Abstract

Membrane backwash is one of the most widely used membrane regeneration techniques in large-scale desalination operations and water and wastewater treatment applications. It is necessary in order to enhance membrane life and is critical in managing the cost of pure water. The estimation of optimal backwash conditions is very important for improved hollow fiber membrane (HFM) operations. A unique feature in HFM backwash operations is the lumen side pressure drop, which leads to variation in backwash water flow across the fiber length. In this work, the effect of fiber diameter, membrane structural properties and backwash water pressure and temperature on flow distribution across the membrane length is studied for HFM modules. An analytical model for HFM backwash is developed, and model predictions are compared with measured backwash water flow variation over the fiber length. Experimental results show that the backwash flow variation over the fiber length is minimized by maintaining low backwash water pressure. Simultaneously, minimum backwash flux for effective cleaning may be achieved by increasing backwash water temperature. Homogeneous flow distribution during backwash improves backwash efficiency or forward flow rejuvenation, which can also be achieved by optimizing fiber diameter and membrane permeability. The validated mathematical model may be used for the optimization of backwash operating conditions and HF dimensions to achieve homogeneous backwash flow distribution across the membrane length.

HIGHLIGHTS

  • The effect of the backwash operating condition in hollow fiber (HF) ultrafiltration (UF) membrane.

  • HFUF membrane backwash optimization.

  • Modeling of the HFUF membrane backwash operation.

  • The analytical solution for the HFUF model.

  • Regeneration of the HFUF membrane fouled by raw Brahmaputra river water and synthetic greywater.

INTRODUCTION

Hollow fiber (HF) ultrafiltration (UF) membranes are often used for desalination pretreatment and in water and waste water purification applications. The HF configuration provides a high membrane surface area per unit module volume and is extensively used in applications like water purification units (Xia et al. 2004). Membrane fouling is an unavoidable phenomenon that occurs on membrane surfaces, which leads to the decline of membrane flux. Therefore, it is essential in almost all the applications that the membrane is cleaned at regular intervals by using appropriate advance process control algorithms (Chew et al. 2018). Typical membrane maintenance activities are (i) physical cleaning, (ii) chemical cleaning and (iii) partial or full replacement of the membrane module (Lin et al. 2010).

Physical cleaning methods use mechanical forces to dislodge and remove deposits from the membrane surfaces and pores. Conventional physical methods include forward and reverse flushing, backwashing, air flushing, sponge ball cleaning and CO2 back permeation (Ebrahim 1994). Unconventional methods for physical cleaning include ultrasound (Muthukumaran et al. 2005) and electrical and magnetic fields (Tarazaga et al. 2006). The unconventional methods are yet to be proven in operating water and waste water treatment plants or in domestic water purification applications.

Chemical cleaning involves both chemical and physical processes (Porcelli & Judd 2010). Chemical cleaning of membranes involves three steps, namely (i) mass transfer of the cleaning chemical from the bulk solution to the fouling layer, (ii) the chemical reaction between the cleaning chemical and the deposits in the membrane fouling layer and (iii) mass transfer of the deposits from the fouling layer to the bulk solution, which is controlled by hydrodynamics. The efficiency of cleaning chemicals strongly depends on the chemical reactivity of the cleaning chemicals since the third step of mass transfer can take place only after the deposit–membrane and deposit–deposit bonds have been weakened by the chemical reaction. Optimized membrane cleaning processes require the identification of suitable cleaning chemicals, its composition and flow conditions. Excessive use of cleaning chemicals may also cause early degradation of the membranes (Hajibabania et al. 2012; Shi et al. 2014). For example, it has been reported that excessive exposure to sodium hypochlorite led to early degradation of the polyvinylidene difluoride (PVDF) membrane (Hajibabania et al. 2012).

If physical or chemical cleaning is unable to restore membrane flux, then partial or full replacement of membrane modules is undertaken (Regula et al. 2014). The partial membrane module replacement is recommended, wherein multiple membrane modules are connected in series. The cost of water production and energy consumption increases when membranes foul/scale.

Among the above alternatives for membrane system maintenance, physical techniques are applied more frequently to regenerate the membrane. In domestic water purification applications, chemical cleaning is not a practical solution from an implementation perspective and frequent replacement of the membrane may not be acceptable. Among physical cleaning methods, it has been proven that backwashing with permeate water is robust and easy to implement in domestic water purification applications (Hajibabania et al. 2012). In HF membrane backwash, the water is circulated opposite to the regular water flux direction. Many researchers have reported studies comparing the membrane regeneration efficiency of physical cleaning methods. In general, membrane regeneration efficiency is estimated by comparing the membrane flux before and after performing membrane cleaning.

Nigel et al. (2013) studied membrane fouling and physical cleaning techniques for membrane regeneration on PVDF HFUF membranes. They report that air-assisted backwash in combination with a short period of relaxation might be beneficial for removing deposits while minimizing net productivity losses. HF membrane (HFM) packing density is another important factor to be considered to achieve energy-efficient membrane backwash.

Günther et al. (2010, 2012) calculated the velocity and flux profile for both forward filtration and backwash operations of UF HFMs. They report variation in fouling rate over the fiber length with respect to fiber packing density due to the pressure gradient across HF membrane bore. When the density of fibers is high, filtration occurs preferentially at the bottom of the fiber. In contrast, when the packing density is low, the permeate flow is higher at the top of the fiber. Their experiments conclude that in the high fiber packing density HF module, higher particle deposition is found at the bottom of the fiber for both outside/in and inside/out filtration modes. Their results show that a moderate packing density leads to higher filtration flux and a homogeneous flux profile along the fiber length. Fiber length is also another important parameter to achieve the homogeneous flux profile over the fiber length.

Yoon et al. (2004) have optimized the fiber diameter and length to minimize both filtration and backwash energy consumption using an experimental method. They observed that a longer HFUF module with larger fiber diameter can achieve better efficiency in case of membrane bioreactors. In systems operated under suction from one end, they report an optimum fiber diameter of 2.16–3.25 mm for a fiber length of 1.03–1.90 m, at 30 kPa trans-membrane pressure (TMP). In case of suction applied at both ends, they report a fiber diameter of 1.34–2.05 mm for a fiber length of 1.01–1.90 m optimal for the same flux and TMP.

Wang et al. (2014) reported the importance of backwash parameters through a mathematical model and experimental studies. They propose that to achieve a better membrane backwash efficiency, the optimal parameters should meet the requirement of delivering the backwash liquid to the most distant fiber surfaces. To achieve their objective, the filtration and backwashing experiments were performed with different HF length membranes. They report that when the fiber length was increased, the backflushing time also increased due to the non-homogeneous flux profile along the fiber length. This variation of backwash flux was verified by comparing model predictions with experimental values of backwash flux.

Venkatesh & Senthilmurugan (2016) reported the performance of UF membrane backwash for greywater to recycle the application. In order to regenerate the UF membrane, backwash was performed with pure water by keeping the treated water tank at a height of 1 foot above the UF module. Pure water is supplied through the fiber lumen side using gravity forces such that the UF membrane is backwashed. They observed that after each backwash cycle, the membrane was able to revert to values nearby to the original flux conditions.

The design and operating condition optimization of the UF HF membrane system has been reported by Guadix et al. (2004) based on detailed models of membrane fouling; however, membrane cleaning models are not incorporated. However, in practice, the membrane cleaning is an essential step of the membrane operations (Yoon et al. 2004). In UF processes, membrane cleaning also consumes a considerable amount of energy and chemicals with respect to the overall plant operation. The energy consumed for the physical cleaning process is also one of the critical factors that have to be considered while optimizing specific energy consumption of the overall water or wastewater treatment plant. Also, to the best of the author's knowledge, a validated mathematical model for the HFUF membrane backwash process is not reported in the literature.

The main objective of this work is to study HFUF membrane backwash efficiency experimentally with respect to backwash operating conditions, HF membrane dimensions and membrane permeability. The study includes the identification of optimal backwash conditions to achieve uniform backwash water flux across the fiber length. The mathematical model of HFUF backwash and its analytical solution is presented in this work. The proposed model is validated with experimental data generated from a different section of HF length. Finally, the validated model is used to calculate the optimal backwash operating condition (such as backwash water pressure, flow and temperature) for different HF dimensions such as length and diameter. The efficiency of the optimized backwash parameter for the HFUF membrane is validated by purifying the synthetic greywater.

THEORY

HFUF membrane module

HFM modules are typically available in two configurations. In the first, the fiber is open at one end, while in the second, both ends of the fiber are open. Modules open at both ends are commonly used in two-phase contactor applications, while single end open modules are used in UF, microfiltration (MF), nanofiltration (NF) and reverse osmosis (RO) applications. Recent research suggests that for larger fiber lengths, both ends open modules can give improved energy-efficient filtration and improved backwash due to the minimal pressure drop across the fiber lumen (Chen et al. 2003; Peiris et al. 2013). In this work, HFUF modules used in domestic water purification applications (Yancheng Yongkang Filter Equipment Co., Ltd, Yancheng, China) are used, and its configuration is given in Supplementary material, Figure A1.

One end open HFM modules having three different specifications are used in this experimental work, and the specifications are shown in Table 1. They are commercially available and are used either for pretreatment before reverse osmosis membrane units or as inline water purifiers. In this work, home-use HF membrane modules are chosen since membrane backwash studies with such filters have not been reported in the literature. Considering the growing popularity of these modules for a variety of domestic water purification applications, understanding backwash effectiveness in these modules is of interest. Since fiber diameter is very small, these modules are operated with flow direction from out to in to avoid irreversible clogging of the fiber lumen side. During filtration, the water flux is out to in, while it is in to out during backwash.

Table 1

HFUF membrane module dimensions

HFUF membrane module parametersMembrane module 1Membrane module 2Membrane module 3
Fiber inner diameter (mm) 0.3 0.7 0.3 
Fiber outer diameter (mm) 0.5 0.9 0.5 
Fiber length (mm) 400 191 215 
Number of fibers (–) 1,400 1,400 1,300 
HFUF membrane module parametersMembrane module 1Membrane module 2Membrane module 3
Fiber inner diameter (mm) 0.3 0.7 0.3 
Fiber outer diameter (mm) 0.5 0.9 0.5 
Fiber length (mm) 400 191 215 
Number of fibers (–) 1,400 1,400 1,300 

Three different HF membrane module configurations with different permeabilities and different physical dimensions are chosen for this experimental study. The membrane module dimensions are given in Table 1.

Model

The mathematical model of the HF membrane module is developed by applying both mass and momentum balance across the HF length. The water is transported through the membrane fiber from outside to lumen side during the normal membrane operation and in the reverse direction during backwash (refer to Supplementary material, Figure A2). Pure water flux across the membrane at a given point is proportional to TMP and can be written as follows.

For filtration mode:
formula
(1)
For backwash mode:
formula
(2)
where A is the membrane water permeability, Pi(x) is the pressure at the fiber lumen side, and Po is the pressure outside the fiber. Under the assumption that water flow through the UF membrane is pore flow, the membrane permeability is thus a function of membrane pore diameter, membrane pore length and water viscosity and can be expressed as
formula
(3)
where – membrane structural parameter, dp – membrane pore diameter, Lp – pore length and m – viscosity of water.
formula
where
The ratio between pore diameter and pore length is unknown and is very difficult to be estimated by direct methods. The ratio can be estimated indirectly by measuring water flux at different pressure conditions for a flat sheet membrane. However, in the case of HF membranes, exact fiber lumen pressure is unknown. Therefore, the simultaneous estimation of the fiber lumen pressure profile and the above ratio is very important for HF modeling. The fiber lumen side pressure drop can be described by applying the Hagen–Poiseuille equation (Chatterjee et al. 2004), where no slip at the lumen surface is assumed.
formula
(4)
where vf – the velocity of water in fiber lumen, df – fiber inside diameter and m – viscosity of water. The water mass balance across the HFM length for filtration mode is written as follows:
For filtration mode:
formula
(5)
For backwash mode:
formula
(6)

To calculate the filtration flux, Equations (1), (2), (4) and (5) are solved, and similarly to calculate the backwash flux, Equations (1), (3), (4) and (6) are solved.

Analytical solution

From the above model equations
formula
(7)
For backwash mode:
formula
(8)
Differentiating the Hagen–Poiseuille equation with respect to x gives,
formula
(9)
Substituting Equation (6) in Equation (7)
formula
(10)
formula
(11)
Let
formula

Boundary conditions: at and at .

Applying boundary conditions, the solution is
formula
(12)
Substituting Equation (9) in Equation (1), the model equation for membrane flux is
formula
(13)
To calculate the backwash flow rate,
formula
formula
(14)

As described in the model Equation (12), the pressure variation along the fiber length is a function of inlet backwash pressure, membrane structural parameters, the viscosity of backwash water and fiber dimensions. The viscosity of water is a function of water temperature. Therefore, homogeneous flux distribution can be achieved by optimizing fiber diameter, backwash water pressure and temperature. The selection of these parameters is subjected to constraints like fiber mechanical strength and minimum homogeneous backwash flux (m3/m2 s) required to regenerate the membrane. As shown in Figure 1(a) and 1(b), the predicted backwash water flux profile along the membrane length can be made uniform by increasing the fiber diameter and decreasing the membrane structural parameter.

Figure 1

(a) Effect of fiber diameter on flux distribution at a backwash inlet pressure of 27.6 KPa, temperature 30 °C and membrane structural parameter 1.5298 × 10−12 m, (b) effect of the membrane structural parameter on flux distribution at backwash pressure 27.6 KPa, temperature 30 °C and fiber diameter 300 μm, (c) effect of backwash temperature on flux distribution at backwash pressure 27.6 Pa, fiber diameter 300 μm and membrane structural parameter 1.5298 × 10−12 m and (d) effect of inlet backwash pressure on backwash flow distribution at 30 °C backwash temperature, fiber diameter 300 μm and membrane structural parameter 1.5298 × 10−12 m.

Figure 1

(a) Effect of fiber diameter on flux distribution at a backwash inlet pressure of 27.6 KPa, temperature 30 °C and membrane structural parameter 1.5298 × 10−12 m, (b) effect of the membrane structural parameter on flux distribution at backwash pressure 27.6 KPa, temperature 30 °C and fiber diameter 300 μm, (c) effect of backwash temperature on flux distribution at backwash pressure 27.6 Pa, fiber diameter 300 μm and membrane structural parameter 1.5298 × 10−12 m and (d) effect of inlet backwash pressure on backwash flow distribution at 30 °C backwash temperature, fiber diameter 300 μm and membrane structural parameter 1.5298 × 10−12 m.

Similarly, the effects of backwash operating conditions such as backwash pressure and backwash water temperature on flux distribution are shown in Figure 1(c) and 1(d). Further results conclude that the homogeneous flux distribution cannot be achieved by varying the temperature alone. Therefore, for optimized membrane dimensions, the desired homogeneous backwash flux can be achieved by decreasing the backwash pressure. Concurrently, the minimum backwash flux required for efficient membrane cleaning can be maintained by increasing backwash water temperature at lower backwash pressure (Figures 1(d) and 2). For example, at a backwash pressure of ∼50 KPa, a similar homogeneous flux distribution curve is achieved for both the backwash temperatures (30 and 60 °C), but the minimum backwash flux increased by 66.7%, i.e. from 0.6 × 10−5 to 1 × 10−4 m3/m2 s at 30 and 60 °C, respectively.

Figure 2

Effect of inlet backwash pressure on backwash flow distribution at 60 °C backwash temperature, fiber diameter 300 μm and membrane structural parameter 1.5298 × 10−12 m.

Figure 2

Effect of inlet backwash pressure on backwash flow distribution at 60 °C backwash temperature, fiber diameter 300 μm and membrane structural parameter 1.5298 × 10−12 m.

Further, to find an optimal l for a given membrane length, the term flux deviation is defined to quantify homogeneous flux distribution. The total flux deviation of a membrane is defined as the summation of absolute relative error between actual flux to average flux along the membrane length. If flux deviation is equal to zero, then flux along the length is homogeneous. As shown in Figure 3, the estimated flux deviation is found to be less than 8% when l < 0.2 and fiber length <0.4. Therefore, while designing the HFUF system, one can achieve homogeneous flux (less than 8% flux deviation) by satisfying HF design parameters such as l < 0.2 and fiber length <0.4.

Figure 3

Flux deviation with respect to membrane design parameters for any Pinlet, df and back wash temperature.

Figure 3

Flux deviation with respect to membrane design parameters for any Pinlet, df and back wash temperature.

EXPERIMENTAL

Pure water flow distribution measurement across fiber length

The flow measurement along the HFUF membrane length is very important to validate the above model for backwash and filtration operations. The experimental setup for flow distribution is shown in Figure 4. Here, pure water was used for the experiment that is pumped from the backwash water tank to the HFUF membrane lumen side. The backwash water tank was equipped with a temperature controller unit having an inbuilt temperature sensor and heating element. The backwash water pressure and flow rate were controlled by manipulating a bypass valve. The six identical HFUF membranes of various lengths were used to measure flow distribution across the different section of the HF length. The lengths of six different membranes are as follows: 6, 12.5, 18.5, 25, 31 and 40 cm. The pure water backwash water flux is a measure for all six different length membrane modules. Thus, pure water flow for six different sections (0–6, 6–12.5, 12.5–18.5, 18.5–25, 25–31 and 31–40 cm) are obtained by calculating the difference between two consecutive lengthy membranes. Since the membrane modules are fabricated manually, the exact 6 cm section could not be maintained. The flow rate with respect to the different lengths was measured and recirculated back to the backwash water tank. The experiments were performed by varying backwash pressure and backwash water temperature for three types of the membrane module reported in Table 1.

Figure 4

Experimental setup for HFUF membrane backwash flux distribution measurement.

Figure 4

Experimental setup for HFUF membrane backwash flux distribution measurement.

Fouling and backwash experiments

After measuring the backwash flow distribution across the membrane length with new membrane modules, the membrane fouling and backwash experiments were performed to evaluate the effect of backwash water temperature and pressure on the membrane backwash efficacy. The experimental setup used for membrane fouling and backwash operation is shown in Figure 5. Two different types of feed water source are used to study the effect of the backwash parameter, namely (i) Brahmaputra river water and (ii) synthetic greywater. In the first case, the raw Brahmaputra river water was used as feed water for the membrane fouling experiments. This water has its unique combination of high suspended particulate matter, low total dissolved solids and high organic content. Feed water having around 100 NTU turbidity was stored in feed water tanks with provisions to avoid settling of particulate matter. The feed water tank was maintained at 30 °C using a temperature control unit. A low-pressure centrifugal pump along with a pressure control unit was used to feed water to the HFUF membrane module. The membrane module was placed in a membrane casing, which is commonly used in domestic water purification applications. The feed water was supplied from outside the HFM module, and product water was collected from the fiber lumen side. The treated product water flow rate was measured by an inline rotameter fixed before the product water tank. The UF HFM module was operated continuously until product water flow decreased to around 85% of the original membrane flow. Then, the membrane operation was stopped to perform the backwash operation. During membrane backwash, the treated pure water from the product water tank was pumped at constant pressure to fiber lumen to create a flux opposite to the direction of the normal membrane operation.

Figure 5

Membrane fouling and cleaning experimental setup.

Figure 5

Membrane fouling and cleaning experimental setup.

To study the effect of backwash water physical parameters, two membrane fouling and backwash experiments were performed on two identical hollow membrane modules fouled using the same feed water. Membrane backwash was performed at two different backwash water temperatures, i.e. 30 and 50 °C. One set of membrane fouling and backwash process is called one cycle, and a total of three cycles of experiments were performed to study the effect of water temperature and pressure on membrane regeneration efficiency. Feed and product turbidity were measured by using a ‘Eutech’ portable turbidity meter (model number TN100) with a range of 0.01–19.99, 20–99.9, 100–1,000 NTU. Feed and product water total dissolved solids (TDS) were measured using ‘Eutech’ bench conductivity/TDS/temperature meter (model number ECCON70043S).

In the second case, a similar experiment was performed by using synthetic greywater. It was formulated using 15 g of Ariel (Procter and Gamble Company, India) detergent dissolved in 50 L of freshwater together with CaCl2. An increasing amount (1, 2, 3, 4, 5, 6, 7, 8, 9, 10 mL) of CaCl2 solution (at 10,000 ppm) was dissolved in the detergent solution (200 mL), and solution turbidity was measured for precipitation. Active precipitation was observed when 6 mL of CaCl2 was mixed in 200 mL of the detergent solution. Hence, a lower CaCl2 concentration (≈250 ppm) was used maintained to formulate a synthetic greywater solution for further fouling studies. The characteristics of synthetic greywater used in this study are as follows: conductivity – 1,084 μS/cm, TDS – 706 ppm, turbidity – 57.1 NTU, chemical oxygen demand (COD) – 234 ppm, biochemical oxygen demand (BOD) – 14 ppm and pH – 9.6.

RESULTS AND DISCUSSION

As shown in Table 1, three different HFUF membrane modules were used in this study. Modules 1 and 3 were chosen with the same fiber diameter but different fiber length. Modules 2 and 3 were chosen with similar length but different fiber diameters. The effect of pressure and temperature on backwash water flux distribution was studied at 2–4 psig and 30–50 °C, respectively.

Effect of backwash conditions on flow distribution

The backwash flow distribution across the membrane length for all three membranes is presented in Table 2. As described in the experimental procedure, the number of sections per membrane module will be varying with respect to module total length. For example, the membrane module 1 has six sections (6, 6.5, 6, 6.5, 6, 9 cm, respectively), and membrane modules 2 and 3 have three sections (7.1, 6.5, 5.5 cm and 7.5, 6.5, 7.5 cm, respectively).

Table 2

Backwash flow distribution with variation in backwash water temperature and pressure

Temperature (°C)Backwash flow distribution % per unit length membrane 1
Backwash pressure 2 psig
Section 1Section 2Section 3Section 4Section 5Section 6
30 100.00 81.67 79.27 49.91 48.89 44.43 
50 100.00 88.42 83.89 51.73 39.14 32.34 
 Backwash pressure 4 psig 
30 100.00 68.47 66.05 60.49 39.70 37.01 
50 100.00 69.78 67.11 62.73 33.60 29.72 
Temperature (°C)Backwash flow distribution % per unit length membrane 1
Backwash pressure 2 psig
Section 1Section 2Section 3Section 4Section 5Section 6
30 100.00 81.67 79.27 49.91 48.89 44.43 
50 100.00 88.42 83.89 51.73 39.14 32.34 
 Backwash pressure 4 psig 
30 100.00 68.47 66.05 60.49 39.70 37.01 
50 100.00 69.78 67.11 62.73 33.60 29.72 

Backwash flow distribution % = 100 * Backwash flow per unit length at section n/section 1.

Backwash flow per unit length of the modules is found to be decreasing with the length of the membrane modules. Similar results are reported in the literature (Yoon et al. 2004). For example, as shown in Figure 6, the predicted fiber pore pressure for membrane 1 is found to be decreasing from 2 to 1.08 psig from the initial section to the final section, and the corresponding backwash flow rate per unit fiber length declines from 0.422 to 0.156 mL/s cm.

Figure 6

Backwash water flow and fiber lumen pressure profile for membrane module 1 at 2 psig backwash pressure and 30 °C temperature.

Figure 6

Backwash water flow and fiber lumen pressure profile for membrane module 1 at 2 psig backwash pressure and 30 °C temperature.

Figure 7

Normalized water flow rate with respect to fouling and backwash conditions for Brahmaputra river water.

Figure 7

Normalized water flow rate with respect to fouling and backwash conditions for Brahmaputra river water.

A new term called ‘backwash flow distribution percent (%) per unit length’ is introduced to study the flow distribution improvement with respect to backwash parameters such as pressure and temperature for different fiber dimensions. The backwash flow distribution % per unit length of section n is defined as the percentage of the ratio between backwash flow per unit length at section n and the backwash flow per unit length at section 1. Since the flow at the entry (i.e. section 1) is always expected to be greater than the flow in the other sections, the backwash flow distribution % per unit length is always 100 for section 1 and decreases for other downstream sections.

As shown in Table 2, improved backwash flow distribution % per unit length is observed when the backwash water temperature is increased. For example, while increasing the backwash water temperature from 30 to 50 °C at constant backwash pressure, the water viscosity is reduced by 31.42%, which leads to increased membrane permeability. Therefore, improved overall backwash flow distribution % along the fiber length is achieved. Similar flux distribution behavior is observed in membranes 2 and 3.

The effect of backwash pressure on improved backwash flow distribution % per unit length is presented in Table 2, and it can be observed that the backwash flow distribution is more uniform when backwash pressure is decreased. For example, in the case of membrane 1, the backwash flow from section 2 to 6 is higher, and a better backwash water distribution is achieved for a backwash pressure of 2 psig in comparison with 4 psig. Hence, backwash flow distribution is found to become homogeneous for lower backwash pressure and higher backwash temperature.

Effect of fiber dimensions on flow distribution

The effect of fiber length on flow distribution across various membrane lengths is shown in Table 3. The slope of the flow distribution along the fiber length is found to be low in membrane module 1 compared with membrane module 3. This may be due to the higher permeability of membrane 3. Estimated permeability values for three different membrane modules are reported in Table 6.

Table 3

Flow distribution with respect to fiber length

Membrane nameBackwash flow distribution % at 2 psig and 30 °C
Section 1Section 2Section 3Section 4Section 5Section 6
Membrane 1
length 400 mm
di = 0.3 mm,
do = 0.5 mm 
100.00 81.67 79.27 49.91 48.89 44.43 
Membrane 3Section 1Section 2Section 3
length 215 mm
di = 0.3 mm,
do = 0.5 mm 
100.00 23.29 11.33    
Membrane nameBackwash flow distribution % at 2 psig and 30 °C
Section 1Section 2Section 3Section 4Section 5Section 6
Membrane 1
length 400 mm
di = 0.3 mm,
do = 0.5 mm 
100.00 81.67 79.27 49.91 48.89 44.43 
Membrane 3Section 1Section 2Section 3
length 215 mm
di = 0.3 mm,
do = 0.5 mm 
100.00 23.29 11.33    

Backwash flow distribution % = 100 * Backwash flow per unit length at section n/section 1.

As shown in Table 4, the flow distribution is found to be improved in membrane module 2 having a larger fiber diameter compared with membrane 3 having a smaller fiber diameter. As expected, smaller fiber diameter leads to higher pressure drop per unit fiber length, and thus, the initial (entry) sections of fiber have larger flow compared with the end sections of the fiber. The optimization of fiber parameters is a very important factor in achieving a uniform backwash flow distribution.

Table 4

Flow distribution with respect to fiber diameter

Membrane nameBackwash flow distribution % at 2 psig backwash pressure
Backwash temperature 30 °C
Backwash temperature 50 °C
Section 1Section 2Section 3Section 1Section 2Section 3
Membrane 2
di = 0.7 mm,
do = 0.9 mm 
100.00 57.29 22.00 100.00 54.16 33.31 
Membrane 3
di = 0.3 mm,
do = 0.5 mm 
100.00 23.29 11.33 100.00 24.44 15.17 
Membrane nameBackwash flow distribution % at 2 psig backwash pressure
Backwash temperature 30 °C
Backwash temperature 50 °C
Section 1Section 2Section 3Section 1Section 2Section 3
Membrane 2
di = 0.7 mm,
do = 0.9 mm 
100.00 57.29 22.00 100.00 54.16 33.31 
Membrane 3
di = 0.3 mm,
do = 0.5 mm 
100.00 23.29 11.33 100.00 24.44 15.17 

Effect of water temperature on backwash efficiency

As shown in Tables 3 and 4, improved backwash water flow distribution can be achieved by using optimal fiber dimensions and optimal physical backwash conditions. To verify the effect of backwash water temperature on corresponding backwash efficiency, the fouling and backwash experiments were performed in membrane modules (type 3) as previously described. Raw Brahmaputra river water was used as feed water for the fouling experiments. The feed water and operation conditions were as follows: turbidity – 100 NTU, TDS – 40 ppm, feed pressure – 5 psig and temperature – 30 °C. During the membrane operation, the product water turbidity was found to be less than 0.4 NTU. The backwash operation was performed at 2 psig pressure with 30 and 50 °C permeate water for 1 min. The experimentally normalized forward flow with respect to time is reported for three cycles of the operation, as shown in Figure 7.

The experimental results suggest that increasing the temperature of the backwash water resulted in an improved normalized product water flow after every backwash operation. For example, the normalized product water flow after three cycles of the operation is measured as 0.84 and 0.92 for backwash water temperatures of 30 and 50 °C, respectively. The backwash efficiency is verified by measuring the turbidity of the collected backwash water. To verify the deposit removal efficiency, relative fouling removal for different backwash temperatures is reported in Table 5. It is evident that the turbidity of collected backwash water is decreasing with increasing the backwash number, and it is due to partial removal of deposits from the fouled membrane surface. The relative deposit removal is decreasing with the backwash number but improves significantly with backwash water temperature.

Table 5

Backwash water turbidity data

Backwash temperature (°C)Turbidity NTU
Backwash 1Backwash 2Backwash 3Backwash 4
30 677 651 615 548 
50 738 733 722 656 
Backwash temperature (°C)Turbidity NTU
Backwash 1Backwash 2Backwash 3Backwash 4
30 677 651 615 548 
50 738 733 722 656 

Similarly, for greywater purification experiments, the backwash operation was carried out at 21 and 50 °C, at a constant backwash water flow rate of 60 L/h. The normalized flow rate for each of the operations is represented in Figure 8, which clearly shows the effect of backwash water temperature and backwash pressure on improved permeate flow rate. The membrane which was backwashed with cold water is again backwashed with hot water at the end of the 34th cycle to explore the effect of temperature on backwash. It can be observed that the normalized flow rate increases from 0.36 to 0.50 due to low pressure and hot water backwash. The hot water backwash efficiency can be measured by comparing the following two parameters: (1) membrane flux reduction % after 700 min of the operation in terms of normalized flux and (2) the number of the backwash operation carried in 700 min of the operation. For example, the experimentally measured normalized flux at 700 min is 0.409 for cold water at high pressure and 0.70 for hot water at the low-pressure backwash condition.

Figure 8

Normalized water flow rate with respected to fouling and backwash conditions for greywater.

Figure 8

Normalized water flow rate with respected to fouling and backwash conditions for greywater.

Similarly, the number of hot and cold water backwash performed were 25 and 34, respectively. The hot water backwash is able to improve 29.1% in terms of membrane flux reduction after 700 min of the operation. Similarly, hot water, low-pressure backwash can minimize the downtime equivalent of 9 backwash cycle time in 700 min of the operation. Therefore, for applications like the HFUF membrane operated under gravity flow, the effective membrane regeneration can be achieved by estimating optimal fiber diameter, gravity head and optimal backwash condition, considering the practical limitation such as fiber strength and durability of the membrane at high temperature.

Parameter estimation and model validation

The membrane permeability is unknown, and thus, it was estimated by calculating the backwash membrane flow distribution across the length of the modules. To calculate the backwash flow distribution along the fiber length, the analytical solution was used and was solved using MS Excel 2013. The membrane permeability is estimated by minimizing the error between experimental and model predicted backwash water flow along the membrane length.
formula
(15)
where is the cumulative backwash water flow at length L from the model prediction and is the experimental cumulative backwash water flow at length L.

The viscosity of backwash water was estimated using empirical relation reported in the literature (Green & Perry 2008). For a set of input backwash water conditions such as backwash water temperature and pressure, the A*m factors are estimated for all three membrane modules by minimizing the error and values are reported in Table 6. Membrane 3 has very high permeability, and membrane 2 has very low permeability among others.

Table 6

Estimated membrane parameter values

ParameterMembrane 1Membrane 2Membrane 3
(m) 15.3 × 10−13 7.89 × 10−13 77.6 × 10−12 
l (1/m) Actual 0.543 6.065 
df (mm) at l = 0.2 1.7 1.4 2.9 
ParameterMembrane 1Membrane 2Membrane 3
(m) 15.3 × 10−13 7.89 × 10−13 77.6 × 10−12 
l (1/m) Actual 0.543 6.065 
df (mm) at l = 0.2 1.7 1.4 2.9 

As shown in Figure 9, to achieve homogeneous flux distribution, l should be less than 0.2, but the estimated l for all the membranes are found be greater than 0.2. To achieve homogeneous flux distribution, one can increase the fiber diameter to decrease the l value. The estimated HF membrane inside diameter for achieving nearby homogeneous flux distribution is reported in Table 6. The model predicted versus experimental results for membrane module 1 are shown in Figure 8, and the model can predict well within the error limit. However, a small deviation between model and experimental data is found for few cases, which might have resulted due to the experimental error in flow measurement in the last section.

Figure 9

Experimental versus model prediction of backwash water flow rate for membrane 1.

Figure 9

Experimental versus model prediction of backwash water flow rate for membrane 1.

CONCLUSION

The effect of backwash operating conditions and its significance on membrane backwash efficiency is studied. The experimental results show that the selection of optimal backwash operating conditions, fiber diameter and membrane structural parameter is very important to achieve homogeneous backwash water flow distribution along the fiber length. The data presented in this work suggest a strong correlation between homogeneous backwash flow distributions and permeate flow rate recovery during the regular filter operation after backwash. For example, by increasing backwash water temperature from 30 to 50 °C at a constant lower backwash pressure, it is observed that a better homogeneous backwash water flow distribution along the fiber length is achieved, which resulted in 9.5% improvement in product water flow after three cycles of the membrane operation for Brahmaputra river water.

Similarly, hot water backwash can minimize the process downtime, i.e. equivalent of 9 backwash cycle time in 700 min of the operation. In addition, experimental results suggest that optimizing both backwash conditions and membrane fiber configuration may provide better backwash flow distribution. The analytical solution for membrane backwash flow distribution presented in this work is able to predict the flow distribution along the fiber length for the given input water operating conditions, fiber configurations and membrane permeability. The analytical model may be used for the optimization of HFM design and backwash parameters to achieve improved membrane performance and membrane life. In the situations where the backwash operation is performed using gravity flow force, i.e. without a pump, the backwash efficiency can be increased by performing hot water backwash at a lower pressure using low-priced waste heat or using solar energy sources.

ACKNOWLEDGEMENT

This work was supported by the Indian Institute of Technology Guwahati startup Grant through Project Number SG/CLE/P/SS/01 and Unilever Research Centre Bangalore through contract MA-2014-01505.

DATA AVAILABILITY STATEMENT

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

REFERENCES

Chatterjee
A.
Ahluwalia
A.
Senthilmurugan
S.
Gupta
S. K.
2004
Modeling of a radial flow hollow fiber module and estimation of model parameters using numerical techniques
.
J. Membr. Sci.
236
,
1
16
.
https://doi.org/10.1016/j.memsci.2004.01.006
.
Chen
J. P.
Kim
S. L.
Ting
Y. P.
2003
Optimization of membrane physical and chemical cleaning by a statistically designed approach
.
J. Membr. Sci.
219
,
27
45
.
https://doi.org/10.1016/S0376-7388(03)00174-1
.
Chew
C. M.
Aroua
M. K.
Hussain
M. A.
2018
Advanced process control for ultrafiltration membrane water treatment system
.
J. Cleaner Prod.
179
,
63
80
.
https://doi.org/10.1016/j.jclepro.2018.01.075
.
Green
D. W.
Perry
R. H.
2008
Perry's Chemical Engineers’ Handbook
, 8th edn.
McGraw-Hill
,
Delhi
.
Guadix
A.
Sørensen
E.
Papageorgiou
L. G.
Guadix
E. M.
2004
Optimal design and operation of continuous ultrafiltration plants
.
J. Membr. Sci.
235
,
131
138
.
https://doi.org/10.1016/j.memsci.2004.02.003
.
Günther
J.
Schmitz
P.
Albasi
C.
Lafforgue
C.
2010
A numerical approach to study the impact of packing density on fluid flow distribution in hollow fiber module
.
J. Membr. Sci.
348
,
277
286
.
https://doi.org/10.1016/j.memsci.2009.11.011
.
Günther
J.
Hobbs
D.
Albasi
C.
Lafforgue
C.
Cockx
A.
Schmitz
P.
2012
Modeling the effect of packing density on filtration performances in hollow fiber microfiltration module: a spatial study of cake growth
.
J. Membr. Sci.
389
,
126
136
.
https://doi.org/10.1016/j.memsci.2011.10.055
.
Hajibabania
S.
Antony
A.
Leslie
G.
Le-clech
P.
2012
Relative impact of fouling and cleaning on PVDF membrane hydraulic performances
.
Sep. Purif. Technol.
90
,
204
212
.
https://doi.org/10.1016/j.seppur.2012.03.001
.
Lin
J. C.-T.
Lee
D.-J.
Huang
C.
2010
Membrane fouling mitigation: membrane cleaning
.
Sep. Sci. Technol.
45
,
858
872
.
https://doi.org/10.1080/01496391003666940
.
Muthukumaran
S.
Kentish
S.
Lalchandani
S.
Ashokkumar
M.
Mawson
R.
Stevens
G. W.
Grieser
F.
2005
The optimisation of ultrasonic cleaning procedures for dairy fouled ultrafiltration membranes
.
Ultrason. Sonochem.
12
,
29
35
.
https://doi.org/10.1016/j.ultsonch.2004.05.007
.
Nigel
P.
Souza
D.
Basu
O. D.
2013
Comparative analysis of physical cleaning operations for fouling control of hollow fiber membranes in drinking water treatment
.
J. Membr. Sci.
436
,
28
35
.
https://doi.org/10.1016/j.memsci.2013.02.014
.
Peiris
R. H.
Budman
H.
Moresoli
C.
Legge
R. L.
2013
Fouling control and optimization of a drinking water membrane filtration process with real-time model parameter adaptation using fluorescence and permeate flux measurements
.
J. Process Control
23
,
70
77
.
https://doi.org/10.1016/j.jprocont.2012.10.001
.
Porcelli
N.
Judd
S.
2010
Chemical cleaning of potable water membranes: a review
.
Sep. Purif. Technol.
71
,
137
143
.
https://doi.org/10.1016/j.seppur.2009.12.007
.
Regula
C.
Carretier
E.
Wyart
Y.
Gésan-Guiziou
G.
Vincent
A.
Boudot
D.
Moulin
P.
2014
Chemical cleaning/disinfection and ageing of organic UF membranes: a review
.
Water Res.
6
.
https://doi.org/10.1016/j.watres.2014.02.050
.
Shi
X.
Tal
G.
Hankins
N. P.
Gitis
V.
2014
Fouling and cleaning of ultrafiltration membranes: a review
.
J. Water Process Eng.
1
,
121
138
.
https://doi.org/10.1016/j.jwpe.2014.04.003
.
Tarazaga
C. C.
Campderrós
M. E.
Padilla
A. P.
2006
Physical cleaning by means of electric field in the ultrafiltration of a biological solution
.
J. Membr. Sci.
278
,
219
224
.
https://doi.org/10.1016/j.memsci.2005.11.004
.
Venkatesh
T.
Senthilmurugan
S.
2016
Grey water treatment and simultaneous surfactant recovery using UF and RO process
.
Sep. Sci. Technol.
1
12
.
https://doi.org/10.1080/01496395.2016.1273244
.
Wang
J.
Cui
Z.
Jia
H.
Zhang
H.
2014
The effect of fiber length on non-uniform and hysteresis phenomenon in hollow fiber membrane back flushing
.
Desalination
337
,
98
108
.
https://doi.org/10.1016/j.desal.2013.11.016
.
Xia
S.
Nan
J.
Liu
R.
Li
G.
2004
Study of drinking water treatment by ultrafiltration of surface water and its application to China
.
Desalination
170
,
41
47
.
https://doi.org/10.1016/j.desal.2004.03.014
.
Yoon
S.-H.
Kim
H.
Yeom
I.
2004
Optimization model of submerged hollow fiber membrane modules
.
J. Membr. Sci.
234
,
147
156
.
https://doi.org/10.1016/j.memsci.2004.01.018
.

Supplementary data