies based on real petroleum effluents is the low membrane permeability due to the
gradual build-up of hydrophobic matter on or within the membrane material. Enhanced
permeability recovery has been demonstrated both by adjustment of the physical cleaning
(i.e. backflushing) conditions during the filtration cycle (Silalahi and Leiknes, 2011) and by
sequential base-acid cleaning (Weschenfelder et al, 2015b). The current study suggests that
further improvement in permeability recovery can be attained through backflushing during
the CIP, yielding a final sustainable flux of around 700 LMH and a corresponding
permeability of ~1200 LMH/bar – somewhat higher than previously reported values (Table
3).

6
 ACCEPTED MANUSCRIPT

Zsirai et al, Cleaning of ceramic membranes for produced water filtration

Table 3: Studies of MF/UF membrane filtration of PW and other petroleum effluents

Fin perm
Pore Init flux, Fin flux, time,
Oil (TSS) concn mg/L, water Scale Material TMP, bar (range), T, ˚C
size µm LMH LMH h
source LMH/bar Reference
200-1000, tank dewatering effl b(m) AlO 0.2 128 28 1 28 2 60 Ebrahimi et al, 2010

PT
200-1000, tank dewatering effl b(m) TiO 0.05 80 4 1 <5 2.5 60 Ebrahimi et al, 2010
200-1000, tank dewatering effl b(m) 120 30 1 120-30 2 60 Ebrahimi et al, 2010

RI
OFPW p TiO, SiC 0.01-0.1 200 - 0.5-3.5 60 120 -
Pedenaud et al, 2011
0.25-1.5 135

SC
221-722 (7-17), OFPW p SiC 0.04-0.1 - 135-590 0.35-0.95 450-1020 26 -
Prado‐Rubio et al, 2012
20 (2.9), OFPW 0.6 520

U
3-25 (24-74), OFPW p SiC 0.1-0.5 25-120 50 0.3-1.5 150 12-14 45 CoMeTas, 2011
52-458, gas field PW p Al-Zr 0.05 - 170-255 - 190-240 600 25

AN
Subramani et al, 2011
-, SAGD effl p AlO 0.05 200 - 1.52 132 - 45 Guirgis et al, 2015
~250, refinery effl. b ZrO 0.1 1000 290 1.5 193 - 45 Weschenfelder et al, 2015

M
9-43, PW p AlO 0.2 - 295-312 2.5 118-125 - 35-60 Reyhania & Meighani, 2015
KEY LMH litres per m2 per hr; SAGD Steam-assisted gravity drainage; OFPW oilfield produced water; TMP transmembrane pressure; TSS total suspended solids.

D
b bench; b(m) bench, based on membrane module; p pilot
AlO aluminium oxide; SiC silicon carbide; TiO titanium oxide; ZrO zirconium oxide

TE
C EP
AC

7
Zsirai et al, Cleaning of ceramic membranes for produced water filtration
ACCEPTED MANUSCRIPT
There are three aspects to the technical viability of the process for the duty envisaged:
a. Attainment of required water quality,
b. Limitation to a maximum footprint, and
c. Impact on operation of other PW process technologies.
Determination of the footprint demands certain assumptions regarding process technology
design/configuration and the flow and loads on the oil platform respectively.

4.1 Water quality

It is evident from all measured water quality data for the membranes tested that a high treated

PT
water quality is sustained regardless of the membrane material or pore size, the residual oil
assumed to be in the dissolved form. Permeate quality values reported from other studies
suggest a total suspended solids (TSS) concentration below 5 mg/L, unless the membrane has

RI
been made irreversibly hydrophobic (Prado Rubia et al, 2011) causing breakthrough of free
oil. The permeate water is otherwise potentially appropriate for reinjection into the reservoir
or overboard discharge. The retentate stream would be expected to have a similar

SC
composition to that of the return stream of the upstream IGF, and could thus be managed in a
similar manner.

U
4.2 Footprint
The footprint of classical and tertiary PW treatment processes has been determined with
AN
reference to flow normalised against the area (FA) and volume (FV) occupied. FA thus takes
units of m/h: m3/h per square metre of floor area in m2. FV correspondingly has units of h-1, it
being the flow in m3/h per m3 volume occupied. A ceramic membrane skid based on
vertically-aligned 1 m modules has been determined to have an area and volume footprint of
M

~7 m/h and ~2.6 h-1 respectively based on literature information (Judd et al, 2014), consistent
with a recently-published study (Weschenfelder et al, 2016) in which a value of 6.5 m/h was
determined for FA. This value is comparable with the nutshell filter (NSF) alternative tertiary
D

technology in terms of floor area but more compact with respect to volume.
TE

To provide a more accurate estimation of the footprint demands reference to existing

commercial ceramic membrane technologies, of which there are two which have been
implemented for PW filtration: the Liqtech International M99 skid and the Veolia Water
EP

Technologies ROSS™ system based on the company’s Ceramem membranes. The Liqtech
technology comprises 99 x 250 mm diameter 0.328 m2 membrane area modules, compared
with the Veolia technology which has 52 x 142 mm dia. 10.5 m2 modules. These data
coupled with the skid dimensions for the respective technologies (Table 4) allow the packing
C

density, the membrane surface area per unit skid footprint or volume, to be determined. As
can be seen, the two technologies are similar in specification.
AC

Table 4: Technical specification of two commercial ceramic membrane filtration technologies

Skid dimensions, m Module Skid Packing density*
Technology L H W A, m2 No Am, m 2
m2/m2 m2/m3
Veolia 18.3 2.7 1.8 33.5 52 10.7 556 17 6.1
Liqtech 1.2 4.5 1.5 1.80 99 0.328 32.5 18 4.0
A: skid floor area; packing density: membrane surface area per unit skid footprint or volume
Am: membrane surface area. *With reference to skid.

Determination of the FA and FV values for the two technologies (Table 5) demands
assumptions regarding both the conversion of feedwater into permeate (taking account of the
use of permeate for backflushing) and the redundancy arising from the requirement for taking
a stream off line for periodic chemical cleaning. Redundancy can be assumed to be

8
Zsirai et al, Cleaning of ceramic membranes for produced water filtration
ACCEPTED MANUSCRIPT
manifested as three skids each providing 50% of the flow capacity (i.e. duty-duty-standby),
adding 50% to the footprint. Based on backflushing for 1% of the time at 150% of the
operating flux, the adjusted conversion is calculated as being 92.5% assuming a conversion
of 96% for the forward filtration cycle. The resultant net FA and FV values are calculated as
being 7.2-7.8 m/h and 1.7-2.6 h-1 respectively. These figures roughly corroborate the values
of 7.0 m/h and 2.6 h-1 for FA and FV originally estimated in the review article by Judd et al
(2014).

Table 5: Footprint determination, based on two commercial ceramic membrane filtration technologies
Corrected for Corrected for

PT
Uncorrected redundancy Backflushing data backflushing
Techno Flux FA FV FA FV Freq Flow Vol Conv- FA FV
-logy m/h m/h h-1 m/h h h/d m3/h m3 ersion* m/h h-1

RI
Veolia 0.7 11.6 4.2 7.8 2.8 0.24 584 140 92.5% 7.2 2.6
Liqtech 0.7 12.6 2.8 8.4 1.9 0.24 34.1 8.2 92.5% 7.8 1.7
Redundancy: three skids each providing 50% of the flow capacity (i.e. duty-duty-standby)

SC
Ancillary equipment excluded: assumed to be comparable for each technology
*Assumes filtration cycle conversion of 96%, based on the experimental data for a conversion of 20% for the
300 mm element, feedwater pre-concentrated by a factor of ~2.5, and two stages of 850 mm-long full-scale
modules.

4.3 Impact
U
AN
It would be expected that the treated water could either be discharged overboard, provided
the regulated discharge limits are based solely on suspended matter and/or total oil (of 15
mg/L or more in the case of the latter), or used for reinjection (PWRI). Whilst the high
permeate water quality generated is generally considered to be appropriate to PWRI, the
M

compatibility is dependent on the precise permeability and other characteristics of the

reservoir. These include the formation water chemistry, and specifically the likelihood of the
precipitation of supersaturated scalants.
D

In addition to the treated water, a further stream (the retentate) is created which comprises
TE

4.5% of the feed flow and >90% of the solids and organic load. This would need to be
managed in the same was as the concentrate streams from the upstream secondary treatment
processes (the HC and IGF steps). The impact of the additional concentrate stream will
EP

therefore be dependent on the capacity of the management system, but the additional
hydraulic load from the membrane return stream would be unlikely to add more than 10% to
the hydraulic load. This is based on the assumption that the HC and IGF concentrate streams
comprise around 5% of the feed flow. Further analysis of the impact of the backflush stream
C

demands flows and loads information of the PW as managed on an individual platform. A

further stream which requires management is the wastewater from the chemical clean, which
AC

makes up no more than 0.5% of the feed flow.

Finally, the rate of chemical consumption is determined both by the total membrane area and
the cleaning frequency (assumed to be daily). Chemical storage requirements depend upon
shipping costs and any logistical constraints, but given that the reagents would be stored as
concentrates the storage volume and associated footprint would be relatively low.

5 Conclusions
The operation of a pilot plant fitted with silicon carbide (SiC) microfiltration (MF) ceramic
membranes and fed with real produced water (PW) derived from an Arabian Gulf oilfield has
indicated that a flux of 700 LMH with a corresponding permeability of 1400 LMH/bar can be

9
Zsirai et al, Cleaning of ceramic membranes for produced water filtration
ACCEPTED MANUSCRIPT
sustained over a number of filtration cycles. This flux is attained through the application of
backflushing during the clean in place (CIP) between filtration cycles, and appears to be
sustained despite the tendency of the ceramic membrane to become hydrophobic as the
number of filtration cycles (and so exposure to the PW free oil) increases.

A consideration of two commercially-available SiC-based technologies, each fitted with

vertical membrane modules, indicates that a flow per unit membrane skid area footprint of
12-14 m/h (or 2.8-4.2 h-1 per unit volume occupancy) can be attained at the ~700 LMH flux
measured for the MF membrane. These values decreases to 7.2-7.8 m/h and 1.7-2.6 h-1
respectively if 50% redundancy is assumed and filtration cycle conversion and backflushing

PT
are taken into consideration which reduce the overall conversion to 92.5%. On this basis the
footprint incurred by the ceramic membrane technology is comparable to values previously
estimated, though this excludes ancillary equipment such as the valve and pump skids, the

RI
control panel and the cleaning in place holding tanks and related equipment. Whilst the
footprint and process efficacy, with respect to water purification capability, suggest the
process to be viable for offshore applications, benchmarking against the alternative media-

SC
based tertiary filtration process (the nutshell filter) is needed to assess the process viability
on-shore. Moreover, benchmarking should necessarily take account of the economic
implications of technology selection (i.e. a technoeconomic analysis). Planned future work in
this area will encompass a sensitivity analysis of individual operation determinants related to

U
the cleaning cycles of both technologies so as to assess the relative impact of enhanced
cleaning and flux (or filtration velocity) on overall cost with reference to reagent use and
AN
residuals generation and management.

Acknowledgement
M

This work was funded by Maersk Oil Qatar, whose help and support is gratefully
acknowledged, and was completed as part of project number QUEX-CENG-CHE-13-14-03,
D

Qatar University
TE

References
Alzahrani, S., Mohammad, A. W., Hilal, N., Abdullah, P., and Jaafar, O. (2013).
EP

Comparative study of NF and RO membranes in the treatment of produced water-part I:

Assessing water quality. Desalination, 315, 18-26.
APHA (2001). American Public Health Association, Standard methods for the examination
C

of water and wastewater, 22nd Ed, Washington, DC, USA.

Cometas (2011). Commissioning of the SemCoMem unit, Company report, Project no. 110-
AC

11.
Dickhout, J. M., Moreno, J., Biesheuvel, P. M., Boels, L., Lammertink, R. G. H., and de Vos,
W. M. (2017). Produced water treatment by membranes: A review from a colloidal
perspective. J. Coll. Interface Sci. 487, 523-534.
Ebrahimi, M., Willershausen, D., Ashaghi, K.S., Engel, L., Placido, L., Mund, P., Bolduan,
P., and Czermak (2010). Investigations on the use of different ceramic membranes for
efficient oil-field produced water treatment. Desalination, 250, 991-6.
Guirgis, A., Gay-de-Montella, R., and Faiz, R. (2015). Treatment of produced water streams
in SAGD processes using tubular ceramic membranes. Desalination, 358, 27-32.

10
Zsirai et al, Cleaning of ceramic membranes for produced water filtration
ACCEPTED MANUSCRIPT
Judd S., Qiblawey, H. Al-Marri, M. Clarkin, C., Watson, S., Ahmed, A., Bach, S. (2014). The
size and performance of offshore produced water oil-removal technologies for reinjection,
Sep. Purif. Technol. 134 241–246.
Lee, J.M., and Frankiewicz, T. (2005). Treatment of produced water with an ultrafiltration
(UF) membrane - A field trial. SPE Annual Technical Conference Proceedings
Munirasu, S., Haija, M. A., & Banat, F. (2016). Use of membrane technology for oil field and
refinery produced water treatment - A review. Process Safety Env. Protect. 100, 183-202.
Pedenaud, P., Heng, S., Evans, W., and Bieonneau, D. (2011). Ceramic membrane and core
pilot results for produced water, OTC-22371-PP, paper presented at Offshore Technology

PT
Conference, 4–6 Oct, Rio de Janeiro.
Prado‐Rubio, O.A., Cardona, D., Svendsen, T., and Yuan L. (2012). SiC membrane pilot
plant ultrafiltration test for produced water treatment, Ocelote field – Hocol (COLOMBIA).

RI
Company report, Liqtech.
Reyhani, A., and Mashhadi Meighani, H. (2016). Optimal operating conditions of micro- and

SC
ultra-filtration systems for produced-water purification: Taguchi method and economic
investigation. Desal. Water Treat. 57(42) 19642-19654.
Silalahi, S.H.D., and Leiknes, T., (2009). Cleaning strategies in ceramic microfiltration

U
membranes fouled by oil and particulate matter in produced water. Desalination 236 (1-3),
160-169.
AN
Subramani, A., Schlicher, R., Long, J., Yu, J., Lehman, S., and Jacangelo, J. G. (2011).
Recovery optimization of membrane processes for treatment of produced water with high
silica content. Desal. Water Treat. 36(1-3), 297-309.
M

Weschenfelder S.E., Louvisse A.M.T., Borges C.P., Meabe E., Izquierdo J., and Campos J.C.
(2015). Evaluation of ceramic membranes for oilfield produced water treatment aiming
reinjection in offshore units. J. Petrol. Sci. Eng. 131, 51-57.
D

Weschenfelder, S. E., Fonseca, M. J. C., Borges, C. P., and Campos, J. C. (2016). Application
of ceramic membranes for water management in offshore oil production platforms: Process
TE

design economics. Sep. Purif. Technol. 171, 214-220.

Xu, J., Ma, C., Cao, B., Bao, J., Sun, Y., Shi, W., & Yu, S. (2016). Pilot study on
hydrophilized PVDF membrane treating produced water from polymer flooding for reuse.
EP

Process Safety Env. Protect. 104, 564-570.

Zsirai, T., Al-Jamal A. K., Qiblawey, H., Al-Marri, M., Ahmed, A., Bach, S., Watson, S.,
Judd S (2016). Ceramic membrane filtration of produced water: impact of membrane module,
C

Sep. Purif. Technol., 165 214-221.

AC

11
Zsirai et al, Cleaning of ceramic membranes for produced water filtration
ACCEPTED MANUSCRIPT
Highlights

Pilot-scale testing of SiC ceramic membranes challenged with real produced water

Two membrane pore sizes tested and various chemical cleaning protocols assessed

Outcomes used to determine area and volumetric footprint of full-scale installation

PT
Backflushing during chemical cleaning found to be critical in recovering permeability

Calculated footprint values for optimised cleaning comparable to published estimates

RI
U SC
AN
M
D
TE
C EP
AC