Electrocoagulation Treatment of Petroleum Refinery Wastewater: Optimization through RSM

Electrocoagulation Treatment of Petroleum Refinery Wastewater: Optimization through RSM

Saidat Olanipekun Giwa1, Abdulwahab Giwa2, Zehra Zeybek1 and Hale Hapoglu1

1Department of Chemical Engineering, Faculty of Engineering, Ankara University, Ankara, Turkey 2Department of Chemical Engineering, Faculty of Engineering, Middle East Technical University, Ankara, Turkey

ABSTRACT

The optimization of the electrocoagulation process used to accomplish the treatment of a turbid petroleum refinery wastewater has been carried out in this work.The turbidity removal and the operating cost of the process were chosen as the response (dependent) variables whilethe current density, the initial wastewater conductivity, the initial pH and the electrolysis time were selected as the input (independent) variables. A set of 30 experimental runs were designed using Central Composite Design of Response Surface Methodology. The designed experiments were run in the experimental setup of the process from where the data used for model development were obtained. The developed models were, thereafter, analyzed and optimized to obtain the optimum values of the electrocoagulation system. The results obtained revealed that the turbidity removal and the operating cost were largely affected by the current density and the electrolysis time. Also, the analyses of variance (ANOVA) of the models showed that the developed models for the turbidity removal and the operating cost were significant with p-values of 0.0003 and less than 0.0001, respectively. In addition, the estimated optimum conditions of the treatment system were 9.9 mA/cm2, 5 mS/cm, 9 and 18 min for the current density, the conductivity, the pH and the electrolysis time, respectively. Using the estimated optimum values to run the experimental system, 96.94% of turbidity and 78.77% of chemical oxygen demand (COD) removals were achieved at the operating cost of 0.654 US$/m3.

Keywords: Electrocoagulation, Petroleum refinery wastewater, Optimization, Response Surface Methodology (RSM), Central Composite Design, ANOVA.

1. INTRODUCTION

   Petroleum refinery generates significant volume of wastewater that are in the range of 0.4-1.6 times the amount of the crude oil processed (Coelho et al., 2006). Petroleum refinery wastewater is characterized by high concentrations of aliphatic and aromatic hydrocarbons, which usually have detrimental and harmful effects on plant, aquatic life as well as surface and ground water sources (El- Naas et al., 2009). Prior to biological treatments, petroleum refinery wastewater is usually treated using physicochemical and mechanical methods. Physicochemical process, such as coagulation, generates large amount of sludge. The sludge treatment cost can increase the total cost of the wastewater treatment. Biological method, on the other hand, cannot efficiently treat wastewater containing non-biodegradable pollutants. Also, mechanical method may require additional maintenance and operation costs.

   Electrocoagulation is a wastewater treatment method that is based on electrolytic generation of coagulant in aqueous medium. The commonly used electrodes for the process are usually made up of aluminum and iron. The use of titanium (Chen and Deng, 2012) and stainless steel (Olmez, 2009) electrodes have also been reported. The main three steps involved in this process are: (1) electrolytic oxidation of anode electrode – on passage of electric current, metallic anode dissolves to form metallic ions (e.g., Al3+) with simultaneous production of hydroxyl ion (OH-) at the cathode; (2) formation of coagulant (e.g., aluminum hydroxide) from reaction between metallic ion and hydroxyl ion; (3) destabilization of pollutants which occurs through the adsorption of the pollutants on the surface of coagulant or by oxidation.

   The advantage of this method include high treatment efficiency, requirement of simple equipment and ease of operation. Moreover, it can be carried out without addition of any chemical. Thus, the amount of sludge generated by this process is usually lower than that generated by chemical coagulation.

   This method has been used to treat a wide variety of wastewaters including paper mill wastewater (Katal and Pahlavanzadeh, 2011), synthetic diary effluents (Tchamango et al., 2010), black liquor from paper industry (Zaied and Bellakhal, 2009). However, the efficiency of electrocoagulation process depends on factors such as current density, electrolysis time, initial pH of the wastewater, conductivity of the solution or supporting electrolyte concentration, pollutant initial concentration and temperature.

   Conventionally, the efficiency of a multi variable dependent process is studied by varying one factor at a time while other factors are kept constant. This method normally ignores the interactions occurring among the factors. Thus, it may not actually give the best conditions that givethe optimum efficiency of the process under investigation. Consequently, Response Surface Methodology (RSM) has been discovered as an effective statistical method of optimizing a process using designs such as Central Composite Design (CCD), Box-Behnken design and D-optimal design. Response Surface Methodology, apart from revealing the true optimum conditions with minimal number of experiments compared to the conventional method, gives the mathematical model(s) defining the relationships between the response(s) and the factors.

   Therefore, in this work, using Central Composite Design, Response Surface Methodology has been applied to optimize the electrocoagulation process used to treat a petroleum refinery wastewater. In order to achieve the aim of this work, the turbidity removal efficiency and the operating cost of the treatment were taken as the responses of the system while the four chosen independent variables were the current density, the conductivity, the pH and the electrolysis time.

2. MATERIALS AND METHOD

   The electrocoagulation experiments of this work were carried out in batch mode using four aluminum electrodes contained in a plexi glass made reactor having a capacity of 1.5 L, as shown in Figure 1. The 45mm x 60mm x 3mm plates having total effective area of 96 cm2 were placed vertically in the reactor at a distance of 1.5 cm apart. Before and after each experimental run, the electrodes were thoroughly rinsed with distilled water to remove clingy impurities from their surfaces. To ensure uniform concentration of the solution in the reactor, the solution was gently stirred using a laboratory stirrer (MTOPS, MS-3020). Conductivity and pH of the solution were measured using conductivity and pH meter (Mettler Toledo M200 easy), respectively. The pH and the conductivity adjustments were also done respectively by adding H2SO4/NaOH (0.5 M) and NaCl. During the treatment,the temperature of the wastewater was controlled by circulating cold water round the reactor using a refrigerated and heated circulation bath (Hoefer RCB20-PLUS). Thus, all the experiments were carried out at room temperature.

   For each experimental run 1 L of petroleum refinery wastewater was used. The measured characteristics of the wastewater are given in Table 1. The experimental conditions for each run are also given in the design matrix contained in Table 3. The turbidity of the wastewater was measured using a water analysis system (Orbeco-Hellige, Model 975-MP) and its chemical oxygen demand (COD) was analyzed using open reflux method (standard method 5220 B).

   Shown in Equation (1) is the expression used to estimatethe turbidity and the COD removal efficiencies of the treatment.The treatment operating cost (OC, US $/m3) for each experimental run was also calculated using Equation (2).

   Y %

   Co Ct 100

   Co

   (1)

   OC aCenergy bCelectrode

   (2)

   In Equation (1) above, Y, Co and Ct are the pollutant removal efficiency, the initial concentration and the concentration at a specific time t. Also, in Equation (2), Cenergy (kWh/m3) and Celectrode (kg Al/m3) are the quantities of energy and the electrodes consumed respectively for the treatment. The energy consumption was calculated using Equation (3) and the quantity of Al used was determined by deducting the final weights of the electrodes from their initial weights. Coefficients

   and are the industrial energy and the wholesale electrode prices which were obtainedrespectively to be 0.098064 US$/kWh and 3.9852 US $/kgAl.

   Cenergy

   IVt

   v

   (3)

   In Equation (3) I, V, t and v are the applied current (A), the voltage (V), the time (h) and the volume of the treated wastewater (m3), respectively.

   Figure 1. The schematic diagram of the experimental set up

   In order to optimize the turbidity removal efficiency of the electrocoagulation system and its operating cost, 30 experimental runs were designed based on Central Composite Design. The experimental matrix comprised16 factorial runs, 6 center point runs and 8 axial runs (see Table 3). The details of the design matrix including the levels used are given in Table 2.

   The experimental data were analyzed using a reduced quadratic polynomial model and the regression coefficients were obtained. The optimization was done using numerical approach. The goal of the optimization was set to finding the operating conditions that would give the maximum turbidity removal efficiency at the minimum operating cost. The experimental design, the statistical analysis and the optimization were accomplished with the aid of Design-Expert 7.0.0.

   Table 1. The measured characteristics of the petroleum refinery wastewater

   Pollutant parameters	Quantity
   Turbidity (FTU)	44.5
   COD (mg/L)	130
   conductivity (mS/cm)	1.96
   pH	7.48

   Table 2. The levels of the factor in the design matrix

   Actual variable, unit	Factor	levels
   		–	-1	0	1	+
   Current density, mA/cm2	X1	4.17	7.29	10.42	13.54	16.67
   Conductivity, mS/cm	X2	2	3	4	5	6
   pH	X4	6	7	8	9	10
   Electrolysis time, min	X3	10	15	20	25	30

3. RESULTS AND DISCUSSIONS

   3.1 Experimental Findings

   The results of the electrocoagulation experiments obtained are given in Table 3. The turbidity removal efficiency and operating cost were found to be affected by the variation of the current density, the conductivity, the pH and the electrolysis time. The center point experiments (conditions:

   10.42 mA/cm2, 4 mS/cm, 9 (pH) and 18 min) gave average turbidity removal of 92.6% with operating

   cost approximately equal to 1US$/m3. In the axial experiments, increasing the current density and the electrolysis time caused increase in the turbidity removal efficiency and the operating cost. Increasing the current density by 6.25 mA/cm2 led to turbidity removal and operating cost of 94.22% and 1.3259 US$/m3, decreasing by the same amount resulted in turbidity removal and operating cost of 30.11% and 0.4481 US$/m3, respectively. Also, as seen in Table 3, at the negative and the positive electrolysis time axial experiments, the turbidity removal were 82.83% and 94.54% with operating costs of 0.4886 US$/m3 and 1.7485 US$/m3, respectively. Though, axial variation of pH and conductivity had little significant on the responses.

   Looking at Table 3, in the factorial experiments, the single effect of all the factors were not very significant but they were found to affect the two responses considered in this work in interactive forms. For instance, at 3 mS/cm, 9 pH and 25 min, when the current density was increased from 7.29 mA/cm2 to 13.54 mA/cm2, the turbidity removal decreased from 98.36% to 79.21%. However, at this conditions, the operating cost increased from 0.9135 US$/m3 to 1.4859 US$/m3. Also, when the current density and the electrolysis time were simultaneously varied in such a way that at the maximum current density, the electrolysis time was minimum, the significant effects of these factors were noted on the turbidity removal efficiency and the operating cost. This can be seen clearly in Table 3 by comparing run numbers 16 and 23. Also, the variations of the current density and the pH, and the current density and the conductivity in the same manner led to increase in the turbidity removal and the operating cost. According to the results obtained, one of the factorial experimentscarried out led to maximum turbidity removal efficiency of 98.36% with operating cost of 0.9135 US $/m3.

   Table 3. Experimental design matrix and the petroleum refinery wastewater treatment results

   	Factors				Responses
   Run	X1, mA/cm2	X2, mS/cm	X3	X4, min	T, %	OC, US $/m3
   1	10.42	4	8	20	92.61	0.7780
   2	13.54	3	9	25	79.21	1.4859
   3	7.29	5	9	25	89.53	1.9019
   4	7.29	5	9	15	92.11	0.5908
   5	10.42	4	10	20	92.54	0.9394
   6	13.54	3	9	15	86.70	0.8144
   7	10.42	6	8	20	94.83	0.9250
   8	7.29	3	9	15	90.16	0.9693
   9	7.29	3	7	25	91.35	0.8434
   10	10.42	4	6	20	92.00	1.1698
   11	16.67	4	8	20	94.22	1.3259
   12	10.42	4	8	20	92.58	1.1514
   13	10.42	4	8	10	82.83	0.4886
   14	13.54	3	7	25	79.46	2.3947
   15	13.54	5	7	25	79.19	1.8683
   16	13.54	5	7	15	87.12	1.8534
   17	10.42	4	8	20	92.63	1.1606

   18	4.17	4	8	20	30.11	0.4881
   19	7.29	3	9	25	98.36	0.9135
   20	10.42	4	8	30	94.54	1.7485
   21	13.54	5	9	25	91.46	1.8527
   22	10.42	4	8	20	92.67	0.9657
   23	7.29	5	7	25	73.93	1.06402
   24	7.29	5	7	15	40.45	0.56288
   25	10.42	4	8	20	92.52	0.97767
   26	10.42	2	8	20	91.30	1.39421
   27	13.54	5	9	15	96.58	0.98259
   28	10.42	4	8	20	92.70	1.11715
   29	7.29	3	7	15	64.83	0.95358
   30	13.54	3	7	15	95.71	1.38424

   3.2 Statistical Studies Findings

   As seen in the previous section,where the results of the experiments carried with the design of Central Composite Design were presented, it was found that Central Composite Designnormally combines factors in such a way that they are easily understood by the experimenter. Thus, it enhances easy study of both the single and the interactive effects of the factors on the chosen response(s) after the experiments by merely looking at the table containing the design matrix and the results, even before any statistical analysis. This seems to be one of thesignificant advantages ofthis experimental design methodology (Central Composite Design).

   The reduced quadratic model obtained for the turbidity removal and the operating cost are given in Equations (1) and (2), respectively. Results of the analysis of variance (ANOVA) showed that these models were significant with p-values of 0.0003 (see Table4) and less than 0.0001 (see Table 5), respectively. The high R-square values of the models confirm their agreements with the experimental data.

   1

   1

   For the turbidity removal, the significant model terms were found to be X1, X3, X1X3, X1X4 and X 2. This revealed thatthe turbidity removal data predicted by the models were affected by single variation of current density, pH, interactively affected by combination of current density and pH,

   current density and electrolysis time as well as by the quadratic term of the current density. The contour plots given in Figures 2(b) and 3(b) revealed that simultaneous increase in current density and pH or electrolysis time led to increase in turbidity removal. As seen in the 3D surface graphs of the results, the maximum turbidity removal efficiency was achieved within the design points by varying the current density and the pH (Figure 2(a)) and the current density and the electrolysis time (Figure 3(a)).

   Similarly, the significant model terms for the operating cost model were obtained to be X1, X4 and X1X3. This impliedthat the operating cost was singly affected by the current density and the electrolysis time, and interactively influenced by the current density and the pH. It was also discovered from the contour plot shown in Figure 4(b) that the pH of the wastewater affected the operating cost almost in parabolic manner with increase in the current density. It was then noted that the effect to current density over weighted that of the pH because the graph moved rightward more. The 3D surface plot of the results of the operating cost model also revealed that the minimum operating cost was within the design points.

   T 99.82521 35.75256X1

   1.2391X1 X 2 1.74247X1 X 3

   1

   1

   0.085674X 2 X 4 0.77473X 2

   48.75449X 2 6.38474X 3 4.35887X 4

   0.40962X1 X 4 4.11938X 2 X 3

   (1)

   C2 3.14056 0.64422X1 0.68707X 2 0.26964X3 0.056109X 4

   1

   1

   0.066331X1 X3 0.085776X 2 X3 0.0012837X 2

   (2)

   Table 4. Results of analysis of variance (ANOVA) for turbidity removal model

   Source	Sum of squares	Degree of freedom	Mean square	f-value	P-value
   Model	5367.81	10	536.78	6.24	0.0003
   X1	1394.52	1	1394.52	16.2	0.0007
   X2	33.46	1	33.46	0.39	0.5404
   X3	533.42	1	533.42	6.2	0.0222
   X4	113.74	1	113.74	1.32	0.2646
   X1X2	239.9	1	239.9	2.79	0.1114
   X1X3	474.41	1	474.41	5.51	0.0299
   X1X4	655.42	1	655.42	7.61	0.0125
   X2X3	271.51	1	271.51	3.15	0.0918
   X2X4	2.94	1	2.94	0.034	0.8554
   2 X1	1648.51	1	1648.51	19.15	0.0003
   Residual	1635.7	19	86.09
   Lack of fit	1635.67	14	116.83	27874.73	< 0.0001
   Pure error	0.021	5	0.00419
   Total cor.	7003.51	29
   R-squared= 0.7664 Adj R-squared = 0.6435

   Table 5. Results of analysis of variance (ANOVA) for operating cost model

   Source	Sum of squares	Degree of freedom	Mean square	F- value	p-value
   Model	4.61	7	0.66	7.99	< 0.0001
   X1	1.77	1	1.77	21.44	0.0001
   X2	1.79E-05	1	1.79E-05	2.18E-0	0.9884
   X3	0.15	1	0.15	1.78	0.1963
   X4	1.89	1	1.89	22.91	< 0.0001
   X1X3	0.69	1	0.69	8.34	0.0085
   X2X3	0.12	1	0.12	1.43	0.2448
   2 X1	4.53E-03	1	4.53E-03	0.055	0.8169
   residual	1.81	22	0.082
   Lack of fit	1.7	17	0.1	4.57	0.0504
   Pure error	0.11	5	0.022
   Totalcor.	6.43	29
   R-squared = 0.7178 Adj R-squared = 0.6280

   Design-Expert® Software

   Turbidityremoval 98.3596

   30.1124

   X1 = A: current density X2 = C: pH

   Turbidity removal

   Turbidity removal

   Actual Factors

   B: conductivity= 4.00

   D: electrolysis time = 20.00

   97

   80.25

   63.5

   46.75

   30

   9.00

   8.50

   8.00

   10.41

   11.98

   13.54

   C: pH (X3)

   7.50

   7.00

   7.29

   8.85A: current density (X1)

   (a)

   Design-Expert® Software Turbidityremoval

   Design Points 98.3596

   30.1124

   X1 = A: current density X2 = C: pH

   Actual Factors

   B: conductivity= 4.00

   D: electrolysis time = 20.00

   9.00

   8.50

   8.00

   Turbidity removal

   86.3253 91.3618

   6

   81.2889

   C: pH (X3)

   C: pH (X3)

   7.50 76.2524

   71.2159

   7.00

   7.29 8.85 10.41 11.98 13.54

   A: current density (X1)

   (b)

   Figure 2. The 3D surface graph (a) and contour plot (b) for interactive effect of current density and pH on turbidity removal

   Design-Expert® Software

   Turbidity removal 98.3596

   30.1124

   X1 = A: current density X2 = D: electrolysis time

   Actual Factors

   B: conductivity = 4.00 C: pH = 8.00

   96

   Turbidity removal

   Turbidity removal

   79.5

   63

   46.5

   30

   25.00

   22.50

   20.00

   10.41

   11.98

   13.54

   D: electroly sis time (X41)7.50

   15.00

   7.29

   8.85A: current density (X1)

   Design-Expert® Software Turbidity removal

   Design Points 98.3596

   30.1124

   25.00

   D: electrolysis time (X4)

   D: electrolysis time (X4)

   22.50

   (a)

   Turbidity removal

   91. 1639

   X1 = A: current density X2 = D: electrolysis time

   Actual Factors

   B: conductivity = 4.00 C: pH = 8.00

   20.00

   81. 8031

   86. 4835

   91. 1639

   6

   77. 1227

   17.50

   72. 4423

   15.00

   7.29 8.85 10.41 11.98 13.54

   A: current density (X1)

   (b)

   Figure 3. The 3D surface graph (a) and contour plot (b) for interactive effect of current density and electrolysis time on turbidity removal

   Design-Expert® Software

   operating cost 2.39473

   0.48813

   X1 = A: current density X2 = C: pH

   operating cost

   operating cost

   Actual Factors

   B: conductivity= 4.00

   D: electrolysis time = 20.00

   1.8

   1.45

   1.1

   0.75

   0.4

   9.00

   8.50

   8.00

   10.41

   11.98

   13.54

   C: pH (X3)

   7.50

   7.00

   7.29

   8.85A: current density (X1)

   Design-Expert® Software operating cost

   Design Points 2.39473

   0.48813

   9.00

   8.50

   (a)

   operating cost

   X1 = A: current density X2 = C: pH

   Actual Factors

   B: conductivity= 4.00

   D: electrolysis time = 20.00

   8.00

   C: pH (X3)

   C: pH (X3)

   0.925354

   1.0849 1.24444

   6

   1.40399

   7.50

   1.56353

   7.00

   7.29 8.85 10.41 11.98 13.54

   A: current density (X1)

   (b)

   Figure 4. The 3D surface graph (a) and contour plot (b) for interactive effect of current density and pH on operating cost

   From the numerical optimization that was carried out, 9.9 mA/ cm2, 5 mS/cm, 9 (pH) and 18 min were estimated as the optimum conditions. Under these conditions, the predicted maximum turbidity removal was found to be 96.50% and the minimum operating cost to achieve this was also calculated to be 1.0627 US$/m3. The optimization resultsobtained were validated experimentally by using the obtained optimum input parameters to run the experimental setup and it was discovered from the validation experiment carried out that the optimum conditions of 96.94% turbidity and 78.77% COD removals were achieved at the operating cost of 0.654 US$/m3.

4. CONCLUSIONS

   From the results obtained in this work,it has been discovered that Central Composite Design of Response Surface Methodology has been successfully applied to the electrocoagulation process used for the treatment of a petroleum refinery wastewater. The turbidity removal and the operating cost were found to be largely affected by the current density and the electrolysis time. Also, the initial pH of the wastewater was found to influence the system responses (the turbidity removal and the operating cost) significantly. The results of the ANOVA carried out showed that the turbidity removal and the operating cost models were significant with p-values of 0.0003 and less than 0.0001,

   respectively. In addition, good correlation coefficients of 0.76 and 0.72 were obtained respectively for the turbidity removal and the operating cost models. The optimum conditions of the treatment were obtained to be 9.9 mA/cm2, 5 mS/cm, 9 and 18 min for the current density, the conductivity, the pH and the electrolysis time, respectively. Under these conditions, 96.94% of turbidity and 78.77% of chemical oxygen demand (COD) removals were achieved experimentally at the operating cost of 0.654 US$/m3.

   ACKNOWLEDGEMENTS

   Abdulwahab Giwa and Saidat Olanipekun Giwa wish to acknowledge and appreciate the supports received from the Prime Ministry of The Republic of Turkey, Presidency for Turks Abroad and Related Communities for their Postdoctoral Research and PhD programmes, respectively. In addition, Saidat Olanipekun Giwa wishes to thank the Scientific Research Project Office of Ankara University (Ankara Ãœniversitesi Bilimsel Aratrma Projeleri) for providing the financial support for her PhD Research.

   REFERENCES

   1. Chen, X. and Deng, H. 2012. Removal of humic acids from water by hybrid titanium-based electrocoagulation with ultrafiltration membrane processes. Desalination, 300, 51-57.

   2. Coelho, A., Castro, A.V., Dezotti, M. and SantAnna Jr., G.L. 2006. Treatment of petroleum refinery sourwater by advanced oxidation processes. Journal of Hazardous Materials, 137(1), 178-184.

   3. El-Naas, M.H., Al-Zuhair, S., Al-Lobaney, A. and Maklouf, S. 2009. Assessment of electrocoagulation for the treatment of petroleum refinery wastewater. Journal of Environmental Management, 91(1), 180-185.

   4. Katal, R. and Pahlavanzadeh, H. 2011. Influence of different combinations of aluminum and iron electrode on electrocoagulation efficiency: Application to the treatment of paper mill wastewater. Desalination, 265(1-3), 199205.

   5. Olmez, T. 2009. The optimization of Cr(VI) reduction and removal by electrocoagulation using response surface methodology. Journal of Hazardous Materials, 162(2-3), 1371-1378

   6. Tchamango, S., Charles P. Nanseu-Njiki, C.P., Ngameni, E., Hadjiev, D. and Darchen, A. 2010. Treatment of dairy effluents by electrocoagulation using aluminium eletrodes. Science of the Total Environment, 408(4), 947-952.

   7. Zaied, M. and Bellakhal, N. 2009. Electrocoagulation treatment of black liquor from paper industry. Journal of Hazardous Materials, 163(2-3), 995-1000