Abstract

Increased pollution caused by socio-economic development has led to compound-contaminated high-hardness water pollution. In this study, laboratory-scale nanofiltration (NF) treatment of such water was investigated. Response-surface methodology was used to optimize the NF operating parameters, and a regression model with desalination rate and transmembrane pressure changes as response values was established. The NF membrane efficiencies in contaminant removal from groundwater and surface water with compound-contaminated high hardness and the membrane-fouling characteristics during long-term operation were investigated. The results show that the optimal operating parameters for the NF membrane in the removal of inorganic salts from groundwater are as follows: influent pH 8, influent pressure 1 MPa, and water yield 27.976%. The removal rates for groundwater total hardness, total alkalinity, total soluble solids, K+, Na+, Ca2+, Mg2+, , Cl, , and were 99.4, 90.3, 84.7, 63.2, 56.8, 99.6, 95.2, 99.6, 68.3, 86.1, and 65.9%, respectively. Surface water contains more complex components; therefore, membrane fouling during surface water is more serious. The NF membrane was operated continuously for more than 35 days under the optimal operating conditions with no serious membrane fouling.

HIGHLIGHTS

  • Response-surface methodology was used to optimize the NF operating parameters.

  • NF membrane was operated continuously for more than 35 days under the optimal operating conditions with no serious membrane fouling.

  • The membrane fouling by hardness ion was analyzed by SEM-EDS.

INTRODUCTION

Water hardness is generally caused by high concentrations of calcium and magnesium ions in the water. These ions can cause either temporary or permanent hardness. Industrial and domestic pollution, the discharge of inadequately treated wastewaters, and landfill leachates can increase the hardness of groundwater and surface water. Some water sources are, therefore, contaminated with other pollutants in addition to having a high total hardness (Tang et al. 2006). We define such water as compound-contaminated high-hardness water. This includes temporary hardness, permanent hardness, and hardness caused by sulfate, chloride, and other inorganic ions (Li et al. 2016). The water-supply systems of most cities in Northern China use both underground and surface sources. The high hardness and inorganic ion concentrations of such drinking water greatly reduce the residents’ quality of life (Baghoth et al. 2011).

The removal of water hardness is also known as water softening. Membrane separation technology has now become a common method of hardness reduction. Broad-spectrum separation by reverse osmosis (RO) is currently widely used to remove inorganic salts from water, but the low recovery rate, high power consumption, and strict pre-treatment requirements make this process expensive (Cheng et al. 2015). They have a looser surface layer structure than RO membrane, therefore NF is more suitable as the core treatment process for softening of drinking water, because of its smaller difference in effluent quality and lower operating pressure (Rui et al. 2020). At the same time, most of the nanofiltration membrane is charged type, NF can effectively remove all types of divalent and multivalent ions with molecular weights greater than 200 (Chen et al. 2017). In addition, NF retains some beneficial elements during salt removal from groundwater and has good potential for use in the remediation of water that is both polluted and has high hardness (Gorzalski & Coronell 2014; Oatley-Radcliffe et al. 2017). Srivastava et al. used response-surface methodology (RSM) for the performance evaluation of pilot-scale hybrid nanofiltration to optimize the treatment of brackish ground water (Srivastava et al. 2021); Dizge studied that RSM was used to investigate NF performance for removing contaminant (Dizge 2011). But few studies of the optimization of NF membranes for treatment of compound-contaminated high-hardness water have been reported.

In common with other membrane separation techniques, NF suffers from membrane fouling and this prevents its widespread adoption (She et al. 2016). The NF operating parameters have important effects on membrane fouling (Park et al. 2017). In this study, we used the RSM in the Design-Expert 10.0 software package to determine the optimal NF operating parameters to minimize membrane fouling and achieve good operational results. A regression model with the desalination rate and transmembrane pressure (TMP) change as the response values was developed, and the optimal process conditions were determined. The NF membrane was operated continuously under the optimal conditions in the treatment of groundwater and surface water, and the pollutant removal efficiency, changes in the operating parameters, and membrane fouling were monitored. The differences between the surface water and groundwater treated by NF were investigated. This work will apply the idea of optimal operating parameter for NF membrane removal of compound-contaminated high-hardness water and the influence of this water to membrane fouling, which can provide a reference for water plant optimization operation.

MATERIALS AND METHODS

Water quality experiments

Raw water obtained from the groundwater and surface water distribution system in Jinan City, China, was supplied to a bench-scale NF membrane device. Calcium chloride, magnesium chloride, and sodium sulfate were added to the raw water to increase the inorganic salt content and total hardness of the raw water. Total hardness, electrical conductivity (EC), turbidity, total dissolved solids (TDS), total alkalinity other water quality data on the quality of the supplied water are shown in Table 1.

Table 1

Test water quality for RSM

Detection indicatorTotal hardness (mg/L)pHEC (μs/cm)Turbidity (NTU)TDS (mg/L)Total alkalinity (mg/L)TOC (mg/L)
Groundwater 590–600 6.2–8.4 1,429–1,440 0.439–0.504 1,020–1,050 200–210 1.221–1.652 
Surface water 588–600 7.2–8.1 1,632–1,653 0.908–1.506 1,180–1,220 140–160 4.264–5.103 
Detection indicatorTotal hardness (mg/L)pHEC (μs/cm)Turbidity (NTU)TDS (mg/L)Total alkalinity (mg/L)TOC (mg/L)
Groundwater 590–600 6.2–8.4 1,429–1,440 0.439–0.504 1,020–1,050 200–210 1.221–1.652 
Surface water 588–600 7.2–8.1 1,632–1,653 0.908–1.506 1,180–1,220 140–160 4.264–5.103 

Filtration tests

Membrane

NF1-4021 (Keensen, China) membrane elements were used to construct the NF membrane system. The membrane material was aromatic polyamide, and the effective membrane area was 3.4 m2. The membrane achieved 90–95% salt rejection when the test water contained 2000 ppm MgSO4. The membrane gave selective desalination with high removal rates of divalent and multivalent ions (heavy metal ions) and partial retention of monovalent ions that are beneficial to human health.

Bench-scale filtration

Experiments were performed with a full-scale cross-flow system (Figure 1). The raw water in the feed tank was pumped into the NF system by a centrifugal pump. The raw-water temperature was controlled at 2.0 ± 0.5 °C with a cooling-water circulator. The NF system was constructed by spiral winding NF1-4021 membrane elements. The NF system was operated in total-recycling mode in which the permeate and concentrate stream were returned to the feed tank after filtration. Hydrochloric acid (1 mol/L) and sodium hydroxide (1 mol/L) were used to adjust the pH of the water supply.

Figure 1

Schematic of the NF experimental system. 1 – cooling-water circulator; 2 – feed tank; 3 – NF membrane module; 4 – concentrate pipe; 5 – permeate pipe.

Figure 1

Schematic of the NF experimental system. 1 – cooling-water circulator; 2 – feed tank; 3 – NF membrane module; 4 – concentrate pipe; 5 – permeate pipe.

The membranes used in the tests were chemically cleaned before feed-water processing to remove foulants from the membrane surfaces. Before each experiment, the membrane was immersed in deionized water for 12 h and pressurized with pure water at 1.2 MPa for at least 1 h to ensure structural stability. After each test, the NF membrane was washed sequentially with water, alkali, and acid to restore the membrane permeability.

Experimental protocols

Response-surface methodology

The RSM is used to predict how a target response variable will be affected by multiple dependent variables. The most commonly used software for RSM is Design-Expert. The operating parameters of the NF device were optimized by using an RSM Box–Behnken design (BBD). The performance of the NF membrane was optimized by using a three-factor BBD routine in the Design-Expert 10.0 software package. This third-order design provides an effective estimate of the second-order quadratic polynomial and yields the optimal response values in a three-dimensional observation space. The effects of influent pH, influent pressure, and water yield on the discharge performance were studied. Previous studies have shown that these parameters have the most significant effects on the performance. The independent variables were coded into three levels: −1, 0, and +1. The values for each coding level and codes of the independent variables are shown in Table 2.

Table 2

Levels and code experimental variables based on RSM

Independent variablesSymbolsLevels
− 10+ 1
pH 6.0 7.0 8.0 
Inlet pressure (MPa) 0.6 0.8 1.0 
Water yield (%) 20 30 40 
Independent variablesSymbolsLevels
− 10+ 1
pH 6.0 7.0 8.0 
Inlet pressure (MPa) 0.6 0.8 1.0 
Water yield (%) 20 30 40 

Membrane-fouling experiments

Continuous and stable operation of the NF membrane system under the optimal working conditions, which were obtained by using the RSM, was investigated by monitoring changes in the TMP difference, membrane flux, and other parameters. The operating parameters of the NF membrane system and changes in the pollutant removal efficiency were investigated to evaluate the NF membrane performance.

Analytical methods

The concentrations of TOC and UV254 were analyzed by a total organic carbon analyzer (TOC-L CPH CN200, Shimadzu, Japan) and an ultraviolet-visible spectrophotometer (TU-1810, Purkinje, Beijing, China), respectively. The surface morphology of fouled membranes was observed by a scanning electron microscope (SEM; JSM-6301, JEOL, Japan). Elements on fouled membranes were detected by an energy-dispersive spectrometer (EDS; Link 300, Oxford University Press, Oxford, UK). Before SEM and EDS test, it should be noted that the membranes were needed to be dried under natural conditions in a clean Petri dish for 48 h. In addition, the surface of the membrane needs sputtering Au to give the film a certain level of EC. Total hardness and total alkalinity were tested by the EDTA titration method and acid-base indicator titration, respectively. Ion chromatography (ICS-90, DIONEX, USA) was adopted to detect the concentration of inorganic ions. Conductivity and TDS of test water were measured by a conductivity meter (DDSJ-318, Shanghai, China).

RESULTS AND DISCUSSION

Optimization of NF operating parameters

Experimental design and results

Seventeen groups of experiments determined by the BBD method were performed to investigate the purification of compound-contaminated high-hardness water by NF membranes. The results are shown in Table 3.

Table 3

The experimental design and results of response-surface analysis

RunIndependent variables
ΔTMP (kPa)Desalination rate (%)
pHInlet pressure (MPa)Water yield (%)
0.6 30 0.37 97.57 
0.6 30 0.39 98.01 
30 0.4 98.31 
30 0.13 98.61 
0.8 20 0.44 98.66 
0.8 20 0.4 98.66 
0.8 40 0.52 97.74 
0.8 40 0.4 97.87 
0.6 20 0.42 98.31 
10 20 0.45 98.55 
11 0.6 40 0.42 97.60 
12 40 0.43 98.12 
13 0.8 30 0.35 98.45 
14 0.8 30 0.37 98.45 
15 0.8 30 0.34 98.60 
16 0.8 30 0.31 98.58 
17 0.8 30 0.32 98.65 
RunIndependent variables
ΔTMP (kPa)Desalination rate (%)
pHInlet pressure (MPa)Water yield (%)
0.6 30 0.37 97.57 
0.6 30 0.39 98.01 
30 0.4 98.31 
30 0.13 98.61 
0.8 20 0.44 98.66 
0.8 20 0.4 98.66 
0.8 40 0.52 97.74 
0.8 40 0.4 97.87 
0.6 20 0.42 98.31 
10 20 0.45 98.55 
11 0.6 40 0.42 97.60 
12 40 0.43 98.12 
13 0.8 30 0.35 98.45 
14 0.8 30 0.37 98.45 
15 0.8 30 0.34 98.60 
16 0.8 30 0.31 98.58 
17 0.8 30 0.32 98.65 

Regression model analysis

Multivariate regression fitting of the response values and the various factors in Table 3 was performed to obtain quadratic polynomial regression equations for the TMP change and desalination rate. The equations are as follows:
formula
formula
where Y1 is the change in TMP (kPa), Y2 is the desalination rate (%), X1 is influent pH, X2 is the raw-water pressure (MPa), and X3 is the water production rate (%).

P values indicate the significance of the regression model. Table 4 shows the variance analysis results for the regression equation. For Y1 (TMP change), P is less than 0.05, which indicates that the model has statistical significance. The independent variables in the model are X1 (first-order term) and X1X2 and X32 (second-order term). The lack-of-fit values, F, indicate how well the model fits the experimental results, i.e., the degree of difference between them. The value of the missing term in the model is 6.65, which indicates that the term missing from the model is not significant and is profitable to the model. The regression equation can, therefore, be used instead of the experimental data for further analysis.

Table 4

Analysis of variance of regression model

SourceSum of squares
DF
Mean square
F value
P value
ModelΔTMPDRΔTMPDRΔTMPDRΔTMPDRΔTMPDR
0.0912.23990.0110.256.2010.520.01830.0026
X1 – Influent pH 0.021 0.092 0.021 0.092 12.93 3.92 0.0134 0.0881 
X2 – Inlet pressure 4.513 × 10−003 0.55 4.513 × 10−003 0.55 2.78 23.50 0.1721 0.0019 
X3 – Water production rate 4.500 × 10−004 1.02 4.500 × 10−004 1.02 0.28 43.49 0.6457 0.0003 
X1X2 0.021 0.004998 0.021 0.004998 12.94 0.21 0.0134 0.6589 
X1X3 1.600 × 10−003 0.004380 1.600 × 10−003 0.004380 0.98 0.19 0.3952 0.6792 
X2X3 1.000 × 10−003 0.020 1.000 × 10−003 0.020 0.062 0.86 0.8273 0.3834 
X12 3.184 × 10−003 0.12 3.184 × 10−003 0.12 0.20 4.91 0.9019 0.0622 
X22 6.845 × 10−003 0.27 6.845 × 10−003 0.27 0.91 11.44 0.5722 0.0117 
X32 0.046 0.093 0.046 0.093 25.27 3.93 0.0018 0.0878 
Residual 0.014 0.16 1.951 × 10−003 0.024     
Misstated item 0.011 0.13 3.792 × 10−003 0.044 6.65 5.33 0.0493 0.0699 
Pure error 0.00228 0.033 5.700 × 10−004 0.008244     
Total variation 0.11 2.39 16 16       
SourceSum of squares
DF
Mean square
F value
P value
ModelΔTMPDRΔTMPDRΔTMPDRΔTMPDRΔTMPDR
0.0912.23990.0110.256.2010.520.01830.0026
X1 – Influent pH 0.021 0.092 0.021 0.092 12.93 3.92 0.0134 0.0881 
X2 – Inlet pressure 4.513 × 10−003 0.55 4.513 × 10−003 0.55 2.78 23.50 0.1721 0.0019 
X3 – Water production rate 4.500 × 10−004 1.02 4.500 × 10−004 1.02 0.28 43.49 0.6457 0.0003 
X1X2 0.021 0.004998 0.021 0.004998 12.94 0.21 0.0134 0.6589 
X1X3 1.600 × 10−003 0.004380 1.600 × 10−003 0.004380 0.98 0.19 0.3952 0.6792 
X2X3 1.000 × 10−003 0.020 1.000 × 10−003 0.020 0.062 0.86 0.8273 0.3834 
X12 3.184 × 10−003 0.12 3.184 × 10−003 0.12 0.20 4.91 0.9019 0.0622 
X22 6.845 × 10−003 0.27 6.845 × 10−003 0.27 0.91 11.44 0.5722 0.0117 
X32 0.046 0.093 0.046 0.093 25.27 3.93 0.0018 0.0878 
Residual 0.014 0.16 1.951 × 10−003 0.024     
Misstated item 0.011 0.13 3.792 × 10−003 0.044 6.65 5.33 0.0493 0.0699 
Pure error 0.00228 0.033 5.700 × 10−004 0.008244     
Total variation 0.11 2.39 16 16       

Notes:P ≤ 0.0001, for highly significant; P ≤ 0.05, for significant; P > 0.05, for not significant.

For Y2 (desalination rate), the P value for the model is less than 0.05, which indicates that the model is statistically significant. The independent variables in the model are primary X1, X2, and X3 (primary); X12 and X22 (secondary). The value of the lack-of-fit term is 3.96, which indicates that the missing term in the model is not significant and is profitable to the model. This regression equation can also be used as a substitute for the experimental data in subsequent analyses.

Response-surface analysis and optimal operating conditions

The predictive model equation can be visualized as a two-dimensional contour map or a three-dimensional response-surface map. Contour maps show how the independent variables affect the response values and the interactions between independent variables. The effects of corresponding independent variables increase with increasing steepness of the curve. In the contour maps and response-surface maps, the results become more significant with increasing darkness of the color (Luo & Wan 2013). The contour maps and response surfaces in Figure 2 show how various factors affect the removal of inorganic salts from groundwater by the NF membrane.

Figure 2

Contour map and response surface of various factors affecting TMP changes.

Figure 2

Contour map and response surface of various factors affecting TMP changes.

At a constant water yield, the color of the contour graph of Y1 (TMP change) becomes lighter and the response-surface graph becomes steeper with increasing pH. This indicates that the TMP changes substantially. At a constant influent pressure and pH, the TMP increases gradually with increasing water yield. The TMP increases with increasing influent pressure, as indicated by the darker color and steeper gradient. At the same influent pressure, the color of the figure is darkest, the gradient steepest, and the change in TMP decreases with increasing pH. This indicates that the water yield has the greatest effect on the TMP change. These simulations show that the order of importance of the effects of the variables on the TMP change is pH > influent pressure > water yield.

Figure 3 shows that for Y2 (desalination rate), at a constant water yield, the color of the contour plot becomes lighter, and the response-surface plot becomes less steep, with increasing pH. This indicates that the conductivity is lower. At a constant influent pressure and pH, the conductivity increases gradually with increasing water yield. The color becomes darker, the gradient becomes steeper, and the change in TMP increases with increasing influent pressure. Under the same pH condition, the water yield increases, the figure color is darkest and steepest, and the conductivity is lowest. This indicates that the water yield has the greatest impact on the conductivity. The order of importance of the effects of the variables on the EC is water yield > influent pressure > pH.

Figure 3

Contour map and response surface of various factors affecting desalination rate.

Figure 3

Contour map and response surface of various factors affecting desalination rate.

The optimal membrane parameters for purification of water with composite hardness were selected in accord with the Box–Behnken model. The optimal parameters are pH 8, influent pressure 1 MPa, and water yield 27.976%. Three experiments were performed under these conditions and the results were averaged. The experimental results are shown in Table 5. The experimental TMP was 0.18 kPa and the predicted value was 0.183 kPa, i.e., a relative error of less than 1.7%. The experimental desalination rate of 98.252% differed from the predicted value of 98.513% by less than 0.3%. This shows that the response-surface model is reliable and that the optimized process parameters can achieve the set test objectives.

Table 5

Validation of response values under optimal working conditions

Factor
Response value, ΔTMP
Response value, desalination rate
pHInlet pressure (MPa)Water yield (%)TruePredictTruePredict
1.0 27.976 0.18 0.183 98.252 98.513 
Factor
Response value, ΔTMP
Response value, desalination rate
pHInlet pressure (MPa)Water yield (%)TruePredictTruePredict
1.0 27.976 0.18 0.183 98.252 98.513 

Analysis of decontamination efficiency

Conventional indicators

The changes in membrane conductivity and the turbidity removal rate for NF-treated groundwater and surface water after 35 days of continuous operation are shown in Figure 4. The conductivity and turbidity removal rate for the groundwater decreased by 1.9 and 77%, respectively, whereas the values for the surface water decreased by 2.6 and 76%, respectively. There are two reasons for this. First, pollutants gradually accumulate on the membrane surface. This causes membrane fouling, which leads to a decline in the interception performance of the membrane. Secondly, the full-reflux operating mode of the NF membrane system, in which both the produced water and concentrate are returned to the raw-water tank and some pollutants are deposited on the membrane surface, results in a gradual improvement in the quality of the raw water. The concentration of pollutants is reduced, resulting in a lower removal rate. On the 18th day, the conductivity and turbidity removal rate increased because some raw water was removed for sampling and replaced by fresh raw water with a higher concentration of contaminants. From this time onwards, the conductivity removal rate decreased more slowly. This indicates that fouling of the NF membrane was not significant and that its ability to remove pollutants did not change substantially.

Figure 4

Change of turbidity and EC-removal rate.

Figure 4

Change of turbidity and EC-removal rate.

Inorganic salt indicators

Table 6 shows that the NF process effectively removed inorganic pollutants. The removal rates for groundwater total hardness, total alkalinity, total soluble solids, K+, Na+, Ca2+, Mg2+, , Cl, , and were 99.4, 90.3, 84.7, 63.2, 56.8, 99.6, 95.2, 99.6, 68.3, 86.1, and 65.9%, respectively. The corresponding removal rates from the surface water were 99.4, 89.3, 84.9, 75.8, 57.0, 95.4, 92.0, 99.7, 62.8, 83.6, and 62.8%. The NF process also effectively removed inorganic pollutants from the raw water, especially hardness-related Ca2+, Mg2+, , and other substances. The total hardness of the raw groundwater and surface water was greater than the total alkalinity. They are typical of compound-composited hard-water sources, in which the hardness of the raw water is mainly permanent hardness. After treatment with the NF membrane, the total hardness of the product water was less than the total alkalinity, its temporary hardness was equal to its total hardness, and its permanent hardness was 0, i.e., the permanent hardness of the raw water was almost completely removed. This shows that under the optimal operating conditions, the NF process effectively removed water hardness and was particularly effective in almost completely removing permanent hardness.

Table 6

Variation of inorganic salts by nanofiltration treatment

Total hardness (mg/L)Total alkalinity (mg/L)TDS (mg/L)K+ (mg/L)Na+ (mg/L)Ca2+ (mg/L)Mg2+ (mg/L) (mg/L)Cl (mg/L) (mg/L) (mg/L)
Groundwater Raw water 592 206 1,045 1.9 103 136 30.4 252 167 251 10.8 
Permeate water 3.6 20 160 0.7 44.5 0.54 1.45 52.9 34.8 3.68 
Surface water Raw water 596 150 1,201 14.9 118 122 40.8 363 256 183 0.86 
Permeate water 3.8 16 181 3.6 50.7 5.6 3.25 95.2 30 0.32 
Total hardness (mg/L)Total alkalinity (mg/L)TDS (mg/L)K+ (mg/L)Na+ (mg/L)Ca2+ (mg/L)Mg2+ (mg/L) (mg/L)Cl (mg/L) (mg/L) (mg/L)
Groundwater Raw water 592 206 1,045 1.9 103 136 30.4 252 167 251 10.8 
Permeate water 3.6 20 160 0.7 44.5 0.54 1.45 52.9 34.8 3.68 
Surface water Raw water 596 150 1,201 14.9 118 122 40.8 363 256 183 0.86 
Permeate water 3.8 16 181 3.6 50.7 5.6 3.25 95.2 30 0.32 

Removal by NF membranes is mainly based on screening and charge effects. Substances with diameters larger than the membrane pores are intercepted, whereas substances with smaller diameters pass through the membrane (Vatankhah et al. 2018). However, because of electrostatic interactions between charges on the membrane surface and charge-carrying pollutants in the water, substances with diameters smaller than the membrane pore will be partially intercepted. A combination of these effects increases the removal of inorganic pollutants by NF membranes (Mondal & De 2010).

Analysis of membrane-fouling characteristics

Variations in operating parameters

Figure 5 shows how the membrane flux and TMP changed during continuous operation. The TMP changes were mainly caused by changes in the water production pressure. The changes in the membrane flux were directly caused by changes in the water production flowrate. Figure 5 shows that the TMP gradually decreased to a stable value during continuous operation. This is primarily because of membrane fouling, which causes an increase in the water production pressure. The actual change in the TMP was between 0.4 and 0.6 kPa, which indicates that the degree of membrane fouling was slight. The membrane flux decreased by between 7 and 8 L/(m2h) because of the decrease in the water yield, which was caused by the increase in the water production pressure. The membrane flux decreased rapidly between the 15th and 20th days and then became stable after 23 days. This indicates that most of the membrane fouling occurred between 15 and 20 days of continuous operation. Pollutants gradually adhere to the NF membrane surface during operation. This results in a decrease in the water production flowrate and an increase in the water production pressure. After 20 days of continuous operation, these effects led to a significant decline in the membrane flux.

Figure 5

Changes in membrane flux and TMP during long-term operation of nanofiltration process.

Figure 5

Changes in membrane flux and TMP during long-term operation of nanofiltration process.

In the groundwater tests, the membrane flux decreased by 7 L/(m2 h) and the TMP difference increased by 0.4 kPa. In the surface water tests, the membrane flux decreased by 8 L/(m2 h) and the TMP difference increased by 0.6–0.7 kPa. In addition, TMP reached a plateau on the 9th day of continuous operation in the surface water tests, which was faster than groundwater testing (12 days). These effects are small under the optimal operating conditions, but the effect of the composite hardness of surface water on the membrane performance is stronger than that of the groundwater hardness. It can be seen from Figure 5 that the membrane flux of surface water is greater than that of groundwater, because groundwater contains more ions to neutralize the membrane surface charge, thus reducing the surface charge of NF membrane and increasing the pore size of the membrane.

Scanning electron microscopy-energy-dispersive X-ray spectroscopy (SEM-EDS) analysis of membrane surface

A combination of SEM and EDS methods provides a powerful tool, which can be used to examine the surface morphology of a sample and also determine the composition of a micro area. Two points on the filter element of the NF membrane were examined by SEM-EDS to observe the morphology of scaling on the membrane surface and determine the structural composition. The results are shown in Figures 6 and 7.

Figure 6

SEM images of fouled membranes onto surface after groundwater tests: (a) entrance at 500× magnification; (b) entrance at 2,000× magnification; (c) entrance at 8,000× magnification; (d) exit at 500× magnification; (e) exit at 2,000× magnification; and (f) exit at 8,000× magnification.

Figure 6

SEM images of fouled membranes onto surface after groundwater tests: (a) entrance at 500× magnification; (b) entrance at 2,000× magnification; (c) entrance at 8,000× magnification; (d) exit at 500× magnification; (e) exit at 2,000× magnification; and (f) exit at 8,000× magnification.

Figure 7

Normalized mass percentage of element contents in scaling on membrane surface.

Figure 7

Normalized mass percentage of element contents in scaling on membrane surface.

Figure 6 shows SEM images at ×500, ×2,000, and ×8,000 magnifications of the entrances and exits of the NF membrane samples after the contaminated groundwater test. The foulants adsorbed on the membrane surface consisted of irregular, compact, small particles. These are either monomers or inorganic compounds and mainly come from inorganic salt pollutants in the raw water. No biofilm was visible on the membrane surface. The fouling layer was mainly concentrated at the exit. The small amount of flocculent at the entrance possibly arose from a small number of suspended impurities in the raw water. Figure 6(d) shows that, in general, few pollutants were deposited on the NF membrane surface. This indicates that the degree of membrane pollution was low.

EDS was used to analyze the deposits on the membrane surface. The normalized mass percentages of each element on the membrane surface are shown in Figure 7. The main elements that fouled the membrane entrance were C, O, Na, Mg, Al, Si, Cl, Ca, Fe, and Cu. The foulant at the exit also contained K, Ti, and Cr. This is consistent with the SEM results.

EDS analysis of the pristine membrane showed that the mass percentages of C, O, and S were 60.58, 24.51, and 12.86%, respectively. The SEM images show that the pollutants did not completely cover the NF membrane surface, and the C and O contents were not significantly different from those for the pristine membrane. The O content increased slightly and the C content decreased. This verifies that the pollutants on the NF membrane were predominantly inorganic. The high Si content can be attributed to the presence of dissolved, colloidal, or suspended SiO2 in the water (Zhou 2000). There are two possible scaling mechanisms. One is a heterogeneous nucleation process, in which monomeric silicic acid directly scales the membrane surface (Mi & Elimelech 2013). The other is a deposition mechanism, i.e., monomeric SiO2 polymerizes and forms a colloid on the membrane surface (Gill 1993). The Ca content was higher than those of other metal cations. This indicates that the scale on the NF membrane consists mainly of calcium salts, the most common being CaCO3 and CaSO4. The Fe content was relatively high and was concentrated at the exit. This may be because Fe forms bridging complexes with organic matter in the water during membrane filtration, which results in Fe deposition of the membrane surface, i.e., deposition of inorganic matter on the film surface.

Figure 8 shows SEM micrographs at ×500, ×2,000, and ×8,000 magnifications of the inlets and outlets of NF membrane samples after filtering surface water of composite hardness.

Figure 8

SEM images of fouled membranes onto surface after surface water tests: (a) entrance at 500× magnification; (b) entrance at 2,000× magnification; (c) entrance at 8,000× magnification; (d) exit at 500× magnification; (e) exit at 2,000× magnification; and (f) exit at 8,000× magnification.

Figure 8

SEM images of fouled membranes onto surface after surface water tests: (a) entrance at 500× magnification; (b) entrance at 2,000× magnification; (c) entrance at 8,000× magnification; (d) exit at 500× magnification; (e) exit at 2,000× magnification; and (f) exit at 8,000× magnification.

The pollution of the membrane surface is more serious than that caused by groundwater treatment, and the types of pollutants are different. Figure 8 shows that the foulants adsorbed on the membrane surface were tightly arranged, unevenly distributed, and contained more organic contaminants. However, no significant biofilm formed on the membrane surface. The SEM images also show that the fouling layer on the membrane surface was mainly concentrated at the exit. The main reason is that organic and inorganic pollutants are forced to the back of the membrane under the action of the inlet pressure and gradually adhere to it.

The normalized mass percentages of the elements found at each site are shown in Figure 9. Carbon and inorganic elements, mainly oxygen, sodium, magnesium, aluminum, silicon, sulfur, chlorine, potassium, calcium, titanium, iron, and copper, were present on the surface of the membrane at both the inlet and exit. Surface water contains complex pollutants; therefore, the effect on the NF membrane is greater. Gold sprayed on the surface to enable SEM observations was not considered in this analysis.

Figure 9

Normalized mass percentages of elements detected in scaling on membrane surface.

Figure 9

Normalized mass percentages of elements detected in scaling on membrane surface.

The S content of the used membrane was significantly lower than that of the pristine membrane, but the C and O contents changed less. This indicates the presence of more organic components on the NF membrane. This is consistent with the SEM results. The Si content at the entrance was higher than that at the exit, similarly to the results for the NF membrane used to treat surface water. Ping et al. reported that Si can be effectively removed by NaOH (Ping et al. 2020). The contents of Mg, Al, Fe, Cu, and other elements at the entrance were lower than those at the exit. This could be because during water treatment, cations decrease the surface charge on natural organic matter (NOM) and the film, and form an intermolecular bridge between NOM and the film. This makes NOM deposition on the film surface easier and increases membrane scaling.

Comparison of membrane fouling by two water sources

There are a number of important differences between the groundwater and surface water test membranes. SEM-EDS shows that the foulant content of the surface water test film was significantly higher than that of the groundwater test film, and it contained organic matter particles. The pollutants on the groundwater test membrane were mainly inorganic and concentrated mainly at the back of the membrane. The surface water test film consisted of composite pollutants, which were formed from organic and inorganic substances, and the areas covered at the entrance and exit of the film were similar. Most of the fouling near the entrances of the surface water test membrane consisted of metal impurities and inorganic particles. These experiments show that the membrane fouling caused by the complex hardness of surface water is more severe than that caused by hard groundwater. The cations in the raw water have different effects on organic pollution, and there is sufficient evidence that membrane pollution caused by dissolved organic matter is significantly enhanced in the presence of divalent cations such as Ca2+ (Jucker & Clark 1994). In addition, we can see from Figures 7 and 8 that the content and degree of contaminants on the membrane fouling is not related to the concentration of contaminants in the water.

CONCLUSIONS

A regression model of the TMP changes and desalination rate was developed by BBD RSM in the Design-Expert software package. The order of importance of the factors that affect TMP was pH > influent pressure > water yield. For the desalination rate, the order of importance was influent yield > influent pressure > pH. Modeling showed that the optimal operating parameters for removing inorganic salts from groundwater by an NF membrane were influent pH 8, influent pressure 1 MPa, and water yield 27.976%. These values were verified experimentally.

The NF process effectively removed inorganic pollutants from groundwater and surface water with compound-contaminated high-hardness water, the removal rates of total hardness, total alkalinity, total soluble solids, K+, Na+, Ca2+, Mg2+, , Cl, , and from groundwater were 99.4, 90.3, 84.7, 63.2, 56.8, 99.6, 95.2, 99.6, 68.3, 86.1, and 65.9%, respectively. For surface water, the corresponding removal rates were 99.4, 89.3, 84.9, 75.8, 57.0, 95.4, 92.0, 99.7, 62.8, 83.6, and 62.8%. Permanent hardness in surface water was almost completely removed. The NF membrane was operated continuously for more than 35 days under the optimal operating conditions with no serious membrane fouling.

SEM-EDS analysis of scaling on the surface of the NF membrane showed that the contaminants were mainly deposited at the membrane exit. The surface water composite hardness was more complex than that of the groundwater. SEM-EDS spectra showed that membrane fouling during treatment of the surface water with composite hardness was greater than that during treatment of groundwater with composite hardness.

ACKNOWLEDGEMENTS

This study was supported by the National Major Projects on Water Pollution Control and Management Technology (No. 2015ZX07406005) and the Taishan Scholar Post Construction Project of Shandong Province.

DATA AVAILABILITY STATEMENT

All relevant data are included in the paper or its Supplementary Information.

REFERENCES

Chen
J.
,
Zhang
J.
,
Jin
Y.
,
Sun
Y.
,
Song
X.
&
Jianguo
Y. U.
2017
Nanofiltration and electrodialysis on separation of monovalent and divalent ions from concentrated seawater
.
The Chinese Journal of Process Engineering
17
(
3
),
491
499
.
Cheng
X. Q.
,
Liu
Y.
,
Guo
Z.
&
Shao
L.
2015
Nanofiltration membrane achieving dual resistance to fouling and chlorine for ‘green’ separation of antibiotics
.
Journal of Membrane Science
493
,
156
166
.
Dizge
T. N.
2011
Multi objective performance analysis of nanofiltration process to loading parameters by response surface approach
.
Desalination
272
,
164
169
.
Gill
J. S.
1993
Inhibition of silica – silicate deposit in industrial waters
.
Colloids & Surfaces A Physicochemical & Engineering Aspects
74
,
101
106
.
Gorzalski
A. S.
&
Coronell
O.
2014
Fouling of nanofiltration membranes in full- and bench-scale systems treating groundwater containing silica
.
Journal of Membrane Science
468
(
468
),
349
359
.
Jucker
C.
&
Clark
M. M.
1994
Adsorption of aquatic humic substances on hydrophobic ultrafiltration membranes
.
Journal of Membrane Science
97
,
37
52
.
Li
K.
,
Wang
J.
,
Liu
J.
,
Wei
Y.
&
Chen
M.
2016
Advanced treatment of municipal wastewater by nanofiltration: operational optimization and membrane fouling analysis
.
Journal of Environmental Sciences
43
(
5
),
106
117
.
Luo
J.
&
Wan
Y.
2013
Effects of pH and salt on nanofiltration – a critical review
.
Journal of Membrane Science
438
,
18
28
.
Mi
B.
&
Elimelech
M.
2013
Silica scaling and scaling reversibility in forward osmosis
.
Desalination
312
,
75
81
.
Oatley-Radcliffe
D. L.
,
Walters
M.
,
Ainscough
T. J.
,
Williams
P. M.
,
Mohammad
A. W.
&
Hilal
N.
2017
Nanofiltration membranes and processes: a review of research trends over the past decade
.
Journal of Water Process Engineering
191
,
64
171
.
Park
M.
,
Anumol
T.
,
Simon
J.
,
Zraick
F.
&
Snyder
S. A.
2017
Pre-ozonation for high recovery of nanofiltration (NF) membrane system: membrane fouling reduction and trace organic compound attenuation
.
Journal of Membrane Science
523
,
255
263
.
Ping
X.
,
Yubin
Z.
,
Shuxuan
D.
&
Feng
X.
2020
Case analysis of foulant component on ultrafiltraion membrane and its chemical cleaning effect in a drinking water plant
.
Chinese Journal of Environmental Engineering
14
(
5
),
274
281
.
Rui
W.
,
Weipeng
G.
,
Yan
Z.
,
Jianyu
Z.
,
Zhendong
G.
,
Heng
L.
&
Guibai
L.
2020
Application of nanofiltration in shallow groundwater treatment
.
Water & Wastewater Engineering
56
(
479(11)
),
10
15
+ 20
.
She
Q.
,
Wang
R.
,
Fane
A.
&
Tang
C.
2016
Membrane fouling in osmotically driven membrane processes: a review
.
Journal of Membrane Science
499
,
201
233
.
Tang
K.
,
Hou
J.
&
Tang
Y.
2006
Assessment of groundwater quality in China: I. Hydrochemical characteristics of groundwater in plain area
.
Water Resources Protection
22
(
3
),
1
5
.
Vatankhah
H.
,
Murray
C. C.
,
Brannum
J. W.
,
Vanneste
J.
&
Bellona
C.
2018
Effect of pre-ozonation on nanofiltration membrane fouling during water reuse applications
.
Separation and Purification Technology
205
,
203
211
.
Zhou
R. S.
2000
Performance of RO membranes in silica bearing waters
.
Desalination
132
(
1
),
337
344
.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY-NC-ND 4.0), which permits copying and redistribution for non-commercial purposes with no derivatives, provided the original work is properly cited (http://creativecommons.org/licenses/by-nc-nd/4.0/).