Sunteți pe pagina 1din 8

2f

13/7/00

2:39 pm

Page 135

Z. Ujang* and G.K. Anderson**


*Institute of Environmental and Water Resource Management, Faculty of Civil Engineering,
Universiti Teknologi Malaysia, KB 791, 80990 Johor Bahru, Malaysia
**Department of Civil Engineering, University of Newcastle upon Tyne, NE1 7RU, UK
Abstract This paper describes an investigation on the effect of operating parameters on the separation of
metal chelates using LPROM. The objective of this study was to optimise the process of metal chelates
removal by a LPROM using statistical factorial design. Factorial experiment by statistical design was applied,
in which a central composite factorial design (half replicate) was used, then followed up by a star design to
give a central composite factorial design. A bench-scale spiral wound configuration of sulphonated
polysulphone LPROM was used at various operating conditions, i.e. operating conditions, solute
concentrations, EDTA, pH and temperature. It has been shown experimentally that the effect of pressure is
non-linear with respect to percentage of metal removal at different concentrations of metal ions in the feed
solution. Observation of the response surface implies that the operating pressure was not the significant
parameter in determining the percentage of zinc removal in the LPROM. The operating pressure,
temperature and concentration of EDTA in the feed solution were the most significant parameters for
permeate flux variation. It can also be concluded that for a given set of feed and experimental conditions,
permeate flux increased linearly with operating pressure and temperature. The other parameters, i.e. the
concentration of zinc in the feed solution and pH, were not statistically significant.
Keywords Statistical factorial design; half replicate; heavy metals; low pressure reverse osmosis
membrane; metal chelates removal

Water Science and Technology Vol 41 No 1011 pp 135142 IWA Publishing 2000

Effect of the operating parameters on the separation of


metal chelates using low pressure reverse osmosis
membrane (LPROM)

Introduction

There is an increasing demand for effective industrial wastewater treatment, reclamation


and reuse with low operating cost. Reverse osmosis (RO) has been widely used in this area,
however, with relatively high operating costs due to the high-pressure system and massive
cleaning procedures. Therefore low-pressure RO membrane (LPROM) has been re-introduced to the water and wastewater industry in the past several years. Actually LPROM is
relatively not a new concept in membrane technology. Its inception could be traced back to
the 1960s. At that time, the low-pressure system was not an attractive system because of low
flux and non-reliable membrane materials.
Recently, due to the advancement in materials science for membrane manufacturing,
LPROM has been reconsidered. Nemeth (1998) reviewed the innovative system with special emphasis on performance optimisation. Ujang and Anderson (1998) presented the performance of LPROM for separating mono- and divalent ions. The feasibility of using
LPROM below 100 psig for effective removal of metal ions from their solvents was also
investigated (Ujang and Anderson, 1996). Most of the studies done so far on LPROM have
been focused on bench-scale feasibility approach using various pollutants. The application
of LPROM for heavy metal removal, particularly to investigate the effect of operating
parameters on separation of metal chelates, however, has not been studied very extensively
up to date.

135

2f

13/7/00

2:39 pm

Page 136

Experimental design and experiment

Z. Ujang and G.K. Anderson

The aim of this paper was to study the effect of the operating parameters on the separation of
metal chelates using LPROM. The specific objective was to optimise the process of metal
chelates removal by a LPROM by using statistical factorial design.
The LPROM system has been described in our earlier papers (Ujang and Anderson, 1996
and 1998). Table 1 illustrates the membrane specifications. The LPROM test rig was
designed based on recommendations by the American Standard of Testing Materials
(ASTM: D4194-89). All materials of construction were selected from high-quality plastic
or stainless steel for all wetted parts to prevent contamination of the feed solutions by corrosion products. Two modules of sulphonated polysulphone negatively-charged LPROMS
were supplied by NWW-Acumem Ltd., Britain.
The performance of LPROMs was evaluated by response parameters, i.e. (a) permeate
flux (unit used: l/hr), and (b) the percentage of metal removal, under varying experimental
conditions such as operating pressure, feed concentration of metal ions, and temperature, at
a constant pH and 40% recovery. Preliminary experiments were carried out to establish data
on permeate flux decline with time and flux changes in solute rejection with time, using
500 mg/l NaCl at 60 psi (414 kPa), 25C and 40% recovery.
Experimental procedures

A known quantity of ZnCl2 was weighed before adding to the EDTA solution (disodium
salt). The pH was adjusted according to the design experiment between pH 2.5 to 6.5 by
adding NaOH or HCl. The total EDTA to total metal ratio was varied in the range of 0.5 to
2.0. Sampling procedures and record-keeping were carried out according to procedures recommended by American Standard for Testing Materials (ASTM: D4472-89). Collected
samples were preserved according to Standard Methods for the Examination of Water and
Wastewater (1992).
Preliminary experiments were carried out to establish data on flux decline with time and
flux changes in solute rejection with time. The data are useful for interpreting the various
changes that would take place, especially those due to membrane compaction, hydrolysis or
fouling. The objectives of preliminary experiments were as follows:
i. to characterise the LPROMs in terms of their water transport properties,
ii. to quantify as much as possible short-term compaction effects on membrane transport,
and
Table 1 Specification of LPROM unit

136

Membrane material:

Sulphonated polysulphone

Membrane configuration:

Spiral-wound

Mode of operation:

Continuous

Active surface area:

0.465 m2

pH range:

211 for continuous operation; 113 for short term exposure

Temperature:

245C

Pressure:

Maximum of 125 psi

Rejection@60 psi,25C:

9597% for softened tap water; 9094% for hard tap water

Charge:

Negative

Thickness:

150175 micron (<1 micron active layer)

Max. feed flow rate:

0.454 m3/s

Expected life span:

3 years

Special characteristics:

Chlorine resistant and low-pressure operation

Manufacturer:

NWW-Acumem Ltd. (UK)

2f

13/7/00

2:39 pm

Page 137

iii. to determine the time that the system needed to reach steady state operation after
changing operating parameters.

Z. Ujang and G.K. Anderson

The LPROMs were used with distilled water for 24 hours at 414 kPa, 25C and 40%
recovery. After the distilled water flux data had been collected, the LPROMs were emptied
and NaCl solutions were used. The LPROMs were washed, rinsed and filled with distilled
water and run for a further one hour. The distilled water flux data were used as the reference.
All analytical procedures were carried out according to Standard Methods for the
Examination of Water and Wastewater (1992). Metal analysis of feed, concentrate and permeate samples were carried out by Atomic Adsorption Spectrophotometer (AAS: Model
UNICAM 929). The analysis of total complexing agents was determined by TOC (total
organic carbon) measurements. Conductivity measurement was taken from samples by
using a Conductivity Monitor. The pH of the sample solutions was monitored using a
standard laboratory pH meter.
Experimental design

Factorial experiment by statistical design was applied, in which a central composite factorial design (half replicate) was used, then followed up by a star design to give a central
composite factorial design.
Procedures

Step 1. Flux of permeate and the percentage heavy metal removal were chosen to be
maximised.
Step 2. The following parameters in Table 2 were selected which may affect the response
parameters, and the levels were determined (known here as control parameter):
Table 2 Parameters and level for experimental design
Operating pressure (X1):

60

Heavy metal conc. (X2):


EDTA conc. (X3):

0.0

pH (X4):
Temperature (X5):

80

90

1.0

1.5

2.0

0.5

1.0

1.5

2.0

2.5
5

70

0.5

15

3.5
25

4.5
35

100

5.5

2.5
6.5

45

Step 3. The appropriate experimental design was chosen (central composite factorial
design): 2k=25=32 runs of experiments.
Step 4. Experimental materials were randomised to runs and the order of runs, subject to the
constraints imposed by the experimental design.
Thirty two possibilities of experiment for 25 central composite factorial design were
determined. Sixteen possibilities or 12 (32) for half replicate of 251 were determined and
tabulated as described by Montgomery (1991).
Step 5. The experiments as shown in the first six columns in Table 3 were conducted to
obtain the value of permeate flux and the percentage of heavy metal removal.
Step 6. Regression analysis and modelling were carried out using MINITAB (Minitab
Stastistical Software: Release 9 by Minitab Inc., U.S.A.)
Step 7. ANOVA were determined by using MINITAB.

137

2f

13/7/00

2:39 pm

Page 138

Step 8. The surface response graphs were plotted by using Uniras-Unimap graphical
package.
Step 9. A conclusion was drawn from the optimum combination of factors from the surface
response graphs to determine the very significant control parameters.
Z. Ujang and G.K. Anderson

Step 10. The other experiments of the half replicate were investigated by changing centre
point.
Step 11. A star design as suggested by Metcalfe (1994) was conducted, which consists of
12 runs.
Step 12. Then, followed by steps 6 to 9.
Results and discussion

Table 3, particularly columns 7 and 8, summarises the percentage of zinc removal and zinc
flux from zinc chloride feed solutions for 28 runs (16 runs of half replicate factorial design,
Table 3 Results on half replicate factorial design and a star design for zinc chloride

Run

138

Press

InZn

EDTA

Temp

ZnRem

ZnFlux

(psi)

(mM)

(mM)

pH

(C)

(%)

(l/hr)

70

0.5

3.5

35

98.36

1.75

90

0.5

3.5

15

99.11

1.95

70

0.5

3.5

15

97.75

1.4

90

0.5

3.5

35

98.73

1.85

70

1.5

3.5

15

99.33

1.35

90

1.5

3.5

35

99.32

2.10

70

1.5

3.5

15

98.59

1.35

90

1.5

3.5

15

98.81

1.65

70

0.5

5.5

15

98.98

1.45

10

90

0.5

5.5

35

99.37

2.0

11

70

0.5

5.5

35

97.14

1.75

12

90

0.5

5.5

15

97.55

1.8

13

70

1.5

5.5

35

98.43

1.5

14

90

1.5

5.5

35

97.28

1.75

15

70

1.5

5.5

15

97.06

1.2

16

90

1.5

5.5

35

98.6

1.75

17

80

1.5

4.5

25

99.1

1.8

18

80

1.5

4.5

25

99.1

1.75

19

100

1.5

4.5

25

99.3

20

60

1.5

4.5

25

99.0

1.5

21

80

2.5

4.5

25

97.1

1.7

22

80

0.5

4.5

25

98.4

1.8

23

80

1.5

4.5

25

99.4

1.75

24

80

1.5

4.5

25

97.3

1.8

25

80

1.5

6.5

25

99.3

1.8

26

80

1.5

2.5

25

97.2

1.7

27

80

1.5

4.5

45

98.7

1.9

28

80

1.5

4.5

97.2

1.3

Note: Press = pressure; InZn = influent concentration of ZnCl2; EDTA = concentration of


ethylidiaminetetraacetic acid; Temp = temperature; ZnRem = zinc removal; ZnFlux = permeate flux

2f

13/7/00

2:39 pm

Page 139

ZnRem = 98.1 + 0.0146 Press 0.694 InZn + 0.386 EDTA 0.067 pH + 0.0056 Temp

(1)

where Press is the operating pressure, InZn is the influent concentration of zinc, EDTA is the
concentration of EDTA, and Temp is the temperature.
Table 4 shows the summary statistics, including the estimated coefficients for the independent parameters in the experiments, their StDev (the estimated standard deviation), tratio (the ratio of the coefficient to its standard deviation), the probability value p (the
proportion of occurrences of the sample point in a long series of experiments), s value (the
estimated standard deviation of all predictors), and R-sq or R2 (the proportion of the corrected sum of squares of the y accounted for by the regression. In this case y is ZnRem) and Rsq(adj) (adjusted R-Sq) values for the fitted equation. Table 5 shows the summary of
analysis of variance (ANOVA) which consists of DF (the degree of freedom), SS (sum of
squares), MS (mean of square), F (F-ratio, the ratio of the regression mean square to the
error mean square), and p values. From the summaries of the multiple regression analyses,
which were shown in both Tables 4 and 5, it can be concluded that the concentration of zinc
in the feed solution was the most significant parameter in this study, since the p value was
0.053. In general, the p value should not exceed 0.05 for 95 percent degree of confidence.
For this regression, various attempts have been made to reduce the p value, but it was a failure due to the nature of the existing experimental data. The other parameters were not statistically significant since the p values were 0.399, 0.261, 0.696 and 0.751 for the operating
pressure, EDTA, pH and temperature respectively (s value 0.8199).
The effect of the controlled parameters (i.e. those which could be changed) on the percentage of zinc removal was further investigated by the 3-D response surface using a
regression model from Equation 1. Figure 1 illustrates the response surface which shows
that the effect of pressure is non-linear with respect to percentage of metal removal at different concentrations of metal ions in the feed solution. As can be seen from the peak
response surface, the highest percentage metal removal reached 99.5% at 98 psi (667 kPa).
The lowest surface in Figure 1 shows that an operating pressure between 70 and 75 psi (483
to 518 kPa) resulted in zinc removals of between 97.9 and 98%. Observation of the response

Z. Ujang and G.K. Anderson

plus 12 runs for a star design) with the corresponding experimental conditions. All results
shown in Table 3 were for an average of eight samples, collected hourly. Table 3 was further
analysed by multiple regression analyses using the Minitab statistical package. The main
objective of the analysis was to formulate a model to show the combined effect of the independent parameters investigated for the percentage of zinc removal and zinc flux (Metcalfe,
1994 and Montgomery, 1991).
The regression equation for predicting the percentage of zinc removal from zinc chloride
feed solutions is shown as follows:

Table 4 Summary of multiple regression analysis for zinc removal


Predictor

Coefficient

StDev

t-ratio

Constant

98.060

1.665

Press

0.01461

58.88

0.000

0.01699

0.86

0.399
0.053

InZn

0.6938

0.3398

2.04

EDTA

0.3858

0.3347

1.15

0.261

0.0673

0.1699

0.40

0.696

0.01748

0.32

0.751

pH
Temp

0.00561

s = 0.8199 R-sq = 24.0% R-sq(adj) = 6.7%


Note: StDev=standard deviation; p=probability value; s=s value; R-sq=residual2;
R-sq(ajd)=adjusted residual2

139

2f

13/7/00

2:39 pm

Page 140

Table 5 Analysis of variance for zinc removal


Source

DF

SS

MS

Regression
Error
Total

5
22
27

4.6687
14.7888
19.4575

0.9337
0.6722

1.39

0.267

Z. Ujang and G.K. Anderson

Source

DF

SeqSS

Press
InZn
EDTA
pH
Temp

1
1
1
1
1

0.5797
3.0460
0.8932
0.0805
0.0693

Note: DF=degree of freedom; SS=sum of square; MS=mean of square;


F=F-ratio; SeqSS=sequential sum of square

Figure 1 3D response surface for zinc removal

140

Figure 2 3D response surface for zinc flux

2f

13/7/00

2:39 pm

Page 141

surface implies that the operating pressure was not the significant parameter in determining
the percentage of zinc removal in the LPROM.
The regression equation for predicting the flux from zinc chloride feed solutions is as
follows:
ZnFlux = 0.497 + 0.0151 Press 0.0689 InZn
(2)

Table 6 shows a summary of the statistics including the estimated coefficients for the independent parameters in the experiments, their StDev, t-ratio, p and R-sq values for the fitted
Eq. 2. Table 7 shows a summary of the analysis of variance (ANOVA). The results of R-sq
(adj) indicate that around 77.6% of the variation in permeate flux could be explained by the
fitted regression Eq. 2. This percentage value is adequate for the prediction of the response
parameter, i.e. the permeate flux. On the other hand, 22.4% of the variation remains unexplained which may be due to the variations in experimental conditions. From the summaries
of the multiple regression analysis which are shown in Tables 6 and 7, it can be concluded
that the operating pressure, temperature and concentration of EDTA in the feed solution
were the most significant parameters, since p values were 0, 0 and 0.015 respectively (s
value 0.1085). The other parameters, i.e. the concentration of zinc in the feed solution and
pH ( p values were 0.139 and 0.390 respectively), were not statistically significant.
The effect of the most significant controlled operating parameters, i.e. pressure and temperature, on the percentage of zinc flux was further investigated by the 3-D response surface
using a regression model from Eq. 2. Figure 2 shows that the effect of the operating pressure

Z. Ujang and G.K. Anderson

0.117 EDTA 0.0197 pH + 0.0118 Temp

Table 6 Summary of multiple regression analysis for zinc flux


Predictor

Coefficient

StDev

t-ratio

Constant

0.4970

Press

0.015114 0.002247

0.2203

2.26

0.034

6.73

0.000

InZn

0.06894

0.04494

1.53

0.139

EDTA

0.11667

0.04428

2.63

0.015

pH

0.01970

0.02247

0.88

0.390

5.11

0.000

Temp

0.011818 0.002312

s = 0.1085 R-sq = 81.8% R-sq(adj) = 77.6%


Note: StDev = standard deviation; p = probability value; s = s value;
R-sq = residual2; R-sq(ajd) = adjusted residual2

Table 7 Analysis of variance for zinc flux


Source

DF

SS

MS

Regression
Error
Total

5
22
27

1.15977
0.25880
1.41857

0.23195
0.01176

19.72

0.000

Source

DF

SeqSS

Press
InZn
EDTA
pH
Temp

1
1
1
1
1

0.70042
0.07042
0.08167
0.00000
0.30727

Note: DF=degree of freedom; SS=sum of square; MS=mean of square;


F=F-ratio; SeqSS=sequential sum of square

141

2f

13/7/00

2:39 pm

Page 142

Z. Ujang and G.K. Anderson

and temperature on permeate flux is linear. The peak response surface clearly shows that the
permeate flux increases as a function of the operating pressure and temperature, in particular between 90 to 100 psi (621 to 690 kPa) at a temperature between 20 and 30C. The figure
also demonstrates that operating pressures below 80 psi (552 kPa) and temperatures below
18C, however, produce the lowest response surface i.e. a permeate flux below 1.5 l/hr.
From Figure 2, it can be concluded that, for a given set of feed and experimental conditions,
permeate flux increased linearly with operating pressure and temperature.
Conclusion

Using half replicate factorial design, it has been experimentally shown that the effect of
pressure is non-linear with respect to percentage of metal removal at different concentrations of metal ions in the feed solution. Observation of the response surface implies that the
operating pressure was not the significant parameter in determining the percentage of zinc
removal in the LPROM. On the other hand, that the operating pressure, temperature and
concentration of EDTA in the feed solution were the most significant parameters for permeate flux variation. It can also be concluded that for a given set of feed and experimental conditions, permeate flux increased linearly with operating pressure and temperature. The
other parameters, i.e. the concentration of zinc in the feed solution and pH, were not
statistically significant.
Acknowledgement

This study would have been impossible without technical support from Dr. Peter Cardew
from NWW-Acumem Ltd. and Dr. Andrew V. Metcalfe from the Department of
Mathematics, University of Newcastle upon Tyne, England., as well as financial support
from the Universiti Teknologi Malaysia.
References
Metcalfe, A.V. (1994). Statistics in Engineering. Chapman and Hall, London.
Montgomery, D.C. (1991). Design and Analysis of Experiments, 3rd ed. John Wiley & Sons, Singapore.
Nemeth, J.E. (1998). Innovative system designs to optimise performance of ultra-pressure RO membranes.
Desalination, 118, 6371.
Standard Methods for the Examination of Water and Wastewater (1992). APHA, AWWA and WRF.
Washington, D.C.
Ujang, Z. and Anderson, G.K. (1996). Application of low-pressure reverse osmosis membrane for Zn2+ and
Cu2+ removal from wastewater. Wat. Sci. Tech., 34(9), 247253.
Ujang, Z. and Anderson, G.K. (1998). Performance of low pressure reverse osmosis membrane for
separating mono- and divalent ions. Wat.,Sci.Tech., 38(45), 521528.

142

S-ar putea să vă placă și