Search

The evolution of the cake during dead-end microfiltration of different compositions of bidisperse feeds was investigated using network modeling and three-dimensional (3D) optical coherence tomography (OCT) imaging. The foulant particles were latex of two different sizes (namely, 3 and 5 µm) and the membrane was polycarbonate track-etched with nominal diameter of 2 µm. The results from the network model indicate that (i) the deposition of the smaller particles on the non-porous area was reduced by the presence of larger particles, whereas that of the larger particles was not affected by the concentration of the smaller ones; and (ii) monodisperse small particles gave a greater specific cake resistance than monodisperse large particles, and the specific cake resistance of the bidisperse mixture can only be reduced with a sufficient proportion of larger particles. OCT results substantiated the model by revealing that (i) bidisperse feeds tended to form homogeneous cakes, but the presence of a higher concentration of large particles led to heterogeneous cakes; (ii) the presence of a lower concentration of larger particles enhanced clustering, while a higher concentration reduced that; and (iii) in the longer-term filtration, the monodisperse feed of small particles dynamically deposited and detach continuously, while the bidisperse feed with the highest concentration of larger particles exhibited negligible deposition and detachment. Economic Development Board (EDB) Ministry of Education (MOE) We acknowledge funding from the Ministry of Education - Singapore Academic Research Funds Tier 2 (MOE2014-T2-2-074; ARC16/15) and Tier 1 (2015-T1-001-023; RG7/15), the GSK (GlaxoSmithKline) – EDB (Economic Development Board - Singapore) Trust Fund, and the Joint Singapore-Germany Research Project Fund (SGP-PROG3-019).

To obtain a better backwashing strategy, effects of released air on the effective backwashing length of dead-end hollow fiber membranes were investigated in this study. Ultrasonic spectrum and AGU-Vallen Wavelet software were imported to ensure a more intuitive understanding on effects of released air from echoes. Modeling and experimental results in this study showed that the air accumulation at the most distant membrane point was a recombination process which contained air release and dissolution. The recombination process contained air release as the pressure type changed from inside-outside to outside-inside and air dissolution in the outside-inside pressure condition. Moreover, longer fiber was more prone to the air accumulation in the most distant membrane lumen. The mathematical model predicted the volume of final remaining air that depended on the instantaneous pressure difference, backwashing flux, backwashing duration, membrane surface area, water viscosity and temperature. Experimental verification of ultrasonic waveforms was consistent with the model prediction. To overcome the effects brought by released air and maintain backwashing effectiveness, a novel method was developed by adding a segment of gas filtration membrane at the end of the hollow fiber which had finally extended the effective backwashing length. These studies were expected to have implications for the design and operation of hollow fiber membrane backwashing.

The present study deals with the specific energetic consumption evaluation of dead-end and tangential ultrafiltration (UF) (100 kDa) of salted suspensions on polysulfone membrane in regards of backwash and aeration conditions. For TMP ranging from 0.2 to 0.8 bar, it was found that specific energy consumption during tangential UF was always higher than in dead-end mode. During dead-end UF at 0.4 bar with a backwash frequency equal to 15 min, the 30 s backwash duration allowed obtaining the highest net permeate flux (27 L h −1 m −2 ) and the lowest specific energy consumption (146 Wh m −3 ). Changing the backwash frequency to respectively 5 or 60 min did not induce lower energy consumption. The bubbling located at the bottom of fiber bundle accounted for more than 68% of the total specific energy consumption and enhanced the net permeate flux only in the presence of backwash. During the UF at 0.4 bar with 30 s and 15 min of backwash time and frequency respectively, the total energy consumption appeared lower (35 Wh m −3 ) without bubbling. Finally, the lowest specific energy consumption (16 Wh m −3 ) was obtained for dead-end UF at 0.4 bar without backwashing or bubbling; in these operating conditions, stabilized net permeate flux was equal to 16 L h −1 m −2 .

This paper describes the dead-end ultrafiltration of humic substances with hollow fiber modules. We show that the resistance of a fouling layer decreases with increasing flux. The lower resistance of the fouling layer is caused by a loss in retention of the membrane for humic acid at these higher fluxes. The ideal cake filtration model does not fit the filtration resistance appropriately even considering a retention smaller than unity. The extended model falls short since it does not incorporate an axially dependent retention. Using five small modules in series, as well as a specially designed flat sheet cell, we prove that both the flux and retention in a fiber module is a strong function of the axial coordinate (fiber length). The experimental results with the small modules gave a complete picture of the axial features in dead-end filtration of humic substances with fibers, and the results can be used to explain the observed behavior in the large module. The main contributor of the axial phenomena is hypothesized to be charge interactions between the membrane and the solute combined with the small size of the foulants causing diffusive concentration gradients of the humic acids.

In this paper, a membrane lifetime model is developed and experimentally validated. The lifetime model is based on the Weibull probability density function. The lifetime model can be used to determine an unambiguous characteristic membrane lifetime. Experimental results showed that membrane lifetime shortens if the average membrane fouling status increases. The lifetime modeling results are then used to determine the economic lifetime of membranes. Subsequently, the economic lifetime of a membrane is used to optimize membrane lifetime, i.e. minimizing the total costs. Based on the experimental results it can be concluded that the total costs are minimal if the average membrane fouling status is approximately 1.7 x the membrane resistance. (c) 2008 Elsevier B.V. All rights reserved.

The properties of dead-end microfiltration were explored under constant pressure using two types of activated sludge controlled under the condition of different air flow rates. The activated sludge cultured at the air flow rate of 0.15 L min −1 (the anaerobic condition) exhibited a significant flux decline compared with the case of the air flow rate of 2.33 L min −1 (the aerobic condition). It was found from the results of microfiltration of the supernatant separated by centrifugation that the constituents in the supernatant caused a major cake resistance in microfiltration of the activated sludge. The average specific filtration resistance for filtration of the activated sludge was closely consistent with that for filtration of the supernatant at low pressure (49 kPa). However, the cake resistance of the microbial floc in microfiltration of the activated sludge became substantial with increasing filtration pressure because of high compressibility of the microbial floc. Moreover, the foulant and the fouling mechanism in microfiltration of the supernatant were evaluated from both microfiltration test of the supernatant and microfiltration test of the filtrate collected thereby. As a result, the effects of the pore size and material of the microfiltration membrane on the flux decline behaviors in dead-end microfiltration were reasonably elucidated.

Dead-end microfiltration of microbial suspensions is reviewed with particular emphasis on the factors affecting the specific cake resistance. The effects of cell size and shape, cell surface properties (including charge), ionic environment, fermentation medium components and ageing effects are reviewed in detail. The measurement of specific resistance and correlation of cake compressibility data is described and current understanding of the underlying mechanisms of cake compression is discussed. Factors affecting the packing arrangement of cells in filter cakes are also reviewed and discussed in the context of the Carman-Kozeny equation.

A pilot study for pretreatment of RO in reclamation of municipal wastewater effluent has been conducted at Ulu Pandan, Singapore using a ultrafiltration system with the capacity of 150 m3/day developed by NORIT. The objective of this study was to establish the process of alum injection at the hollow fiber UF plant to achieve a stable operation in dead-end flow mode. Trial runs on various concentrations of alum dosage and water recoveries were conducted. The pilot results showed that the concentration of alum dosage was critical and the optimum concentration was determined at 2.5 mg/L as Al. A stable operation of the plant in dead-end flow has been achieved by optimizing the operating conditions. The pilot plant was operated stably for 4 weeks under the optimum conditions with TMP between 0.22 and 0.26 bar and normalized flux of UF membrane of (200–250) × 10−5 L m−2 h−1 Pa−1 at 20 °C. Product water with average SDI of 1.6 and turbidity of 0.13 NTU was produced from the plant at water recovery over 90%. The pilot results also showed that particle (1–15 μm) in the UF permeate was 2 counts/mL. Total coliform was not detectable. Total bacteria was 69 cfu/mL. The total power consumption was 0.194 kWh/m3. Of which, 0.100 kWh/m3 was for the operation of UF unit and 0.094 kWh/m3 was for the submersible pump and the air compressor. The chemical cost was 1.3 cents (US)/m3 product water at a capacity of 5 m3/h. The results demonstrated that the process of alum injection established at the UF plant achieved a stable operation in dead-end mode with low cost and the UF membrane used could be a suitable pre-treatment option prior to RO for producing NEWater from the secondary treated effluent.

A series of dead-end filtration experiments are compared with a recently proposed pore blocking model that contains only one experimentally fitted parameter. This parameter has physical meaning and it can be verified by an independent check of the experimental data. The challenge suspension used was near monosized latex particles of ∼0.45 μm filtered on a track-etched membrane with similar sized pores of ∼0.4 μm. The filtered suspension concentration ranged from 0.00006 to 0.01% (w/w) and the transmembrane pressures varied from 1000 to 20,000 Pa. During the experiments three stages of microfiltration were observed. There was an initial stage that fitted a sieving (or pore blocking) model until the mass of latex deposited per unit membrane area (specific mass) reached 2×10 −3 kg m −2 . The model required a single parameter that was found to fit all the data under different experimental operating conditions. The second stage of pore blocking could also be fitted using a second parameter and the predicted flux at the end of the first stage of pore blocking. When the deposited particles had reached a depth of approximately 12 particle layers, or specific mass of 4.8×10 −3 kg m −2 , the flux then showed normal cake filtration behaviour.

The effects of a direct current electric field on the transient and dynamically balanced filtration rate in dead-end inclined and downward ultrafiltration were explored under constant pressure using protein (bovine serum albumin) solution. It is found that, in downward electro-ultrafiltration, the dynamically balanced filtration rate qe is directly proportional to the electric field strength E. Also, in inclined electro-ultrafiltration, qe increases with E above the critical electric field strength Ec. However, qe becomes invariant below Ec. The model was developed to describe the processes occurring during inclined and downward electro-ultrafiltration, and it accounted well for the variation in filtration rate with time.