1 1. CLEAN-SOIL AIR WATER Bimonthly
1 1. CLEAN-SOIL AIR WATER Bimonthly
1 1. CLEAN-SOIL AIR WATER Bimonthly
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
SCIENCE CITATION INDEX EXPANDED JOURNAL LIST<br />
Search terms: 1863-0650<br />
Total journals found: 1<br />
<strong>1.</strong> <strong>CLEAN</strong>-<strong>SOIL</strong> <strong>AIR</strong> <strong>WATER</strong><br />
<strong>Bimonthly</strong><br />
ISSN: 1863-0650<br />
WILEY-V C H VERLAG GMBH, PO BOX 10 11 61, WEINHEIM, GERMANY, D-69451<br />
<strong>1.</strong> Science Citation Index Expanded<br />
2. Current Contents - Agriculture, Biology & Environmental Sciences<br />
3. Zoological Record<br />
4. BIOSIS Previews<br />
Abstracting and Indexing Information<br />
Aqualine Abstracts (CSA/CIG)<br />
ASFA: Aquatic Sciences & Fisheries Abstracts<br />
(CSA/CIG)<br />
BIOBASE (Elsevier)<br />
Biological Abstracts® (Thomson ISI)<br />
BIOSIS Previews® (Thomson ISI)<br />
CAB Abstracts® (CABI)<br />
Cambridge Scientific Abstracts (CSA/CIG)<br />
Chemical Abstracts Service/SciFinder (ACS)<br />
CSA Environmental Sciences & Pollution<br />
Management Database (CSA/CIG)<br />
Current Awareness in Biological Sciences<br />
(Elsevier)<br />
Current Contents®/Agriculture, Biology &<br />
Environmental Sciences (Thomson ISI)<br />
Environment Abstracts (LexisNexis)<br />
FISHLIT (NISC)<br />
FIZ Karlsruhe Databases (FIZ Karlsruhe)<br />
FLUIDEX/Fluid Abstracts (Elsevier)<br />
GEOBASE/Geographical & Geological Abstracts<br />
(Elsevier)<br />
Journal Citation Reports/Science Edition<br />
(Thomson ISI)<br />
PASCAL Database (INIST/CNRS)<br />
Science Citation Index Expanded (Thomson<br />
ISI)<br />
SCOPUS (Elsevier)<br />
<strong>WATER</strong>LIT (NISC)<br />
Web of Science® (Thomson ISI)<br />
Zoological Record (Thomson ISI)
186 Clean 2007, 35 (2), 186 – 192<br />
Ahmet Baylar 1<br />
Davut Hanbay 2<br />
Emrah Ozpolat 1<br />
1 Firat University, Civil Engineering<br />
Department, Elazig, Turkey.<br />
2 Firat University, Electronic and<br />
Computer Science Department, Elazig,<br />
Turkey.<br />
1 Introduction<br />
Research Article<br />
Stepped cascade flows are characterized by the strong turbulent<br />
mixing, the large residence time, and the substantial air bubble<br />
entrainment. Stepped flows can be classified into skimming flow,<br />
transition flow, and nappe flow. For narrow steps or larger discharges<br />
such as the design discharge the water skims over the step<br />
corners and recirculating zones develop in triangular niches<br />
formed by the step faces and the pseudo-bottom, as shown in Fig. 1a.<br />
In skimming flow the water flows as a coherent stream over the<br />
pseudo-bottom formed by the step corners. For a range of intermediate<br />
discharges, a transition flow regime takes place. The dominant<br />
feature is stagnation on the horizontal step face associated with significant<br />
splashing and a chaotic appearance (see Fig. 1b). For nappe<br />
flow the steps act as a series of overfalls with the water plunging<br />
from one step to another (see Fig. 1c). Generally speaking, the nappe<br />
flow is found for low discharges and wide steps [1].<br />
Self-aeration on stepped cascades is now recognized for its substantial<br />
contribution to the air-water transfer of atmospheric gases<br />
such as oxygen and nitrogen. Stepped cascades are very efficient<br />
means of aeration because of the strong turbulent mixing, the large<br />
residence time, and the substantial air bubble entrainment.<br />
Stepped cascades are used in water treatment for re-oxygenation,<br />
denitrification, or VOC removals. In the treatment of drinking<br />
water, cascade aeration may be used to remove chlorine and to eliminate<br />
or reduce offensive taste and odor [2].<br />
Chanson and Toombes [3] conducted gas-liquid interface measurements<br />
in a stepped cascade. Local void fractions, bubble count<br />
rates, bubble size distributions, and gas-liquid interface areas were<br />
measured simultaneously in the air-water flow region using resistivity<br />
probes. However, they stated that future work is needed to compare<br />
aeration efficiencies estimated with detailed interfacial area<br />
data and based upon dissolved gas measurements. Recently, Baylar<br />
Correspondence: Prof. A. Baylar (abaylar@firat.edu.tr), Firat University,<br />
Civil Engineering Department, Elazig, Turkey.<br />
Modeling Aeration Efficiency of Stepped Cascades<br />
by Using ANFIS<br />
The physical process of oxygen transfer or oxygen absorption from the atmosphere<br />
acts to replenish the used oxygen, a process termed re-aeration or aeration. Aeration<br />
enhancement by macro-roughness is well-known in water treatment and one form is<br />
the aeration cascade. The macro-roughness of the steps significantly reduces the flow<br />
velocities and leads to flow aeration along the stepped cascade. In this paper, the aeration<br />
efficiency in stepped cascade aerators was modeled by using the Adaptive Network<br />
Based Fuzzy Inference System (ANFIS). The obtained model was tested with experimental<br />
data. Test results showed that ANFIS can be used to estimate the aeration<br />
efficiency in stepped cascade aerators.<br />
Keywords: ANFIS; Aeration efficiency; Modeling; Oxygen transfer;<br />
Received: February 14, 2007; revised: March 9, 2007; accepted: March 12, 2007<br />
DOI: 10.1002/clen.200700019<br />
Figure <strong>1.</strong> Flow regimes above stepped cascades: a) Skimming flow; b)<br />
Transition flow; c) Nappe flow [22].<br />
and Emiroglu [4], Emiroglu and Baylar [5], Baylar et al. [6], Emiroglu<br />
and Baylar [7], and Baylar et al. [8 – 10] did some detailed experiments<br />
on the aeration efficiency of stepped cascades. It was found<br />
that water can trap a lot of air when passing through the steps and<br />
that the oxygen content in the water body is increased. So stepped<br />
cascades can be used as highly effective aerators in streams, rivers,<br />
constructed channels, fish hatcheries, water treatment plants, etc.<br />
Intelligent techniques are used in various areas of water-related<br />
research [11 – 14]. Recently, Hanbay et al. [15,16] modeled the aeration<br />
efficiency of weirs by using the Adaptive Network Based Fuzzy<br />
Inference System (ANFIS) and by using wavelet and neural network.<br />
In this study, the aeration efficiency in stepped cascade aerators is<br />
modeled by using ANFIS. After reviewing some basic properties of<br />
i 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.clean-journal.com
Clean 2007, 35 (2), 186 – 192 Modeling of Aeration Efficiency 187<br />
the oxygen transfer posses and the ANFIS, the effectiveness of the<br />
proposed intelligent system is demonstrated.<br />
2 Preliminaries<br />
The oxygen transfer process and the theoretical foundations for the<br />
intelligent system used in the presented study are given in the following<br />
subsections.<br />
2.1 Oxygen Transfer Process<br />
The rate of oxygen mass transfer, i. e., from the gas (air bubbles) to<br />
the liquid phase (water) is governed by the term described below:<br />
dC<br />
dt<br />
¼ KL<br />
A<br />
V ðCs CÞ ð1Þ<br />
Where C is the dissolved oxygen (DO) concentration, KL is the<br />
liquid film coefficient for oxygen, A is the surface area associated<br />
with the volume, V, over which the transfer occurs, Cs is the saturation<br />
concentration, and t is the time. The term A/V is often called the<br />
specific surface area, a, or surface area per unit volume.<br />
Eq. (1) does not consider sources and sinks of oxygen in the water<br />
body because their rates are relatively slow compared to the oxygen<br />
transfer that occurs at most hydraulic structures due to the increase<br />
in free-surface turbulence and the large quantity of air that is normally<br />
entrained into the flow.<br />
The predictive relations assume that Cs is constant and determined<br />
by the water-atmosphere partitioning. If that assumption is<br />
made, Cs is constant with respect to time, and the oxygen transfer<br />
efficiency (aeration efficiency), E, may be defined as [17]:<br />
E ¼ Cd Cu<br />
Cs Cu<br />
¼ 1<br />
1<br />
r<br />
where u and d are subscripts indicating upstream and downstream<br />
locations, respectively, and r is the oxygen deficit ratio:<br />
[(Cs – Cu)/(Cs –Cd)]<br />
A transfer efficiency value of <strong>1.</strong>0 means that the full transfer up to<br />
the saturation value has occurred at the structure surface. No transfer<br />
would correspond to E = 0.0. The saturation concentration in distilled,<br />
deionized water may be obtained from charts or equations<br />
what is an approximation as the saturation DO concentration for<br />
natural waters is often different from that of distilled, deionized<br />
water due to the salinity affects.<br />
Comparative evaluations of oxygen uptake at hydraulic structures<br />
require that the aeration efficiency is corrected to a reference<br />
temperature. To provide a uniform basis for comparison of measurement<br />
results, the aeration efficiency is often normalized to a 208C<br />
standard. Gulliver et al. [17] proposed the following equation to<br />
describe the influence of the temperature:<br />
1–E20 =(1–E) 1/f (3)<br />
Where E is the transfer efficiency at the actual water temperature,<br />
E20 is the transfer efficiency for 208C, and f is an exponent described<br />
by:<br />
f = <strong>1.</strong>0 + 2.1610 –2 (T –20) + 8.26610 –5 (T–20) 2 (4)<br />
ð2Þ<br />
In this study, the aeration efficiency was normalized to 208C<br />
using Eq. (3).<br />
2.2 Adaptive Network Based Fuzzy Inference System<br />
(ANFIS)<br />
The architecture and learning rule of ANFIS have been described in<br />
detail in [18]. ANFIS is a Multilayer feed forward network where<br />
each node performs a particular function on incoming signals. Both<br />
square and circle node symbols are used to represents different<br />
properties of adaptive learning. To perform desired input-output<br />
characteristics, adaptive learning parameters are updated based on<br />
gradient learning rules [18 – 20]. The ANFIS architecture has six<br />
inputs and one output. This architecture is formed by using five<br />
layer and 64 if-then rules, a sample rule base is described in Eq. (5):<br />
If (x is A1, y is B1, …, zz is F1)<br />
then (f1 = p1x +q1y+… + pp1zz + ri) (5)<br />
Where, x, y, …, zz, are inputs and p, q, r, …, pp, u are linear output<br />
parameters. The node functions in the same layer are similar and<br />
described as below:<br />
Layer-1: Every node i in this layer is a square node with a node<br />
function.<br />
O 1 i ¼ lAiðxÞ ð6Þ<br />
Where, x1, x2, …, x5, x6 are inputs to node i, and Ai, Bi, Ci, …, Hi are<br />
linguistic labels associated with this node function. O1 i is the membership<br />
function of Ai. Usually lAi(x) is chosen as a bell-shaped function<br />
with a maximum value equal to 1 and a mininmum value equal<br />
to 0.<br />
Layer-2: Every node in this layer is a circle node labeled P which<br />
multiplies the incoming signals and sends the product out. For<br />
instance:<br />
wi = lAi(x)*lBi(y)*lCi(z)*lDi(xx)*lEi(yy)*lEi(yy)<br />
i =1,2,3,…, 64 (7)<br />
Each node output represents the firing strength of a rule.<br />
Layer-3: Every node in this layer is a circle node labeled N. The i th<br />
node calculates the ratio of the i th rules firing strength to the sum of<br />
all rule's firing strengths.<br />
wi ¼<br />
wi<br />
w1 þ w2:::w64<br />
i ¼ 1; 2; 3; :::; 64 ð8Þ<br />
Layer-4: Every node i in this layer is a square node with a node<br />
function:<br />
O 4 i ¼ wifi ¼ wiðpix þ qiy þ :::: þ riÞ i ¼ 1; 2; 3; :::; 64 ð9Þ<br />
Where, wi is the output of layer 3 and {pi, qi, ri, …, ppi, ui} is the<br />
parameter set. Parameters in this layer will be referred to as consequent<br />
parameters.<br />
Layer-5: The single node in this layer is a circle node labeled S that<br />
computes the overall output of ANFIS as the summation of all<br />
incoming signals.<br />
X<br />
O 5 i<br />
¼ overall output ¼ X<br />
i 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.clean-journal.com<br />
i<br />
wifi ¼<br />
i<br />
X<br />
i<br />
wifi<br />
wi<br />
ð10Þ
188 A. Baylar et al. Clean 2007, 35 (2), 186 – 192<br />
Figure 2. Experimental arrangement for the stepped cascade model.<br />
Figure 3. ANFIS model structure.<br />
3 Experimental Results and Application<br />
3.1 Experimental Arrangement<br />
The data used in this study were taken from studies conducted by<br />
Baylar et al. [4, 6] on a large model of a stepped cascade. Schematic<br />
representation of the experimental setup used in these studies is<br />
shown in Fig. 2. All experiments were conducted in a prismatic rectangular<br />
channel with 0.30 m width and 0.50 m deepth. The side<br />
walls were made of transparent methacrylate to follow the flow<br />
regime. Water was pumped from the storage tank to the stilling<br />
tank, from which it entered the stepped channel through an<br />
approach channel.<br />
The discharge was measured by means of a flow meter installed<br />
in the supply line. All experimental runs were carried out in unit<br />
discharges ranging between 16.67 and 166.67 L/s m. For the stepped<br />
channel, the downstream channel was 3.0 m long, 0.35 m wide, and<br />
Figure 4. Training performance of the ANFIS model.<br />
Figure 5. Test performance of the ANFIS model.<br />
0.45 m deep. The slopes of the stepped channel were varied: 14.488,<br />
18.748, 22.558, 308, 408, and 508. For all slopes tested, steps with a<br />
height, h, of 5, 10, and 15 cm were used.<br />
i 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.clean-journal.com
Clean 2007, 35 (2), 186 – 192 Modeling of Aeration Efficiency 189<br />
Table <strong>1.</strong> Data of stepped cascade for h = 0.05 m.<br />
q<br />
(m 2 /s610 –3 )<br />
h<br />
(m)<br />
a<br />
(deg.)<br />
L<br />
(m)<br />
Tap water was used throughout the present experiments. The<br />
water was changed for each experiment. The water in the tank was<br />
deoxygenated by the sodium sulfite method. Theoretically, 7.9 g/m 3<br />
of sodium sulfite is required to remove 1 g/m 3 of DO. Based on the<br />
DO of the test tap water, the approximate sodium sulfite requirements<br />
are estimated (a 10 – 20% excess is used). Usually, addition of<br />
cobalt(II) chloride is required at a dosage of 3.3 g/m 3 as a catalyst for<br />
the deoxygenation reaction. In this study, 70 g/m 3 of sodium sulfite<br />
and 3.3 g/m 3 of cobalt(II) chloride were added. The salt content of<br />
the tap water used for all of the experiments reported in this paper<br />
was low and monitored constantly during the experiments to<br />
ensure no significant buildup of residues caused by deoxygenation<br />
of the chemicals added to the water. Therefore, it is unlikely that<br />
N E20<br />
(–)<br />
Flow Regime<br />
16.67 0.05 14.48 5.00 25 0.60 Nappe<br />
33.33 0.05 14.48 5.00 25 0.58 Transition<br />
50.00 0.05 14.48 5.00 25 0.55 Skimming<br />
66.67 0.05 14.48 5.00 25 0.45 Skimming<br />
100.00 0.05 14.48 5.00 25 0.30 Skimming<br />
133.33 0.05 14.48 5.00 25 0.26 Skimming<br />
166.67 0.05 14.48 5.00 25 0.23 Skimming<br />
16.67 0.05 30.00 5.00 50 0.81 Nappe<br />
33.33 0.05 30.00 5.00 50 0.82 Skimming<br />
50.00 0.05 30.00 5.00 50 0.74 Skimming<br />
66.67 0.05 30.00 5.00 50 0.70 Skimming<br />
100.00 0.05 30.00 5.00 50 0.62 Skimming<br />
133.33 0.05 30.00 5.00 50 0.59 Skimming<br />
166.67 0.05 30.00 5.00 50 0.57 Skimming<br />
16.67 0.05 18.74 3.89 25 0.60 Nappe<br />
33.33 0.05 18.74 3.89 25 0.57 Transition<br />
50.00 0.05 18.74 3.89 25 0.52 Skimming<br />
66.67 0.05 18.74 3.89 25 0.44 Skimming<br />
100.00 0.05 18.74 3.89 25 0.28 Skimming<br />
133.33 0.05 18.74 3.89 25 0.22 Skimming<br />
166.67 0.05 18.74 3.89 25 0.16 Skimming<br />
16.67 0.05 40.00 3.89 50 0.74 Transition<br />
33.33 0.05 40.00 3.89 50 0.75 Skimming<br />
50.00 0.05 40.00 3.89 50 0.72 Skimming<br />
66.67 0.05 40.00 3.89 50 0.70 Skimming<br />
100.00 0.05 40.00 3.89 50 0.63 Skimming<br />
133.33 0.05 40.00 3.89 50 0.59 Skimming<br />
166.67 0.05 40.00 3.89 50 0.56 Skimming<br />
16.67 0.05 22.55 3.26 25 0.68 Nappe<br />
33.33 0.05 22.55 3.26 25 0.61 Transition<br />
50.00 0.05 22.55 3.26 25 0.53 Skimming<br />
66.67 0.05 22.55 3.26 25 0.42 Skimming<br />
100.00 0.05 22.55 3.26 25 0.32 Skimming<br />
133.33 0.05 22.55 3.26 25 0.29 Skimming<br />
166.67 0.05 22.55 3.26 25 0.24 Skimming<br />
16.67 0.05 50.00 3.26 50 0.79 Transition<br />
33.33 0.05 50.00 3.26 50 0.77 Skimming<br />
50.00 0.05 50.00 3.26 50 0.75 Skimming<br />
66.67 0.05 50.00 3.26 50 0.74 Skimming<br />
100.00 0.05 50.00 3.26 50 0.72 Skimming<br />
133.33 0.05 50.00 3.26 50 0.66 Skimming<br />
166.67 0.05 50.00 3.26 50 0.64 Skimming<br />
the results were measurably affected by the presence of chemicals<br />
or pollutants.<br />
Each experiment was started by filling the storage tank with<br />
water by adding Na2SO3 and CoCl2 for chemical deoxygenation. During<br />
the experiments, DO measurements upstream and downstream<br />
of the stepped cascade were taken using calibrated portable HANNA<br />
Model HI 9142 oxygen meters at the locations identified in Fig. 2.<br />
Measurements were made by submersing the probe to a depth of<br />
approximately 0.20 m at the sampling points. The DO meters were<br />
calibrated daily according to local atmospheric pressure, prior to<br />
use, by the air calibration method. Calibration procedures followed<br />
those recommended by the manufacturer. The calibration was performed<br />
in humid air under ambient conditions. In this study, the<br />
i 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.clean-journal.com
190 A. Baylar et al. Clean 2007, 35 (2), 186 – 192<br />
Table 2. Data of stepped cascade for h = 0.10 m.<br />
q<br />
(m 2 /s610 –3 )<br />
saturation concentrations were determined by the chart of McGhee<br />
[21].<br />
3.2 Modeling<br />
h<br />
(m)<br />
a<br />
(deg.)<br />
L<br />
(m)<br />
The experimental results for the aeration efficiency in the stepped<br />
cascade aerators are given in Tabs. 1 – 3.<br />
The ANFIS model used in this study is shown in Fig. 3. The parameters<br />
considered in the study are unit discharge (q), step height (h),<br />
channel slope (a), channel length (L), total number of steps (N), flow<br />
regime information and aeration efficiency at 208C(E20). The parameters,<br />
q, h, a, L, N, and the flow regime information were used as<br />
inputs to the ANFIS model for the estimation of E20. In the seventh<br />
column, the numbers 1, 2, and 3 used to represent the flow regimes,<br />
nappe, transition, and skimming, respectively. 100 of 126 experimental<br />
data sets were used to train the ANFIS model and 26 experimental<br />
data sets were used for the validation test. The 26 test data<br />
were randomly selected among the whole data.<br />
Before training and testing, the training input and output values<br />
were normalized using Eq. (11):<br />
xi ¼ xi xmin<br />
xmax xmin<br />
N E20<br />
(–)<br />
Flow Regime<br />
16.67 0.10 14.48 5.00 12 0.55 Nappe<br />
33.33 0.10 14.48 5.00 12 0.54 Nappe<br />
50.00 0.10 14.48 5.00 12 0.54 Nappe<br />
66.67 0.10 14.48 5.00 12 0.52 Transition<br />
100.00 0.10 14.48 5.00 12 0.44 Transition<br />
133.33 0.10 14.48 5.00 12 0.41 Skimming<br />
166.67 0.10 14.48 5.00 12 0.34 Skimming<br />
16.67 0.10 30.00 5.00 25 0.80 Nappe<br />
33.33 0.10 30.00 5.00 25 0.79 Nappe<br />
50.00 0.10 30.00 5.00 25 0.77 Nappe<br />
66.67 0.10 30.00 5.00 25 0.75 Transition<br />
100.00 0.10 30.00 5.00 25 0.72 Skimming<br />
133.33 0.10 30.00 5.00 25 0.67 Skimming<br />
166.67 0.10 30.00 5.00 25 0.60 Skimming<br />
16.67 0.10 18.74 3.89 12 0.58 Nappe<br />
33.33 0.10 18.74 3.89 12 0.58 Nappe<br />
50.00 0.10 18.74 3.89 12 0.55 Nappe<br />
66.67 0.10 18.74 3.89 12 0.55 Transition<br />
100.00 0.10 18.74 3.89 12 0.47 Transition<br />
133.33 0.10 18.74 3.89 12 0.41 Skimming<br />
166.67 0.10 18.74 3.89 12 0.37 Skimming<br />
16.67 0.10 40.00 3.89 25 0.74 Nappe<br />
33.33 0.10 40.00 3.89 25 0.76 Nappe<br />
50.00 0.10 40.00 3.89 25 0.77 Transition<br />
66.67 0.10 40.00 3.89 25 0.76 Transition<br />
100.00 0.10 40.00 3.89 25 0.70 Skimming<br />
133.33 0.10 40.00 3.89 25 0.66 Skimming<br />
166.67 0.10 40.00 3.89 25 0.63 Skimming<br />
16.67 0.10 22.55 3.26 12 0.62 Nappe<br />
33.33 0.10 22.55 3.26 12 0.59 Nappe<br />
50.00 0.10 22.55 3.26 12 0.57 Nappe<br />
66.67 0.10 22.55 3.26 12 0.55 Transition<br />
100.00 0.10 22.55 3.26 12 0.46 Skimming<br />
133.33 0.10 22.55 3.26 12 0.39 Skimming<br />
166.67 0.10 22.55 3.26 12 0.30 Skimming<br />
16.67 0.10 50.00 3.26 25 0.77 Nappe<br />
33.33 0.10 50.00 3.26 25 0.74 Transition<br />
50.00 0.10 50.00 3.26 25 0.74 Transition<br />
66.67 0.10 50.00 3.26 25 0.73 Transition<br />
100.00 0.10 50.00 3.26 25 0.71 Skimming<br />
133.33 0.10 50.00 3.26 25 0.68 Skimming<br />
166.67 0.10 50.00 3.26 25 0.65 Skimming<br />
ð11Þ<br />
Where xmin and xmax denote the minimum and maximum of the<br />
stage and discharge data. Each input has two bell-shaped membership<br />
functions. The desired sum squared error value was selected as<br />
0.000<strong>1.</strong> The ANFIS model performed this value at 300 epochs.<br />
i 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.clean-journal.com
Clean 2007, 35 (2), 186 – 192 Modeling of Aeration Efficiency 191<br />
Table 3. Data of stepped cascade for h = 0.15 m.<br />
q<br />
(m 2 /s610 –3 )<br />
h<br />
(m)<br />
The training performance of the ANFIS model is graphically<br />
showed in Fig. 4. Moreover, the test performance of the ANFIS model<br />
is graphically presented in Fig. 5. As it can be seen from Fig. 5, the<br />
ANFIS model estimates the measured values with a high accuracy.<br />
The regression coefficient (R 2 ) was calculated as R 2 = 0.999 for a training<br />
performance of the ANFIS model and R 2 = 0.985 for a test performance<br />
of the model.<br />
4 Conclusions<br />
a<br />
(deg.)<br />
L<br />
(m)<br />
Hydraulic structures can increase dissolved oxygen levels by creating<br />
turbulent conditions where small air bubbles are carried into<br />
the bulk of the flow. Stepped chute aeration is a particular instance<br />
of this. In this study, an ANFIS model was used in estimation of the<br />
aeration efficiency of stepped cascade aerators. There was a good<br />
agreement between the measured aeration efficiencies and the<br />
values computed from the ANFIS model. This is evidenced by a very<br />
high regression coefficient, R 2 = 0.985. Therefore, ANFIS can be used<br />
to estimate the aeration efficiency in stepped cascade aerators.<br />
References<br />
N E20<br />
(–)<br />
Flow Regime<br />
16.67 0.15 14.48 5.00 8 0.49 Nappe<br />
33.33 0.15 14.48 5.00 8 0.50 Nappe<br />
50.00 0.15 14.48 5.00 8 0.48 Nappe<br />
66.67 0.15 14.48 5.00 8 0.46 Nappe<br />
100.00 0.15 14.48 5.00 8 0.43 Nappe<br />
133.33 0.15 14.48 5.00 8 0.40 Transition<br />
166.67 0.15 14.48 5.00 8 0.40 Transition<br />
16.67 0.15 30.00 5.00 16 0.78 Nappe<br />
33.33 0.15 30.00 5.00 16 0.76 Nappe<br />
50.00 0.15 30.00 5.00 16 0.75 Nappe<br />
66.67 0.15 30.00 5.00 16 0.75 Nappe<br />
100.00 0.15 30.00 5.00 16 0.73 Nappe<br />
133.33 0.15 30.00 5.00 16 0.72 Transition<br />
166.67 0.15 30.00 5.00 16 0.71 Skimming<br />
16.67 0.15 18.74 3.89 8 0.57 Nappe<br />
33.33 0.15 18.74 3.89 8 0.58 Nappe<br />
50.00 0.15 18.74 3.89 8 0.53 Nappe<br />
66.67 0.15 18.74 3.89 8 0.52 Nappe<br />
100.00 0.15 18.74 3.89 8 0.47 Nappe<br />
133.33 0.15 18.74 3.89 8 0.43 Transition<br />
166.67 0.15 18.74 3.89 8 0.39 Transition<br />
16.67 0.15 40.00 3.89 16 0.76 Nappe<br />
33.33 0.15 40.00 3.89 16 0.76 Nappe<br />
50.00 0.15 40.00 3.89 16 0.77 Nappe<br />
66.67 0.15 40.00 3.89 16 0.76 Nappe<br />
100.00 0.15 40.00 3.89 16 0.71 Transition<br />
133.33 0.15 40.00 3.89 16 0.69 Transition<br />
166.67 0.15 40.00 3.89 16 0.68 Skimming<br />
16.67 0.15 22.55 3.26 8 0.56 Nappe<br />
33.33 0.15 22.55 3.26 8 0.56 Nappe<br />
50.00 0.15 22.55 3.26 8 0.53 Nappe<br />
66.67 0.15 22.55 3.26 8 0.52 Nappe<br />
100.00 0.15 22.55 3.26 8 0.51 Nappe<br />
133.33 0.15 22.55 3.26 8 0.47 Transition<br />
166.67 0.15 22.55 3.26 8 0.41 Transition<br />
16.67 0.15 50.00 3.26 16 0.77 Nappe<br />
33.33 0.15 50.00 3.26 16 0.75 Nappe<br />
50.00 0.15 50.00 3.26 16 0.74 Nappe<br />
66.67 0.15 50.00 3.26 16 0.74 Transition<br />
100.00 0.15 50.00 3.26 16 0.72 Transition<br />
133.33 0.15 50.00 3.26 16 0.70 Skimming<br />
166.67 0.15 50.00 3.26 16 0.69 Skimming<br />
[1] H. Chanson, The Hydraulics of Stepped Chutes and Spillways, Balkema,<br />
Lisse, The Netherlands 200<strong>1.</strong><br />
[2] L. Toombes, H. Chanson, Air-Water Flow and Gas Transfer at Aeration<br />
Cascades: A Comparative Study of Smooth and Stepped Chutes,<br />
Proc. of the Int. Workshop on Hydraulics of Stepped Spillways, Zurich, Switzerland,<br />
March, 22 – 24, 2000, 77 – 84.<br />
i 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.clean-journal.com
192 A. Baylar et al. Clean 2007, 35 (2), 186 – 192<br />
[3] H. Chanson, L. Toombes, Experimental Study of Gas-Liquid Interfacial<br />
Properties in a Stepped Cascade Flow, Envir. Fluid Mech. 2002, 2 (3),<br />
241 – 263.<br />
[4] A. Baylar, M. E. Emiroglu, Study of Aeration Efficiency at Stepped<br />
Channels, Proc.-Inst. Civ. Eng.-Water Marit. Eng. 2003, 156 (WM3), 257 –<br />
263.<br />
[5] M. E. Emiroglu, A. Baylar, An Investigation of Effect of Stepped<br />
Chutes with End Sill on Aeration Performance, Water Qual. Res. J. Can.<br />
2003, 38 (3), 527 – 539.<br />
[6] A. Baylar, M. E. Emiroglu, T. Bagatur, An Experimental Investigation<br />
of Aeration Performance in Stepped Spillways, Water Environ. J. 2006,<br />
20 (1), 35 – 42.<br />
[7] M. E. Emiroglu, A. Baylar, Self-Aeration in Smooth and Stepped<br />
Chutes, Int. J. Sci. Technol. 2006, 1 (2), 105 – 113.<br />
[8] A. Baylar, T. Bagatur, M. E. Emiroglu, Aeration Efficiency with Nappe<br />
Flows over Stepped Cascades, Proc. of the Institution of Civil Engineers-<br />
Water Management, 2007, 160 (1), 43 – 50.<br />
[9] A. Baylar, T. Bagatur, M. E. Emiroglu, Prediction of Oxygen Content<br />
of Nappe, Transition and Skimming Flow Regimes in Stepped-Channel<br />
Chutes, J. Environ. Eng. Sci. 2007, DOI: 10.1139/S06-048.<br />
[10] A. Baylar, M. E. Emiroglu, T. Bagatur, Influence of Chute Slope on<br />
Oxygen Content in Stepped Waterways, Int. J. Sci. Technol. 2007.<br />
[11] A. Baylar, O. Kisi, Aeration Efficiency Estimation in Stepped Cascade<br />
Aerators Using Neural Network Approach, J. Hydraul. Eng., ASCE 2007.<br />
[12] R. S. Govindaraju, ASCE Task Committee on Application of Artificial<br />
Neural Networks in Hydrology, Artificial Neural Network in Hydrology,<br />
J. Hydrol. Eng., ASCE 2000, 5 (2), 115 – 137.<br />
Manage your access easily with “MY PROFILE”<br />
17610502_vo <br />
31441611_vo<br />
[13] O. Kisi, River Flow Modeling Using Artificial Neural Networks, J.<br />
Hydrol. Eng. ASCE 2004, 9 (1), 60 – 63.<br />
[14] O. Kisi, Multi-Layer Perceptions with Levenberg-Marquardt Optimization<br />
Algorithm for Suspended Sediment Concentration Prediction<br />
and Estimation, Hydrol. Sci. J. 2004, 49 (6), 1025 – 1040.<br />
[15] D. Hanbay, I. Turkoglu, Y. Demir, Complex Systems Modeling by<br />
Using ANFIS, Int. 15th Turkish Symposium on Artificial Intelligence and<br />
Neural Networks, Turkey 2006, 83 – 90.<br />
[16] D. Hanbay, I. Turkoglu, Y. Demir, A Wavelet Neural Network for<br />
Intelligent Modeling, Int. 15th Turkish Symposium on Artificial Intelligence<br />
and Neural Networks, Turkey 2006, 175 – 182.<br />
[17] J. S. Gulliver, J. R. Thene, A. J. Rindels, Indexing Gas Transfer in Self-<br />
Aerated Flows, J. Envir. Eng., ASCE 1990, 116 (3), 503 – 523.<br />
[18] J. S. R. Jang, ANFIS: Adaptive Network-Based Fuzzy Inference Systems,<br />
IEEE, Trans. Syst., Man. Cybern. 1993, 23 (3), 665 – 685.<br />
[19] L. A. Zadeh, Fuzzy Sets and Applications: Selected Papers (Ed: L. A. Zadeh),<br />
John Wiley & Sons, New York 1987.<br />
[20] E. Avci, I. Turkoglu, Modelling of Tunnel Diode by Adaptive-Network-Based<br />
Fuzzy Inference System, Int. 12th Turkish Symposium on<br />
Artificial Intelligence and Neural Networks, Turkey 2003, T-1, 39 – 4<strong>1.</strong><br />
[21] T. J. McGhee, Water Supply and Sewerage, McGraw-Hill, New York 199<strong>1.</strong><br />
[22] Y. Yasuda, M. Takahashi, I. Ohtsu, Energy Dissipation of Skimming<br />
Flows on Stepped-Channel Chutes, Proc. of the 28th Congress of IAHR,<br />
Beijing, China, September, 2 – 18, 2001, 531 – 536.<br />
Manage your access easily with “MY PROFILE”<br />
Simply register. Registration is fast and free to all internet users.<br />
Easy Access<br />
Save Titles, Articles & Queries for quick<br />
access<br />
Set up roaming access to access content<br />
outside of your institutions network<br />
Get free online sample copies<br />
Get free online trial subscriptions<br />
View a complete list of your subscriptions<br />
and accessible products<br />
www.interscience.wiley.com<br />
Enhanced Tools<br />
Receive E-Mail Alerts when<br />
new content is available<br />
Purchase Article Select Tokens<br />
online<br />
Purchase individual articles online<br />
with Pay-Per-View<br />
i 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.clean-journal.com
Editor-in-Chief Mufit Bahadir, Braunschweig (DE)<br />
Editors<br />
Soil Nicola Senesi, Bari (IT)<br />
Robert Kreuzig, Braunschweig (DE)<br />
Kyoungphile Nam, Seoul (KR)<br />
Air Igor Agranovski, Brisbane, QLD (AU)<br />
Water Gheorghe Duca, Chisinau (MD)<br />
Rainer Lohmann, Narragansett, RI (US)<br />
Bülent Topkaya, Antalya (TR)<br />
Green Chemistry and<br />
Technology<br />
Joseph J. Bozell, Knoxville, TN (US)<br />
Buxing Han, Beijing (CN)<br />
Ayrton Figueiredo Martins, Santa Maria, RS (BR)<br />
Advisory Board M. Batarseh, Karak (JO)<br />
J. de Boer, IJmuiden (NL)<br />
W. Brewitz, Braunschweig (DE)<br />
J. Cornejo, Sevilla (ES)<br />
E. Dabek-Zlotorzynska, Ottawa, ON, (CA)<br />
A. S. Demir, Ankara (TR)<br />
N. Dichtl, Braunschweig (DE)<br />
I. Holoubek, Brno (CZ)<br />
W. Ivens, Heerlen (NL)<br />
B. Jastorff, Bremen (DE)<br />
B. Kamm, Teltow (DE)<br />
B. König, Regensburg (DE)<br />
D. Kotzias, Ispra (IT)<br />
J. Krahl, Coburg (DE)<br />
K. Kümmerer, Freiburg (DE)<br />
J. Metzger, Oldenburg (DE)<br />
B. Moran, Rhode Island (US)<br />
A. Newton, Faro (PT)<br />
A. Piccolo, Portici/Napoli (IT)<br />
G. Pilidis, Ioannina (GR)<br />
G. K. Prakash, Los Angeles, CA, (US)<br />
X. Quan, Delian (CN)<br />
S. Raisuddin, Seoul (KR)<br />
S. J. Santosa, Yogyakarta (ID)<br />
D. Sarigiannis, Bruxelles (BE)<br />
M. Tiris, Gebze/Kocaeli (TR)<br />
V. Viman, Baia Mare (RO)