Filtration efficiency in a conventional water treatment system was analyzed in the context of pre-hydrolyzed coagulant overdosing. Two commercial coagulants of different aluminum speciation were tested. A study was carried out at a water treatment plant supplied with raw water of variable quality. The lack of stability of water quality caused many problems with maintaining the optimal coagulant dose. The achieved results show that the type of coagulant had a very strong influence on the effectiveness of filtration resulting from the application of an improper coagulant dose. The overdosing of high basicity coagulant (PAC85) caused a significant increase of fine particles in the outflow from the sedimentation tanks, which could not be retained in the filter bed due to high surface charge and the small size of hydrolysis products. When using a coagulant of lower basicity (PAC70), it was much easier to control the dose of coagulant and to adjust it to the changing water quality.
Coagulation followed by sedimentation and filtration are the basic processes in the treatment of surface water. In recent years, polyaluminum chlorides (PACls), partially hydrolyzed products of Al (III), have been widely used to remove particulate, colloidal and dissolved substances. Pre-hydrolyzed aluminum coagulants have many advantages over conventional aluminum coagulants. Their effectiveness is less dependent on temperature and the pH. PACls contain highly charged polymeric aluminum species as well as monomers and the Al13 polymer has been shown to be the dominant polymeric species.
The Al13 polymer is the most effective destabilizer of negatively charged colloids and it maintains high speciation stability (dissolved and polymerized form) over a wider pH range compared to aluminum sulfate. The rate of hydrolysis of polycationic products to Al(OH)3 is significantly slower than for alum. Hence, PACls neutralize the negative electrokinetic potential of the colloids more efficiently than alum. However, the transformation process of monomers to Al13, which occurs at the slightly acid pH 6.0–6.5, should not be underestimated. Al13 formed in situ may reveal more effective charge neutralization than preformed Al13 (Wang et al. 2004; Yan et al. 2008a, b; Świderska-Bróż & Rak 2009).
Alum and PACls have been shown to form different solids in water treatment coagulation. Al-Ferron research indicates that the hydrolysis precipitates of alum are composed of amorphous Al(OH)3, but those of PACls are composed of aggregates of Al13 (Van Benschoten & Edzwald 1990; Duffy & van Loon 1994; Pernitsky & Edzwald 2006). Therefore, the precipitate formed from solutions of polymeric aluminum is often designated as Al(OH)*3(am) in the literature to differentiate it from the Al(OH)3(am) precipitate formed with monomeric Al solutions.
Although the ﬂocs have received a lot of attention, little research has focused on the impact of the properties of ﬂocs formed by different Al species on the effectiveness of post-sedimentation processes (McCurdy et al. 2004; Yan et al. 2008a, b; Zhao et al. 2010). PACl flocs show a different particle size distribution from hydrolyzing coagulants. Floc properties are essentially dependent on the distribution of Al species, which is pH dependent. Comparing the size development trajectory of hydrolysis precipitates, the floc size of the hydrolyzing coagulant is larger than that of PACls and shows higher mechanical strength (Yan et al. 2007; Chen et al. 2009; Xu et al. 2011; Hu et al. 2012).
The main objective of this paper was to examine the effect of a type and a dose of PACl used in coagulation on the effectiveness of rapid filtration. The possibility of controlling the filtration process with a particle counter was analyzed.
The study was conducted on a technical scale at the selected water treatment plant (WTP). The technological system consisted of a coagulation process carried out in the rapid mixing tanks with mechanical stirring, followed by flocculation in the hydraulic mixing tanks combined with the vertical settling tanks. Sedimentation was followed by filtration at the rate 6 m/h in the gravity rapid filters. The filtrate was conveyed to the clean water tanks through the ultraviolet (UV) lamp system. In addition to the UV disinfection process, chlorine compounds were also used to protect water from the bacteriological contamination during its distribution.
The dual media filter bed used consisted of anthracite and sand with the following characteristics:
granulation of sand: 0.5–0.8 mm (d10 = 0.6 mm) and a layer height of 0.6 m;
granulation of anthracite: 0.8–1.6 mm (d10 = 0.9 mm, d60/d10 < 1.4) and a layer height of 0.6 m.
Reagents and analysis
In the study, two commercial reagents of different basicity were tested. The Al speciation of the coagulants was identified by the ferron method (Wang et al. 2004). A K-value-based ferron assay using nonlinear least squares analysis was applied to determine Al speciation (Ye et al. 2009).
The results of speciation measurement show that for PAC85 (basicity 85%), the fraction of Al13 and Al monomers was 40 and 5%, respectively. For PAC70 (basicity 70%), the fraction of Al13 was very low, i.e. 5%, and the Al monomer fraction was ca. 40%.
The doses of coagulants were selected automatically based on the measurement of the turbidity of water supplied to the treatment plant. In addition, the accuracy of selected doses was checked periodically using jar tests. A coagulant dose was optimized for minimum settled water turbidity and absorbance UV254.
The operation procedures for the jar tests were as follows: 0.7 L of raw water was transferred into a 1.2-L square beaker with a sampling port 3 cm below the water surface; the jar tester was started at rapid mixing of 270 rpm; after 30 s coagulant was added, followed by a mixing speed of 200 rpm for 2 min; 30 rpm for 20 min; and then after 30 min of quiescent settling, samples were taken for water quality measurement.
After settling, the residual absorbance UV254 and turbidity of the supernatant were measured. Zeta potential and a particle size distribution of precipitates forming during the coagulation stage were measured via a laser zeta analyzer, Zetasizer Nano Range (Malvern Instruments Ltd, Malvern, UK), and a particle counter, IPC (Kamika Instruments, Warsaw, Poland). Turbidity was measured using a turbidimeter, 2100AN (HACH Company, Loveland, CO, USA), and absorbance via a spectrophotometer, DR5000 (HACH, Company, Loveland, CO, USA).
At technical scale, the same type of spectrophotometer and turbidimeter were applied to analyze absorbance and turbidity of the filter influent and the effluent. A particle number was analyzed online by a particle counter, ARTI WPC21 (HACH Company, Loveland, CO, USA).
The need to use two different coagulants resulted from the rapid pH changes of the water supplied to the WTP. The average pH value of the treated water was in the range of 7.3–7.5. However, in spring and summer, an increase of the pH to a level as high as 9.0 was observed. Coagulation with PAC70, which was the primary coagulant used at the WTP, was then ineffective. The variable pH caused interference with flocculation, resulting in an increase of turbidity and a decrease in the efficiency of contaminant removal. Therefore, in the period of the increased pH, a coagulant of higher basicity (PAC85) was used.
The study was conducted in the period when the pH of the raw water was close to average values. The water characteristics are shown in Table 1.
|Parameter||Unit||Max. value||Min. value||Median value|
|Total organic carbon||mg C/L||9.8||3.2||4.8|
|Two micrometer particle number||1/mL||19,050||3,500||7,200|
|Parameter||Unit||Max. value||Min. value||Median value|
|Total organic carbon||mg C/L||9.8||3.2||4.8|
|Two micrometer particle number||1/mL||19,050||3,500||7,200|
The measuring system used at technical scale consisted of two particle counters. One of the devices was installed on a pipeline that supplied clarified water from sedimentation tanks to one of the selected rapid filters and the other on a filtrate pipeline. Each counter measured a number of particles of the preliminary stated size at 30 measurements per minute.
The size of particles was specified by analyzing a particle size distribution in the filtrate. The analysis showed that most of the particles were about 2 µm. These results were consistent with the earlier studies on coagulation with pre-hydrolyzed coagulants (Gumińska & Kłos 2011; Kłos 2014). Taking into account the above-mentioned results, particle counters in the technical system were set to measure 2 µm particles.
The technical rapid filters were not equipped with turbidimeters working online. Therefore, every 2 h, samples of the filter influent and the effluent were collected to measure turbidity. In addition, microbiological analysis was carried once daily out to determine the presence of Escherichia coli bacteria in the filtrate.
The measurements were made in the series of measurement compatible with the duration of filtration runs. The study was carried out for 6 weeks.
RESULTS AND DISCUSSION
The use of particle number measurement to control the quality of filtrate
The results of the study show that in the treatment system where coagulation is the basic treatment process, the application of turbidity measurements to control rapid filtration may not be reliable. The results of the tests suggest that any significant correlation between turbidity and the number of 2 µm particles could not be determined. It was particularly evident for low turbidity of the filtrate (less than 0.2 NTU) (Kłos 2014). Figure 1 presents the changes of the filtrate quality for one exemplary filter run. In that stage of the study, PAC85 was used as a coagulant. The filtrate quality was analyzed on the basis of the number of 2 µm particles and turbidity. As shown in Figure 1, the curves illustrating the change of turbidity and the number of particles do not overlap. At the 21st hour of the filtration run, a particle number in the filtrate increased from 40 in 1 mL to more than 70 in 1 mL. The longer the filtration time, the higher the number of particles observed, exceeding the value of 150 mL−1 at the 24th hour of the filter run. The value of 150 particles in 1 mL was defined as the maximum acceptable value due to the fact that at higher values E. coli bacteria appeared in the filtrate. The continuous increase in the number of particles was observed in the filtrate until the 27th hour of the filter run, when it reached a value of more than 800 particles in 1 mL and then it ranged from 400 to 1,000 in 1 mL. Changes in the number of particles did not influence the filtrate turbidity. Turbidity of the filtrate was less than 0.05 NTU up to the 25th–26th hour of the filter run. At the 27th hour of the filter run, turbidity was 0.1 NTU and then increased until the end of the filter run, reaching a value of about 0.5 NTU. This means that in the filtration cycle, the time delay between the change of the effluent quality controlled by the number of 2 µm particles and turbidity was approximately 5–6 h. For other tested rapid filters, this time ranged from 2 to up to 10 h.
The observed problems resulted directly from the variable quality of clarified water. The analysis of water quality supplying the rapid filter showed that at the 20th hour of the filter run, the number of particles in the feed water increased, and the increase was observed until the 27th hour of the filter run. At the same time, the increase in particle number was not accompanied by increased turbidity. The increase of turbidity in the feed water was recorded only at the 27–28th hour of the filter run, followed by an improvement of the quality of water after sedimentation to the previous level.
Changes of treatment effectiveness resulted from the periodic sudden changes in water quality supplied to the WTP and from the lack of proper adjustment of a coagulant dose to these changes. The coagulant dosing control system used at the WTP was based on the correlation between the dose of coagulant and turbidity. When turbidity increased from 17 to 41 NTU, the dose of coagulant was increased by about 35%. The recorded increase in the number of particles indicated that such a dose was far too high in relation to the optimal dose. A significant number of non-agglomerated particles of precipitated coagulant hydrolysis products were observed. Only manual correction of the coagulant dose to the previous value caused an improvement in water quality feeding the rapid filter. Coagulant overdosing and the resultant appearance of a large number of fine particles were confirmed in jar tests carried out at different doses of coagulant. Figure 2 shows the impact of the PAC85 dose on zeta potential and on the average size of particles noted during coagulation (a rapid titration method for predicting the optimal coagulant dose). As shown in Figure 2, the optimal dose of PAC85 was approximately 0.85 mg Al/L. The dose accuracy selection was confirmed by jar test results. The PAC85 dose (determined by the coagulant dosing control system) used in the technical system was much higher – 2 mg Al/L.
Fluctuations in the filter influent quality due to the inadequacy of coagulant dose to the quality of raw water had a very strong impact on filtration efficiency. Until the change of coagulant dose, the filter operated correctly. The excessive dose of coagulant resulted in a large number of precipitated particles in the filter influent. The particles were adsorbed on the surface of grains of the filter bed. Considering the effective size of the sand bed, a filtration process based on a sieving mechanism allowed the retention of particles of a size greater than ca. 40–50 µm. Smaller particles were retained through interaction with the surface of grains and earlier deposited suspensions. Owing to the fact that those particles were not electrically neutral, they formed a coating of positively charged aluminum precipitates on the surface of filter grains. That coating repelled particles present in the filter inflow, preventing them from retention. Therefore, in the period of coagulant overdosing, when a large number of particles were noticed in the outflow from the settling tank, the filter was not able to recover its efficiency, despite improvement of the quality of the filter influent. The process of filtration was canceled to wash the filter bed.
The influence of the type of coagulant on the rapid filter operation
The results of the measurements of filtrate quality showed that the impact of a coagulant type on the efficiency of filtration was dominant. Figure 3 presents the variation in the average amount of 2 μm particles in the filtrate. Measurements were made during the filter cycle, which followed coagulation with the use of two coagulants. At first, coagulation was conducted with PAC85 and then PAC70 was applied. The change of coagulant was made at the 21st hour of the filter run. The theoretical hydraulic retention time in the settling tanks was approximately 7 h.
As shown in Figure 3, at the 22nd hour of the filter run, an increase in the number of particles was observed, and a few hours later, turbidity also increased. At the 36th hour of the filter run, a sharp drop in the number of particles was noticed, despite high raw water turbidity and the increased coagulant dose. It was approximately 14 h since the change of coagulant type, which corresponded to the duration of the double exchange of water in the settling tanks. This means that the improved quality of the filtrate was related to the quality of the supernatant in a sedimentation tank after a coagulant change. At the 40th hour of the filter run, backwashing started. During the subsequent filter run, when coagulation with PAC70 was conducted, the filtrate quality significantly improved (Figure 4). The average number of 2 μm particles was much lower than during the previous filter cycle and ranged from 5 to 30 in 1 mL. This low value was maintained during the entire filter run, despite the changes in raw water quality and the change in coagulant dose.
The results show that type of coagulant may have a very strong impact on the effectiveness of filtration deriving from a problem with maintaining the optimum dose of coagulant. High basicity coagulants of high Al13 fraction are characterized by high surface charge of hydrolysis products. For this reason, it is very difficult to maintain the optimum dose of coagulant in periods of sudden changes in raw water quality. As a result of coagulant overdosing, a significant number of fine particles appears in the outflow from the sedimentation tanks, which cannot be retained in the filter bed due to their high surface charge. Consequently, the filtration process has to be stopped despite the inexhaustible bed capacity. When using a coagulant with lower basicity and lower Al13 fraction, it is easier to control the dose of coagulant and to adjust it to the changing water quality.
A particle counter is a better tool to control the filtrate quality than turbidity measurements. Owing to the selectivity of turbidimeters, only a particle counter allows for an immediate identification of the negative results of coagulant overdosing noted as the increase of the number of 2 μm particles. To ensure the correct operation of rapid filters in a conventional treatment system, an effective system of coagulant dose selection should be provided, depending on raw water quality changes. Simple methods based on correlation of coagulant dose and turbidity do not allow for the selection of an optimal dose in the case of sudden changes of quality of water supplied to the WTP. The application of a particle counter to control the quality of filtrate is particularly recommended for technological systems where pre-hydrolyzed coagulants are used. When these systems are not equipped with a system for coagulant dose prediction (e.g., a streaming current detector), a particle counter allows for the identification of problems arising from the lack of adjustment of coagulant dose to water quality. Hence, the application of a particle counter enables adaptation of the operation schedule of filters to the current situation.