Introduction

Capable of retaining beneficial mineral elements to an appropriate extent in the permeat, the nanofiltration technology is currently recognized as the most optimal process for the advanced treatment of drinking water in China, gradually replacing the position of reverse osmosis (Fig. 1). Chinese large-scale water supply plants primarily choose to use spiral wound nanofiltration membranes. Worthwhile to be noticed, the liquid channels of spiral wound nanofiltration membranes are screwed in spiral shapes with relatively large perimeters, tending to generate high resistance and easily causing serious loss of water potential, which is unfavorable for suppressing the concentration polarization. As a result, the feed water entering spiral wound nanofiltration membrane elements requires to be carefully pretreated to mitigate the membrane fouling phenomena involving microbial fouling and macromolecular organic fouling. Additionally, pretreatment agents such as reductants and antiscalants are also necessary due to the sensitivity of spiral wound nanofiltration membranes to chlorine oxidation as well as their tendency to scale, whose addition to the feed water increases the influent impurity content and then the operating expense.

Fig. 1
Fig. 1
Full size image

Graphical abstract.

In recent years, hollow fiber nanofiltration membranes have been gradually accepted by the water treatment market. Made from antioxidant polymer materials such as cellulose acetate, polysulfones, polyetherimides, and ceramics, a hollow fiber nanofitration membrane element is typically constituted by several hollow fiber membrane filaments arranged in parallel, each of which possesses a double-layer structure. As the support layer is used to enhance the mechanical strength, the separating layer is coated layer-by-layer with outer preliminary filtration layer simulating microfiltration membranes or ultrafiltration membranes and inner nanofiltration layer intercepting micromolecular pollutants. The pore size of the inner nanofiltration layer generally ranges from 1 to 10 nm, slightly larger than the one of traditional dense spiral wound nanofiltration membranes1,2,3. As an important advantage, hollow fiber nanofitration membranes own favorable fluid dynamic conditions for their short and straight hollow liquid channels are capable of reducing the loss of water potential and accommodating concentrate reflux, effectively alleviating the concentration polarization. In addition, the surfaces of hollow fiber nanofitration membranes are backwashable, which help achieve better membrane cleaning effects, and their larger specific areas help reduce the occupied area of the membrane system4,5,6.

In summary, the hollow fiber nanofitration technology possesses prominent advantages compared with the traditional spiral wound nanofiltration technology based on its anti-fouling performance, high packing density and unnecessity of pretreatment processes or pretreatment agents, while its main disadvantage lies in the relatively weak retention effect on ionic contaminants. Nevertheless, the national standards of China strictly limit the content of suspended solids, organic contaminants and microorganisms in the drinking water, whose restrictions on the content of classic inorganic contaminants are relatively loose compared to those prescribed by the United States, the European Union or Japan (Table 1). Thus, the corresponding policies of drinking water in China are theoretically compatible with the hollow fiber nanofitration technology7,8,9,10.

Table 1 Transnational comparison of the standards for drinking water quality.
Full size table

Before 2022, several pilot-scale hollow fiber nanofitration researches involving the advanced treatment of micro-polluted surface water have been conducted worldwide except in China. Despite highlighting some environmental factors which are often overlooked during lab-scale researches, these pilot studies focused more on the constituent complexity of natural feed streams while seldom paying attention to the variation and influences of fundamental water properties, such as water temperature11,12. Considering the aforementioned aspects, we put forth the hypothesis that the hollow fiber nanofitration technology is applicable for producing drinking water using natural surface water from Beijing-Hangzhou Grand Canal, and implemented the very first large-scale pilot testing in China in order to explore the potential technological challenges for treating the certain type of surface water with seasonably variable temperature and find methods to solve them.

Methods and materials

Conditions of pilot testing site

The pilot testing was situated in Jiangsu Province within the basin of Beijing-Hangzhou Grand Canal, which is located in an inland plain surrounded by fertile soils without the existence of saline-alkaline lands or inwelling phenomena. Therefore, the source water of the research was able to be regarded as non-brackish water. With villages and farmlands centralized around, the industry of the region where the research was conducted was supported by agriculture, leading to the discharge of organic-abundant irrigation water, graziery wastewater, aquacultural wastewater, straw sanies and rural domestic sewage towards the water source. On the other hand, the source water experienced considerable temperature differences across different seasons, with the water temperatures reaching as high as 40 °C in summer and dropping to as low as 5 °C in winter.

Excluding the sterilization unit, the original technological process of the water supply plant where the pilot test was implemented was the combination of biological aerated filters, coagulation tanks, sedimentation tanks, sand filters and ozonized activated carbon filters. As the disinfection byproducts in the drinking water production was ignorable, the ferrum ions (Fe3+) originating from the iron-containing flocculants could possibly result in the excessive iron content of the drinking water products. According to Table 2, the quality of current water products can no longer met the requirements of local standards for high-quality drinking water7,13. Thus, the hollow fiber nanofitration system was considered to be applied to replace the technological units subsequent to the sedimentation tank (Fig. 2).

Table 2 Latest quality standards for local drinking water.
Full size table
Fig. 2
Fig. 2
Full size image

Schematics of original technological process and potential reconstruction plan.

Configuration of pilot equipment

The component devices and their specifications of the pilot equipment are illustrated in Fig. 3.

Fig. 3
Fig. 3
Full size image

Diagram of pilot equipment.

Designed operating conditions of pilot testing

The pilot testing utilized the effluent of sedimentation tank of the local water supply plant as the feed water (Fig. 1), whose water quality status is shown in Table 3, and the preliminarily designed operating parameters are listed in Table 4.

Table 3 Quality of feed water.
Full size table
Table 4 Designed operating parameters of pilot equipment.
Full size table

Monitoring methods for analytical index

The water quality indexes listed in Table 3 were monitored by online instruments with instantaneous values automatically recorded and used for data analysis, except that UV254, TDS, Fe3+ and total number of bacterial colony were tested by the methods prescribed by Standard Examination Methods for Drinking Water (GB/T 5750, China).

The details of corresponding analytical instruments are listed in Tables 5 and 6.

Table 5 Online detection instruments for water quality indexes.
Full size table
Table 6 Laboratory detection instruments for water quality indexes.
Full size table

The molecular weight distribution of organic contaminants in both influent and permeat of the hollow fiber nanofitration unit were analyzed by the ultraviolet spectrometer (Metash, TU-1901, China), and the fouling layers of hollow fiber nanofitration membranes were analyzed by the scanning electron microscope (SEM) and the energy-dispersive X-ray (EDX) spectrometer (Zhongkekeyi, KYKY-EM6900, China).

The transmembrane pressure (TMP, bar), membrane flux (J, LMH), recovery rate (R, %) and reflux ratio of concentrate (r) were respectively calculated through Eqs. (1), (2), (3) and (4).

$${\text{TMP }} = \, \left( {{\text{IP}} + {\text{CP}}} \right) \, /{ 2} - {\text{ PP}}$$
(1)
$${\text{J }} = { 1}000{\text{ PQ }}/{ 43}$$
(2)
$${\text{R }} = {\text{ PQ }}/{\text{ IQ}} \times {1}00$$
(3)
$${\text{r }} = {\text{ CQ }}/{\text{ IQ}}$$
(4)

Results and discussion

Comprehensive performance of hollow fiber nanofitration

Membrane flux and recovery rate of hollow fiber nanofitration

During the pilot testing, the membrane flux and recovery rate of hollow fiber nanofitration system was mainly controlled by adjusting the influent pressure with auxiliary adjustment of the concentrate reflux ratio being made synchronously, which ultimately resulted in seven stages with significant differences (Fig. 4).

Fig. 4
Fig. 4
Full size image

Variation of operating parameters.

From Stage I to Stage II, the average influent pressure increased from 3.26 bar to 3.95 bar with membrane flux rising from no more than 20 LMH to a higher range whose upper bound was 25 LMH, while the recovery rate continuously remained between 55 and 65% without showing noticeable changes. From Stage III to Stage IV, the average influent pressure successively increased to 4.17 bar and 4.27 bar with recovery rate correspondingly raised to the range of 65–70% and the range of 70%-75%. On the foundation of Stage III, the average influent pressure during Stage V and Stage VI successively declined to 4.15 bar and 4.05 bar with recovery rate furtherly raised to the range of 75–80% and then to the range of 80–90%. Lastly during Stage VII, the average influent pressure increased dramatically to 4.52 bar with recovery rate remaining consistently above 85%. In addition, the membrane flux was stably maintained within the range of 23 LMH to 25 LMH from Stage III to Stage VII.

The experimental results manifested that the impact of the osmotic differential pressure was neglectable when the average influent pressure was lower than 4 bar, thus the membrane flux still had the potential to rise and the permeat quantity could increase with the influent quantity, which disabled the improvement of recovery rate14. When the average influent pressure reached the first threshold value of approximately 4.0 bar, the membrane flux approached the maximum value of 25 LMH and ceased to present drastic fluctuations, indicating that the permeat quantity (PQ) tended to stabilize ever since and the variation trend of recovery rate (R) and influent quantity (IQ) started to go in opposite directions referring to Eq. (4).

Theoretically, IQ is positively correlated with the influent pressure (IP) and negatively correlated with the mass transfer resistance (MTR) of hollow fiber nanofitration membranes, which can be calculated through Eq. (5). As a supplementary, f in Eq. (5) represents the friction factor and Re represents the Reynold coefficient, while ρ and μ respectively represent the density and the viscosity of influent, whose changes can be negligible, and d and l respectively represent the diameter and the length of liquid channels, which are fixed values decided by the specification of hollow fiber nanofitration membranes15,16,17.

$${\text{MTR }} = {2}\left( {{\text{IQ}}} \right) \times \rho \times {\text{f}} \times {\text{l}}/ \, \left( {\pi {\text{d}}^{{3}} } \right)$$
$${\text{f}} \propto {1}/{\text{Re}}$$
$${\text{Re }} = { 4}\left( {{\text{IQ}}} \right) \times \rho / \, \left( {\pi {\text{d}}\mu } \right)$$
(5)

As explained by the aforementioned theories, the IQ tended to increase with the rising IP when it approximately remained between 4.0 bar and 4.5 bar, during which the MTR showed no apparent change as the growth quotient of IQ and the reduction quotient of f counterbalanced each other, and thus the R decreased with IP increasing. After exceeding the second threshold value of approximately 4.5 bar, the growth quotient of IQ surpassed the reduction quotient of f, causing a rapid increase of MTR and then a rapid decrease of IQ before accomplishing a dynamic equilibrium at a lower value, and thus the R was elevated on the contrary.

Furthermore, the variation trend of the transmembrane pressure was essentially the same as that of the influent pressure, whose upper limit was obviously lower than the maximum allowable value of 6.0 bar ensuring the normal operation of the pilot testing. Nevertheless, there was a continuous fluctuation of the concentrate reflux ratio existing under the premise that the membrane flux and recovery rate were controlled within the preset ranges, which was revealed by the deviation between the variation trend of concentrate reflux quantity and that of influent quantity and will be discussed in section "Prediction model for the reflux ratio of concentrate".

Retention of contaminants

Retention of suspended solids, organic contaminants and microorganisms

With the pilot testing beginning in March and ending in November, the feed water temperature rose firstly and fell subsequently with the maximum temperature of 37 °C and the minimum temperature of 9 °C. It was manifested that the turbidity and content of bacteria of the feed water were hardly affected by the change of water temperature, which oscillated randomly within an ordinary range, and the removal rate of turbidity was maintained above 90% with that of bacteria consistently remaining as 100%. Benefiting from the aforementioned aspects, the turbidity of permeat steadily remained beneath 0.2 NTU and the existence of bacterial colony in permeat was not detected (Fig. 5).

Fig. 5
Fig. 5
Full size image

Retention effect of suspended solids, organic contaminants and microorganisms.

As the temperature of the surface water of Beijing-Hangzhou Grand Canal was relatively high during spring and summer, it tended to evaporate and be preserved in the humid air before eventually being transformed to surface runoff again through precipitation and carrying organic contaminants from surrounding farmlands into the water source. Therefore, the COD of feed water obviously increased with water temperature gradually rising and subsequently decreased as the water temperature dropped during autumn and winter. On the other hand, it was noticed that the removal rate of COD was barely influenced by the feed water temperature, fluctuating randomly between 60 and 90% which guaranteed the COD of permeat to be kept 1.5 mg/L.

At the end of the pilot testing, which was after 5741 h of operation precisely, ultraviolet spectroscopic analysis of organic content were taken for both an influent sample and a permeat sample from the hollow fiber nanofitration system. As shown in Fig. 6, it was estimated based on the variation of peak areas that the content of organic substances whose molecular weight was less than 600 Da merely decreased by approximately 20% after being processed by the hollow fiber nanofitration system, while the content of organic substances whose molecular weight was 600 Da–1000 Da decreased by approximately 75% and organic substances with a molecular weight higher than 1000 Da could hardly pass through hollow fiber nanofitration membranes. The total removal rate of organic substances was also available according to Fig. 5, which was approximately 70% and extremely close to the actual removal rate of COD. Thus, it was confirmed that the molecular weight cut-off of hollow fiber nanofitration membranes approximately lied between 600 and 1000 Da, moderately higher than that of traditional dense spiral wound nanofiltration membranes which was 500 Da, and the removal rate of COD was primarily impacted by the molecular weight distribution of organic contaminants in feed water18,19.

Fig. 6
Fig. 6
Full size image

Molecular weight distribution of organic contaminants.

In conclusion, the retention effect of the hollow fiber nanofitration system on suspended solids, organic pollutants and microorganisms is sufficient to satisfy the local standards even if the organic contamination of feed water was intensified seasonally and relevant standards strictly limit the content of aforementioned substances in permeat.

Retention of classic inorganic contaminants

It was observed that the classic inorganic salt indexes of the feed water, which was also the effluent of sedimentation tank in this case, possessed higher levels during autumn and winter compared with that during other warmer seasons. Targeting at this phenomenon, we inferred that the effectiveness of coagulation process and sedimentation process was significantly weakened when the feed water temperature was relatively low, resulting from the reduced probability of particulate collision which was caused by the enhanced viscosity of water slowing the velocity of particulate movement, and the recession of the ability of charge neutralization and bridging adsorption which was caused by the reduced hydrolysis rate of flocculants (Fig. 7)20,21.

Fig. 7
Fig. 7
Full size image

Retention effect of classic inorganic contaminants.

After hollow fiber nanofitration system operating for more than 1000 h, the removal rates of TDS, SO42, total hardness and Fe3+ started to present a certain degree of reduction, which showed no tendency of rebound within at least 4000 h, while the removal rates of turbidity, COD and bacteria did not exhibit a similar variation trend (Fig. 4). Because of the charged sulfonic groups fixed on the interface of hollow fiber nanofitration membranes which were made from polyethersulfones, Donnan potential equilibrium was supposed to be generated and impeded the passage of both cations and anions. However, the multivalent ions in the feed water, which failed to be sufficiently captured by the coagulation tank during the low-temperature period, were able to gradually weaken the charge interaction by binding sulfonic groups through coordinate bonds, altering the surface properties of hollow fiber nanofitration membranes and damaging their original characteristics of desalination. Although carrying negative charges, the majority of organic contaminants were constituted by molecules which were much larger than the membrane pores and thus the removal rate of COD could barely be affected by the interaction of charged groups22,23,24,25.

The upgrade of chemical cleaning process became inevitable when the content of Fe3+ in permeat tended to exceed the limitation of local drinking water standard, which will be concretely interpreted in section "Improvement of chemical cleaning process", and prompted the salt rejection rate to recover to its initial level. In detail, the removal rate of TDS recovered from 5–10% to 15–20%, with the removal rate of SO42- recovering from 45–50% to 80–85%, the removal rate of total hardness recovering from 5–20% to 20–25%, and the removal rate of Fe3+ recovering from 50–55% to 85–90%.

Additionally, the TDS, content of SO42− and total hardness of the effluent of sedimentation tank were adequate to meet the local standards for drinking water even though the retention effect of the hollow fiber nanofitration system on the aforementioned substances were limited, benefiting from the non-brackish source water and the relatively lenient standards for inorganic contaminants. Concretely speaking, the TDS, content of SO42−, and total hardness of permeat were respectively kept beneath 450 mg/L, 35 mg/L and 250 mg/L, furtherly optimizing the taste of drinking water.

However, the content of Fe3+ in the effluent of sedimentation tank was possible to get 2–3 times higher than that required by the local standard for drinking water as a result of the detachment and dispersion of iron-containing flocculant radicals, making hollow fiber nanofitration a necessary treatment process whose removal rate of Fe3+ was requested to exceed 65%. The experimental results proved that the removal rate of Fe3+ consistently remained over 80% when hollow fiber nanofitration membranes were not fouled by coordinates or the fouling extent was effectively alleviated, which means that the content of Fe3+ in permeat could easily satisfy the relevant standard under the premise of adopting suitable chemical cleaning strategies.

Retention of emerging contaminants

Featured by risk concealment, environmental persistence and bioaccumulation, emerging contaminants are defined as toxic and harmful organic substances which are capable of posing severe health risks even if ingested in trace amounts, against which enough precautions still haven’t been taken. Selecting from typical emerging contaminants including organic odorants, pharmaceutical and personal care products, exocrine disrupters, antibiotics, etc., the research finally focused on monitoring the retention effect of the hollow fiber nanofitration system on geosmin and 2-MIB, whose molecular weight was less than 200 Da and relevant standards were explicit. It was demonstrated that the removal rates of aforementioned organic odorants were approximately 40–50% with content of geosmin in permeat maintained under 10 ng/L and content of 2-MIB under 9 ng/L, which were acceptable for local drinking water standards (Fig. 8).

Fig. 8
Fig. 8
Full size image

Retention effect of emerging contaminants.

Comparison of performance between hollow fiber nanofitration and spiral wound nanofiltration

The performance differences between hollow fiber nanofitration system and spiral wound nanofiltration system were directly compared based on the status of pilot testing using spiral wound nanofiltration membranes (NF270-400/34i) drawing feed water from an identical source during the same period. According to Table 7, the influent pressure demanded by the hollow fiber nanofitration system to guarantee a favorable recovery rate was noticeably lower than that demanded by the spiral wound nanofiltration system, while the upper limits of membrane flux and recovery rate achieved by two systems were quite similar. In terms of the retention effects of contaminants, the removal rates of suspended particulates and microorganisms of the hollow fiber nanofitration system were basically equal to those of spiral wound nanofiltration system, while the removal rate of organic contaminants was moderately lower and the salt rejection rate declined by approximately 50% compared with those of spiral wound nanofiltration system, indicating that the hollow fiber nanofitration technology is more adaptable for non-brackish surface water which is limitedly polluted with inorganic salts or minerals.

Table 7 Comparison of performance between hollow fiber nanofitration and spiral wound nanofiltration.
Full size table

Prediction model for the reflux ratio of concentrate

The concentration polarization refers to the phenomenon where intercepted substances gradually accumulate at the membrane-liquid interface, spread from the membrane surface to the liquid under the influence of a concentration gradient and ultimately form a stagnant layer with partially-high osmotic pressure contiguous to the membrane surface. Despite of being interfered by the aforementioned stagnant layer, the membrane flux and recovery rate are still able to be restored and stabilized by the concentrate reflux which can adequately suppress the formation of the stagnant layer by elevating the fluid velocity and potential within the liquid channels of hollow fiber nanofitration membranes and diluting the intercepted substances. As the major components of the stagnant layer are suspended particulates and colloids, whose diffusion rate are supposed to be directly affected by water temperature, the influencing factors of the concentrate reflux quantity as well as concentrate reflux ratio theoretically consist of turbidity and temperature of the influent26,27,28,29.

Accordingly, the absolute values of membrane flux (J, LMH), recovery rate (R, %), turbidity of influent (TU, NTU) and temperature of influent (t, ℃) were hypothetically set as independent variables with concentration reflux ratio (r) as the dependent variable, and the analysis of variance (ANOVA) was conducted using R software to verify the assumptions.

As shown in Table 8, nine types of parameter combination were worth to be paid attention to with properly high F values and limited P values which were less than 0.05, indicating qualified confidence intervals and significant correlations between r and the nine types of independent variable combination. Furthermore, linear regression fitting of the nine groups of parameters was performed using R software.

Table 8 Results of ANOVA.
Full size table

According to Table 9 and Fig. 9, the prediction model for concentrate reflux ratio was permitted to be created as a linear regression model, with the optimal one exhibited by Eq. (6).

$$\begin{aligned} {\text{r }} & = { 34}.{43 } - { 1}.{\text{4339J }} - { 1}0.{\text{2787 TU }} - { 1}.{\text{7233 t }} \\ & \quad+ \, 0.{489}0{\text{ J}} \times {\text{TU }} + \, 0.0{\text{778 J}} \times {\text{t }} + 0.{\text{5383 TU}} \times {\text{t }} \\ & \quad- \, 0.0{\text{248 J}} \times {\text{TU}} \times {\text{t}} \\ \end{aligned}$$
(6)
Table 9 Results of linear regression.
Full size table
Fig. 9
Fig. 9
Full size image

Model diagnostic graph of r ~ J*TU*t.

The multiple linear regression model adopted membrane flux, turbidity of influent and temperature of influent as independent variables, whose degree of fitting appeared to be approximately 0.25, which was moderately deviated from the ideal degree of fitting (≥ 0.8). Overall, the attained model can function for judging the probable range and the variation trend of concentrate reflux ratio and reduce the frequency of manual adjustments.

Based on Eqs. (2) and (6), an estimation formula for straightforwardly calculating the concentrate reflux quantity (CQ, m3/h) was furtherly acquired and shown by Eq. (7).

$$\begin{aligned} {\text{CQ }} & \approx \, ({34}.{43 } - { 333}.{\text{4651PQ }} - { 1}0.{\text{2787 TU }} \\ & - { 1}.{\text{7233 t }} + { 113}.{72}0{\text{9 PQ}} \times {\text{TU }} + {18}.0{93}0{\text{ PQ}} \times {\text{t }} \\ & + \, 0.{\text{5383 TU}} \times {\text{t }} - \, 0.{\text{5767 PQ}} \times {\text{TU}} \times {\text{t}}) \times {\text{IQ}} \\ \end{aligned}$$
(7)

Interaction between membrane cleaning process and membrane fouling form

Selection methodology of hydraulic cleaning mode

Mostly prevented from entering the small membrane pores of hollow fiber nanofitration membranes, biopolymers and macromolecular humic substances tend to form cake layers over the surfaces of membranes, which becomes the most common pattern of hollow fiber nanofitration membrane fouling. Meanwhile, a part of micromolecular organics including humic acid are possible to pass through the twisted channels within the cake layers and then adhere inside the membrane pores, leading to irreversible membrane fouling patterns such as pore narrowing and pore blockage30,31,32.

According to the preliminary experimental results shown in Table 10, the fouling pattern of hollow fiber nanofitration membranes could be tentatively characterized by the differential between the influent pressure increment (ΔIP) and the transmembrane pressure increment (ΔTMP) during one hydraulic cleaning cycle. In general, the membrane fouling pattern was dominated by cake layers when ΔIP − ΔTMP ≥ 0, which was otherwise dominated by pore narrowing and pore blockage when ΔIP − ΔTMP < 0.

Table 10 SEM image of hollow fiber nanofitration membrane surface.
Full size table

Compared to spiral wound nanofiltration membranes which can only be forwardly flushed, more hydraulic cleaning modes are optional for hollow fiber nanofitration membranes. However, it might reduce the total output of permeat as well as raising the energy consumption per ton of permeat during a long-term operation if the hydraulic cleaning process is unreasonably lengthened, which leads to the necessity of establishing an appropriate selection methodology of hydraulic cleaning mode.

According to Table 10, all hydraulic cleaning modes presented acceptable effects with the membrane fouling pattern dominated by cake layers, while only the mode of backwash presented obvious effect with the membrane fouling pattern dominated by pore narrowing and pore blockage. As a result, it is more feasible to choose the mode of backwash when ΔIP-ΔTMP < 0 and to choose the mode of forward flush or air scouring whose energy consumption is more limited when ΔIP-ΔTMP ≥ 0. Practically for the pilot testing, both forward flush and air scouring were carried out simultaneously when ΔIP-ΔTMP ≥ 0 to guarantee the cleaning effect.

Improvement of chemical cleaning process

To solve the problem of salt rejection rate attenuation mentioned in section "Retention of classic inorganic contaminants", an attempt of using the chelating agent of ethylene diamine tetraacetic acid (EDTA) was made to remove the ions originally complexed with sulfonic groups. Notionally, the four carboxyl groups in one EDTA molecule can dehydrogenize in an alkaline environment, allowing them to combine divalent and trivalent ions along with the two amine groups and consequently generate soluble hexacoordinate compounds, whose coordinate bonds are supposed to have greater adsorption strength compared with the monocoordinate bond of one sulfonic group33,34,35,36,37.

As shown in Fig. 10 and Table 11, alkalized EDTA was able to present the optimal cleaning effect for the complex layer collected during the low-temperature period which failed to be eroded by normal chemical cleaning agents, promoting the detachment of calcium ions and silicon dioxide. It is proved in Fig. 7 that the removal rate of inorganic contaminants of hollow fiber nanofitration membranes apparently recovered after adding 1% EDTA to the conventional alkaline agent at the point of 5522-h operation.

Fig. 10
Fig. 10
Full size image

SEM image of cleanout fluid after dehydration. (a) 4000 mg/L citric acid, (b) 1% EDTA, (c) 1% EDTA + 250 mg/L NaOH.

Table 11 EDX analysis of cleanout fluid after dehydration.
Full size table

Evaluation of direct operating expense

As shown in Table 12, the direct operating expense of the hollow fiber nanofitration system was relatively the lowest when the average influent pressure was approximately 4.05 bar with recovery rate ranging between 80 and 90%, which was approximately 0.5278 RMB per ton of permeat. Therefore, the operating condition of stage VI was regarded as the most economically viable one. Worthwhile to be emphasized, the expenditure of adding EDTA was not taken into account for calculating the agent consumption since it only needed to be intermittently dosed when the salt rejection rate of the hollow fiber nanofitration system severely declined.

Table 12 Statistics of direct operating expense.
Full size table

The detailed calculation processes of energy consumption and agent consumption are elaborated in Tables 13 and 14 taking stage VI as an example.

Table 13 Energy consumption accounting of hollow fiber nanofitration.
Full size table
Table 14 Agent consumption accounting of hollow fiber nanofitration.
Full size table

According to Tables 15 and 16, the direct operating expense of spiral wound nanofiltration system was approximately 0.5483 RMB per ton of permeat as the relevant pilot testing was conducted at the same location during the same period possessing the recovery rate of 80–90%, which was slightly higher than the direct operating expense of hollow fiber nanofitration system gaining the same recovery rate.

Table 15 Energy consumption accounting of spiral wound nanofiltration.
Full size table
Table 16 Agent consumption accounting of spiral wound nanofiltration.
Full size table

Scheme of relative full-scale demonstration

Currently, the drinking water advanced treatment demonstration to be constructed at the location of pilot testing has come into the period of engineering design. Based on the conclusions of the pilot study, the average membrane flux is designed as 24 LMH and the average recovery rate is designed as 85%, with the processing scale of 7200 m3/d and production scale of no less than 6000 m3/d. The blueprint of the core equipment is shown in Fig. 11.

Fig. 11
Fig. 11
Full size image

Cross-section diagram of hollow fiber nanofitration membrane universal platform for demonstration.

Considering that hollow fiber nanofitration membrane elements with diverse specifications and models are available on the market, the hollow fiber nanofitration membrane universal platform is designed to be compatible with different membrane elements of various surface area and operation modes including inside-out cross-flow filtration and outside-in cross-flow filtration. Separated by end plates, the pressure vessel of hollow fiber nanofitration membrane universal platform consists of top chamber, middle chamber and bottom chamber, while the middle chamber is prepared for containing membrane elements whose water distribution pipes are interconnected with the top chamber or the bottom chamber. When the operation mode of inside-out cross-flow filtration is demanded, the feed water inflows through the bottom inlet and bottom chamber before entering hollow fiber nanofitration membrane elements, while permeat outflow through the middle chamber and middle outlet with concentrates outflowing through the top chamber and top outlet. When the operation mode of outside-in cross-flow filtration is demanded, the feed water inflows through the middle inlet and fills the middle chamber before entering hollow fiber nanofitration membrane elements, while permeat outflow through the top/bottom chambers and top/bottom outlets with concentrates outflowing through another middle outlet38.

Conclusion

With the permeat-production scale of approximately 25 m3/d, the micro-polluted surface water sourced from Beijing-Hangzhou Grand Canal, which was pretreated by biological aerated filters and coagulation sedimentation tanks, was utilized as the feed water of the pilot-scale hollow fiber nanofitration system. During the 240-day pilot testing, the operating conditions of a certain stage can be considered as the optimal ones with average influent pressure of approximately 4.05 bar and recovery rate ranging between 80 and 90%, whose operating expense was approximately 0.5279 RMB per ton of permeat and was assessed to be lower than that of spiral wound nanofiltration system presenting the same performance.

Meeting both national standards and local standards of Jiangsu province for high-quality drinking water, the turbidity, CODMn, total number of bacterial colony, TDS, content of SO42-, total hardness, content of Fe3+, content of geosmin and content of 2-MIB of the permeat of the hollow fiber nanofitration system were respectively kept beneath 0.2 NTU, 1.5 mg/L, 0 CFU/ml, 450 mg/L, 35 mg/L, 250 mg/L, 0.2 mg/L, 10 ng/L and 9 ng/L. Accordingly, the removal rates of suspended particulates and microorganisms of the hollow fiber nanofitration system were basically equal to those of spiral wound nanofiltration system, while the removal rate of organic contaminants was moderately lower and the salt rejection rate declined by approximately 50% compared with those of spiral wound nanofiltration system, indicating that the hollow fiber nanofitration technology is more adaptable for non-brackish surface water which is limitedly polluted with inorganic salts or minerals.

Due to the complicated factors such as the seasonal variation of water temperature, several problems occurred during the research involving the unpredictable concentrate reflux ratio with violent fluctuations, the attenuation of salt rejection rate resulting from the membrane fouling layer which were unable to be decomposed by conventional chemical cleaning agents, etc. Corresponding solutions were explored through the pilot testing whose shortcomings were still concerned about. For example, the fitting degree of the multiple linear regression model targeting at predicting the concentrate reflux ratio was relatively limited, which might induce non-negligible errors, and the agent consumption for restoring the salt rejection rate was considerably higher than that for normal chemical cleaning processes. As a result, corrections and improvements need to be made for the aforementioned technical proposals through the long-term data analysis and auxiliary experiments during the demonstration.