Removal of COD and SO₄²⁻ from Oil Refinery Wastewater Using a Photo-Catalytic System—Comparing TiO₂ and Zeolite Efficiencies

Abstract

Advanced oxidation processes (AOPs) have many prospects in water and wastewater treatment. In recent years, AOPs are gaining attention as having potentials for the removal of different ranges of contaminants from industrial wastewater towards water reclamation. In this study, the treatability efficiencies of two photo-catalysts (TiO₂ and zeolite) were compared on the basis of the removal of chemical oxygen demand (COD) and SO₄²⁻ from oil refinery wastewater (ORW) using photo-catalytic system. The effects of three operating parameters: catalyst dosage (0.5–1.5 g/L), reaction time (15–45 min), mixing rate (30–90 rpm) and their interactive effects on the removal of the aforementioned contaminants were studied using the Box–Behnken design (BBD) of response surface methodology (RSM). Statistical models were developed and used to optimize the operating conditions. An 18 W UV light was incident on the system to excite the catalysts to trigger a reaction that led to the degradation and subsequent removal of contaminants. The results obtained showed that for almost the same desirability (92% for zeolite and 91% for TiO₂), TiO₂ exhibited more efficiency in terms of mixing rate and reaction time requirements. At the 95% confidence level, the model’s predicted results were in good agreement with experimental data obtained.

1. Introduction

Issues of water scarcity and how it can be curbed continues to be a challenge to the world. There is continuous aggravation of water pollution due to population growth and its corresponding effects on industrialization, agriculture, and domestic activities. This has led to radical exploration of alternative and sustainable means of obtaining water from unconventional means, such as water re-use through wastewater treatment [1]. Because wastewater production is inevitable, wastewater treatment for re-use has become the most convenient way of stemming water scarcity. One of the major industrial water pollution sources is the oil refinery industry. Reports indicate that for every 1 barrel of oil processed, about 10 barrels of oily wastewater is produced. In addition, large volumes of oily wastewater is generated during transport and refinery of oil [2]. Apart from salts being the highest in concentration (1000–3500 mg/L) of oily wastewater, organic compounds follow with a concentration range of 15–1500 mg/L, then oil and grease ranging from 2 to 565 mg/L [3]. Consequently, high levels of chemical oxygen demand (COD), total suspended solids (TSS), total dissolved solids (TDS), soap oil and grease (SOG), and traces of metals are found in the oily waste streams, which make their treatment problematic and cause huge pollution problems [4].

Over the years, several treatment methods have been employed to treat oily wastewater. These include membrane filtration, coagulation-flocculation, hydro-cyclones, and biological treatment systems, among others [5,6]. With a focus on organic component removal, biological systems have been preferred due to their low cost of operation and eco friendliness [7]. However, biological treatment processes have shown limitations at high contaminant concentration. In addition, some recalcitrant organic pollutants (such as inert COD) show high resistance to biological degradation processes [8].

With advancements in research, advanced oxidation processes (AOPs) are being explored for many wastewater treatment applications and to fill the gaps left by conventional treatment processes. AOPs are chemical processes that make use of hydroxyl radicals to effect oxidation [9]. Figure 1 shows a brief description of contaminant degradation in an AOP (photo-catalysis). Photo-catalysis is essentially the activity that occurs when a light source interacts with a semiconductor metal oxide [9]. During photo-catalysis, an incident light (UV) causes excitation on the surface of a photo-catalyst. This happens when electrons in the valence band absorb energy higher than their band gap energy and therefore move into the conduction band. When this happens, a hole (h⁺ in the valance band) and electron (e⁻ in the conduction band) are produced simultaneously. These two species (e⁻ and h⁺) then either recombine later or they react with oxygen molecules to form peroxide or with water molecules to form hydroxyl species. The peroxide and hydroxyl formed then attack and degrade organic pollutants into more manageable compounds such as water or carbon dioxide [10,11].

Photo-catalysts such as TiO₂ have been used extensively in wastewater treatment due to their ability to function at ambient temperature, their accessibility, cost effectiveness, and their ability to oxidize most organic contaminants to safer and more manageable compounds such as CO₂ and H₂O [12]. In an experiment to treat olive mill wastewater, El Hajjouji et al. [13] used a TiO₂/UV system to degrade 94% phenol over a 24 h period with 1 g/L of the catalyst. In a similar study, Chatzisymeon et al. [14] studied the effects of operating parameters, including initial organic load of influent, contact time, and concentration of TiO₂ in media on the removal of COD, total phenols, and color from black table olive processing wastewater. The findings showed that the efficiency of the TiO₂ photo-catalysis increased with increasing contact time, catalyst dosage, and decreasing initial organic load.

Zeolites are mainly known for their peculiar ion-exchange, molecular sieving, and adsorption abilities [15,16]. Zeolites have been applied in diverse ways in water purification. They have been applied in heavy metal removal from wastewater [17,18] and for dye removal for textile effluent [19,20]. Zeolites act as photo-catalysts when the zeolite framework is photo-activated or incorporated with other materials such as semiconductors [21]. Their unique ability of promoting stabilization of photochemically generated redox species and controlling charge transfer and electron transfer processes makes them worth exploring for many photo-catalytic applications [22,23].

In this study, the efficiencies of zeolite (photo-activated) and TiO₂ were studied and compared on the basis of the removal of COD and SO₄²⁻ from oily wastewater in a photo-catalytic system. These two contaminants are higher than the permitted discharge limits of 1000 ppm (COD) and 500 ppm (SO₄²⁻), respectively [1,2,3,24]. Three operating parameters were varied catalyst dosage (0.5–1.5 g/L), reaction time (15–45 min), and mixing rate (30–90 rpm). Response surface methodology (RSM) was used to optimize and study the interactions between the operating parameters. The effluent used for the experiment was collected from a local oil refinery effluent treatment plant in South Africa.

2. Materials and Methods

2.1. Effluent Sample and Analytical Methods

Effluents for the experiment were collected from the sewer of the effluent treatment plant of a local oil refinery. Before getting to the sewer, the effluent passes through some preliminary treatment processes. Residual oil is recovered at the oil–water separator and dissolved air floatation units. Other contaminants such as suspended solids are removed using coagulation-flocculation systems. Characterization of the effluent for SOG and COD was performed according to standard methods, as used by [24]. Sulphate determination was done using a HI 83099 COD meter and multi-parameter photometer from Hanna Instruments (Limean, Vaneto, Italy). Conductivity was measured using a TDS meter, model C7 from HJM electronics (Edenval, Gauteng, South Africa). Sulphide and pH analysis was done using HI 4222 research grade bench pH and ion specific (ISE) meter from Hanna Instruments at ambient temperature.

Table 1. Properties of oily wastewater before treatment. COD: chemical oxygen demand, SOG: soap oil and grease.

Component	Value
pH	8.09
Conductivity (mg/L)	613.38
Sulphates (mg/L)	895.00
Sulphide (mg/L)	1.26
COD (mg/L)	1226.8
SOG (mg/L)	33.23

gives the properties of the effluent.

2.2. Chemicals Used

TiO₂ nanomaterial used in this study was supplied by Huntsman Tioxide, South Africa (Pty). The SACHTLEBEN R-KB-6 is a micronized rutile titanium dioxide pigment treated with alumina and zirconia compounds. Detailed properties of the TiO₂ nanomaterial can be found in

Table 2. Properties of TiO₂ nanomaterial.

Properties	Value
White powder content	94% purity
Phase mixture	Rutile 94%, anatase 6%
Surface treatment	Alumina, zirconia
Organic treatment	Present
Surface gravity	4.1 g/cm3
Crystal size	10 nm
Loss at 105 °C	0.60%
Bulk density	1.1 g/cm3
Oil absorption	18 cm3 /100 g pigment
Durability	Highly durable
ISO 591 classification	R2
Chemical Abstract Service (CAS) Number	13463-67-7

. Commercial zeolite powder (95–98% clinoptilolite balance opaline silica) used in this study was supplied by Sigma-Aldrich, South Africa. 1M H₃PO₄ (analytical grade) was used to regulate the pH of the influent at 6. Figure 2 shows the flow diagram for this study.

2.3. Experimental Setup

The experiment was carried out in a six-place jar testing apparatus (JLT 6, flocculation tester, Velp Scientifica, New York, NY, USA) as shown in Figure 3. The setup consist of six identical beakers of 1 L, all equipped with a stirrer. Two 1 inch-diameter radiant fluorescent T8 black light blue bulbs, wavelength (400 nm), with a power rating of 18 W were used as the UV light source to excite the catalysts in order to trigger a reaction. The light was incident on the beakers equally to ensure uniformity of light distribution, as shown in Figure 1.

As per the experimental design, the three operating parameters (catalyst dosage, mixing speed, and reaction time) were varied, and their interactive effects on contaminant removal were analyzed. Catalyst dosage was varied at 0.5, 1.0, and 1.5 g/L. Mixing speed was varied at 30, 60, and 90 rpm, whereas reaction time was varied at 15, 30, and 45 min. The volume of effluent used for each beaker was 1 L. After each experimental run, the mixer was turned off, and the contents were allowed to settle for 30 min. After settling, 10 mL samples of each supernatant were analyzed of COD and SO₄²⁻. Contaminant removal efficiency was calculated using Equation (1).

$$\eta =\frac{A_o-A}{A_o} \times 100\%$$

(1)

where η is contaminant removal efficiency, and A_o and A are the initial and final concentrations (mg/L) before and after treatment, respectively.

2.4. Response Surface Methodology (RSM)

Response surface methodology (RSM) was used in this study. RSM is a useful optimization tool in developing and studying the impact of independent and interactive factors of a process. These include process optimization, enhancement, and development of products [25]. Basically, RSM establishes the relationship between a variable and its effect (response), with the advantage of producing a mathematical model. In this experiment, the Box–Behnken design (BBD) was used. BBD is one of the response surface designs that is widely used apart from the central composite design (CCD). Even though CCD provides high prediction over the entire design space, BBD was preferred for this experiment because with the same number of factors, it generates a smaller number of experimental runs, which translates into less time and resource requirements without compromising accuracy [25].

3. Results and Discussion

By utilizing the BBD (Design Expert software V.11.1.0.1, Stat-Ease Inc., Minneapolis, MA, USA), 15 experimental runs were generated for each catalyst, including three replications. The three different normalized central levels were coded as −1, 0, +1, corresponding to the minimum, central point, and maximum for the factors considered. These experimental runs were randomized in order to eliminate bias.

Table 3. Coded matrix and response values for zeolite.

	Factor A (Coded)	Factor B (Coded)	Factor C (Coded)	Response 1 (COD) %		Response 2 (SO42−) %
Run	Catalyst Dosage (g/L)	Reaction Time (min)	Mixing Rate (rpm)	Actual	Predicted	Actual	Predicted
1	−1	1	0	90.1	90.07	82.87	82.89
2	−1	−1	0	90.28	90.31	82.87	82.81
3	−1	0	−1	91.38	91.35	84.26	84.24
4	0	0	0	88.84	88.84	82.87	82.89
5	0	−1	1	87.07	87.07	84.26	84.33
6	1	1	0	90.1	90.01	82.87	82.81
7	1	0	1	89.08	89.11	87.06	87.12
8	1	−1	0	90.04	90.01	83.2	83.06
9	0	−1	−1	93.24	93.24	81.47	81.54
10	0	1	1	91	91.06	84.26	84.25
11	0	1	−1	88.94	89.01	81.47	81.45
12	1	0	−1	88.37	88.4	78.68	78.74
13	−1	0	1	86.55	86.52	81.47	81.45
14	0	0	0	88.84	88.84	82.87	82.89
15	0	0	0	88.84	88.84	82.87	82.89

and

Table 4. Coded matrix and response values for TiO₂.

	Factor A (Coded)	Factor B (Coded)	Factor C (Coded)	Response 1 (COD) %		Response 2 (SO4) %
Run	Catalyst Dosage (g/L)	Reaction Time (min)	Mixing Rate (rpm)	Actual	Predicted	Actual	Predicted
1	0	−1	1	90.59	90.59	80.45	80.42
2	−1	0	−1	91.51	91.57	81.14	81.11
3	1	0	−1	90.41	90.47	83.94	83.97
4	−1	0	1	91.17	91.11	82.54	82.51
5	0	0	0	88.51	88.51	81.38	81.38
6	0	1	1	89.63	89.76	79.05	79.08
7	1	0	1	90.88	90.82	79.75	79.78
8	−1	1	0	91.25	91.19	82.54	82.54
9	0	0	0	88.51	88.51	81.38	81.38
10	−1	−1	0	89.96	90.02	81.14	81.21
11	0	−1	−1	90.01	89.88	81.84	81.81
12	1	−1	0	90.49	90.55	83.94	83.94
13	0	0	0	88.51	88.51	81.38	81.38
14	1	1	0	89.33	89.27	80.01	79.94
15	0	1	−1	90.59	90.59	80.44	80.47

show the experimental results for zeolite and TiO₂, respectively. Analysis of variance (ANOVA) was carried out to check the validity and accuracy of the fitted model.

3.1. Model Fitting and Statistical Analysis

The reduced form of the models generated by the Design Expert software for contaminant removal in coded form is represented by Equations (2)–(5). Equations (2), (3), and (5) are quadratic models, whereas Equation (4) is a two-factor interaction (2FI) model. The model reduction and modification was necessary to enhance the predictability of the response variables. The models’ automatic selection and modification was based on Akaike information criterion (AICc) with a correction for a small design. AICc is multiple-model selection method and criteria used to choose the best model required. The models present a direct relationship between the dependent variables (COD and SO₄²⁻ removal) and the independent variables (catalyst dosage (A), reaction time (B), and mixing rate (C)). With these models, COD and SO₄²⁻ removal can be predicted within the designed space.

$$COD \left(Z\right)=88.84-0.09A-0.06B-1.03C+0.06\mathit{AB}+1.39\mathit{AC}+2.06\mathit{BC}+1.25B²,$$

(2)

$$COD \left(Ti{O}_2\right)=88.51-0.35A-0.03B-0.03C-0.61AB+0.20AC-0.39BC+1.27A²+0.48B²+1.22C²,$$

(3)

$$S{O}_4 \left(Z\right)=82.89+0.04A-0.04B+1.40C-0.08AB+2.79AC,$$

(4)

$$S{O}_4 \left(Ti{O}_2\right)=81.38+0.03A-0.67B-0.70C-1.33AB-1.40AC+0.96A²-0.43B²-0.50C².$$

(5)

The extent to which terms of the models can affect the response or contaminant removal are associated with the positive and negative coefficient of the terms. Negative coefficients specify undesirable effects of the factors, whereas positive coefficients show that a factor or combination of factors contribute favorably to the contaminant removal. Again, the magnitudes of the coefficients correlate with the degree to which the response variable is affected.

To check the statistical accuracy and validity of the model, analysis of variance (ANOVA) was performed for each model, as shown by

Table 5. ANOVA for COD removal by zeolite—reduced quadratic model. CV: coefficient of variation.

Source	Sum of Squares	Degree of Freedom	Mean Square	F-Value	p-Value
Model	39.05	7	5.58	1614.82	<0.0001	significant
A—Catalyst dosage	0.0648	1	0.0648	18.76	0.0034
B—Reaction time	0.03	1	0.03	8.69	0.0215
C—Mixing rate	8.47	1	8.47	2451.01	<0.0001
AB	0.0144	1	0.0144	4.17	0.0805
AC	7.67	1	7.67	2221.24	<0.0001
BC	16.93	1	16.93	4902.02	<0.0001
B2	5.87	1	5.87	1697.88	<0.0001
Residual	0.0242	7	0.0035
Lack of fit	0.0242	5	0.0048
Pure error	0	2	0
Cor total	39.07	14
R2 0.9994	Adjusted R2 0.9988	CV% 0.0657	Predicted R2 0.9955	Adeq. Pr 156.67	Mean 89.51	SD 0.0588

,

Table 6. ANOVA for SO₄²⁻ removal by zeolite—reduced two-factor interaction (2FI) model.

Source	Sum of Squares	Degree of Dreedom	Mean Square	F-Value	p-Value
Model	46.84	5	9.37	1764.91	<0.0001	significant
A—Catalyst dosage	0.0144	1	0.0144	2.72	0.1334
B—Reaction time	0.0136	1	0.0136	2.56	0.1438
C—Mixing rate	15.6	1	15.6	2938.04	<0.0001
AB	0.0272	1	0.0272	5.13	0.0498
AC	31.19	1	31.19	5876.09	<0.0001
Residual	0.0478	9	0.0053
Lack of fit	0.0478	7	0.0068
Pure error	0	2	0
Cor total	46.89	14
R2 0.999	Adjusted R2 0.9984	CV% 0.0879	Predicted R2 0.9953	Adeq. Pr 181.80	Mean 82.89	SD 0.0729

,

Table 7. ANOVA for COD removal by TiO₂—quadratic model.

Source	Sum of Squares	Degree of Freedom	Mean Square	F-Value	p-Value
Model	14.15	9	1.57	120.91	<0.0001	significant
A—Catalyst dosage	0.9661	1	0.9661	74.28	0.0003
B—Reaction time	0.0078	1	0.0078	0.6007	0.4733
C—Mixing rate	0.0078	1	0.0078	0.6007	0.4733
AB	1.5	1	1.5	115.39	0.0001
AC	0.164	1	0.164	12.61	0.0164
BC	0.5929	1	0.5929	45.59	0.0011
A2	5.93	1	5.93	456.12	<0.0001
B2	0.8507	1	0.8507	65.41	0.0005
C2	5.45	1	5.45	419.12	<0.0001
Residual	0.065	5	0.013
Lack of fit	0.065	3	0.0217
Pure error	0	2	0
Cor total	14.22	14
R2 0.9954	Adjusted R2 0.9872	CV% 0.1266	Predicted R2 0.9268	Adeq. Pr 32.9036	Mean 90.09	SD 0.114

and

Table 8. ANOVA for SO₄²⁻ removal by TiO₂—reduced quadratic model.

Source	Sum of Squares	Degree of Freedom	Mean Square	F-Value	p-Value
Model	27.82	8	3.48	1187.01	<0.0001	significant
A—Catalyst dosage	0.0098	1	0.0098	3.35	0.1171
B—Reaction time	3.55	1	3.55	1212.33	<0.0001
C—Mixing rate	3.88	1	3.88	1323.96	<0.0001
AB	7.1	1	7.1	2424.66	<0.0001
AC	7.81	1	7.81	2666.98	<0.0001
A2	3.42	1	3.42	1167.76	<0.0001
B2	0.6987	1	0.6987	238.52	<0.0001
C2	0.9231	1	0.9231	315.13	<0.0001
Residual	0.0176	6	0.0029
Lack of fit	0.0176	4	0.0044
Pure error	0	2	0
Cor total	27.83	14
R2 0.9994	Adjusted R2 0.9985	CV% 0.0665	Predicted R2 0.9919	Adeq. Pr 116.61	Mean 81.39	SD 0.054

. Typical features of ANOVA are the F-values, p-values, R², and predicted R² values. These values show whether the model is statistically valid or not. As shown in the tables, for all models, the F-values were greater than their respective p-values for each term in the model, indicating the significance of each interaction term in the model. In all instances, the probability for the F-value occurring due to noise was 0.01%. Furthermore, the p-values for all model terms were less than 0.05, which showed the significance of these model terms and can therefore be generalized to a broader range of interest.

The correlation coefficient (R²) defines how close a set of data are to the fitted regression line. This value is of immense statistical significance, as it gives a clear indication of how good a dataset fits a model [26]. The closer the R² value is to 1, the more meaningful the model. As seen from the tables, the R² values (0.9994, 0.999, 0.9954, and 0.9994) and their respective adjusted R² values (0.9988, 0.9984, 0.9872, and 0.9985) are close to 1, suggesting that the developed statistical model fitted well with the data obtained.

Another way to determine the significance of a model is by measuring its adequate precision (Adeq. Pr) and coefficient of variation (CV). Adequate precision is the signal-to-noise ratio which measures the range of the predicted response relative to its associated error. An adequate precision of greater than 4 is generally desired [26]. From the tables above, the values obtained for adequate precision prove the adequacy of the model. The CV is the ratio of the standard deviation to the mean. This value, expressed as a percentage, shows the level of dispersion around the mean. Lower values of CV suggest more precise models [27]. Evidently, the CV values given in the tables above are low enough to ensure significant models.

3.2. Interactive Effects of Parameters

The cross-factor interactive effects of the factors (catalyst dosage, reaction time, and mixing rate) on the response (COD and SO₄²⁻ removal) were studied. Figure 4, Figure 5, Figure 6 and Figure 7 represent the 2D (contour) and 3D (surface) plots resulting from the cross factor interactions on the contaminant removal by the photo-catalysts.

Figure 4 represents the interaction between reaction time and mixing rate (BC). The combined effects of these two parameters have the greatest effect in the COD removal by zeolite. It can be seen, that with a reaction time of 15 to 21 min and mixing rate of 30–50 rpm, an optimum removal efficiency of 93% was achieved with minimum catalyst dosage. As catalyst dosage increased, removal efficiency reduced. This can be explained by the fact that higher catalyst dosage causes lumping of particles, consequently leading to poor light penetration into the system [28]. In Figure 5, the effect of the interactions between catalyst dosage and mixing rate (AC) on COD removal by TiO₂ is shown. AC was proven to have the highest effect on this system, as shown by the quadratic model in Equation (2). It can be inferred from the figure that as catalyst dosage increased, mixing rate also increased. This corresponding increase in mixing rate was needed in order to enhance proper homogenizing of the catalyst in the reacting media.

Figure 6 and Figure 7 represent the plots for cross factor interactive effects on SO₄²⁻ removal by zeolite and TiO₂, respectively. As shown in Figure 6, AC had the highest influence on SO₄²⁻ removal by zeolite, whereas catalyst dosage and reaction time (AB) had the highest influence on SO₄²⁻ removal by TiO₂, as shown in Figure 7.

4. Optimum Conditions

Numerical optimization was done to determine the optimum conditions of the three parameters for contaminant removal. The numerical optimization technique explores the entire design space on the basis of the developed models to detect the optimum factor conditions for the given range. Equations (2)–(5) served as the objective functions, whereas the three independent variables served as constrains. These constrains were set within the given range. The goal for the optimization was to achieve maximum contaminant removal.

Figure 8 represents a ramp graph showing the optimum conditions for the operating parameters and the desirability obtained from the zeolite experiment. As can be inferred from the graph, to achieve maximum COD removal of 92% and SO₄²⁻ removal of 87%, all operating parameters were at the maximum with a desirability of 91.7% contaminant removal. This translates into maximum time and energy requirements to achieve the set goal of contaminant removal for the given range of factors.

Figure 9 shows the ramp graph for the optimum conditions of the operating parameters for the TiO₂ experiment. Unlike zeolite, a catalyst dosage of 1.5 g/L, reaction time of 15 min, and mixing rate of 30 rpm achieved a removal efficiency of 91.21% of COD and 85.5% removal efficiency of SO₄²⁻, leading to a desirability of 91%, which translates into lesser energy and time requirements for achieving maximum contaminant removal for the given range of factors.

Table 9. Comparative studies on photo-catalytic degradation of petroleum refinery effluent.

Photo-Catalytic Degradation of Petroleum Refinery Effluent
Contaminant Investigated	Conditions	Results (Removal %)	Reference
* TCOD	pH (4), catalyst concentration (100 mg/L), temperature (45 °C), and reaction time (120 min)	Over 83	[29]
** DOC	Catalyst concentration (0.2 g/L), reaction time (60 min), reactor volume (100 mL)	21	[30]
COD	Temperature (45 °C), catalyst concentration (100 mg/L), pH (3), reaction time (240 min)	90	[31]
COD	Catalyst concentration (1 g/L), pH (3), temperature (50 °C)	60	[32]
COD	Catalyst concentration (1.5 g/L), pH (6), reaction time (15 min), temperature (25 ± 2 °C), mixing rate (30 rpm)	92	* This study
SO42−	Catalyst concentration (1.5 g/L), pH (6), reaction time (45 min), temperature (25 ± 2 °C), mixing rate (90 rpm)	87

* Total chemical oxygen demand, ** dissolved organic carbon.

shows a comparative study of work done by other authors in photo-catalytic degradation of petroleum refinery effluent. It can be seen that much focus was given to only the organic (COD) aspects of the petroleum refinery effluent without consideration for the possibility of removing other inorganics such as salts simultaneously. This study however considered the possibility of removing both organics (COD) and inorganics (SO₄²⁻) simultaneously from oil refinery effluent. The conditions of the experiment and the results obtained in each case are shown in the table.

5. Conclusions

This study compared COD and SO₄²⁻ (contaminants) removal efficiencies of zeolite and TiO₂ from oily wastewater in a photo-catalytic system. A 2FI model and a quadratic model were developed for SO₄²⁻ and COD removal by zeolite, respectively, whereas quadratic models were developed for SO₄²⁻ and COD removal by TiO₂. The optimum conditions for zeolite were found to be a mixing rate of 90 rpm, reaction time of 45 min, and catalyst dosage of 1.5 g/L. For TiO₂ the optimum conditions were found to be a mixing rate of 30 rpm, reaction time of 15 min, and catalyst dosage of 1.5 g/L. Validity of the models were demonstrated by analysis of variance (ANOVA) and were found to be in agreement with predicted values. The desirability of using zeolite and TiO₂ were almost the same (92% and 91% respectively). Between the two photo-catalysts, TiO₂ was found to be more promising in the treatment of oily wastewater than the zeolite. Therefore, incorporating TiO₂ in a photo-catalytic system optimized with RSM has a bright future in the water and wastewater setting for both organic and inorganic contaminant removal.