A New Approach for Detecting the Onset of Fouling during

Aug 6, 2005 - Spectra: A New Approach for. Detecting the Onset of Fouling during Microfiltration of Paper Mill. Effluent. RON D. SANDERSON,* JIANXIN L...
0 downloads 0 Views 558KB Size
Download PDF
Environ. Sci. Technol. 2005, 39, 7299-7305

Fourier Wavelets from Ultrasonic Spectra: A New Approach for Detecting the Onset of Fouling during Microfiltration of Paper Mill Effluent RON D. SANDERSON,* JIANXIN LI,† DIETER K. HALLBAUER, AND SWAPAN K. SIKDER UNESCO Associated Center for Macromolecules & Materials, Department of Chemistry and Polymer Science, University of Stellenbosch, P. Bag X1, Matieland 7602, South Africa

Nondestructive (NDT) and noninvasive ultrasonic techniques have long been used to evaluate the properties and especially the thickness of thin layers. Here, we use this technique adding a new approach to investigate microfiltration of paper mill wastewater, which gives an unexpected sensitivity in the detection of membrane fouling. In situ ultrasonic reflections data can indicate an early fouling deposition at about 30 s after filtration starts, evident by an initial decline in permeate flux. By producing differential signals, obtained by comparing reference and test waveforms, the fouling process can be detected and monitored. A linear relationship between fouling resistance and the amplitude of differential signals exists. In the case of fouling layer thickness, the resolution exceeds the theoretical limit of h/λ g 0.25, where h is the layer thickness and λ is the wavelength. When using differential signals, excellent results for thickness measurements were obtained, down to h/λ ) 0.04. Measurements on wavelet transforms support the findings and add quantitative information on other physical properties such as density and porosity of fouling layers and the fouling process. Measurement of early fouling allows (automated) remedial methods to be applied so as to maintain a high flux and therefore improve the filtration process.

Introduction From desalination to the clarification of beverages and cleaning of industrial effluents, membrane filtration is considered to be one of the best separation options. However, membrane fouling, that is, the blocking of the porous filters, is universally accepted as one of the most critical problems limiting the wider application of membranes in liquid separations. The occurrence of fouling is usually inferred from a marked decrease in permeate flux and quality. Yet these can change, because of factors other than fouling. Consequently, the development of an independent method that is sensitive only to changes in the condition of the membrane surface is of considerable interest. Direct analysis of the fouling deposits on membranes has been carried out * Corresponding author phone: +27(21) 8083172; fax: +27(21) 8084967; e-mail: [email protected]. † Present address: School of Material Science & Chemcal Engineering, Tianjin Polytechnic University, Tianjin, 300160, P.R. China. 10.1021/es050414y CCC: $30.25 Published on Web 08/06/2005

 2005 American Chemical Society

using scanning electron microscopy (SEM), X-ray diffraction (XRD), and X-ray fluorescence (XRF) (1, 2). Although these invasive methods can supply some information on the fouling mechanism, they provide little information regarding the build-up of a fouling layer on a membrane surface. The development and utilization of a suitable nondestructive and noninvasive technique for the prediction and on-line monitoring of fouling in industrial and laboratory membrane applications should enable the effectiveness of fouling remediation and cleaning strategies to be improved and quantified. A number of noninvasive observation techniques, using microelectrode measurements (3), laser interferometry (4), light reflectometry (5), radioisotopes (6), micro-array of semiconductor photosets (6), optical methods (7), photography (8, 9), and NMR micro-imaging techniques (10), have been used to measure fouling. However, these measurements are limited by the spatial resolution and sensitivity of the various techniques. Ultrasonic techniques have been shown to be an accurate and a potentially useful tool for nondestructive testing or imaging (11-13). With the development of a high-frequency digital and computer technique, ultrasonic methods have been successfully applied to the monitoring of cake thickness during a ceramics slip-casting process (14), chemical reactions (15), film formation (16, 17), glue processes or crystallization in polymers (18), and the determination of coating parameters such as elastic modulus, density, signal attenuation, and thickness (19, 20). Ultrasonic time-domain reflectometry (UTDR) is a more recent, versatile, nondestructive, and noninvasive technique that has been used for insitu monitoring of the growth of a fouling layer (21-26). A direct correlation between the change in ultrasonic signal amplitude and fouling layer initiation has been observed with this technique. The thickness of fouling layer and densities for various foulants could be measured without the use of Fourier transforms or clear evidence of its fouling peak in most cases. However, further investigation of its sensitivity and resolution for detecting thin nondistinguishable fouling layers on membranes is of considerable interest. The paper mill effluent gives a very difficult fouling layer to measure as its density matches that of the membrane used for the separation. Other systems, especially those containing higher density inorganic materials, are more easily measured and interpreted. (Additional literature on ultrasonic measurements of thin layers is provided in the Supporting Information.) In summary, therefore, the present paper addresses the sensitivity and resolution of ultrasonic testing of a thin fouling layer on membranes, in particular to fouling caused by an industrial effluent from the processing of paper pulp. A pulseecho technique was used in this study to satisfy the physical conditions of testing. Ultrasonic frequency spectra, simplified modeling, and signal processing by wavelet transform were applied to demonstrate the process of fouling layer growth. It is normally accepted that the theoretical resolution is about λ/4. This can be exceeded and improved by about 1 magnitude (radar resolution technique) (27).

Experimental Section Apparatus. In our UTDR experiments, a Biodyne A (nylon) membrane (supplied by Pall Corp., Pensacola, FL) with a nominal pore size of 0.2 µm was placed into a rectangular flat-bed test cell made of polymethyl methacrylate (perspex). The inside of the module with the exposed membrane was 100 mm long, 20 mm wide, and 3 mm high. The ultrasonic measurement system consisted of a 7.5 MHz ultrasonic VOL. 39, NO. 18, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

7299

FIGURE 1. Ultrasonic signal responses from 0 to 60 min of fouling operation with paper mill effluent. The straight line indicates the approximate position of the top surface of the porous membrane layer, whereas the stronger peak at 0 min signifies the fibrous support in the porous membrane. (Curves shifted on amplitude axis for easy viewing; the original superimposed graph is given in Supporting Information Figure S4). transducer (Panametrics V120), a pulser-receiver (Panametrics 5058PR), and a digital oscilloscope (HP model 54602B) with sweep speeds from 5 s/div to 1 ns/div and 1 mv/div sensitivity. The transducer was externally mounted in contact with the top plate by using food-grade honey as the coupling agent. This coupled to the fact that the frequency used does not disturb the fouling layer ensures that this technique is noninvasive. To clean the fouling layer, new frequencies and higher amplitude are required (24). A peristaltic pump was used to feed the cell at a constant flow. Feed Effluent. The effluent from a wastewater treatment plant of Mondi Kraft Mills, Mondi Ltd., South Africa, was selected as the feed solution. (The characteristics of the effluent and schematic of experimental setup are given in the Supporting Information, Table T1 and Figure S1, respectively.) Procedure. Cross-flow microfiltration was carried out at a cross-flow velocity of 32 cm/s (Reynolds’ number 1163) across the membrane, applied pressure of 100 kPa, and a temperature of 20 °C. The permeate flux was measured by timing and weighing on an electric balance. The run was stopped when the flux dropped down to a value too low to measure (60-80 min). Scanning electron microscopy (SEM) analysis of the fouled membrane was performed by using a LEO S440 SEM machine, and the samples were viewed at 10 kV and 20 mm working distance. Atomic force microscopy (AFM) analysis was done by using a TopoMetrix 2000 Explorer, where the measurement was done in contact mode. Data Acquisition and Processing. The pulser-receiver generated the required voltage signal to trigger the transducer to send out an ultrasonic wave. The oscilloscope captured and displayed the signal. Each signal contained 500 data points. A computer was connected to the oscilloscope to store the signal data as required. Ultrasonic signal changes were recorded at ever-increasing intervals, depending on the fouling layer growth. The recorded data were analyzed by using MS Excel and AGUVallen Wavelet (Vallen-Systeme GmbH, Germany) software. (The principle and theory of ultrasonic measurement and 7300

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 18, 2005

FIGURE 2. Flux decline with operation time in a paper mill fluid fouling experiment. The flow rate is 32 cm/s, and the applied pressure is 100 kPa. wavelet transform are provided in the Theoretical section of the Supporting Information.)

Results and Discussion Ultrasonic Reflection Signals. Actual time-domain ultrasonic responses (or waveforms) are shown in Figure 1, in relation to an increased fouling. It should be noted that membranes are usually composed of several layers, including one or two support layers apart from the actual porous filter material. In the case reported here, the nylon membrane consisted of an upper layer of porous nylon (80 µm thick), a central support layer (40 µm) of solid nylon fibers, and a lower porous layer (40 µm). The different echoes from the top of the membrane and the central support layer can be clearly distinguished on the waveform of “fouling 0 min” (lower part of Figure 1). This is the clean membrane signal and will be used as the “reference signal or waveform” to detect fouling. Progress of fouling caused an apparent movement of the “top” echo to earlier arrival times and ended with two clearly separated echoes after 60 min of run time. However, this signal is actually the convoluted waveform of the top-of-membrane echo and the signal reflected off the fouling layer. It should be noted that all signals were treated with a Fourier-type filter to remove spurious high-frequency components.

FIGURE 3. Ultrasonic response signal of the clean nylon membrane and differential signals from 0 to 60 min of fouling operation after subtraction of the clean membrane signal.

FIGURE 4. Amplitude of differential signals versus fouling resistance during paper mill effluent fouling experiment. The resolution in Figure 1 between the signals from the top of the membrane and the support layer is of the order of λ/4, that is, the normal limit of detection of about 50 µm under the conditions of the experiment. The sampling interval was 4 ns. It is very difficult to measure fouling layer thickness, as the density change from the top of the membrane to the top of the fouling layer is too small to create a separate peak. Other physical changes during fouling of the membrane, and concomitant with ultrasonic recordings, were monitored as changes in permeate flux (Figure 2). At the beginning of the fouling experiment, a rapid decline in permeate flux was observed (up to 3 min), after which it reached a near steadystate flux within 5 min. This rapid decline in flux would result primarily from particle deposition or the formation of a hemicellulose/lignin-rich fouling layer (23) that led to reduced porosity and blocked the membrane pores and surface. The shear stress was apparently too low and did not sweep away the material that accumulated on the membrane surface. The gradual flux decline during the period of steady state is

FIGURE 5. Arrival time of differential signals and thickness of fouling layer versus operation time during fouling experiment. presumed to occur as a result of the slow growth and consolidation of the cake layer (23). The ultrasonic signal responses revealed corresponding changes in time and amplitude domain to flux decline as fouling proceeded (Figure 1). The appearance of a new echo was observed at 5 min of fouling operation because of both filling of membrane pores (membrane 63% porous) and formation of a fouling layer on the membrane surface. The growth (cake density) and movement (cake thickness) of the new echo was seen from 10 to 60 min of fouling operation as fouling proceeded (25). However, the fouling layer is so thin and the acoustic impedance difference so small that two successive echoes (fouling layer and membrane surface) cannot be separated. The fouling echo combined with the clean membrane echo to form the new echo. This makes biofouling hard to monitor, and hence data manipulation is necessary to enhance resolution and detection. Ultrasonic Differential (Separated) Signals. During the progress of fouling, no change in signal shape was possible VOL. 39, NO. 18, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

7301

FIGURE 6. Wavelet transform and superimposed waveform (red) for an ultrasonic signal reflected off a nylon membrane (T, top of membrane; S, nylon suport). until about 10 min of running time. However, an almost immediate response to fouling was visible on the differential signal after 30 s (Figure 3). The differential signals were obtained by subtracting the reference signal from the subsequent signals. The differential signal represents an echo signal of a fouling layer on a membrane surface. It is shown in Figure 3 that a differential signal appeared within 0.5 min of fouling operation, and grew over the period 1, 5, 15, and 30 min, as fouling proceeded. A sharper signal emerged after 60 min of fouling operation due to a denser fouling layer on the membrane surface (25, 26). Figure 3 also demonstrates that the amplitude of the differential signals increased as fouling proceeded (see also Figure S5 in the Supporting Information). The fouling build-up and growth results in an increase in acoustic impedance of the cake layer (24). This means that the density of the fouling covering the membrane surface increases due to particle deposition and perhaps increasing dehydration of the initial viscous deposit and fouling layer compressibility as fouling proceeds (26). The fouling resistance was calculated from our data using the resistance model (see Supporting Information) and presented graphically in Figure 4. A linear relationship between differential signals and fouling resistance exists. It suggests that fouling resistance increases with operation time, resulting from the deposition and build-up of the fouling layer. The increasing amplitude also indicates a density increase of this fouling layer. Figure 3 also shows that a time-domain movement as a result of an increase in the thickness of the fouling layer (26) can be observed and quantified. The arrival time of an echo from the front surface of the membrane is 6.36 µs. From the difference in arrival times between differential echoes and the echo of the front membrane surface, the thickness of the cake layer (breakdown products of lignin or lignosulfonate (23)) on the membrane surface can be calculated (using eq 2, see Supporting Information). The sound velocity in the fouling matter was measured to be 1536 m/s by relating the 7302

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 18, 2005

FIGURE 7. Cross-section (microtome cut) through a slightly fouled nylon membrane showing (from top down) the fouling layer (F), the porous membrane filter (P), and the nylon support layer (Ny). The total membrane thickness is about 160 µm. ultrasonic waveform to a theoretical waveform (24). The thickness data in Figure 5 are representative of arrival times. This is the standard technique in nondestructive testing to measure distances and hence thicknesses (see ref 24). As shown in Figure 5, the thickness of the fouling layer in real time was about 6 and 9 µm after 0.5 and 1 min of fouling operation, respectively. This results in a resolution of h/λ ) hf/V ) 0.04 (where f ) 7.5 MHz). This thickness value depends on calculating the density of the fouling layer to give a sound velocity; however, it cannot easily be confirmed by a second technique, because removing the fouled membrane and allowing partial drying distorts the water swollen fouled layer. Yet reasonable values are obtained, for example, shrinkage to about one-half by drying (see Figure 7) after 10 min of operation time. Ultrasonic Signal Processing by Wavelet Transform. The waveform (A-scan) of a reflected signal from a clean nylon membrane (about 160 µm thick) was Fourier transformed into wavelet (using AGU-Vallen Wavelet). The wavelet

FIGURE 8. Wavelet transform and superimposed waveform (red) for an ultrasonic signal reflected off a fouled nylon membrane after 60 min (T, top of membrane; S, nylon support; F, top of fouling layer). transform (WT) is presented in Figure 6 together with its waveform. Irrespective of the sign of the waveform, the wavelet transform shows all deviations from zero as positive magnitude. In the waveform, a small shoulder in front of the main echo indicates a reflecting surface of lower sonic impedance. Comparison with microscopic observations and measurements identifies it as reflection from the top layer of the membrane. On the wavelet, it becomes a small peak (T), clearly separated from the signal received from the supporting layer (S). As illustration, a microtome cross-section of a slightly fouled nylon membrane is shown in Figure 7, consisting of a fouling layer (F), a highly porous layer (P), and a fibrous supporting layer (Ny), followed by another porous layer. It appears that in the case of a clean membrane the first echo would come from the porous surface while the second and stronger echo would be produced by the nylon support. The overall density of a dry nylon membrane was measured using a pyknometer and gave an average value of 1.25 g/cm3. The density of a water-saturated membrane was found to be about 1.1 g/cm3, and these two values gave a porosity of 63%. Using the sound velocities in nylon and pure water of 2200 and 1483 m/s, respectively, it also can be calculated that in a water-saturated membrane it is of the order of 1700 m/s. The small shoulder on the A-scan in Figure 6 can thus be explained by the different sonic impedances of the middle layer of solid nylon fibers (support structure) and the water-logged membrane. The time difference between the surface reflection and the echo from the nylon support (0.095 µs) can be accurately determined on the wavelet transform (Figure 6), which can be translated into a thickness of about 80 µm (the porous top layer, P, in Figure 7). Such a degree of accuracy would not be possible from A-scan measurements. It should also be noted that the energy content (color coding in Figure 6) of the top reflection is lower than that of the support layer, thus indicating the higher density of the latter. The absorption of higher frequencies (probably by scattering) by the porous top layer is likewise well documented by the WT.

FIGURE 9. Changes in signal level (wavelet magnitude) for the “T” and “S” peaks with progress of fouling. During the progress of fouling, an additional peak was forming adjacent to the “T” peak. At a later stage in the fouling process, this appeared to be a completely separated peak (peak F). This new peak is shown on the wavelet transform of the 60-min fouling waveform (Figure 8). This peak could be identified as a signal coming from the top of the fouling layer. Because this layer also had physical properties different from those of the porous membrane, the reflected signal had a slightly higher frequency component. The time difference between the newly formed peak “F” and the peak “T” was used to determine the thickness of the fouling layer. At the time of 60 min (Figure 8), this was about 21 µm. During the progress of fouling, the peaks “F”, “T”, and “S” could then be monitored and their wavelet magnitude measured. In the present experiment, the latter peaks showed a decrease in wavelet magnitude, concomitant with fouling (Figure 9), indicating less ultrasonic energy reaches these surfaces due to reflection from the fouling layer on top of the membrane. The relative position of the peaks, however, remained constant. The decline in wavelet magnitude (Figure 9) is mirrored by a similar decline in permeate flux (Figure VOL. 39, NO. 18, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

7303

FIGURE 10. 2D Fourier spectra for the time windows “T” (top of membrane) (left) and “S” (nylon fibrous support) (right), combined for all stages of fouling. 2), showing that the wavelet transform of the ultrasonic signals reflects the physical processes that accompany the progress of membrane fouling. One of the advantages of the wavelet transform is that for any specific time or frequency, the Fourier spectrum can be given in a two-dimension plot dissecting the area of interest. In our case, the amplitude changes were examined for the time windows of the “T” and “S” peaks to highlight the influence of fouling on energy loss and frequency changes. The results for both time windows are presented in Figure 10 (left, 0.35 µs; right, 0.45 µs). The time window for peak “S” (at 0.45 µs) shows a steady decline in WT magnitude, accompanied by a shift in maximum frequency from about 5500 to 4500 kHz, in direct response to an increase in fouling. Because the surface properties (mainly porosity and roughness) of the membrane change almost instantaneously with the onset of fouling, less sound energy is available for reflection from the nylon support layer (“S”). We attribute the change in frequency of the reflected signal to the properties of the fouling layer, which could be less porous than the membrane and would absorb less of the higher frequency components, thus passing a lower proportion of high frequency to the support layer. This is apparent from the build-up of a second frequency maximum, in relation to fouling, next to the “T” peak. The top, porous part of the membrane, on the other hand, retains its original absorbing properties for high frequency and thus retains its frequency distribution in the spectrum, but decreases its energy content according to the losses caused by the formation of the fouling layer. Samples of the membrane in various stages of fouling were also examined by atomic force microscopy (AFM) and scanning electron microscopy (SEM). (AFM and SEM micrographs are provided in the Supporting Information, Figures S6 and S7, respectively.) The results show a rapid change from the porous membrane to a smooth surface. This would collaborate the argument presented above. The surface roughness as measured by AFM showed values of 132 nm for the clean membrane, 94 nm after 10 s, 69 nm after 1 min, and 42 nm after 5 min. Thus, the application of Fourier wavelet transform to the processing of ultrasonic signals shows individual peaks for the different surfaces and so increases resolution even further while also helping to create a better understanding of the 7304

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 18, 2005

processes taking place on and within the membrane during filtration. Rapid changes can be detected in the behavior of the membrane components in response to fouling. The results encourage the application and utilization of such techniques for an automatic cleaning process and quality control in liquid separation processes using membrane filters.

Acknowledgments We gratefully acknowledge the financial support of the Water Research Commission of South Africa. Special thanks go to Professor A. G. Fane of the University of New South Wales, Australia, for his helpful suggestions. We would also like to thank Vallen Systems GmbH, Germany, for the use of and support with their wavelet software, and Prof. Sharff, IFU, for assisting in creating a modified Plugging Index Meter based on this technology. Finally, we acknowledge early assistance from Prof. A. Greenberg and Prof. Bill Krantz.

Supporting Information Available The theory of ultrasonic measurement and the wavelet transform procedure, and AFM and SEM, etc., photographs of membrane fouling to corroborate the ultrasonic data. This material is available free of charge via the Internet at http:// pubs.acs.org.

Literature Cited (1) Butt, F. H.; Rahman, F.; Baduruthamal, U. Identification of scale deposits through membrane autopsy. Desalination 1995, 101, 219-230. (2) Dudley, L. Y.; Darton, E. G. Membrane autopsysa case study. Desalination 1996, 105, 135-141. (3) Hendricks, T. J.; Williams, F. A. Diffusion-layer structure in reverse osmosis channel flow. Desalination 1971, 9, 155-180. (4) Altmann, J.; Ripperger, S. Particle deposition and layer formation at the crossflow microfiltration. J. Membr. Sci. 1997, 124, 119128. (5) Ethier, C. R.; Lin, D. C. Refractometric measurement of polarized layer structure: studies of hyaluronic acid ultrafiltration. J. Membr. Sci. 1992, 68, 249-261. (6) Mc Donogh, R. M.; Bauser, H.; Stroh, N.; Grauschopf, U. Experimental in situ measurement of concentration polarisation during ultra- and micro-filtration of bovine serum albumin and Dextran Blue solutions. J. Membr. Sci. 1995, 104, 51-63. (7) Mores, W. D.; Davis, R. H. Direct visual observation of yeast deposition and removal during microfiltration. J. Membr. Sci. 2001, 189, 217-230.

(8) Wakeman, R. J. Visualization of cake formation in crossflow microfiltration. Trans. Inst. Chem. Eng., Part A 1994, 72, 530540. (9) Li, H.; Fane, A. G.; Coster, H. G. L.; Vigneswaran, S. Direct observation of particle deposition on the membrane surface during crossflow microfiltration. J. Membr. Sci. 1998, 149, 8397. (10) Yao, S.; Costello, M.; Fane, A. G.; Pope, J. M. Noninvasive observation of flow profiles and polarisation layers in hollow fibre membrane filtration modules using NMR micro-imaging. J. Membr. Sci. 1995, 99, 207-216. (11) Lynnworth, L. C. Ultrasonic measurement for process control: theory, techniques, applications; Academic Press: Boston, 1989. (12) Ensminger, D. Ultrasonics; Marcel Dekker: New York, 1998. (13) Smith, J. L.; Evans, D. H.; Bell, P. R. F.; Naylor, A. R. Time domain analysis of embolic signals can be used in place of highresolution Wigner analysis when classifying gaseous and particulate emboli. Ultrasound: Its Appl. Med. Biol. 1998, 24, 989-993. (14) Haerle, A. G.; Haber, R. A. Real-time monitoring of cake thickness during slip casting. J. Mater. Sci. 1993, 28, 5679-5683. (15) Parthun, M. G.; Johari, G. P. Dynamics of a molecule’s growth: Ultrasonic relaxation studies. J. Chem. Phys. 1995, 102, 63016307. (16) Aimar, P.; Baklouti, S.; Sanchez, V. Membrane-solute interactions: influence on pure solvent transfer during ultrafiltration. J. Membr. Sci. 1986, 29, 207-224. (17) Kools, W. F. C.; Konagurthu, S.; Greenberg, A. R.; Bond, L. J.; Krantz, W. B.; Boomgaard, Th. Van den; Strathmann, H. Use of ultrasonic time-domain reflectometry for real-time measurement of thickness changes during evaporative casting of polymeric films. J. Appl. Polm. Sci. 1998, 69, 2013-2019. (18) Alig, I.; Tadjbakhsch, S. Film formation and crystallization kinetics of polychloroprene studies by an ultrasonic shear wave reflection method. J. Polym. Sci., Part B: Polym. Phys. 1998, 36, 1730-1741. (19) Lavrentyev, A. I.; Rokhlin, S. I. An ultrasonic method for determination of elastic moduli, density, attenuation and

(20)

(21)

(22)

(23)

(24)

(25)

(26)

(27)

thickness of a polymer coating on a stiff plate. Ultrasonics 2001, 39, 211-221. Lavrentyev, A. I.; Rokhlin, S. I. Determination of elastic moduli, density, attenuation, and thickness of a layer using ultrasonic spectroscopy at two angles. J. Acoust. Soc. Am. 1997, 102, 34673477. Mairal, A. P.; Greenberg, A. R.; Krantz, W. B.; Bond, L. J. Realtime measurement of inorganic fouling of RO desalination membranes using ultrasonic time-domain reflectometry. J. Membr. Sci. 1999, 159, 185-196. Mairal, A. P.; Greenberg, A. R.; Krantz, W. B. Investigation of membrane fouling and cleaning using ultrasonic time-domain reflectometry. Desalination 2000, 130, 45-60. Li, J.; Sanderson, R. D.; Jacobs, E. P. Noninvasive visualization of the fouling of microfiltration membranes by ultrasonic timedomain reflectometry. J. Membr. Sci. 2002, 201, 17-29. Li, J.; Hallbauer, D. K.; Sanderson, R. D. Direct monitoring of membrane fouling and cleaning during ultrafiltration using a noninvasive ultrasonic technique. J. Membr. Sci. 2003, 215, 3352. Li, J.; Sanderson, R. D.; Hallbauer, D. K.; Hallbauer-Zadorozhnaya, V. Y. Measurement and modelling of organic fouling deposition in ultrafiltration by ultrasonic transfer signals and reflections. Desalination 2002, 146, 177-185. Li, J.; Hallbauer-Zadorozhnaya, V. Y.; Hallbauer, D. K.; Sanderson, R. D. Cake-layer deposition, growth and compressibility during microfiltration, measured and modelled using a noninvasive ultrasonic technique. Ind. Eng. Chem. Res. 2002, 41, 4106-4115. Private communication. Dr. P. J. Cilliers, Hermanus Magnetic Observatory, NRF Hospital Street, Box 32, Hermanus, South Africa. Telephone: +27-28-3121196.

Received for review February 28, 2005. Revised manuscript received June 27, 2005. Accepted July 5, 2005. ES050414Y

VOL. 39, NO. 18, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

7305