Chapter 6

Analysis of membrane lifetimedata

In this chapter, membrane lifetime was evaluated by means of acceleratedageing experiments. A pressure pulse unit was used to perform the ageingexperiments in an accelerated way. An experimental design has been setup and 4 ageing factors were varied at two levels. The four ageing factorsstudied were: fouling status of the membrane, cleaning agent concentration,magnitude of the back pulse and number of applied back pulses. The in-tegrity of the membrane modules was evaluated by means of permeabilitytesting, pressure decay tests and bubble tests. Also tensile tests were per-formed to investigate the mechanical properties of the membrane modules.The collected data was used for an analysis of variance to determine whichageing factors and/or which combinations of ageing factors influence mem-brane lifetime. The outcome of this analysis showed that the fouling statusin combination with the number of applied pressure pulses were significantageing factors. Additional tensile tests confirmed these results.

6.1 Introduction

In general, failure (breach, leaks, etc) of fibers is believed to be the result ofexposure to cleaning chemicals, mechanical stress and the membrane foulingstatus.

Research on fiber failure was performed by Huisman and Williams [1], who

99

100 CHAPTER 6. ANALYSIS OF MEMBRANE LIFETIME DATA

analyzed the causes of membrane damage for ultrafiltration membranes usedto treat wastewater. Autopsy results, bubble point tests and tensile strengthtests revealed that the membranes were not damaged by chlorine, otherhalogens or cleaning agents, but merely by mechanical forces, e.g. pressureshocks. The effects of mechanical force were shown to be stronger if themembranes were fouled.

Work of Gijsbertsen-Abrahamse et al. [2], focused on the relationship be-tween fiber failure and fiber properties. They concluded that fiber failuredepends on the strength of the membrane material, the operating conditionsand incidents. In this research was found that on average fiber failure occursannually 1 to 10 times on 1 million fibers.

Also Childress et al. [3] researched fiber failure with respect to fiber pro-perties. In an evaluation of different membranes the material, the symmetry,the modulus of elasticity, diameter, thickness, potting technique and moduleflow pattern (inside-out, outside-in) were investigated. This research showedthat stress at the juncture of the potting is likely to lead to the formationof fractures

In recent studies the influence of sodium hypochlorite in a changing pH en-vironment has been tested on PES/PS UF membranes [4; 5; 6; 7]. It wasfound that in certain pH regions hypochlorite reacts more aggressive. Otherauthors investigated the so-called ageing phenomenon for other membranematerials and applications [8; 9; 10; 11; 12; 13; 14; 15; 16]. Statistical analy-sis and modeling based on data collected by means of accelerated ageing testswas done in different fields of science and engineering [17; 18; 19; 20; 21; 22].However, little has been published on accelerated ageing tests of UF mem-branes. In this study factors which influence membrane life time are sys-tematically identified. A 4x2 (four factors, two levels) experimental designis constructed and each experiment is performed in threefold. With the col-lected data an ANOVA (Analysis Of Variance) is performed to investigatewhich factors and which combinations of factors significantly influence mem-brane integrity. Also the influence of ageing on the membrane mechanicalproperties is evaluated by means of tensile testing.

6.2. THEORY 101

Table 6.1: Factors and levels. *At a frequency of around 25 pulses perminute. 100.000 back pulses correspond to 3 years of operation.Factor Symbol Low (-1) High (+1)Fouling status (-) A Clean module Fouled moduleNaClO Conc. (ppm) B 0 ppm 1000 ppmMagn. of the back pulse (bar) C 0.5 bar 2.0 barNumber of back pulses* (#) D 100.000 300.000

6.2 Theory

6.2.1 Design of experiments

In this study, four influence factors are varied at two levels. In table 6.1the factors A, B, C and D are denoted, respectively as: the fouling status,the sodium hypochlorite concentration, the magnitude of the back pulsesand the number of back pulses. Table 6.1 also shows the values for high-and low experimental settings. Based on the number of applied pulses, theoverall test time is for 100.000 pulses around 60-85 hours and for 300.000around 170 to 250 hours. In hypochlorite exposure such times correspond to60.000 to 85.000 ppm.h and 170.000-250.000 ppm.h, respectively. Assumingthat in normal membrane operation a backwash is performed four times perhour, the membranes were aged after 100.000 pulses nearly 3 years, and after300.000 pulses nearly 9 years.

A full factorial experimental design scheme is presented in table 6.2. It isnoted that experiments as shown in the experimental design matrix normallyare executed in random order.

6.2.2 ANOVA method

This method is used to compare the magnitude of the effects of factors withthe magnitude of experimental error and is called analysis of variance. Ifthe magnitude of a factor effect is large in comparison to the experimentalerror, the changes in the selected response cannot occur by coincidence andthose changes in the response are considered to be the effects of the influencefactors. The factors causing a variation in the response are called significant.In this study, the so-called F-test was used in the analysis of variance and

102 CHAPTER 6. ANALYSIS OF MEMBRANE LIFETIME DATA

Table 6.2: The experimental design matrix.

Combination A B C D1 -1 -1 -1 -1a 1 -1 -1 -1b -1 1 -1 -1ab 1 1 -1 -1c -1 -1 1 -1ac 1 -1 1 -1bc -1 1 1 -1abc 1 1 1 -1d -1 -1 -1 1ad 1 -1 -1 1bd -1 1 -1 1abd 1 1 -1 1cd -1 -1 1 1acd 1 -1 1 1bcd -1 1 1 1abcd 1 1 1 1

6.2. THEORY 103

Table 6.3: Signs for calculating the contrasts, according to [23].

exp A B AB

C AC

BC

AB

C

D AD

BD

AB

D

CD AC

D

BC

D

AB

CD

1 -1 -1 1 -1 1 1 -1 -1 1 1 -1 1 -1 -1 1a 1 -1 -1 -1 -1 1 1 -1 -1 1 1 1 1 -1 -1b -1 1 -1 -1 1 -1 1 -1 1 -1 1 1 -1 1 -1ab 1 1 1 -1 -1 -1 -1 -1 -1 -1 -1 1 1 1 1c -1 -1 1 1 -1 -1 1 -1 1 1 -1 -1 1 1 -1ac 1 -1 -1 1 1 -1 -1 -1 -1 1 1 -1 -1 1 1bc -1 1 -1 1 -1 1 -1 -1 1 -1 1 -1 1 -1 1abc 1 1 1 1 1 1 1 -1 -1 -1 -1 -1 -1 -1 -1d -1 -1 1 -1 1 1 -1 1 -1 -1 1 -1 1 1 -1ad 1 -1 -1 -1 -1 1 1 1 1 -1 -1 -1 -1 1 1bd -1 1 -1 -1 1 -1 1 1 -1 1 -1 -1 1 -1 1abd 1 1 1 -1 -1 -1 -1 1 1 1 1 -1 -1 -1 -1cd -1 -1 1 1 -1 -1 1 1 -1 -1 1 1 -1 -1 1acd 1 -1 -1 1 1 -1 -1 1 1 -1 -1 1 1 -1 -1bcd -1 -1 1 -1 1 -1 1 -1 1 -1 1 -1 1 -1 1abcd 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

given as:

FA =S2

A

S2ERROR

(6.1)

where S2ERROR is the variance of the overall error and where S2

A is the vari-ance with respect to factor A, estimated according to:

s2A =

KA

φA

(6.2)

φA is the degree of freedom with respect to A and KA is the sum of squareswith respect to A. For an experimental design of 4x2, the sum of squareswith respect to A is calculated from its contrast according to:

KA =C2

A

24n(6.3)

where n is the number of experiment repetitions and CA is the contrastwith respect to A . The contrast is calculated by use of the scheme in table6.3. If X1 is the experimental outcome of experiment (1), and Xa is the

104 CHAPTER 6. ANALYSIS OF MEMBRANE LIFETIME DATA

Table 6.4: The randomized experimental scheme.

Exp. Combination Exp. Combination1 b 9 ab2 bc 10 abc3 bd 11 abd4 bcd 12 abcd5 1 13 ad6 c 14 acd7 d 15 [a]8 cd 16 [ac]

outcome of experiment (a), and Xb is the outcome experiment (b), etc.,than the contrast with respect to A can be calculated as a summation of theoutcomes of the experiments while the signs of the experimental values aretaken from from table 6.3.

CA = −X1 + Xa −Xb + · · · −Xbcd + Xabcd (6.4)

6.3 Materials and methods

The experimental design matrix of table 6.2 is quasi randomized and theexperiments are executed in the order of table 6.4. Each experiment isperformed in threefold. After the experiment, the membrane is tested forpermeability, pressure decay and defect fibers. Tensile tests were performedto investigate the mechanical properties of the membranes. The methodsused to determine these parameters are briefly discussed below. It shouldbe noted that run 15 and 16 could not be performed, due to the limitationof available membrane modules.

6.3.1 The pressure pulse unit

In fig. 6.1 and fig. 6.2 the PPU (Pressure Pulse Unit) is presented. Themain part of the PPU is the membrane pump, which can pump sodiumhypochlorite - or water - at a pressure of 0 to 3 bar through membranemodules with a frequency of around 20 to 30 pulses per minute. Using

6.3. MATERIALS AND METHODS 105

two pressure restrictions, two sets of experiments can be simultaneouslyperformed at two different pressures. In the experiments, industrial sodiumhypochlorite, colloquially known as caustic bleach (15 weights%, pH between12-12.5) of ViVo Chemicals B.V. was used and diluted with demineralizedwater to a 1000 ppm solution. It is known that in the pH range 11-13the concentration of chlorine ions in the hypochlorite is stable. The pH ofthe solution used in the accelerated ageing experiments was around 11.5.This value corresponds to industrially applied pH ranges of hypochloritecleanings. The solution was refreshed on a daily basis. The used PESmembrane modules were Norit-Xiga RX300 PSU hollow fiber UF moduleswith a membrane surface of 0.07m2. The Xiga fibers have a poly sulphonehousing and PES/PVP flow distributors (fibers). Every test module contains100 fibers with a length of 30 cm. Potting procedures and materials for cleanand fouled fibers were the same.Fouled modules were assembled from fibers that were taken from a 4 inchmodule that was operated in a pilot with surface water over a six monthperiod. Cleaning was performed on the membrane every 24 hours, onlyapplying caustic cleaning with sodium hydroxide 0.1M and acidic cleaningwith 0.1 mol.l−1 hydrochloric acid. PES fibers are known to withstandcleanings with these chemicals in high concentrations.

Sodium hypo chlorite

tank

Membrane pump

30 pulses/min 0 - 3.0 bar

Pressure restriction with manometers

1" membrane modules

Figure 6.1: Schematic representation of the pressure pulse unit.

6.3.2 Permeability test

The permeability test is a commonly used measure of the overall performanceof a membrane. Permeability of the membrane modules is determined at the

106 CHAPTER 6. ANALYSIS OF MEMBRANE LIFETIME DATA

beginning of the experiment Lp,0 and is determined after 100.000 and 300.000back pulses. To determine the permeability Lp , the flux J and viscosity,µ are determined as a function of the trans-membrane pressure. In theseexperiments flux and viscosity were evaluated at 4 different trans-membranepressures in the range of 0.2 to 1.0 bar. The permeability of the membranecan easily be calculated from Darcy’s equation:

µJ = Lp∆P (6.5)

From a graph where µJ is plotted versus ∆P , the slope (permeability, Lp)can be determined.

Figure 6.2: Picture of the bench scale setup used to perform permeabilitymeasurements. On the left side a pressurized feed tank can be seen, in themiddle the test module and on the right side a balance to determine thefiltration flow.

The permeabilities of the membrane at respectively 100.000 and 300.000back pulses are normalized with respect to the initial permeability of themembrane. The initial permeabilities of the fouled fibers was averagely 4times lower as clean fibers. Fouled fibers had an average permeability of5.10−13(m) and clean fibers had an average permeability of 2.10−12 (m).

6.3. MATERIALS AND METHODS 107

Figure 6.3: Picture of the pressure pulse unit and the ageing frame.

6.3.3 Pressure decay test

The PDT (pressure decay test) can be used to check the integrity of a mem-brane unit. The main principle is based on measurement of the pressuredrop on the feed side after draining and pressurizing. In general the ap-plied pressure is approximately 1 bar. The overall duration of the test isapproximately 15 minutes. Acceptable pressure decreases are around 1.5mbar.min−1.

6.3.4 Bubble test

The bubble test can be used to locate damaged membrane fibers. The prin-ciple is based on the pressurization of the membrane below bubble pointpressure in a water filled basin. Bubble tests are performed with test pres-sures around 0.3-0.5 bar. The overall duration of a test is around 10 to 15minutes. After location of the damaged fiber, normally the fiber is repairedwith use of a special repair equipment.

108 CHAPTER 6. ANALYSIS OF MEMBRANE LIFETIME DATA

6.3.5 Tensile testing

One of the most fundamental tests to determine the mechanical propertiesof a material is the tensile test. In this study a tensile test apparatus ofZwick is used to determine the modulus of elasticity and the rupture point.A 10N load cell is used to strain the test fibers.

6.4 Results and discussion

Fig. 6.4, 6.5 and 6.6 show the results for the 14 performed experimental runs.Especially for runs (ad), (abd), (acd) and (abcd) the increase in permeability,PDT values and number of defect fibers is striking. Those experimentsencompass fouled modules. Another striking result that can immediatelybe derived from the data is that clean modules can be operated under highpressures (2.0 bar) with high cleaning agent dosings (1000 ppm) for a highnumber of pressure pulses (300.000) without major change in permeabilityand without detectable fiber leakage. It should be further noted that thevariation in experimental outcomes of the PDT for tests (abd) and (acd)and (abcd) may be caused by difference in qualities of the modules, basedon their initial permeabilities.

6.4. RESULTS AND DISCUSSION 109

Permeability test results

0.00

1.00

2.00

3.00

4.00

5.00

6.00

1 A B AB C AC BC ABC D AD BD ABD CD ACD BCD ABCD

Combination

Lp

/Lp

0

EXP1 EXP2 EXP3

Figure 6.4: Experimental values for the permeability tests.

Pressure decay test results

1

10

100

1000

1 A B AB C AC BC ABC D AD BD ABD CD ACD BCD ABCD

Combination

(mb

ar/m

in)

EXP1 EXP2 EXP3

Figure 6.5: Experimental values for the pressure decay tests.

110 CHAPTER 6. ANALYSIS OF MEMBRANE LIFETIME DATA

Bubble test results

0

1

2

3

4

5

6

7

1 A B AB C AC BC ABC D AD BD ABD CD ACD BCD ABCD

Combination

Nu

mb

er o

f d

efec

t fi

ber

s

EXP1 EXP2 EXP3

Figure 6.6: Experimental values for the bubble tests.

The analysis of variance for the different membrane integrity measurementsyield the results shown in tables 6.5, 6.6 and 6.7. The factors A (foulingstatus), D (number of back pulses) and the combination AD, yield F-valuesthat exceed the critical F-values for all three integrity methods. If F > 4.49,the effect of the influence factor is 95% significant, and if F > 8.53, theeffect of the influence factor is 99% significant. It is further noted thatthe calculated F-value is proportional to the square of its contrast. Thecontrast is directly calculated from the experimental values. However, thecalculation scheme used to calculate the contrasts, influences the outcomesby the positive of negative sign added in the summation. For example, fromfigure 6.4 can be seen that the experiments ad, abd and acd have the highestchanges in permeability. Those values contribute mostly to the calculationof its contrast:

CA = · · ·+ Xad · · ·+ Xabd · · ·+ Xacd · · · = 4.95

CB = · · · −Xad · · ·+ Xabd · · · −Xacd · · · = 1.43

CAB = · · · −Xad · · ·+ Xabd · · · −Xacd · · · = 0.02

For this reason, the factor A can be significant, but a combination whichincludes this factor, such as AB is not. The ANOVA results are graphicallypresented in fig. 6.7, 6.8 and 6.9. The critical F-values are denoted by thedotted line (95% significant) and the dashed line (99% significant)

6.4. RESULTS AND DISCUSSION 111

Table 6.5: ANOVA results for the permeability tests.

Source K φ S2 F Critical F-valuesA 4.95 1 4.95 12.09 8.53 0.99B 1.43 1 1.43 3.50 4.49 0.95AB 0.02 1 0.02 0.05C 0.02 1 0.02 0.05AC 2.61 1 2.61 6.38BC 0.04 1 0.04 0.10ABC 2.25 1 2.25 5.50D 24.27 1 24.27 59.32AD 19.88 1 19.88 48.59BD 1.11 1 1.11 2.72ABD 2.13 1 2.13 5.21CD 0.39 1 0.39 0.95ACD 1.16 1 1.16 2.84BCD 0.34 1 0.34 0.83ABCD 1.14 1 1.14 2.78Error 6.55 16 0.41Total 68.28 47

112 CHAPTER 6. ANALYSIS OF MEMBRANE LIFETIME DATA

Table 6.6: ANOVA results for the pressure decay tests.

Source K φ S2 F Critical F-valuesA 718586 1 718586.02 7.77 8.53 0.99B 12192 1 12192.19 0.13 4.49 0.95AB 13906 1 13906.02 0.15C 14043 1 14042.52 0.15AC 12449 1 12448.52 0.13BC 5440 1 5440.02 0.06ABC 4238 1 4237.52 0.05D 736809 1 736808.52 7.97AD 722998 1 722997.52 7.82BD 12773 1 12772.69 0.14ABD 14249 1 14248.52 0.15CD 14387 1 14386.69 0.16ACD 12256 1 12256.02 0.13BCD 5398 1 5397.52 0.06ABCD 4351 1 4351.02 0.05Error 1479503 16 92468.92Total 3783574 47

6.4. RESULTS AND DISCUSSION 113

Table 6.7: ANOVA results for the bubble tests.

Source K φ S2 F Critical F-valuesA 27.00 1 27.00 6.97 8.53 0.99B 0.08 1 0.08 0.02 4.49 0.95AB 1.33 1 1.33 0.34C 1.33 1 1.33 0.34AC 0.08 1 0.08 0.02BC 1.33 1 1.33 0.34ABC 0.08 1 0.08 0.02D 36.75 1 36.75 9.48AD 27.00 1 27.00 6.97BD 0.08 1 0.08 0.02ABD 1.33 1 1.33 0.34CD 1.33 1 1.33 0.34ACD 0.08 1 0.08 0.02BCD 1.33 1 1.33 0.34ABCD 0.08 1 0.08 0.02Error 62.00 16 3.88Total 161.25 47

114 CHAPTER 6. ANALYSIS OF MEMBRANE LIFETIME DATA

Permeabilities

0.00

10.00

20.00

30.00

40.00

50.00

60.00

70.00

D AD A

AC

ABC

ABD B

ACD

ABCD BD

CD

BCD BC

AB C

Factor

F-v

alu

e

Figure 6.7: F-values for the permeability tests.

PDT

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

8.00

9.00

10.00

D AD A

CD

ABD C AB

BD AC

ACD B BC

BCD

ABCD ABC

Factor

F-v

alu

e

Figure 6.8: F-values for the pressure decay tests.

6.4. RESULTS AND DISCUSSION 115

Bubble test

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

8.00

9.00

10.00

D A AD

AB C BC

ABD CD

BCD B AC

ABC BD

ACD

ABCD

Factor

F-v

alu

es

Figure 6.9: F-values for the bubble tests.

At full scale level, similar results are obtained. Although there is muchvariation. Currently integrity testing is done ’semi-online’, where at frequentintervals an airflow test is performed. Test data shows for an average plantthat airflow values exceed their maximum once or twice per year. Estimatesfor membrane lifetime are around 4 to 7 years. This corresponds to resultspresented in this research. However, for clean fibers membrane lifetimeis considerably higher than for fouled fibers. Tensile tests have also been

0 10 20 30 40 50 600

2

4

6

8

10

12

Strain, ε (%)

Sta

ndar

d fo

rce,

σ (

MP

a)

Clean modules pulsed with waterFouled modules pulsed with NaClOClean modules pulsed with NaClOClean modules before pulse testingFouled modules before pulse testing

Figure 6.10: Tensile test curves for different recorded experiments.

performed with non-defected membrane fibers from experiment (cd) (cleanmodules, back pulsed with water), (bcd) (clean modules, back pulsed with

116 CHAPTER 6. ANALYSIS OF MEMBRANE LIFETIME DATA

Table 6.8: Tensile test values

Type E (MPa) εr (%) σm (N)Clean (Initial) 226 (s=17) 50 (s=5) 2.70 (s=0.18)Fouled (Initial) 230 (s=10) 23 (s=4) 2.19 (s=0.07)Clean (Pulsed with water) 243 (s=19) 51 (s=4) 2.80 (s=0.16)Fouled (Pulsed with NaClO) 228 (s= 9) 22 (s=3) 2.16 (s=0.08)Clean (Pulsed with NaClO) 227 (s= 5) 50 (s=1) 2.94 (s=0.06)

NaClO), (abcd) (fouled modules back pulsed with NaClO) and with cleanand fouled fibers that were not used in the pulse tests. During the tensiletests, fibers broke in the middle and not at the grip. The tensile tests wereperformed in six fold and the average results are plotted in fig. 6.10 Fig.6.9 shows the standard force or stress, σ (MPa) as a function of the strain ε(%). Striking is the difference in elongation εr (the rupture point or strainat break) for clean modules and fouled modules. The mechanical propertiesof clean and fouled fibers after the pulse tests were conducted are similar tothe clean and fouled fibers before pulse tested were conducted. Mechanicalproperties do not seem to change much after intense exposure to NaClO, butdrastically change if the fibers are fouled. In table 6.8 the elastic modulus, E,the elongation, εr and the maximal force, σm are shown with the calculatedstandard deviations.

6.5 Conclusions

A systematic study of factors that influence membrane lifetime was per-formed. An experimental design was developed and executed and the resultswere analyzed by means of ANOVA . The study revealed that the foulingstatus of a membrane, the number of applied back pulses and the combi-nation of those two factors influence membrane integrity significantly. Italso showed that the effect of sodium hypochlorite and the magnitude of theapplied pressure pulse, do not significantly influence membrane integrity.Additional tensile tests were performed and showed that the rupture pointof fouled modules is almost half of the value for clean modules. Mechanicalproperties of clean modules back pulsed with water and sodium hypochloritewere similar.

6.5. CONCLUSIONS 117

List of symbols

Symbol Meaning Unit (trivial) [SI]C Contrast (-)εr Elongation (%)E Elastic modulus (MPa) [N.m−2]F Value of F-test (-)φ Degree of freedom (-)J Flux (m.s−1)K Sum of squares (-)Lp Permeability (m)n Number of experiment repetitions (-)µ Viscosity (Pa.s) [N.s.m−2]∆P Trans membrane pressure (Pa) [N.m2]S Variance (-)s Standard deviation (-)σ Standard force (MPa) [N.m−2]σm Maximal force (N)X Experimental value (-)

118 CHAPTER 6. ANALYSIS OF MEMBRANE LIFETIME DATA

Bibliography

[1] Huisman I.H. and Williams K, (2004) ”Autopsy and failure analysis ofultrafiltration membranes from a waste-water treatment system”, Desali-nation 165 p. 161.

[2] Gijsbertsen-Abrahamse A.J. , Cornelissen E.R. and Hofman J.A.M.H.,(2006) ”Fiber failure frequency and causes of hollow fiber integrity loss”,Desalination 194 p. 251.

[3] Childress A.E. , Le-Clech P., Daugherty J.L., Chen C. and Lesly G.L.,(2005) ”Mechanical analysis of hollow fiber membrane integrity in waterreuse applications”, Desalination 180 p. 5.

[4] Begoin L., Rabiller-Baudry M., Chaufer B., Hautbois M. and Doneva T.,(2006) ”Ageing of PES industrial spiral-wound membranes in acid wheyultrafiltration”, Desalination 192, p. 25.

[5] Rouaix S. , Causserand C. and Aimar P., (2006) Experimental study ofthe effects of hypochlorite on polysulfone membrane properties”, Journalof Membrane Science, Vol. 277, p. 137.

[6] Thominette F. , Farnault O., Gaudichet-Maurin E. , Machinal C. andSchrotter J. (2006) Ageing of polyethersulfone ultrafiltration membranesin hypochlorite treatment, Vol 282 (1-2), p. 198

[7] Gaudichet-Maurin E. and Thominette F., (2006) ”Ageing of polysulfoneultrafiltration membranes in contact with bleach solutions”, Journal ofMembrane science, Vol. 282(1-2), p. 198.

[8] Benavente J. and Vazquez M.I. (2004), Effect of age and chemical treat-ments on characteristic parameters for active and porous sublayers of poly-meric composite membranes, Journal of Colloid and Interface Science, 273p. 547.

[9] Shung Chung T., Khean Teoh S., (1999) The ageing phenomenon ofpolyethersulphone hollow fibre membranes for gas separation and theircharacteristics, Journal of Membrane Science, 152, p. 175.

[10] Platt S. , Nystrom M., Bottino A. and Capannelli G., (2004) Stabilityof NF membranes under extreme acidic conditions, Journal of MembraneScience, 239 p. 91.

BIBLIOGRAPHY 119

[11] Tessaro I.C., da Silva J.B.A. and Wada K., (2005) Investigation of someaspects related to the degradation of polyamide membranes: aqueous chlo-rine oxidation catalyzed by aluminum and sodium laurel sulfate oxidationduring cleaning, Desalination, 181 p. 275.

[12] Zhang S. , Shen J., Qiu X., Weng D. and Zhu W., (2005) ESR and vi-brational spectroscopy study on poly(vinylidene fluoride) membranes withalkaline treatment, Journal of Power Sources, Vol. 153(2), p. 234.

[13] Nghiem L.D. and Schafer A.I., (2005) Diagnosing Fouling of HollowFibre MF Membranes in Wastewater Reclamation, Integrated Conceptsin Water Recylcing, Peer-reviewed by ICWR2005

[14] Gijsbertsen-Abrahamse A.J., (2004) What is causing fibre failure?,Membrane filtration symposium - Integrity, Amsterdam.

[15] Kruithof J.C., (1999), Membrane integrity monitoring at the UF/ROHeemskerk Plant, Technical report PWN Water Supply Company NoordHolland.

[16] Causserand C., Kinetics of degradation of polysulfone membranes in re-lation to cleaning conditions using chlorine, Laboratoire de Genie Chim-ique, miniposter

[17] Crevecoeur G.U., (1993) A model for the integrity Assessment of Ageingrepairable systems, IEEE Transactions on reliability, Vol. 42, No. 1.

[18] Lee L., (1980) Testing adequacy of the Weibull and log linear rate modelsfor a Poisson process, Technometrics,. Vol. 22, No. 2, p. 195.

[19] Ryu D. and Chang S., (2005) Novel concepts for reliability technology,Microelectronics reliability, 45 p. 611.

[20] Wang K.S., Hsu F.S. and Liu P.P. (2002), Modeling the bathub shapehazard rate function in terms of reliability, Reliability Engineering andSystem Safety, 75, p. 397.

[21] Chen Z., (2000) A new two-parameter lifetime distribution with bathtubshape or increasing failure rate function, Statistics & Probability Letters,49, p. 155.

[22] R. Jiang R., Ji P. and Xiao X., (2003) Aging property of unimodal failurerate models, Reliability Engineering and System Safety, 79, p. 113.

120 CHAPTER 6. ANALYSIS OF MEMBRANE LIFETIME DATA

[23] Box G.E.P., Hunter W.G. and Hunter J.S. (1978), Statistics for ex-perimenters, an introduction to design data analysis and model building,published by Wiley Interscience.