Distribution and deposition of organic fouling on the microfiltration membrane evaluated by high-frequency ultrasound

Abstract

A 50 MHz high-frequency ultrasound and analysis method were developed to further improve the in situ assessment of deposition and distribution of organic fouling on the polyvinylidene fluoride (PVDF) membranes. Measurements of fouling depositions were performed from PVDF membranes filtrated with aqueous humic acid solutions (HAS) of 2 and 4 ppm concentrations in a flat-sheet module. Ultrasound signals reflected from the PVDF membranes, following filtrations at various durations including 0, 5, 15, 30, 60, and 100 min, were acquired. The thickness and distribution of fouling estimated and assessed by peak-to-peak echo voltage (V_pp) and C-mode images were found to be non-homogeneously deposited on the membranes. Following the filtrations with 2 and 4 ppm HAS for 100 min, the corresponding thickness of fouling deposition increased from 1.81±9 to 2.4571.57 mm, respectively; those average V_pp decreased from 2.05±07 to 1.13±16 V and from 2.11±08 to 0.94±15 V. These results demonstrated that the deposition and distribution of organic fouling could be sensitively and rapidly evaluated by high-frequency ultrasound image incorporated with the analysis method.

1. Introduction

Microfiltration (MF) membrane is frequently applied to remove particles or microorganisms from fluid suspensions in various industrial applications, such as natural and waste water processing, food processing, paper mill, and biotechnology [1–3]. The performance of filtration, in terms of fluid separations and usage lifetime, depends greatly on the degree of membrane fouling deposition [4]. Specifically, the filtration flux is usually declining associated with such complications as the blockage or plugging of pores within the membrane as well as the concentration polarization and cake layer formation on the membrane surface [2,5,6]. In general, the formation of fouling is resultant from a serial of complex process corresponding to several factors, including the operating pressure, flow velocity, temperature, concentration and pH value of the feed solution, material and structure of filtration membrane, and characteristics of particle suspensions (such as shape, size, surface charge, compressibility and chemical composition) [7,8]. It thus is essential to in situ assess the formation and distribution of fouling deposition on the filtration membranes and to further comprehend the mechanism of fouling for extensive control applications.

The deposition and mechanism of fouling formation were typically investigated by filtration flux and morphological pictures taken from scanning electron microscopy (SEM) [9]. The flux-decline information was acquired and that however may not sufficient to be applied to thoroughly elucidate the mechanism of fouling formation associated with the simultaneous occurrence of concentration polarization, membrane compaction and degradation [9–11]. SEM picture certainly provides a substantial high resolution, which is capable of directly analyzing and characterizing the fouling deposition on the membrane. Nevertheless, in addition to being not cost effective, the implementation of SEM requires a relative time-consuming and complex procedure to prepare the surface of specimen for being electrically conductive. Moreover, measurement with SEM is not only invasive to the testing sample, but also incapable of performing long-term and continuous observations on the progress of fouling formation. Therefore, it is important to develop non-destructive techniques to sensitively measure the fouling formation during filtration process in real-time and to fast characterize the properties and mechanism about the growth of fouling layer associated with the flux decline [7].

In addition to other modalities being utilized, ultrasound technique is commonly applied to evaluate materials nondestructively. As a mechanical wave, the propagation of ultrasound in the medium may be characterized by such parameters as pressure, medium velocity, particle displacement, density, and temperature [12]. When the propagating ultrasound wave encounters with an interface between two media of different acoustic impedances (defined as the product of density and sound velocity), part of the incident energy will be reflected and refracted [13]. With several advantages, such as real-time imaging, non-ionizing radiation, and low cost, ultrasonic technique has been frequently applied to clinical diagnoses, nondestructive testing of materials, and under water applications. An interesting example is to utilize ultrasonic time-domain reflectometry (UTDR) to measure the deposition of calcium sulfate on the commercial reverse osmosis (RO) membrane in a flat-sheet filtration module [7,14]. It achieved a remarkable correlation between the decline of ultrasonic amplitude and the progress of fouling layer deposition, which also validated the capability of UTDR for noninvasively monitoring the membrane fouling in real-time. UTDR was further implemented to detect the growth of fouling on MF of an effluent liquid of paper mill [4]. A new echo indicating the feed/fouling interface was found as that of the fouling layer was thick enough. Subsequently, the thickness of fouling layer was measured by postmortem SEM analysis and UTDR. Nevertheless, the echoes reflected from both interfaces of feed/fouling and fouling/membrane remain difficult to be resolved as the thickness of fouling is smaller than the axial resolution of employed ultrasonic transducer. Another technique, called differential signal method, was proposed for improving the sensitivity of fouling thickness detection, which was implemented by comparing the ultrasonic signals reflected from a reference material and testing object [15]. Typically, ultrasonic signal reflected from a membrane filtrated with pure water was employed as the reference. The differential signal method has been applied to estimate the fouling layer formation, growth and cake layer compressibility during cross-flow MF, and ultrafiltration of paper mill effluents [2,15]. The differential signal method was validated to be capable of detecting not only the state and progress of the fouling layer, but also the non-uniform distribution of cake layer covering on the membrane surface. A number of applications using UTDR, including measurement of deposited bovine serum albumin properties corresponding to different pH values [16], calcium sulfate deposition on RO membranes under the cross-flow and dead-end operations [17], and the effect of electromagnetic field on calcium carbonate scale deposition on the membrane surface during cross-flow nanofiltration [18], were also explored.

In addition to being able to measure the fouling layer in a flat-sheet membrane module, the UTDR has also been adopted to successfully monitor the fouling of bovine serum albumin in the tubular ultrafiltration module [19], calcium sulfate deposition on RO membrane in the spiral-wound module [20,21], and deposited kaolin fouling in the hollow fiber membrane module [22]. Moreover, the UTDR was employed to evaluate the cleaning efficiency from the membrane corresponding to various cleaning techniques, including forward flushing, reverse flushing, ultrasonic cleaning, and the combination of ultrasound and forward flushing [9,23–25]. Another study to combine ultrasound and forward flushing demonstrated to have a better efficiency in cleaning membrane than other approaches [24].

UTDR has been verified to be able to non-invasively and nondestructively measure the fouling growth during filtration process in real-time. However, most of previous studies measured samples only at a certain position utilizing ultrasound with the frequency less than 10 MHz. These pose complications not only on the insufficient ultrasonic resolution to sensitively differentiate the fouling of a fine thickness, but also obtaining limited information regarding to the spatial distribution of the membrane. Certainly, the detectability of fouling deposition with ultrasonic technique may be readily improved by increasing the implementation of ultrasonic frequency and focusing technique. In addition, C-mode scanning is able to discern the spatial information of fouling deposition associated with a constant depth or surface of a sample [26,27]. In the present study, a high-frequency ultrasound system with a 50 MHz ultrasound transducer and f-number of 1.5 capable of achieving the axial and lateral resolution to be better than 50 and 100 μm, respectively, was developed for detecting the fouling layer on the MF membrane in a flat-sheet module. Experiments were arranged and carried out to validate the system performance and investigate the distribution of fouling deposition in response to various filtration conditions. The high-frequency ultrasound C-mode images and reflected signals were acquired and analyzed for further assessment.

2. Materials and methods

2.1. Microfiltration system

A flat-sheet MF module, shown in Fig. 1(a), consisted of two acrylic plates in the top and bottom sides, a silicon rubber plate, a porous stainless-steel plate, and a MF membrane was arranged for the study. Both top and bottom acrylic plates, with dimensions of 220 × 80 × 30 mm (length × width × thickness), are transparent that allow the variations inside the module to be readily observed. The silicon rubber plate was made for preventing fluid leakage and formation of cross-flow channel, with dimension of 100 × 9 × 5 mm³ (length × width × thickness), between the acrylic plates [28]. The polyvinylidene fluoride MF membrane (GVWP14250, Millipore Co., MA, USA), with nominal pore size of 0.22 μm and 70% porosity in conjunction with the maximum usage of water flow rate of 6.7 ml/min cm², was fixed on the porous stainless-steel plate. The effective filtration area of MF membrane is 40 × 9 mm (length × width). The experimental system of MF, as the schematic diagram shown in Fig. 1(b), was designed to mainly consist of a feeding stream and a retentating and permeating stream to allow the filtration fluid to flow in or out of the filtration module. The circulation of filtration fluid was driven by a peristaltic pump (Model 7553-02, Masterflex, Cole-Parmer, IL, USA). The flow rate and operation pressure were controlled by the rotation speed of peristaltic pump and retentate valve. The flow feeding into the MF module was measured by a flow meter (Model S-111, McMillan, TX, USA). A manometer connected between the inlet and outlet of the flow channel was utilized to measure operation pressure. The weight of permeated fluid for calculating the filtration flux was measured by an electronic balance (XS4250C, Precisa Gravimetrics AG, Dietikon, Switzerland).

Fig. 1.

(a) Diagram of the constructed flat-sheet MF module. (b) Schematic representation of filtration system.

2.2. High-frequency ultrasound system

A schematic diagram detailing the arrangement of high-frequency ultrasound system is shown in Fig. 2. A 50 MHz single-element ultrasound transducer (NIH Ultrasonic Transducer Resource Center, USC, LA, USA) was employed for the generation and reception of ultrasound waves. The transducer characteristics are summarized in Table 1. The transducer was driven by a pulser/receiver (Model 5900PR, Panametrics, Waltham, MA, USA) and coupled with an expander (Matec Instruments Company, MA, USA) to eliminate electrical noise. The received ultrasonic signals were amplified by a low-noise amplifier (Model LN1000A, Amplifier Research, PA, USA) via a connection with an electronic limiter (Matec Instruments Company, MA, USA) for protecting the device from damage. Subsequently, radio frequency (RF) ultrasonic signals were filtered by a bandpass filter (Model BIF-50, Mini-Circuits, NY, USA) and then digitized by an 8-bit analog-to-digital converter (PXI 5152, National Instruments, TX, USA) at 500 MHz sampling frequency. The transducer mounted on the piezoceramic motor (HR8, Nanomotion Ltd., Yokneam, Israel) was swept scanned to different locations of each sample where corresponding ultrasonic signals were acquired. In addition, the transducer was also allowed to flexibly move in two directions controlled by two axes of stepping motors (CM1-C-17L30A, Cool Muscle, Osaka, Japan) and actuators (KR2602A, THK, Tokyo, Japan). All of these motor stages were controlled by a motor controller (DMC-1842, Galil Motion Control Inc., California, USA). The program to acquire data and control motor was developed using LabVIEW software (National Instruments, TX, USA). The developed high-frequency ultrasound system is able to perform A-mode, B-mode and C-mode imaging. A-mode imaging displays the amplitude of ultrasound echoes as a function of time corresponding to a certain propagation depth. The B-mode imaging was realized by a serial of signal processing procedures, including filtering, Hilbert transform, and logarithmic compression, to detect envelopes of a sequence of ultrasonic echoes acquired from different positions of a sample. Subsequently, the magnitude of signals was converted into brightness for gray scale display [13]. The brightness of C-mode image also corresponded to amplitude of the acquired ultrasonic signals from an interrogated sample of a constant depth. A certain area of C-mode image was typically achieved by moving the ultrasonic transducer with raster scanning, indicated in Fig. 3.

Fig. 2.

Schematic diagram of the high-frequency ultrasound system.

Fig. 3.

Procedure of C-mode scanning with high-frequency ultrasound imaging system.

2.3. Calibration of ultrasonic transducer resolution

The resolution of high-frequency ultrasound transducer was calibrated by signals reflected from a 10-mm-diameter tungsten wire. Calibration measurements were carried out in the degassed and deionized water tank. The distance between the tungsten wire and transducer was arranged to be 6 mm corresponding to transducer's focal length. The acquired B-mode image in Fig. 4(a) is with dimension of 1 × 0.6 mm² and is composed of 1000A-lines with 1 mm interval between each adjacent scanning line. The -6 dB axial and lateral resolutions of high-frequency ultrasonic transducer, based on point spread function [29] in Fig. 4(b) and (c), were estimated to be 24 and 48 μm, respectively.

Fig. 4.

(a) Ultrasonic B-mode image of a wire phantom. (b) Axial point-spread function and (c) lateral point-spread function of the 50 MHz ultrasonic transducer.

2.4. Experimental procedures and data analysis

The aqueous humic acid solution (HAS) for filtration experiments were prepared by solving the humic acid powder (53680, Sigma-Aldrich, Mo, USA) in phosphate buffer saline. The HAS of 2 and 4 ppm concentrations, and pH value of 7 was prepared. Prior to the beginning of each experiment, pure water was circulated, with an average shear rate of 97.8 s^-1, in the MF system for 30 min to stabilize the membrane surface and to remove possible air trapped in the MF membrane. This procedure is to avoid the interference of air bubbles in the received ultrasound waves. Subsequently, the HAS was circulated, while the permeated valve C in Fig. 1 was turned off, with the flow rate of 0.22 l/min under 1.2 bar pressure and that of the temperature was maintained at 25 °C by a hotplate/magnetic stirrer. The permeated valve was then turned on as the flow of feed solution in the filtering channel became stable. Subsequently, the filtration process was lasted for 100 min, where the fluid in the feed tank was stirred continuously. The filtered fluid was weighed by a computer-controlled electronic balance and that of permeated flux as a function of time was measured. The top acrylic plate and silicon rubber plate were removed carefully in the end of each filtration at durations of 0, 5, 15, 30, 60, and 100 min. Subsequently, C-mode imaging of each sample covering an area of 6 × 8 mm filtration membrane was measured, where the distance between MF membrane and transducer was maintained at 6 mm. A total of 400 B-mode images, per each image was composed of 300 scan lines, were acquired.

The acquired peak-to-peak voltage (V_pp) corresponding to the intensity of ultrasonic signals [30] can be calculated by:

$$V_{\mathrm{pp}}=V_{\max }-V_{\min }$$ (1)

where V_max and V_min denote, respectively the maximum and minimum amplitudes of echo signals. During the formation of fouling, the V_pp of membrane surface tends to decrease accordingly due to fouling deposition would increase and decrease, respectively the ultrasound attenuation and the difference of acoustic impedance between the fouling and membrane. The V_pp was sampled from the gated signals ranged from 4.05 to 4.11 μs. Subsequently, the thickness of fouling layer, DS, can be determined by

$$\Delta S=\frac{1}{2}c(t_2-t_1)$$ (2)

where c represents the sound velocity of fouling; t₁ and t₂ denote the propagation time traveling from water/membrane interface to membrane/porous stainless-steel plate interface and that of from feed solution/fouling interface to membrane/porous stainless-steel plate interface, respectively. The detection of propagation time between water/membrane and feed solution/fouling interfaces were achieved by measuring the time of the first cycle of echo signals which is larger than the assigned threshold. The sediments of 0.2 g/ml HAS to simulate the fouling deposition were formed by a centrifuge operated at 3000 rpm. The HAS sediments were then filled into a 5 × 10 × 15 mm³ (length width high) acrylic container, of which the sound velocity for estimating the thickness of fouling during MF filtration was subsequently measured as 1554±2.3 m/s. The algorithms to estimate both V_pp and ΔS for C-mode image reconstruction were implemented by Matlab software (The MathWorks, MA, USA). The fouled membrane was fractured in liquid nitrogen and treated with Au sputtering after ultrasound measurements. Subsequently, the thickness of fouling was estimated by SEM (S-3000, Hitachi High-Technologies Co., Japan).

3. Results and discussion

The permeation flux as a function of filtration time, corresponding to the circulation of HASs with 2 and 4 ppm concentrations, is given in Fig. 5. The variations of permeation flux in general decreased exponentially with the increase of filtration duration. Specifically, the permeation flux tended to decline faster in the feed solution of 4 ppm HASs than that of 2 ppm. The rate of permeation flux decline varied rapidly in the beginning of filtration process and then became slower after 30 min of filtration. The instantaneous flux decline was found from previous reports [4,14,31] to be associated primarily with several factors, including the concentration polarization, formation of fouling layer, and pore plugging. The main factor to slow down the rate of flux decline is presumably due to the slow growth of fouling layer [4,14,31]. In the present study, some apparent increases of permeation fluxes, indicated with arrow symbols in Fig. 5, were found to occur at 15, 30 and 60 min during filtration experiments. These variations were due to the destabilization of feed flow in the MF module and the relaxation of fouling layer corresponding to the interruptions of filtration process when the fouling layer were measured by high-frequency ultrasound system. Consequently, the flow rate and shear rate of the filtration would change in accompanying with the re-start of the filtration experiment, which led to the decreases of fouling layer thickness, partial dissolution of the fouling layer, and fouling density [4]. Furthermore, the immersion of fouling in the water tank during high-frequency ultrasonic measurement might also vary with the thickness and density of fouling layer. All of these factors account for a significant increase of the permeation flux associated with the formation of fouling layer on the membrane surface.

Fig. 5.

Permeation flux, with HAS of 2 and 4 ppm concentrations, as a function of filtration time. The arrow symbols indicate the moments that significant increases of permeation fluxes were occurred.

The high-frequency ultrasonic signals of MF membranes, which were filtrated with 4 ppm HAS at different fouling durations, are shown in Fig. 6, in which signals A and B indicated in Fig. 6(a) correspond, respectively to the interfaces of water/MF membrane and MF membrane/porous stainless-steel plate before the filtration with HAS. Following the filtration procedure for 5 min, the maximum amplitude of reflected signal A decreased from 0.97 to 0.84 V (with respect to that of A⁰), as shown in Fig. 6(b). The maximum amplitude of reflected signal A⁰ decreased further to 0.45 V after the HAS filtration for 100 min, shown in Fig. 6(f). Yet, it is still difficult to distinguish the fouling from the acquired echo signals due to the thickness of fouling in this study was smaller than the axial resolution of employed ultrasonic transducer. Moreover, the signals reflected from the MF membrane decreased progressively during the formation of fouling, which is owing to the increase of ultrasound attenuation, decrease of the difference of acoustic impedance between the fouling and MF, and the roughness of fouling layer surface. Furthermore, a new echo corresponding to the interface between feed solution and fouling will be observed as the fouling layer is thick enough to be resolved [25]. In the present study, the new ultrasonic echo reflected from the feed solution/fouling interface was occurred after 100 min of feed filtration, which is indicated by an arrow in Fig. 6(f). Generally, the differential signal method could be utilized to enhance the fouling echoes for further analysis as the signal reflected from the feed solution/fouling interface is weak during the early phase of feed filtration. Nevertheless, this method is difficult to be realized in the present study due to the occurrence of slight undulations and inconsistency of MF membrane may complicate the reference signals acquired from the clean membrane to be adopted. B-mode images acquired from different positions of MF membranes at different filtration durations are shown in Fig. 7, in which the first interface corresponds to the surface of MF membranes. The distance between the first and second interfaces was estimated to be 0.12 mm, which is consistent with the thickness of MF membrane, as shown in Fig. 7(a). Corresponding to the formation of fouling, the image brightness of membrane surface was declining accordingly, as shown in Fig. 7(b)–(d). Moreover, the image of clean membrane surface tended to be more uniform than that of with fouling deposition, in which the membrane surface was getting rough especially after 30 min of feed filtration.

Fig. 6.

High-frequency ultrasonic signals of MF membrane correspond to the filtration processes with 4 ppm HAS for 0, 5, 15, 30, 60, and 100 min. The reflected echoes, A, B, and A′, associate with the interfaces of water/MF membrane, MF membrane/porous stainless-steel plate, and water/fouling layer combined MF membrane. The arrow symbol in (f) depicts an ultrasonic echo reflected from the new water/fouling interface.

Fig. 7.

Partial high-frequency ultrasonic B-mode images of MF membrane after (a) 0, (b) 5, (c) 15, (d) 30, (e) 60, and (f) 100 min of filtration with 4 ppm HAS.

To further analyze the distribution of fouling layer, ultrasonic C-mode images of MF surfaces, which were filtrated with 2 and 4 ppm HASs at durations from 0 to 100 min, were acquired, as shown in Figs. 8 and 9. These images clearly disclose information about that the fouling deposition and density were not homogeneously distributed on the membranes in response to filtrations of HAS at different concentrations and durations. After 5 min of feed filtration, mesh-type distributions of C-mode images, as shown in Figs. 8(b) and 9(b), were displayed which agree well with the patterns of porous stainless-steel plate. These images also illustrate that the fouling could easily deposit at the pore positions corresponding to the porous stainless-steel plate beneath the membrane. Images of Figs. 8(b)–(d) and 9(b)–(d) are two apparent stripe shapes of fouling depositions, indicated by arrow signs at the locations around 4.5 and 6.5 mm along the flow direction, that may be resultant from the turbulent feed flow due to the position of feed channel is higher than that of the flow channel in the MF module. The V_pp of membrane echo as a function of filtration duration is shown in Fig. 10, in which the average V_pp tends to linearly decrease with the increase of filtration duration. Specifically, the V_pp associated with the HASs of 2 and 4 ppm for 100 min filtration tended to decrease from 2.05±07 V (mean7standard deviation) to 1.13±16 V and from 2.11±08 to 0.94±15 V, respectively. The decline rate of average V_pp acquired from the feed filtration with 4 ppm HAS tended to be faster than that of with 2 ppm HAS. Furthermore, due to concentration polarization and fouling layer formation [4,9], the average V_pp decreased rapidly at 5 min after feed filtration, which agreed to the decline tendency of permeation flux given in Fig. 5. In addition, the average V_pp was higher for the filtration with 4 ppm HAS for 60 min than that for 30 min, which reflected to the variation of permeation flux.

Fig. 8.

C-mode images depicting the V_pp of membrane echo after (a) 0, (b) 5, (c) 15, (d) 30, (e) 60, and (f) 100 min of filtration with 2 ppm HAS. The arrow symbols indicate the stripe shapes where the fouling is not homogeneous distributed significantly.

Fig. 9.

C-mode images depicting the V_pp of membrane echo after (a) 0, (b) 5, (c) 15, (d) 30, (e) 60, and (f) 100 min of filtration with 4 ppm HAS. The arrow symbols indicate the stripe shapes where the fouling is not homogeneous distributed significantly.

Fig. 10.

Average V_pp of membrane echo under various filtration durations.

The sequences of C-mode images, as given in Figs. 11 and 12, detail the fouling thickness corresponding to the filtration with 2 and 4 ppm HASs at durations from 5 to 100 min, respectively. Due to the occurrence of slight undulations and discrepancy of MF membrane, it is not practical to estimate the fouling thickness directly using changes of the arrival time from echoes of the fouling layer. An alternative may be applied to calculate fouling thickness using Eq. (2). The fouling thickness was not distributed homogeneously on the membrane, and that the V_pp and roughness of membrane tended to, respectively decrease and increase with the increase of filtration durations. The average fouling thickness of membrane as a function of filtration duration is shown in Fig. 13, in which an exponential relationship could be empirically achieved. After the filtrations with 2 and 4 ppm HASs for 100 min, the average fouling thicknesses estimated by ultra-sound images were 1.81±9 and 2.4571.57 mm, respectively; while those of measured by cross-sectional SEM images, as shown in Fig. 14, were 0.898±053 and 1.62±27 mm, respectively. The thickness of the fouling layer estimated by high-frequency ultrasound system seems to be larger than that of by SEM due to partially that the sample was prepared in dry condition before taking SEM picture [4]. Nevertheless, the rate of average fouling thickness during feed filtration with 4 ppm HAS always increased faster than that of with 2 ppm HAS. The rapid increase of average fouling thickness at the 5 min moment after feed filtration corresponds well with that of the decrease of average V_pp.

Fig. 11.

C-mode images depicting the thickness of fouling that the filtration with 2 ppm HAS after the durations of (a) 5, (b) 15, (c) 30, (d) 60, and (e) 100 min.

Fig. 12.

C-mode images depicting the thickness of fouling that the filtration with 4 ppm HAS after the MF duration of (a) 5, (b) 15, (c) 30, (d) 60, and (e) 100 min.

Fig. 13.

Average thickness of fouling under various filtration durations.

Fig. 14.

Cross-sectional SEM images of MF membrane after filtration process with (a) 2 ppm and (b) 4 ppm HAS for 100 min.

Previous study [14] had indicated the UTDR technique remains difficult to be implemented to monitor the organic and biological fouling due to low impedance contrast between these two substances. The humic acid is an organic substance, that is with dark brown color and is soluble in the alkaline solution [32] leading to fouling problems in water filtration. Other studies have shown that membrane fouling containing humic substances may be affected by the characteristics of humic substance, hydrodynamic conditions, chemical composition of the feed solution, and properties of membranes [31–33]. With high resolution capability, high-frequency ultrasound system in the present study was validated to be able to sensitively measure the fouling distribution on the membrane after filtration with humic acid.

Nevertheless, the depth penetration of high-frequency ultrasound is still limited to 1 cm due to the accompanying attenuation is tremendously increased with the increase of ultrasonic frequency. This penetration limitation would lead the filtration process to be interrupted for detecting the formation of fouling within current flat-sheet MF module using high-frequency ultrasound system. This hurdle conceptually could be resolved using low-frequency ultrasound (< 10 MHz) for the measurement. Yet, it consequently decreases the sensitivity for detecting the finer fouling. Therefore, it is essential to further design a proper filtration module for alleviating the effect of attenuation associated with the use of high-frequency ultrasound.

4. Conclusion

In this study, a 50 MHz high-frequency ultrasound system was utilized to measure the fouling formation and distribution on MF membrane. The distribution of fouling formed from HAS was assessed using ultrasonic C-mode images incorporating with quantitative V_pp of membrane echo and fouling thickness. The permeation flux was found to decrease exponentially with the increase of filtration durations, in which the decline of permeation flux generally was faster for the filtration of MF module with HAS of 4 ppm feed concentration than that of 2 ppm feed concentration. The flux decline is rapid at the beginning phase of filtration process and then is slower after the filtration for 30 min. Results of C-mode images indicated that the distribution of fouling is not homogeneous and that is presumably due to the influence corresponding to turbulent feed flow and porous stainless-steel plate. After 100 min filtration with 2 and 4 ppm HAS, the average V_pp decreased from 2.05±07 to 1.13±16 V and from 2.11±08 to 0.94±15 V, respectively, and those of corresponding average fouling thickness were 1.81±9 and 2.4571.57 mm. Both the decline rate of average V_pp and increase rate of fouling thickness are faster for the feed filtration with 4 ppm HAS than that of with 2 ppm HAS. Fouling deposition was found to be a process of both temporal- and spatial-dependence, and that may be feasible to be sensitively and rapidly evaluated by high-frequency ultrasound image incorporated with the analysis method. It certainly is essential to improve the high-frequency ultrasound system by alleviating the complication of attenuation and to design a proper filtration module for allowing current technique to be further applied in real-time and in-situ analysis of fouling behavior.