Sirocco Fan Design for Residential Ventilation ...

Proceedings of ASME Turbo Expo 2012 GT2012 June 11-15, 2012, Copenhagen, Denmark

GT2012-68528

SIROCCO FAN DESIGN FOR RESIDENTIAL VENTILATION THROUGH MULTIOBJECTIVE OPTIMIZATION TO ENHANCE AERODYNAMIC PERFORMANCE Jin-Hyuk Kim Dept. of Mechanical Engineering, Inha University Incheon, Republic of Korea [email protected]

Kyung-Hun Cha Dept. of Mechanical Engineering, Inha University Incheon, Republic of Korea [email protected]

ABSTRACT A multi-objective optimization of a sirocco fan for residential ventilation has been carried out in the present work. A hybrid multi-objective evolutionary algorithm combined with response surface approximation is applied to optimize the totalto-total efficiency and total pressure rise of the sirocco fan for residential ventilation. Three-dimensional Reynolds-averaged Navier-Stokes equations with the shear stress transport turbulence model are discretized by finite volume method and solved on hexahedral grids for the flow analysis. Numerical results are validated with the experimental data for the total-tototal efficiency and total pressure. The total-to-total efficiency and total pressure rise of the sirocco fan are used as objective functions for the optimization. In order to improve the total-tototal efficiency and total pressure rise of the sirocco fan, four variables defining the scroll cut-off angle, scroll diffuser expansion angle, hub ratio and the blade exit angle, respectively, are selected as the design variables in this study. Latinhypercube sampling as design-of-experiments is used to generate the design points within the design space. A fast nondominated sorting genetic algorithm with an ε–constraint strategy for the local search is applied to determine the global Pareto-optimal solutions. The trade-off between two objectives is determined and discussed with respect to the representative clustered optimal solutions in the Pareto-optimal solutions compared to the reference shape. NOMENCLATURE b1 Impeller height C Regression analysis coefficient D1 and D2 Inside and outside diameters, respectively EFFI Total-to-total efficiency EXP Experimental data hr Hub ratio LE Leading edge

Kwang-Yong Kim Dept. of Mechanical Engineering, Inha University Incheon, Republic of Korea [email protected]

n Number of design variables P Total pressure PRE Total pressure rise P1-5 Control points generated from B-spline curve Q Volumetric flow rate R2 Correlation coefficient in least squares surface fitting R2adj Adjusted correlation coefficient Rc Cut-off radius R1-4 Scroll expansion angles TE Trailing edge x Set of design variables Greek Symbols α Diffuser expansion angle b1 and b2 Blade inlet and outlet angles, respectively h Total-to-total efficiency θc Cut-off angle t Torque W Angular velocity Subscripts in Inlet out Outlet INTRODUCTION With the recent focus on qualitative improvement in the standard of living, desire of modern people for a pleasant life is gradually being increased. However, residential space occupying their life is mostly exposed to air pollutions such as incomplete gases, cooking odors, smoke, steam, and so on. Thus, the role of ventilation fans is very important to create a pleasant residential environment. Among various fans for residential ventilation, sirocco fan which represents multi-blades centrifugal fan has smaller size, lower noise, and larger flow rate. Due to these advantages, sirocco fans are widely being applied to residential ventilation. The flow field in a sirocco fan is extremely complicated

1

Copyright © 2012 by ASME

owing to three-dimensional unsteady vortical flow structure induced by the interaction between multi-blades impeller and scroll. However, with the recent developments in computational fluid dynamics and computing power, numerical analysis method using three-dimensional Reynolds-averaged NavierStokes (RANS) equations becomes possible to analyze and explore the flow field in a sirocco fan. Thus, the RANS analysis is effectively being used to the design of sirocco fan by many researchers. Adachi et al. [1] studied the effects of flow in the volute casing on the performance of a sirocco fan through numerical analysis using three-dimensional RANS equations. Hah et al. [2] reported the characteristics of internal flow field in a sirocco fan using three-dimensional incompressible RANS equations. Younsi et al [3] conducted a systematic study to understand the flow phenomena in a sirocco fan through both three-dimensional and two-dimensional numerical analyses based on unsteady RANS approach. Jung and Baek [4] performed a numerical study on the unsteady flow behavior and the performance of an automotive sirocco fan using threedimensional unsteady RANS equations. Behzadmehr et al. [5] carried out a numerical sensitivity analysis to study and understand the effects of the entrance design parameters of a backward-inclined sirocco fan on the efficiency. Systematic optimization techniques associated with RANS analysis have recently become a practical tool for turbomachinery designs. Numerical optimization strategies coupled with various surrogate models have been widely applied to the design of turbomachinery to enhance aerodynamic performance. In particular, response surface approximation (RSA) [6] as the surrogate model has been applied to the optimization by many researchers [7-9]. The RSA model can use information collected from various sources and by different tools; thus, it is effective for both single and multioptimization problems. Kim and Seo [10] maximized the efficiency of a centrifugal fan having forward-curved-blade with four design variables through a numerical design optimization in conjunction with the RSA model. Marjavaara et al. [11] employed multiple surrogate model approximations including the RSA model for shape optimization of a hydraulic turbine diffuser. Yi et al [12] applied an optimization method that combines a hybrid multi-objective evolutionary algorithm (MOEA) [13, 14] with the RSA model to a transonic axial compressor rotor for two different optimization objectives. Kim et al. [15] conducted a hybrid MOEA coupled with the RSA model to understand the coupled effects of diverse variables on the enhancement of aerodynamic performance for an axial-flow ventilation fan. Kim et al. [16] also performed an optimization to improve operating stability of a transonic axial compressor through a hybrid MOEA coupled with the RSA model. This work presents a numerical optimization procedure for design of a sirocco fan for residential ventilation. The optimization uses a hybrid MOEA combined with the RSA model for surrogate modeling. For the multi-objective optimization, the total-to-total efficiency and total pressure rise are employed as the objective functions with four design variables defining the scroll cut-off angle, scroll diffuser

Fig. 1 Schematic diagram of sirocco fan expansion angle, hub ratio, and blade exit angle, respectively. The objective functions are numerically assessed through threedimensional RANS analysis at each design point sampled by Latin hypercube sampling (LHS). The optimal solutions are analyzed and discussed to compare its aerodynamic performance with that of a reference shape. SIROCCO FAN MODEL A sirocco fan model for residential ventilation has been investigated in this work. The sirocco fan model shown in Fig. 1 consists of an impeller with forty-five blades and a scroll, and is operated at the speed of 2,119 rpm. The blade sections in an impeller are defined by circular-arc. The motor for the operation of this fan is installed inside of the impeller due to the volume restriction, as shown in Fig. 1. The volumetric flow rate and exit total pressure at the design point are 0.06 m3/s and 237.36 Pa with the total-to-total efficiency, 59.58%, respectively. The detailed specifications of the sirocco fan model are listed in Table 1. Table 1 Design specifications for sirocco fan Impeller D2, mm

70.5

Blade shape

Cir. arc

D1/D2

0.852

Chord length, mm

10.64

b1, mm

58

Thickness, mm

1.4

b1, b2, deg.

58, 141

Number of blades

45

Scroll

2

R1, mm

133

θc, deg.

R2, mm

115

Rc, mm

92 10 2

R3, mm

101

Outlet area, m

0.007

R4, mm

91

α, deg.

34


NUMERICAL METHODS The flow field in the sirocco fan is analyzed using the commercial code ANSYS CFX 11.0 [17]. Blade-Gen and Design-Modeler are employed to create the blade profile and the scroll definition, respectively, and Turbo-Grid and ICEM-CFD are used to generate the computational meshes for the blade and scroll, respectively. CFX-Pre, CFX-Solver, and CFX-Post are applied for defining boundary conditions, solving, and postprocessing, respectively. The three-dimensional steady-state incompressible RANS equations are discretized using the finite volume method (FVM). A high-resolution scheme that is second-order accurate in space is used to solve the convectiondiffusion equations [18]. The k-ω-based shear stress transport (SST) model [19] is used as a turbulence closure model for accurate prediction of flow separation under adverse pressure gradient [20]. This model uses k-ω model in the near-wall region and k-ε model in the bulk domain, and a blending function ensures smooth transitions between these two models. Accuracy of the numerical scheme in prediction of turbulent flows depends strongly on treatment of wall shear stress. In this study, to benefit from the SST model, high resolution of the boundary layer is assured with more than 10 mesh points, and the nearwall grid resolution is adjusted to keep y+ ≤ 2 to accurately capture wall shear stress and also to implement low-Reynoldsnumber SST model [17]. As shown in Fig. 2, the whole inside domain of the sirocco fan consisting of an impeller with forty-five blades and a scroll is considered as the computational domain for the numerical analysis. This computational domain includes a rotating impeller domain and two stationary domains (suction and scroll domains). Air at 25 °C is considered as the working fluid. The total pressure and design mass flow rate are set at the inlet and outlet of the computational domain, respectively. The solid surfaces in the computational domain are considered to be hydraulically smooth with adiabatic and no-slip conditions. A frozen rotor method [17] is used at the interface between the stationary and rotating domains. A hexahedral grid system is constructed in the computational domain, which has O-type grids near the blade surfaces and H-type grids in other regions as shown in Fig. 2. The suction and scroll domains are constructed using approximately 200,000 and 300,000 grid points, respectively, whereas the rotating domain is constructed using approximately 1,340,000 grid points. Thus, the total optimum grid system selected by the grid-independency test has approximately 1,840,000 grid points as reported by the previous work [21]. Root-mean-square (RMS) values of the residuals of the governing equations for convergence criteria are specified to be −6 at least 10 for all equations. The physical time scale is set to 1/W, where W is the angular velocity of the blades. The converged solutions are obtained after about 800 iterations. The computations are performed on a personal computer with an Intel Core I7 CPU having clock speed of 2.67 GHz. The computational time per each calculation is approximately 10 hours.

Fig. 2 Hexahedral grid system in the computational domain OPTIMIZATION TECHNIQUES The purpose of the present multi-objective optimization process is to maximize both the total-to-total efficiency (η) and the pressure rise (Pr), which are defined, respectively, as follows:

h=

( Pout - Pin ) × Q t ×W

Pr = Pout - Pin

(1) (2)

where, P, Q, τ, and W are the total pressure, volumetric flow rate, torque, and angular velocity, respectively. The subscripts in and out represent the inlet and outlet, respectively. Thus, the total-tototal efficiency and total pressure rise are expected to be maximized simultaneously through the multi-objective optimization. From the previous literatures [1-5, 22, 23], many studies on the sirocco fans have been performed to understand the effects of design parameters on fan performance. For example, Adachi et al. [1] studied the effects of the diverse variables related to the scroll of a sirocco fan. Adachi and Sugita [22] also demonstrated the effects of the blade inlet and outlet angles of a sirocco fan through experimental tests. Han and Maeng [23] maximized the efficiency of a sirocco fan with the design variables related to the scroll tongue. On the basis of the previous literatures, the present study employs four design variables, i.e., the scroll cut-off angle (θc), scroll diffuser expansion angle (α), hub ratio (hr) of the impeller, and blade exit angle (b2). Fig. 3(a) shows definitions of the scroll cut-off and diffuser expansion angles. The scroll cut-off angle is defined as an angle between y-axis and the tongue. The scroll diffuser expansion angle is defined as an angle between xaxis and the slope of diffuser expansion as shown in Fig. 3(a). The hub ratio is defined as the ratio of the inside diameter (D1) to the outside diameter (D2) of the impeller; D2 is fixed to limit the size of the impeller, and only D1 is varied to change the hub ratio. Here, the hub ratio is changed with the blade stacks

3


including the hub, mid span and tip. Fig. 3(b) shows an example of the changed hub ratio with the variation of D1. The angle, β is defined as an angle between the axis of rotation and a tangent of camber line of the blade. β distribution of the blade is illustrated in Fig. 3(c). In this study, β distribution is changed on the blade of the fixed meridional geometry by the control points represented by the B-spline curve as shown in Fig. 3(c). The advantage of using the B-spline curve for blade shape parameterization is that only control points located along the curves can control the curves. When one control point in the B-spline curve is moved vertically, the others are being kept fixed. Thus, each control point is controlled independently, and these all points can be considered as design variables. In this study, to change the blade exit angle (b2), all control points are being kept except the control point P5. That is to say, the blade exit angle is changed by the control point P5. For design optimization, it is important to find the feasible design space formed by ranges of the design variables, which are determined by sensitivity tests. The design ranges of the design variables are decided as shown in Table 2. Thirty-five design points are generated within the design space with the help of LHS [24] as design-of-experiment. The objective functions at these design points are evaluated by RANS analysis. Multi-objective optimization based on evolutionary algorithms combined with a surrogate model requires many evaluations of the objective functions in the search for optimal solutions. In this study, the RSA [6] is employed as the surrogate model and applied to predict the objective function values in the design space. The RSA is a methodology of fitting a polynomial function to discrete responses obtained from numerical calculations. It represents the association between design variables and response functions. The constructed response of a second-order polynomial RSA can be expressed as follows: n

f ( x) = C 0 +

åC j =1

n jxj

+

åC j =1

n 2 jj x j

+

åå C

ij x i x j

(a) Definition of the scroll cut-off and diffuser expansion angles

(b) Definition of the hub ratio

(3)

i¹ j

where, C, n, and x represent the regression analysis coefficients, number of design variables, and a set of design variables, respectively, and the number of regression analysis coefficients (C0, Ci, etc.) is (n + 1) × (n + 2)/2. The RSA model is constructed to approximate the objective function values for an evolutionary algorithm, and a hybrid MOEA is used to obtain global Pareto-optimal solutions (POSes) [14]. Approximate POSes are obtained by using the real-coded fast and elitist non-dominated sorting genetic algorithm (NSGA-II) developed by Deb [13] for two objective functions, the efficiency and the pressure rise. Here, “realcoded” indicates that crossovers and mutations are performed in real space to obtain a response from NSGA-II. The POSes are then refined by searching for a local optimal solution for each objective function over all the NSGA-II-derived optimal solutions; the search uses sequential quadratic programming (SQP) [25] with NSGA-II solutions as initial guesses. SQP is a generalization of Newton’s method, which is a gradient-based

(c) b angle distribution by B-spline curve Fig. 3 Definition of design variables Table 2 Ranges of the design variables

4

Variables

Lower

Reference

Upper

θc, deg.

84.268

92.268

98.268

α, deg.

18.34

33.67

40.34

hr

0.8233

0.8517

0.8801

b2, deg.

120.914

140.914

165.914


RESULTS AND DISCUSSION To evaluate the accuracy of the numerical analysis, the results of the numerical analysis should be validated prior to the design optimization. The validity of the numerical results has been verified in comparison with the experimental data in the previous work [21]. The sirocco fan shown in Fig. 1 has been used for this validation, and in this work this fan is regarded as the reference fan model. Fig. 4 shows results of the validation for the system curves for the total-to-total efficiency and total pressure. The numerical results show good agreements with experimental data throughout the whole range of volume flow rate except for efficiency values in the low volume flow region. To perform multi-objective optimization of a sirocco fan, the RSA model is trained for both objective functions, the totalto-total efficiency and total pressure rise, by using RANS analysis results at the design points selected by LHS. In the RSA model, an analysis of variance (ANOVA) and a regression analysis including t-statistics [6] are conducted to measure the uncertainty in a set of coefficients in a polynomial. The values of R2 and R2adj for second-order curve-fitting and the root mean square error (RMSE) for the RSA model are listed in Table 3. Here, R2 and R2adj indicate the correlation coefficient in least squares surface fitting and adjusted correlation coefficient, respectively. The values of R2adj for each objective function are 0.939 and 0.995, respectively. These values are reliable according to the 0.9 < R2adj< 1.0 range suggested by Giunta [26] for accurate prediction of RSA models. Leave-one-out crossvalidation (CV) [27] is also conducted to assess the accuracy of the RSA models. Although it is uncertain how well the CV is correlated with the accuracy of the RSA model, the estimation of the generalization errors is nearly unbiased, as it takes into account the CV of the RSA model at every design point. The estimated CV errors are shown in Table 3. The functional forms of both objective values that are obtained by the RSA model can be expressed in terms of normalized design variables, respectively, as follows:

h = -0.5847+0.0015x1-0.0418x2-0.0051x3 -0.0364x4+0.0111x1x2+0.019x1x3+0.0149x1x4

600

70

500

60 50

400 40 300 30 200

EFFI-EXP EFFI-RANS PRE-EXP PRE-RANS

100 0

0

0.015

0.03

0.045

20 10 0.06

Total-to-total efficiency [%]

Total pressure [Pa]

optimization technique. Two approaches are usually applied to perform a local search [14]. In one approach, all the objectives are combined into a single composite objective, and the optimum is sought. In the other, one objective is optimized by treating the others as equality constraints, and the process is repeated for all objectives. In this study, the first objective is optimized, whereas the second objective is fixed. Then the local search is repeated for the second objective by keeping the first objective at a fixed value. This process produces two new sets of optimal solutions, which are then merged with the NSGA-II solutions. From these solutions, dominated solutions are discarded, and then duplicate solutions are removed to produce the global POSes. Subsequently, a local search is conducted to improve the quality of the POSes.

0 0.09

0.075

3

Volume flow rate [m /s]

Fig. 4 Validation of the numerical results [21]

Table 3 Results of ANOVA and regression analysis Objective functions η

R2

R2adj

RMSE

CV errors

0.965

0.939

3.98×10-3

5.26×10-3

Pr

0.997

0.995

1.35×10-2

1.92×10-2

+0.0117x2x3+0.0278x2x4-0.014x3x4+0.0066x12 +0.021x22+0.0194x32-0.0165x42 Pr = -199.944-3.0572x1-34.4638x2-29.293x3 -43.5411x4+0.8932x1x2+9.9572x1x3+9.9954x1x4 +11.1459x2x3+10.4813x2x4-2.0171x3x4-1.7402x12 +14.4648x22+28.3121x32-29.5495x42

(4)

(5)

where, x1, x2, x3, and x4 indicate θc, α, hr, and b2, respectively. The hybrid MOEA based on the above response functions is used to obtain the global POSes through the real-coded NSGAII. The real-coded NSGA-II is invoked to obtain well-spread approximate POSes with 250 generations and 100 populations, and crossover and mutation probabilities of 0.95 and 0.25, respectively, are chosen. The crossover and mutation parameters are set to 10 and 50, respectively. Here, these parameters are adjusted one by one to suit the nature of the problem. Fig. 5 shows the global POSes that were generated by a hybrid MOEA through the RSA model. As both objectives are to be maximized, the Pareto optimal front composed of POSes shows a convex shape. Every POS is thought to have its own optimized conditions for the objective functions. Extreme ends of the Pareto optimal front represent a pair of the highest value of one objective function and the lowest value of the other objective function. Any improvement in one objective function leads to the deterioration of the other. A trade-off analysis shows an obvious correlation between the total-to-total efficiency and total pressure rise. Higher totalto-total efficiency is obtained at a lower total pressure rise and

5


288 287.5

Total pressure rise [Pa]

vice versa. Meanwhile, the POSes are grouped through K-means clustering [28] to determine the representative solution for a group of solutions. In this study, four representative clusters are formed and distributed equally among the POSes. They are also numerically evaluated through RANS analysis. The values of the design variables corresponding to the clustered optimal solutions (COSes) are shown in Table 4. Two design variables, α and hr, tend to increase as the solution moves from A to D on the curve of Pareto optimal front, while θc and b2 have almost the same values at every COS. The values of the objective functions at the COSes and the objective function values calculated by RANS analysis are shown in Table 5. RANS calculations of the objective functions are also compared with the reference values. It can be seen that the COSes are superior to the reference values for both the efficiency and pressure rise. The results for COSes A and D show improvements in the total-to-total efficiency by 4.136 and 4.417%, respectively, and improvements in the total pressure rise by 53.360 and 48.029 Pa, respectively in comparison with the values for the reference shape. To find the main factors responsible for the improvement of the performance, the internal flow fields of the representative COSes A and D are compared with that of the reference shape. As shown in Fig. 5 and Table 5, COSes A and D have the highest total pressure rise and the highest total-to-total efficiency among the COSes in the global POSes, respectively. Thus, COSes A and D are selected as the representative highpressure and high-efficiency designs of the sirocco fan, respectively. Fig. 6 shows the pressure distributions along observation lines on scroll surface for the reference and COSes A and D. Observation lines, 1, 2 and 3, are located at 10, 50, and 90% spans, respectively, as shown in Fig. 6(a). On all observation lines, a rapid pressure drop is observed at the cut-off region for the reference and COSes A and D. It is noted that the loss occurs mainly at the scroll cut-off region. However, COSes A and D at the cut-off region have slightly lower pressure values than those of the reference shape. In Fig. 6, the pressure distributions for COSes A and D are similar, except in the diffuser expansion region, and these values are much higher than those for the reference on most of the observation lines. COS A, which features the highest overall pressure rise, shows the highest pressure values at the diffuser expansion region. Fig. 7 shows the velocity vector distributions near the cutoff region at 10% span. For COSes A and D, sizes of the separation zones near the pressure surfaces of the blades in the impeller are reduced in comparison with the reference shape. Especially for COS D which features the highest overall efficiency, this phenomenon is remarkable as shown in Fig. 7(c). Reducing these separation zones contributes to the enhancement of the efficiency. Fig. 8 shows the streamlines at 90% span for the reference and COSes A and D. In the reference shape, a large separation zone is observed on the upper wall of the exit duct. However, this separation zone almost disappears in COSes A and D, as shown in Fig. 8(b) and (C).

A B

287 286.5

C 286 285.5 Global Pareto-optimal solutions Clustered optimal solutions

285 284.5 63.8

63.9

64

64.1

D

64.2

64.3

64.4

Total-to-total efficiency [%]

Fig. 5 Global Pareto-Optimal Solutions Table 4 Clustered Optimal Solutions Designs

Design variables θc, deg.

α, deg.

hr

b2, deg.

Reference

92.268

33.67

0.8517

140.914

COS A

84.268

27.44

0.8473

165.914

COS B

84.268

30.03

0.8477

165.914

COS C

84.268

33.66

0.8478

165.914

COS D

84.268

36.40

0.8479

165.914

Table 5 Objective function values of the COSes Designs Reference

MOEA

RANS

h, %

Pr, Pa

-

-

h, %

Pr, Pa

59.582 237.357

Increment

h, %

Pr, Pa

-

-

COS A

63.933 287.624 63.718 290.717 4.136

53.360

COS B

64.115 287.366 63.862 289.135 4.280

51.778

COS C

64.276 286.336 63.923 287.154 4.341

49.797

COS D

64.322 285.022 64.00 285.386 4.417

48.029

Fig. 9 shows the isosurfaces having low velocity of 2 m/s. In Fig. 9, the low velocity regions are formed just downstream of the impeller and upstream of scroll exit. As discussed earlier, the low velocity regions observed downstream of the impeller represent the separation zones. As shown in Figs. 9(b) and (c), the low velocity regions in COSes A and D are remarkably reduced in all blade passages compared to the reference shape as discussed above for Fig. 7. The low velocity regions formed in the exit duct are also reduced considerably in comparison with the reference shape. This is also consistent with that discussed for Fig. 8. These results show clearly the reason for performance enhancement by the present optimization.

6


(a) Locations of observation lines 350 300 ¯

Cut-off region

(a) Reference shape

200 150

¯

Pressure [Pa]

250

Diffuser expansion region

100 50

REF COS A COS D

0 -50 -100

A

A'

Position

(b) Line 1 at 10% span (b) COS A

300 250 ¯

Cut-off region

150 ¯

Pressure [Pa]

200

100


50 0

REF COS A COS D

-50 -100

A

A'

Position

(C) COS D

(c) Line 2 at 50% span 300 ¯

Cut-off region

Fig. 7 Velocity vectors near the cut-off region at 10% span ¯

Pressure [Pa]

200 100


0 -100

REF COS A COS D

-200

A

Position

A'

(d) Line 3 at 90% span Fig. 6 Pressure distributions along observation lines on scroll

CONCLUDING REMARKS A multi-objective optimization of a sirocco fan for residential ventilation has been performed by using a hybrid MOEA combined with RSA surrogate modeling through threedimensional RANS analysis. In order to improve the total-tototal efficiency and total pressure rise of the sirocco fan, four variables defining the scroll cut-off angle, scroll diffuser expansion angle, hub ratio, and the blade exit angle, respectively, are selected as the design variables for the present optimization. The optimum designs, COSes A and D as the representatives of the high-pressure and high-efficiency designs

7


(a) Reference shape (a) Reference shape

(b) COS A (b) COS A

(C) COS D

(C) COS D Fig. 8 Streamlines at 90% span of scroll

Fig. 9 Isosurface having low velocity of 2 m/s

among the Pareto-optimal solutions, respectively, show improvements in the total pressure rise of 53.360 and 48.029 Pa, respectively, and improvements in the total-to-total efficiency of 4.14 and 4.42%, respectively, in comparison with the reference shape. By analyzing the internal flow fields of COSes A and D, it is found that separation zones in the impeller and exit duct are reduced remarkably in the optimum designs compared to the reference design, which contributes to the enhancement of efficiency.

ACKNOWLEDGMENTS This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MEST) (No. 2011-0015903). REFERENCES [1] Adachi, T., Sugita, N., and Ohomori, S., 2004, “Study on the Effects of Flow in the Volute Casing on the Performance of a Sirocco Fan,” Journal of Thermal

8


Science, Vol. 13, No. 3, pp. 199-206. [2] Hah, J. H., Moon, Y. J., and Park, J. M., 2001, “Computational Analysis of the Three-Dimensional Flow Fields of Sirocco Fan,” International Journal of AirConditioning and Refrigeration, Vol. 9, No. 2, pp. 44-50. [3] Younsi, M., Bakir, F., Kouidri, S., and Rey, R., 2007, “Numerical and Experimental Study of Unsteady Flow in a Centrifugal Fan,” Proceedings of the Institution of Mechanical Engineers, Part A-Journal of Power and Energy, Vol. 221, No. 7, pp. 1025-1036. [4] Jung, Y. H., and Baek, J. H., 2008, “A Numerical Study on the Unsteady Flow Behavior and the Performance of an Automotive Sirocco Fan,” Journal of Mechanical Science and Technology, Vol. 22, Np. 10, pp. 1889-1895. [5] Behzadmehr, A., Mercadier, Y., and Galanis, N., 2006, “Sensitivity Analysis of Entrance Design Parameters of a Backward-Inclined Centrifugal Fan Using DOE Method and CFD Calculations,” ASME Journal of Fluids Engineering, Vol. 128, No. 3, pp. 446-453. [6] Myers, R. H., and Montgomery, D. C., 1995, Response surface methodology: process and product optimization using designed experiments, John Wiley & Sons Inc, New York, USA. [7] Jang, C. M., and Kim, K. Y., 2005, “Optimization of a Stator Blade Using Response Surface Method in a SingleStage Transonic Axial Compressor,” Proceedings of The Institution of Mechanical Engineers, Part A-Journal of Power and Energy, Vol.219, No.8, pp.595-603. [8] Chen, B., and Yuan, X., 2008, “Advanced Aerodynamic Optimization System for Turbomachinery,” ASME Journal of Turbomachinery, Vol. 130, No. 2, pp. 021005-1-12. [9] Keskin, A., Swoboda, M., Flassig, P. M., Dutta, A. K., and Bestle, D., 2008, “Accelerated Industrial Blade Design Based on Multi-Objective Optimization Using Surrogate Model Methodology,” ASME Turbo Expo 2008, Berlin, Germany, GT2008-50506. [10] Kim, K. Y., and Seo, S. J., 2004, “Shape Optimization of Forward-Curved-Blade Centrifugal Fan with NavierStokes Analysis,” ASME Journal of Fluids Engineering, Vol. 126, No. 5, pp. 735-742. [11] Marjavaara, B. D., Lundstrom, T. S., Goel, T., Mack, Y., and Shyy, W., 2007, “Hydraulic Turbine Diffuser Shape Optimization by Multiple Surrogate Model Approximations of Pareto Fronts,” ASME Journal of Fluids Engineering, Vol. 129, No. 9, pp. 1228-1240. [12] Yi, W., Huang, H., and Han, W., 2006, “Design Optimization of Transonic Compressor Rotor Using CFD and Genetic Algorithm,” ASME Turbo Expo 2006, Barcelona, Spain, GT2006-90155. [13] Deb, K., 2001, Multi-objective optimization using evolutionary algorithms, 1st ed., John Wiley & Sons Inc., Chichester, UK. [14] Goel, T., Vaidyanathan, R., Haftka, R. T., Shyy, W., Queipo, N. V., and Tucker, K., 2007, “Response Surface Approximation of Pareto Optimal Front in Multi-Objective

[15]

[16]

[17] [18]

[19]

[20]

[21]

[22]

[23]

[24]

[25]

[26]

[27]

[28]

9

Optimization,” Computer Methods in Applied Mechanics and Engineering, Vol. 196, No. 4-6, pp. 879-893. Kim, J. H., Kim, J. W., and Kim, K. Y., 2011, “Axial-Flow Ventilation Fan Design Through Multi-Objective Optimization to Enhance Aerodynamic Performance,” ASME Journal of Fluids Engineering, Vol. 133, No. 10, pp. 101101-1-12. Kim, J. H., Choi, K. J., and Kim, K. Y., 2011, “Optimization of a Transonic Axial Compressor Considering Interaction of Blade and Casing Treatment to Improve Operating Stability,” ASME Turbo Expo 2011, Vancouver, Canada, GT2011-45404. ANSYS CFX-11.0, 2006, ANSYS CFX-Solver Theory Guide, ANSYS Inc. Kim, J. H., Choi, J. H., and Kim, K. Y., 2010, “Surrogate Modeling for Optimization of a Centrifugal Compressor Impeller,” International Journal of Fluid Machinery and Systems, Vol. 3, No. 1, pp. 29-38. Menter, F. R., 1994, “Two-Equation Eddy-Viscosity Turbulence Models for Engineering Applications,” AIAA Journal, Vol. 32, No. 8, pp. 1598-1605. Bardina, J. E., Huang, P. G., and Coakley, T. J., 1997, “Turbulence Modeling Validation,” 28th AIAA Fluid Dynamics Conference, Snowmass Village, USA, AIAA1997-2121. Kim, J. H., Song, W. S., Lee, S., and Kim, K. Y., 2010, “A Study on Aerodynamic and Noise Characteristics of a Sirocco Fan for Residential Ventilation,” Journal of Fluid Machinery (in Korea), Vol. 13, No. 2, pp. 19-24. Adachi, T., and Sugita, N., 2004, “Study on the Performance of a Sirocco Fan (Flow Around the Runner Blade),” International Journal of Rotating Machinery, Vol. 10, No. 5, pp. 415-424. Han, S. Y., and Maeng, J. S., 2003, “Shape Optimization of Cut-off in a Multi-Blade Fan/Scroll System Using Neural Network,” International Journal of Heat and Mass Transfer, Vol. 46, No. 15, pp. 2833-2839. McKay, M. D., Beckman, R. J., and Conover, W. J., 2000, “A Comparison of Three Methods for Selecting Values of Input Variables in the Analysis of Output from a Computer Code,” Technometrics, Vol. 42, No. 1, pp. 55-61. MATLAB® 7.0, 2004, The Language of Technical Computing, Release 14, The Mathworks Inc., Natick, MA, USA. Giunta, A. A., 1997, “Aircraft Multidisciplinary Design Optimization Using Design of Experimental Theory and Response Surface Modeling Methods,” Ph.D. thesis, Department of Aerospace Engineering, Virginia Polytechnic Institute and State University, Blacksburg, VA. Queipo, N. V., Haftka, R. T., Shyy, W., Goel, T., Vaidyanathan, R., and Tucker, P. K., 2005, “SurrogateBased Analysis and Optimization,” Progress in Aerospace Science, Vol. 41, No. 1, pp. 1-28. JMP, 2005, The Statistical Discovery Software, Version 6.00, SAS Institute Inc., Cary, NC, USA.
