Fenton oxidative treatment of petroleum refinery wastewater: process optimization and sludge characterization

Abstract

Sludge generation is one of the major concerns in Fenton treatment of recalcitrant wastewaters. In this work, mineralization (reduction of Dissolved Organic Carbon (DOC)) of petroleum refinery effluent (PRE) and the accompanying sludge generated in treating PRE was investigated. The mineralization level of 53% at optimum conditions produced 0.16 L of wet sludge per litre of wastewater. Further investigation of the sludge properties was conducted to assess the ease of handling and potential reuse of the sludge. For this purpose, sludge generated at studied conditions exceeding 30% DOC reduction were characterized by sludge volume index (SVI), sludge settling rates (V_s) and volumes of settled sludge within the first 30 minutes (SSV₃₀). High V_s (≈0.16 cm s⁻¹) and low SVI (<100 mL g⁻¹) and SSV₃₀ (<50 min s⁻¹) values indicated that the sludge has good settling and compaction properties. The results confirmed that the sludge generated could be managed systematically with appropriate operating conditions.

---

1. Introduction

Petroleum refinery effluent (PRE) contains hundreds of various organic pollutants ranging from aliphatic to aromatic compounds. Many of these compounds resist microbial degradation and are not treatable biologically. This results in low biodegradability of this wastewater due to high Chemical Oxygen Demand (COD) and low Biochemical Oxygen Demand (BOD₅).^1–3 Generally, the organic compounds present in PRE are carcinogenic which causes considerable damage to the ecosystem and human health.^4,5 Disposal of improperly treated PRE into water-bodies results in decreased algae productivity and depleted oxygen level in receiving water bodies which affect the food chain.^6–8 The environmental concern of these discharges is related to their large volume, high toxicity and lack of biodegradability.^1,9

A critical review of the processes currently adopted for treatment of PRE shows that these methods are bedevilled with many problems.² The primary drawbacks that are associated with these treatment methods are (i) inability to mineralize contaminants;¹⁰ (ii) phase transfer of contaminants from one medium to another; (iii) generation of a large amount of sludge; (iv) low treatment efficiencies and (v) slow reaction rates.^11,12 In this regard, Fenton oxidation, an advanced oxidation process (AOPs), that generates highly aggressive hydroxyl radicals (˙OH) through catalytic decomposition of hydrogen peroxide by iron species is expected to offer better treatment efficiencies. Literature is replete with the equations governing Fenton oxidation.^13–15 In this oxidation system: the reagents are easy to handle, the treatment process is relatively inexpensive, the reactors are simple and minimal control and operation are required.^16,17 A large number of studies have used this strong oxidative method for treatment of sludge¹⁸ and various industrial wastewaters.^13,19,20 However, based on the articles that have been reviewed, literature on Fenton oxidative treatment of PRE is scarce.² Recently, our group has reported the feasibility of Fenton oxidative mineralisation of this wastewater, omitting the pre-treatment stage¹⁷ and in another study, assessed the effectiveness Fenton process in treatment of PRE when combined with a sequencing batch reactor.²¹ However, generation of ferric hydroxide sludge at pH above 4.0 is an environmental and economic issue associated with this process. Several strategies such as recycling as a coagulant, using in construction or industry or disposing through land filling²² that have been used for sewage sludge management can be studied for this iron-containing sludge to address this matter. In addition, Fenton sludge can be used as a base material for goethite production^23,24 which is a potential heterogeneous catalyst for water purification.²⁵ Though, the characteristics of the generated sludge in terms of settling and compaction should be identified under the influence of initial carbon load of the wastewater and the amount of Fenton reagents utilized in the treatment system. Considering that a typical PRE is recalcitrant, it appears imperative to quantify and study the sludge characteristics to properly establish the suitability of Fenton oxidation treatment of PRE. The main characteristics of good sludge include (i) low volume;¹⁰ (ii) high density and (iii) easily settleable and dewaterable. The important parameters necessary to establish the sludge characteristics are: sludge settleability index (SVI), the volume of the settled sludge in the first 30 minutes (SSV₃₀) and sludge settling rate (V_s). Sludge generation can be minimized by optimizing the operating conditions.²⁶

One-factor-at-a-time (OFAT) approach is a conventional method to optimize the multifactor experiments by varying one factor while keeping the others constant. In this method, the interaction effects among the experimental variables are not evaluated and a large number of experiments is required. Therefore, it is slow and expensive. Response surface methodology (RSM) has thus been introduced to solve these limitations by optimizing the reaction variables by taking into account the multi-dimensional interactions of the experimental factors. Although this approach has been widely used to design Fenton experiments for treatment of various wastewaters and optimization of the main parameters,^26–28 the reported works on generated sludge from Fenton treatment of industrial waste streams are scarce.

Many studies focus on improving the economic feasibility of Fenton oxidation for wastewater treatment. One area of the focuses is minimization of sludge generation through optimization of the reaction. Prior to sludge management studies, the characteristics of the generated sludge should be identified. So far, there has not been any comprehensive report on synergistic effects of Fenton reagents and organic load on quantity and characteristics of generated sludge. Herein, the effects of [H₂O₂] : [COD], [H₂O₂] : [Fe²⁺] molar ratios and Fenton reaction duration on dissolved organic carbon (DOC) removal efficiency (%) and produced sludge (mL L⁻¹) were investigated using Central Composite Design (CCD). Subsequently, two statistical models were developed for both the target responses and their significance was verified according to the defined optimum conditions. Lastly, the characteristics of the generated sludge in terms of settling and compaction were determined. The present work aims at presenting (i) the efficiency of Fenton oxidation process in treating raw PRE;¹⁰ (ii) the trend of sludge generation in the treated wastewater and (iii) the sludge characterization of the treated wastewater. The main objective is to further assert the suitability of Fenton process as a green technology for PRE treatment based on the reduction in sludge volume and easy separation of sludge from the Fenton reaction medium under optimum condition.

2. Materials and methods

2.1. Chemicals

All the reagents were of analytical grade and used without further purification. Hydrogen peroxide (H₂O₂, 30% w/w), ferrous sulphate (FeSO₄·7H₂O), sodium hydroxide (NaOH) and sulphuric acid (H₂SO₄) were purchased from Merck. Catalase (10 000–40 000 units per mg protein) was purchased from Sigma-Aldrich. The ferrous (Fe²⁺) catalyst was added to the reaction solution by dissolving FeSO₄ in acidified wastewater solution.

2.2. Fenton experiments

PRE was sourced from Malaysian National Refiner. The initial physiochemical properties of the PRE are given in Table 1.

Table 1 The PRE physico-chemical characteristics

Parameter	BOD5	COD	DOC	BOD5/COD	pH	TSS	Turbidity	Oil & grease
a Dimensionless.b NTU.
Value (mg L^−1)	280	930	332	0.3	8.2^a	124	194^b	233

---

Fenton experiments were carried out in batch using a 250 mL borosilicate glass beaker with 100 mL of PRE per batch at the initial pH of 3.0 ± 0.1 (known as the optimum pH¹³) and temperature of 25 °C. The PRE was acidified using sulphuric acid and it was measured using a Cyber Scan pH meter (Eutech, Thermo Fisher Scientific). At the next step, a fixed amount of the catalyst was added and Fenton reaction was initiated by addition of predetermined quantities of H₂O₂ solution under constant magnetic stirring to homogenize the mixture and avoid heat build-up.²⁹ At the end of the stipulated time period, the pH was readjusted to 10.0 ± 0.5, leading to ferric hydroxide precipitation. A sedimentation period of 2 hours was allowed under quiescent conditions. The residual H₂O₂ was removed from the solution using a few drops of catalase enzyme (10 v/v) and confirmed using H₂O₂ test strips (Merckoquant, Merck). A 0.22 μm Millipore filter was used to remove the remnants of the ferric hydroxide precipitate before the dissolved organic carbon (DOC) reduction was monitored. The DOC values were obtained by a TOC analyser (Shimadzu TOC-V_CSH, Japan) equipped with an auto-sampler. The measurements were done at least in duplicate and the results were expressed as mean values of the measurements with an experimental error below 5%.

2.3. Sludge characteristics

The methodology reported by Mahiroglu et al.³⁰ and Besra et al.³¹ was adopted for sludge characterization. It basically involves two stages: a sedimentation test followed by a filtration stage. In the first stage, after the coagulation step, the treated wastewater was transferred to a 100 mL graduated cylinder and allowed to settle after the measuring cylinder was inverted five times. The V_s, SSV₃₀, SSV and SVI were determined as follows:

(i) The height of the sludge sample settling in the graduated cylinder was monitored at regular time intervals and the sludge settling velocity (V_s) was estimated from the slope of the settling height versus time.

(ii) The volume of the sludge recorded at the end of the initial 30 minutes of settling time represented the SSV₃₀.

(iii) The sludge was then allowed to continue settling for 24 hours (without disturbance) after which the volume of the settled sludge (SSV) was obtained. Then, the final volume occupied by the settled sludge was read directly from the graduated cylinder at the end of the settling time.

(iv) The sludge volume index (SVI) was measured by dividing the wet volume (mL L⁻¹) of the sludge by the dry weight concentration of mixed liquor suspended solids (MLSS) in grams per L. The MLSS content of the sludge was determined gravimetrically and used in SVI calculation. This parameter is generally used to express the sludge dewatering potential.

2.4. Design of experiments (DOE)

The amount of sludge generated and the related quality parameters are essential to determine the suitability of excluding the pre-treatment step. To achieve this, Design Expert Software (version 9.0.3) was used for the statistical design of the experiments, development of regression models, data analysis and optimization of Fenton reaction. Response surface methodology was used to investigate the individual and interactive effects of three independent variables: (A) [H₂O₂] : [COD] molar ratio, (B) [H₂O₂] : [Fe²⁺] molar ratio and (C) the reaction time on DOC removal efficiency (%) and the amount of generated sludge (mL L⁻¹). For the three variables, a face centred CCD that consisted of 8 factorial points, 6 axial points and 3 replicates at the centre points were employed. The centre points were used to estimate the experimental error and the duplicability of the data. The axial points were placed at (±α, 0, 0), (0, ±α, 0) and (0, 0, ±α) where the distance of the axial point from the centre is α. The value of α depends on the number of points in the factorial portion of the design. In this study, the star points were at the centre of each face of the factorial space, so α = ±1. The ranges between [H₂O₂] : [COD] molar ratio of 5–20, [H₂O₂] : [Fe²⁺] molar ratio of 2–20 and an oxidation reaction time of 10–180 min were chosen based on the preliminary studies and the literatures.^17,21

A total of 17 experiments were performed that three of them were at the centre point: (A) = 12.5, (B) = 11 and (C) = 95 min. The following second-order polynomial equation (eqn (1)) was used to predict the studied variable factors as a function of independent variables and the interaction among them:

where, Y is the predicted dependent variable, b₀ is constant coefficient, b_i, b_ii and b_ij are regression coefficients, i and j are index numbers, k is number of patterns, X′_is are independent variables and ε is the random error. The analysis of variance (ANOVA) was used to assess the significance and adequacy of the model. The fitness of the polynomial models was expressed by the coefficients of determination, R², R_adj² and R_pred². The main indicators that were used to show the significance of the model were Fisher variation ratio (F-value), probability value (Prob > F) with 95% confidence level and adequate precision. The final model for each response was obtained after elimination of insignificant terms (p > 0.05) based on F-test and the 3-D plots were presented. Furthermore, the optimum values for independent variables were identified and further experiments were performed to verify the regression models.

3. Results and discussion

3.1. Experimental design and ANOVA analysis

The results obtained from the Fenton experiments and the predicted values by the developed models for the studied dependent variables are presented in Table 2.

Table 2 Experimental design matrix and response based on experimental runs and predicted values

Run no.	Experimental design			Results
				Experimental		Predicted
	[H2O2] : [COD] (A)	[H2O2] : [Fe^2+] (B)	Time (min) (C)	DOC (%)	Sludge (mL L^−1)	DOC (%)	Sludge (mL L^−1)
1	5	20	180	27	205	28	219
2	20	20	180	58	180	59	171
3	5	11	95	35	305	36	276
4	12.5	11	10	41	240	44	233
5	20	20	10	43	195	44	198
6	12.5	20	95	45	210	44	202
7	20	11	95	69	280	66	290
8	12.5	11	95	55	235	56	241
9	5	2	180	35	330	35	331
10	12.5	11	95	54	225	56	241
11	20	2	10	51	480	50	470
12	5	20	10	14	265	12	264
13	12.5	11	180	63	200	59	188
14	12.5	2	95	52	405	51	394
15	5	2	10	20	380	20	394
16	20	2	180	61	420	63	425
17	12.5	11	95	55	225	56	240

---

From the interaction results of the variables, two multivariate models were derived to describe the DOC reduction and the accompanying sludge generation (eqn (2) and (3)):

DOC = −3.20861 + 3.98933[H₂O₂] : [COD] + 1.68979[H₂O₂] : [Fe²⁺] + 0.20742 × time + 5.55556 × 10⁻³[H₂O₂] : [COD] × [H₂O₂] : [Fe²⁺] − 5.88235 × 10⁻⁴[H₂O₂] : [COD] × time + 4.90196 × 10⁻⁴[H₂O₂] : [Fe²⁺] × time − 0.079249[H₂O₂] : [COD]² − 0.098244[H₂O₂] : [Fe²⁺]² − 6.16989 × 10⁻⁴ × time² (2)

Sludge = +476.59834 − 12.53592[H₂O₂] : [COD] − 20.06745[H₂O₂] : [Fe²⁺] + 0.39127 × time − 0.52778[H₂O₂] : [COD] × [H₂O₂] : [Fe²⁺] + 6.86275 × 10⁻³[H₂O₂] : [COD] × time + 5.71895 × 10⁻³[H₂O₂] : [Fe²⁺] × time + 0.74491[H₂O₂] : [COD]² + 0.70249[H₂O₂] : [Fe²⁺]² − 4.23510 × 10⁻³ × time² (3)

The synergistic or antagonistic effects are shown by positive and negative signs in the equations respectively. Based on these models, DOC removal efficiency (%) and the amount of sludge can be predicted as a function of the [H₂O₂] : [COD] and [H₂O₂] : [Fe²⁺] ratios and the oxidation time. All of the model terms were found to be statistically significant (p < 0.05) and therefore included in the models. To assess the “goodness of Fit” of the polynomial models, Analysis of Variance (ANOVA) was performed (Table 3). Based on the results, it can be seen that the p-values for both responses were less than 0.05 (<0.0001), which indicated the models were significant and could be used for response prediction. Regression coefficients, R², adjusted R² and predicted R², were used to evaluate the quality of the developed equations. The adjusted R² values that display the total variation of the responses were 0.9678 and 0.9569 for DOC removal (%) and the generated sludge respectively. The R² value close to 1.0 is desirable as it shows the acceptable adjustment of the suggested model with the experimental data. The DOC removal (%) and sludge regression coefficients (R² = 0.9859 and 0.9812) and predicted R² (0.8693 and 0.8404) indicated that the models were highly reliable in terms of repetition of the experiments and also adequate in representing the actual relationship between the responses and variables. In addition, Pred-R² and Adj-R² were in good agreement where the difference between them was less than 0.2.³² Besides, the obtained results for adequate precision that is the signal to noise ratio, were 25.64 and 20.69 respectively, which are greater than 4.0 (desirable value). This indicates adequate signal and shows that the models could be used to navigate the design space.

Table 3 Analysis of variance (ANOVA) results for the studied responses

Source	Sum of squares	Degrees of freedom	Mean square	F-value	p-value (Prob > F)
DOC removal
Model	3757.42	9	417.49	54.48	<0.0001
Residual	53.64	7	7.66
Lack of Fit	52.97	5	10.59	31.78	0.0308
Pure error	0.67	2	0.33
Cor total	3811.06	16
Sludge generation
Model	1.299 × 10^5	9	14 431.37	40.50	<0.0001
Residual	2494.15	7	356.31
Lack of Fit	2427.48	5	485.50	14.56	0.0655
Pure error	66.67	2	33.33
Cor total	1.324 × 10^5	16

---

The lack of fit shows the variation of data around the fitted model.³³ In the sludge model, the “Lack of Fit F-value” of 14.56 implied that the model was not significant relative to the pure error. The insignificant lack of fit was good and it showed that this model was fit to predict the amount of the produced sludge within the studied range of variables.³² On the contrary, although the “Lack of Fit” was significant (“F-value” was 31.78) for the DOC removal model, the model could still be used for design space navigation defined by the CCD, due to good agreement between the adjusted and predicted R² values.²⁷

To ascertain the adequacy of the models in exploring the response surface arising from the variability observed in the data owing to random error, residuals were evaluated in addition to the regression coefficients. The plots of (i) studentized residuals and the predicted values,¹⁰ (ii) predicted versus actual values and (iii) normal plot of residuals were of interest in this study.

The plot in Fig. 1 shows the studentized residuals versus predicted values for (a) DOC% and (b) sludge generation. The figure shows that there is random scattering of the points instead of a funnel-shaped pattern. It suggests that the variance of original observations were constant for all values of the response.^34,35 It was found that all the studentized residuals were within the range of ±3.00 for both models. This shows that the model approximation of the response surface was satisfactory and not associated with data recording error.

Fig. 1 Plot of studentized residuals against the predicted values for (a) DOC% and (b) sludge generation.

The plot of the predicted vs. actual values for (a) DOC removal (%) and (b) generated sludge are presented in Fig. 2. The proximity of the points to the line shows the effectiveness of the quadratic models in capturing the correlation between the concentrations of Fenton reagents, the carbon load of the waste stream and the reduction in DOC (%) and sludge generation.

Fig. 2 Predicted vs. actual values plot in terms of (a) DOC removal (%) and (b) sludge generation.

Lastly, normal probability plots that is a suitable graphical method for judging the normality of the residuals, as indicated in literature.³⁶ Fig. 3 shows the diagnostic plots of the studentized residuals for DOC (%) and sludge generation. The exemplified figures show that the residual behaviour followed a normal distribution. Normal distribution indicates that the qualitative and quantitative agreement are satisfactory, which is an important assumption for validating statistical modelling.³⁶ From the plots (Fig. 1–3), the assumptions of normality, independence and randomness of the residuals were satisfactory.

Fig. 3 Normal probability plot of studentized residuals of the linear model for mineralisation for (a) DOC% and (b) sludge generation.

3.2. Mineralization

The synergistic effects of [H₂O₂] : [COD] and [H₂O₂] : [Fe²⁺] molar ratios on PRE mineralization are shown in Fig. 4. The figure shows that the increase in the [H₂O₂] : [COD] ratio brought about an increase in the rate of PRE mineralization and the maximum mineralization was achieved by a simultaneous increase in both ratios. However, at a [H₂O₂] : [COD] ratio exceeding 12.5, mineralization was slower as a consequence of higher concentration of H₂O₂ within the reaction mixture. The suppression of oxidation reactions with an increase in the oxidant concentration has been observed by many authors.^26,37 It is generally attributed to the scavenging of hydroxyl radicals according to reaction (4):

H₂O₂ + ˙OH → H₂O + HO₂˙ (4)

Fig. 4 The 3-D response surface of the effect of [H₂O₂] : [Fe²⁺] and [H₂O₂] : [PRE] on DOC removal (%).

The perhydroxyl radical (HO₂˙) produced in eqn (1) potentially reduces the available active hydroxyl radicals by further scavenging ˙OH, as shown in eqn (5):

HO₂˙ + ˙OH → H₂O + O₂ (5)

For [H₂O₂] : [Fe²⁺] ratio, an increased amount of catalyst was favoured for the interaction with a high oxidant concentration in the mixture by enhancing the formation of ˙OH radicals.³⁷ However, concentrations above the optimal value resulted in ˙OH radical scavenging and formation of a large amount of sludge at the end of the treatment¹³ (eqn (6)):

˙OH + Fe²⁺ → Fe³⁺ + OH⁻ (6)

The observed low mineralization with low amount of Fenton reagent is related to poor generation of hydroxyl radicals³⁸ and quenching of available free radicals by a high number of probe molecules. Under these conditions, the generated ˙OH radical is insufficient to treat PRE that is composed of aromatics and conjugated C=C double bonds which are difficult to be oxidized in comparison to aliphatic or single bond compounds.³⁹ However, increasing the Fenton reagents up to optimal values increases production of ˙OH and subsequent ring opening to promote mineralization.³⁹

The optimal values suggested by the DOE software for simultaneous high DOC removal and low sludge production were: [H₂O₂] : [COD] = 12.5, [H₂O₂] : [Fe²⁺] = 15 and reaction time = 180 min. With a fixed value for the initial COD of the effluent (930 mg L⁻¹), increase in [H₂O₂] : [COD] ratio contributed to higher amount of H₂O₂ and consequently, lower values for [H₂O₂] : [Fe²⁺] ratio indicated higher iron concentrations in the treatment. Therefore, the criteria to choose optimal values were to maximize the amount of mineralization with the least sludge production by using as low concentrations of Fenton reagents as possible for effective and lucrative treatment. The values of 55% DOC removal and 156 mL L⁻¹ sludge production were predicted at the optimum points. Two additional experiments were conducted under the CCD-predicted optimum conditions to confirm the validity of the results and accuracy of the models. The results obtained from the experiments were 53% and 160 mL L⁻¹ with 3.6 and 2.5 deviation errors (<5%) respectively which were in a close agreement with the predicted values.

The biodegradability of wastewater was effectively enhanced (from 0.3 to 0.47) at optimum conditions with DOC reduction of 53% and an attractive value of 160 mL L⁻¹ for generated sludge. As concluded by Ballesteros Martin et al.,⁴⁰ biodegradability enhancement started beyond 30% of mineralization. This could be the minimum value for initiating further biological oxidation. Therefore, as observed in Table 2, there were several operating conditions under which sufficient mineralization could be achieved. Some of these conditions may then be adopted to allow minimal mineralization followed by coupling with the existing traditional biological treatment for refining the wastewater until the legal discharge limits are achieved.³⁸ Many researchers have adopted this combined process and demonstrated its effectiveness in further reducing the wastewater contamination level.^{21,38,41–44}

3.3. Sludge generation

Fenton treatment of wastewater involves four stages: oxidation, neutralization, coagulation and sedimentation.^14,30,45 Amongst them, the oxidation step, which is responsible for wastewater mineralization, and the coagulation step, where turbidity removal mainly occurs are of concern.¹⁴ Termination of the oxidation reaction by an increase in pH serves to remove the Fe³⁺ formed during the process in the form of iron oxyhydroxide (FeO(OH)·H₂O) precipitate.¹⁴ From Table 2, a strong variation existed in the results obtained for the volume of generated sludge. Although the initial values of the studied wastewater parameters were constant, the variation in the sludge volume was mainly related to the differences in the amount of Fenton reagents introduced to the system for each run. Accordingly, the differences in sludge generation could be based on the extent of organic content mineralization²⁶ and the amount of iron ions introduced.

Fig. 5 illustrates the response surface plots of sludge generation as a function of the combined synergistic or antagonistic effects of [H₂O₂] : [COD] and [H₂O₂] : [Fe²⁺] molar ratios. It was observed that the reduction in sludge production improved with an increase in the [H₂O₂] : [COD] ratio and/or decrease in iron load. This was associated with sufficient oxidant concentration maintained with an increasing ratio of [H₂O₂] : [COD], thereby consuming enough catalytic iron and resulting in a corresponding reduced final sludge volume.^14,46 A similar observation was made by Benatti et al.,²⁶ who found minimal formation of iron sludge with sufficient oxidation of organic matter. It should be noted that in most studies, the amount of applied [H₂O₂] : [COD] ratio is generally higher than the stoichiometric value, considering the effect of other factors such as the presence of Cl⁻ ions on H₂O₂ consumption.¹³

Fig. 5 The response surface plot of the sludge generation as a function of [H₂O₂] : [COD] and [H₂O₂] : [Fe²⁺] molar ratios.

In addition, based on Fig. 5, regardless of the [H₂O₂] : [COD] ratio, sludge generation steadily decreased with an increasing [H₂O₂] : [Fe²⁺] ratio up to a value of 11. This indicated that at lower ratios, the Fe²⁺ available within the mixture was sufficient to catalyse ˙OH production. Although there might not be sufficient ˙OH radicals for attaining higher mineralization under certain reagent combinations, they were generally sufficient to initiate the Fenton reaction.³⁰

3.4. Sludge characterization

The most important parameters to establish the characteristics of the produced sludge are mainly SVI, V_s and SSV₃₀.⁴⁷ These indices indicate the relationship between the weights of the sludge and the volume occupied by the sludge. They also express the dewatering potential by simple settlement and the extent of sludge separation from the treated wastewater.³⁰ Sludge volume index (SVI₃₀) is the volume in mL occupied by one gram of a suspension after 30 min of settling. As for the clarifier design, SVI is a commonly used variable that provides an insight into obtaining clear wastewater discharge without significant carryover of sludge; as it indicates the sludge settleability in the final clarifier and can be used to detect any change in the sludge settling characteristics. In the present study, as the mineralization studies preceded the sludge analysis, the optimal conditions and runs considered attractive for industrial applications were determined while the sludge analysis was limited to the runs with mineralization exceeding 30% (justification highlighted in 3.2).

The SSV₃₀, SVI and V_s results for the 10 batches analysed are given in Table 4. The SVI values were found to range from 18 to 48 mL g⁻¹, which may have accounted for the high sample settling rates of 0.09–0.16 cm s⁻¹. In general, lower SSV and SVI were accompanied with higher settling rates. This indicated the need for optimization of operating conditions to minimize sludge generation as it affects the settling properties. For instance, a decreased SVI value was observed with higher settling rate, as reflected in run 2 and run 13, which had similar SVI values.

Table 4 Summary of the sludge characterization

Experimental run	SSV30 (mL L^−1)	SVI (mL g^−1)	Vs (cm s^−1)
2	165	18	0.1582
4	280	36	0.1211
5	185	21	0.1309
6	195	26	0.0912
7	265	29	0.1174
11	450	48	0.1055
13	180	18	0.1601
14	360	43	0.1172
16	360	46	0.0985
17	205	28	0.1096

---

Low SSV and SVI values suggest ease of handling of the formed sludge.³⁰ The literature reports that a good SVI value for sludge should be below 100 mL g⁻¹ (ref. 48) or range between 50–150 mL g⁻¹.⁴⁹ This is against the backdrop that sludge with a low SVI presents good sedimentation characteristics while values higher than 100 indicate there is bulking sludge.^47–50 In this work, all the SVI values were found to fall below 100 which indicate good sludge sedimentation properties with good settling and compaction characteristics. On the drying and mechanical dewaterability of the Fenton sludge, Dewil et al.⁵¹ demonstrated that Fenton oxidation improved these properties.

The settling rates were generally fast for most of the samples in which a clear liquid–slurry interface was found to form shortly after shaking the treated wastewater in a measuring cylinder. A similar observation was made by Atmaca et al.⁴⁷ where Fenton sludge possessed high settling properties and was generated in small quantities. This is positive for Fenton oxidation, as it eliminates the need to assist the settling process. Besra et al.³¹ reported an increase in the settling rate by 10-fold by adding polyacrylamide (PAM-C) as a cationic flocculent. However, in their study, the mode of application in the presence of surfactants led to complex interactions and possible precipitate formation in the mixture. Another study suggested recycling of Fenton sludge to promote the formation of relatively dense particle structures, thereby enhancing the settling properties.⁵²

From the findings presented in Table 4, it was evident that the sludge concentration and the extent of mineralization were insignificant, as the variation in SVI was not correlated to V_s and higher sludge volumes did not improve the settling rate.

For PRE mineralization, the choice of operating conditions to yield less sludge is also important. This is against the backdrop that the sludge from Fenton oxidation is effective when recycled in the treatment system as it does not add to the organic load of the coagulation step.⁵² Therefore, it potentially aids in organic matter treatment since additional organic load reduction is achieved at the coagulation stage.¹⁴ This characteristic supports the idea of recycling Fenton sludge instead of using a coagulant to achieve COD reduction.

4. Conclusions

The efficiency of Fenton oxidation in treating raw petroleum refining wastewater with a biodegradability value of 0.3 was investigated. RSM was used to quantify the interaction effects of three parameters: the mass ratios of [H₂O₂] : [COD] and [H₂O₂] : [Fe] and Fenton oxidation time on DOC removal (%) and sludge generation (mL L⁻¹). The optimization of wastewater treatment over 180 min showed an optimal mineralization of 53% with 160 mL sludge per L of treated wastewater. This corresponded to mass ratios of 12.5 and 15 for [H₂O₂] : [COD] and [H₂O₂] : [Fe³⁺], respectively.

In terms of sludge characterization, the results obtained for the volume of the settled sludge within the first 30 minutes (SSV₃₀), sludge settleability index (SVI) and sludge settling rate (V_s) showed that the sludge settled well, with good sedimentation characteristics and compaction properties; thus, it was easily handled. Under the optimum conditions, sludge with low SVI (18 mL g⁻¹) and high settling velocity (0.16 cm s⁻¹) was obtained. Both responses were found to be statistically relevant to the treatment process as reflected in the ANOVA results.

This study could be a principal step for managing Fenton generated sludge. Future studies in this field should focus on various strategies that can be employed to handle this iron-containing sludge in an economic and environmental friendly manner.

Abbreviations

AOPs	Advanced oxidation processes
ANOVA	Analysis of variance
BOD5	Biochemical oxygen demand
BOD5/COD	Biodegradability ratio
CCD	Central composite design
COD	Chemical oxygen demand
DOC	Dissolved organic carbon
MLSS	Mixed liquor suspended solids
OFAT	One factor at a time
PAM-C	polyacrylamide
PRE	Petroleum refinery effluent
RSM	Response surface methodology
SSV30	Volume of settled sludge in the first 30 minutes
SVI	Settleability index
TOC	Total organic carbon
TSS	Total suspended solids
Vs	Sludge settling rate
3-D	Three-dimensional

Acknowledgements

This research was supported by a Postgraduate Research Fund (PPP) with project number PG115-2012B from the University of Malaya.

References

1. A. Coelho, A. V. Castro, M. Dezotti and G. L. Sant'Anna, J. Hazard. Mater., 2006, 137, 178–184.
2. B. H. Diya'uddeen, W. M. A. W. Daud and A. R. Abdul Aziz, Process Saf. Environ. Prot., 2011, 89, 95–105.
3. Y. Yavuz, A. S. Koparal and Ü. B. Öğütveren, Desalination, 2010, 258, 201–205.
4. S. K. Pardeshi and A. B. Patil, Sol. Energy, 2008, 82, 700–705.
5. O. Abdelwahab, N. K. Amin and E. S. Z. El-Ashtoukhy, J. Hazard. Mater., 2009, 163, 711–716.
6. M. H. El-Naas, S. Al-Zuhair, A. Al-Lobaney and S. Makhlouf, J. Environ. Manage., 2009, 91, 180–185.
7. F. K. Attiogbe, M. Glover-Amengor and K. T. Nyadziehe, J. Appl. Ecol., 2007, 11, 155–164.
8. S. W. Poulton, M. D. Krom, J. V. Rijn and R. Raiswell, Water Res., 2002, 36, 825–834.
9. A. Alva-Argáez, A. C. Kokossis and R. Smith, Chem. Eng. J., 2007, 128, 33–46.
10. M. Minella, G. Marchetti, E. de Laurentiis, M. Malandrino, V. Maurino, C. Minero, D. Vione and K. Hanna, Appl. Catal., B, 2014, 154–155, 102–109.
11. M. S. Kuyukina, I. B. Ivshina, M. K. Serebrennikova, A. B. Krivorutchko, E. A. Podorozhko, R. V. Ivanov and V. I. Lozinsky, Int. Biodeterior. Biodegrad., 2009, 63, 427–432.
12. N. A. Laoufi, D. Tassalit and F. Bentahar, Global NEST J., 2008, 10, 404–418.
13. S. Rahim Pouran, A. R. Abdul Aziz and W. M. A. Wan Daud, J. Ind. Eng. Chem., 2015, 21, 53–69.
14. M. Umar, H. A. Aziz and M. S. Yusoff, Waste Manag., 2010, 30, 2113–2121.
15. J. J. Pignatello, E. Oliveros and A. MacKay, Crit. Rev. Environ. Sci. Technol., 2006, 36, 1–84.
16. B. Bianco, I. de Michelis and F. Vegliò, J. Hazard. Mater., 2011, 186, 1733–1738.
17. D. u. B. Hasan, A. R. Abdul Aziz and W. M. A. W. Daud, Chem. Eng. Res. Des., 2012, 90, 298–307.
18. X. Zhou, G. Jiang, Q. Wang and Z. Yuan, RSC Adv., 2014, 4, 50644–50652.
19. S. Rahim Pouran, A. A. Abdul Raman and W. M. A. Wan Daud, J. Cleaner Prod., 2014, 64, 24–35.
20. C. Minero, M. Lucchiari, V. Maurino and D. Vione, RSC Adv., 2013, 3, 26443–26450.
21. B. H. Diya'uddeen, S. R. Pouran, A. R. A. Aziz, S. M. Nashwan, W. M. A. W. Daud and M. G. Shaaban, J. Ind. Eng. Chem., 2015, 25, 186–191.
22. B. M. Cieślik, J. Namieśnik and P. Konieczka, J. Cleaner Prod., 2015, 90, 1–15.
23. H.-H. Huang, M.-C. Lu and J.-N. Chen, Water Res., 2001, 35, 2291–2299.
24. M. Mohapatra and S. Anand, Int. J. Eng. Sci. Res. Technol., 2010, 2, 127–146.
25. I. R. Guimarães, L. C. A. Oliveira, P. F. Queiroz, T. C. Ramalho, M. Pereira, J. D. Fabris and J. D. Ardisson, Appl. Catal., A, 2008, 347, 89–93.
26. C. T. Benatti, C. R. G. Tavares and T. A. Guedes, J. Environ. Manage., 2006, 80, 66–74.
27. S. Rahim Pouran, A. R. Abdul Aziz, W. M. A. Wan Daud and S. Shamshirband, Measurement, 2015, 59, 314–328.
28. I. Arslan-Alaton, G. Tureli and T. Olmez-Hanci, J. Photochem. Photobiol., A, 2009, 202, 142–153.
29. F. J. Rivas, F. J. Beltran, O. Gimeno and J. Frades, J. Agric. Food Chem., 2001, 49, 1873–1880.
30. A. Mahiroglu, E. Tarlan-Yel and M. F. Sevimli, J. Hazard. Mater., 2009, 166, 782–787.
31. L. Besra, D. K. Sengupta, S. K. Roy and P. Ay, Sep. Purif. Technol., 2003, 30, 251–264.
32. A. Arami-Niya, W. M. A. Wan Daud, F. S. Mjalli, F. Abnisa and M. S. Shafeeyan, Chem. Eng. Res. Des., 2012, 90, 776–784.
33. S. Ghafari, H. A. Aziz, M. H. Isa and A. A. Zinatizadeh, J. Hazard. Mater., 2009, 163, 650–656.
34. D. C. Montgomery, Design and Analysis of Experiments, John Wiley & Sons, New York, USA, 7th edn, 2009.
35. R. H. Myers, D. C. Montgomery and C. M. Anderson-Cook, Response surface methodology: Process and product optimization using designed experiments: Process and Product Optimization Using Designed Experiments, John Wiley & Sons, New York, 2nd edn, 2002.
36. K. Cruz-González, O. Torres-Lopez, A. M. García-León, E. Brillas, A. Hernández-Ramírez and J. M. Peralta-Hernández, Desalination, 2012, 286, 63–68.
37. P. Shukla, S. Wang, H. Sun, H.-M. Ang and M. Tadé, Chem. Eng. J., 2010, 164, 255–260.
38. R. C. Martins, A. F. Rossi and R. M. Quinta-Ferreira, J. Hazard. Mater., 2010, 180, 716–721.
39. E. Neyens and J. Baeyens, J. Hazard. Mater., 2003, 98, 33–50.
40. M. M. Ballesteros Martin, J. L. Casas Lopez, I. Oller, S. Malato and J. A. Sanchez Perez, Ecotoxicol. Environ. Saf., 2010, 73, 1189–1195.
41. C. S. D. Rodrigues, L. M. Madeira and R. A. R. Boaventura, J. Hazard. Mater., 2009, 172, 1551–1559.
42. G. Sudarjanto, B. Keller-Lehmann and J. Keller, J. Hazard. Mater., 2006, 138, 160–168.
43. D. Suryaman, K. Hasegawa and S. Kagaya, Chemosphere, 2006, 65, 2502–2506.
44. N. P. Tantak and S. Chaudhari, J. Hazard. Mater., 2006, 136, 698–705.
45. B. Kiril Mert, T. Yonar, M. Yalili Kiliç and K. Kestioğlu, J. Hazard. Mater., 2010, 174, 122–128.
46. Y. Deng and J. D. Englehardt, Water Res., 2006, 40, 3683–3694.
47. E. Atmaca, J. Hazard. Mater., 2009, 163, 109–114.
48. M. Lamine, L. Bousselmi and A. Ghrabi, Desalination, 2007, 215, 127–132.
49. W. Janczukowicz, M. Szewczyk, M. Krzemieniewski and J. Pesta, Pol. J. Environ. Stud., 2001, 10, 15–20.
50. G. Tchobanoglous, F. Burton and H. Stensel, Wastewater Engineering: Treatment and Reuse, Metcalf & Eddy, New York, 2003.
51. R. Dewil, J. Baeyens and E. Neyens, J. Hazard. Mater., 2005, 117, 161–170.
52. H. C. Yoo, S. H. Cho and S. O. Ko, J. Environ. Sci. Health, Part A: Toxic/Hazard. Subst. Environ. Eng., 2001, 36, 39–48.

---