Centrifugal Pump Research in Twente

1
Centrifugal Pump Research in Twente

• dr.ir. Niels P. KruytEngineering Fluid
  Dynamics, Department of Mechanical Engineering,
  University of Twente

Holland Pomp Groep, Hengelo, 24/04/07
2
Group Engineering Fluid Dynamics(prof. Harry
Hoeijmakers)

• Rotating-flow machines
• centrifugal pumps
• wind turbines
• Aero-acoustics
• Aerodynamics, gas dynamics and flows with
  phase-transitions
• Thin-film flows
• Fluid-structure interaction and aero-elasticity
• Bio-physical flows

3
Overview

• Why is fluid dynamics important for centrifugal
  pump design?
• What is Computational Fluid Dynamics (CFD)?
• Centrifugal pump research in Twente
• inverse-design methods
• optimization methods

4
Characteristics of centrifugal pumps
5
Flow field performance
6
Basics of pump design/analysis

• One-dimensional flow model
• Euler pump relation

• Slip factor is empirical
• Hydraulic efficiency is empirical

7
Need for knowledge of flow field

• Performance characteristics from first principles
• Insight into flow provides pointers for improved
  design
• Flow field
• pressure
• velocity
• loss

8
What is Computational Fluid Dynamics?

• Determination of flow
• Analytical impossible
• Experiments expensive
• Numerical CFD(computer test-rig)

9
Benefits of CFD for pump design

• Improved designs
• More reliable design methods
• Cheaper design process

10
Components of CFD

• Model formulation
• geometry
• flow model
• boundary conditions
• Numerical solution
• grid/mesh generation
• discretisation of governing equations
• solution of discretised equations
• Interpretation

11
Selection of modelled geometry

• Single channel of impeller
• Full pump impeller volute/diffusor
• steady
• unsteady
• Leakage-flow region
• Piping system / pump intake
• Single stage vs. multi-stage

12
Turbulent flow
13
Flow models

• Stream-surface methods
• Potential-flow model
• Euler-flow model
• RANS-based models
• Large-eddy simulations (LES)
• Direct Navier-Stokes simulations (DNS)

Increasing complexity
14
Sources of errors in CFD-predictions

• Modelling errors
• Geometrical uncertainties
• Limited validity of adopted flow model
• Uncertain boundary conditions
• Numerical errors
• Discretisation error due to finite grid-size
• Lack of convergence in iterative solution process
• Insufficient mesh/grid quality
• User/programmer errors

15
Choice of flow model
Around design point
16
Intermezzo

• Importance of fluid dynamics for centrifugal pump
  design
• Components of CFD
• Research in Twente
• inverse-design methods (with R.W. Westra)
• optimization methods (with R.W. Westra)

17
Performance prediction with potential-flow model
van Esch Kruyt (2001)
18
Inverse-design method

• Previous method
• Specify geometry compute performance
• Inverse-design method
• Specify performance compute geometry

19
Objective inverse-design method

• Specify
• design conditions (Q, H, O)
• meridional plane
• blade loading
• Obtain
• blade angles

20
Meridional plane

• Hub and shroud contours specified
• Positions of leading and trailing edge specified
• Number of blades also specified

21
Blade loading

• Euler equation (1D)
• Mean-swirl distribution

22
Example of inverse-design method
Westra et al. (2005)
23
Review inverse-design method

• Obtain desired head
• Obtain correct incidence
• Fairly rapid method
• Very sensitive to prescribed mean-swirl
• Single duty
• No blade thickness
• Constraints to design can not be incorporated

24
Optimization methods

• Improve existing pump design
• variations in geometry
• select best geometry
• Also suited to obtaining new design

25
Components of optimization methods

• Cost function
• objectives
• multi-point
• Parametrization of the geometry
• Optimization algorithm
• Differential Evolution

26
Cost function

• Multi-objective
• head
• cavitation characteristics
• losses
• ¼
• Multi-point
• evaluated at various flowrates, qi
• different weights are employed

27
Cost function head (fH)

• A target head is set Hd
• Geometries that do not meet the target head get a
  penalty to the cost function

28
Cost function NPSHi (fk)

• Cavitation (inception) characteristics
• Cavitation number

29
Cost function losses (f?)

• Power loss estimated from inviscid velocities
• Hydraulic power
• Loss coefficient ?

30
Parametrization
Blade angles
Shroud contour

• 13 parameters
• blade angles (10)
• meridional shroud curve (2)
• number of blades (1)

31
Weight factors for cost function
32
Example of use optimization method

• Selection of impeller for test facility
• Starting point impeller with 7 blades
• New impeller must be for experimental facility
  few blades (4-6)

33
Convergence historyDifferential Evolution
Foriginal 1.69
Foptimum 1.35
30.0 bad meshes
34
Optimized design

• Original (blue) and optimized (red) blades and
  meridional shroud curve

35
Optimized impeller
36
Optimized design head
37
Optimized design losses
38
Optimized design NPSHinc
39
Review optimization methods

• Improve existing designs and obtain new designs
• Multi-point method
• Constraints can be incorporated easily
• Easily extended
• additional objectives
• additional geometrical parameters
• Long computing time

40
Test facility(under construction)

• Validation of developed methods
• Measurement of velocities
• Particle Image Velocimetry (PIV)
• Pressures

41
Conclusions

• Fluid dynamics is important to centrifugal pump
  design
• CFD is a powerful tool
• many pitfalls present
• CFD does not replace a smart designer
• Recent developments in Twente
• inverse-design methods
• optimization methods

42
Future work

• Improvement of inverse-design method
• other blade loadings
• incorporate blade thickness
• Improvement of optimization method
• cavitation model
• increase computational speed
• Validation using test facility

43
Questions

• Thank you for your attention!

• Niels KruytEngineering Fluid Dynamics,
  Department of Mechanical Engineering,
  University of Twente, P.O. Box 217, 7500 AE
  EnschedeTel 053-489 2528n.p.kruyt_at_utwente.nlww
  w.ts.ctw.utwente.nl/kruyt/