Optimization of Whey Treatment in Rotating Biological Contactor: Application of Taguchi Method

Industrial waste management is the main concern of sustainable environment, and the proper control and operation of wastewater treatment plants for efficient removal of pollutants are certainly important. In this study, the L16 orthogonal array of the Taguchi method was applied to determine the optimum condition of some controlling parameters such as organic loading rate, hydraulic retention time, number of stages, rotational speed and recycling for treating whey effluent in a rotating biological contactor (RBC). In order to determine the impacts of levels and factors, the optimum condition of the experiment was predicted and determined. QUALITEK-4 has reported 76.25% COD removal based on a desired experimental condition (COD: 50000 mg L−1, HRT: 24 h, No. of stages: 3). To improve the performance of the RBC, second optimum condition (COD: 50000 mg L−1, HRT: 24 h, recycling: positive, rotational speed: 10 rpm, No. of stages: 3) was also investigated. According to the above-stated conditions, the highest COD removal was found to be 98%. Furthermore, two sets of test experiments were also conducted in lab scale under optimum conditions, and the prediction of other combinations of factors/levels was evaluated.


Iranica Journal of Energy & Environment
Journal Homepage: www.ijee.net Environmental protection is one of the most important concerns of mankind.Discharge of industrial effluents may cause serious environmental pollution.Based on the nature of industrial effluents, suitable treatment methods should be employed.The disposal of whey produced in process of cheese production has always created a major problem in the dairy industry.Cheese effluents exhibited high COD values in the interval of 0.8-102 g L −1 and BOD values in the range of 0.6-60 g L −1 leading to a high consumption of dissolved oxygen in water bodies [1,2].In the last few decades, technological and economic advancement made it possible to recover soluble proteins from cheese whey [3].However, whey reuse is not applicable with small dairies which produce cheese and their whey is discharged as waste with the rest of wastewater.As a result, their waste discharge loaded with organic material and other pollutants may cause environmental problems especially pollution in groundwater sources.Generally, biological treatment processes are mainly used to treat cheese whey wastewater [1,[4][5][6][7][8][9].
Based on microorganism growth, two distinct types of biological treatment processes exist known as, attached growth and suspended growth.Attached growth process is more stable than suspended growth; particularly when the wastewater includes high concentration of organic matter (up to 60 g COD L −1 ), and high biodegradability.
Very low bicarbonate alkalinity of cheese whey wastewater limits its direct treatment in high-rate anaerobic reactors that is probably due to problems encountered in maintaining stable operation [1,[8][9][10].
Aerobic rotating biological contactor (RBC) has several advantages such as short retention time, low power consumption, high biomass concentration, handling a wide range of flows and low sludge generation [8,11].Performance of RBC reactor was operated for high-strength wastewaters of poultry, palm oil mill effluent and landfill leachate [12][13][14][15].
The appropriate operation and maintenance of wastewater treatment plants became a major issue due to increasing concern for environmental pollution.In order to enhance the total efficacy of wastewater treatment plants, it is essential to use a robust model to predict certain key parameters of the influent based on previous observations.A Full factorial design (FFD) is a method of experiment design in which the effect(s) of two or more factors\variables, each with discrete values\levels and their possible interactions are determined.In this method, experimental units take on all possible combinations of these levels across all such factors.In most cases, use of FFD is extremely time-consuming, especially when the number of factors or levels increases.Taguchi as a family of FFE matrices was developed to minimize and control variation in a product or process and the number of tests required [16][17][18][19].Using this method, the optimum experimental conditions can be easily determined.Up to now, Taguchi method has been extensively used to obtain the optimum process parameters in different areas such as developing space vehicles [20], agricultural products [21], physical activities [22], telecommunications [23], wastewater treatment [24,25], structure [26], power [27,28], metals manufacturing [29,30], milling [31], engineering [27,32,33] and medical science [34].
In this study, the influence of some operating conditions such as organic loading rate, hydraulic retention time, number of stages, rotational speed and recycling in the treatment of whey effluent by RBC was investigated.Moreover, in order to optimize the design of an existing process, it is necessary to identify which factor has the greatest influence; thus some experiments were conducted using Taguchi experimental design.

RBC experiment
A 3-stage RBC similar to our previous study was built [2,4,8].The RBC consisted of three equal-size compartments (75×35×30 cm 3 /each), separated by fixed baffle plates with 16 lightweight PVC discs in each compartment.The total volume of RBC was 78.75 L. Discs mounted on a galvanized metal shaft had diameter of 32 cm, 3 mm thickness and interspacing of 8 mm.The total surface area available for microorganism growth on discs was 7.1 m 2 .A changeable speed motor (NORD motor, model SK 63/4, Germany) was used to rotate the shaft with mounted discs.Based on steady-state assumption, the samples were taken for each volumetric flow rate after two times of corresponding retention time.The fresh raw wastewater was collected in a 20 liters container, then transported to the laboratory on a daily basis and finally refrigerated and stored at 4 o C to prevent from acidification and undesirable changes in their chemical composition.Diluted whey was fed into the reactor at pH 6.5 during the adaptation phase.Several dilutions of cheese whey were prepared using distilled water based on requirements of the tests.For the adjustment of feed pH to 6.5, a 6 M sodium hydroxide solution was used.The cheese whey used in this study was obtained from "Gela Factory" (Amol, Iran), which was collected from ultra-filtration process for the production of cheese.
The RBC reactor was inoculated with 10 L of the sludge with MLSS value of 3000 mg L -1 which was obtained from the mentioned wastewater treatment plant.In order to form an initial accumulation of biofilm, the reactor was fed with raw whey containing 0.4 g COD L -1 d -1 at rotating speed of 4 rpm and HRT of 36 h.The amount of biofilm formed on the rotating discs was observed on the seventh day and then a fast biofilm accumulation was observed.The COD removal rate achieved 81% on the 12th days; that suggested the startup of the RBC was completed.During the operation, the heavy biomass/ biofilm deposited on net-like discs at the three stages were 383.4,311.1 and 175.2 g m -2 , respectively.The fabricated pilot unit with and without biofilm on the surface of discs are shown in Figure 1.The reactor design and operation parameters are summarized in Table 1.

Analytical methods
The collected samples were settled for 30 minutes to analyze COD and biomass concentration.Standard methods for the examination of water and wastewater were used to measure TSS and DO indicators [35].In order to measure COD, a colorimetric method with closed reflux method was developed.Spectrophotometer (UNICO 2100, USA) at 600 nm was used to measure the absorbance of samples.Biofilm weight was subsequently quantified by a specified methodology: first, the biomass flashed off from a piece of meshed material by water, and then this effluent filtered with 0.45 µm membrane and dried for at least 24 h at 105 о C. DO and pH meter (Hanna HI9024) were used to measure pH and dissolved oxygen.The reactor temperature was maintained basically at the room temperature (22 о C).

DOE (Design of Experiment) Experiment No.1
The most important stage in the design of an experiment is the selection of the control factors from the major influential items involved; in this study, from both theoretical and empirical points of view and set their respective levels accordingly and appropriately.Table 2 indicates all factors and levels that might affect our experiment.

Signal-to-noise ratio
In Taguchi designs, the signal to noise ratio (S/N ratio), was used to identify control factors that reduce variability in the process by minimizing the effects of uncontrollable factors (noise factors).The term "signal" represents the desirable control factors during the process and "noise" represents the undesirable effect.The S/N ratio can be calculated by the following equation [36]: Where "MSD" is the Mean Square Deviation, which was obtained using equation (2) [36].
Where 'n' is the number of repetitions and 'Yi' is the observed data obtained from the present experimental work.Note that there are three main categories of the quality characteristics, they are: i. smaller is better, ii.nominal is better and iii.bigger is better.To obtain high percentage of COD removal, the 'bigger is better' quality characteristic was chosen.Among the levels of each factor, there is only one level which is the most effective level than others.It demonstrates the importance of noted levels with higher effectiveness rates in our study; these are essential to pursuing "bigger is better" characteristic.The knowledge of the contribution of individual parameters is critically important for the control of the final response.In order to study the significance of the parameters in affecting the quality characteristic of COD removal, ANOVA was performed.Table 5 shows the results of analysis of variance (ANOVA) for the COD removal.
Where DOF stands for the degree of freedom, SS is the sum of squares and the variance is mean square factor.Also, F-ratio for each process parameter is simply a ratio of the mean of the squared deviations to the mean of squared errors.Percent is contribution percentage of each factor on the response which indicates the influences of factors and interactions assigned to the column to the variations of the results.The row labeled other/error, contains information about the sources of variability of the results which indicates information about the influence from three sources.1. Experimental error 2. Factors which are not included in the experiment 3. Uncontrollable noise factors [37].Thus, based on the S/N ratio and ANOVA analyses, the optimum condition of the conducted experiment can be predicted and reported.At the optimum condition (COD= 50000 mg L −1 , HRT=24, Stage No. = 3) QUALITEK-4 has predicted COD removal of 76.25%.Implementation of the above optimal condition has resulted in 71% COD removal.The small deviation (error of 6.8%) was generally caused by the factor which was devoted by noise factor.Such error caused insufficient COD removal which was due to the limitation in oxygen mass transfer and disc surface area for the attached growth system.The fact is, COD removal efficiency increased with an increase in HRT and the system achieved the highest COD removal efficiency at the HRT of 24 h.Normally, high HRT provides enough contact time for the biodegradation of organic matter in the reactor and hence, a long contact time between support media and wastewater; that may enhance the pollutant removal efficiency.The justification might be the result of enough time for attachment of microorganisms on disk's surface and hence, developing a very active biomass layer on disk's surface.Furthermore, a slight decline in the trend of COD removal from HRT of 24 to 36 h could be attributed to the increasing of biofilm layer thickness presented on the disk's surface at highest HRT, which resulted in reducing oxygen mass transfer.For having sufficient COD removal, the experiment was improved and the other parameters for a desired contact time of microorganisms were considered in experiment No. 2.

Experiment No. 2
Two effective items were suggested to be added to Table 2.These items include rotational speed and recirculation rate.Since it was observed in previous studies, these items have an effect on reactor performance in COD removal [38,39].Therefore, the alternative optimum result for our new DOE conditions was investigated and the results are shown in Table 6.L16 arrays were selected to design and calculate based on the average value of the results.The result of each experiment is shown in Table 7.In the second optimum condition COD: 50000, HRT: 24, recycling: positive, rotational speed: 10 rpm, stage No.: 3 expected COD removal was 98 %.The findings confirmed that the result of the second experiment by means of determining the robust design condition was mostly similar with experiment No. 1.However, the outcome of the second condition is the much elegant condition.The result indicates that the COD removal improves when rotational speed increases from 4 to 10 rpm.The main reason was due to an increase in dissolved oxygen (DO) level when the rotational speed increased to 10 rpm.In fact, in low rotational speed, the rate of aeration was not as necessarily enough to treat the wastewater.Therefore, after few days of operation, the bad odor was spread out due to the oxygen depletion.In addition, COD removal efficiency was firmly affected by recirculation.It was concluded that the recirculation may be beneficial when the inlet organic loading rate of chemical shock was extremely high.
Similar to the previous result, using these conditions for COD removal from experiment resulted in 92% COD removal, only 6.1% less than obtained result by Taguchi model, and very close to experiment No. 1.The obtained results prove that the proposed method with fair accuracy can be employed for the different experimental conditions.

CONCLUSION
Using Taguchi method, the optimum conditions of whey treatment using RBC reactor were investigated.This study examined the power of Taguchi's optimization method to predict the COD removal in response to different combinations of factors/levels through conducting a very limited number of experiments with an acceptable level of accuracy.The results showed that Taguchi method was successfully able to model the COD removal in RBC reactor.Moreover, the percentage contribution of each studied factor to whey treatment was also determined using the procedure explained by the method.The information generated in this study might help environmental engineers and managers to come up with suitable decisions in dealing with waste treatment.It is highly recommended to conduct further studies with other determinant factors for COD removal; that might allow drawing the comprehensive conclusion.

Figure 1 .
Figure 1.The pilot RBC before and after developing biofilm

Figure 2 .
Figure 2. Effect of different factors on S/N ratio in experiment No. 1

Figure 3 .
Figure 3.Effect of different factors on S/N ratio in experiment No. 2

TABLE 1 .
The reactor design and operation parameters

TABLE 2 .
Factors and their levels considered in the Taguchi DOE Taguchi method demonstrated how each factor quantitatively or even each level of them can influence the results of our experiments.The mean values of S/N ratios for COD removal are shown in Figure 2.

TABLE 3 .
Result of experiment, M16 orthogonal array

TABLE 4 .
Results of designed experiments

TABLE 5 .
Analysis of variance in ANOVA table

TABLE 6 .
2actors and levels in experiment no.2The control factors indicate the trend of influence and variations of factors with their levels.Figure3displays plots which show the influence of each factor in the studied levels.COD, HRT, and stage No. had the same behavior as experiment No. 1.As it was discussed earlier, recycling plot remains no discussion on its positive effect in experiment No. 2. Also, rotational speed comes with 2 levels, 4 rpm and 10 rpm.The plot shows that high speed (10 rpm) improved the results.Table 8 summarized optimum conditions in new design experiment No. 2.

Table 8 .
Optimum condition in experiment No. 2