Print this page Email this page
Users Online: 365
Home About us Editorial board Search Browse articles Submit article Instructions Subscribe Contacts Login 

Previous article Browse articles Next article 
ORIGINAL ARTICLE
Int J Env Health Eng 2018,  7:5

Modeling of extracellular polymeric substances production at different Carbon/Nitrogen ratio and solid retention time by artificial neural network


1 Department of Environmental Health Engineering, School of Health, Isfahan University of Medical Sciences; Student Research Committee, School of Health, Isfahan, Iran
2 Department of Environmental Health Engineering, School of Health, Isfahan University of Medical Sciences; Environment Research Center, Research Institute for Primordial Prevention of Noncommunicable Disease, Isfahan, Iran

Date of Submission02-Dec-2017
Date of Acceptance07-Feb-2018
Date of Web Publication28-Jun-2018

Correspondence Address:
Ali Fatehizadeh
Department of Environmental Health Engineering, School of Health, Isfahan University of Medical Sciences, Isfahan
Iran
Login to access the Email id

Source of Support: None, Conflict of Interest: None


DOI: 10.4103/ijehe.ijehe_10_17

Rights and Permissions
  Abstract 


Aims: The ability of extracellular polymeric substances (EPS) production was observed in many species of heterotrophic microorganisms through the biological wastewater treatment systems. Materials and Methods: The batch experiments at different carbon/nitrogen and solid retention time (SRT) were carried out to investigate the effects of initial nitrogen concentration and SRT on EPS production and chemical oxygen demand (COD) removal efficiency. The artificial neural network (ANN) was developed to modeling of obtained data. Results: The results showed that: (i) with increasing SRT, the COD removal improves; (ii) initially, the amount of carbohydrate increases as SRT increases; however, with further increase of SRT, it declines; (iii) the protein/carbohydrate ratio improves as SRT decreases; (iv) the carbohydrate and protein concentration of soluble EPS increased with increasing initial nitrogen concentration from 0 to 10 mg/L; and (v) further increase of initial nitrogen concentration lead to depletion of carbohydrate production. Conclusion: The highest yield (Y) value was calculated at low sludge age and deficient initial nitrogen concentration, which may be due to the application of EPS production mechanism. The ANN model moderately predicted effluent COD concentration, carbohydrate, and protein production.

Keywords: Artificial neural network model, carbohydrate, extracellular polymeric substances, nitrogen, protein


How to cite this article:
Taheri E, Amin MM, Ghasemian M, Rafiei N, Fatehizadeh A. Modeling of extracellular polymeric substances production at different Carbon/Nitrogen ratio and solid retention time by artificial neural network. Int J Env Health Eng 2018;7:5

How to cite this URL:
Taheri E, Amin MM, Ghasemian M, Rafiei N, Fatehizadeh A. Modeling of extracellular polymeric substances production at different Carbon/Nitrogen ratio and solid retention time by artificial neural network. Int J Env Health Eng [serial online] 2018 [cited 2018 Sep 24];7:5. Available from: http://www.ijehe.org/text.asp?2018/7/1/5/235433




  Introduction Top


The ability of extracellular polymeric substances (EPS) production was observed in many species of heterotrophic microorganisms through the biological wastewater treatment systems.[1] The microorganisms can release EPS that consists of some polymers with carbohydrates and proteins as dominant component. The EPS can be categorized as soluble and insoluble that separated by the centrifugation method. The polymers in the supernatant being soluble EPS and those founded in the microbial pellets being insoluble or bound EPS.[1],[2] As previously mentioned, the presence of EPS depicts the positive effect on the sedimentation properties of sludge, biofilm formation, and bacteria resistance to biocides or other harmful effect, and its presence is essential to strength sludge structure.[2],[3],[4] The negative effects such as the membrane biofouling, poor bioflocculation, and greater cell erosion are related to higher EPS content of sludge.[5],[6]

The EPS production may be influenced by (i) substrate type: the carbon source is the important factor that affect the EPS production, for example, using glucose or soybean oil as carbon source, the highest EPS production was noticed;[2],[7] (ii) the nutrient content: phosphorus, nitrogen, and carbon/nitrogen (C/N) ratio affects the EPS content of sludge;[1] and (iii) other parameters such as the microbial species, growth stage, dissolved oxygen, ionic strength, solid retention time (SRT), and food to microorganism (F/M) ratio can also affect the EPS formation.[1],[8],[9]

Previous researches demonstrate many contradictions about EPS effects on the sludge properties and EPS production as a function of SRT and C/N ratio, etc. Despite EPS effects on the sludge settling characteristics, Ye et al. reported that the contents of loosely bond protein in EPS is directly related to the flocculation, settleability, and dewaterability of sludge; however, there is no relationship with total amount of EPS.[10] Furthermore, a similar contradiction was reported by Xuan et al. and Sponza about the effect of EPS content on the sludge volume index (SVI) as the sludge settleability indicator.[3],[4] Liao et al. depict the no-significant correlation between the amount of EPS and SRT, but Sesay et al. demonstrate that the total amount of EPS in activated sludge significantly increases as SRT increases.[11],[12] In respect to C/N ratio effect on the EPS production, Miqueleto et al. found that the EPS production is favored by a high C/N ratio,[2] although Ye et al. reported that the carbohydrate and protein contents in EPS are not influenced by C/N ratio.[10]

Considering these contradictions and the effect of other mentioned factors, prediction of EPS production at different situations has been difficult. This study aimed to investigate (i) the influence of SRT and C/N ratio on EPS production; (ii) simulates the results using artificial neural network (ANN) (MATLAB software, R2014a); and (iii) the effect of EPS on sludge sedimentation properties.


  Materials and Methods Top


Batch experiments

The batch experiments were performed using a cylindrical reactor of 15.5 cm and 16.5 cm (D and H) with a working volume 2 L and aerated with the bottom air stone spray to achieve DO 2–5 mg/L. The sludge inoculum for each batch test was obtained from another sequencing batch reactor. Each set was operated at steady state condition with SRT of 2, 5, 10, and 15 d. At each SRT, the experiment was done in duplicate with C/N ratio of 100:0, 100:1, 100:2.5, 100:5, and 100:10. The initial chemical oxygen demand (COD) and pH for all experiment were maintained to 1000 ± 100 mg/L and 7–8, respectively. The batch experiments were operated for 12 h per cycle at 27°C. During batch experiments, variation of pH was monitored but not controlled.

Synthetic substrate

The synthetic wastewater by organic concentration with COD 1000 ± 100 mg/L was feed to batch tests. The glucose was used as carbon source and macro and micronutrient elements for microorganism growth were also added to the synthetic substrate.[13]

Analytical procedure

DO, pH, EC, and turbidity were measured using a DO membrane probe (Eutech Cyborscan DO300), precalibrated glass body pH probe (CG 824 SCHOTT), HACH digital conductivity meter (SENSION 5), and EUTECH instruments turbidimeter (TN-100), respectively. At the end of batch experiments, the supernatant and mixed liquor samples were collected from batches and analyzed for COD, total suspended solids, volatile suspended solids (VSS), and SVI. The standard methods for water and wastewater examination were adopted to measure these parameters.[14]

Extracellular polymeric substances analysis

The EPS extracted using formaldehyde + sodium hydroxide (NaOH) method adapted and the carbohydrate and protein in the sludge and effluent was measured as EPS and soluble EPS (sEPS), respectively.[15] To extract EPS and sEPS, the sludge was stored in 4°C for 1.5 h. Then, the supernatant was discarded and residual sludge centrifuged at 6000 rpm for 30 min at 4°C. After this, the supernatant was removed and pellet was resuspended by distilled water and used for EPS extraction using formaldehyde + NaOH method and extracted liquid subjected to EPS analysis. For sEPS analysis, the effluent wastewater sample was filtered using 0.45 μ cartridge filter then centrifuged (6000 rpm, 30 min) at 4°C and supernatant for sEPS analysis was used. The carbohydrate and protein content in EPS was measured using phenol–sulfuric acid assay procedure with glucose as standard and Lowry method using bovine serum albumin as standard solution, respectively.[16],[17]

Statistical analysis

Tests for significant differences between EPS concentration in variable C/N and F/M ratio were determined using one-way ANOVA (or Mann–Whitney). Pearson correlation coefficients were used to examine the relationship between the parameters and were considered significant when P < 0.05.

Artificial neural network modeling

As part of this study, the ANN was used to predict carbohydrate, protein production, and effluent COD concentration.

For this purpose, a three-layer ANN with a tangent sigmoid transfer function (tansig) at hidden layer, and a linear transfer function (purelin) at output layer were used. Levenberg–Marquardt back propagation (trainlm) was used to train the designed networks.

To avoid numerical overflows due to very large or small weights, the data are converted to normalized values using Equation 1.



Where xnorm is the normalized value of xi, the xmax and xmin are the maximum and minimum value of xi, respectively.

The results of various network structures and training procedures were compared using the mean squared error (MSE) and the coefficient of determination (R2) defined by Equations 2 and 3, respectively.





Where ypre,i is the predicted value from the ANN model, yexp,i is the experimental value, N is the number of data, and yave is the average value of the experimental value.


  Results and Discussion Top


Artificial neural network modeling

The data obtained from batch experiments were used to train the network. The input layer had three neurons consist of influent COD, SRT, and influent nitrogen. The output layer had one neuron consists of effluent COD concentration (or carbohydrate and protein production). Same approach was adopted to the prediction of carbohydrate and protein production. All data were randomly split into three groups (70% for training, 15% for validation, and 15% to test the model). To obtain the optimum number of neurons (N) in the hidden layer, trial and error method was used. Therefore, different number of neurons in the range of 1–10 was tested in the hidden layer. The optimum hidden layer size was determined based on the minimum value of MSE and R2 of the predicted results. As seen in [Figure 1], as the number of neurons in the hidden layer increase, MSE decreases promptly and then rises.
Figure 1. Comparison of 10 neurons in the hidden layer for value estimating by ANN model development with the Levenberg-Marquardt algorithm; (a) Effluent COD concentration, (b) Carbohydrate production and (c) Protein production

Click here to view


As shown in [Figure 1], in this study, three-layered feed forward back propagation neural network including 3:2:1, 3:5:1, and 3:6:1 were used to model the effluent COD concentration, carbohydrate, and protein production, respectively.

Plot of MSE versus the number of iterations for optimal ANN model [Figure 2] show that the training was stopped after 10, 7, and 8 epochs for effluent COD concentration, carbohydrate, and protein production, respectively. [Figure 3] showed that the training and testing performance of the ANN model. It shows that the ANN model is able to predict the removal of zinc ions from wastewater using activated almond shell with MSE <0.05. With regard to the fact that many factors involve in EPS production and also the level of uncertainty involves, the proposed modeling methodology can be used to data prediction.
Figure 2: MSE versus the number of epochs; (a) Effluent COD concentration, (b) Carbohydrate production and (c) Protein production

Click here to view
Figure 3: Experimental data versus the predicted data of normalized value from ANN: (a) Effluent COD concentration, (b) Carbohydrate production and (c) Protein production

Click here to view


[Figure 1] presents the normalized results of ANN model versus normalized experimental data for the effluent COD, carbohydrate, and protein production. The coefficient of determination (R2) was >0.5 for the training, validation, and test set. Although the coefficients of determinations for the testing data are lower than the training values, their difference is acceptable and the proposed model is recommended.

Chemical oxygen demand removal

The COD removal efficiency throughout all batch experiments is presented in [Figure 4]a. The results show that with increase of SRT, the COD removal is enhanced. It was expected that the COD removal controlled by two processes including biomass assimilation/accumulation and oxidation, in which these processes were strongly influenced by SRT. The ANOVA test showed that the COD removal efficiency related to 2-day SRT was significantly lower than those at SRT 5, 10, or 15 days (P< 0.05). However, there was no significant difference between COD removal efficiency at other SRTs (5, 10, and 15 days). Furthermore, no significant relationship between COD removal and initial N concentration was observed, and the results demonstrated that by changing of initial N concentration (increasing or decreasing), the COD removal efficiency remains relatively constant. Perhaps, it means that bacteria can uptake carbon as an electron donor in both N deficit and abundance. In these tests, at C/N ratio of 100:0, 100:1, 100:2.5, 100:5, and 100:10, the average of COD removal efficiency was 95.9%, 95.35%, 93.79%, 93.14%, and 94.66%, respectively. The small variation of COD removal at different initial N concentration is in line with Feng et al. They reported that as COD/N ratio decreases from 10 to 5, the average efficiency of COD removal fluctuates from 94.3% to 94.5%, respectively.[18]
Figure 4: Influence of SRT and initial concentration of N on (a) COD removal, (b) EPS production and (c) sEPS production

Click here to view


Extracellular polymeric substances production

The variation of carbohydrate and protein content in the sludge and effluent wastewater as a function of initial N concentration and SRT was evaluated and presented in [Figure 4]b and [Figure 4]c. The results showed that with increasing SRT from 2 days to other studied SRT values (5, 10, and 15 days) led to significant decrease in carbohydrate of sludge (Pearson correlation coefficient: −0.685, P < 0.05). It may be related to high carbohydrate production by bacteria at low sludge age. The protein/carbohydrate ratio was improved as SRT increases. The results did not prove that as the bacteria age increases, the EPS layer form the exterior of the bacterial cell as demonstrated by a previous study [19] and depicted that the EPS production may be controlled by other operating conditions such as nitrogen concentration and electron donor availability. The previous studies report different outcome for EPS production at various SRT. Sesay et al. demonstrate the polymers production significantly amplify as SRT increases. With SRT increasing from 4 to 8 days, the carbohydrates amount was constant and also at SRT higher than 8, the carbohydrates concentration was augmented.[12] In addition, Ahmed reported that the EPS content and SRT are negatively correlated.[5] Other studies reported that the amount of EPS production is independent of SRT.[11],[20] Liao et al. found that varying SRT from 4 to 12 days, led to increase of proteins/carbohydrates ratio and remain constant for SRTs above 12 days.[11] It appears that the SRT variation is an important factor on EPS production.

In this study, statistical analysis demonstrated that the sEPS production has no significant correlation with SRT. These results are in line with the previous study.[11],[20] Huang reported that as SRT increases, the soluble microbial product content in the matrix raises.[20]

Furthermore, the influence of initial N concentration on EPS concentration was surveyed and the results are presented in [Figure 4]b. As seen in [Figure 4]b and Mann–Whitney test depicted that with increasing initial N concentration from 0 to 10 mg/L as NO3, the carbohydrates concentration diminishes (P< 0.05) and then increased (P< 0.05). The variation of initial N concentration showed no significant effect on protein content. Amount of EPS production may have been affected by several conditions,[8] and presumably, N deficiency influences microbial species to produce more carbohydrate.

According to [Figure 4]c, the carbohydrate and protein concentration of sEPS was improved by increasing initial N concentration from 0 to 10 mg/L. Furthermore, further increase of initial N concentration (from 10 to 100 mg/L as NO3) led to depletion of carbohydrate production. The statistical analysis showed strong relationship between concentration of protein and COD in effluent. As soluble protein of effluent increases, the amount of COD in the effluent also amplified. This may be due to that the protein content of matrix has influenced on the effluent COD. Therefore, the appropriate level of N in influent wastewater is essential for good quality of effluent.

Sludge sedimentation properties

The sludge settling characteristics can be examined with microscopic observation of sludge or SVI. For this purpose, the variations of SVI during batch experiments were survived. Following the EPS and sEPS variations by initial nitrogen and SRT fluctuations, SVI was varied as shown in [Figure 5]. The amount of protein and protein-to-carbohydrate ratio in EPS has positively affected the value of SVI (Pearson correlation coefficient; 0.478 and 0.509, P < 0.05).
Figure 5: Variation of sludge volume index, protein content of extracellular polymeric substances, and carbohydrate of soluble extracellular polymeric substances as function of initial N concentration and solid retention time

Click here to view


Furthermore, the carbohydrate concentration in EPS showed negative effect on SVI but statistically insignificant (P > 0.05). The value of SVI fluctuated from below 70 mL/g to above 70 mL/g as carbohydrate in sEPS was reduced which was statistically significant (P< 0.05).

As EPS have high amount of protein and sEPS contain low concentration of carbohydrate, the SVI was increased and resulted in the weak sludge settleability (viscous bulking). With respect to this, the bacteria have negative charge and EPS consist of proteins with positive charge as amino function groups. Furthermore, carbohydrate with carboxyl and phosphate function groups carried negative charges.[11],[19] As EPS contain high level of protein (carried positive charge), the EPS covered bacterial cell and resulted in positive charge on the cell. Furthermore, low concentration of carbohydrate (negative charge) cannot neutralize protein positive charge and resulted in accumulation of positive charge in the environment. Under this circumstance, generation of repulsive force between same charge lead to poor sludge settleability. A similar result was noted by Liao et al. Furthermore, Sheng et al. reported the negative effect of EPS on SVI.[1],[11]

Kinetic coefficients

The different nitrogen concentration resulted in different biological kinetic coefficients. For this purpose, the Y coefficient (mg VSS/mg sCOD) was calculated for the studied initial nitrogen concentration base on the previous study.[21] The Y coefficient at initial N concentration of 0, 10, 25, 50, and 100 mg/L as NO3− was 0.32, 0.11, 0.23, 0.23, and 0.24 gVSS/gCOD, respectively. The highest value of Y was obtained at lowest initial concentration of nitrogen (N: 0 mg/L as NO3).

As mentioned before, the carbohydrate amount of EPS at zero initial nitrogen concentration was higher than other studied initial concentration of nitrogen. The highest Y value in lowest N concentration was apparent and not related to cell generation but was due to carbohydrate accumulation in matrix. In fact, the bacteria need all elemental components such as nitrogen for cell generation. In an environment with nitrogen deficiency, the bacteria cell imposed to change its normal metabolism way and produced high amount of compounds including carbohydrate. These results are in line with Eckenfelder.[22] Increase of EPS production as nitrogen decreases was observed in a study by Miqueleto et al., who reported that the Y value decreases as nitrogen concentration in matrix improve.[2]


  Conclusions Top


  • The results demonstrated that by changing of initial N concentration (increasing or decreasing), the COD removal efficiency remain constant
  • The results showed that with increasing SRT from 2 days to other studied SRT values (5, 10, and 15 days) led to significant decrease in carbohydrate of sludge
  • Statistical analysis demonstrated that the sEPS production has no significant correlation with SRT
  • With increasing initial N concentration from 0 to 10 mg/L as NO3, the carbohydrates concentration diminishes and then increased. The variation of initial N concentration showed no significant effect on protein content
  • The amount of protein and protein to carbohydrate ratio in EPS has positively affected the value of SVI. The carbohydrate concentration in EPS showed negative effect on SVI
  • The Y coefficient at initial N concentration of 0, 10, 25, 50, and 100 mg/L as NO3− was 0.32, 0.11, 0.23, 0.23, and 0.24 gVSS/gCOD, respectively. The highest value of Y was obtained at lowest initial concentration of nitrogen (N: 0 mg/L as NO3)
  • The proposed ANN modeling methodology can be used to predict the results.


Acknowledgment

The authors would like to thank Student Research Committee of Isfahan University of Medical Sciences, Isfahan, Iran.

Financial support and sponsorship

This study was financially supported by Isfahan University of Medical Sciences under research project No. 293229.

Conflicts of interest

There are no conflicts of interest.



 
  References Top

1.
Sheng GP, Yu HQ, Li XY. Extracellular polymeric substances (EPS) of microbial aggregates in biological wastewater treatment systems: A review. Biotechnol Adv 2010;28:882-94.  Back to cited text no. 1
    
2.
Miqueleto AP, Dolosic CC, Pozzi E, Foresti E, Zaiat M. Influence of carbon sources and C/N ratio on EPS production in anaerobic sequencing batch biofilm reactors for wastewater treatment. Bioresour Technol 2010;101:1324-30.  Back to cited text no. 2
    
3.
Xuan W, Bin Z, Zhiqiang S, Zhigang Q, Zhaoli C, Min J, et al. The EPS characteristics of sludge in an aerobic granule membrane bioreactor. Bioresour Technol 2010;101:8046-50.  Back to cited text no. 3
    
4.
Sponza DT. Investigation of extracellular polymer substances (EPS) and physicochemical properties of different activated sludge flocs under steady-state conditions. Enzyme Microb Technol 2003;32:375-85.  Back to cited text no. 4
    
5.
Ahmed Z, Cho J, Lim BR, Song KG, Ahn KH. Effects of sludge retention time on membrane fouling and microbial community structure in a membrane bioreactor. J Memb Sci 2007;287:211-8.  Back to cited text no. 5
    
6.
Yang SF, Li XY. Influences of extracellular polymeric substances (EPS) on the characteristics of activated sludge under non-steady-state conditions. Proc Biochem 2009;44:91-6.  Back to cited text no. 6
    
7.
Linton J, Jones D, Woodard S. Factors that control the rate of exopolysaccharide production by Agrobacterium radiobacter NCIB 11883. J Gen Microbiol 1987;133:2979-87.  Back to cited text no. 7
    
8.
Wang H, Deng H, Ma L, Ge L. Influence of operating conditions on extracellular polymeric substances and surface properties of sludge flocs. Carbohydr Polym 2013;92:510-5.  Back to cited text no. 8
    
9.
Liu Y, Liu H, Cui L, Zhang K. The ratio of food-to-microorganism (F/M) on membrane fouling of anaerobic membrane bioreactors treating low-strength wastewater. Desalination 2012;297;97-103.  Back to cited text no. 9
    
10.
Ye F, Ye Y, Li Y. Effect of C/N ratio on extracellular polymeric substances (EPS) and physicochemical properties of activated sludge flocs. J Hazard Mater 2011;188:37-43.  Back to cited text no. 10
    
11.
Liao BQ, Allen DG, Droppo IG, Leppard GG, Liss SN. Surface properties of sludge and their role in bioflocculation and settleability. Water Res 2001;35:339-50.  Back to cited text no. 11
    
12.
Sesay ML, Ozcengiz G, Dilek Sanin F. Enzymatic extraction of activated sludge extracellular polymers and implications on bioflocculation. Water Res 2006;40:1359-66.  Back to cited text no. 12
    
13.
Taheri E, Khiadani MH, Amin MM, Nikaeen M, Hassanzadeh A. Treatment of saline wastewater by a sequencing batch reactor with emphasis on aerobic granule formation. Bioresour Technol 2012;111;21-6.  Back to cited text no. 13
    
14.
Rice EW, American Public Health Association, American Water Works Association, Water Environment Federation. Standard Methods for Examination of Water and Wastewater 2012. American Public Health Association; 2012.  Back to cited text no. 14
    
15.
Adav SS, Lee DJ. Extraction of extracellular polymeric substances from aerobic granule with compact interior structure. J Hazard Mater 2008;154:1120-6.  Back to cited text no. 15
    
16.
Brummer Y, Cui SW. Understanding carbohydrate analysis. Food Carbohydrates: Chemistry, Physical Properties and Applications. Taylor and Francis: CRC; 2005.  Back to cited text no. 16
    
17.
Waterborg JH. The lowry method for protein quantification. The Protein Protocols Handbook. New Jersey, USA: Springer; 2009. p. 7-10.  Back to cited text no. 17
    
18.
Feng S, Zhang N, Liu H, Du X, Liu Y, Lin H. The effect of COD/N ratio on process performance and membrane fouling in a submerged bioreactor. Desalination 2012;285:232-8.  Back to cited text no. 18
    
19.
Li XY, Yang SF. Influence of loosely bound extracellular polymeric substances (EPS) on the flocculation, sedimentation and dewaterability of activated sludge. Water Res 2007;41:1022-30.  Back to cited text no. 19
    
20.
Huang Z, Ong SL, Ng HY. Submerged anaerobic membrane bioreactor for low-strength wastewater treatment: Effect of HRT and SRT on treatment performance and membrane fouling. Water Res 2011;45:705-13.  Back to cited text no. 20
    
21.
Amin MM, Khiadani MH, Fatehizadeh A, Taheri E. Validation of linear and non-linear kinetic modeling of saline wastewater treatment by sequencing batch reactor with adapted and non-adapted consortiums. Desalination 2014;344:228-35.  Back to cited text no. 21
    
22.
Eckenfelder WW. Industrial Water Pollution Control. University of Michigan, USA: McGraw-Hill; 1989.  Back to cited text no. 22
    


    Figures

  [Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5]



 

Top
Previous article  Next article
 
  Search
 
Similar in PUBMED
   Search Pubmed for
   Search in Google Scholar for
 Related articles
Access Statistics
Email Alert *
Add to My List *
* Registration required (free)

 
  In this article
Abstract
Introduction
Materials and Me...
Results and Disc...
Conclusions
References
Article Figures

 Article Access Statistics
    Viewed264    
    Printed30    
    Emailed0    
    PDF Downloaded70    
    Comments [Add]    

Recommend this journal