Title Indonesian Journal of Environmental Management and Sustainability e-ISSN:2598-6279 p-ISSN:2598-6260 Research Paper Box Behken design for optimization of COD removal from Palm oil mill effluent (POME) using Reverse osmosis (RO) membrane Muhammad Said1*, Muneer M. Ba-Abbad2, Siti Rozaimah Sheikh Abdullah3, Abdul Wahab Mohammad3,4 1Department of chemistry, Faculty of Mathematics and Science,Sriwijaya University, Indralaya, 30662,Sumatera Selatan, Indonesia 2Department of chemical Engineering, Faculty of Engineering and Petroleum, Hadhramout University of Science &Technology, Mukalla, Hadhramout, Yemen 3Department of Chemical and Process Engineering, Faculty of Engineering and Built Environment, Universiti Kebangsaan Malaysia, 43600 Bangi, Selangor, Malaysia 1Research Centre for Sustainable Process Technology, Faculty of Engineering and Built Environment, Universiti Kebangsaan Malaysia, 43600, Bangi, Selangor, Malaysia *Corresponding author e-mail: saidusman2011@gmail.com Abstract The optimization of COD removal from palm oil mill effluent (POME) using the Reverse Osmosis (RO) membrane was investigated. Experimental conditions for reduce the COD value of POME were achieved successfully using the Box Behken design. The values of affecting factors (POME concentration, pH and Transmembrane pressure were optimized according to the polynomial regression model. The predicted conditions to produce lower COD values were found to be POME concentration (vol. %) =28.30, pH =10.75 and Transmembrane pressure= 0.69 kPa. The predicted of COD value was 24.137 mg/l which in good agreed with experiment value as 25.763 mg/l was obtained. Keywords POME, RO membrane, Box Behken, COD Received: 18 February 2019, Accepted: 1 March 2019 https://doi.org/10.26554/ijems.2019.3.1.39-46 1. INTRODUCTION Palm oil mill effluent (POME) has been reported as a main wastewater produced from industry of palm oil mill. The main process of crude palm oil extraction from fruit is re- quired to use huge amount of water which estimated 5-7.5 tons of water for produce of 1 tons of crude palm oil. Un- fortunately, more than 50% of the water has been produced as POME which causes to contaminate the environment (Ahmad, 2003). The common method to handle POME is an integrated anaerobic and aerobic pond which disadvan- tages large area and long residence times have been reported earlier (Yejian, 2008). Last few years, some of previous stud- ies are trying to still this problem by including biological and physical-chemical (Chan, 2010; Mohammed, 2014; Shak, 2014). Recently, the new method of membrane technology has been applied for POME treatment which improved the quality of final process effluents (Said, 2005). The application of membrane especially Reverse Osmosis (RO) for POME treatment has many advantages such as ca- pability to produce clear water in relatively short amount of time, small area and energy consumption. Higher efficiency to reduce many parameters related to POME treatment process quality such as BOD, COD, and, TSS to acceptable levels set by regulatory agency (Ahmad, 2006; Wu, 2010). Effects of several parameters such the pressure and con- centration on the POME treatment have been investigated experimentally for each parameter which led to increase the experiment runs and consuming time (Wu, 2007). In carrying out an experiments POME treatment which con- taining many affect variables, a tool such as Response Sur- face Methodology (RSM) is required to optimize the process response (Mohajeri, 2010). The Response Surface Methodology (RSM) has been applied as a statistical technique to study of affecting process variables and build of experimental model with interactive variables (Muneer, 2013b). As main advantages to apply RSM is, the higher interaction between all process variables and lower runs of experiment which attributed to consuming time of the process compared the traditional optimizations (Box and Draper, 1987). For this propose, the Box Behken design with lowest run of experiment has been selected for other chemical process which exhibited more efficient and accurate for the final process response (Ismail, 2005). The https://doi.org/10.26554/ijems.2019.3.1.39-46 Said et. al. Indonesian Journal of Environmental Management and Sustainability, 3 (2019) 39-46 Figure 1. Experiment stages of POME treatment process main objective of this work is to investigate the ability of RSM based on the Box Behken design to optimize the affected process variables to reduce the COD using Reverse Osmosis (RO) membrane. 2. EXPERIMENTAL SECTION 2.1 Materials and method Raw POME was collected from a local palm oil mill in Selangor, Malaysia. To adjust of POME pH within the treatment process, sodium hydroxide and hydrochloric acid from R & M chemical, Malaysia and Merck Company respec- tively, were used. The POME treatment process consists of two main stages as the first pre-treatment and second optimization was designed as shown in Fig. 1.The first stage of pre-treatment includes Adsorption and Ultrafiltration membrane (UF) was used to reduce suspended solids from the raw POME. About 10 L contains different volume % (vol. %) of raw POME concentration was feed to adsorp- tion column using dosing pump under constant flow rate at 2 ml/min. The POME flows downward along gravity, and exits at the bottom of the column which was feed into the UF membrane. The ultrafiltration pretreatment were carried out in cross flow unit with a hollow fiber membrane. Permeate from the UF membrane also simultaneously serves as feed into the RO membrane. In the stage of RO mem- brane, the POME feed was pumped through a spiral wound (RE2012-LPF, CSM filter) and recycled back to the UF feed reservoir. The COD values of the POME feed sample and permeate were performed and analysed using a DR/2010 portable data logging spectrophotometer (HACH, USA). 2.2 Statistical design of experiments In this study, the Box Behken was applied based on the Design Expert software version 6.0. Determination of the DOE aims to reduce the number of experiments and obtain the optimum response (Y) according to the interaction of all the factors (Xi) were involved. The response (Y) was related with three factors can be described by polynomial Equation (1): Y = β0 + k∑ i=1 βiXi + k∑ i=1 βiiX 2 i + k∑ i=1 βijXiXij + ε (1) The relationship among the three factors mathematically based on the second order as given in Equation (2) is: Y = β0 + β1X1 + β2X2 + β3X3 + β11X12 + β22X22 +β33X32 + β12X1X2 + β13X1X3–β23X2X3 (2) where,Y is the predicted response, βi is the coefficients, βii is coefficients of the quadratic terms, βij is the coefficients of the interactions of factors ε is random error. The response of this study was value of COD (Y) while the affecting factors were concentration of POME (X1), pH of solution(X2) and Transmembrane pressure (X3). All fac- tors and their levels affects on POME treatment process were selected as concentration varying between 10 to 90 and of solution adjusted between 3.0 to 11.0 using hydrochloric acid and sodium hydroxide, while the Transmembrane pres- sure between 0.5 to 2.5 kPa as summarized in Table 1.The total number of experimental runs was 17 based on the Box design and given in Table 1.The effects of the interaction among all the factors to reduce COD value were evaluated through an analysis of variance (ANOVA) according to ex- perimental results. Moreover, it was important to check the adequacy of the model using diagnostic graphs and validity of model by comparing the predicted to experimental results as the main steps (Muneer, 2013a,c). Table 1. Parameters and levels of Box Behken design Independent Xi -1 levels Factors 0 1 POME Concentration, (vol. %) X1 10 50 90 PH X2 3 7 11 TMP, (kPa) X3 0.5 1.5 2.5 3. RESULTS AND DISCUSSION 3.1 Model Fitting of Box Behken design The effects of POME concentration, pH of solution and Transmembrane pressure on the final value of COD were in- vestigated using the quadratic polynomial model. The math- ematical model of POME treatment optimization was esti- mated based on the experimental results using Box Behken design with the respective coefficients as given in Equation (3): Y COD = 78.36563 − 0.49438X1 − 8.46437X2 + 0.40000X3 +0.011198X21 + 0.38547X 2 2 − 1.25000X32 − 0.027594X12 +0.10831X13 + 0.52063X23 (3) © 2019 The Authors. Page 40 of 46 Said et. al. Indonesian Journal of Environmental Management and Sustainability, 3 (2019) 39-46 The final model creates by the Box Behken design was involved all the coefficients which shown as a quadratic re- gression. The theoretical (predicted) values of final COD using the created model were compared to experimental re- sults as listed in Table 2. The analysis of variance (ANOVA) results of the Box Behken model was presented in Table 3. The significance degree of the model and all the factors (X1, X2 and X3) are estimated according to the P-value. If this value that less than 0.050 which was considered to be significant, and any other value that is greater than 0.050 was not significant (Jafarzadeh, 2011). The higher signifi- cant factors of POME concentration (X1), the pH (X2) and then Transmembrane pressure (X3) well as the interaction between the POME concentration and the pH (X21) were observed. However, the lack of fit of Box Behken model with F-value of 0.9193 exhibits to be insignificant relative to the pure error (Zhang, 2009). Additionally, Other important terms are the accuracy and variability of the Box Behken model, which can be estimated according to the R-Squared (R2) value which is between 0 to 1, with a value closer to1 being a better predic- tion of the response (Karacan et al., 2007). The R2 of the Box Behken model showed a higher value, which was 0.9692. On the other hand, the Adj R-Squared (Radj) coefficient was also found to be 0.9297, which was agreed to the R2 value. However, adequate precision term of Box Behken was used for evaluated the predicted range of responses relative to the associated error. For this process, the value greater than 4 as 18.019 was found which attributed to support the fitness of the final model (Korbahti and Rauf, 2009).These values shows a good correlation between the all factors of the POME treatment process using the Box design. 3.2 Adequacy Check of the Box Behken Model Some plots to investigate the optimization POME treatment process using the Box Behken design are given in Fig.2. The normality term which indicates the relationship between the student zed residuals with normal probability (Fig 2 a) showed all points were close to the line. This result confirmed there no obvious problems with the normality of the design based on experiments result of COD. Fig.2 b show the plot between the studentized residuals and the predicted COD values which as random scattering of all the points rather than a funnel-shaped pattern was obtained. This results, confirms that the response was an original observation of variance and no problem with the response [20]. The values of the studentized residuals were almost at intervals of between -3.5 to +3.5 as shown in Fig.2 c, and the observed response value was not considered for any value beyond these values. This model was a studentized residual value that was lower than ± 3.5, which gives a good fitting of the model to the response surface (Rauf et al., 2008). The outlier of the experimental runs POME treatment process clearly gave that all the points in the range of the outlier Figure 2. All diagnostic plots of optimization of COD using Box Behken design, (a) Normality, (b) studentized residuals, (c ) Outlier T, (d) Actual and predicted. as a good distribution for the Box Behken design model. The actual value of the final COD from the experimental was nearly the same as the value predicted by the model as shown in Fig2 d, which due to the higher values of R2 and R2adj terms. 3.3 Response Surface Plotting and Optimization of Box Behken Model 3.3.1 Effects of concentration and pH solution on COD value The effects of the POME concentration and pH on the COD value of POME treatment were investigated by the RSM based on a Box design. The 3-D response surfaces and con- tour graphs were used to explain the effects of the interaction between two factors as shown in Fig. 3 (a,b). It can be observed that the final COD value decreases as the pH was increased from 3.0 to 11.0. This behaviour leads to changing of surface properties of impurities in POME. At higher pH of solution indicates to the charge of the impurities could be equal to the charge on the surface of the membrane. The similarity of these charges gives more hydrophilic nature of the membrane which leads to the impurities not stick to the membrane surface and the trapped to the bulk solution (Ahmad et al., 2005). However, at lower pH of solution, the attraction force between the impurities and surface of membrane was increased. This phenomenon guides the im- purities to easily attach to membrane surface and then pass through the pores. The impurities was not only included the solid particles but also the organic molecule. The existence of organic molecule can be assumed as the COD value in solution (Said et al., 2014). The liner relationship between the final COD and con- © 2019 The Authors. Page 41 of 46 Said et. al. Indonesian Journal of Environmental Management and Sustainability, 3 (2019) 39-46 Table 2. Experimental runs of Box Behken design with Actual and Predicted results of COD response Std Run No. (X1) (X2) (X3) COD(mg/L) COD(mg/L) (Actual) Predicted 1 4 10 3 1.5 54.67 53.55 2 12 90 3 1.5 108 109.96 3 10 10 11 1,5 35 33.04 4 3 90 11 1.5 70.67 71.79 5 6 10 7 0.5 32 34.5 6 14 90 7 0.5 74 73.42 7 5 10 7 2.5 36.67 37.25 8 2 50 7 2.5 96 93.5 9 17 50 3 0.5 60.33 58.96 10 9 50 11 0.5 26 25.46 11 16 50 3 2.5 65.67 66.21 12 11 50 11 2.5 39.67 41.04 13 8 50 7 1.5 48.67 43 14 7 50 7 1.5 37.33 43 15 1 50 7 1,5 33.33 43 16 13 50 7 1.5 51.67 43 17 15 50 7 1.5 44 43 Table 3. ANOVA results for quadratic model based on the Box Behken design Source Sum Degree Mean F-value P-value of squares of freedom square Model 8245.45 9 916.16 24.51 0.0002 X1 4528.19 1 4528.19 121.12 < 0.0001* X2 1720.79 1 1720.79 46.03 0.0003* X3 260.83 1 260.83 6.98 0.0334* X21 1351.73 1 1351.73 36.16 0.0005* X22 160.16 1 160.16 4.28 0.0772 X23 6.58 1 6.58 0.18 0.6874 X12 77.97 1 77.97 2.09 0.1919 X13 75.08 1 75.08 2.01 0.1994 X13 17.35 1 17.35 0.46 0.5176 Residual 261.7 7 37.39 Lack of fit 27.72 3 9.24 0.16 0.9193** Pure error 233.98 4 58.49 Total 8507.15 16 *Significant at < 0.05% level; ** Not significant, R2= 0.9692, R2adj =0.9297, Std. Dev. = 6.11, Mean=53.75, C.V= 11.38, Adeq Precision=18.019. © 2019 The Authors. Page 42 of 46 Said et. al. Indonesian Journal of Environmental Management and Sustainability, 3 (2019) 39-46 Figure 3. The effect of the concentration and pH on COD value of POME treatment (a) 3-D, (b) contour, (c) interaction centration was observed. By increasing of concentration of POME, the final COD after treatment was increased because higher concentration of POME contained more of impurities and organic molecules. The effects of interaction between the POME concentration and pH to produce on lower value of COD were shown in Fig. 3(c). Lower value of COD at lower concentration of POME was found which less than 50 mg/l at higher pH of 11.00. However, the interaction between the POME concentration and pH exhibits more affecting at all values which also was confirmed by significant of ANOVA of concentration analysis in Table 3 as p-value, 0.0005 < 0.05. 3.3.2 Effects of concentration and Transmembrane pressure on COD value The effects of concentration of POME and trans membrane pressure on COD value were estimated with different sev- eral values as presented in Fig.4. Slightly affecting of trans membrane pressure on COD with different values especially at lower concentration of POME (10 vol.%) was reported in Fig. 4 (a,b). This result is probably due to the lower concentration of POME as indicator to presence of smaller impurities. In addition, the increase in trans membrane pressure makes higher chance of the small particles to pass through the membrane pores (Said et al., 2014). At the same time, the COD value was found to increase as the concentration of POME increased due to the presence more amount of impurities in POME solution. To explain that higher of COD value resulting under all values of trans- membrane pressure for concentration more than 50 vol.% was found to be because to presence of more impurities. Additionally, at higher transmembrane pressure causes to throughout membrane more amount of impurities which attributed to increase the final COD as reported earlier [23]. Figure 4. The effect of the concentration and Transmembrane pressure on COD value of POME treatment (a) 3-D, (b) contour, (c) interaction The effects of interaction between the POME concentration and transmembrane pressure to produce on lower value of COD were shown in Fig. 4(c). At lower POME concentra- tion (10 vol.%), the changing in transmembrane pressure was not affected on COD value due to lower of impurities and organic molecules. The lower COD at lower POME con- centration than 50 vol.% under transmembrane pressure of 0.5 kPa was observed. However, the interaction between the POME concentration and transmembrane pressure showed affecting at higher values. 3.3.3 Effects of Transmembrane pressure and pH on COD value Effects of pH and transmembrane pressure on COD value were investigated under several varying values as shown in Fig.5. The lower COD values of POME at higher of 11.0 and lower of 0.5 kPa for pH and transmembrane pres- sure respectively were observed. At lower pH of 3.0 and all transmembrane pressure range, the higher COD values were produced. Influence of pH on COD value refers to changing the properties of impurities in POME and mem- brane surfaces, which at lower pH of solution, the attraction force between the impurities and surface of membrane was increased. This result leads to easily attach of impurities to membrane surface and then pass through the pores which attributed to increasing of COD value. However, the in- fluence of transmembrane pressure also showed the COD value increases as the pressure was increased. This results because under higher transmembrane pressure leads to more impurities pass through the pore membrane compare to under lower pressure which presence of small amount of impurities as reported in previous study (Zinatizadeh and Mohamed, 2007). The pH of solution shows as the main © 2019 The Authors. Page 43 of 46 Said et. al. Indonesian Journal of Environmental Management and Sustainability, 3 (2019) 39-46 Figure 5. The effect of the Transmembrane pressure and pH on COD value of POME treatment (a) 3-D, (b) contour, (c) interaction factor was affected on the performance of POME treatment process which reported for other treatment process earlier (Ba-abbad et al., 2013, 2012). 3.4 Model validation of POME optimization condi- tions To check of Box Behken model validation is very important to produce more stable model for POME treatment process. The main important term use to investigate the validity of the predicted model is desirability function (Box et al., 2005). The maximum value of the desirability function (D=1.000 or D≈1.000) was selected to optimize of POME treatment process. To determine factors to produce lower value of COD within the experiment process, the optimization criteria of RSM had five options: none, maximum, minimum, target and within range as shown in Table 4. The ANOVA analysis of this model, all factors of POME concentration, pH and Transmembrane pressure were signif- icant which each factor contributed affecting within their range option. The minimum value of COD according to the factor as POME concentration =28.30 vol. %, pH =10.75 and Transmembrane pressure= 0.69 kPa was to be 24.137 mg/l as shown in Fig.6. To investigate the model validity predicted by Box Behken design, a triplicate of experimental runs under optimum con- ditions of all factors was achieved. The average COD results was 25.763 mg/l was obtained which in good agreed with the predicted by model. Accordingly, the optimization results (comparison between the predicted and experimental results) showed more effective and reliable to apply the Box Behken design for reduces of COD value which was attributed to a good interaction between the selected factors with their ranges. Figure 6. Predicted COD, as obtained from the RSM based on Box Behken design under optimal conditions 4. CONCLUSIONS In this study, the POME treatment process using reverse osmosis membrane was optimized by Appling the response surface method (RSM) based on the Box Behken design. Effects of POME treatment process factors with their in- teractions were estimated by ANOVA. Good coefficients of R2 as 0.9692, R2adj as 0.9297 of the predicted model were obtained. Optimum conditions for lower value of response COD were POME concentration =28.30 vol.%, pH = 10.75 and Transmembrane pressure= 0.69 kPa. The predicted value of COD under optimal condition was 25.763 mg/l which showed in good agreement with the predicted by model as 24.137 mg/l. This result evidences to support the validity of the model was created by Box Behken de- sign which also showed as suitable way for optimizing the conditions POME treatment process in future. 5. ACKNOWNLEDGMENT The authors wish to thank the Research Centre For Sustain- able Process Technology (CESPRO) , Faculty of Engineering and Built Environment, Universiti Kebangsaan Malaysia for support this study under project KK-2013-003. One of the authors (M. M. Ba-Abbad) is grateful to the Hadhramaut University of Science &Technology, Yemen for their financial support for his PhD study. Also one of authors (Muhammad Said) is grateful to Universiti Kebangsaan Malaysia (UKM) for supporting through the Beasiswa Zamalah Universiti Penyelidikan. REFERENCES Ahmad, A. L., S. Ismail, and S. Bhatia (2005). Membrane treatment for palm oil mill effluent: effect of transmem- brane pressure and crossflow velocity. Desalination, 179; 245–255 Ahmad, C. M. B.-S. I. S., A. L. (2006). Drinking water reclamation from palm oil mill effluent (POME) using membrane technology. Desalination, 191; 35–44 Ahmad, I. S. B. S., A.L. (2003). Water recycling from palm oil mill effluent (POME) using membrane technology. Desalination, 157; 87–95 © 2019 The Authors. Page 44 of 46 Said et. al. Indonesian Journal of Environmental Management and Sustainability, 3 (2019) 39-46 Table 4. Optimization details of POME treatment process term target Lower limit Upper limit Lower weight Upper weight POME conc. (vol.%) In the range 10 90 1 1 pH In the range 3 11 1 1 Transmembrane In the range 0.5 2.5 1 1 pressure (kPa) COD(mg/l) Minimum 26 108 1 1 Ba-abbad, M. M., A. A. H, Kadhum, A. bakar Mohamad, M. S, and K. Takriff (2013). Photocatalytic degradation of chlorophenols under direct solar radiation in the presence of ZnO catalyst. Research on Chemical Intermediates, 39; 1981–1996 Ba-abbad, M. M., A. A. H, Kadhum, A. bakar Mohamad Mohd S, and K. S. Takriff (2012). Synthesis and Cat- alytic Activity of TiO2 Nanoparticles for Photochemical Oxidation of Concentrated Chlorophenols under Direct Solar Radiation. International Journal of Electrochemical Science, 7; 4871 – 4888 Box, G. and N. Draper (1987). Wiley Series in Probabil- ity and Mathematical Statistics: Applied Probability and Statistics. John Wiley & Sons Inc Box, J. G., S. Hunter, and W. Hunter (2005). Statistics for Experimenters: Design, Innovation, and Discovery, second ed. New Jersey: Wiley Interscience Chan, C. M. L.-C., Y. J. (2010). Biological treatment of anaerobically digested palm oil mill effluent (POME) using a Lab-Scale Sequencing Batch Reactor SBR). Journal of Environmental Management, 91; 1738–1746 Ismail, E.-M. A. A.-A. E. . E.-S. H., A.A. (2005). Applica- tion of statistical design to optimize the preparation of ZnO nanoparticles via hydrothermal technique. Materials Letters, 59; 1924–1928 Jafarzadeh, S. S. H.-S. N. . R.-F., N. K. (2011). Statisti- cal optimization of process conditions for photocatalytic degradation of phenol with immobilization of nano TiO2 on perlite granules. Korean Journal of Chemical Engi- neering, 28; 531–538 Karacan, F., U. Ozden, and S. Karacan (2007). Optimiza- tion of manufacturing conditions for activated carbon from Turkish lignite by chemical activation using response surface methodology. Applied Thermal Engineering, 27; 1212–1218 Korbahti, B. and M. Rauf (2009). Determination of optimum operating conditions of carmine decoloration by UV/H2O2 using response surface methodology. Journal of Hazardous Materials, 161; 281–286 Mohajeri, A. H. I. M.-Z. M. A. M., S. (2010). Statistical optimization of process parameters for landfill leachate treatment using electro-Fenton technique. Journal of Hazardous Materials, 176; 749–758 Mohammed, C. M., R. R. (2014). Treatment and decoloriza- tion of biologically treated Palm Oil Mill Effluent (POME) using banana peel as novel biosorbent. J. Environ. Manag, 132; 237–249 Muneer, K. A.-M. A. T. M.-K., M B A. (2013a). Visible light photocatalytic activity of Fe3+-doped ZnO nanoparticles prepared via sol–gel technique. Chemosphere, 91; 1604– 1611 Muneer, K. A. M.-A. T. M. K., M. B.A. (2013b). Optimiza- tion of process parameters using D-optimal design for syn- thesis of ZnO nanoparticles via sol–gel technique. Journal of Industrial and Engineering Chemistry, 19; 99–105 Muneer, K. A. M.-A. T. M. K., M. B.A. (2013c). Optimiza- tion of process parameters using D-optimal design for syn- thesis of ZnO nanoparticles via sol–gel technique. Journal of Industrial and Engineering Chemistry, 19; 99–105 Rauf, M., N. Marzouki, and B. Korbahti (2008). Photolytic decolorization of Rose Bengal by UV/H2O2 and data optimization using response surface method. Journal of Hazardous Materials, 159; 602–609 Said, M., A. Ahmad, A. Mohammad, M. Nor, and S. Ab- dullah (2014). Blocking mechanism of PES membrane during ultrafiltration of POME. J. Indust. Eng. Chem, 21; 182–188 Said, M. A. N. M.-A. S. H. H., M. (2005). Investigation of Three Pre-treatment Methods Prior to Nanofiltration Membrane for Palm Oil Mill Effluent Treatment. Sains Malaysiana, 44(3); 421–427 Shak, W. T., K. P. Y. (2014). Coagulation-flocculation treat- ment of high-strength agro-industrial wastewater using natural Cassia obtusifolia seed gum: Treatment efficiencies and flocs characterization. Chem. Eng. J., 256; 293–305 Wu, M. A. J. J. A. N., T. Y. (2010). Pollution control tech- nologies for the treatment of palm oil mill effluent (POME) through end-of-pipe processes. Journal of Environmental Management, 91; 1467–1490 Wu, M. A. J. J. A. N., T.Y. (2007). Palm oil mill effluent (POME) treatment and bioresources recovery using ul- trafiltration membrane: Effect of pressure on membrane fouling. Biochemical Engineering Journal, 35(3); 309–317 Yejian, L. Y. L. C. X. L. Z. M.-Z. Z., Z. (2008). Startup and operation of anaerobic EGSB reactor treating palm oil mill effluent. Journal of Environmental Sciences, 20(6); 658–663 Zhang, Z. H., Z. (2009). Optimization for decolorization of © 2019 The Authors. Page 45 of 46 Said et. al. Indonesian Journal of Environmental Management and Sustainability, 3 (2019) 39-46 azo dye green 20 by ultrasound and H2O2 using response surface methodology. Journal of Hazardous Materials, 172(3); 1388–1393 Zinatizadeh, A. A. L. and A. R. e. a. Mohamed (2007). Optimization of pre-treated palm oil mill effluent digestion in an up-flow anaerobic sludge fixed film bioreactor: A comparative study.”. Biochemical Engineering Journal 35(2):, 35(2); 226–237 © 2019 The Authors. Page 46 of 46 INTRODUCTION EXPERIMENTAL SECTION Materials and method Statistical design of experiments RESULTS AND DISCUSSION Model Fitting of Box Behken design Adequacy Check of the Box Behken Model Response Surface Plotting and Optimization of Box Behken Model Effects of concentration and pH solution on COD value Effects of concentration and Transmembrane pressure on COD value Effects of Transmembrane pressure and pH on COD value Model validation of POME optimization conditions CONCLUSIONS ACKNOWNLEDGMENT