CFD and wave and current induced loads on offshore wind turbines

Transcription

1 CFD and wave and current induced loads on offshore wind turbines O. M. Faltinsen and M. Greco, CeSOS, NTNU SPH predictions by CNR-Insean

2 Relevant offshore wind turbines Relevant hydrodynamic problems are similar as for other offshore structures

3 Problems of Practical Interest Slowly-varying (Slow-drift) motions: Moored offshore platforms Resonance caused by second-order wave-body interaction effects Wave-drift and viscous damping

4 Transient ringing response of monotowers and TLPs Springing of TLP Resonance caused by second-order wave-body interaction effects in operating conditions Viscous damping Caused by 3rd and 4th-order wave-body interaction effects in survival conditions in deep water Ringing in shallow water due to plunging breaking waves Damping is secondary

5 Wetdeck slamming Wave run up and slamming Global and local effects

6 Accidental load: Breaking wave impact on platform column Experiments The resulting shear forces can cause structural failure of a concrete structure Steep and breaking deep-water waves. Scale 1:40 & 1:125 Correspond to extreme events in severe storms (H s =18m, T p =17s)

7 Breaking wave forces in shallow water Truss support structure for wind turbines on the ThorntonBank, Belgian Coast. Design by Reinertsen A/S. Truss structure in shallow water A. Chella

8 Mathieu-type instability Damping is important Time-dependent change of waterplane area shape and position Haslum Model test with scaled 202.5m draft Spar in 8.5m wave amplitude and 22.5s wave period Maximum response: Heave amplitude=20m Pitch amplitude=20 o Time-dependent difference between centers of gravity and buoyancy

9 Vortex Induced Motions (VIM) and Vibrations (VIV)

10 Marine installation operations using a floating vessel/jack-up Resonance and viscous damping Lin Li, T. Moan

11 Access to offshore wind turbines Minimize relative motion between ship and wind turbine support structure by automatic control Wave run up at the structure matters Use of high-speed Surface Effect Ship

12 Tools Experiments Analytical Solutions CFD Methods Approximate Computational Tools

13 Tools Complex models Complex geometries Visualization Global & local estimates Peric, CFD Workshop, Trondheim (2007) CFD Methods (Navier-Stokes solvers) - Reliability -Costs

14 Survey by International Towing Tank Conference (ITTC) (2011) What are the difficulties and limitations of CFD to achieve wider use and acceptance to marine hydrodynamic problems? The accuracy of CFD is a main problem (Industries and universities) Grid generation is a major problem (Model basins) CFD Methods (Navier-Stokes solvers)

15 Main Features of a CFD Solver 1. Modelling 2. Discretization 3. Reliability analysis

16 1) Modelling Define the problem

17 Define the Problem Air Body Water Space: P Time : t Computational Domain,

18 1) Modelling Define the problem Define the fluid/flow properties

19 Define Fluid/Flow Properties Examples: Compressible or incompressible fluid? Air? Water

20 Compressibility is secondary for water, but may be important for air

21 Define Fluid/Flow Properties Examples: Viscous or inviscid fluid? Water?

22 Define Fluid/Flow Properties Examples: Laminar or turbulent flow? t 0 t 1 t 2 t 3? t 0 t 1 t 2 t 3 Path of a fluid particle in laminar flow Mean path of a fluid particle in turbulent flow

23 Turbulence Turbulence involves low space-time scales. Proper modelling is still an open question. t 0 t 1 t 2 t 3 Mean path of a fluid particle

24 Viscous effects must be considered when flow separation occurs Occurs for structures with small cross-sectional area Water

25 Turbulent flow is secondary with flow separation from sharp corners Water

26 1) Modelling Define the problem Define the fluid/flow properties Identify the variables of the problem

27 Identify the variables of the Problem - Assume no air pockets - Air irrelevant for water evolution - Incompressible fluid Pressures Integrated hydrodynamic loads Motions Body Water u, p Surrounding fluid Fluid point

28 1) Modelling Define the problem Define the fluid/flow properties Identify the variables of the problem Find the equations

29 Governing Equations They are given by fluid conservation properties: 1) Conservation of fluid mass 2) Conservation of fluid momentum Navier-Stokes equations

30 Governing Equations Conservation of fluid mass: (Incompressible Fluid) Volume = constant Zero net flux B

31 Initial Conditions - Unsteady problem Body (u, p) at t = 0 Water

32 Boundary Conditions Body Wetted hull Free surface Water Control surface

33 1) Modelling Remark Modelling implies assumptions and they may be wrong.

34 2) Discretization h : discretization size 1 h Body i Water N

35 CFD Scenarios Basic Assumptions Potential Flow Methods BEM, HPC Field Equations Navier-Stokes Methods Field Methods Interfaces Grid Methods (FVM/FDM/FEM) Forms of the Governing Eqs Boundary-Fitted Grids Not Boundary- Fitted Grids Lagrangian Gridless Methods Moving particles (SPH, MPS) Fixed Nodes (RKPM) Immersed in the Comput. Domain body modeled numerically Grid methods: inside body problem, body capturing.. Gridless methods: body force/particles, ghost particles.. Interface capturing (VOF, LS, MAC, CIP) Body Gas-liquid Interface Domain Boundary body naturally tracked Interface tracking Eulerian Solution view point Eulerian Lagrangian

36 Potential Flow Methods Incompressible inviscid water with irrotational motion Equations with unknowns on the boundary Equations with unknowns in the domain Computational Domain Computational Domain Boundary Element Methods Field Methods

37 Potential Flow Methods Harmonic Polynomial Cell (HPC) method wave maker water surface Features: 2D sketch of the wave tank Water divided by cells Within each cell, the physical variables are expressed by harmonic polynomials; Benefits: Efficient (sparse matrix) Accurate (higher order) Easy to implement 3D wave-body interaction Yanlin Shao

38 Solution Point of View Lagrangian : The solution is found through moving grid/particles. u

39 Solution Point of View Eulerian : The solution is found at fixed locations.

40 Solution Point of View Eulerian : The solution is found at fixed locations. When Eulerian strategy is used for the field equations an additional technique is required to predict free-surface evolution

41 Strategies for Free-Surface Evolution Lagrangian strategy t 1 t 0 Eulerian strategy f j f i f j f i t 0 t 1

42 Free-surface Evolution VoF Method Hp: The quantity f is the fraction of water volume < 1 < 1 Free surface 1 1

43 Free-surface capturing methods (Level Set, VOF, Colour functions) Initial artificial change between water and air must remain small

44

45 CFD Scenarios Potential Flow Methods Hybrid Methods BEM, HPC Field Equations Navier-Stokes Methods Field Methods Interfaces Grid Methods (FVM/FDM/FEM) Fixed Grids Boundary-Fitted Grids Eulerian Lagrangian Gridless Methods Meshless FEM Particle Methods (SPH, MPS) Time evolution Higher-order methods Immersed in the Comput. Domain body modeled numerically Grid methods: inside body problem, body capturing.. Gridless methods: body force/particles, ghost particles.. Eulerian: interface capturing (VOF, LS, MAC, CIP) Body Air-Water Interface Explicit Semi-Implicit Implicit Domain Boundary body naturally tracked Lagrangian : interface tracking

46 Hybrid method Vortex shedding causes damping of resonant motions Coupled potential flow (linear) and local viscous solutions with flow separation with significant reduction of computational speed

47 Hybrid method Linear or weakly nonlinear potential-flow solver Efficiency Accuracy Navier-Stokes solver Breaking Fragmentation Air entrainment Viscosity

48 CFD and CeSOS BEM (Boundary Element Method) HPC (Harmonic Polynomial Cell) SPH (Smoothed Particle Hydrodynamics) FDM (Finite Difference Method)+Level Set CIP(Constrained Interpolation Profile): FDM with CIP used during advection step. Combined with colour functions OpenFoam (FVM+VOF) Maxwell-Boltzmann method and use of GPU Hybrid methods for locally separated flow (HPC,FVM) Hybrid methods for green water on deck

49 Reliability Analysis 1) Verification: Solving the equations right? The check of 1) means comparing with other benchmark numerical tests and analytical solutions using the same model. 2) Validation: Solving the right equations? The check of 2) means comparing with model tests. Experiments have also errors Error analysis is needed in both experiments and computations

50 Is this correct?

51 Frame Jan 2013 contour lines Numerical wave tank wave maker water surface Wave beach Numerical damping can cause too quick decay of waves on a coarse grid Numerical dispersion can cause too large error in the dispersion relation between wave length and wave speed

52 Wave focusing due to uneven seafloor Frame May d contour Water surface Y Z X Sea floor HPC results agree well with experiments

53 Verification and validation of ringing loads Comparisons with numerical (Ferrant) and experimental (Huseby&Grue) results Linear Force Analytical Experiment Present Ferrant Wave slope ka Sum Frequency Force Analytical Experiment Present Ferrant Wave slope ka Triple Frequency Force Analytical Experiment Present Ferrant Wave slope ka Quadruple Frequency Force Experiment Ferrant Present Wave 0.25 ka slope Time-domain nonlinear wave forces on a cylinder with exact free-surface conditions by the HPC method Yanlin Shao

54 Ringing of monotower in survival condition Steep local waves propagating on the two sides of the cylinder The waves start on the upstream hull side when there is a wave trough

55 Ringing of monotower in survival condition The two steep propagating waves will later collide Big splash

56 Water entry of circular cylinders Relevant for impact loads on vertical cylinders due to steep waves Wienke&Oumeraci

57 Non-viscous flow separation on a curved surface by BEM t = 0.315s t = 0.390s t = 0.500s t = 0.410s

58 Convergence studies for water entry of circular cylinder with commercial CFD code based on FVM and VOF The effect of the time step size dt on the slamming force coefficient C F / Experiments: C = 5.15 / ( Vt / R) V t / R S S 3 RV 2 C S Neither the force nor the force impulse converge CIP code developments at CeSOS: X. Zhu : Converged force impulse T. Vestbøstad: Satisfactory force Vt R

59 Verification and validation of numerical predictions of local slamming Temarel (2009) Viscous effects can be neglected and potential flow assumed Simple case Difficult benchmark tests: Small local deadrise angles Ventilation Experimental errors (bias and precision errors) Bias examples: 3D flow, oscillatory rig motions, elastic rope forces

60 p p 0.5 V 0 2 Difficult benchmark test Water entry of wedge with deadrise angle 7.5 degrees and constant velocity V Slamming predictions by commercial code (FVM+VOF) Commercial code Human errors? Similarity solution BEM Composite Wagner solution Polynomial fit of average commercial code results z Vt

61 Summary There exists a broad variety of CFD methods Computational time is of concern for wave and current induced loads Physical simplifications by potential flow and hybrid methods reduce computational time Reliability of CFD results All physical problems such as turbulence is not fully understood Convergence Verification and validation

62 Some Useful CFD References Brebbia C. A. many books on Boundary Elements, WIT Press. Ferziger J.H., Peric M. Computational methods for fluid dynamics, Springer. Faltinsen, O. M., Timokha, A. (2009) Sloshing, Chapter 10, Cambridge University Press. Hirt C.W., Nichols B.D. (1988)"VolumeofFluid(VOF)MethodfortheDynamicsof Free Boundaries," J. of Comput. Phys., 39, 201, Johnson C. Numerical Solution of Partial Differential Equations by the Finite Element Method, Cambridge. Monaghan J.J. (1994) "Simulating free surface flows with SPH", J. Comp. Physics, 110, Sussman M., Smereka P., Osher S. (1994) A level set method for computing solutions to imcompressible two-phase flow, J. Comput. Phys, 119, Unverdi S.O., Tryggvason G. (1992) "A Front Tracking Method for Viscous Incompressible Flows, J. of Comput.Phys., 100, Yabe T., Ogata Y., Takizawa K., Kawai T., Segawa A., Sakurai K. (2002) The next generation CIP as a conservative semi-lagrangian solver for solid, liquid, and gas, J. Comput. Appl. Math., 149,