Progress in Developing Hybrid Models

Transcription

1 Progress in Developing Hybrid Models Shiqiang Yan, Jinghua Wang & Qingwei Ma City, University of London A Zonal CFD Approach for Fully Nonlinear Simulation of Two vessels in Launch and Recovery Operation SUTGEF Meeting, 25 th Jan 2017 L&R Project Meeting

2 Challenges of Modelling wave-structure interaction in high sea state Extreme Wave Field Modelling Low resolution (~10s km) of forecasting/hindcasting date Wave with low probability of occurrence (e.g. 10,000years) Typical large-scale (~km),long duration (~3 hrs) modelling

3 Challenges of Modelling wave-structure interaction in high sea state Extreme Wave Field Modelling Low resolution (~10s km) of forecasting/hindcasting date Wave with low probability of occurrence (e.g. 10,000years) Typical large-scale (~km),long duration (~3 hrs) modelling Structure responses & nearfield wave condition Significant turbulent effects Breaking wave impact, slamming Aeration Large structure motion, significant hydro-elasticity, multi-body interaction Progress in Developing Hybrid Models Typical small to micro-scale (~m to ~mm) modelling Simulation is limited to small domain due to low efficiency of C-NS

4 Challenges of Modelling wave-structure interaction in high sea state Extreme Wave Field Modelling Low resolution (~10s km) of forecasting/hindcasting date Wave with low probability of occurrence (e.g. 10,000years) Typical large-scale (~km),long duration (~3 hrs) modelling Structure responses & nearfield wave condition Significant turbulent effects Breaking wave impact, slamming Aeration Large structure motion, significant hydro-elasticity, multi-body interaction Typical small to micro-scale (~m to ~mm) modelling Simulation is limited to small domain Progress in Developing Hybrid Models A multi-scale problem, May be accurately predicted using a single model, e.g. C-NS, with high computing demand, which is practically unacceptable Requires a robust hybrid solutions combining multiple models with different simplifications, e.g. FNPT,I- NS and C-NS

5 Zonal Coupling strategy in the Multi-Model Multi- Scale Hybrid models MLPG-R Progress in Developing Hybrid Models NLSE-ESBI-QSBI Robust QALE-FEM to cover large domain away from floating bodies; Strong couple between the FNPT model and the NS model using various technics; Self-adaptive wave-maker is applied at outer boundaries of the FNPT domain for wave generation/absorption How to deal with the wave breaking in the FNPT domain? Coupling with one-phase meshless methods; What is the target wave to be generated in the wave paddle? Coupling with other wave models;

6 MLPG-R NLSE-ESBI-QSBI Hybrid Model for large-scale and long-time simulation FNPT based QALE-FEM is efficient but not sufficient Purpose is to reproduce the wave condition based on low-resolution wave forecasting/hindcasting data To be coupled with QALE- FEM by specifying the wave condition at its wave paddle

7 Hybrid Model for large-scale and long-time simulation Combine NLSE, ESBI and QSBI Communication scheme: Transform between envelope with free surface and velocity potential in order to guarantee the data exchange between ENLSE-5F, QSBI with ESBI Nonlinearity monitoring scheme: Enable the hybrid model to switch between ENLSE-5F, QSBI and ESBI automatically, according to the wave nonlinearities Nonlinearity becomes stronger ENLSE- 5F QSBI ESBI Nonlinearity becomes weaker

8 Hybrid Model for large-scale and long-time simulation (NLSE-ESBI-QSBI) Validation (Docruzet, 2007) Spreading spectrum S k, θ = S k G θ Where S k is the JONSWAP spectrum, and the spread function G θ is expressed as G θ = 2cos2 θ π Domain: 42 x 42 peak wave lengths ( 40 km 2 ) Resolution: 1024 x 512 Duration: 250 peak periods ( 42 min) CPU: Intel(R) Xeon(R) E5620@2.4GHz Total CPU time: 11.9 hours Docruzet, G., Modélisation des processus non-linéaires de génération et de propagation d'états de mer par une approche spectrale (Doctoral dissertation). Ecole Centrale de Nantes: Université de Nantes. Probability distribution of the free surface elevation at T/T 0 = 200. Gaussian distribution; --- Results in (Docruzet, 2007); Results by using hybrid model

9 Hybrid Model for large-scale and long-time simulation (NLSE-ESBI-QSBI) Fully nonlinear simulation of 3-hour sea state k 0 H s = 0.05 & m = 25 based on the Wallops spectrum embedded with a rogue wave of 2H s, domain size 125L0, need 969s 16min to finish the simulation (single thread computation on Intel Xeon E GHz )

10 Hybrid Model for large-scale and long-time simulation (NLSE-ESBI-QSBI) 3D simulations of Rogue (extreme) waves embedded in random sea Domain: 32 x 32 peak wave lengths ( 23 km 2 ) Resolution: 1024 x 1024 Duration: 100 peak periods ( 17 min) CPU: Intel(R) Xeon(R) E5620@2.4GHz (8 cores) Total elapsed time: 3.2 hours

11 MLPG-R NLSE-ESBI-QSBI Self-adaptive wave paddle to generate the incoming wave and absorb the reflected wave Why not directly couple the hybrid wave model (NLSE- ESBI-QSBI) with QALE-FEM? Requires periodic boundary condition Only directly predicts the free surface data Only cover the entire domain, but feed the wave elevation/velocity at the position where the wave paddle located NLSE-ESBI-QSBI cover a larger domain including FNPT, I-NS and C-NS domain NLSE-ESBI-QSBI provides target waves at the wave paddles in the FNPT domain->predicting the paddle motion using self-corrective wavemaker theory- >reproduce the target wave in the FNPT domain The disturbance on wave by the floating bodies is expected to be absorbed by wave paddle

12 Self-adaptive wave absorber for highly nonlinear water waves Challenges High sea state: highly nonlinear incident waves Large motion of vessels, breaking, aeration, hydro-elasticity: highly nonlinear radiation/diffraction waves due to the vessels Wave spectrum is developing and may be changed rapidly due to wavewave interaction or wind-wave interaction in high sea state Existing theory Using wavemaker motion to absorb unexpected waves Based on linear wavemaker theory Efficiency is low for highly nonlinear wave, especially shallow waves Wave spectrum shall be specified at prior

13 Self-adaptive wave absorber for highly nonlinear water waves Main Idea for improvement: track the local wave frequency Does not require any pre-specification to the wavemaker, e.g. the wave spectrum Frequency tracking using extended Kalman filter

14 Self-adaptive wave absorber for highly nonlinear water waves Works for both monochromic wave and steep wave group k 0 a 0 =0.15,k 0 d = 0.6 PM spectrum, ω 0 4.3s 1, 210 components random amplitude approach k 0 H s 0.2,k 0 η max 0.13 Water depth: 2.93m To be extended to 3D wave basin

15 MLPG-R NLSE-ESBI-QSBI Two-way coupling between QALE-FEM & onephase MLPG-R Support particles QALE-FEM provides velocity or pressure Boundary of QALE-FEM domain Requires Pressure or velocity from MLPG-R Interpolated from pressure or velocity at ISPH particles surrounding it Boundary of MLPG-R domain QALE-FEM provides velocity or pressure Overlap(translational) zone QALE-FEM requires velocity at free surface from MLPG-R A weighting function is used for smooth transition Similar work may need to do in the MLPG-R

16 Two-way coupling between QALE-FEM & onephase MLPG-R Validation Progress in Developing Hybrid Models 2 nd order piston wavemaker Translational zone Self-adapted wave absorber QALE-FEM domain 1 MLPG-R domain QALE-FEM domain 2 Wave tank: 110 m long, 2.2m wide and 2m deep. 0.7m mean water depth; Wave generator: piston wave paddle based on 2 nd order wavemaker theory; Wave gauges and pressure sensors Unidirectional focusing waves ( 32 components; Frequency range: 0.34 Hz to 1.02 Hz with different wave height)

17 (m) (m) Two-way coupling between QALE-FEM & onephase MLPG-R Validation Progress in Developing Hybrid Models Good agreement between the hybrid modelling results and the experimental data (a) x=4.975 Exp QALE-FEM Exp Velocity Feed Pressure Feed time(s) time(s)

18 Two-way coupling between QALE-FEM & onephase MLPG-R Hybrid model for breaking jet treatment Progress in Developing Hybrid Models When the local wave slope is higher than threshold value, a local area near the crest is defined (bounded by red) The nodes in this area used by the QALE-FEM are converted to particles, saving CPU time on local particle generation Coupling the QALE-FEM with MLPG_R using the concept of overset grid

19 Two-way coupling between QALE-FEM & onephase MLPG-R Hybrid model for breaking jet treatment Progress in Developing Hybrid Models Good agreement with experimental time history of the wave elevation

20 MLPG-R NLSE-ESBI-QSBI Coupling between QALE-FEM & two-phase incompressible NS solver One way coupling INS domain Damping zone QALE-FEM solution (velocity, pressure and wave elevation) Weighted summation of the INS solution and the QALE-FEM solution Boundary of INS QALE-FEM provides velocity, pressure and wave elevation Require large damping zone to absorb the reflection from the structures OpenFOAM is used for providing the INS solution

21 One-way coupling between QALE-FEM & twophase incompressible NS solver Validation: Fixed cylinder in extreme waves Physical Models and Validations Model test at Franzius-Institute, Leibniz Universitat Hannover Wave tank: 110 m long, 2.2m wide and 2m deep. 0.7m mean water depth; Wave generator: piston wave paddle based on 2 nd order wavemaker theory; Cylinder: 0.22m diameter; Wave gauges and pressure sensors Progress in Developing Hybrid Models Unidirectional focusing waves: 32 components; Frequency range: 0.34 Hz to 1.02 Hz with different wave height. Cylinder Moving towards the wave paddle with speed ranging from 0.25 to 0.75m/s Ensure the cylinder reaches the focusing point at focusing time

22 OpenFOAM domain: length 6D centred at the cylinder cylinder surface facing incident wave (f l = 0.34 Hz to f u = 1.02 Hz. G a = 0.003), QALE-FEM OpenFOAM only QALE-FEM with OpenFOAM CPU time Desktop PC 4 cores ~4 hours Newmann cluster 192 cores ~1 week(mmu) Intel(R) Xeon(R) E5620@2.4GHz (8 cores): ~10 hours

23 One-way coupling between QALE-FEM & twophase incompressible NS solver Validation Fixed cylinder in extreme waves Progress in Developing Hybrid Models Exp QALE-FEM/OpenFOAM 30 pressure(100pa) time(s) Time history of pressure at different locations on the cylinder surface facing incident wave (f l = 0.34 Hz to f u = 1.02 Hz. G a = 0.002, cylinder moves towards the wave paddle( 0.25m/s) Acceptable agreement between the hybrid modelling results and the experimental data of the pressure recorded on the cylinder surface

24 Two-way coupling between QALE-FEM & twophase incompressible NS solver y(m) x(m) Physical Models and Validations Whole domain: FNPT model Near the floating bodies: a local mesh is generated and moves together with the floating bodies Two-phase incompressible NS solver is used in the local area Overlap area: a robust quadric interpolation scheme is developed and used

25 Two-way coupling between QALE-FEM & twophase incompressible NS solver Validation: moving cylinder in extreme waves Physical Models and Validations cylinder surface facing incident wave (f l = 0.34 Hz to f u = 1.02 Hz. G a = 0.002, cylinder moves towards the wave paddle with speed of 0.25m/s

26 Two-way coupling between QALE-FEM & twophase incompressible NS solver Validation: moving cylinder in extreme waves (m) 0.05 Exp Physical Models and Validations 0 (a) x=4.975 QALE-FEM (m) time(s) 0 (b) x= time(s) Exp QALE-FEM Wave elevation recorded at different locations realative to the wave paddle (f l = 0.34 Hz to f u = 1.02 Hz. G a = 0.002, cylinder moves towards the wave paddle with speed of 0.25m/s) Excellent agreement on the wave elevation recorded by gauges near the wave paddle

27 pressure(100pa) pressure(100pa) Progress in Developing Hybrid Models Two-way coupling between QALE-FEM & twophase incompressible NS solver Validation: moving cylinder in extreme waves Exp QALE-FEM 28.5cm below MWL Physical Models and Validations cm below MWL cm above MWL time(s) 40 Exp QALE-FEM 28.5cm below MWL cm above MWL 8.5cm below MWL time(s) Time history of pressure at different locations on the cylinder surface facing incident wave (f l = 0.34 Hz to f u = 1.02 Hz. G a = 0.002, cylinder moves towards the wave paddle (left) 0.25m/s and (right) 0.75m/s) Acceptable agreement between the QALE-FEM prediction and the experimental data of the pressure recorded on the cylinder surface

28 MLPG-R NLSE-ESBI-QSBI Progress has been made on model coupling NLSE-ESBI-QSBI Couple QALE-FEM with single-phase MLPG-R Couple QALE-FEM with two-phase VOF solver Self-adaptive wavemaker based on frequency tracking Working closely to form an integrated multiscale, multi-model numerical framework Shiqiang Yan, Jinghua Wang & Qingwei Ma City, University of London Northampton Square London EC1V 0HB United Kingdom

29 WP3 Coupling fully nonlinear potential method with multiphase models Wave-structure interactions: multi-scale problems Different spatial scales Turbulent mixing and bubbles in centimetre Vortex shedding in meters Wave length in hundred meters Change of energy spectrum in kilometres Different temporal scales Impacts measured by millisecond Structural natural periods in centi-seconds Water wave periods in a few seconds Internal waves in minutes Different physical features Viscos effect is less significant for nonbreaking waves but significant for breaking waves Compressibility is important for impact; but not for others Air entrapment may be considerable near free surface but not otherwise Zonal approach coupling different numerical models aims to maximise the computational robustness

30 WP3 Coupling fully nonlinear potential method with multiphase models Zonal Coupling strategy in the Multi-Model Multi- Scale Hybrid models MLPG-R Robust QALE-FEM to cover large domain away from floating bodies; Strong couple between the FNPT model and the NS model using various technics; Self-adaptive wave-maker is applied at outer boundaries of the FNPT domain for wave generation/absorption; A single-phase meshless method (MLPG-R) may be coupled with QALE-FEM in the FNPT domain to resolve breaking waves; QALE-FEM will be coupled with the I-NS (wsifoam)

31 WP3 Coupling fully nonlinear potential method with multiphase models Self-adaptive wave absorber for highly nonlinear water waves Main Idea for improvement: track the local wave frequency Does not require any pre-specification to the wavemaker, e.g. the wave spectrum Frequency tracking using extended Kalman filter

32 WP3 Coupling fully nonlinear potential method with multiphase models Self-adaptive wave absorber for highly nonlinear water waves Works for both monochromic wave and steep wave group k 0 a 0 =0.15,k 0 d = 0.6 PM spectrum, ω 0 4.3s 1, 210 components random amplitude approach k 0 H s 0.2,k 0 η max 0.13 Water depth: 2.93m To be extended to 3D wave basin

33 WP3 Coupling fully nonlinear potential method with multiphase models Coupling QALE-FEM with OpenFoam Whole domain: FNPT model Near the floating bodies: a local mesh is generated and moves together with the floating bodies Two-phase incompressible NS solver is used in the local area Overlap area: a robust quadric interpolation scheme is developed and used