Ceramic microfiltration – influence of pretreatment on operational performance

Since about 2000, there has been an increasing need to treat surface water for drinking water production and treat wastewater for reuse, for which colloid removal is a necessity. Micro- and ultra-filtration (MF and UF) are often used because they provide an absolute barrier against particles greater than the pore size. Polymeric membranes still dominate this sector of the water industry but ceramic membranes have some unique resilience properties that make them a good option. Ceramic membranes are less fragile than polymeric membranes, have a longer life, and can withstand heavy pollutant and solid loads, vigorous backwashing and a variety of chemical types and concentrations.

For membrane application, a major obstacle is the potential for fouling. Membrane operations often require pretreatment to reduce the treated water's fouling potential. Huang et al. (2009) summarized pretreatment technologies as: coagulation, adsorption, pre-oxidation, pre-filtration, dissolved air flotation, ion exchange, or some combination of them. Coagulation is much the most widely adopted pretreatment for membrane filtration. Coagulated water causes less fouling than un-coagulated water in most applications, although the opposite can happen, with coagulation leading to a higher level of irreversible fouling. The fouling mechanism seems to differ at different locations and is not completely understood. Most studies suggest that coagulation controls colloidal fouling (i.e., pore blockage) and removes the HMW fraction of natural organic matter (NOM), thus reducing NOM fouling. Gray et al. (2008) argued that coagulation's fouling control mechanism was the removal of LMW organics (with an adsorption peak at 220 nm), which ‘glue’ colloids to the membrane surface. Galjaard et al. (2005) suggested that coagulation removes HMW organics, while introducing or forming metal organic complexes. These could interact with smaller organics and the membrane, resulting in film formation on the membrane surface and irreversible fouling.

Full-scale application of other pretreatment techniques like adsorption, ion exchange and pre-oxidation is still very limited, and requires further assessment. One purpose of this study is to compare the impact of various pretreatment methods on ceramic membrane fouling.

A better understanding of membrane fouling mechanisms is crucial when determining the optimal pre-treatment strategy for particular waters. After extensive water treatment experience, Galjaard et al. (2005) suggested that the irreversible – i.e. persistent – fouling is caused by NOM-film attachment to the membrane surface. They believe that HMW organics interact at high concentrations at the membrane surface, forming long polymers. LMW organics – e.g. carboxylic acids and humics – combine with the HMW species by electrostatic forces, this interaction accelerating film formation in the same way that organic metal complexes do. This can result in rapid, irreversible fouling, if the film and membrane have opposite charges, because the film is adsorbed by the membrane. On this basis, two solutions exist to reduce the water's fouling potential: (1) remove the humics and carboxylic acids using ion exchange, and avoid metal organic complex formation by not using coagulants; and, (2) reduce the membrane's negative surface charge or create an opposite surface charge, to promote electrostatic exclusion of the film. The wish to remove humics and carboxylic for fouling control led to the development of both a novel ion exchange technology called suspended ion exchange (SIX^®) and a ceramic membrane process called CeraMac^®. Water sources are, however, unique and differ in their fouling characteristics. It is valuable, therefore, to verify the fouling hypothesis for other source waters and study how different pretreatment strategies impact the feasibility of using ceramic MF.

This paper presents the results of several pilot studies using ceramic MF. At least two pretreatment strategies were investigated at each location. Membrane performance with these pretreatments is discussed in comparison to the ‘NOM film’ fouling hypothesis.

The CeraMac^® process, developed by PWN Technologies, reduces ceramic membrane system installation costs so that they are cost competitive with polymeric membrane systems. Rather than having ceramic membrane modules in individual stainless steel casings, up to 192 modules can be housed in a single stainless steel vessel. This reduces the amount of stainless steel and number of valves significantly, while increasing productivity (i.e., all elements are backwashed simultaneously, reducing downtime during backwash [BW], from 10 minutes to a few seconds).

In all cases the ceramic microfilter is operated in dead-end mode and water is fed upwards into a vertically set module. BW occurs after a prescribed operating time. Water is forced through the membranes by air-pressure, built up in the BW tank by pumping water into it, not by a BW pump.

This creates an air-spring effect when the valve opens. The BW water rushes from the vessel and the valve closes, but the tank remains under pressure. Virtually no air volume is lost during BW, apart from that dissolved in the water. BW takes a few seconds, and forces water from the permeate- to the feed- side of the membrane. The BW water leaves the vessel through a separate backwash water port at the bottom. The flow used for BW is 3 L/m². The process takes 3 to 5 seconds when the membrane is clean, but up to 30 seconds when it is fouled. After BW, forward flush (FF) is automatic, from the top to the bottom of the feed channels of the membrane module, and is also forced over the membrane by air (which is generated by a compressor and stored in a separate FF tank). The air pushes a fixed volume of flush water (stored on top of the modules and in the membrane feed channels) out, thus emptying the whole feed side of the membranes.

Two types of chemically enhanced backwash (EBW) were tried during these studies, one with chlorine and the other low pH/peroxide. These were run at prescribed intervals (i.e., after a fixed number of backwashes), the chlorine EBW generally being used more frequently than the low pH/peroxide EBW. A typical EBW pattern was a chlorinated EBW after every five to 15 BWs, with low pH/peroxide EBW after every five chlorinated EBWs.

The EBW flow rate was the same as for normal BW, but the chemical was added while the BW tank was filling. During the EBW, the BW tank drained for four minutes through a separate, smaller, EBW outlet. Then, while the tank was re-filling, the membrane(s) soaked in the EBW solution for approximately five minutes. The sequence ended with a standard BW and FF. The whole EBW sequence lasted approximately 10 minutes.

The pilot studies were performed at different locations treating different sources, three surface waters and one secondary effluent from a municipal wastewater treatment plant.

The first pilot study was conducted on Ijssel Lake water for drinking water production in Andijk, the Netherlands. The lake is fed by the river Rhine and is the biggest fresh water lake in Holland. For the Andijk pilot work, in-line coagulation and ion exchange were used as pretreatments, with ferric chloride as the coagulant. The ion exchange process was SIX^® with LanXess VPOC 1071 anion resin.

The second pilot was at Singapore's Choa Chu Kang Waterworks (CCKWW), which has three raw water sources – the Kranji, Pandan and Western Catchments reservoirs. Here, the membrane system treated clarified water with and without pre-ozonation. The clarified water had been treated by screening, aeration, coagulation and clarification in the existing full scale plant. The target ozone contact time was as short as possible to reduce the contactor volume and dose as little ozone as possible. The ozone contactors' maximum capacity was not known when they were designed, so a conservative value of around five minutes at a maximum capacity of 110 m³/h was used before the water entered the membrane vessel. This yielded an initial ozone dose of approximately 1.3 to 1.5 mg/L, similar to the current dose at CCKWW after the existing sand filters. The target ozone concentration on the membrane surface was 0.8 to 1.1 mg/L.

The third pilot was at South West Water's (SWW) Crownhill Water Treatment Works (WTW) in the UK. The raw water came from Burrator Reservoir, combined occasionally with water from the rivers Tamar and Tavy. The study included four pre-treatments: (1) SIX^® (by PWN Technologies); (2) clarification (from Crownhill WTW); (3) clarification by the WTW followed by SIX^®; and, (4) SIX^® followed by pilot-scale, in-line coagulation. The anion resin used at Crownhill was LanXess S5128 (Germany).

The last pilot study was treating secondary effluent with a ceramic membrane for a confidential client. The effluent was first strained and chlorinated. Four pretreatments were evaluated, direct treatment (no pretreatment), in-line coagulation, ozonation, and ozonation followed by in-line coagulation.

The operational parameters were logged automatically in all pilot studies and included, but not limited to: time, feed water temperature, flow rate, and membrane feed- and permeate- side pressures. The transmembrane pressure (TMP) was calculated from the difference between the feed and permeate pressures. The membrane operating conditions were different for each pilot run – see below. More details of the studies can also be found in previous publications (Galjaard et al. 2013; Zheng & Galjaard 2013; Shorney-Darby et al. 2014).

The organic matter was analyzed by size-exclusion chromatography – liquid chromatography – organic carbon detection (SEC-LC-OCD) at ‘Het Water Laboratorium HWL’ (the Netherlands). The method was developed by DOC Lab (Germany) and its principles are described by Huber et al. (2011). The zeta potential measurements were made in the University of Twente (the Netherlands), using a Malvern Zetasizer nano.

Table 1 presents the TOC and DOC concentrations, and the UV Transmission at 254 nm (UVT) of the four source waters. The TOC and DOC data for the clarified surface water in Singapore were measured using USEPA's 415.1 method. The other TOC and DOC concentrations were obtained by SEC-LC-OCD. The values in the table show the range of TOC, DOC and UVT measured during the long term pilot tests, the values in brackets are their averages.

Water from Ijssel Lake yields a huge humics UVD peak. In the secondary effluent, the LMW fractions show high UV adsorbing properties. Combining the observations from Figures 1 and 2 shows that the organic matter from different sources was very different, not only in quantity, but also in composition and properties. LC-OCD analyses were performed regularly and three representative samples are shown. For Ijssel Lake water and the secondary effluent, there is some variability over time according to the spectra, but the results are generally fairly consistent. The fluctuation is much more significant for water from Burrator Reservoir.

Figure 6 shows fluctuations in TMP, attributed to seasonal feed water changes and operating errors (e.g., no EBWs in January). In general, however, it was very stable, with a TMP increase of 0.01 kPa/day. The first increase in TMP (June 2012) was caused by moving the 6 mg/L peroxide dosing, required for the advanced oxidation process with UV, from its initial location upstream of the membrane to a new location downstream (Galjaard et al. 2011). (Ceramic membranes can withstand high concentrations of peroxide.). This change resulted in an immediate TMP increase. Several experiments showed that dosing peroxide prior to the ceramic membranes could increase membrane permeability by around 20% (Zheng & Galjaard 2013). Ion exchange pretreatment was selected for the full scale plant, mainly because of its ability to remove nitrate as well as DOC, both of which improve membrane operation and downstream AOP processes (Martijn et al. 2010).

For the CCKWW CeraMac^® demo-plant (Galjaard et al. 2013), the ceramic membrane treated first stage clarified water produced by the existing plant. DOC concentration in the clarified water was between 2 and 3 mg/L, and comparable with the DOC concentration after pre-treatment in Andijk. DOC was determined in a local lab using USEPA method 415.1. No LC-OCD analyses were carried out.

Four pretreatment methods were tested in the secondary effluent treatment study: no pretreatment, in-line coagulation, ozonation and a combination of pre-ozonation and in-line coagulation. NOM analysis indicated that in-line coagulation worked fairly well to remove some of the biopolymers and humics. Ozonation alone had a limited impact on the total biopolymer concentration in the membrane feed water (Table 2) but did alter the character of the organic matter (according to the LC-UVD). This can also be seen in a change in the total amount of biopolymers in the permeate. Retention of these biopolymers by the membrane rose compared to in-line coagulation. Ozonation combined with in-line coagulation largely enhanced organic removal. (More details of the NOM analysis of the raw water will be published in due course).

Table 2 shows the different biopolymer rejection performances for the various pretreatment types. The membrane rejects more biopolymer when filtering untreated and ozonated secondary effluent, associated with significant zeta potential change after membrane filtration. However, a remarkably low rejection of biopolymer was observed when filtering the in-line coagulated water, associated with minor zeta potential change.

The fouling of low pressure membranes is often attributed to the deposition of dissolved HMW organic carbon – e.g. polysaccharides and polyhydroxyaromatics – on the membrane surface. However, as shown in the study carried out with water from Ijssel Lake, it is not necessary to remove the HMW NOM fraction to achieve a sustainable membrane operation. In both the ion exchange and ceramic MF process, it seems that the membrane rejected biopolymer fractions. The biopolymers were not glued/linked together and could be removed easily by BW. This contributed to the pre-removal of the humics and LMW acids by ion exchange. Kim & Dempsey (2008, 2010) also found that, by removing organic acid with ion exchange, almost no fouling occurred when treating secondary effluent. How humics and LMW acids affect membrane fouling remains unclear but they might be absorbed directly onto the membrane surface. The mass percentage of this group is small (<10%), however, and the molecules are often smaller than the membrane pores, so it is unlikely to be the only reason. It seems likely that the humics and LMW acids ‘glue’ the HMW organics, to the divalent and trivalent cations. The ‘crosslink’ effect of metal ions in developing irreversible fouling has been demonstrated by Elimelech and others (Hong & Elimelech 1997; Li & Elimelech 2004) for nanofiltration, and by Li for ultrafiltration (2011).

The combination of ozonation with ceramic MF significantly improved the CCKWW system's performance. When treating the clarified water, the fouling film formed gradually from organic and inorganic species. Ozone can affect/break the fouling film by reacting with its organic and inorganic components. Once the fouling film had been loosened, it was easily washed off the membrane surface. This can be considered a scrubbing/cleaning effect (Sartor et al. 2008). After the membrane had cleared, the continued presence of ozone prevented fouling film formation.

In the Plymouth case, the biopolymer fraction can be considered the primary foulant, because the biopolymer is sticky and difficult to remove by BW. Remarkably, removal of the LMW fraction by ion exchange alone did not lead to the stable performance seen at Andijk. The nature or origin and properties of the biopolymers seem to be different. This is only evident due to the data from the LC-OCD as compared with the LC-UVD. At Plymouth, in-line coagulation for the removal of this fraction seems to be essential to control fouling.

The biopolymer concentration in the secondary effluent is relatively very high compared to that in the other sources and removal of this fraction by coagulation seems essential. Coagulation also altered the charge properties of the biopolymer and consequently led to different biopolymer rejection by the ceramic membrane. A high level of biopolymer rejection was observed for untreated effluent and that which was only ozonated (78% rejection in both cases, without coagulation), with moderate rejection in the coagulated feed (50%), and low rejection effluent ozonated followed by in-line coagulation. This could suggest that the coagulation neutralized the biopolymer change and promoted electrostatic exclusion between the membrane and the biopolymer. Because of charge exclusion, fouling was reduced.

Pretreatment was needed for all the study feeds, to control membrane fouling, and establish a technically and economically feasible process. In all cases, the HMW fraction of the DOC – e.g. biopolymers – and its LMW fraction (humics/acids) seem to be responsible for the increase in TMP after a filtration cycle followed by BW (i.e. irreversible fouling). Removing one of these fractions often results in more stable operation. The decision about which fraction to remove – i.e. ion exchange to remove humics or in-line coagulation to remove some of the biopolymers – is often fixed by secondary needs or other restrictions (e.g. conditioning for downstream processes or removing other species than suspended matter). Ozonation led to much improved operation in all cases, but is not always cost effective. The feasibility of ozonation as a pretreatment depends largely on the initial ozone demand, and whether or not there are secondary treatment targets (e.g. greater virus removal, taste, odor, etc). SEC-LC-OCD is a powerful tool for water analysis in combination with UV detection, because it provides quantitative information as well the organic matter's characteristics.

The slow implementation of ceramic membranes offers an opportunity to develop new pretreatment technologies or cleaning strategies, especially in combination with strong oxidants and in-line coagulation. This is possible because of the membrane's superior ability to withstand strong oxidizing agents or heavy solids loads.

The authors wish to acknowledge the following for their contribution(s) to this study:

- PWN Water Supply Company, North-Holland, the Netherlands;

- PUB Singapore's National Water Agency (Wui Seng Ang and Mong Hoo Lim)

- Metawater;

- South West Water, United Kingdom (Chris Rocky and David Metcalfe);

- Shane Snyder Group, University of Arizona United States;

- Technical University Twente, the Netherlands;

- Het Water Laboratorium, the Netherlands.