Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Electrodialytic desalination of brackish water: determination of optimal experimental parameters using full factorial design

Electrodialytic desalination of brackish water: determination of optimal experimental parameters... Appl Water Sci (2017) 7:4563–4572 https://doi.org/10.1007/s13201-017-0609-2 ORIGINAL ARTICLE Electrodialytic desalination of brackish water: determination of optimal experimental parameters using full factorial design 1 1 2 1 • • • • Soumaya Gmar Nawel Helali Ali Boubakri Ilhem Ben Salah Sayadi 1 1 Mohamed Tlili Mohamed Ben Amor Received: 22 March 2017 / Accepted: 29 August 2017 / Published online: 15 September 2017 The Author(s) 2017. This article is an open access publication Abstract The aim of this work is to study the desalination Introduction of brackish water by electrodialysis (ED). A two level- three factor (2 ) full factorial design methodology was used The shortage of drinking water is a major problem in the to investigate the influence of different physicochemical southern communities of Tunisia. In these regions, the parameters on the demineralization rate (DR) and the traditional sources of fresh water are insufficient to meet specific power consumption (SPC). Statistical design the demand and are being stressed by competing uses, such determines factors which have the important effects on ED as irrigation and industrial needs (Walha et al. 2007). performance and studies all interactions between the con- However, in recent times and with the development of sidered parameters. Three significant factors were used desalination processes, there has been an increasing interest including applied potential, salt concentration and flow in using brackish waters as a source of potable water. rate. The experimental results and statistical analysis show Desalination is a process that removes dissolved minerals that applied potential and salt concentration are the main from seawater or brackish water or treated waste water. effect for DR as well as for SPC. The effect of interaction These processes create more valuable water by converting between applied potential and salt concentration was saline waters into a resource (Reig et al. 2014; McGovern observed for SPC. A maximum value of 82.24% was et al. 2014a; Tanaka et al. 2015). Several desalination obtained for DR under optimum conditions and the best methods have been developed to obtain fresh drinking -1 value of SPC obtained was 5.64 Wh L . Empirical water. There are mainly two families of desalination regression models were also obtained and used to predict technologies used throughout the world today. They the DR and the SPC profiles with satisfactory results. The include thermal (evaporative) and membrane technologies. process was applied for the treatment of real brackish water Membrane methods are less energy intensive than thermal using the optimal parameters. methods and since energy consumption directly affects the cost effectiveness and feasibility of using desalination Keywords Demineralization rate  Desalination  technology, membrane methods such as reverse osmosis Electrodialysis  Full factorial design  Specific power (RO) and electrodialysis (ED) are attracting great attention consumption lately (Sadrzadeh and Mohammadi 2008; McGovern et al. 2014b). ED is a membrane process for the separation of ions across charged membranes from one solution to another under the influence of an electrical potential dif- & Soumaya Gmar ference used as a driving force (Sadrzadeh and Moham- soumayagmar@yahoo.com madi 2008). This process has been widely used to produce drinking Laboratory of Natural Water Treatment, Center of water from brackish and sea water (Fidaleo and Moresi Researches and Water Technologies, P.B 273, 8020 Soliman, 2005; Lee et al. 2006; Jing et al. 2012; Galama et al. 2014; Tunisia Zourmand et al. 2015; Reig et al. 2016a, b; Monohar et al. Laboratory of Waste Water Treatment, Center of Researches 2017). It has been also used in treatment of industrial and Water Technologies, P.B 273, 8020 Soliman, Tunisia 123 4564 Appl Water Sci (2017) 7:4563–4572 effluents (Ghyselbrecht et al. 2013), purification of amino different concentrations for checking the optimal condi- acids and other organic compounds (Elisseeva et al. 2002) tions. Then, real brackish water was treated by ED using and to remove heavy metals from waste water (Nemati the optimal conditions. et al. 2017). Many factors influence the ED performance such as applied potential (Elmidaoui et al. 2002; Banasiak et al. Materials and methods 2007), salt concentration (Banasiak et al. 2007; Shady et al. 2012), flow rate of dilute compartment and temperature Chemicals reagents (Sadrzadeh and Mohammadi 2008; Ben Sik Ali et al. 2010a; Shady et al. 2012). Analytical grade sodium chloride (NaCl) and sodium A literature survey revealed that DR was increased with sulphate Na SO are used to produce solution with 2 4 the applied potential but it was decreased at high values of known amounts of salts and electrode rinse solution, salt concentration (Banasiak et al. 2007; Sadrzadeh and respectively. KCl is used to calibrate conductivity cell. Mohammadi 2008; Shady et al. 2012). Sadrzadeh and Ethylene diamine tetraacetic acid (EDTA), sulphuric Mohammadi (2008) and Ben Sik Ali et al. (2010a) acid (H SO ,), barium chloride (BaCl ), hydrochloric 2 4 2 observed that the salt percent removal increases when the acid (HCl) sodium hydroxide (NaOH), sodium fluoride flow rates decrease. They suggested that for low flow rates (NaF) and glacial acid acetic (CH COOH) are used to the residence time of ions in the dilute compartment analyze the real water. Each solution was prepared using increases. On the other hand, Elmidaoui et al. (2002) distilled water. All chemicals were purchased from demonstrated that DR increased with increasing flow rates. Sigma-Aldrich. The authors attributed this result to the decrease in the thickness of the boundary layers adjacent to the membrane Real water sample surfaces with increasing solution velocity. In most previous studies, the effect of some parameters Brackish water studied was sampled from the south of on ED process is determined by varying one parameter by Tunisia during February 2013. The physicochemical time, maintaining all the other parameters constant (Elmi- characteristics of the sample water are given in Table 1.It daoui et al. 2002; Kabay et al. 2008; Ben Sik Ali et al. is a brackish water of low salinity (total dissolved salts -1 2010a, b). Then the best value achieved by this procedure (TDS)\3000 mg L ). The fluoride concentration largely -1 is fixed and other parameters are varied by time. The dis- exceeds 1.5 mg L , the recommended value by World advantage of this univariate procedure is that the best Health Organization (WHO). Moreover, the recommended -1 -1 conditions cannot be attained, because the interaction values of 400 mg L for sulphate and 250 mg L for effects between the parameters are discarded. Moreover, chloride are also exceeded. Therefore, this water is not conventional methods are time consuming and require a suitable for drinking. large number of experiments to determine the optimum conditions. These drawbacks of the conventional methods Electrodialysis equipment and membranes can be eliminated by studying the effect of all parameters using a factorial design. In fact, this methodology deter- The ED setup consists of a power DC, a concentrate mines which factors have significant effects on a response reservoir, a dilute reservoir, a rinsing electrode reservoir as well as how the effect of one factor varies according to and three pumps (Heidolph D-93309) equipped each with a the level of the other factors (Meski et al. 2011; Balbasi flow-meter (PC Cell GmbH) and three valves to control the 2013; Azza et al. 2015). Its most important advantages are feed flow rate in the compartment of ED cell. Figure 1 not only the effects of individual parameters but also the shows a simplified scheme of ED setup working in batch interaction of two or more variables can also be derived recirculation mode. (Montgomery 2001). This is not possible in a classical one The ED cell was a PC Cell ED 64-004 (Germany) used factor at a time of experiment. as a conventional electrodialysis unit with two compart- Therefore, the aim of this work is to study the perfor- ments: the dilute and the concentrate. ED cell was made by mance of the ED process on brackish water desalination. two polypropylene blocks supporting electrodes. One To be made, a 2 full factorial design was used to inves- electrode was made of Pt/Ir-coated Ti stretched (anode) tigate the effects of operating parameters (applied poten- and the other of Ti stretched metal (cathode). The mem- tial, flow rate and salt concentration) on the ED efficiency. branes and spacers were stacked between the two elec- This efficiency is evaluated by the demineralization rate trode-end blocks. The ED stack was formed by 10 (DR) and the specific power consumption (SPC). Experi- repeating sections called cell pairs. A cell pair consists of ments were carried out with synthetic solutions of NaCl at the following: 123 Appl Water Sci (2017) 7:4563–4572 4565 Table 1 Physicochemical characteristics of brackish water sample separately. For each membrane, the active surface area was 64 cm . The flow channel width between two membranes Physicochemical Parameters Brackish Recommended was 0.5 mm determined by the thickness of intermembrane water values by WHO sample (Fewtrell and spacers. The main characteristics of used membranes are Bartram 2008) given in Table 2, which were supported by the manufac- -1 turer. The stack was equipped with three separate external Conductivity at 25 C (mS cm ) 2.5 0.5 plastic reservoirs: the first served to concentrate solution, pH 7.2 6.5–8.5 -1 the second to dilute solution and the third to rinse electrode TDS (mg L ) 2133 500 - -1 solution. The fluid circulation was achieved using three Cl (mg L ) 300 250 pumps equipped with flow-meters. Experiments were per- - -1 HCO (mg L ) 197 – formed in batch recirculation mode at room temperature. 2- -1 SO (mg L ) 917.46 400 - -1 F (mg L ) 3.66 1.5 Experimental procedure - -1 NO (mg L ) 54.2 50 ? -1 K (mg L ) 8.075 12 During all experiments, the volume of dilute, concentrate ? -1 Na (mg L ) 292.4 250 and rinsing electrode solutions was 1 L each. 0.1 M Na 2- 2? -1 Ca (mg L ) 188.8 – SO was used as electrode rinse solution circulating in 2? -1 Mg (mg L ) 211.9 – electrode compartment, to prevent the generation of chlo- rine or hypochlorite, which could be hazardous for the electrodes. Flow rate of electrode rinse solution was fixed -1 at 100 L h for all experiments. However, the dilute flow -1 rate solution was varied between 20 and 90 L h and the -1 concentrate one was fixed at 90 L h for all experiments. Before the onset of the desalination test, NaCl aqueous solution at the same concentration was introduced in dilute and concentrate compartments. The experiment started at time of the potential application, which was varied between 5 and 12 V. For these potentials, the ED system operates under the limiting current. Ionic conductivity was recorded in time. It was measured using a consort D 292 conduc- tivity meter equipped with a D292 conductivity cell. Prior to ED experiment, the conductivity cell was calibrated at 298 K with KCl standard solution at 0.01 and 0.1 M of 1.4 -1 and 12.67 mS cm , respectively (cell con- -1 stant = 0.5 cm ). Dilute and concentrate solutions were circulated through the ED cell until the desired product -1 conductivity (&0.5 mS cm ) was achieved in the dilute one. This value is equivalent to the good quality water. After every experiment, ED cell was cleaned with circu- lation of 0.1 M HCl solution during 15 min to remove any deposits, followed by circulation of distilled water. Fig. 1 Scheme of the ED installation Analytical method • cation exchange membrane (PC-SK); ? ? Na and K were analyzed by atomic emission spec- • dilute flow spacer (0.5 mm); troscopy using a ‘‘Sherwood 410’’ spectrophotometer. • anion exchange membrane (PC-SA); 2? 2? Ca and Mg amounts were determined using a con- • concentrate flow spacer (0.5 mm). ventional colorimetric EDTA titration. HCO was deter- Spacers were made in plastic and were placed between mined using a conventional colorimetric sulphuric acid the membranes to form the flow paths of the dilute and (H SO ) titration. Nitrate concentration was measured by 2 4 concentrate streams. The spacers were designed to mini- UV spectrophotometric method. Chloride analysis was mize boundary layer effects and were arranged in the stack measured by potentiometric titration using an automatic so that all the dilute and concentrate streams are manifold titrator (Metrohm 809). Sulphate concentration was 123 r 4566 Appl Water Sci (2017) 7:4563–4572 Table 2 Characteristics of the PC cell standard cation and anion exchange membranes Membranes Thickness Ion exchange capacity Chemical stability Permselectivity Functional Membrane resistance -1 2 (lm) (meq g ) (pH) groups (X cm ) PC-SK 130 &1 0–11 0.96 –SO 0.75–3 PC-SA 90–130 &1.5 0–9 0.93 –NH 1–1.5 determined by gravimetric analysis using BaCl in acidi- combinations of the experimental parameters levels. This fied medium. Fluoride concentration was determined using statistical design methodology allows measuring not only ion selective electrode (ISE 6.0502.150 fluoride ion elec- the main effect of each parameter, but also the interaction trode) in conjunction with a standard reference electrode effect among all the parameters. The determination of connected to a Metrohm 781 pH/Ion-meter. To avoid interaction effects of parameters may be important for possible interference resulting from changes in solution pH successful system optimization (Montgomery 2001). and conductivity, a total ionic strength adjustment buffer Today, the most used experimental design is the 2 facto- (TISAB) solution was used. It contained 58 g of NaCl and rial designs, where each variable is investigated at two 57 mL of glacial acetic acid and their pH was regulated at levels. 5.5 value using NaOH. The fluoride samples and the flu- In this study, a 2 full factorial design was carried out to oride standard were diluted by addition of TISAB solution investigate the performance of the ED process to reduce with a molar ratio of 1:1. pH meter (consort D 291) was salt concentration from brackish water. Initial salt con- used for measuring pH solutions. centration (C), dilute feed flow rate (Q) and applied potential (E) were chosen as a relevant parameters for ED Data analysis optimization. The responses were expressed in terms of percent of demineralization rate (DR) and specific power To investigate the influence of the salt concentration, consumption (SPC). Operating parameters, experimental applied potential and flow rate on the ED efficiency, the range and coded levels are given in Table 3. DR was calculated after 12 min of ED application using A total of 12 experiments were performed according to a the following equation (Elmidaoui et al. 2001): two level-three factor (2 ) full factorial (8 points of the factorial design and 4 center points to establish the DR (%Þ¼ 100 1  ; ð1Þ experimental errors). The chosen variables for this work were set at two levels and coded as (?1) and (-1) for high -1 where S (mg L ) is the salinity in the dilute compartment and low level, respectively. Since interactions between -1 and S (mg L ) is the initial salinity in the feed phase. The 0 these factors could be important, a linear polynomial model salinity was calculated from conductivity (Rodier et al. with first order was postulated by the following equation 2009). Eq. (3): The specific power consumption (SPC) is also an Y ¼ b0 þ b E þ b C þ b Q þ b E:C þ b E:Q important parameter of electrodialytic desalination. It can 1 2 3 12 13 þ b C:Q þ b E:C:Q; ð3Þ be described as the energy needed to treat unit volume of 23 123 solution. The SPC was calculated for each experimental where Y is the response, b is the constant term, b , b and 0 1 2 condition using the following equation (Kabay et al. 2008). b , are the linear coefficients which indicate the effect of E ItðÞdt applied potential (E), salt concentration (C) and flow rate SPC ¼ ; ð2Þ V (Q), respectively. Coefficients b , b , b describe the D 12 13 23 interacting effects of applied potential-salt concentration, where E is the applied potential, I is the current, V is the applied potential-flow rate and salt concentration-flow rate. volume of dilute stream and t is the time. Coefficient b implies the interacting effect of applied potential-salt concentration-flow rate, while the E, C and Statistical method Q are the independent coded variables (Turan et al. 2011). The analysis of experimental results was achieved with Factorial design determines the effect of multiple variables statistical and graphical analysis software (Minitab Release on a specific response and it can be used to reduce the 16, 2006). This software was used for regression analysis number of experiments in which multiple factors must be of the data obtained and to estimate the coefficients of investigated simultaneously (Montgomery 2001). In regression equations. experimental design, responses are measured at all 123 Appl Water Sci (2017) 7:4563–4572 4567 Table 3 Experimental range and levels of independent variables The p value is the probability value that is used to determine the effects in the model that are statistically Variable real values of coded levels significant. The significance of the data is judged by its Low (-1) Central point (0) High (?1) p value being closer to zero. For a 95% confidence level the p value should be less than or equal to 0.05 for the effect to E (V) 5 8.5 12 -1 be statistically significant (Alimi et al. 2014). The Pareto C (g L ) 1 5.5 10 -1 plot presents the absolute values of the effects of main Q (L h )20 55 90 factors and the effects of interaction of factors. A reference line is drawn to indicate that factors which extend past this Results and discussions line are potentially important (Antony 2003). The effects that are above the reference line are statistically significant Statistical analysis and modeling at 95% confidence level. It can be seen from Figs. 2 and 3 that applied potential had the greatest effect on the DR and A series of experiments were conducted by considering the SPC. 2 full factorial design. Table 4 presents the experimental Based on data presented in Table 5 and graphical Pareto responses measured at two levels of the studied parameters. chart in Fig. 2, the effect of interaction of two factors As shown by Table 4, the best combination of the fac- which were statistically insignificant was discarded. The tors for the highest demineralization rate occurs at run 6 final empirical model for DR in term of coded parameters where a higher applied potential, a lower salt concentration is given by Eq. (6): and a higher flow rate are used. This result agrees with that DR ¼ 47:06 þ 20:92E  13:05C þ 4:17Q þ 3:51E:C:Q obtained in the previous studies (Kabay et al. 2002; ð6Þ Banasiak et al. 2007; Kabay et al. 2008; Shady et al. 2012). Concerning the SPC, the values varied between 0.4 and And based on data presented in Table 6 and graphical -1 15.96 Wh L . The lowest value of the SPC was obtained Pareto chart in Fig. 3, the final empirical model for SPC in during the runs 3 and 5. The increase of salt concentration term of coded variables is given by Eq. (7): and applied potential determined an increase of SPC. SPC ¼ 5:4437 þ 4:9062E þ 2:2063C  0:3862Q Similar result was observed by Ben Sik Ali et al. (2010a) þ 2:1438E:C  0:4237E:Q  0:1687C:Q and Kabay et al. (2002). Whereas a slight variation of SPC 0:2812E:C:Q ð7Þ was observed when flow rate varied from low to high value. This result was in accordance with those of Kabay The goodness of fit of the model was evaluated by the et al. (2002) which have reported that there is no any 2 coefficient of determination (R ). The determination of considerable effect of flow rate on the SPC. 2 very useful R is allowed by calculation of the ratio of the A linear regression model was fitted for the experi- sum of squares of the predicted responses to the sum of mental data using the Minitab statistical software. It was squares of the observed responses (Srinivasan and used to investigate the main effects of factors, the inter- 2 Viraraghavan 2010). It is suggested that R should be actions, the coefficient standard deviations and various close to 1 for a good fit model (Boubakri et al. 2013). The statistical parameters of the fitted models. These parame- estimated model for both DR and SPC had satisfactory R ters, for each response (DR and SPC) are shown in more than 99%. In the case of DR, fitting is very good Tables 5 and 6. (R = 99.75%) and only 0.25% of total variance was not The effect is the difference between the responses of two 2 explained by the model. For the SPC (R = 99.99%), levels (high and low level) of factors; the regression model which presents a high value and only 0.01% of a total coefficients are obtained by dividing the effects by two. variance was not explained by the model. The standardized effects (T) are obtained by dividing the regression coefficients by the standard error coefficient Main effects plot (Alimi et al. 2014). Substituting the coefficients b,in Eq. (3) by the respective values from Tables 5 and 6,we The main effects are shown in Figs. 4 and 5, for DR and get: SPC, respectively. It indicates the relative strength of DR ¼ 47:06 þ 20:92E  13:05C þ 4:17Q  0:73E:C effects of various factors. A main effect is present when the þ 0:5E:Q þ 0:67C:Q þ 3:51E:C:Q ð4Þ mean response changes across the level of a factor. The sign of the main effect indicates the direction of the effect SPC ¼ 5:4437 þ 4:9062E þ 2:2063C  0:3862Q (Srinivasan and Viraraghavan 2010). þ 2:1438E  0:4237EQ  0:1687C:Q As shown in Fig. 4, the potential had a positive effect on 0:2812E:C:Q ð5Þ desalination efficiency. In fact an increase of applied 123 4568 Appl Water Sci (2017) 7:4563–4572 Table 4 Full factorial design matrix for desalination efficiency -1 -1 -1 Run order E (V) C (g L ) Q (L h ) DR (%) SPC (Wh L ) 1 5 1 20 31.95 0.55 2 12 1 20 81.26 6.36 3 5 10 20 13 0.45 4 12 10 20 45.35 15.96 5 5 1 90 44.95 0.4 6 12 1 90 82.24 5.64 7 5 10 90 14.64 0.75 8 12 10 90 63.05 13.44 9 8.5 5.5 55 64.39 5.35 10 8.5 5.5 55 61.71 5.18 Fig. 2 Pareto chart for standardized effects for DR 11 8.5 5.5 55 59.09 5.44 12 8.5 5.5 55 62.3 5.35 Table 5 Estimated effects and coefficients for DR (coded units) Term Effect Coefficient Tp value Constant 47.06 60.98 0.000 E 41.84 20.92 27.11 0.000 C -26.09 -13.05 -16.91 0.000 Q 8.33 4.17 5.40 0.012 E.C -1.46 -0.73 -0.95 0.414 E.Q 1.01 0.50 0.65 0.560 C.Q 1.34 0.67 0.87 0.449 E.C.Q 7.02 3.51 4.55 0.020 Fig. 3 Pareto chart for standardized effects for SPC Standard error coefficient for all cases = 0.7716 hydrodynamic and electrical conditions an increase of the R = 0.9975 initial salt concentration leads to a decrease of the DR. This result can be explained by the concentration polarization phenomenon which is more important at high concentra- Table 6 Estimated effects and coefficients for SPC (coded units) tion (Sadrzadeh and Mohammadi 2008). As demonstrated Term Effect Coefficient Tp value in previous studies (Kabay et al. 2002; Banasiak et al. 2007; Sadrzadeh and Mohammadi 2008; Ben Sik Ali et al. Constant 5.4437 141.74 0.000 2010a), the number of ions transported through the mem- E 9.8125 4.9062 127.75 0.000 branes are almost the same but total amounts of salts are C 4.4125 2.2063 57.45 0.000 quite different from the different treated solution. As Q -0.7725 -0.3862 -10.06 0.002 known, the calculation of DR depends strongly on the E.C 4.2875 2.1438 55.82 0.000 initial feed concentration and the amount of transported E.Q -0.8475 -0.4237 -11.03 0.002 ions. So, the DR evolves reciprocally to the initial feed C.Q -0.3375 -0.1687 -4.39 0.022 concentration at some hydrodynamic and electrical condi- E.C.Q -0.5625 -0.2812 -7.32 0.005 tions. Increasing concentration from low to high level Standard error coefficient for all cases = 0.03841 resulted in 26% decrease in DR (Fig. 4). R = 0.9999 At high flow rate, the increase in the DR with flow rate may be attributed to the decrease in the thickness of the boundary layers adjacent to the membranes surfaces with potential from low to high level resulted in increasing DR increasing solution velocity. In the present case, Fig. 4 by 41.84%. shows a slight increase (8.33%) in DR when flow rate At higher salt concentration values, it can be observed increases from low to high level likely because the thick- that the DR has a considerable dependence on the feed -1 ness of the boundary layers adjacent to the membranes solution in the range of 5.5–10 g L . Effectively at some 123 Appl Water Sci (2017) 7:4563–4572 4569 Fig. 4 Main effects plot for DR Fig. 6 Interaction effects plot for DR Fig. 5 Main effects plot for SPC Fig. 7 Interaction effects plot for SPC surfaces does not change significantly when flow rates vary concentration. But a negative interactive effect was -1 from 20 to 90 L h . Similar results were demonstrated by observed between applied potential and flow rate as well as Kabay et al. (2002). between salt concentration and flow rate. An increase of -1 In the case of SPC, applied potential and the initial salt the concentration value from 1 to 10 g L increased the concentration have a positive effect on this response. But -1 -1 SPC by 11 Wh L (from 4 to 15 Wh L )at12 V. -1 the flow rate had a slight negative one. As a general trend, Increasing the flow rate from 20 to 90 L h enhances the an increase of applied potential and initial salt concentra- -1 -1 decrease of SPC by 2 Wh L (from 12 to 10 Wh L )at tion from low to high level resulted in increasing SPC by 12 V. Also, the increase of flow rate from low to high level -1 9.81 and 4.41 Wh L , respectively. -1 decreases the SPC value by 1 Wh L (from 8 to -1 -1 7WhL )at10gL . Interaction effects plot Normal probability plot of residuals The interaction plot is a graphical tool which plots the mean response of two factors at all possible combination of One of the key of assumptions for the statistical analysis of their settings. If the lines are non-parallel, it is an indication data from experiments is that the data that come from a of interaction between the two factors (Antony 2003). normal distribution (Antony 2003). The normality of the Parallel lines indicate that there is no interaction between data can be checked by plotting a normal probability plot two factors. The interaction effect plots are shown in of the residuals. If the points on the plot fall fairly close to a Figs. 6 and 7, for DR and SPC, respectively. straight line, the data are normally distributed (Antony In the case of DR, there are no significant interactions 2003). The normal probability plot of the residuals with a between all factors. In the case of SPC, Fig. 7 shows 95% confidence level for DR and SPC are shown in Figs. 8 positive interaction between applied potential and salt and 9. It can be seen that for DR and SPC, the experimental 123 4570 Appl Water Sci (2017) 7:4563–4572 Fig. 8 Normal probability plot of the residuals for DR Fig. 9 Normal probability plot of the residuals for SPC -1 -1 points fall fairly close to the straight line. Therefore, the 12 V at 90 L h or 18 V at 40 L h , the value of limit data from the experiments come from a normally dis- current density was not reached. pH variation due to the ? - tributed population, and they were reliable. reaction of water dissociation into H O and OH is then avoided and this limits the probability of fouling and/or Treatment of the real water sample scaling formation. Desalination of brackish water was achieved and the Finally, the application of electrodialysis was performed on concentrations of different species in the obtained treated the real brackish water (Table 1) using the optimal water are below the amount recommended by WHO. An parameters. The flow rate and applied potential were fixed 85.5% of DR was obtained after 24 min of ED application -1 -1 at 90 L h and 12 V, respectively. The physicochemical with 14.76 Wh L of SPC for E = 18 V and -1 characteristics of treated water are given in Table 8. Then, Q = 40 L h . Whereas, DR tends to 84% obtained after -1 the results were compared with those obtained using clas- 27 min of ED application with 6.72 Wh L of SPC for -1 -1 sical method of optimization (E = 18 V, Q = 40 L h ) E = 12 V and Q = 90 L h . So, we can clearly observe (Table 7). the advantage of full factorial design which manifests in As shown in Fig. 10 which describes the polarization decreasing SPC. curve for the real water sample, for the applied potential of 123 Appl Water Sci (2017) 7:4563–4572 4571 Table 7 Optimization of factors influencing the ED efficiency by the classical method optimization -1 -1 Effect of applied potential for C = 3gL and Q = 40 L h E (V) DR (%) 10 57.37 15 81.81 18 86.88 -1 Effect of flow rate for C = 3gL and E = 18 V -1 Q (L h ) DR (%) 20 80 30 90 40 90 50 89.47 Fig. 10 Polarization curves, I = f (E) -1 Effect of salt concentration for E = 18 V and Q = 40 L h -1 C (g L ) DR (%) 1 81.48 applied potential and feed flow rate were performed to 1.5 73.91 optimize the demineralization rate and the specific power consumption. 2 73.44 2.5 60 The applied potential and the salt concentration have a significant effect on the process efficiency and mainly on demineralization rate. It was also found that the decrease of salt concentration induces better performance. On the other hand, the specific power consumption was mostly influenced by initial salt concentration and applied Table 8 Physicochemical characteristics of treated water potential. The significant interactions found are between Physicochemical Sample treated Sample treated Recommended applied potential and salt concentration for the SPC. The characteristics at E = 12 V, at E = 18 V, values by WHO factorial experiment design method is undoubtedly a good -1 -1 Q = 90 L h Q = 40 L h technique for studying the influence of major process Conductivity 0.5 0.5 0.5 parameters on response factors by significant reducing the -1 (mS cm ) number of experiment and henceforth, saving time, pH 6.8 6.7 6.5–8.5 energy and money. During this study we were able to -1 TDS (mg L ) 340 308 500 obtain high values of demineralization rate going to - -1 Cl (mg L ) 94.66 44.37 250 82.24%. HCO 00 – 3 Electrodialysis process was applied for the treatment of -1 (mg L ) real brackish water sample. The concentrations of different 2- -1 SO (mg L ) 152.44 80.75 400 species in the obtained treated water are below the amounts - -1 F (mg L ) 0.42 1.05 1.5 recommended by World Health organization for drinking - -1 NO (mg L ) 0.8 0.81 50 water. ? -1 K (mg L ) 7 0.78 12 ? -1 Na (mg L ) 20 21.44 250 Acknowledgements The funding was provided by certe (Grant no. 2? -1 ?216 79325122). Ca (mg L ) 37.33 20 – 2? -1 Mg (mg L ) 32 39.6 – Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http:// creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give Conclusions appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The results from the presented study showed that the desalination of brackish water using electrodialysis process is depending on several parameters. A series of full fac- Publisher’s Note Springer Nature remains neutral with regard to torial design experiments varying initial salt concentration, jurisdictional claims in published maps and institutional affiliations. 123 4572 Appl Water Sci (2017) 7:4563–4572 McGovern RK, Weiner AM, Sun L, Chambers CG, Zubair SM, References Lienhard JH (2014a) On the cost of electrodialysis for the desalination of high salinity feeds. Appl Energy 136:649–661 Alimi F, Boubakri A, Tlili MM, Amor MB (2014) A comprehensive McGovern RK, Zubair SM, Lienhard VJH (2014b) Hybrid electro- factorial design study of variables affecting CaCO scaling under dialysis reverse osmosis system design and its optimization for magnetic water treatment. Water Sci Technol 70:1355–1362. treatment of highly saline brines. J Desalt Water Reuse 6:15–23 doi:10.2166/wst.2014.377 Meski S, Ziani S, Khireddine H, Boudboub S, Zaidi S (2011) Antony J (2003) Design of experiments for engineers and scientists. Factorial design analysis for sorption of zinc on hydroxyapatite. Butterworth-Heinemann, New York J Hazard Mater 186:1007–1017 Azza A, Hamdon A, Darwish NAD, Hilal N (2015) The use of Monohar M, Das AK, Shahi VK (2017) Alternative preparative route factorial design in the analysis of air-gap membrane distillation for efficient and stable anion exchange membrane for water data. Desalination 367:90–102 desalination by electrodialysis. Desalination 413:101–108 Balbasi M (2013) Application of full factorial design method to Montgomery DC (2001) Design and analysis of experiments, 5th edn. silicate synthesis. Mater Res Bull 48:2908–2914 Wiley, New York Banasiak LJ, Kruttschnitt TW, Schafer AI (2007) Desalination using Nemati M, Hosseini SM, Shabanian M (2017) Novel electrodialysis electrodialysis as a function of voltage and salt concentration. cation exchange membrane prepared by 2-acrylamido-2-methyl- Desalination 205:38–46 propane sulfonic acid; heavy metal ions removal. J Hazard Mater Ben Sik Ali M, Mnif A, Hamrouni B, Dhahbi M (2010a) Electro- 337:90–104 dialytic desalination of brackish water: effect of process Reig M, Casas S, Aladjem C, Valderrama C, Gibert O, Valero F, parameters and water characteristics. Ionics 16:621–629 Centeno CM, Larrotcha E, Cortina JL (2014) Concentration of Ben Sik Ali M, Hamrouni B, Dhahbi M (2010b) Electrodialytic NaCl from seawater reverse osmosis brines for the chlor-alkali defluoridation of brackish water: effect of process parameters industry by electrodialysis. Desalination 342:107–117 and water characteristics. Clean-Soil Air Water 38(7):623–629 Reig M, Casas S, Valderrama C, Gibert O, Cortina JL (2016a) Boubakri A, Helali N, Tlili M, Amor MB (2013) Fluoride removal Integration of monopolar and bipolar electrodialysis for val- from diluted solutions by Donnan dialysis using full factorial orization of seawater reverse osmosis desalination brines: design. Korean J Chem Eng 10:1–6. doi:10.1007/s11814-013- production of strong acid and base. Desalination 398:87–97 0263-9 Reig M, Casas S, Gibert O, Valderrama C, Cortina LJ (2016b) Elisseeva TV, Shaposhnik VA, Luschik IG (2002) Demineralization Integration of nanofiltration and bipolar electrodialysis for and separation of amino acids by electrodialysis with ion- valorization of seawater desalination brines: production of exchange membrane. Desalination 149:405–409 drinking and waste water treatment chemicals. Desalination Elmidaoui A, Elhannouni F, Sahli MAM, Chay L, Elabbassi H, Hafsi 382:13–20 M, Largeteau D (2001) Pollution of nitrate in Moroccan ground Rodier J, Legube B, Merlet N (2009) The analysis of water, 9th edn. water: removal by electrodialysis. Desalination 136:325–332 Dunod, Paris, pp 78–81 Elmidaoui A, Elhanouni F, Taky M, Chay L, sahli MA, Echihabi L, Sadrzadeh M, Mohammadi T (2008) Sea water desalination using Hafsi M (2002) Optimization of nitrate removal operation from electrodialysis. Desalination 221:440–447 ground water by electrodialysis. Sep Purif Technol 29:235–244 2? Shady AA, Peng C, Bi J, Xu H, Almeria J (2012) Recovery of Pb Fewtrell L, Bartram J (2008) Guidelines for drinking water quality, and removal of NO from aqueous solutions using integrated 3rd edn. World Health Organization, Geneva, pp 375–492 electrodialysis, electrolysis and adsorption process. Desalination Fidaleo M, Moresi M (2005) Optimal strategy to model the 286:304–315 electrodialytic recovery of a strong electrolyte. J Membr Sci Srinivasan A, Viraraghavan T (2010) Oil removal from water by 260:90–111 fungal biomass: a factorial design analysis. J Hazard Mater Galama AH, Saakes M, Bruning H, Rijnaarts HHM, Post JW (2014) 175:695–702 Seawater predesalination with electrodialysis. Desalination Tanaka Y, Reig M, Casas S, Aladjem C, Cortina JL (2015) Computer 342:61–69 simulation of ion-exchange membrane electrodialysis for salt Ghyselbrecht K, Huygebaert M, Bruggen BV, Ballet R, Meesschaert concentration and reduction of RO discharged brine for salt B, Pinoy L (2013) Desalination of an industrial saline water with production and marine environment conservation. Desalination conventional and bipolar membrane electrodialysis. Desalination 367:76–89 318:9–18 Turan NG, Elevli S, Mesci B (2011) Adsorption of copper and Zinc Jing G, Du W, Guo Y (2012) Studies on prediction of separation ions on illite: determination of the optimal conditions by the percent in electrodialysis process via BP neural networks and statistical design of experiments. Appl Clay Sci 52:392–399 improved BP algorithms. Desalination 291:78–93 ´ ´ Walha K, Amar RB, Firdaous L, Quemeneur F, Jaouen P (2007) Kabay N, Demircioglu M, Ersoz E, Kurucaovali I (2002) Removal of Brackish groundwater treatment by nanofiltration, reverse calcium and magnesium hardness by electrodialysis. Desalina- osmosis and electrodialysis in Tunisia: performance and cost tion 149:343–349 comparison. Desalination 207:95–106 Kabay N, Arar O, Samatya S, Yuksel U, Yuksel M (2008) Separation Zourmand Z, Faridirad F, Kasiri N, Mohammadi T (2015) Mass of fluoride from aqueous solution by electrodialysis: effect of transfer modeling of desalination through an electrodialysis cell. process parameters and other ionic species. J Hazard Mater Desalination 359:41–51 153:107–113 Lee HJ, Strathmann H, Moon SH (2006) Determination of the limiting current density in electrodialysis desalination as an empirical function of linear velocity. Desalination 190:43–50 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Applied Water Science Springer Journals

Electrodialytic desalination of brackish water: determination of optimal experimental parameters using full factorial design

Download PDF
10 pages

Loading next page...
1
 
/lp/springer_journal/electrodialytic-desalination-of-brackish-water-determination-of-0o0B4nfpac

References (43)

Publisher
Springer Journals
Copyright
Copyright © 2017 by The Author(s)
Subject
Earth Sciences; Hydrogeology; Water Industry/Water Technologies; Industrial and Production Engineering; Waste Water Technology / Water Pollution Control / Water Management / Aquatic Pollution; Nanotechnology; Private International Law, International & Foreign Law, Comparative Law
ISSN
2190-5487
eISSN
2190-5495
DOI
10.1007/s13201-017-0609-2
Publisher site
See Article on Publisher Site

Abstract

Appl Water Sci (2017) 7:4563–4572 https://doi.org/10.1007/s13201-017-0609-2 ORIGINAL ARTICLE Electrodialytic desalination of brackish water: determination of optimal experimental parameters using full factorial design 1 1 2 1 • • • • Soumaya Gmar Nawel Helali Ali Boubakri Ilhem Ben Salah Sayadi 1 1 Mohamed Tlili Mohamed Ben Amor Received: 22 March 2017 / Accepted: 29 August 2017 / Published online: 15 September 2017 The Author(s) 2017. This article is an open access publication Abstract The aim of this work is to study the desalination Introduction of brackish water by electrodialysis (ED). A two level- three factor (2 ) full factorial design methodology was used The shortage of drinking water is a major problem in the to investigate the influence of different physicochemical southern communities of Tunisia. In these regions, the parameters on the demineralization rate (DR) and the traditional sources of fresh water are insufficient to meet specific power consumption (SPC). Statistical design the demand and are being stressed by competing uses, such determines factors which have the important effects on ED as irrigation and industrial needs (Walha et al. 2007). performance and studies all interactions between the con- However, in recent times and with the development of sidered parameters. Three significant factors were used desalination processes, there has been an increasing interest including applied potential, salt concentration and flow in using brackish waters as a source of potable water. rate. The experimental results and statistical analysis show Desalination is a process that removes dissolved minerals that applied potential and salt concentration are the main from seawater or brackish water or treated waste water. effect for DR as well as for SPC. The effect of interaction These processes create more valuable water by converting between applied potential and salt concentration was saline waters into a resource (Reig et al. 2014; McGovern observed for SPC. A maximum value of 82.24% was et al. 2014a; Tanaka et al. 2015). Several desalination obtained for DR under optimum conditions and the best methods have been developed to obtain fresh drinking -1 value of SPC obtained was 5.64 Wh L . Empirical water. There are mainly two families of desalination regression models were also obtained and used to predict technologies used throughout the world today. They the DR and the SPC profiles with satisfactory results. The include thermal (evaporative) and membrane technologies. process was applied for the treatment of real brackish water Membrane methods are less energy intensive than thermal using the optimal parameters. methods and since energy consumption directly affects the cost effectiveness and feasibility of using desalination Keywords Demineralization rate  Desalination  technology, membrane methods such as reverse osmosis Electrodialysis  Full factorial design  Specific power (RO) and electrodialysis (ED) are attracting great attention consumption lately (Sadrzadeh and Mohammadi 2008; McGovern et al. 2014b). ED is a membrane process for the separation of ions across charged membranes from one solution to another under the influence of an electrical potential dif- & Soumaya Gmar ference used as a driving force (Sadrzadeh and Moham- soumayagmar@yahoo.com madi 2008). This process has been widely used to produce drinking Laboratory of Natural Water Treatment, Center of water from brackish and sea water (Fidaleo and Moresi Researches and Water Technologies, P.B 273, 8020 Soliman, 2005; Lee et al. 2006; Jing et al. 2012; Galama et al. 2014; Tunisia Zourmand et al. 2015; Reig et al. 2016a, b; Monohar et al. Laboratory of Waste Water Treatment, Center of Researches 2017). It has been also used in treatment of industrial and Water Technologies, P.B 273, 8020 Soliman, Tunisia 123 4564 Appl Water Sci (2017) 7:4563–4572 effluents (Ghyselbrecht et al. 2013), purification of amino different concentrations for checking the optimal condi- acids and other organic compounds (Elisseeva et al. 2002) tions. Then, real brackish water was treated by ED using and to remove heavy metals from waste water (Nemati the optimal conditions. et al. 2017). Many factors influence the ED performance such as applied potential (Elmidaoui et al. 2002; Banasiak et al. Materials and methods 2007), salt concentration (Banasiak et al. 2007; Shady et al. 2012), flow rate of dilute compartment and temperature Chemicals reagents (Sadrzadeh and Mohammadi 2008; Ben Sik Ali et al. 2010a; Shady et al. 2012). Analytical grade sodium chloride (NaCl) and sodium A literature survey revealed that DR was increased with sulphate Na SO are used to produce solution with 2 4 the applied potential but it was decreased at high values of known amounts of salts and electrode rinse solution, salt concentration (Banasiak et al. 2007; Sadrzadeh and respectively. KCl is used to calibrate conductivity cell. Mohammadi 2008; Shady et al. 2012). Sadrzadeh and Ethylene diamine tetraacetic acid (EDTA), sulphuric Mohammadi (2008) and Ben Sik Ali et al. (2010a) acid (H SO ,), barium chloride (BaCl ), hydrochloric 2 4 2 observed that the salt percent removal increases when the acid (HCl) sodium hydroxide (NaOH), sodium fluoride flow rates decrease. They suggested that for low flow rates (NaF) and glacial acid acetic (CH COOH) are used to the residence time of ions in the dilute compartment analyze the real water. Each solution was prepared using increases. On the other hand, Elmidaoui et al. (2002) distilled water. All chemicals were purchased from demonstrated that DR increased with increasing flow rates. Sigma-Aldrich. The authors attributed this result to the decrease in the thickness of the boundary layers adjacent to the membrane Real water sample surfaces with increasing solution velocity. In most previous studies, the effect of some parameters Brackish water studied was sampled from the south of on ED process is determined by varying one parameter by Tunisia during February 2013. The physicochemical time, maintaining all the other parameters constant (Elmi- characteristics of the sample water are given in Table 1.It daoui et al. 2002; Kabay et al. 2008; Ben Sik Ali et al. is a brackish water of low salinity (total dissolved salts -1 2010a, b). Then the best value achieved by this procedure (TDS)\3000 mg L ). The fluoride concentration largely -1 is fixed and other parameters are varied by time. The dis- exceeds 1.5 mg L , the recommended value by World advantage of this univariate procedure is that the best Health Organization (WHO). Moreover, the recommended -1 -1 conditions cannot be attained, because the interaction values of 400 mg L for sulphate and 250 mg L for effects between the parameters are discarded. Moreover, chloride are also exceeded. Therefore, this water is not conventional methods are time consuming and require a suitable for drinking. large number of experiments to determine the optimum conditions. These drawbacks of the conventional methods Electrodialysis equipment and membranes can be eliminated by studying the effect of all parameters using a factorial design. In fact, this methodology deter- The ED setup consists of a power DC, a concentrate mines which factors have significant effects on a response reservoir, a dilute reservoir, a rinsing electrode reservoir as well as how the effect of one factor varies according to and three pumps (Heidolph D-93309) equipped each with a the level of the other factors (Meski et al. 2011; Balbasi flow-meter (PC Cell GmbH) and three valves to control the 2013; Azza et al. 2015). Its most important advantages are feed flow rate in the compartment of ED cell. Figure 1 not only the effects of individual parameters but also the shows a simplified scheme of ED setup working in batch interaction of two or more variables can also be derived recirculation mode. (Montgomery 2001). This is not possible in a classical one The ED cell was a PC Cell ED 64-004 (Germany) used factor at a time of experiment. as a conventional electrodialysis unit with two compart- Therefore, the aim of this work is to study the perfor- ments: the dilute and the concentrate. ED cell was made by mance of the ED process on brackish water desalination. two polypropylene blocks supporting electrodes. One To be made, a 2 full factorial design was used to inves- electrode was made of Pt/Ir-coated Ti stretched (anode) tigate the effects of operating parameters (applied poten- and the other of Ti stretched metal (cathode). The mem- tial, flow rate and salt concentration) on the ED efficiency. branes and spacers were stacked between the two elec- This efficiency is evaluated by the demineralization rate trode-end blocks. The ED stack was formed by 10 (DR) and the specific power consumption (SPC). Experi- repeating sections called cell pairs. A cell pair consists of ments were carried out with synthetic solutions of NaCl at the following: 123 Appl Water Sci (2017) 7:4563–4572 4565 Table 1 Physicochemical characteristics of brackish water sample separately. For each membrane, the active surface area was 64 cm . The flow channel width between two membranes Physicochemical Parameters Brackish Recommended was 0.5 mm determined by the thickness of intermembrane water values by WHO sample (Fewtrell and spacers. The main characteristics of used membranes are Bartram 2008) given in Table 2, which were supported by the manufac- -1 turer. The stack was equipped with three separate external Conductivity at 25 C (mS cm ) 2.5 0.5 plastic reservoirs: the first served to concentrate solution, pH 7.2 6.5–8.5 -1 the second to dilute solution and the third to rinse electrode TDS (mg L ) 2133 500 - -1 solution. The fluid circulation was achieved using three Cl (mg L ) 300 250 pumps equipped with flow-meters. Experiments were per- - -1 HCO (mg L ) 197 – formed in batch recirculation mode at room temperature. 2- -1 SO (mg L ) 917.46 400 - -1 F (mg L ) 3.66 1.5 Experimental procedure - -1 NO (mg L ) 54.2 50 ? -1 K (mg L ) 8.075 12 During all experiments, the volume of dilute, concentrate ? -1 Na (mg L ) 292.4 250 and rinsing electrode solutions was 1 L each. 0.1 M Na 2- 2? -1 Ca (mg L ) 188.8 – SO was used as electrode rinse solution circulating in 2? -1 Mg (mg L ) 211.9 – electrode compartment, to prevent the generation of chlo- rine or hypochlorite, which could be hazardous for the electrodes. Flow rate of electrode rinse solution was fixed -1 at 100 L h for all experiments. However, the dilute flow -1 rate solution was varied between 20 and 90 L h and the -1 concentrate one was fixed at 90 L h for all experiments. Before the onset of the desalination test, NaCl aqueous solution at the same concentration was introduced in dilute and concentrate compartments. The experiment started at time of the potential application, which was varied between 5 and 12 V. For these potentials, the ED system operates under the limiting current. Ionic conductivity was recorded in time. It was measured using a consort D 292 conduc- tivity meter equipped with a D292 conductivity cell. Prior to ED experiment, the conductivity cell was calibrated at 298 K with KCl standard solution at 0.01 and 0.1 M of 1.4 -1 and 12.67 mS cm , respectively (cell con- -1 stant = 0.5 cm ). Dilute and concentrate solutions were circulated through the ED cell until the desired product -1 conductivity (&0.5 mS cm ) was achieved in the dilute one. This value is equivalent to the good quality water. After every experiment, ED cell was cleaned with circu- lation of 0.1 M HCl solution during 15 min to remove any deposits, followed by circulation of distilled water. Fig. 1 Scheme of the ED installation Analytical method • cation exchange membrane (PC-SK); ? ? Na and K were analyzed by atomic emission spec- • dilute flow spacer (0.5 mm); troscopy using a ‘‘Sherwood 410’’ spectrophotometer. • anion exchange membrane (PC-SA); 2? 2? Ca and Mg amounts were determined using a con- • concentrate flow spacer (0.5 mm). ventional colorimetric EDTA titration. HCO was deter- Spacers were made in plastic and were placed between mined using a conventional colorimetric sulphuric acid the membranes to form the flow paths of the dilute and (H SO ) titration. Nitrate concentration was measured by 2 4 concentrate streams. The spacers were designed to mini- UV spectrophotometric method. Chloride analysis was mize boundary layer effects and were arranged in the stack measured by potentiometric titration using an automatic so that all the dilute and concentrate streams are manifold titrator (Metrohm 809). Sulphate concentration was 123 r 4566 Appl Water Sci (2017) 7:4563–4572 Table 2 Characteristics of the PC cell standard cation and anion exchange membranes Membranes Thickness Ion exchange capacity Chemical stability Permselectivity Functional Membrane resistance -1 2 (lm) (meq g ) (pH) groups (X cm ) PC-SK 130 &1 0–11 0.96 –SO 0.75–3 PC-SA 90–130 &1.5 0–9 0.93 –NH 1–1.5 determined by gravimetric analysis using BaCl in acidi- combinations of the experimental parameters levels. This fied medium. Fluoride concentration was determined using statistical design methodology allows measuring not only ion selective electrode (ISE 6.0502.150 fluoride ion elec- the main effect of each parameter, but also the interaction trode) in conjunction with a standard reference electrode effect among all the parameters. The determination of connected to a Metrohm 781 pH/Ion-meter. To avoid interaction effects of parameters may be important for possible interference resulting from changes in solution pH successful system optimization (Montgomery 2001). and conductivity, a total ionic strength adjustment buffer Today, the most used experimental design is the 2 facto- (TISAB) solution was used. It contained 58 g of NaCl and rial designs, where each variable is investigated at two 57 mL of glacial acetic acid and their pH was regulated at levels. 5.5 value using NaOH. The fluoride samples and the flu- In this study, a 2 full factorial design was carried out to oride standard were diluted by addition of TISAB solution investigate the performance of the ED process to reduce with a molar ratio of 1:1. pH meter (consort D 291) was salt concentration from brackish water. Initial salt con- used for measuring pH solutions. centration (C), dilute feed flow rate (Q) and applied potential (E) were chosen as a relevant parameters for ED Data analysis optimization. The responses were expressed in terms of percent of demineralization rate (DR) and specific power To investigate the influence of the salt concentration, consumption (SPC). Operating parameters, experimental applied potential and flow rate on the ED efficiency, the range and coded levels are given in Table 3. DR was calculated after 12 min of ED application using A total of 12 experiments were performed according to a the following equation (Elmidaoui et al. 2001): two level-three factor (2 ) full factorial (8 points of the factorial design and 4 center points to establish the DR (%Þ¼ 100 1  ; ð1Þ experimental errors). The chosen variables for this work were set at two levels and coded as (?1) and (-1) for high -1 where S (mg L ) is the salinity in the dilute compartment and low level, respectively. Since interactions between -1 and S (mg L ) is the initial salinity in the feed phase. The 0 these factors could be important, a linear polynomial model salinity was calculated from conductivity (Rodier et al. with first order was postulated by the following equation 2009). Eq. (3): The specific power consumption (SPC) is also an Y ¼ b0 þ b E þ b C þ b Q þ b E:C þ b E:Q important parameter of electrodialytic desalination. It can 1 2 3 12 13 þ b C:Q þ b E:C:Q; ð3Þ be described as the energy needed to treat unit volume of 23 123 solution. The SPC was calculated for each experimental where Y is the response, b is the constant term, b , b and 0 1 2 condition using the following equation (Kabay et al. 2008). b , are the linear coefficients which indicate the effect of E ItðÞdt applied potential (E), salt concentration (C) and flow rate SPC ¼ ; ð2Þ V (Q), respectively. Coefficients b , b , b describe the D 12 13 23 interacting effects of applied potential-salt concentration, where E is the applied potential, I is the current, V is the applied potential-flow rate and salt concentration-flow rate. volume of dilute stream and t is the time. Coefficient b implies the interacting effect of applied potential-salt concentration-flow rate, while the E, C and Statistical method Q are the independent coded variables (Turan et al. 2011). The analysis of experimental results was achieved with Factorial design determines the effect of multiple variables statistical and graphical analysis software (Minitab Release on a specific response and it can be used to reduce the 16, 2006). This software was used for regression analysis number of experiments in which multiple factors must be of the data obtained and to estimate the coefficients of investigated simultaneously (Montgomery 2001). In regression equations. experimental design, responses are measured at all 123 Appl Water Sci (2017) 7:4563–4572 4567 Table 3 Experimental range and levels of independent variables The p value is the probability value that is used to determine the effects in the model that are statistically Variable real values of coded levels significant. The significance of the data is judged by its Low (-1) Central point (0) High (?1) p value being closer to zero. For a 95% confidence level the p value should be less than or equal to 0.05 for the effect to E (V) 5 8.5 12 -1 be statistically significant (Alimi et al. 2014). The Pareto C (g L ) 1 5.5 10 -1 plot presents the absolute values of the effects of main Q (L h )20 55 90 factors and the effects of interaction of factors. A reference line is drawn to indicate that factors which extend past this Results and discussions line are potentially important (Antony 2003). The effects that are above the reference line are statistically significant Statistical analysis and modeling at 95% confidence level. It can be seen from Figs. 2 and 3 that applied potential had the greatest effect on the DR and A series of experiments were conducted by considering the SPC. 2 full factorial design. Table 4 presents the experimental Based on data presented in Table 5 and graphical Pareto responses measured at two levels of the studied parameters. chart in Fig. 2, the effect of interaction of two factors As shown by Table 4, the best combination of the fac- which were statistically insignificant was discarded. The tors for the highest demineralization rate occurs at run 6 final empirical model for DR in term of coded parameters where a higher applied potential, a lower salt concentration is given by Eq. (6): and a higher flow rate are used. This result agrees with that DR ¼ 47:06 þ 20:92E  13:05C þ 4:17Q þ 3:51E:C:Q obtained in the previous studies (Kabay et al. 2002; ð6Þ Banasiak et al. 2007; Kabay et al. 2008; Shady et al. 2012). Concerning the SPC, the values varied between 0.4 and And based on data presented in Table 6 and graphical -1 15.96 Wh L . The lowest value of the SPC was obtained Pareto chart in Fig. 3, the final empirical model for SPC in during the runs 3 and 5. The increase of salt concentration term of coded variables is given by Eq. (7): and applied potential determined an increase of SPC. SPC ¼ 5:4437 þ 4:9062E þ 2:2063C  0:3862Q Similar result was observed by Ben Sik Ali et al. (2010a) þ 2:1438E:C  0:4237E:Q  0:1687C:Q and Kabay et al. (2002). Whereas a slight variation of SPC 0:2812E:C:Q ð7Þ was observed when flow rate varied from low to high value. This result was in accordance with those of Kabay The goodness of fit of the model was evaluated by the et al. (2002) which have reported that there is no any 2 coefficient of determination (R ). The determination of considerable effect of flow rate on the SPC. 2 very useful R is allowed by calculation of the ratio of the A linear regression model was fitted for the experi- sum of squares of the predicted responses to the sum of mental data using the Minitab statistical software. It was squares of the observed responses (Srinivasan and used to investigate the main effects of factors, the inter- 2 Viraraghavan 2010). It is suggested that R should be actions, the coefficient standard deviations and various close to 1 for a good fit model (Boubakri et al. 2013). The statistical parameters of the fitted models. These parame- estimated model for both DR and SPC had satisfactory R ters, for each response (DR and SPC) are shown in more than 99%. In the case of DR, fitting is very good Tables 5 and 6. (R = 99.75%) and only 0.25% of total variance was not The effect is the difference between the responses of two 2 explained by the model. For the SPC (R = 99.99%), levels (high and low level) of factors; the regression model which presents a high value and only 0.01% of a total coefficients are obtained by dividing the effects by two. variance was not explained by the model. The standardized effects (T) are obtained by dividing the regression coefficients by the standard error coefficient Main effects plot (Alimi et al. 2014). Substituting the coefficients b,in Eq. (3) by the respective values from Tables 5 and 6,we The main effects are shown in Figs. 4 and 5, for DR and get: SPC, respectively. It indicates the relative strength of DR ¼ 47:06 þ 20:92E  13:05C þ 4:17Q  0:73E:C effects of various factors. A main effect is present when the þ 0:5E:Q þ 0:67C:Q þ 3:51E:C:Q ð4Þ mean response changes across the level of a factor. The sign of the main effect indicates the direction of the effect SPC ¼ 5:4437 þ 4:9062E þ 2:2063C  0:3862Q (Srinivasan and Viraraghavan 2010). þ 2:1438E  0:4237EQ  0:1687C:Q As shown in Fig. 4, the potential had a positive effect on 0:2812E:C:Q ð5Þ desalination efficiency. In fact an increase of applied 123 4568 Appl Water Sci (2017) 7:4563–4572 Table 4 Full factorial design matrix for desalination efficiency -1 -1 -1 Run order E (V) C (g L ) Q (L h ) DR (%) SPC (Wh L ) 1 5 1 20 31.95 0.55 2 12 1 20 81.26 6.36 3 5 10 20 13 0.45 4 12 10 20 45.35 15.96 5 5 1 90 44.95 0.4 6 12 1 90 82.24 5.64 7 5 10 90 14.64 0.75 8 12 10 90 63.05 13.44 9 8.5 5.5 55 64.39 5.35 10 8.5 5.5 55 61.71 5.18 Fig. 2 Pareto chart for standardized effects for DR 11 8.5 5.5 55 59.09 5.44 12 8.5 5.5 55 62.3 5.35 Table 5 Estimated effects and coefficients for DR (coded units) Term Effect Coefficient Tp value Constant 47.06 60.98 0.000 E 41.84 20.92 27.11 0.000 C -26.09 -13.05 -16.91 0.000 Q 8.33 4.17 5.40 0.012 E.C -1.46 -0.73 -0.95 0.414 E.Q 1.01 0.50 0.65 0.560 C.Q 1.34 0.67 0.87 0.449 E.C.Q 7.02 3.51 4.55 0.020 Fig. 3 Pareto chart for standardized effects for SPC Standard error coefficient for all cases = 0.7716 hydrodynamic and electrical conditions an increase of the R = 0.9975 initial salt concentration leads to a decrease of the DR. This result can be explained by the concentration polarization phenomenon which is more important at high concentra- Table 6 Estimated effects and coefficients for SPC (coded units) tion (Sadrzadeh and Mohammadi 2008). As demonstrated Term Effect Coefficient Tp value in previous studies (Kabay et al. 2002; Banasiak et al. 2007; Sadrzadeh and Mohammadi 2008; Ben Sik Ali et al. Constant 5.4437 141.74 0.000 2010a), the number of ions transported through the mem- E 9.8125 4.9062 127.75 0.000 branes are almost the same but total amounts of salts are C 4.4125 2.2063 57.45 0.000 quite different from the different treated solution. As Q -0.7725 -0.3862 -10.06 0.002 known, the calculation of DR depends strongly on the E.C 4.2875 2.1438 55.82 0.000 initial feed concentration and the amount of transported E.Q -0.8475 -0.4237 -11.03 0.002 ions. So, the DR evolves reciprocally to the initial feed C.Q -0.3375 -0.1687 -4.39 0.022 concentration at some hydrodynamic and electrical condi- E.C.Q -0.5625 -0.2812 -7.32 0.005 tions. Increasing concentration from low to high level Standard error coefficient for all cases = 0.03841 resulted in 26% decrease in DR (Fig. 4). R = 0.9999 At high flow rate, the increase in the DR with flow rate may be attributed to the decrease in the thickness of the boundary layers adjacent to the membranes surfaces with potential from low to high level resulted in increasing DR increasing solution velocity. In the present case, Fig. 4 by 41.84%. shows a slight increase (8.33%) in DR when flow rate At higher salt concentration values, it can be observed increases from low to high level likely because the thick- that the DR has a considerable dependence on the feed -1 ness of the boundary layers adjacent to the membranes solution in the range of 5.5–10 g L . Effectively at some 123 Appl Water Sci (2017) 7:4563–4572 4569 Fig. 4 Main effects plot for DR Fig. 6 Interaction effects plot for DR Fig. 5 Main effects plot for SPC Fig. 7 Interaction effects plot for SPC surfaces does not change significantly when flow rates vary concentration. But a negative interactive effect was -1 from 20 to 90 L h . Similar results were demonstrated by observed between applied potential and flow rate as well as Kabay et al. (2002). between salt concentration and flow rate. An increase of -1 In the case of SPC, applied potential and the initial salt the concentration value from 1 to 10 g L increased the concentration have a positive effect on this response. But -1 -1 SPC by 11 Wh L (from 4 to 15 Wh L )at12 V. -1 the flow rate had a slight negative one. As a general trend, Increasing the flow rate from 20 to 90 L h enhances the an increase of applied potential and initial salt concentra- -1 -1 decrease of SPC by 2 Wh L (from 12 to 10 Wh L )at tion from low to high level resulted in increasing SPC by 12 V. Also, the increase of flow rate from low to high level -1 9.81 and 4.41 Wh L , respectively. -1 decreases the SPC value by 1 Wh L (from 8 to -1 -1 7WhL )at10gL . Interaction effects plot Normal probability plot of residuals The interaction plot is a graphical tool which plots the mean response of two factors at all possible combination of One of the key of assumptions for the statistical analysis of their settings. If the lines are non-parallel, it is an indication data from experiments is that the data that come from a of interaction between the two factors (Antony 2003). normal distribution (Antony 2003). The normality of the Parallel lines indicate that there is no interaction between data can be checked by plotting a normal probability plot two factors. The interaction effect plots are shown in of the residuals. If the points on the plot fall fairly close to a Figs. 6 and 7, for DR and SPC, respectively. straight line, the data are normally distributed (Antony In the case of DR, there are no significant interactions 2003). The normal probability plot of the residuals with a between all factors. In the case of SPC, Fig. 7 shows 95% confidence level for DR and SPC are shown in Figs. 8 positive interaction between applied potential and salt and 9. It can be seen that for DR and SPC, the experimental 123 4570 Appl Water Sci (2017) 7:4563–4572 Fig. 8 Normal probability plot of the residuals for DR Fig. 9 Normal probability plot of the residuals for SPC -1 -1 points fall fairly close to the straight line. Therefore, the 12 V at 90 L h or 18 V at 40 L h , the value of limit data from the experiments come from a normally dis- current density was not reached. pH variation due to the ? - tributed population, and they were reliable. reaction of water dissociation into H O and OH is then avoided and this limits the probability of fouling and/or Treatment of the real water sample scaling formation. Desalination of brackish water was achieved and the Finally, the application of electrodialysis was performed on concentrations of different species in the obtained treated the real brackish water (Table 1) using the optimal water are below the amount recommended by WHO. An parameters. The flow rate and applied potential were fixed 85.5% of DR was obtained after 24 min of ED application -1 -1 at 90 L h and 12 V, respectively. The physicochemical with 14.76 Wh L of SPC for E = 18 V and -1 characteristics of treated water are given in Table 8. Then, Q = 40 L h . Whereas, DR tends to 84% obtained after -1 the results were compared with those obtained using clas- 27 min of ED application with 6.72 Wh L of SPC for -1 -1 sical method of optimization (E = 18 V, Q = 40 L h ) E = 12 V and Q = 90 L h . So, we can clearly observe (Table 7). the advantage of full factorial design which manifests in As shown in Fig. 10 which describes the polarization decreasing SPC. curve for the real water sample, for the applied potential of 123 Appl Water Sci (2017) 7:4563–4572 4571 Table 7 Optimization of factors influencing the ED efficiency by the classical method optimization -1 -1 Effect of applied potential for C = 3gL and Q = 40 L h E (V) DR (%) 10 57.37 15 81.81 18 86.88 -1 Effect of flow rate for C = 3gL and E = 18 V -1 Q (L h ) DR (%) 20 80 30 90 40 90 50 89.47 Fig. 10 Polarization curves, I = f (E) -1 Effect of salt concentration for E = 18 V and Q = 40 L h -1 C (g L ) DR (%) 1 81.48 applied potential and feed flow rate were performed to 1.5 73.91 optimize the demineralization rate and the specific power consumption. 2 73.44 2.5 60 The applied potential and the salt concentration have a significant effect on the process efficiency and mainly on demineralization rate. It was also found that the decrease of salt concentration induces better performance. On the other hand, the specific power consumption was mostly influenced by initial salt concentration and applied Table 8 Physicochemical characteristics of treated water potential. The significant interactions found are between Physicochemical Sample treated Sample treated Recommended applied potential and salt concentration for the SPC. The characteristics at E = 12 V, at E = 18 V, values by WHO factorial experiment design method is undoubtedly a good -1 -1 Q = 90 L h Q = 40 L h technique for studying the influence of major process Conductivity 0.5 0.5 0.5 parameters on response factors by significant reducing the -1 (mS cm ) number of experiment and henceforth, saving time, pH 6.8 6.7 6.5–8.5 energy and money. During this study we were able to -1 TDS (mg L ) 340 308 500 obtain high values of demineralization rate going to - -1 Cl (mg L ) 94.66 44.37 250 82.24%. HCO 00 – 3 Electrodialysis process was applied for the treatment of -1 (mg L ) real brackish water sample. The concentrations of different 2- -1 SO (mg L ) 152.44 80.75 400 species in the obtained treated water are below the amounts - -1 F (mg L ) 0.42 1.05 1.5 recommended by World Health organization for drinking - -1 NO (mg L ) 0.8 0.81 50 water. ? -1 K (mg L ) 7 0.78 12 ? -1 Na (mg L ) 20 21.44 250 Acknowledgements The funding was provided by certe (Grant no. 2? -1 ?216 79325122). Ca (mg L ) 37.33 20 – 2? -1 Mg (mg L ) 32 39.6 – Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http:// creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give Conclusions appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The results from the presented study showed that the desalination of brackish water using electrodialysis process is depending on several parameters. A series of full fac- Publisher’s Note Springer Nature remains neutral with regard to torial design experiments varying initial salt concentration, jurisdictional claims in published maps and institutional affiliations. 123 4572 Appl Water Sci (2017) 7:4563–4572 McGovern RK, Weiner AM, Sun L, Chambers CG, Zubair SM, References Lienhard JH (2014a) On the cost of electrodialysis for the desalination of high salinity feeds. Appl Energy 136:649–661 Alimi F, Boubakri A, Tlili MM, Amor MB (2014) A comprehensive McGovern RK, Zubair SM, Lienhard VJH (2014b) Hybrid electro- factorial design study of variables affecting CaCO scaling under dialysis reverse osmosis system design and its optimization for magnetic water treatment. Water Sci Technol 70:1355–1362. treatment of highly saline brines. J Desalt Water Reuse 6:15–23 doi:10.2166/wst.2014.377 Meski S, Ziani S, Khireddine H, Boudboub S, Zaidi S (2011) Antony J (2003) Design of experiments for engineers and scientists. Factorial design analysis for sorption of zinc on hydroxyapatite. Butterworth-Heinemann, New York J Hazard Mater 186:1007–1017 Azza A, Hamdon A, Darwish NAD, Hilal N (2015) The use of Monohar M, Das AK, Shahi VK (2017) Alternative preparative route factorial design in the analysis of air-gap membrane distillation for efficient and stable anion exchange membrane for water data. Desalination 367:90–102 desalination by electrodialysis. Desalination 413:101–108 Balbasi M (2013) Application of full factorial design method to Montgomery DC (2001) Design and analysis of experiments, 5th edn. silicate synthesis. Mater Res Bull 48:2908–2914 Wiley, New York Banasiak LJ, Kruttschnitt TW, Schafer AI (2007) Desalination using Nemati M, Hosseini SM, Shabanian M (2017) Novel electrodialysis electrodialysis as a function of voltage and salt concentration. cation exchange membrane prepared by 2-acrylamido-2-methyl- Desalination 205:38–46 propane sulfonic acid; heavy metal ions removal. J Hazard Mater Ben Sik Ali M, Mnif A, Hamrouni B, Dhahbi M (2010a) Electro- 337:90–104 dialytic desalination of brackish water: effect of process Reig M, Casas S, Aladjem C, Valderrama C, Gibert O, Valero F, parameters and water characteristics. Ionics 16:621–629 Centeno CM, Larrotcha E, Cortina JL (2014) Concentration of Ben Sik Ali M, Hamrouni B, Dhahbi M (2010b) Electrodialytic NaCl from seawater reverse osmosis brines for the chlor-alkali defluoridation of brackish water: effect of process parameters industry by electrodialysis. Desalination 342:107–117 and water characteristics. Clean-Soil Air Water 38(7):623–629 Reig M, Casas S, Valderrama C, Gibert O, Cortina JL (2016a) Boubakri A, Helali N, Tlili M, Amor MB (2013) Fluoride removal Integration of monopolar and bipolar electrodialysis for val- from diluted solutions by Donnan dialysis using full factorial orization of seawater reverse osmosis desalination brines: design. Korean J Chem Eng 10:1–6. doi:10.1007/s11814-013- production of strong acid and base. Desalination 398:87–97 0263-9 Reig M, Casas S, Gibert O, Valderrama C, Cortina LJ (2016b) Elisseeva TV, Shaposhnik VA, Luschik IG (2002) Demineralization Integration of nanofiltration and bipolar electrodialysis for and separation of amino acids by electrodialysis with ion- valorization of seawater desalination brines: production of exchange membrane. Desalination 149:405–409 drinking and waste water treatment chemicals. Desalination Elmidaoui A, Elhannouni F, Sahli MAM, Chay L, Elabbassi H, Hafsi 382:13–20 M, Largeteau D (2001) Pollution of nitrate in Moroccan ground Rodier J, Legube B, Merlet N (2009) The analysis of water, 9th edn. water: removal by electrodialysis. Desalination 136:325–332 Dunod, Paris, pp 78–81 Elmidaoui A, Elhanouni F, Taky M, Chay L, sahli MA, Echihabi L, Sadrzadeh M, Mohammadi T (2008) Sea water desalination using Hafsi M (2002) Optimization of nitrate removal operation from electrodialysis. Desalination 221:440–447 ground water by electrodialysis. Sep Purif Technol 29:235–244 2? Shady AA, Peng C, Bi J, Xu H, Almeria J (2012) Recovery of Pb Fewtrell L, Bartram J (2008) Guidelines for drinking water quality, and removal of NO from aqueous solutions using integrated 3rd edn. World Health Organization, Geneva, pp 375–492 electrodialysis, electrolysis and adsorption process. Desalination Fidaleo M, Moresi M (2005) Optimal strategy to model the 286:304–315 electrodialytic recovery of a strong electrolyte. J Membr Sci Srinivasan A, Viraraghavan T (2010) Oil removal from water by 260:90–111 fungal biomass: a factorial design analysis. J Hazard Mater Galama AH, Saakes M, Bruning H, Rijnaarts HHM, Post JW (2014) 175:695–702 Seawater predesalination with electrodialysis. Desalination Tanaka Y, Reig M, Casas S, Aladjem C, Cortina JL (2015) Computer 342:61–69 simulation of ion-exchange membrane electrodialysis for salt Ghyselbrecht K, Huygebaert M, Bruggen BV, Ballet R, Meesschaert concentration and reduction of RO discharged brine for salt B, Pinoy L (2013) Desalination of an industrial saline water with production and marine environment conservation. Desalination conventional and bipolar membrane electrodialysis. Desalination 367:76–89 318:9–18 Turan NG, Elevli S, Mesci B (2011) Adsorption of copper and Zinc Jing G, Du W, Guo Y (2012) Studies on prediction of separation ions on illite: determination of the optimal conditions by the percent in electrodialysis process via BP neural networks and statistical design of experiments. Appl Clay Sci 52:392–399 improved BP algorithms. Desalination 291:78–93 ´ ´ Walha K, Amar RB, Firdaous L, Quemeneur F, Jaouen P (2007) Kabay N, Demircioglu M, Ersoz E, Kurucaovali I (2002) Removal of Brackish groundwater treatment by nanofiltration, reverse calcium and magnesium hardness by electrodialysis. Desalina- osmosis and electrodialysis in Tunisia: performance and cost tion 149:343–349 comparison. Desalination 207:95–106 Kabay N, Arar O, Samatya S, Yuksel U, Yuksel M (2008) Separation Zourmand Z, Faridirad F, Kasiri N, Mohammadi T (2015) Mass of fluoride from aqueous solution by electrodialysis: effect of transfer modeling of desalination through an electrodialysis cell. process parameters and other ionic species. J Hazard Mater Desalination 359:41–51 153:107–113 Lee HJ, Strathmann H, Moon SH (2006) Determination of the limiting current density in electrodialysis desalination as an empirical function of linear velocity. Desalination 190:43–50

Journal

Applied Water Science – Springer Journals

Published: Sep 15, 2017

There are no references for this article.