MODELING AND MULTIOBJECTIVE OPTIMIZATION OF FORWARDCURVED BLADE CENTRIFUGAL FANS USING CFD AND NEURAL NETWORKS
Abolfazl Khalkhali, Mehdi Farajpoor, Hamed Safikhani
Department of Mechanical Engineering, East Tehran Branch, Islamic Azad University, Tehran, Iran Email: abolfazl.khalkhali@gmail.com
Received May 2010, Accepted January 2011 No. 10CSME27, E.I.C. Accession 3190
ABSTRACT
In the present study, multiobjective optimization of ForwardCurved (FC) blade centrifugal fans is performed in three steps. In the first step, Head rise (H _{R} ) and the Head loss (H _{L} ) in a set of FC centrifugal fan is numerically investigated using commercial software NUMECA. Two metamodels based on the evolved group method of data handling (GMDH) type neural networks are obtained, in the second step, for modeling of H _{R} and H _{L} with respect to geometrical design variables. Finally, using the obtained polynomial neural networks, multi objective genetic algorithms are used for Pareto based optimization of FC centrifugal fans considering two conflicting objectives, H _{R} and H _{L} .
Keywords: forwardcurved blade centrifugal fan; multiobjective optimization; CFD; GMDH; genetic algorithms.
´
MODE LISATION ET OPTIMISATION MULTIOBJECTIF D’UN VENTILATEUR
´
CENTRIFUGE A AUBES INCLINEE S VERS L’AVANT, UTILISANT LA
`
´
´
´
ME CANIQUE DES FLUIDES NUME RIQUES (MFN) ET DES RE SEAUX DE
NEURONES
L’objectif de cette e´tude, est l’exe´cution en trois e´tapes de l’optimisation multiobjectif d’un ventilateur centrifuge a` aubes incline´es vers l’avant. Dans un premier temps, l’augmentation de charge et la perte de charge dans un ensemble de ventilateurs centrifuges a` aubes incline´es, sont examine´es nume´riquement utilisant le logiciel commercial NUMECA. Dans un deuxie`me temps, deux me´tamode`les base´s sur la me´thode de traitement de donne´es par groupe (MTDG) de type de re´seaux de neurones, sont obtenus pour la mode´lisation de l’augmentation de charge et de la perte de charge, par rapport aux variables ge´ome´triques. Finalement, en utilisant les re´seaux de neurones polynoˆmes obtenus, des algorithmes ge´ne´tiques multiobjectifs sont utilise´s pour l’optimisation de Pareto d’un ventilateur centrifuge en prenant en conside´ration ces deux objectifs conflictuels l’augmentation de charge et la perte de charge.
Motscle´s : ventilateur centrifuge a` aubes incline´es vers l’avant; optimisation multiobjectif; MFN; MTDG; algorithmes ge´ne´tiques.
63
Transactions of the Canadian Society for Mechanical Engineering, Vol. 35, No. 1, 2011
1. INTRODUCTION
ForwardCurved (FC) blade centrifugal fans or squirrel cage fans are the group of turbo machines which occur in industry in large scales. Nowadays, increasing demands and competition require the use of good models to describe the operation of FC centrifugal fans. Kim and Seo [1] presented a response surface method using three dimensional NavierStokes analyses to optimize the shape of a forwardcurved blade centrifugal fan and finally improved the efficiency of the fan. Lu et al. [2] numerically investigated the internal flow field of centrifugal fans. They could decrease the head loss and increase the total pressure using splitter blades in centrifugal fans. Optimization of FC fans is indeed a multiobjective optimization problem rather than a single objective optimization problem that has been considered so far in the literature. Sugimura et al. [3] investigated a multiobjective optimization process on centrifugal fans using multiobjective robust design exploration method (MORDE). They tried to determine the design variables which have the minimum turbulent noise level and the maximum efficiency. Besides applications to fan optimization, in recent years there have been many efforts to increase the performance of different types of turbo machines. Safikhani and Nourbakhsh [4] investigated a multiobjective optimization approach on centrifugal pumps. They finally presented the Pareto front for centrifugal pumps and defined five optimum points which had the best efficiency and cavitation behavior. Derakhshan et al. [5, 6] optimized a pump as a turbine machine (PAT) for increasing the efficiency using the genetic algorithms (GAs) and incomplete sensitivities method. In centrifugal fans there are some objective functions which are not independent of each other, like efficiency, head rise and input shaft power, so these parameters are not suitable for multi objective optimization process. Head rise and the head loss are important and independent objective functions which can be used in a multiobjective optimization process. These objective functions are either obtained from experiments or computed using very timely and highcost computational fluid dynamic (CFD) approaches, which cannot be used in an iterative optimization task unless a simple but effective metamodel is constructed over the response surface from the numerical or experimental data. Therefore, modeling and optimization of the parameters is investigated in the present study, by using GMDHtype neural networks and multi objective genetic algorithms in order to maximize the head rise and minimize the head loss. System identification and modeling of complex processes using inputoutput data have always attracted many research efforts. System identification techniques are applied in many fields in order to model and predict the behavior of unknown and/or very complex systems based on given inputoutput data [7]. In this way, softcomputing methods [8], which concern computation in an imprecise environment, have gained significant attention. The main components of soft computing, namely, fuzzy logic, neural network, and evolutionary algorithms have shown great ability in solving complex nonlinear system identification and control problems. Many research efforts have been developed that make use of evolutionary methods as effective tools for system identification [9]. Among these methodologies, Group Method of Data Handling (GMDH) algorithm is a selforganizing approach by which gradually complicated models are generated based on the evaluation of their performance on a
set of multiinputsingleoutput data pairs ð
i Þ ( i5 1, 2, …, M ). The GMDH was first
X i
, y
developed by Ivakhnenko [10] as a multivariate analysis method for complex systems modeling and identification, which can be used to model complex systems without having specific knowledge of the systems. The main idea of GMDH is to build an analytical function in a feed forward network based on a quadratic node transfer function [11] whose coefficients are obtained using regression techniques.
Transactions of the Canadian Society for Mechanical Engineering, Vol. 35, No. 1, 2011
64
In recent years, however, the use of such selforganizing networks leads to successful application of the GMDHtype algorithm in a broad range of areas in engineering, science, and economics [12]. Moreover, there have been many efforts in recent years to deploy GAs to design artificial neural networks since such evolutionary algorithms are particularly useful for dealing with complex problems having large search spaces with many local optima [13]. In this way, Gas have been used in a feed forward GMDHtype neural network for each neuron searching its optimal set of connections with the preceding layer [14]. In the former reference, authors have proposed a hybrid genetic algorithm for a simplified GMDHtype neural network in which the connection of neurons are restricted to adjacent layers. Moreover a multiobjective genetic algorithm has also been recently used by some of authors to design GMDHtype neural networks considering some conflicting objectives [15, 16]. In this paper, the head rise and the head loss in a set of forwardcurved blade centrifugal fans are numerically investigated using NUMECA. Genetically optimized GMDH type neural networks are then used to obtain polynomial models for the effects of geometrical parameters of the FC fans on both H _{R} and H _{L} . This approach of metamodeling of those CFD results allows the use of iterative optimization techniques. The obtained simple polynomial models are then used in a Pareto based multiobjective optimization approach to find the best possible combinations of H _{R} and H _{L} , known as the Pareto front. The corresponding variations of design variables, namely, geometrical parameters, known as the Pareto set, constitute some important and informative design principles.
2. CFD SIMULATION OF FC BLADE CENTRIFUGAL FANS
The governing equations of incompressible flow are as follows:
Continuity equation
LV _{i}
Lx _{i}
~0
Reynolds averaged momentum equation
DV
_{i}
Dt
~{ ^{1}
Lp
r Lx _{i}
zn
Standard k– e model
Dk
^{L}
~
Dt
e
Dt ^{~} Lx _{j}
D
L
x _{j}
^{L}
C
k
C
k
k
^{2}
e
k
^{2}
e
{C e2
e
^{2}
k
ð1 Þ
ð2 Þ
ð3 Þ
The dimensions of the case study in the present paper and some operating conditions for the simulations are shown in Table 1. The simulations are performed using Numeca software. Firstly one blade is modeled in Auto Blade 3.6 and then the Design 3D environment of Numeca automatically generates the database with different design variables.
65
Transactions of the Canadian Society for Mechanical Engineering, Vol. 35, No. 1, 2011
Table 1. Dimensions and operating conditions of FC centrifugal fan case study.
Parameter
Value
Outer diameter ( mm) 
333 
Inner diameter (mm) 
210 
Width of blades (mm) 
150 
Mass flow rate (kg/s) 
0.34 
Rotational velocity (rpm) 
634 
Inlet k (m ^{2} /s ^{2} ) 
5 
Inlet e (m ^{2} /s ^{3} ) 
30000 
Outlet static pressure (atm) 
1 
To parameterize the camber line curve, the simple Bezier method is used [17]. A schematic definition of simple Bezier method is shown in Fig. 1. The design variables in this method are leading edge angle ( b _{1} ), trailing edge angle (b _{2} ) and the stagger angle (c ). In the present paper two sections are defined in the blades, one on hub and one on shroud as shown in Fig. 2. It is supposed that b _{1} , b _{2} and c are equal at hub and shroud section due to the 2D nature of FC blades centrifugal fan, which can mathematically be given by
b _{1}_{H}_{u}_{b} ~ b _{1}_{S}_{h}_{r}_{o}_{u}_{d} ~Design Variable 
ð 
4Þ 
b _{2}_{H}_{u}_{b} ~ b _{2}_{S}_{h}_{r}_{o}_{u}_{d} ~ Design Variable 
ð 
5Þ 
c _{H}_{u}_{b} ~ c _{S}_{h}_{r}_{o}_{u}_{d} ~ Design Variable 
ð 
6Þ 
The design variables and their range of variations are shown in Table 2. By changing the geometrical independent parameters according to the Table 2, various designs will be generated
Fig. 1. Blade camber line parameterization using simple Bezier method.
Transactions of the Canadian Society for Mechanical Engineering, Vol. 35, No. 1, 2011
66
Fig. 2. Defining two sections on centrifugal fan blade.
and evaluated using CFD. Consequently, some metamodels can be optimally constructed using GMDHtype neural networks, which will be further used for multiobjective Pareto based design of such centrifugal fans. In this way, 132 various CFD analyses have been performed due to those different design geometrics. For CFD grid generation, the Auto Grid environment of Numeca is coupled with the Auto Blade environment. To test for grid independency, three grid types (named A , B , and C) with increasing grid density are studied and their details are listed in Table 3. The computational results of three grid types for different mass flow rates are compared in Table 4. As can be seen, the maximum difference between the results is less than 6 % so the grid type ( A) is used for all computations in the present study. Figure 3 shows the details of the computational grid for the centrifugal fans. The physical model used in the solver is the ReynoldsAveraged Navier–Stokes equations and the k e turbulence model. Mass flow, k and e are imposed at the fan inlet. A static pressure outlet boundary condition is used at the outlet and finally periodic boundary condition is applied between two blades. The computation is continued until the solution converged with a total residual of less than 2 5. Samples of numerical results, using CFD are shown in Table 5. A typical pressure contour in one of the simulations is shown in Fig. 4. The results obtained in such CFD analysis can now be used to build the response surface of both the head rise and the head loss for those different 132 geometries using GMDHtype polynomial neural networks. Such metamodels will, in turn, be used for the Paretobased multiobjective optimization of the FC fans. A post analysis using the CFD software NUMECA is also performed to verify the optimum results using the metamodeling approach. Finally, the solutions obtained by the approach of this paper exhibit some important tradeoffs
Table 2. Design variables and their range of variations.
Design Variable 
From 
To 
c (deg) b _{1} (deg) b _{2} (deg) N (no. of blades) 
7 
25 
12 
47 

25 
60 

25 
40 
67
Transactions of the Canadian Society for Mechanical Engineering, Vol. 35, No. 1, 2011
Table 3. Details of 3 grid types used in grid independency test.
Blade to Blade
Grid Type 
Inlet 
Outlet 
Periodic 
Meridional 
Total No. of Cells 
A 
20 
25 
79 
37 
444690 
B 
24 
27 
86 
43 
588632 
C 
29 
33 
104 
55 
762820 
among those objective functions which can be simply used by a designer to optimally compromise among the obtained solutions.
3. MODELING OF H _{R} AND THE H _{L} USING GMDHTYPE NEURAL NETWORK
By means of the GMDH algorithm, a model can be represented as set of neurons in which
different pairs in each layer are connected through a quadratic polynomial and thus produce new neurons in the next layer. Such representation can be used in modeling to map inputs to
^
outputs. The formal definition of the identification problem is to find a function ff that can be
approximately used instead of the actual one, f in order to predict output ^yy for a given input
vector X ~ ð x _{1} , x _{2} , x _{3} ,
n Þ as close as possible to its actual output y. Therefore, given M
observations of multiinputsingleoutput data pairs so that
,
x
y _{i} ~ f ð x
i
1
, x ,x ,
i
2
i
3
,x
in
Þ ð i ~1,2
M Þ ,
ð 7Þ
It is now possible to train a GMDHtype neural network to predict the output values ^yy _{i} for any
given input vector X ~ ð x That is
i 1
,
x _{i} _{2} , x _{i} _{3} ,
,
x in
Þ
^
ð 8Þ
The problem is now to determine a GMDHtype neural network so that the square of difference between the actual output and the predicted one is minimized, that is
^yy _{i} _{~} ff ð x , x ,x ,
i
1
i
2
i
3
,x
in
Þ ð i ~1,2
M Þ ,
M
X
i ~1
h
^
ff ð x
i
1
,x , x ,
i
2
i
3
,
x
in
Þ { y _{i} i ^{2} ?min
ð 9Þ
Table 4. Comparison of H _{R} and H _{L} for 3 different grid types.
H _{R} (m) 
H _{L} (m) 

Mass 
ABC 
Max 
Max 

Flow (kg/s) 
Diff (%) 
^{A}^{B}^{C} 
Diff (%) 

.22 
14.82 
14.05 
14.78 
5.19 
1.22 
1.21 
1.20 
1.63 
.34 
12.10 
12.65 
12.08 
4.50 
0.99 
0.98 
0.98 
1.00 
.46 
10.86 
10.77 
10.93 
1.46 
0.88 
0.88 
0.86 
2.27 
Transactions of the Canadian Society for Mechanical Engineering, Vol. 35, No. 1, 2011
68
Fig. 3. CFD structured grid generation for centrifugal fans.
General connection between inputs and output variables can be expressed by a complicated discrete form of the Volterra functional series in the form of
n
n
y~a _{0} z ^{X} a _{i} x _{i} z ^{X}
n
X
i
~1
i ~1 j ~ 1
n
a ij x i x j z X
n
n
X X
i ~1 j ~ 1 k ~1
a _{i}_{j}_{k} x _{i} x _{j} x _{k} z
ð
10 Þ
Which is known as the KolmogorovGabor polynomial [18]. This full form of mathematical description can be represented by a system of partial quadratic polynomials consisting of only two variables (neurons) in the form of
Table 5. Samples of numerical result using CFD.
_{I}_{n}_{p}_{u}_{t} _{D}_{a}_{t}_{a} 
Output Data 

_{N}_{u}_{m} 
c(deg) 
b _{1} (deg) 
b _{2} (deg) 
N 
H _{R} (m) 
H _{L} (m) 
1 
7.45 
12.14 
42.77 
40 
14.236 
1.559 
2 
7.45 
46.14 
59.77 
40 
18.777 
2.056 
3 
9.54 
46.14 
59.77 
40 
15.587 
1.707 
4 
24.45 
46.14 
59.77 
35 
18.89 
2.008 
5 
7.45 
12.14 
42.77 
35 
13.951 
1.481 
6 
24.45 
29.14 
25.77 
35 
12.530 
1.331 
7 
7.45 
29.14 
25.77 
30 
10.841 
1.168 
8 
24.4 
46.14 
25.77 
30 
11.360 
1.220 
9 
7.45 
12.14 
25.77 
30 
10.373 
1.113 
… 

130 
9.54 
12.14 
59.77 
25 
9.488 
0.745 
131 
7.45 
46.14 
42.77 
25 
15.427 
1.729 
132 
9.54 
12.14 
42.77 
25 
5.768 
0.645 
69
Transactions of the Canadian Society for Mechanical Engineering, Vol. 35, No. 1, 2011
Fig. 4. A typical contour of pressure in one of simulations.
^yy~ _{G}
x i ,x j
^{} ~a _{0} za _{1} x _{i} za _{2} x _{j} za _{3} x _{i} x _{j} za _{4} x _{i} ^{2} z a _{5} x _{j} ^{2}
ð 11Þ
There are two main concepts involved within GMDHtype neural networks design, namely, the parametric and the structural identification problems. In this way, some authors presented a hybrid GA and singular value decomposition (SVD) method to optimally design such polynomial neural networks. The methodology in these references has been successfully used in this paper to obtain the polynomial models of H _{R} and H _{L} . The obtained GMDHtype polynomial models have shown very good prediction ability of unforeseen data pairs during the training process which will be presented in the following sections. The input–output data pairs used in such modeling involve two different data tables obtained from the CFD simulation discussed in Section 2. Both of the tables consist of four variables as inputs, namely, the geometrical parameters of the FC fans c, b _{1} , b _{2} (Fig. 1) and N (number of blades) and outputs, which are H _{R} and H _{L} . The tables consist of a total of 132 patterns, which have been obtained from the numerical solutions to train and test such GMDH type neural networks. However, in order to demonstrate the prediction ability of the evolved GMDH type neural networks, the data in both input–output data tables have been divided into two different sets, namely, training and testing sets. The training set, which consists of 112 out of the 132 input– output data pairs for H _{R} and H _{L} , is used for training the neural network models. The testing set, which consists of 20 unforeseen input–output data samples for H _{R} and H _{L} during the training process, is merely used for testing to show the prediction ability of such evolved GMDH type neural network models. The GMDH type neural networks are now used for such input–output data to find the polynomial models of head rise and head loss with respect to their effective input parameters. In order to design, genetically, such GMDH type neural networks described in the previous section, a population of 10 individuals with a crossover probability (Pc ) of 0.7 and mutation probability (Pm ) 0.07 has been used in 500 generations for H _{R} and H _{L} . The corresponding polynomial representation for Head Rise ( H _{R} ) is as follows:
Transactions of the Canadian Society for Mechanical Engineering, Vol. 35, No. 1, 2011
70
Y _{1} ~{ 27: 356z 0: 338 b _{2} z 1: 677 N 0: 00345 b _{2} ^{2} {0: 0234 N ^{2} z0: 00291 N b _{2} ð 
12a 
Þ 
_{2} ~{ 20: 196886 {0: 00586 b _{1} z1: 7555 N z 0:00054 b _{1} ^{2} {0: 0234 N ^{2} z0: 00198 b _{1} N Y 
ð 12b 
Þ 
_{3} ~{ 19: 7199 {0: 10393c z1: 73866 N z 0:00507 c ^{2} {0 :02340 N ^{2} z0 :0039480 c N Y 
ð 12c 
Þ 
_{4} ~{ 1: 552z 0:18001 b _{1} z0: 4275 b _{2} {0: 001652 b _{1} ^{2} { 0: 003306 b _{2} ^{2} { 0: 00013b _{1} b _{2} Y 
ð 12d 
Þ 
_{5} ~ 8: 78175z 0:502008 Y _{2} { 1:1355Y _{1} { 0: 04922Y _{2} ^{2} zz0 :000709Y _{1} ^{2} z 0:1217 Y _{2} Y _{1} Y 
ð 12e 
Þ 
Y _{6} ~{ 6: 2660z 1: 05304Y _{4} { 0:044003Y _{3} {0: 016115 Y _{4} ^{2} z 0:026433Y _{3} ^{2} z 0:02430 Y _{4} Y _{3} ð 12f Þ
H _{R} ~ 1: 898441{ 0: 45095Y _{5} z 1:18816Y _{6} {0: 14790 Y _{5} ^{2} {0 :17717Y _{6} ^{2} z0: 3359 Y _{5} Y _{6}
ð
12g Þ
Similarly, the corresponding polynomial representation of the model for Head Loss ( H _{L} ) is in the form of
Y 
’ _{1} ~1: 2905{ 0: 0799c z0: 01905 b _{1} z0: 00382 c ^{2} z3: 383e {005 b _{1} ^{2} { 0: 000769c b _{1} 
ð 
13a 
Þ 

Y 
’ _{2} ~{2: 8676 z0: 02454 b _{2} z0: 1849 N { 0:00025 b _{2} ^{2} {0: 002538 N ^{2} z0: 000339 N b _{2} 
ð 13b 
Þ 

H 
_{L} ~{ 2: 93563z4 :11599 Y ’ z0: 4002Y _{2} ’ {1 :4661 Y ’ ^{2} {0: 10065Y _{2} ’ ^{2} z0: 602920 Y ’ _{2} 1 1 1 Y 
’ 
ð 13c 
Þ 
The very good behavior of such GMDH type neural network model for head rise is also depicted in Fig. 5, both for the training and testing data. Such behavior has also been shown for the training and testing data of head loss in Fig. 6. It is evident that the evolved GMDH type neural network in terms of simple polynomial equations successfully model and predict the outputs of the testing data that have not been used during the training process. The models obtained in this section can now be utilized in a Pareto multiobjective optimization of the FC centrifugal fans considering both HR and HL as conflicting objectives. Such a study may unveil some interesting and important optimal design principles that would not have been obtained without the use of a multiobjective optimization approach.
Transactions of the Canadian Society for Mechanical Engineering, Vol. 35, No. 1, 2011
71
Fig. 5. CFD vs. Network for H
4. MULTIOBJECTIVE OPTIMIZATION OF FC CENTRIFUGAL FANS USING POLYNOMIAL NEURAL NETWORK MODELS
Multiobjective optimization, which is also called multi criteria optimization or vector optimization, has been defined as finding a vector of decision variables satisfying constraints to give acceptable values to all objective functions. In these problems, there are several objective or cost functions (a vector of objectives) to be optimized (minimized or maximized) simultaneously. These objectives often conflict with each other so that improving one of them will deteriorate another. Therefore, there is no single optimal solution as the best with respect to all the objective functions. Instead, there is a set of optimal solutions, known as Pareto optimal solutions or Pareto front [19] for multiobjective optimization problems. The concept of Pareto front or set of optimal solutions in the space of objective functions in multiobjective optimization problems (MOPs) stands for a set of solutions that are nondominated to each other but are superior to the rest of solutions in the search space. This means that it is not possible to find a single solution to be superior to all other solutions with respect to all objectives so that changing the vector of design variables in such a Pareto front consisting of these nondominated solutions could not lead to the improvement of all objectives simultaneously. Consequently, such a change will lead to deteriorating of at least one objective. Thus, each solution of the Pareto set includes at least one objective inferior to that of another solution in that Pareto set, although both are superior to others in the rest of search space. Such problems can be mathematically defined as:
Transactions of the Canadian Society for Mechanical Engineering, Vol. 35, No. 1, 2011
72
Fig. 6. CFD vs. Network for H _{L} .
Find the vector X ^{} ~ x ^{} _{2} ,
_{1}
,x ^{}
F ð X Þ ~ ½ f _{1}
ð X Þ ,
f
2
ð X Þ ,
,
,x ^{} ^{T} to optimize
n
f
k
ð X Þ ^{T} ,
Subject to m inequality constraints
g _{i} ð X Þƒ 0, i ~1 to m,
And p equality constraints
h _{j} ð X Þ ~ 0, j ~1 to p,
ð 14 Þ
ð
ð
15
16
Þ
Þ
Where X ^{} [< ^{n} is the vector of decision or design variables, and F ð X Þ [< ^{k} is the vector of objective functions, which must each be either minimized or maximized. However, without loss of generality, it is assumed that all objective functions are to be minimized. In order to investigate the optimal performance of the FC centrifugal fan in different conditions, the polynomial neural network models obtained in section 3 are now employed in a multiobjective optimization procedure. The two conflicting objectives in this study are Head Rise ( H _{R} ) and Head Loss ( H _{L}_{)} that is to be simultaneously optimized with respect to the design variables c , b _{1} , b _{2} (Fig. 1) and N (Number of blades). The multiobjective optimization problem can be formulated in the following form:
73
Transactions of the Canadian Society for Mechanical Engineering, Vol. 35, No. 1, 2011
8
>
>
>
> Minimize
>
>
Maximize Head Rise ð Þ ~f _{1} ð
<
>
>
>
>
>
>
:
Subject to
H
R
Head Loss ð
H
L
12 ^{0} ƒ b _{1} ƒ 47 ^{0}
25
25ƒ N ƒ40
^{0} ƒb _{2} ƒ60 ^{0}
g ,b , b
1
2
, N Þ
Þ ~f _{2} g ,b , b
ð
1
2
,N Þ
ð
17Þ
The evolutionary process of Pareto multiobjective optimization is accomplished by using the recently developed algorithm, namely, the elimination diversity algorithm by some of authors [14] where a population size of 60 has been chosen in all runs with crossover probability P _{c} and mutation probability P _{m} as 0.7 and 0.07 respectively. Figure 7 depicts the obtained nondominated optimum design points as a Pareto front of those two objective functions. There are four optimum design points, namely, A, B, C and D whose corresponding designs variables and objective functions are shown in Table 6. Moreover, for more clarity, the design variables of optimum design points have been superimposed with each other in Fig. 8. These points clearly demonstrate tradeoffs in objective functions head rise and head loss from which an appropriate design can be compromisingly chosen. It is clear from Fig. 7 that all the optimum design points in the Pareto front are nondominated and could be chosen by a designer as optimum FC fan. Evidently, choosing a better value for any objective
Fig. 7. Pareto front of head rise and head loss.
Transactions of the Canadian Society for Mechanical Engineering, Vol. 35, No. 1, 2011
74
Table 6. The values of objective functions and their associated design variables of the optimum points.
Point 
c (deg) 
b _{1} (deg) 
b _{2} (deg) 
N 
HR (m) 
HL (m) 
A 
13.48 
12.36 
25.24 
25 
6.5036 
.52068 
B 
12.54 
12.14 
57.82 
25 
10.695 
.8093 
C 
13.33 
12.01 
57.57 
33 
14.376 
1.156 
D 
17.38 
42.19 
58.57 
38 
18.981 
1.8272 
function in the Pareto front would cause a worse value for anoth er objective. The corresponding decision variables of the Pareto front shown in Fig. 7 are the best possible design points so that if any other set of decision variables is chosen, the corresponding values of the pair of objectives will locate a point inferior to this Pareto front. Such inferior area in the space of the two objectives is in fact bottom/right side of Fig. 7. In Fig. 7, the design points A and D stand for the best head loss and the best head rise. Moreover, the other optimum design point, B can be simply recognized from Fig. 7. The design point, B exhibit important optimal design concepts. In fact, optimum design point B obtained in this paper exhibits an increase in head loss (about 23.58 %) in comparison with that of point A whilst its head rise improves about 30.6 % in comparison with that of A. It is now desired to find a tradeoff optimum design point compromising both objective functions. This can be achieved by the method employed in this paper, namely, the mapping method. In this method, the values of objective functions of all nondominated points are mapped into interval 0 and 1. Using the sum of these values for each nondominated point, the tradeoff point simply is one having the minimum sum of those values. Consequently, optimum design point C is the tradeoff point which has been obtained from the mapping method. There are some interesting design facts which can be used in the design of such FC fans. It is clear from Figs. 9 and 10 that from point A to B , design variables c, b _{1} and N are nearly constant whereas b _{2} varies almost linearly. Similarly, from point B to point C, the geometrical
Fig. 8. Overlay graph of the design variables in optimum points.
75
Transactions of the Canadian Society for Mechanical Engineering, Vol. 35, No. 1, 2011
Fig. 9. Variations of design variables corresponding to the Pareto front of head rise.
Fig. 10. Variations of design variables corresponding to the Pareto front of head loss.
Transactions of the Canadian Society for Mechanical Engineering, Vol. 35, No. 1, 2011
76
Fig. 11. Overlay graph of the obtained optimal Pareto front with the numerical data.
design parameters c, b _{1} and b _{2} are nearly constant and the number of blades (N ) varies almost linearly. Finally from point C to point D , design parameters b _{2} and N are nearly constant whereas c and b _{1} varies almost linearly. These useful relationships that indefeasible between the optimum design variables of FC fans cannot be discovered without the use of multiobjective Pareto optimization process presented in this paper. The Pareto front obtained from the GMDHtype neural network model (Fig. 7) has been superimposed with the corresponding CFD simulation results in Fig. 11. It can be clearly seen from this figure that such obtained Pareto front lies on the best possible combination of the objective values of CFD data, which demonstrate the effectiveness of this paper, both in deriving the model and in obtaining the Pareto front. In a post numerical study, the design points of the obtained Pareto front have been re evaluated using CFD. It should be noted that the optimum design points of the Pareto set are not included in the training and testing sets utilized metamodeling using GMDHtype neural network which makes such reevaluation sensible. The results of such CFD analysis re evaluations have been compared with those of numerical results using the GMDH model in Table 7. As seen the CFD data agree well with the GMDH data.
5. CONCLUSION
Genetic algorithms have been successfully used both for optimal design of generalized GMDH type neural network models of Head Rise and Head Loss in FC centrifugal fans and for multiobjective Pareto based optimization of such processes. Two different polynomial relations for H _{R} and H _{L} have been found by evolved GSGMDH type neural networks using
77
Transactions of the Canadian Society for Mechanical Engineering, Vol. 35, No. 1, 2011
Table 7. Reevaluation of the obtained optimal Pareto front using CFD.
HR (m) 
HL(m) 

Point 
GMDH 
CFD 
Error (%) 
GMDH 
CFD 
Error (%) 
A 
6.5036 
6.351 
2.31 
.52068 
.501 
3.79 
B 
10.695 
10.09 
5.99 
.8093 
.778 
3.98 
C 
14.376 
14.11 
2.87 
1.156 
1.10 
5.09 
D 
18.981 
18.25 
4.01 
1.8272 
1.765 
3.23 
some CFD simulations for input–output data of the fans. The derived polynomial models have been then used in an evolutionary multiobjective Pareto based optimization process so that some interesting and informative optimum design aspects have been revealed for fans with respect to the design variables such as geometrical parameters of c, b _{1} , b _{2} (Fig. 1) and number of blades ( N ). Consequently, some very important tradeoffs in the optimum design of FC centrifugal fans have been obtained and proposed based on the Pareto front of two conflicting objective functions. Such combined application of GMDH type neural network modeling of input–output data and subsequent nondominated Pareto optimization process of the obtained models is a very promising technique for discovering useful and interesting design relationships.
REFERENCES
1. Kim, K. and Seo, S., ‘‘Application of numerical optimization technique to design of forward curved blades centrifugal fan,’’ Japan Society of Mechanical Engineers International Journal Series B , Vol. 49, No. 1, pp. 152–158, 2006.
2. Lu, Y., Lu, Q., Wang, J. and Zhang, J., ‘‘3D Numerical simulation of centrifugal fan inside flow field and modification investigation,’’ Proceedings of the International Conference on Power Engineering, pp. 429–433, 2007.
3. Sugimura, K., Jeong, S., Obayashi, S. and Kimura, T., ‘‘Krigingmodelbased multiobjective robust optimization and tradeoff rule mining of a centrifugal fan with dimensional uncertainty,’’ Journal of computational science and technology, Vol. 3, No. 1, 2009.
4. Safikhani, H. and Nourbakhsh, A., ‘‘Modeling and multiobjective optimization of centrifugal pumps using CFD and genetic algorithms,’’ Engineering Applications of Computational Fluid Mechanics, in press, 2010.
5. Derakhshan, S., Mohammadi, B. and Nourbakhsh, A., ‘‘Incomplete sensitivities for 3D radial turbo machinery blade optimization,’’ Computers & Fluids , Vo. 37, No. 10, pp. 1354–1363, 2008.
6. Derakhshan, S., Mohammadi, B. and Nourbakhsh, A., ‘‘The comparison of incomplete sensitivities and genetic algorithms application in 3D radial turbo machinery blade optimization,’’ Computers & Fluids , Vo. 39, No. 10, pp. 2022–2029, 2010.
7. Astrom, K.J. and Eykhoff, P., ‘‘System identification: A survey,’’ Automatica, Vol. 7, pp. 123–162,
1971.
8. Sanchez, E., Shibata, T. and Zadeh, L.A., Genetic Algorithms and Fuzzy Logic Systems, World Scientific, Riveredge, NJ, 1997.
9. Kristinson, K. and Dumont, G., ‘‘System identification and control using genetic algorithms,’’ IEEE Transactions on Systems, Man and Cybernetics , Vol. 22, No. 5, pp. 1033–1046, 1992.
10. Ivakhnenko, A.G., ‘‘Polynomial theory of complex systems,’’ IEEE Transactions on Systems, Man and Cybernetics, SMC1, pp. 364–378, 1971.
Transactions of the Canadian Society for Mechanical Engineering, Vol. 35, No. 1, 2011
78
11.
Farlow, S.J., SelfOrganizing Method in Modeling: GMDH Type Algorithm, Marcel Dekker Inc.,
1984.
12. Porto, V.W., ‘‘Evolutionary computation approaches to solving problems in neural computa tion,’’ In Handbook of Evolutionary Computation Back, T., Fogel, D. B. and Michalewicz, Z.
(Eds), Institute of Physics Publishing and New York: Oxford University Press, pp. D1.2:1– D1.2:6, 1997.
13. Amanifard, N., NarimanZadeh, N., Borji, M., Khalkhali, A. and Habibdoust, A., ‘‘Modeling and Pareto optimization of heat transfer and flow coefficients in micro channels using GMDH type neural networks and genetic algorithms,’’ Energy Converse Management, Vol. 49, pp. 311–325,
2008.
14. Atashkari, K., NarimanZadeh, N., Golcu, M., Khalkhali, A. and Jamali, A., ‘‘Modeling and multiobjective optimization of a variable valve timing sparkignition engine using polynomial
neural networks and evolutionary algorithms,’’ Energy Converse Management; Vol. 48, No. 3, pp. 1029–1041, 2007.
15. Safikhani, H., Hajiloo, A. and Ranjbar, M.A., ‘‘Modeling and multiobjective optimization of cyclone separators using CFD and genetic algorithms,’’ Computers & Chemical Engineering, in press, Chemical Engineering Research and Design, 2010.
16. Safikhani, H., AkhavanBehabadi, M.A., NarimanZadeh, N. and Mahmoodabadi, M.J., ‘‘Modeling and multiobjective optimization of square cyclones using CFD and neural networks,’’ in press, 2010.
17. Auto Blade v.7 user guide. Numeca, 2006.
18. Coello Coello, C.A. and Christiansen, A.D., ‘‘Multiobjective optimization of trusses using genetic algorithms,’’ Computers & Structures, Vol. 75, pp. 647–660, 2000.
19. Pareto, V., Cours d’e´conomie politique, Lausanne, Switzerland, Rouge, 1896.
79
Transactions of the Canadian Society for Mechanical Engineering, Vol. 35, No. 1, 2011
Bien plus que des documents.
Découvrez tout ce que Scribd a à offrir, dont les livres et les livres audio des principaux éditeurs.
Annulez à tout moment.