Aeroacoustic Optimization of an Axial Fan with Variable Blade - - PowerPoint PPT Presentation

β–Ά
aeroacoustic optimization of an axial fan with
SMART_READER_LITE
LIVE PREVIEW

Aeroacoustic Optimization of an Axial Fan with Variable Blade - - PowerPoint PPT Presentation

Dec 06, 2022 261 likes β€’811 views

Aeroacoustic Optimization of an Axial Fan with Variable Blade Loading Michae ael l Stadler er Michae ael l B. Schmitz Ninsight ebm-papst St. Georgen Graz, Austria St. Georgen, Germany Wolfg fgan ang Laufer er Peter Ragg ebm-papst

slide-1
SLIDE 1

Michae ael l Stadler er Ninsight Graz, Austria Michae ael l B. Schmitz ebm-papst St. Georgen

  • St. Georgen, Germany

Wolfg fgan ang Laufer er ebm-papst St. Georgen

  • St. Georgen, Germany

Peter Ragg ebm-papst St. Georgen

  • St. Georgen, Germany

Aeroacoustic Optimization of an Axial Fan with Variable Blade Loading

ninsight

slide-2
SLIDE 2
  • minimize selected bands of the acoustic spectrum
  • maximize aerodynamic efficiency

Objectives

slide-3
SLIDE 3

𝑠𝑑𝑣 𝑠𝑑𝑣

shroud hub

This is performed for a fan with varying blade geometry (i.e. each blade is loaded differently).

Objectives

  • minimize selected bands of the acoustic spectrum
  • maximize aerodynamic efficiency
slide-4
SLIDE 4

Traditional development cycle for small axial fans

  • 1. Design the blade geometry
  • 2. Assess aerodynamic performance via CFD
  • 3. Assess aeroacoustic performance via physical prototypes
slide-5
SLIDE 5

Sh Shortco tcoming mings: s:

  • Spurious noise of physical prototypes
  • imperfect rotor dynamics
  • eccentricity of the fan
  • Vibration of the bearing system
  • Noise due to secondary flow through the cooling channels
  • Low geometrical resolution of rapid prototyping

Traditional development cycle for small axial fans

  • 1. Design the blade geometry
  • 2. Assess aerodynamic performance via CFD
  • 3. Assess aeroacoustic performance via physical prototypes
slide-6
SLIDE 6

Sh Shortco tcoming mings: s:

  • Spurious noise of physical prototypes
  • imperfect rotor dynamics
  • eccentricity of the fan
  • Vibration of the bearing system
  • Noise due to secondary flow through the cooling channels
  • Low geometrical resolution of rapid prototyping

Traditional development cycle for small axial fans

T

  • avoid this tedious optimization procedure

augmentation of design process with aeroacoustic simulation

  • 1. Design the blade geometry
  • 2. Assess aerodynamic performance via CFD
  • 3. Assess aeroacoustic performance via physical prototypes
slide-7
SLIDE 7

Outline

1

Variable Blade Loading 2 Winglet

3

Turbulator

4

Numerical Model

5

Experimental Setup 6 Results

slide-8
SLIDE 8

Outline

1

Variable Blade Loading 2 Winglet

3

Turbulator

4

Numerical Model

5

Experimental Setup 6 Results

slide-9
SLIDE 9

Variable Blade Loading

𝑠 β‹… 𝑑𝑣 = 𝑂 2𝜌 𝑠 β‹…

2𝜌 𝑂

π‘‘π‘£π‘’πœ„

Blade parameterization by averaged swirl at trailing edge for hub and shroud:

𝑠𝑑𝑣 x

slide-10
SLIDE 10

Variable Blade Loading

𝑠 β‹… 𝑑𝑣 = 𝑂 2𝜌 𝑠 β‹…

2𝜌 𝑂

π‘‘π‘£π‘’πœ„

βˆ†π‘ž = 2𝜌 𝑂 𝜍π‘₯π‘›π‘π‘š πœ–(𝑠 β‹… 𝑑𝑣) πœ–π‘›

From this, the blade loading may be obtained:

𝑠𝑑𝑣 x

Blade parameterization by averaged swirl at trailing edge for hub and shroud:

slide-11
SLIDE 11

Variable Blade Loading

PARAMETERS FOR EVOLUTIONARY OPTIMIZATION 𝑠𝑑𝑣

Averaged swirl at trailing edge for hub and shroud

𝑠𝑑𝑣 𝑠𝑑𝑣

shroud hub

slide-12
SLIDE 12

Variable Blade Loading

𝑠𝑑𝑣 𝑠𝑑𝑣

shroud hub

PARAMETERS FOR EVOLUTIONARY OPTIMIZATION 𝑠𝑑𝑣

Averaged swirl at trailing edge for hub and shroud

slide-13
SLIDE 13

Variable Blade Loading

𝑠𝑑𝑣 𝑠𝑑𝑣

shroud hub

is specified independently for 7 individual blades at hub and shroud 14 parameters

𝑠𝑑𝑣 PARAMETERS FOR EVOLUTIONARY OPTIMIZATION 𝑠𝑑𝑣

Averaged swirl at trailing edge for hub and shroud

slide-14
SLIDE 14

Outline

1

Variable Blade Loading 2 Winglet

3

Turbulator

4

Numerical Model

5

Experimental Setup 6 Results

slide-15
SLIDE 15

Winglet Parameterization

PURPOSE OF THE WINGLET: control the tip vortex

  • Vortex may collide with the adjacent blade β†’ lead to vibration and associated noise production
  • Vortex can restrict the flow β†’ decrease the aerodynamic performance
slide-16
SLIDE 16

Winglet Parameterization

1. Cut the blade with a cylinder of diameter 0.8 D β†’ cross section l

slide-17
SLIDE 17

Winglet Parameterization

1. Cut the blade with a cylinder of diameter 0.8 D β†’ cross section l

slide-18
SLIDE 18

Winglet Parameterization

1. Cut the blade with a cylinder of diameter 0.8 D β†’ cross section l 2. Project l orthogonal to cylindrical surface of diameter D β†’ cross section m

slide-19
SLIDE 19

Winglet Parameterization

1. Cut the blade with a cylinder of diameter 0.8 D β†’ cross section l 2. Project l orthogonal to cylindrical surface of diameter D β†’ cross section m

  • 3. F1(q) : Rotation of m around x β†’ m1
slide-20
SLIDE 20

Winglet Parameterization

1. Cut the blade with a cylinder of diameter 0.8 D β†’ cross section l 2. Project l orthogonal to cylindrical surface of diameter D β†’ cross section m

  • 3. F1(q) : Rotation of m around x β†’ m1
  • 4. F2(z) : Rotation of m1 around c β†’ m2
slide-21
SLIDE 21

Winglet Parameterization

1. Cut the blade with a cylinder of diameter 0.8 D β†’ cross section l 2. Project l orthogonal to cylindrical surface of diameter D β†’ cross section m

  • 3. F1(q) : Rotation of m around x β†’ m1
  • 4. F2(z) : Rotation of m1 around c β†’ m2
  • 5. F3(s) : Translation of m2 along c β†’ m3
slide-22
SLIDE 22

Winglet Parameterization

Leading edge Blade bends towards the suction side Trailing edge Blade bends towards the pressure side Closure of winglet surface by multisection extrusion between {m,l} β†’ winglet surface k

slide-23
SLIDE 23

Winglet Parameterization

Trailing edge Blade bends towards the pressure side β€žConic Winglet Designβ€œ Radius of curvature varies along the winglet spine Leading edge Blade bends towards the suction side Closure of winglet surface by multisection extrusion between {m,l} β†’ winglet surface k

slide-24
SLIDE 24

Outline

1

Variable Blade Loading 2 Winglet

3

Turbulator

4

Numerical Model

5

Experimental Setup 6 Results

slide-25
SLIDE 25

Turbulator Parameterization

Adverse effects of flow separation in axial fans:

  • Increased generation of noise
  • Reduced cross sectional area of the flow channel β†’ degradation of performance

PURPOSE OF THE TURBULATOR: avoid flow separation along blade

slide-26
SLIDE 26

Turbulator Parameterization

General usage of turbulators:

  • Turn laminar flow into turbulent flow (near the leading edge)
  • Increase the energy of an already turbulent boundary layer β†’ move the point of flow

separation further downstream (e.g. ailerons of commercial airliners)

PURPOSE OF THE TURBULATOR: avoid flow separation along blade

Adverse effects of flow separation in axial fans:

  • Increased generation of noise
  • Reduced cross sectional area of the flow channel β†’ degradation of performance
slide-27
SLIDE 27

Turbulator Parameterization

DEFINITION OF THE TURBULATOR GEOMETRY

Turbulator spine = parallel curve to the line of flow separation (offset g ) Turbulator cross section = simple step (height t )

g t

slide-28
SLIDE 28

Turbulator Parameterization

DETERMINATION OF THE LINE OF FLOW SEPARATION

  • either by integral convolution of the velocity field, or
  • by showing the streamlines for the elements adjacent to the blade
slide-29
SLIDE 29

Turbulator Parameterization

TURBULATOR AND RAPID PROTOTYPING

Machine: EOS 390 (selective laser sintering of polyamide) For the turbulator to work properly, sharp edges are required Note: The final product is created by injection die molding (which can easily represent sharp edges). Fine details of turbulator not sufficiently resolved β†’ augmentation

  • f optimization process with numerical

tools necessary!

Rapid prototyping specimen

slide-30
SLIDE 30

Outline

1

Variable Blade Loading 2 Winglet

3

Turbulator

4

Numerical Model

5

Experimental Setup 6 Results

slide-31
SLIDE 31

Numerical Model

STAR-CCM+ 8.02 RANS solver k-e-turbulence model All y+ wall model Discretization of 1/7th of the model (rotational symmetry) STAR-CCM+ 8.02 LES solver WALE subgrid scale Discretization of the complete model

Aerodynamic analysis Aeroacoustic analysis

slide-32
SLIDE 32

Numerical Model

Evolutionary Optimization

  • Parametric geometry generation in the 3D-Modeler of STAR-CCM+
  • Restore an identical operating point throughout the optimization (change rpm accordingly)
  • Objective function evaluation
  • Differential Evolution via Multi-Objective Genetic Algorithm NSGA-II (Non Dominated Sorting GA)
  • Assisted by a metamodel to reduce the number of objective function evaluations
  • Pareto front shows the best possible compromise between noise and efficiency over the selected

design space

Restore

  • perating

point Efficiency Noise NSGA-II Discretization

Parametric Geometry Generator

Metamodel

Objective function evaluation

slide-33
SLIDE 33

Numerical Model

Evolutionary Optimization

  • Parametric geometry generation in the 3D-Modeler of Star-CCM+
  • Restore an identical operating point throughout the optimization (change rpm accordingly)
  • Objective function evaluation
  • Differential Evolution via Multi-Objective Genetic Algorithm NSGA-II (Non Dominated Sorting GA)
  • Assisted by a metamodel to reduce the number of objective function evaluations
  • Pareto front shows the best possible compromise between noise and efficiency over the selected

design space

Restore

  • perating

point Efficiency Noise NSGA-II Discretization Metamodel

STAR-CCM+ + Plugin in

Parametric Geometry Generator

slide-34
SLIDE 34

Numerical Model

Evolutionary Optimization

  • Parametric geometry generation in the 3D-Modeler of STAR-CCM+
  • Restore an identical operating point throughout the optimization (change rpm accordingly)
  • Objective function evaluation
  • Differential Evolution via Multi-Objective Genetic Algorithm NSGA-II (Non Dominated Sorting GA)
  • Assisted by a metamodel to reduce the number of objective function evaluations
  • Pareto front shows the best possible compromise between noise and efficiency over the selected

design space

Restore

  • perating

point Efficiency Noise NSGA-II Discretization

Parametric Geometry Generator

Metamodel

  • Polyhedral discretization
  • Prismatic expansion layer to

resolve boundary layer

slide-35
SLIDE 35

Numerical Model

Evolutionary Optimization

  • Parametric geometry generation in the 3D-Modeler of STAR-CCM+
  • Restore an identical operating point throughout the optimization (change rpm accordingly)
  • Objective function evaluation
  • Differential Evolution via Multi-Objective Genetic Algorithm NSGA-II (Non Dominated Sorting GA)
  • Assisted by a metamodel to reduce the number of objective function evaluations
  • Pareto front shows the best possible compromise between noise and efficiency over the selected

design space

Restore

  • perating

point Efficiency Noise NSGA-II Discretization

Parametric Geometry Generator Evaluation of objective functions is numerically expensive οƒ  Metamodel necessary!

  • Polynomial Response Surface Model (PRSM)
  • Artificial Neural Network (ANN)
  • Radial basis function network (RBF)
  • Kriging

Due to nonlinearity of the problem, PRSM was unsuitable. Here we have chosen RBF.

Metamodel

slide-36
SLIDE 36

Numerical Model

Evolutionary Optimization

  • Parametric geometry generation in the 3D-Modeler of Star-CCM+
  • Restore an identical operating point throughout the optimization (change rpm accordingly)
  • Objective function evaluation
  • Differential Evolution via Multi-Objective Genetic Algorithm NSGA-II (Non Dominated Sorting GA)
  • Assisted by a metamodel to reduce the number of objective function evaluations
  • Pareto front shows the best possible compromise between noise and efficiency over the selected

design space

Restore

  • perating

point Efficiency Noise Discretization

Parametric Geometry Generator

Metamodel

STAR-CCM+ + Java va Plugi gin

NSGA-II

slide-37
SLIDE 37

Numerical Model

OBJECTIVE FUNCTIONS

Aerodynamic Efficiency Soundpressure 𝑔

1 = max Ξ”π‘žπ‘Š

π‘πœ• 𝑔

2 = min 10 log 10 π‘€π‘ž π‘—βˆ’ 𝑀𝐡 𝑗 𝑠 𝑗=1

Sound pressure of frequency band with index i A-weighting associated with frequency band of index i

Here we choose to minimize the tonal noise at BPF1=840 Hz and BPF2=1680 Hz

Evolutionary Optimization

This allows to selectively minimize individual bands of the acoustic spectrum.

slide-38
SLIDE 38

Outline

1

Variable Blade Loading 2 Winglet

3

Turbulator

4

Numerical Model

5

Experimental Setup 6 Results

slide-39
SLIDE 39

Experimental Setup

Physical prototype

slide-40
SLIDE 40

Experimental Setup

Physical prototype Acoustic test rig Aerodynamic test rig

slide-41
SLIDE 41

Outline

1

Variable Blade Loading 2 Winglet

3

Turbulator

4

Numerical Model

5

Experimental Setup 6 Results

slide-42
SLIDE 42

Results

Downstream flow field Q-Criterion = 2e+5 Time step: 5e-5 s n=7200 min-1

slide-43
SLIDE 43

Results

Pareto front

slide-44
SLIDE 44

Results

Pareto front Averaged swirl velocity at trailing edge

SHROUD HUB

𝑠𝑑𝑣

slide-45
SLIDE 45

Results

Pareto front Averaged swirl velocity at trailing edge

SHROUD HUB

𝑠𝑑𝑣

slide-46
SLIDE 46

Results

Pareto front Averaged swirl velocity at trailing edge

SHROUD HUB

𝑠𝑑𝑣

slide-47
SLIDE 47

Results

Pareto front Averaged swirl velocity at trailing edge

SHROUD HUB

𝑠𝑑𝑣

slide-48
SLIDE 48

Results

Pareto front Conlusions

  • For high efficiency, all blades are loaded equally
  • For low tonal noise, all blades are loaded differently

Averaged swirl velocity at trailing edge

SHROUD HUB

𝑠𝑑𝑣

slide-49
SLIDE 49

Results

Typical sound pressure frequency spectrum for a sensor in the Large Eddy Simulation

slide-50
SLIDE 50

Results

Comparison between

BPF1= 840 Hz BPF2= 1680 Hz

(a) physical test β–  (b) simulation β– 

slide-51
SLIDE 51

Results

Comparison between

Rotational frequency RF = 7200 rpm / 60 = 120 Hz Source: ce: Eccentricity of the physical specimen (absent ent in numerical simulation) Represents an advantage, since the spectrum is not polluted by spurious noise BPF1= 840 Hz BPF2= 1680 Hz

(a) physical test β–  (b) simulation β– 

slide-52
SLIDE 52

Results

Comparison between blade geometry (physical tests) (a) identicalβ–  (b) varying β– 

slide-53
SLIDE 53

Summary

1

Variable Blade Loading 2 Winglet

3

Turbulator

4

Numerical Model

5

Experimental Setup 6 Results

ninsight

Mich chael ael Stad adler ler Ninsight Graz, Austria Mich chael ael B. Schmit mitz, z, Wolfga lfgang Lau aufe fer, , Peter er Rag agg ebm-papst

  • St. Georgen, Germany