Abstract

Optimized condition for desalination of the reverse osmosis (RO) rejected stream from Esfahan Oil Refining Company (EORC) using direct contact membrane distillation (DCMD) with polytetrafluoroethylene (PTFE) membrane was investigated here, having designed a set of 34 experiments using response surface methodology (RSM) and full factorial design (FFD) modelling, carried out in a laboratory scale set-up built for this purpose. Statistical criteria for validation, significance, accuracy and adequacy confirmed the suitability of the quadratic polynomial model employed. Response plots and regression equations suggested that the permeate flux response improved with increased feed temperature, reduced permeate temperature and enhanced feed flow rate. Optimizing DCMD process showed that maximum permeate flux of 60.76 L/m2·h could be achieved at the following optimum operational conditions: feed temperature and flow rate of 70 °C and 2 L/min, respectively, as well as the permeate temperature of 15 °C. At this point, the mean annual energy required for 90% water recovery (36 m3/h off the RO brackish rejected stream) at EORC refinery was found to be 96 GJ, which could be supplied using solar or conventional energy systems at Isfahan, facing a very critical water shortage at present.

INTRODUCTION

Water is essential in any crude oil refining process, and as such, its recovery has become paramount due to the ever increasing demand, environmental concerns and the need for sustainable development in the petroleum industry. Where water shortages become critical, as is the case in central Iran, there is great pressure from the public and local authorities on facilities like EORC (Esfahan Oil Refining Company), to recover utmost the consumed water. EORC started operation nearly 4 decades ago, when water shortage was not envisaged at the time and hence, little attention was paid to its recovery in the original design, evidenced by numerous large evaporation ponds built at the site.

EORC is an important refinery at the center of Iran and is situated 25 km northwest of the historical city of Isfahan, based on a German-US joint venture design prepared by Thyssen-Fluor for 200,000 bpd crude oil refining capacity, and operated at excess capacity of 375,000 bpd since the early 1980's due to the Iran-Iraq war. It has played a vital role in producing up to 25% of the nation's petroleum products ever since.

Isfahan province, on the other hand, is amongst the world's most water stressed regions in the Middle East and faces water scarcity as the annual water supply is 900 m3 per person, which is below the 1,000 m3 Falkenmark indicator threshold (Rijsberman 2006). Development of high water consuming heavy industries such as steel, petroleum refining and petrochemicals as well as 8,500 other units has intensified water scarcity in the region with consumption of 188 million m3 per annum in 2014 by these industries, and this is expected to reach 471 million m3 per annum by 2031 (of which 258 million m3 is by steel, oil and petrochemicals), thereby imposing a 1.0 billion m3 shortage annually (Torkashvand 2014). Hence, in 2005, a heavily lobbied development plan to double the EORC's present capacity was blocked and stopped by the local department of environment office amidst the construction phase, based on recent laws prohibiting industrial development within 50 km of the city of Isfahan.

At EORC, there is a 90 m3/h wastewater originated from a crude oil desalting unit, sour water process, cooling towers, wastewater streams and the effluent from the reverse osmosis (RO) unit. The 350 m3/h capacity RO unit produces 40 m3/h brackish water, which is transferred directly to ponds No. 5 and 6, two out of 10 (400 m × 100 m) evaporation ponds at the refinery site (see Figure 1).

Figure 1

Evaporation pond No. 5, one of the RO unit effluent receivers at EORC.

Figure 1

Evaporation pond No. 5, one of the RO unit effluent receivers at EORC.

Various experimental and industrial methods have been suggested for maximum water recovery from the RO rejected stream, most of which are associated with drawbacks including large land required, low productivity, high investment capital required, etc. Membrane distillation (MD), on the other hand, offers advantages for such application in a refinery such as mild operating conditions and high rejection factor (high water recovery).

Direct contact membrane distillation (DCMD) is a thermally driven microfiltration process in which water vapor molecules are allowed to pass through a porous hydrophobic membrane, using their vapor pressure gradient. In its quest for maximum water reclamation at EORC, application of DCMD was investigated as a promising technology (Ramlow et al. 2017). In fact, DCMD is one of the four main configurations in membrane distillation; the others include air gap membrane distillation (AGMD), sweeping gas membrane distillation (SGMD) and vacuum membrane distillation (VMD). Attractive features of these technologies include: low operating temperatures and pressures, cost effective membranes that are less prone to corrosion, combination with other processes such as UF (ultrafiltration) and RO units, and utilization of alternative energy sources. The drawbacks include: low permeate flux, feed conditions dependency (concentration and temperature) and considerable conduction heat losses. In DCMD, hot feed is in direct contact with the hot membrane surface, thereby allowing evaporation of volatile molecules to be moved to the permeate side and condensed inside the module by the corresponding induced vapor pressure difference (Alkhudhiri et al. 2012). DCMD technology has been used in various applications such as: desalination (Hou et al. 2010a), separation and purification (Hou et al. 2010b; Kesieme et al. 2014), crystallization (Quist-Jensen et al. 2016b), fruit juice concentration and separations in the food industry (Quist-Jensen et al. 2016a), and production of chemicals (Tomaszewska & Łapin 2012), etc.

In assessing DCMD application for treatment of the RO rejected stream at EORC, one is faced with many operating parameters that influence its performance. Evaluating these is difficult using the traditional approach in which one independent variable is studied while others are kept constant, as this would lead to an excessive number of tests, long experimentation time and high costs, irrespective of the fact that interactions between these influential parameters are often ignored in such approaches. Modern statistical methods, such as design of experiments (DoE) and response surface methodology (RSM), have been employed with great success in determining the most influential parameters of a process and prediction of its performance by means of factorial design and mathematical modeling. In these methods, all significant parameters vary simultaneously in a set of experiments.

RSM, developed by Box and Wilson in 1951, uses a second degree polynomial equation to propose an optimized response of interest influenced by several variables (Box & Wilson 1951; Myers et al. 2016). Although RSM is widely used in many research studies (Bezerra et al. 2008; Gangil & Pradhan 2017), limited works are reported in the literature on its use in various modes of membrane distillation. Khayet et al. used RSM for effective optimization and prediction of permeate flux in a DCMD operation, involving NaCl solution as feed, having considered NaCl concentration, feed temperature and mixing velocity as independent variables (Khayet et al. 2007). Boubakri et al. employed RSM in a similar DCMD optimization operation, but used independent factors of vapor pressure difference, feed flow rate, permeate flow rate and feed ionic strength at five levels with 3.9% deviation in mean experimental and predicted results (Boubakri et al. 2014). In a different SGMD application involving prediction of sucrose permeate flux, 2nd order polynomials were successfully employed by Cojocora & Khayet (Cojocaru & Khayet 2011). The same authors reported using both DoE and RSM in an AGMD application (Khayet & Cojocaru 2012). In another different SGMD application, Khayet et al. used both DoE and RSM, with independent controlling variables being the temperatures and flow rates of liquid and gas phases, and noted that the optimum permeate flux showed only 2% deviation from the predicted one (Khayet et al. 2012). Full factorial design (FFD) was also employed at two levels in studying influential parameters affecting two responses of evaporation flux and soluble solids content in a process involving apple juice concentration (Onsekizoglu et al. 2010). The models were attained using the independent variables of osmotic agent concentration, flow rate and temperature difference between feed and osmotic agent.

In this work, three main aims are pursued: (a) presenting a mathematical correlation for feed temperature and flow rate and permeate temperature as a function of permeate flux, (b) validating the model by comparing the predicted and empirical results, (c) computing and predicting a set of operational conditions determining a specific permeate flux in the DCMD application.

METHODS

Experimental

Figure 2 illustrates the set-up built for the DCMD experiments for this study. The flat sheet membrane modulus was made from 316 stainless steel. Feed and permeate sides were recirculated using peristaltic pumps (Longer Pump Model: WT600-2 J), with their temperatures being controlled (using an ATBIN NOVA) and illustrated on individual indicators.

Figure 2

Schematics of set-up built and used for DCMD experiments.

Figure 2

Schematics of set-up built and used for DCMD experiments.

For measuring total dissolved solids, an on-line TDS meter (Model: TDS-230, CREATE Co. Korea) was used in the permeate side, while flows were measured using rotameters. The feed container was made from steel to avoid corrosion and associated impurities, while being heated up directly using an electrical controlled heater.

Brackish water effluent (Table 1) was supplied from the RO unit at EORC refinery and used as feed to the set-up.

Table 1

Composition of RO brackish water at EORC refinery

ParametersUnitsValue
pH – 7.25 
TDS mg/L 4,700 
Conductivity  8,400 
Total hardness mg/L 3,200 
Total alkalinity mg/L 100 
Chlorides mg/L 780 
Calcium mg/L 1,940 
Magnesium mg/L 1,260 
Sodium mg/L 1,020 
ParametersUnitsValue
pH – 7.25 
TDS mg/L 4,700 
Conductivity  8,400 
Total hardness mg/L 3,200 
Total alkalinity mg/L 100 
Chlorides mg/L 780 
Calcium mg/L 1,940 
Magnesium mg/L 1,260 
Sodium mg/L 1,020 

Microporous PTFE (polytetrafluoroethylene) membranes with PP (polypropylene) substrate used in the experimental set-up were provided from a commercial local supplier (Iran Membrane Co.) and used as purchased. Table 2 lists the membrane characteristics.

Table 2

Properties of commercial PTFE membrane used in this study

ItemUnitValue
Membrane – PTFE 
Substrate – PP 
Pore size   
Thickness   
Bubble point in IPA   
Air perm   
ItemUnitValue
Membrane – PTFE 
Substrate – PP 
Pore size   
Thickness   
Bubble point in IPA   
Air perm   

Following design, fabrication and testing the set-up, a flat membrane module providing 80 cm2 effective area was designed and built from stainless steel SS-316.

Design of experiments

Using DE 7.1.5 software for modelling, the ‘miscellaneous’ item was selected from the ‘response surface’ menu. For Design type, ‘3-level factorial’ was chosen using three ‘numeric factors’. Considering the industrial application of this study, assessable operation parameters need to be selected as independent variables. Having three parameters of feed temperature , feed flow rate and permeate temperature introduced as independent variables, with permeate flux taken as their corresponding response, a set of 32 tests were proposed by the software. Following the accomplishment of these experiments, the empirical data for were entered into the software and run subsequently. It is worth noting that optimization could be achieved using the DE program. At this stage, a quadratic equation was proposed to correlate independent variables and their corresponding response as follows: 
formula
(1)
Here, Y refers to the response variable, is a constant coefficient, 's are the coefficients, 's are the quadratic coefficients, 's are the interaction coefficients and , refer to the coded values of the independent variables. Considering the aforementioned variables, Equation (1) may be rewritten as: 
formula
(2)
Coded variables of , , and denote feed temperature, permeate temperature and feed flow rate here, respectively, while Y denotes the permeate flux. Experimental conditions and independent variable levels are described in Table 3, while FFD experimental matrix and response values are reported in Table 4.
Table 3

Coded and actual values of full factorial designed variables

FactorSymbolActual value of coded level
− 101
Feed temperature (°C)  50 60 70 
Permeate temperature (°C)  15 25 35 
Feed flow rate (L/min)  1.5 
FactorSymbolActual value of coded level
− 101
Feed temperature (°C)  50 60 70 
Permeate temperature (°C)  15 25 35 
Feed flow rate (L/min)  1.5 
Table 4

Full factorial design of experiments and output DCMD responses

RunFactors

ActualPredicted
−1 −1 −1 19.06 19.17 
−1 −1 30.13 28.99 
+1 −1 −1 42.69 43.89 
−1 −1 14.19 15 
−1 25 23.76 
+1 −1 40.56 37.6 
−1 +1 −1 8.88 10.98 
+1 −1 19.25 18.67 
+1 +1 −1 29.75 31.45 
10 −1 −1 22.06 21.95 
11 −1 30.94 33.78 
12 +1 −1 51.5 50.7 
13 −1 20.56 17.34 
14 27.56 28.11 
15 +1 42 43.96 
16 −1 +1 14.13 12.87 
17 +1 25.94 22.58 
18 +1 +1 36.94 37.36 
19 −1 −1 +1 27.81 27.99 
20 −1 +1 41.88 41.83 
21 +1 −1 +1 63 60.76 
22 −1 +1 24.69 22.93 
23 +1 33.69 35.71 
24 +1 +1 52.69 53.57 
25 −1 +1 +1 14.88 18.01 
26 +1 +1 31.75 29.73 
27 +1 +1 +1 46.69 46.53 
28 27.44 28.11 
29 27.06 28.11 
30 27.81 28.11 
31 27.19 28.11 
32 28.06 28.11 
RunFactors

ActualPredicted
−1 −1 −1 19.06 19.17 
−1 −1 30.13 28.99 
+1 −1 −1 42.69 43.89 
−1 −1 14.19 15 
−1 25 23.76 
+1 −1 40.56 37.6 
−1 +1 −1 8.88 10.98 
+1 −1 19.25 18.67 
+1 +1 −1 29.75 31.45 
10 −1 −1 22.06 21.95 
11 −1 30.94 33.78 
12 +1 −1 51.5 50.7 
13 −1 20.56 17.34 
14 27.56 28.11 
15 +1 42 43.96 
16 −1 +1 14.13 12.87 
17 +1 25.94 22.58 
18 +1 +1 36.94 37.36 
19 −1 −1 +1 27.81 27.99 
20 −1 +1 41.88 41.83 
21 +1 −1 +1 63 60.76 
22 −1 +1 24.69 22.93 
23 +1 33.69 35.71 
24 +1 +1 52.69 53.57 
25 −1 +1 +1 14.88 18.01 
26 +1 +1 31.75 29.73 
27 +1 +1 +1 46.69 46.53 
28 27.44 28.11 
29 27.06 28.11 
30 27.81 28.11 
31 27.19 28.11 
32 28.06 28.11 

As for the model validation, three statistical criteria of F-value, P-value and diagnostic plots were employed in the analysis of variance (ANOVA), as discussed below. Fitting the polynomial model to the experimental results was assessed by R2 (the coefficient of determination) and adjusted R2.

RESULTS AND DISCUSSION

RSM regression model equation

As can be seen in Table 4, both experimental and predicted DCMD permeate fluxes conform well, justifying further statistical analysis to be carried out.

The equation describing permeate flux in L/m2·h as functions of coded variables for the model employed may be written as: 
formula
(3)
Rearranging Equation (3) based on the actual variables obtained in this work yields: 
formula
(4)

Analysis of variance

Statistical significance, evaluated by the estimators in ANOVA (see Table 5), reflects the quality and effectiveness of the model. P-value of less than 0.0001 and F-value of 127.55 confirms that the model is highly significant, as evidenced by the close fit of predicted and actual response values in Table 4.

Table 5

Analysis of variance for the quadratic model using permeate flux as response

SourceModel
Sum of squares 4,549.25 
df 
Mean square 505.47 
F value 127.55 
p-value Prob > F <0.0001 Significant 
Standard deviation 1.99 
Mean 30.49 
C.V. % 6.53 
PRESS 209.69 
R-squared 0.9812 
Adjusted R-squared 0.9735 
Predicted R-squared 0.9548 
Adequate precision 44.730 
SourceModel
Sum of squares 4,549.25 
df 
Mean square 505.47 
F value 127.55 
p-value Prob > F <0.0001 Significant 
Standard deviation 1.99 
Mean 30.49 
C.V. % 6.53 
PRESS 209.69 
R-squared 0.9812 
Adjusted R-squared 0.9735 
Predicted R-squared 0.9548 
Adequate precision 44.730 

The proportion of variation in the permeate flux, reflected by the R2 value and attributed for the proposed regression model, should be as close to 1.0 as possible. Here, the 0.9812 value for R2 indicates that 98.12% of the deviation can be explained by the empirical model, reaffirming the model's significance. Adjusted R2, however, evaluates the model adequacy based on the number of terms used in the model. In other words, the adjusted R2 of 97.35% reflects the percentage variation by only the independent variables that actually affect the response, indicating a highly correlated experimental and predicted data.

Figure 3 compares predicted versus actual permeate flux of the DCMD by plotting a regression line, where clearly a close fit is observed. Statistical analysis presented above shows that the model developed can predict and optimize DCMD process with PTFE membrane for desalinating the RO effluent at EORC refinery confidently, knowing that the process is statistically validated over the experimentation domain envisaged and the permeate flux response.

Figure 3

Experimental DCMD fluxes vs. predicted ones.

Figure 3

Experimental DCMD fluxes vs. predicted ones.

Variables interactions

Effective interactions of independent variables on the permeate flux response could be graphically represented by three dimensional response surface plots prepared from the regression Equation (4), where the interaction of each variable is examined while the other is taken as a fixed constant at its respective zero level. Figure 4(a)–4(c) illustrate such 3D permeate flux response surface plots for this study, where variable interactions and their optimum level for maximum response are examined.

Figure 4

Graphical illustration of variable interactions for the proposed DCMD desalination at EORC refinery.

Figure 4

Graphical illustration of variable interactions for the proposed DCMD desalination at EORC refinery.

The influences of permeate flux and feed flow rate at constant feed temperature on the response, as shown in Figure 4(a), indicate that increasing feed flow rate or decreasing permeate temperature would lead to increased permeate flux, with the maximum permeate flux being evaluated at 41.88 L/m2·h. On the other hand, Figure 4(b) indicates that, at constant permeate temperature, increased feed flow and temperature would improve the response with the maximum permeate flux being 52.69 L/m2·h. For constant feed flow rate, however (see Figure 4(c)), increased feed temperature or decreased permeate temperature would enhance the response with the maximum permeate flux being 51.50 L/m2·h. Therefore, as can be seen from Figure 4, the most influential parameter on the permeate flux is the feed temperature. The corresponding coefficient values associated with independent variables in Equation (3) also confirm this with (the coefficient for the feed temperature, ), being nearly twice as large as those of permeate temperature or feed flow rate .

The influence of feed temperature was investigated by Khayet et al. (2007) for 25.0, 37.5 and 50.0 °C, which showed that feed temperature rise had the highest influence on the permeate flux. The present study also confirmed the same. As regards the influence of feed flow rate, Boubakri et al. (2014) tried five levels ranging from 62.4 to 73.6 L/m2·h and demonstrated that increasing the feed flow rate would also lead to increased permeate flux, be it to a lesser extent compared to the feed temperature. Having tried larger feed flow rates, the same results was concluded in this work, too.

Optimizing operational conditions for DCMD desalinating at EORC refinery

As expressed before, our main aim in this work was to determine optimum DCMD operational conditions for desalination of the RO effluent at EORC using a PTFE membrane. Using a design expert (DE optimizer) software, we can suggest the best process operating conditions having searched through the response surfaces and interaction variables considering their main effects in the model global information. In this work, the optimum maximized permeate flux was found to be 60.76 L/m2·h with 0.959 desirability. The operating conditions at this point are: 70 °C feed temperature, 15 °C permeate temperature and 2 L/min feed flow rate. There was a 3.5% error in the maximum response evaluated, and that of the empirical one. This compared to about 0.9–1.2% in the Khayet et al. (2007) study, and 3.9% in the work of Boubakri et al. (2014) on DCMD structure, and 2.0% (SGMD) and 1.6% (AGMD) of the Khayet & Cojocaru (2012) and Khayet et al. (2012) researches. Also, a minimum permeate flux response of 10.98 L/m2·h, with 0.961 desirability, was attained at the operating conditions of 50 °C feed temperature, 35 °C permeate temperature and 1 L/min feed flow rate.

An important feature of this modeling exercise is its ability to predict a set of operating conditions (or working line) for a target response. Based on EORC needs, different scenarios could be examined with the respective results (optimum conditions, energy expenditure, etc.) However, this requires access to operational management data and strategies from the refinery which are not available at this stage, and should be carefully worked out in future studies. In general, it could be said that if EORC aims to recycle a predetermined permeate flux (or a range of it), considering the climatic conditions and optimum energy consumption at various seasons, once configured, the software could provide various operating conditions for it. On the other hand, for practical purposes, a stable achievable permeate flux may be required at EORC which is expected to be less than the maximized optimum flux response. Extrapolating the independent variables could lead to diminished accuracy of the model. However, interpolation carried out indicated that 70% of the maximum optimum flux (i.e. nearly 45 L/m2·h) might be a practical stable response, arrived at through a set of working lines. Selecting the right point on this working line should be made based on the required climatic conditions, energy restrictions and economic considerations. Figure 5 illustrates an exemplary working line for the mentioned 70% response, which corresponds to 25 °C permeate flux (see Figure 5(a)), 1.5 L/min feed flow rate (see Figure 5(b)), and 69 °C feed temperature (see Figure 5(c)). Many other operational conditions were also proposed by the model for the 45 L/m2·h, which give the operators far more flexibility in planning and reuse of the DCMD desalinated water.

Figure 5

Working line for at (a) (b) (c) .

Figure 5

Working line for at (a) (b) (c) .

Climate is an important influential factor in selecting the operating conditions considering the energy issues. In summer, when working with maximum feed and permeate temperatures are favorable, the maximized optimum permeate flux was 46.53 L/m2·h, attained at 70 °C feed temperature, 35 °C permeate flux and 2.0 L/min feed flow rate. On the contrary, in winter working with the minimal feed and permeate temperature are favorable, with the operating conditions being 63 °C and 15 °C for feed and permeate temperatures, respectively with the same 2.0 L/min feed flow rate yielding 46.83 L/m2·h permeate flux response.

Considering the 40 m3/h mean effluent flow rate of the brackish water from the RO unit at EORC, and 90% recycling requirement by the DCMD unit (i.e. 36 m3/h water reclamation), Table 6 shows the mean average monthly temperature of the city of Isfahan (taken from the past 50 years' national meteorological data) and the energy required in MJ/h to bring the feed temperature to its corresponding maximized optimum. The available energy sources at the refinery include: low pressure (1.0 bar–48 ton/h, 4.0 bar–320 ton/h) operational steam, flare gases and natural solar energy (considering 215 sunny days per annum).

Table 6

Monthly mean temperature and energy required for raising the temperature up to 70 °C

MonthAverage temperature (C)Energy required (MJ/h)
January 2.2 9,966.6 
February 5.2 9,525.6 
March 9.8 8,849.4 
April 15.1 8,070.3 
May 20.4 7,291.2 
June 25.3 6,570.9 
July 28.2 6,144.6 
August 26.8 6,350.4 
September 22.8 6,938.4 
October 16.5 7,864.5 
November 9.9 8,834.7 
December 4.8 9,584.4 
MonthAverage temperature (C)Energy required (MJ/h)
January 2.2 9,966.6 
February 5.2 9,525.6 
March 9.8 8,849.4 
April 15.1 8,070.3 
May 20.4 7,291.2 
June 25.3 6,570.9 
July 28.2 6,144.6 
August 26.8 6,350.4 
September 22.8 6,938.4 
October 16.5 7,864.5 
November 9.9 8,834.7 
December 4.8 9,584.4 

Low pressure operational steams are used elsewhere in the refinery and no surplus energy may be allocated for practical and emergency reasons. Energy recovery from flares has not yet become a realistic target at EORC, and much more upgrading is necessary before it can be considered as a viable option. As for application of a solar system, it is worth noting that the heat requirement for any membrane distillation may account for 90% of the total energy requirement and this increases significantly with feed temperature for DCMD. Solar renewable energy, despite its capital intensive investment, has been considered in the limited non-commercial DCMD desalination studies available in the literature (Suárez et al. 2010; Kim et al. 2013; Shim et al. 2015). Compared to the heavily subsidized conventional energy costs in Iran, employing green energies such as solar systems are not economically justified at present. However, considering the latest corresponding price improvements on fossil fuels, increased supportive policies for industrial green energy system developments, restricted environmental protection implementations, increased public awareness and emphasis by authorities on the sustainability of the oil industry, would most likely make such technological improvements by DCMD and solar desalination economically justifiable in the near future. EORC could play a proactive role in this approach, bearing in mind that there are no technical hindrances, space limitations (due to large evaporation ponds being next to the RO unit), critical water shortages, and policies described above. Hence, EORC could well decide on using fossil fuels for DCMD heat requirement in the meantime.

CONCLUSIONS

Having faced a critical water shortage at EORC, application of DCMD desalination on the brackish water off the RO unit was studied using a PTFE membrane. Determination of optimum operating conditions for the permeate flux was investigated, having designed the experiments with DE7.5.1 software (FFD) and using laboratory data obtained from a DCMD set-up in which the real RO effluent was tested in 34 experiments proposed, where independent parameters of feed and temperatures as well as feed flow rate were selected for a dependent permeate flux parameter. Having validated the results, RSM was used to optimize and predict the operational conditions and their corresponding working lines. R2 = 0.9812 and adjusted R2 = 0.9735 reflect the suitability of calculated regression line equations fitting well with the empirical data. Following optimization, the maximum permeate flux in the experimental domain investigated was found to be 60.76 L/m2·h (desirability = 0.957), having the following operation conditions: feed temperature of 70 °C, permeate temperature of 15 °C and feed flow rate of 2 L/min. The error for the maximum empirical and predicted DCMD permeate flux was 3.5%. Considering 90% water reclamation of the brackish water off the RO effluent (36 m3/h), the variation of the climatic condition at Isfahan, the annual energy required for this hybrid DCMD desalination at EORC is 96 GJ, which can easily be attained from solar or conventional systems.

ACKNOWLEDGEMENTS

The authors gratefully acknowledge Research and Development departments at EORC and ABFA (water and wastewater authorities) Kashan for their collaboration and support.

REFERENCES

REFERENCES
Alkhudhiri
A.
,
Darwish
N.
&
Hilal
N.
2012
Membrane distillation: a comprehensive review
.
Desalination
287
,
2
18
.
Bezerra
M. A.
,
Santelli
R. E.
,
Oliveira
E. P.
,
Villar
L. S.
&
Escaleira
L. A.
2008
Response surface methodology (RSM) as a tool for optimization in analytical chemistry
.
Talanta
76
,
965
977
.
Boubakri
A.
,
Hafiane
A.
&
Bouguecha
S. a. T.
2014
Application of response surface methodology for modeling and optimization of membrane distillation desalination process
.
Journal of Industrial and Engineering Chemistry
20
,
3163
3169
.
Box
J.
&
Wilson
W.
1951
Central composites design
.
JR Stat. Soc.
1
,
1
35
.
Gangil
M.
&
Pradhan
M. K.
2017
Modeling and optimization of electrical discharge machining process using RSM: a review
.
Materials Today: Proceedings
4
(2), Part A
,
1752
1761
.
Hou
D. Y.
,
Wang
J.
,
Qu
D.
,
Luan
Z. K.
,
Zhao
C. W.
&
Ren
X. J.
2010a
Desalination of brackish groundwater by direct contact membrane distillation
.
Water Science and Technology
61
,
2013
2020
.
Hou
D. Y.
,
Wang
J.
,
Wang
B. Q.
,
Luan
Z. K.
,
Sun
X. C.
&
Ren
X. J.
2010b
Fluoride removal from brackish groundwater by direct contact membrane distillation
.
Water Science and Technology
61
,
3178
3187
.
Kesieme
U. K.
,
Milne
N.
,
Cheng
C. Y.
,
Aral
H.
&
Duke
M.
2014
Recovery of water and acid from leach solutions using direct contact membrane distillation
.
Water Science and Technology
69
,
868
875
.
Khayet
M.
,
Cojocaru
C.
&
García-Payo
C.
2007
Application of response surface methodology and experimental design in direct contact membrane distillation
.
Industrial and Engineering Chemistry Research
46
,
5673
5685
.
Khayet
M.
,
Cojocaru
C.
&
Baroudi
A.
2012
Modeling and optimization of sweeping gas membrane distillation
.
Desalination
287
,
159
166
.
Kim
Y. D.
,
Thu
K.
,
Ghaffour
N.
&
Choon Ng
K.
2013
Performance investigation of a solar-assisted direct contact membrane distillation system
.
Journal of Membrane Science
427
,
345
364
.
Myers
R. H.
,
Montgomery
D. C.
&
Anderson-Cook
C. M.
2016
Response Surface Methodology: Process and Product Optimization Using Designed Experiments
.
John Wiley & Sons
,
Hoboken, NJ, USA
.
Onsekizoglu
P.
,
Savas Bahceci
K.
&
Acar
J.
2010
The use of factorial design for modeling membrane distillation
.
Journal of Membrane Science
349
,
225
230
.
Quist-Jensen
C. A.
,
Macedonio
F.
,
Conidi
C.
,
Cassano
A.
,
Aljlil
S.
,
Alharbi
O. A.
&
Drioli
E.
2016a
Direct contact membrane distillation for the concentration of clarified orange juice
.
Journal of Food Engineering
187
,
37
43
.
Quist-Jensen
C. A.
,
Macedonio
F.
&
Drioli
E.
2016b
Membrane crystallization for salts recovery from brine – an experimental and theoretical analysis
.
Desalination and Water Treatment
57
,
7593
7603
.
Ramlow
H.
,
Machado
R. a. F.
&
Marangoni
C.
2017
Direct contact membrane distillation for textile wastewater treatment: a state of the art review
.
Water Science and Technology
76
,
2565
.
Rijsberman
F. R.
2006
Water scarcity: fact or fiction?
Agricultural Water Management
80
,
5
22
.
Shim
W. G.
,
He
K.
,
Gray
S.
&
Moon
I. S.
2015
Solar energy assisted direct contact membrane distillation (DCMD) process for seawater desalination
.
Separation and Purification Technology
143
,
94
104
.
Torkashvand
M.
2014
Investigating Methods of Water Recovery From Effluent of the Reverse Osmosis System at Esfahan Refinery
.
MS Thesis
,
Department of Engineering, University of Kashan
,
Kashan
,
I. R. Iran
.