- Email: [email protected]
An evaluation of tannery industry wastewater treatment sludge gasification by artificial neural network modeling
Journal of Hazardous Materials 263 (2013) 361–366
Contents lists available at ScienceDirect
Journal of Hazardous Materials journal homepage: www.elsevier.com/locate/jhazmat
An evaluation of tannery industry wastewater treatment sludge gasification by artificial neural network modeling Atakan Ongen ∗ , H. Kurtulus Ozcan, Semiha Arayıcı Istanbul University, Faculty of Engineering, Department of Environmental Engineering, 34320 Avcilar, Istanbul, Turkey
h i g h l i g h t s • • • •
We model calorific value of syn-gas from tannery industry treatment sludge. We monitor variation of gas composition in produced gas. Heating value of produced gas is around 1500 kcal/m3 . Model predictions are in close accordance with real values.
a r t i c l e
i n f o
Article history: Received 30 November 2012 Received in revised form 11 March 2013 Accepted 16 March 2013 Available online 1 April 2013 Keywords: Tannery industry Thermochemical treatment Synthesis gas ANN
a b s t r a c t This paper reports on the calorific value of synthetic gas (syngas) produced by gasification of dewatered sludge derived from treatment of tannery wastewater. Proximate and ultimate analyses of samples were performed. Thermochemical conversion alters the chemical structure of the waste. Dried air was used as a gasification agent at varying flow rates, which allowed the feedstock to be quickly converted into gas by means of different heterogeneous reactions. A lab-scale updraft fixed-bed steel reactor was used for thermochemical conversion of sludge samples. Artificial neural network (ANN) modeling techniques were used to observe variations in the syngas related to operational conditions. Modeled outputs showed that temporal changes of model predictions were in close accordance with real values. Correlation coefficients (r) showed that the ANN used in this study gave results with high sensitivity. © 2013 Elsevier B.V. All rights reserved.
1. Introduction Conversion of organic materials to gaseous fuel through thermal processes is regarded to have a significant potential to contribute to energy problem. Since it is estimated that fossil-derived oil and natural gas reserves will be depleted in the future, this potential emphasized the need to use wastes and other sustainable materials as energy sources [1–3]. In recent years, gasification began to be a popular method to dispose of sludge containing large amounts of organic substances; like the derivation of sewage sludge by wastewater treatment [4–6]. Various treatment methods such as thickening, digestion, dewatering, drying, and lime application are used to transform sludge into biosolids, to provide gasification [7]. Although sludge might be composted and spread on soil, it is known to contain heavy metals, other organic micropollutants and polycyclic aromatic hydrocarbons (PAHs). It poses a health risk for humans. Furthermore, it is the combination of various wastes, which lacks proper control. Thus, it can contain traces of highly
∗ Corresponding author. Tel.: +90 212 4737070/17724; fax: +90 212 473 71 80. E-mail addresses: [email protected], [email protected] (A. Ongen). 0304-3894/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jhazmat.2013.03.043
unhealthy pollutants. Minimization of sludge waste volume and reduction of disposal area is possible by gasification. Considering that gasification of biomass involves thermochemical processes, temperature and heating significantly affect weight loss of biomass [8,9]. As gasification is an endothermic process, it requires transfer of heat to the reactor. It involves a distinction between direct-heated gasifiers (authotherm), which use partial oxidation to produce the required heat demand, and indirect-heated gasifiers (allotherm), which are externally heated mostly by the reactor wall of heat carriers. There is a strong link between the heat transfer method and gasification agent and heating value, amount and composition of the product gas. A nitrogen-diluted low-calorific gas is produced by air-blown direct-heated gasifiers, while a medium-calorific gas is produced by oxygen-blown or indirect-heated gasifiers [10]. The temperature and heating rates have significant impacts on the weight loss of biomass as pyrolysis and gasification of biomass are thermochemical processes. Temperatures in excess of 500 ◦ C are generally used to reduce carbon dioxide by carbon to carbon monoxide and thus to obtain maximum yield of char with inert medium flowing, pyrolysis is conducted at temperatures below 500 ◦ C [8,11].
362
A. Ongen et al. / Journal of Hazardous Materials 263 (2013) 361–366
Production of gas during pyrolysis and gasification processes is commonly used to produce electricity at industrial scale. Judex et al. [12] analyzed two pilot plants based in Germany (Balingen and Mannheim) which were fueled by sewage sludge, and reported typical gas data for the mentioned plants. Each plant had producer gas with a calorific value of 765 kcal/m3 and 1123 kcal/m3 respectively. The authors reported that gasification was conducted successfully in fluidized bed and gas cleaning using granular bed filter. Moltó et al. [13] investigated gasification and pyrolysis of physico-chemical (PC) and a biological (BIO) sludge. Low heating rate and parallel flow using both sludges were found to give the highest pyrolysis ratio; maximum CO emission, i.e., the worst combustion conditions, with minimum contact time and high heating rates. Ouadi et al. [14] investigated gasification of pre-conditioned rejects and de-inking sludge mixtures with wood chips in a fixed bed gasifier. The researchers obtained a gas with a composition of 16.24% H2 , 23.34% CO, 5.21% CH4 and having a higher heating value (1744 kcal/m3 ). They reported that paper mill rejects can be gasified in a fixed bed downdraft gasifier. Nipattummakul et al. [15] reported that, following fuel reforming with steam gasification, wastewater sludge can be a good source of fuel. They used a semi-batch gasifier to determine evolutionary behavior of syngas properties by employing calibrated experiments and diagnostic facilities. Nipattummakul et al. [16] used steam as the gasifying agent and reported that when compared to air gasification, this increased hydrogen yield by three times. The researchers compared sewage sludge gasification results with other samples including paper, food wastes, and plastics. Hightemperature steam was used as the gasifying agent to gasify sewage sludge samples. The temperature of the reactor was adjusted to 700, 800, 900, and 1000 ◦ C. It was reported that low reaction rates at 700 ◦ C showed the necessity of using higher temperatures (>800 ◦ C) to carry our sewage sludge gasification. On the other hand, low reaction rates lead to high steam consumption and prolonged gasification times. An artificial neural network (ANN) is a mathematical structure, which is designed to mimic the biological neural system in terms of the information processing functions of neurons. These systems show high parallelism; they use various interconnected units to process information. These units react to inputs via convertible weights, thresholds and mathematical functions [17]. Artificial neural networks especially address to the problems in large data sets which contain complicated, non-linear relationships in a wide variety of inputs. They are used to solve various types of problems in fields of engineering and science, especially where traditional modeling methods fails to be productive. These models can identify complex patterns in datasets, which cannot be described by simple mathematical formulae. When well-trained, an ANN can serve as a predictive model for a specific application including emission analysis [18], nanofiltration applications [19], water quality [20–23], water distribution strategies [24], and risk analysis [25]. This study aimed to analyze the possibility of converting wastewater treatment sludge into a source of energy in the form of syngas, employing ANN modeling to detect the variations in syngas calorific value regarding operational conditions. 2. Experimental 2.1. ANN modeling The most widely used neural network method is the backpropagation training algorithm, which was used in the present study [20,26–29]. Fig. 1 shows this type of model. The aim of an ANN is to construct a data-generating process model to enable the network to generalize and predict outputs from newly introduced
Fig. 1. A representation of a simple 3-layer back-propagation artificial neural network.
inputs. This learning algorithm involves a multi-layered neural network and contains an input layer, hidden layers and an output layer. The neurons in the hidden and output layer multiply each input by a corresponding weight; sum the product and the process the sum by employing a non-linear transfer function [25]. In the present model, the neural network is trained and tested using MATLAB 7.0. A three-layer neural network that consists of an input layer, output layer, and one hidden layer is used. Data (H2 , O2 , CH4 , CO, CO2 , N2 , and calorific value – Cv) obtained during the study period were designed to meet the requirements of training and testing the neural network. This database is divided into training and testing sets, taking the odd-numbered patterns as training data and even-numbered ones as test data. In the data set arrangement, input variables consist on the concentrations of seven parameters (H2 , O2 , CH4 , CO, CO2 , N2 and Cv) during time “t”, and the output variable (Cv) is the concentrations at time “t + 1”. Arranged input and output data sets were used in training and test phases. The training coefficient was chosen as (lr) 0.1. The number of neurons within the hidden layer is an important factor whereas the number cascade connected to the hidden layer is less influential. In the present study, the input, hidden and output layers of the ANN contained 7, 12, and 1 neurons, respectively. 2.2. Sample characterization Sludge derived from treated wastewater from the tannery industry was used as gasifier fuel. The sludge was collected after dewatering. Dewatered samples were dried at 105 ◦ C for 24 h. Ultimate and proximate analyses of sludge samples are shown in Table 1. Each experiment was performed according to Standard Methods [30]. Elemental analyses were carried out using Thermo Finnigan Flash EA 1112 instrument. 2.3. Thermal conversion and calorific value calculation Gasification experiments were performed in an updraft fixedbed steel reactor with a volume of 2 L. The reactor was heated Table 1 Sludge characterization. Parameters, wt%
Tannery sludge
Ash Volatile matter Moisture C H N
58.8 41.1 4.5 20.6 4.5 2.3
A. Ongen et al. / Journal of Hazardous Materials 263 (2013) 361–366 Table 2 Higher (HHV) of some common fuels [31].
CO CH4 H2
363
Table 5 Syngas composition – E3 – 0.2 L/min dried air – 20 g sample.
Density, kg/m3
MJ/kg
MJ/m3
kcal/m3
Time, min
T, ◦ C
H2 , %
O2 , %
CH4 , %
CO, %
CO2 , %
N2 , %
1.25 0.714 0.089
10.1 55.5 141.8
13.6 39.8 12.8
3019 9470 3025
5 10 15 20 25 30 35 40 45 50 55 60 65 90 120
200 300 400 450 500 550 600 650 700 750 800 850 900 950 1000
0.00 0.00 0.00 0.05 0.08 0.27 0.81 2.19 5.11 10.52 11.81 11.06 12.03 9.20 9.20
15.14 17.87 18.49 17.87 17.15 8.53 3.75 1.29 0.38 0.10 0.00 0.08 0.17 1.20 3.20
0.00 0.00 0.00 0.00 0.02 0.14 0.42 1.48 6.55 5.33 4.71 2.47 3.24 4.70 3.70
0.00 0.00 0.00 0.00 0.22 0.97 1.54 5.00 8.80 10.30 14.24 15.24 19.24 9.30 8.56
0.04 0.04 0.24 0.84 1.85 8.56 10.88 15.00 19.00 17.00 18.00 14.50 16.00 18.36 20.36
67.59 63.94 68.52 70.88 69.95 70.75 67.82 59.15 52.75 43.98 38.09 39.26 42.43 48.00 51.23
Table 3 Syngas composition – E1 – 0.05 L/min dried air – 20 g sample. Time, min
T, ◦ C
H2 , %
O2 , %
CH4 , %
CO, %
CO2 , %
N2 , %
5 10 15 20 25 30 35 40 45 50 55 60 65 90 120
200 300 400 450 500 550 600 650 700 750 800 850 900 950 1000
0.00 0.00 0.00 0.00 0.12 0.09 0.12 0.09 0.98 5.38 12.66 11.13 5.68 4.23 5.12
15.00 16.50 16.33 17.50 19.15 17.73 11.44 16.44 6.77 3.01 0.04 0.10 0.00 1.23 2.33
0.00 0.00 0.00 0.00 0.02 0.04 0.08 0.14 0.51 4.46 6.21 5.47 3.39 1.10 1.00
0.00 0.00 0.00 0.00 0.05 0.12 0.16 0.21 1.45 3.53 12.41 16.02 15.44 14.50 13.23
0.00 0.00 0.00 0.00 2.16 7.65 8.15 4.00 3.97 18.65 18.41 16.50 18.90 13.00 16.56
68.50 65.00 63.50 65.50 68.70 68.09 70.65 69.43 65.34 55.00 48.24 45.37 45.16 43.00 54.36
indirectly by a ceramic resistance until the operational temperature of 750 ◦ C was achieved. Approximate calorific values of producer gases were calculated according to H2 , CO and CH4 volumetric percentages determined by a micro-GC (Varian CP-4900). Dried air was used as a gasification agent. Heating values were calculated according to higher heating values (HHV), as shown in Table 2.
Table 6 Syngas composition – E4 – 0.3 L/min dried air – 20 g sample. Time, min
T, ◦ C
H2 , %
O2 , %
CH4 , %
CO, %
CO2 , %
N2 , %
5 10 15 20 25 30 35 40 45 50 55 60 65 90 120
200 300 400 450 500 550 600 650 700 750 800 850 900 950 1000
0.00 0.00 0.00 0.07 0.31 0.59 1.58 4.37 6.16 9.27 8.67 6.55 6.56 6.21 4.75
17.00 16.73 14.68 18.29 17.06 12.56 10.95 10.34 7.10 3.03 3.00 2.35 3.00 2.36 2.20
0.00 0.00 0.00 0.00 0.17 0.33 0.92 2.40 3.29 5.16 4.60 6.50 5.55 5.60 6.10
0.00 0.00 0.00 0.12 1.10 1.49 2.02 2.73 2.50 5.50 6.00 7.50 8.89 7.30 9.20
0.07 0.06 3.39 1.32 1.84 6.38 9.55 17.19 17.98 16.80 16.82 16.07 18.36 21.36 24.50
68.23 67.35 69.96 69.81 70.35 70.17 65.62 59.13 60.09 55.01 55.10 57.04 55.60 55.60 50.30
3. Results and discussion Syngas compositions from each experiment are shown in Tables 3–8. CO, CH4 , and H2 values were used to calculate calorific values for each sampling time. Under the condition of 0.05 L/min dried-air flow, gasification produced 12% H2 , 12% CO and 6% CH4 . The effective temperature for the process exceeded 750 ◦ C. CO2 levels did not exceed 19%. The experiments were carried out for 2 h. Under 0.1 L/min dried air flow, gasification achieved 12% H2 , 17% CO and 4% CH4 . The effective temperature for the process exceeded 750 ◦ C. CO2 levels did not exceed 15%. Syngas composition did not show a significant change when compared to Table 3. As seen in Table 5, the highest CO level was 19%, at more than 800 ◦ C, where CO2 levels also increased. Experiments were carried out with 0.2 L/min dried-air flow, which achieved 12% H2 , 17% CO, Table 4 Syngas composition – E2 – 0.1 L/min dried air – 20 g sample.
Table 7 Syngas composition – E5 – 0.4 L/min dried air – 20 g sample. Time, min
T, ◦ C
H2 , %
O2 , %
CH4 , %
CO, %
CO2 , %
N2 , %
5 10 15 20 25 30 35 40 45 50 55 60 65 90 120
200 300 400 450 500 550 600 650 700 750 800 850 900 950 1000
0.00 0.00 0.00 0.00 0.09 0.09 0.35 0.84 1.75 1.97 5.97 3.54 4.25 4.12 3.58
18.00 16.50 17.50 20.32 18.32 13.32 9.86 5.16 1.76 5.60 6.20 5.36 6.20 5.00 5.68
0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.41 1.09 1.39 3.13 2.30 1.94 2.00 1.50
0.00 0.00 0.00 0.00 0.00 0.00 0.98 1.56 2.20 2.34 3.02 8.68 7.86 10.20 6.50
0.00 0.00 0.05 0.56 0.97 0.97 5.75 6.04 14.87 15.96 15.88 16.40 13.87 18.25 20.10
58.00 60.00 63.00 69.20 70.29 69.29 68.95 70.22 68.95 70.90 62.66 59.64 64.45 56.98 55.23
Table 8 Syngas composition – E6 – 0.5 L/min dried air – 20 g sample.
Time, min
T, ◦ C
H2 , %
O2 , %
CH4 , %
CO, %
CO2 , %
N2 , %
Time, min
T, ◦ C
H2 , %
O2 , %
CH4 , %
CO, %
CO2 , %
N2 , %
5 10 15 20 25 30 35 40 45 50 55 60 65 90 120
200 300 400 450 500 550 600 650 700 750 800 850 900 950 1000
0.00 0.00 0.04 0.08 0.18 0.28 0.50 1.59 4.69 11.78 12.50 12.16 9.00 4.20 3.60
17.50 17.60 18.58 17.37 15.05 10.05 6.16 2.98 0.89 0.21 0.06 0.23 1.00 1.20 3.20
0.00 0.00 0.00 0.00 0.04 0.14 0.22 0.63 1.97 3.97 3.27 2.50 3.97 2.90 3.97
0.00 0.00 0.00 0.08 0.62 1.02 1.29 1.76 2.49 16.50 17.00 16.72 13.82 15.24 14.13
0.00 0.01 0.88 0.71 3.69 2.69 1.00 2.01 18.00 17.00 12.00 13.31 13.55 15.36 14.36
70.50 68.00 68.45 69.01 70.67 69.67 69.49 64.90 55.94 38.11 36.00 38.00 47.91 48.00 49.23
5 10 15 20 25 30 35 40 45 50 55 60 65 90 120
200 300 400 450 500 550 600 650 700 750 800 850 900 950 1000
0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 3.38 2.20 3.08 1.50 2.30 2.00 2.35
18.11 18.00 17.05 18.51 17.00 15.50 13.20 11.00 8.00 4.00 1.00 1.10 1.00 2.30 2.45
0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1.38 1.50 2.00 3.50 3.00 2.50 3.00
0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1.55 1.19 4.16 5.30 5.69 4.50 3.20
0.00 0.00 0.00 0.06 0.06 0.11 0.61 1.05 22.00 23.20 23.00 20.00 26.50 29.30 28.50
69.61 66.14 68.61 66.91 70.00 67.67 68.57 68.97 60.00 58.00 56.00 55.30 54.00 52.36 55.20
A. Ongen et al. / Journal of Hazardous Materials 263 (2013) 361–366
Heating value, kcal/m3
E1
E2
E3
E4
E5
Table 9 Correlation coefficient for ANN outputs.
E6
1500
900
1250
750
1000
600
750
450
500
300
250
150
Temperature, oC
364
0
0 0
15
30
45
60
75
90
105
120
Time, min. Fig. 2. Syngas heating values of each experiment.
and 4% CH4 . The effective temperature for the process exceeded 700 ◦ C. Syngas production during gasification was negatively affected if the amount of dried air exceeded the optimum level (Tables 6–8). Each gas component was analyzed at the lowest level where CO2 reached the highest percentage by 30%. Fig. 2 shows heating values calculated for syngases produced from each experiment. In the experiments, the gasification process used dried air as a gasification agent to convert carbonaceous materials into gaseous products. Since heating value is a function of gas composition, the calculations were based on the gas composition results. In natural gas, the heating value is more dependent on gas composition. If the composition is known, the heating value can be calculated using: hu =
xi × hui
(1)
where xi is the combustible proportion of gas and hui is its heating value.
a
Experiment number
r
E1 E2 E3 E4 E5 E6
0.68 0.85 0.92 0.82 0.97 0.98
As a rule, the proportions follow the sequence CH4 , H2 and CO. In the present study, CO and CH4 contents increased with temperature, whereas CO2 content decreased. The highest heating values (1000–1500 kcal/m3 ) were achieved for E1, E2 and E3, with flow rates of 0.05, 0.1, and 0.2 L/min, respectively, at 700 ◦ C. The operation time was between 45 and 65 min. The calculated values were used for ANN modeling in the second phase of the study. ANN modeling used values for H2 , O2 , CH4 , CO, CO2 , N2 , and Cv, and the output data consisted of the change in Cv concentration at time t + 1. The outputs of the ANN-modeled structure are presented in Figs. 3–5. Correlations between observed data and prediction were calculated, and the results are shown in Table 9. For the ANN structure, the r-values calculated for the testing phase were between 0.68 and 0.98. As these rates are close to 1, it means that it is a good modeling. Correlation coefficients are one of the most important indices used for artificial neural network (ANN) model performance. Many studies in the literature investigated ANN forecast modeling techniques. Pashova and Popova [32] used ANN to forecast daily sea levels at Burgas, Bulgaria. ANN performance was controlled by correlation coefficient and their values found 0.72–0.95. Palani et al. [33], presented an ANN application to predict atmospheric nitrogen deposition. Total and organic nitrogen were predicted by this model
b 1200 observed predicted
900
Calorific Value
Calorific Value
1200
600 300
2
3
4
5
6
7
predicted
600 300
0 1
observed
900
0
8
1
2
3
4
5
6
7
8
7
8
Fig. 3. (a) ANN outputs for predicted calorific value (E1); (b) ANN outputs for predicted calorific value (E2).
a
b 1200
1200 Calorific Value
Calorific Value
observed
predicted
900 600 300
observed predicted
900 600 300
0
0 1
2
3
4
5
6
7
8
1
2
3
4
5
6
Fig. 4. (a) ANN outputs for predicted calorific value (E3); (b) ANN outputs for predicted calorific value (E4).
A. Ongen et al. / Journal of Hazardous Materials 263 (2013) 361–366
a
b 1200
1200
observed predicted
Calorific Value
Calorific Value
365
900 600 300
observed predicted
900 600 300
0
0 1
2
3
4
5
6
7
1
2
3
4
5
6
7
8
Fig. 5. (a) ANN outputs for predicted calorific value (E5); (b) ANN outputs for predicted calorific value (E6).
and correlation coefficients ranged from 0.93 and 0.99 respectively. Feng et al. [34] forecast ozone concentration, and reported correlation coefficients between 0.40 and 0.87. The correlation coefficient of our model (Table 9) shows r values similar to those of previous studies. 4. Conclusion The present study examined the production of syngas by gasification of sludge derived from treatment of tannery wastewater. Dried air was used as a gasification agent, at flow rates between 0.05 and 0.5 L/min. Since heating value is dependent on gas composition, the CO, CH4 , and H2 contents of syngases were analyzed. It was found that combustible gas composition increased with increasing temperature in excess of 600 ◦ C. Calorific values of syngases were calculated for each experiment, and the results were modeled with an ANN. Air-flow rate between 0.05 and 0.1 L/min was found to be optimum for the sludge samples used in the study, and produced the highest heating values of 1000–1500 kcal/m3 . For the ANN structure, the r values of the testing phase were between 0.68 and 0.98. Modeled outputs showed that temporal changes of modeled predictions were in close accordance with observed values. Correlation coefficient values showed that the ANN used in this study gave results in high accuracy. Finally, the proposed ANN model can be used to model gasification products without further adaptation. It can also be implemented in different time series or different model structures. References [1] W-J. Lee, S-D. Kim, B.-H. Song, Steam gasification of coal with salt mixture of potassium and nickel in a fluidized bed reactor, Korean J. Chem. Eng. 18 (5) (2011) 640–645. [2] F. Pinto, R. André, C. Franco, H. Lopes, C. Carolino, R. Costa, I. Gulyurtlu, Cogasification of coal and wastes in a pilot-scale installation. 2: effect of catalysts in syngas treatment to achieve sulphur and nitrogen compounds abatement, Fuel. 89 (2010) 3340–3351. [3] A. Fuente-Cuesta, M.A. Lopez-Anton, M. Diaz-Somoano, A.VanZomeren, M. Cieplik, M.R. Martínez-Tarazona, Leaching of major and trace elements from paper–plastic gasification chars: an experimental and modelling study, J. Hazard. Mater. 244–245 (2013) 70–80. [4] X. Gang, J. Bao-sheng, Z. Zhao-ping, C. Yong, N. Ming-jiang, C. Ke-fa, X. Rui, H. Ya-ji, H. He, Experimental study on MSW gasification and melting technology, J. Environ. Sci. 19 (2007) 1398–1403. [5] T. Malkow, Novel and innovative pyrolysis and gasification technologies for energy efficient and environmentally sound MSW disposal, Waste Manage. 24 (2004) 53–79. [6] L. Peng, Y. Wang, Z. Lei, G. Cheng, Co-gasification of wet sewage sludge and forestry waste in situ steam agent, Bioresour. Technol. 114 (2012) 698–702. [7] P. Mondal, G.S. Dang, M.O. Garg, Syngas production through gasification and cleanup for downstream applications—recent developments, Fuel Process. Technol. 92 (2011) 1395–1410.
[8] V. Kirubakaran, V. Sivaramakrishnan, R. Nalini, T. Sekar, M. Premalatha, P. Subramanian, A review on gasification of biomass, Renew. Sust. Energy Rev. 13 (2009) 179–186. [9] Y. Zhang, B. Li, H. Li, B. Zhang, Exergy analysis of biomass utilization via steam gasification and partial oxidation, Thermochim. Acta 538 (2012) 21–28. [10] U. Zuberbühler, Gasification of biomass—an overview on available technologies, in: 1st European Summer School on Renewable Motor Fuels, Birkenfeld, Germany, 29–31 August, 2005. [11] J.A. Ruiz, M.C. Juarez, M.P. Morales, P. Munoz, M.A. Mendivil, Biomass gasification for electricity generation: review of current technology barriers, Renew. Sust. Energy Rev. 18 (2013) 174–183. [12] J.W. Judex, M. Gaiffi, H.C. Burgbacher, Gasification of dried sewage sludge: status of the demonstration and the pilot plant, Waste Manage. 32 (2012) 719–723. [13] J. Moltó, et al., Gas production during the pyrolysis and gasification of biological and physico-chemical sludges from oil refinery, J. Anal. Appl. Pyrol. (2012), http://dx.doi.org/10.1016/j.jaap.2012.09.012. [14] M. Ouadi, J.G. Brammer, M. Kay, A. Hornung, Fixed bed downdraft gasification of paper industry wastes, Appl. Energy 103 (2013) 692–699. [15] N. Nipattummakul, I. Ahmed, S. Kerdsuwan, A.K. Gupta, High temperature steam gasification of wastewater sludge, Appl. Energy 87 (2010) 3729–3734. [16] N. Nipattummakul, I. Ahmed, S. Kerdsuwan, A.K. Gupta, Hydrogen and syngas production from sewage sludge via steam gasification, Int. J. Hydrogen Energy 35 (2010) 11738–11745. [17] S.A. Rounds, Development of a neural network model for dissolved oxygen in the Tualatin River, Oregon, in: Proceedings of the Second Federal Interagency Hydrologic Modeling Conference, Las Vegas, Nevada, July 29–August 1, 2002, Subcommittee on Hydrology of the Interagency Advisory Committee on Water Information, 2002. [18] B. Ghobadian, H. Rahimi, A.M. Nikbakht, G. Najafi, T.F. Yusaf, Diesel engine performance and exhaust emission analysis using waste cooking biodiesel fuel with an artificial neural network, Renew. Energy 34 (2009) 976–982. [19] H. Al-Zoubi, N. Hilal, N.A. Darwish, A.W. Mohammad, Rejection and modelling of sulphate and potassium salts by nanofiltration membranes: neural network and Spiegler–Kedem model, Desalination 206 (2007) 42–60. [20] K.P. Singh, A. Basant, A. Malik, G. Jain, Artificial neural network modeling of the river water quality—a case study, Ecol. Model. 220 (2009) 888–895. [21] N. Balkaya, H.K. Ozcan, O.N. Ucan, Determination of relationship between hardness and groundwater quality parameters by neural networks, Desalination Water Treat. 11 (2009) 258–263. [22] I.S. Yeon, J.H. Kim, K.W. Jun, Application of artificial intelligence models in water quality forecasting, Environ. Technol 29 (6) (2008) 625–631. [23] I.S. Yeon, K.W. Jun, H.J. Lee, The improvement of total organic carbon forecasting using neural networks discharge model, Environ. Technol 30 (1) (2009) 45–51. [24] R. Jafar, I. Shahrour, I. Juran, Application of artificial neural networks (ANN) to model the failure of urban water mains, Math. Comput. Model. 51 (2010) 1170–1180. [25] B. Pradhan, S. Lee, Landslide risk analysis using artificial neural network model focusing on different training sites, Int. J. Phys. Sci. 4 (1) (2009) 1–15. [26] H.K. Ozcan, O.N. Ucan, U. Sahin, M. Borat, C. Bayat, Artificial neural network modeling of methane emissions at Istanbul Kemerburgaz-Odayeri Landfill Site, J. Sci. Ind. Res. 65 (2) (2006) 128–134. [27] E. Cadenas, W. Rivera, Wind speed forecasting in three different regions of Mexico, using a hybrid ARIMA-ANN model, Renew. Energy 35 (2010) 2732–2738. [28] R. Noori, A. Karbassi, M.S. Sabahi, Evaluation of PCA and Gamma test techniques on ANN operation for weekly solid waste prediction, J. Environ. Manage. 91 (2010) 767–771. [29] A. C¸elekli, F. Geyik, Artificial neural networks (ANN) approach for modeling of removal of Lanaset Red G on Characontraria, Bioresour. Technol. 102 (2011) 5634–5638.
366
A. Ongen et al. / Journal of Hazardous Materials 263 (2013) 361–366
[30] Standard Methods, Standard Methods for the Examination of Water and Wastewater, 21st edition, AWWA/APHA/WEF, 2005. [31] U. Bossel, Well-to-Wheel Studies, Heating Values, in: and the Energy Conservation Principle, European Fuel Cell Forum, Oberrohrdorf, Switzerland, 2003. [32] L. Pashova, S. Popova, Daily sea level forecast at tide gauge Burgas, Bulgaria using artificial neural networks, J. Sea Res. 66 (2011) 154–161.
[33] S. Palani, P. Tkalich, R. Balasubramanian, J. Palanichamy, ANN application for prediction of atmospheric nitrogen deposition to aquatic ecosystems, Mar. Pollut. Bull. 62 (2011) 1198–1206. [34] Y. Feng, W. Zhang, D. Sun, L. Zhang, Ozone concentration forecast method based on genetic algorithm optimized back propagation neural networks and support vector machine data classification, Atmos. Environ. 45 (2001) 1979–1985.
Contents lists available at ScienceDirect
Journal of Hazardous Materials journal homepage: www.elsevier.com/locate/jhazmat
An evaluation of tannery industry wastewater treatment sludge gasification by artificial neural network modeling Atakan Ongen ∗ , H. Kurtulus Ozcan, Semiha Arayıcı Istanbul University, Faculty of Engineering, Department of Environmental Engineering, 34320 Avcilar, Istanbul, Turkey
h i g h l i g h t s • • • •
We model calorific value of syn-gas from tannery industry treatment sludge. We monitor variation of gas composition in produced gas. Heating value of produced gas is around 1500 kcal/m3 . Model predictions are in close accordance with real values.
a r t i c l e
i n f o
Article history: Received 30 November 2012 Received in revised form 11 March 2013 Accepted 16 March 2013 Available online 1 April 2013 Keywords: Tannery industry Thermochemical treatment Synthesis gas ANN
a b s t r a c t This paper reports on the calorific value of synthetic gas (syngas) produced by gasification of dewatered sludge derived from treatment of tannery wastewater. Proximate and ultimate analyses of samples were performed. Thermochemical conversion alters the chemical structure of the waste. Dried air was used as a gasification agent at varying flow rates, which allowed the feedstock to be quickly converted into gas by means of different heterogeneous reactions. A lab-scale updraft fixed-bed steel reactor was used for thermochemical conversion of sludge samples. Artificial neural network (ANN) modeling techniques were used to observe variations in the syngas related to operational conditions. Modeled outputs showed that temporal changes of model predictions were in close accordance with real values. Correlation coefficients (r) showed that the ANN used in this study gave results with high sensitivity. © 2013 Elsevier B.V. All rights reserved.
1. Introduction Conversion of organic materials to gaseous fuel through thermal processes is regarded to have a significant potential to contribute to energy problem. Since it is estimated that fossil-derived oil and natural gas reserves will be depleted in the future, this potential emphasized the need to use wastes and other sustainable materials as energy sources [1–3]. In recent years, gasification began to be a popular method to dispose of sludge containing large amounts of organic substances; like the derivation of sewage sludge by wastewater treatment [4–6]. Various treatment methods such as thickening, digestion, dewatering, drying, and lime application are used to transform sludge into biosolids, to provide gasification [7]. Although sludge might be composted and spread on soil, it is known to contain heavy metals, other organic micropollutants and polycyclic aromatic hydrocarbons (PAHs). It poses a health risk for humans. Furthermore, it is the combination of various wastes, which lacks proper control. Thus, it can contain traces of highly
∗ Corresponding author. Tel.: +90 212 4737070/17724; fax: +90 212 473 71 80. E-mail addresses: [email protected], [email protected] (A. Ongen). 0304-3894/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jhazmat.2013.03.043
unhealthy pollutants. Minimization of sludge waste volume and reduction of disposal area is possible by gasification. Considering that gasification of biomass involves thermochemical processes, temperature and heating significantly affect weight loss of biomass [8,9]. As gasification is an endothermic process, it requires transfer of heat to the reactor. It involves a distinction between direct-heated gasifiers (authotherm), which use partial oxidation to produce the required heat demand, and indirect-heated gasifiers (allotherm), which are externally heated mostly by the reactor wall of heat carriers. There is a strong link between the heat transfer method and gasification agent and heating value, amount and composition of the product gas. A nitrogen-diluted low-calorific gas is produced by air-blown direct-heated gasifiers, while a medium-calorific gas is produced by oxygen-blown or indirect-heated gasifiers [10]. The temperature and heating rates have significant impacts on the weight loss of biomass as pyrolysis and gasification of biomass are thermochemical processes. Temperatures in excess of 500 ◦ C are generally used to reduce carbon dioxide by carbon to carbon monoxide and thus to obtain maximum yield of char with inert medium flowing, pyrolysis is conducted at temperatures below 500 ◦ C [8,11].
362
A. Ongen et al. / Journal of Hazardous Materials 263 (2013) 361–366
Production of gas during pyrolysis and gasification processes is commonly used to produce electricity at industrial scale. Judex et al. [12] analyzed two pilot plants based in Germany (Balingen and Mannheim) which were fueled by sewage sludge, and reported typical gas data for the mentioned plants. Each plant had producer gas with a calorific value of 765 kcal/m3 and 1123 kcal/m3 respectively. The authors reported that gasification was conducted successfully in fluidized bed and gas cleaning using granular bed filter. Moltó et al. [13] investigated gasification and pyrolysis of physico-chemical (PC) and a biological (BIO) sludge. Low heating rate and parallel flow using both sludges were found to give the highest pyrolysis ratio; maximum CO emission, i.e., the worst combustion conditions, with minimum contact time and high heating rates. Ouadi et al. [14] investigated gasification of pre-conditioned rejects and de-inking sludge mixtures with wood chips in a fixed bed gasifier. The researchers obtained a gas with a composition of 16.24% H2 , 23.34% CO, 5.21% CH4 and having a higher heating value (1744 kcal/m3 ). They reported that paper mill rejects can be gasified in a fixed bed downdraft gasifier. Nipattummakul et al. [15] reported that, following fuel reforming with steam gasification, wastewater sludge can be a good source of fuel. They used a semi-batch gasifier to determine evolutionary behavior of syngas properties by employing calibrated experiments and diagnostic facilities. Nipattummakul et al. [16] used steam as the gasifying agent and reported that when compared to air gasification, this increased hydrogen yield by three times. The researchers compared sewage sludge gasification results with other samples including paper, food wastes, and plastics. Hightemperature steam was used as the gasifying agent to gasify sewage sludge samples. The temperature of the reactor was adjusted to 700, 800, 900, and 1000 ◦ C. It was reported that low reaction rates at 700 ◦ C showed the necessity of using higher temperatures (>800 ◦ C) to carry our sewage sludge gasification. On the other hand, low reaction rates lead to high steam consumption and prolonged gasification times. An artificial neural network (ANN) is a mathematical structure, which is designed to mimic the biological neural system in terms of the information processing functions of neurons. These systems show high parallelism; they use various interconnected units to process information. These units react to inputs via convertible weights, thresholds and mathematical functions [17]. Artificial neural networks especially address to the problems in large data sets which contain complicated, non-linear relationships in a wide variety of inputs. They are used to solve various types of problems in fields of engineering and science, especially where traditional modeling methods fails to be productive. These models can identify complex patterns in datasets, which cannot be described by simple mathematical formulae. When well-trained, an ANN can serve as a predictive model for a specific application including emission analysis [18], nanofiltration applications [19], water quality [20–23], water distribution strategies [24], and risk analysis [25]. This study aimed to analyze the possibility of converting wastewater treatment sludge into a source of energy in the form of syngas, employing ANN modeling to detect the variations in syngas calorific value regarding operational conditions. 2. Experimental 2.1. ANN modeling The most widely used neural network method is the backpropagation training algorithm, which was used in the present study [20,26–29]. Fig. 1 shows this type of model. The aim of an ANN is to construct a data-generating process model to enable the network to generalize and predict outputs from newly introduced
Fig. 1. A representation of a simple 3-layer back-propagation artificial neural network.
inputs. This learning algorithm involves a multi-layered neural network and contains an input layer, hidden layers and an output layer. The neurons in the hidden and output layer multiply each input by a corresponding weight; sum the product and the process the sum by employing a non-linear transfer function [25]. In the present model, the neural network is trained and tested using MATLAB 7.0. A three-layer neural network that consists of an input layer, output layer, and one hidden layer is used. Data (H2 , O2 , CH4 , CO, CO2 , N2 , and calorific value – Cv) obtained during the study period were designed to meet the requirements of training and testing the neural network. This database is divided into training and testing sets, taking the odd-numbered patterns as training data and even-numbered ones as test data. In the data set arrangement, input variables consist on the concentrations of seven parameters (H2 , O2 , CH4 , CO, CO2 , N2 and Cv) during time “t”, and the output variable (Cv) is the concentrations at time “t + 1”. Arranged input and output data sets were used in training and test phases. The training coefficient was chosen as (lr) 0.1. The number of neurons within the hidden layer is an important factor whereas the number cascade connected to the hidden layer is less influential. In the present study, the input, hidden and output layers of the ANN contained 7, 12, and 1 neurons, respectively. 2.2. Sample characterization Sludge derived from treated wastewater from the tannery industry was used as gasifier fuel. The sludge was collected after dewatering. Dewatered samples were dried at 105 ◦ C for 24 h. Ultimate and proximate analyses of sludge samples are shown in Table 1. Each experiment was performed according to Standard Methods [30]. Elemental analyses were carried out using Thermo Finnigan Flash EA 1112 instrument. 2.3. Thermal conversion and calorific value calculation Gasification experiments were performed in an updraft fixedbed steel reactor with a volume of 2 L. The reactor was heated Table 1 Sludge characterization. Parameters, wt%
Tannery sludge
Ash Volatile matter Moisture C H N
58.8 41.1 4.5 20.6 4.5 2.3
A. Ongen et al. / Journal of Hazardous Materials 263 (2013) 361–366 Table 2 Higher (HHV) of some common fuels [31].
CO CH4 H2
363
Table 5 Syngas composition – E3 – 0.2 L/min dried air – 20 g sample.
Density, kg/m3
MJ/kg
MJ/m3
kcal/m3
Time, min
T, ◦ C
H2 , %
O2 , %
CH4 , %
CO, %
CO2 , %
N2 , %
1.25 0.714 0.089
10.1 55.5 141.8
13.6 39.8 12.8
3019 9470 3025
5 10 15 20 25 30 35 40 45 50 55 60 65 90 120
200 300 400 450 500 550 600 650 700 750 800 850 900 950 1000
0.00 0.00 0.00 0.05 0.08 0.27 0.81 2.19 5.11 10.52 11.81 11.06 12.03 9.20 9.20
15.14 17.87 18.49 17.87 17.15 8.53 3.75 1.29 0.38 0.10 0.00 0.08 0.17 1.20 3.20
0.00 0.00 0.00 0.00 0.02 0.14 0.42 1.48 6.55 5.33 4.71 2.47 3.24 4.70 3.70
0.00 0.00 0.00 0.00 0.22 0.97 1.54 5.00 8.80 10.30 14.24 15.24 19.24 9.30 8.56
0.04 0.04 0.24 0.84 1.85 8.56 10.88 15.00 19.00 17.00 18.00 14.50 16.00 18.36 20.36
67.59 63.94 68.52 70.88 69.95 70.75 67.82 59.15 52.75 43.98 38.09 39.26 42.43 48.00 51.23
Table 3 Syngas composition – E1 – 0.05 L/min dried air – 20 g sample. Time, min
T, ◦ C
H2 , %
O2 , %
CH4 , %
CO, %
CO2 , %
N2 , %
5 10 15 20 25 30 35 40 45 50 55 60 65 90 120
200 300 400 450 500 550 600 650 700 750 800 850 900 950 1000
0.00 0.00 0.00 0.00 0.12 0.09 0.12 0.09 0.98 5.38 12.66 11.13 5.68 4.23 5.12
15.00 16.50 16.33 17.50 19.15 17.73 11.44 16.44 6.77 3.01 0.04 0.10 0.00 1.23 2.33
0.00 0.00 0.00 0.00 0.02 0.04 0.08 0.14 0.51 4.46 6.21 5.47 3.39 1.10 1.00
0.00 0.00 0.00 0.00 0.05 0.12 0.16 0.21 1.45 3.53 12.41 16.02 15.44 14.50 13.23
0.00 0.00 0.00 0.00 2.16 7.65 8.15 4.00 3.97 18.65 18.41 16.50 18.90 13.00 16.56
68.50 65.00 63.50 65.50 68.70 68.09 70.65 69.43 65.34 55.00 48.24 45.37 45.16 43.00 54.36
indirectly by a ceramic resistance until the operational temperature of 750 ◦ C was achieved. Approximate calorific values of producer gases were calculated according to H2 , CO and CH4 volumetric percentages determined by a micro-GC (Varian CP-4900). Dried air was used as a gasification agent. Heating values were calculated according to higher heating values (HHV), as shown in Table 2.
Table 6 Syngas composition – E4 – 0.3 L/min dried air – 20 g sample. Time, min
T, ◦ C
H2 , %
O2 , %
CH4 , %
CO, %
CO2 , %
N2 , %
5 10 15 20 25 30 35 40 45 50 55 60 65 90 120
200 300 400 450 500 550 600 650 700 750 800 850 900 950 1000
0.00 0.00 0.00 0.07 0.31 0.59 1.58 4.37 6.16 9.27 8.67 6.55 6.56 6.21 4.75
17.00 16.73 14.68 18.29 17.06 12.56 10.95 10.34 7.10 3.03 3.00 2.35 3.00 2.36 2.20
0.00 0.00 0.00 0.00 0.17 0.33 0.92 2.40 3.29 5.16 4.60 6.50 5.55 5.60 6.10
0.00 0.00 0.00 0.12 1.10 1.49 2.02 2.73 2.50 5.50 6.00 7.50 8.89 7.30 9.20
0.07 0.06 3.39 1.32 1.84 6.38 9.55 17.19 17.98 16.80 16.82 16.07 18.36 21.36 24.50
68.23 67.35 69.96 69.81 70.35 70.17 65.62 59.13 60.09 55.01 55.10 57.04 55.60 55.60 50.30
3. Results and discussion Syngas compositions from each experiment are shown in Tables 3–8. CO, CH4 , and H2 values were used to calculate calorific values for each sampling time. Under the condition of 0.05 L/min dried-air flow, gasification produced 12% H2 , 12% CO and 6% CH4 . The effective temperature for the process exceeded 750 ◦ C. CO2 levels did not exceed 19%. The experiments were carried out for 2 h. Under 0.1 L/min dried air flow, gasification achieved 12% H2 , 17% CO and 4% CH4 . The effective temperature for the process exceeded 750 ◦ C. CO2 levels did not exceed 15%. Syngas composition did not show a significant change when compared to Table 3. As seen in Table 5, the highest CO level was 19%, at more than 800 ◦ C, where CO2 levels also increased. Experiments were carried out with 0.2 L/min dried-air flow, which achieved 12% H2 , 17% CO, Table 4 Syngas composition – E2 – 0.1 L/min dried air – 20 g sample.
Table 7 Syngas composition – E5 – 0.4 L/min dried air – 20 g sample. Time, min
T, ◦ C
H2 , %
O2 , %
CH4 , %
CO, %
CO2 , %
N2 , %
5 10 15 20 25 30 35 40 45 50 55 60 65 90 120
200 300 400 450 500 550 600 650 700 750 800 850 900 950 1000
0.00 0.00 0.00 0.00 0.09 0.09 0.35 0.84 1.75 1.97 5.97 3.54 4.25 4.12 3.58
18.00 16.50 17.50 20.32 18.32 13.32 9.86 5.16 1.76 5.60 6.20 5.36 6.20 5.00 5.68
0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.41 1.09 1.39 3.13 2.30 1.94 2.00 1.50
0.00 0.00 0.00 0.00 0.00 0.00 0.98 1.56 2.20 2.34 3.02 8.68 7.86 10.20 6.50
0.00 0.00 0.05 0.56 0.97 0.97 5.75 6.04 14.87 15.96 15.88 16.40 13.87 18.25 20.10
58.00 60.00 63.00 69.20 70.29 69.29 68.95 70.22 68.95 70.90 62.66 59.64 64.45 56.98 55.23
Table 8 Syngas composition – E6 – 0.5 L/min dried air – 20 g sample.
Time, min
T, ◦ C
H2 , %
O2 , %
CH4 , %
CO, %
CO2 , %
N2 , %
Time, min
T, ◦ C
H2 , %
O2 , %
CH4 , %
CO, %
CO2 , %
N2 , %
5 10 15 20 25 30 35 40 45 50 55 60 65 90 120
200 300 400 450 500 550 600 650 700 750 800 850 900 950 1000
0.00 0.00 0.04 0.08 0.18 0.28 0.50 1.59 4.69 11.78 12.50 12.16 9.00 4.20 3.60
17.50 17.60 18.58 17.37 15.05 10.05 6.16 2.98 0.89 0.21 0.06 0.23 1.00 1.20 3.20
0.00 0.00 0.00 0.00 0.04 0.14 0.22 0.63 1.97 3.97 3.27 2.50 3.97 2.90 3.97
0.00 0.00 0.00 0.08 0.62 1.02 1.29 1.76 2.49 16.50 17.00 16.72 13.82 15.24 14.13
0.00 0.01 0.88 0.71 3.69 2.69 1.00 2.01 18.00 17.00 12.00 13.31 13.55 15.36 14.36
70.50 68.00 68.45 69.01 70.67 69.67 69.49 64.90 55.94 38.11 36.00 38.00 47.91 48.00 49.23
5 10 15 20 25 30 35 40 45 50 55 60 65 90 120
200 300 400 450 500 550 600 650 700 750 800 850 900 950 1000
0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 3.38 2.20 3.08 1.50 2.30 2.00 2.35
18.11 18.00 17.05 18.51 17.00 15.50 13.20 11.00 8.00 4.00 1.00 1.10 1.00 2.30 2.45
0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1.38 1.50 2.00 3.50 3.00 2.50 3.00
0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1.55 1.19 4.16 5.30 5.69 4.50 3.20
0.00 0.00 0.00 0.06 0.06 0.11 0.61 1.05 22.00 23.20 23.00 20.00 26.50 29.30 28.50
69.61 66.14 68.61 66.91 70.00 67.67 68.57 68.97 60.00 58.00 56.00 55.30 54.00 52.36 55.20
A. Ongen et al. / Journal of Hazardous Materials 263 (2013) 361–366
Heating value, kcal/m3
E1
E2
E3
E4
E5
Table 9 Correlation coefficient for ANN outputs.
E6
1500
900
1250
750
1000
600
750
450
500
300
250
150
Temperature, oC
364
0
0 0
15
30
45
60
75
90
105
120
Time, min. Fig. 2. Syngas heating values of each experiment.
and 4% CH4 . The effective temperature for the process exceeded 700 ◦ C. Syngas production during gasification was negatively affected if the amount of dried air exceeded the optimum level (Tables 6–8). Each gas component was analyzed at the lowest level where CO2 reached the highest percentage by 30%. Fig. 2 shows heating values calculated for syngases produced from each experiment. In the experiments, the gasification process used dried air as a gasification agent to convert carbonaceous materials into gaseous products. Since heating value is a function of gas composition, the calculations were based on the gas composition results. In natural gas, the heating value is more dependent on gas composition. If the composition is known, the heating value can be calculated using: hu =
xi × hui
(1)
where xi is the combustible proportion of gas and hui is its heating value.
a
Experiment number
r
E1 E2 E3 E4 E5 E6
0.68 0.85 0.92 0.82 0.97 0.98
As a rule, the proportions follow the sequence CH4 , H2 and CO. In the present study, CO and CH4 contents increased with temperature, whereas CO2 content decreased. The highest heating values (1000–1500 kcal/m3 ) were achieved for E1, E2 and E3, with flow rates of 0.05, 0.1, and 0.2 L/min, respectively, at 700 ◦ C. The operation time was between 45 and 65 min. The calculated values were used for ANN modeling in the second phase of the study. ANN modeling used values for H2 , O2 , CH4 , CO, CO2 , N2 , and Cv, and the output data consisted of the change in Cv concentration at time t + 1. The outputs of the ANN-modeled structure are presented in Figs. 3–5. Correlations between observed data and prediction were calculated, and the results are shown in Table 9. For the ANN structure, the r-values calculated for the testing phase were between 0.68 and 0.98. As these rates are close to 1, it means that it is a good modeling. Correlation coefficients are one of the most important indices used for artificial neural network (ANN) model performance. Many studies in the literature investigated ANN forecast modeling techniques. Pashova and Popova [32] used ANN to forecast daily sea levels at Burgas, Bulgaria. ANN performance was controlled by correlation coefficient and their values found 0.72–0.95. Palani et al. [33], presented an ANN application to predict atmospheric nitrogen deposition. Total and organic nitrogen were predicted by this model
b 1200 observed predicted
900
Calorific Value
Calorific Value
1200
600 300
2
3
4
5
6
7
predicted
600 300
0 1
observed
900
0
8
1
2
3
4
5
6
7
8
7
8
Fig. 3. (a) ANN outputs for predicted calorific value (E1); (b) ANN outputs for predicted calorific value (E2).
a
b 1200
1200 Calorific Value
Calorific Value
observed
predicted
900 600 300
observed predicted
900 600 300
0
0 1
2
3
4
5
6
7
8
1
2
3
4
5
6
Fig. 4. (a) ANN outputs for predicted calorific value (E3); (b) ANN outputs for predicted calorific value (E4).
A. Ongen et al. / Journal of Hazardous Materials 263 (2013) 361–366
a
b 1200
1200
observed predicted
Calorific Value
Calorific Value
365
900 600 300
observed predicted
900 600 300
0
0 1
2
3
4
5
6
7
1
2
3
4
5
6
7
8
Fig. 5. (a) ANN outputs for predicted calorific value (E5); (b) ANN outputs for predicted calorific value (E6).
and correlation coefficients ranged from 0.93 and 0.99 respectively. Feng et al. [34] forecast ozone concentration, and reported correlation coefficients between 0.40 and 0.87. The correlation coefficient of our model (Table 9) shows r values similar to those of previous studies. 4. Conclusion The present study examined the production of syngas by gasification of sludge derived from treatment of tannery wastewater. Dried air was used as a gasification agent, at flow rates between 0.05 and 0.5 L/min. Since heating value is dependent on gas composition, the CO, CH4 , and H2 contents of syngases were analyzed. It was found that combustible gas composition increased with increasing temperature in excess of 600 ◦ C. Calorific values of syngases were calculated for each experiment, and the results were modeled with an ANN. Air-flow rate between 0.05 and 0.1 L/min was found to be optimum for the sludge samples used in the study, and produced the highest heating values of 1000–1500 kcal/m3 . For the ANN structure, the r values of the testing phase were between 0.68 and 0.98. Modeled outputs showed that temporal changes of modeled predictions were in close accordance with observed values. Correlation coefficient values showed that the ANN used in this study gave results in high accuracy. Finally, the proposed ANN model can be used to model gasification products without further adaptation. It can also be implemented in different time series or different model structures. References [1] W-J. Lee, S-D. Kim, B.-H. Song, Steam gasification of coal with salt mixture of potassium and nickel in a fluidized bed reactor, Korean J. Chem. Eng. 18 (5) (2011) 640–645. [2] F. Pinto, R. André, C. Franco, H. Lopes, C. Carolino, R. Costa, I. Gulyurtlu, Cogasification of coal and wastes in a pilot-scale installation. 2: effect of catalysts in syngas treatment to achieve sulphur and nitrogen compounds abatement, Fuel. 89 (2010) 3340–3351. [3] A. Fuente-Cuesta, M.A. Lopez-Anton, M. Diaz-Somoano, A.VanZomeren, M. Cieplik, M.R. Martínez-Tarazona, Leaching of major and trace elements from paper–plastic gasification chars: an experimental and modelling study, J. Hazard. Mater. 244–245 (2013) 70–80. [4] X. Gang, J. Bao-sheng, Z. Zhao-ping, C. Yong, N. Ming-jiang, C. Ke-fa, X. Rui, H. Ya-ji, H. He, Experimental study on MSW gasification and melting technology, J. Environ. Sci. 19 (2007) 1398–1403. [5] T. Malkow, Novel and innovative pyrolysis and gasification technologies for energy efficient and environmentally sound MSW disposal, Waste Manage. 24 (2004) 53–79. [6] L. Peng, Y. Wang, Z. Lei, G. Cheng, Co-gasification of wet sewage sludge and forestry waste in situ steam agent, Bioresour. Technol. 114 (2012) 698–702. [7] P. Mondal, G.S. Dang, M.O. Garg, Syngas production through gasification and cleanup for downstream applications—recent developments, Fuel Process. Technol. 92 (2011) 1395–1410.
[8] V. Kirubakaran, V. Sivaramakrishnan, R. Nalini, T. Sekar, M. Premalatha, P. Subramanian, A review on gasification of biomass, Renew. Sust. Energy Rev. 13 (2009) 179–186. [9] Y. Zhang, B. Li, H. Li, B. Zhang, Exergy analysis of biomass utilization via steam gasification and partial oxidation, Thermochim. Acta 538 (2012) 21–28. [10] U. Zuberbühler, Gasification of biomass—an overview on available technologies, in: 1st European Summer School on Renewable Motor Fuels, Birkenfeld, Germany, 29–31 August, 2005. [11] J.A. Ruiz, M.C. Juarez, M.P. Morales, P. Munoz, M.A. Mendivil, Biomass gasification for electricity generation: review of current technology barriers, Renew. Sust. Energy Rev. 18 (2013) 174–183. [12] J.W. Judex, M. Gaiffi, H.C. Burgbacher, Gasification of dried sewage sludge: status of the demonstration and the pilot plant, Waste Manage. 32 (2012) 719–723. [13] J. Moltó, et al., Gas production during the pyrolysis and gasification of biological and physico-chemical sludges from oil refinery, J. Anal. Appl. Pyrol. (2012), http://dx.doi.org/10.1016/j.jaap.2012.09.012. [14] M. Ouadi, J.G. Brammer, M. Kay, A. Hornung, Fixed bed downdraft gasification of paper industry wastes, Appl. Energy 103 (2013) 692–699. [15] N. Nipattummakul, I. Ahmed, S. Kerdsuwan, A.K. Gupta, High temperature steam gasification of wastewater sludge, Appl. Energy 87 (2010) 3729–3734. [16] N. Nipattummakul, I. Ahmed, S. Kerdsuwan, A.K. Gupta, Hydrogen and syngas production from sewage sludge via steam gasification, Int. J. Hydrogen Energy 35 (2010) 11738–11745. [17] S.A. Rounds, Development of a neural network model for dissolved oxygen in the Tualatin River, Oregon, in: Proceedings of the Second Federal Interagency Hydrologic Modeling Conference, Las Vegas, Nevada, July 29–August 1, 2002, Subcommittee on Hydrology of the Interagency Advisory Committee on Water Information, 2002. [18] B. Ghobadian, H. Rahimi, A.M. Nikbakht, G. Najafi, T.F. Yusaf, Diesel engine performance and exhaust emission analysis using waste cooking biodiesel fuel with an artificial neural network, Renew. Energy 34 (2009) 976–982. [19] H. Al-Zoubi, N. Hilal, N.A. Darwish, A.W. Mohammad, Rejection and modelling of sulphate and potassium salts by nanofiltration membranes: neural network and Spiegler–Kedem model, Desalination 206 (2007) 42–60. [20] K.P. Singh, A. Basant, A. Malik, G. Jain, Artificial neural network modeling of the river water quality—a case study, Ecol. Model. 220 (2009) 888–895. [21] N. Balkaya, H.K. Ozcan, O.N. Ucan, Determination of relationship between hardness and groundwater quality parameters by neural networks, Desalination Water Treat. 11 (2009) 258–263. [22] I.S. Yeon, J.H. Kim, K.W. Jun, Application of artificial intelligence models in water quality forecasting, Environ. Technol 29 (6) (2008) 625–631. [23] I.S. Yeon, K.W. Jun, H.J. Lee, The improvement of total organic carbon forecasting using neural networks discharge model, Environ. Technol 30 (1) (2009) 45–51. [24] R. Jafar, I. Shahrour, I. Juran, Application of artificial neural networks (ANN) to model the failure of urban water mains, Math. Comput. Model. 51 (2010) 1170–1180. [25] B. Pradhan, S. Lee, Landslide risk analysis using artificial neural network model focusing on different training sites, Int. J. Phys. Sci. 4 (1) (2009) 1–15. [26] H.K. Ozcan, O.N. Ucan, U. Sahin, M. Borat, C. Bayat, Artificial neural network modeling of methane emissions at Istanbul Kemerburgaz-Odayeri Landfill Site, J. Sci. Ind. Res. 65 (2) (2006) 128–134. [27] E. Cadenas, W. Rivera, Wind speed forecasting in three different regions of Mexico, using a hybrid ARIMA-ANN model, Renew. Energy 35 (2010) 2732–2738. [28] R. Noori, A. Karbassi, M.S. Sabahi, Evaluation of PCA and Gamma test techniques on ANN operation for weekly solid waste prediction, J. Environ. Manage. 91 (2010) 767–771. [29] A. C¸elekli, F. Geyik, Artificial neural networks (ANN) approach for modeling of removal of Lanaset Red G on Characontraria, Bioresour. Technol. 102 (2011) 5634–5638.
366
A. Ongen et al. / Journal of Hazardous Materials 263 (2013) 361–366
[30] Standard Methods, Standard Methods for the Examination of Water and Wastewater, 21st edition, AWWA/APHA/WEF, 2005. [31] U. Bossel, Well-to-Wheel Studies, Heating Values, in: and the Energy Conservation Principle, European Fuel Cell Forum, Oberrohrdorf, Switzerland, 2003. [32] L. Pashova, S. Popova, Daily sea level forecast at tide gauge Burgas, Bulgaria using artificial neural networks, J. Sea Res. 66 (2011) 154–161.
[33] S. Palani, P. Tkalich, R. Balasubramanian, J. Palanichamy, ANN application for prediction of atmospheric nitrogen deposition to aquatic ecosystems, Mar. Pollut. Bull. 62 (2011) 1198–1206. [34] Y. Feng, W. Zhang, D. Sun, L. Zhang, Ozone concentration forecast method based on genetic algorithm optimized back propagation neural networks and support vector machine data classification, Atmos. Environ. 45 (2001) 1979–1985.