1. Airflow modeling for research vessels B. I. Moat 1, M. J. Yelland 1, R. W. Pascal 1, S. R. Turnock 2, S. Popinet 3 1) National Oceanography Centre, UK 2) University of Southampton, UK. 3) National Institute of Water and Atmospheric Research, New Zealand. GOSUD/SAMOS Workshop 2-4 May 2006 Outline 1)Problem 2)Background 3)Introduction to Computational Fluid Dynamics (CFD) 4)CFD in ship design for atmospheric research 5)Examine various CFD codes 6)Summary

2. Problem Finding a measure of the acceleration/deceleration at any given instrument location Determining the effective measurement height (vertical displacement of air) Developing an effective tool that can be used during the ship design phase to best site atmospheric instrumentation Order 50 ships? - balance between time/finanical cost etc GOSUD/SAMOS Workshop 2-4 May 2006

3. Background The effects of flow distortion: –are not that sensitive to variations in wind speeds between 6 to 20 ms -1 (<0.5%) –are very sensitive to relative wind direction and anemometer location (20 %) –are sensitive to the ship design (~5%) GOSUD/SAMOS Workshop 2-4 May 2006 (VOSClim brochure)

4. The 4 stage CFD modelling process Digitize plans (autocad) Femgen (Femsys, UK) 4 WEEKS PER SHIP Or, obtain model from naval architects Significant learning curve 1) Geometry creation GOSUD/SAMOS Workshop 2-4 May 2006

5. The 4 stage CFD modelling process High resolution mesh Historically the most difficult part of CFD modelling Variable mesh sizes and types Different mesh for each wind direction 1) Geometry creation 2) Mesh generation GOSUD/SAMOS Workshop 2-4 May 2006

6. The 4 stage CFD modelling process 10 m/s 15 m/s Run times are machine dependant (1 - 10 days) 64 bit workstation 1- 3 Gbytes memory 1) Geometry creation 2) Mesh generation 3) Compute solution GOSUD/SAMOS Workshop 2-4 May 2006

7. The 4 stage CFD modelling process 10 m/s 15 m/s 1) Geometry creation 2) Mesh generation 3) Compute solution 4) Post-processing Estimate wind speed bias and vertical displacement

8. GOSUD/SAMOS Workshop 2-4 May 2006 Research ship design RRS Charles Darwin 5% wind speed error due to ship structure RRS Discovery Streamlined Wind speed errors (bow- on 1%)

9. GOSUD/SAMOS Workshop 2-4 May 2006 Research ship design RRS James Cook to be delivered in September 2006

10. GOSUD/SAMOS Workshop 2-4 May 2006 Research ship design RRS Charles Darwin RRS James Cook RRS Discovery Wind speed bias (%) Relative wind direction Request a streamlined superstructure Foremast as tall and far forward as possible Uncluttered foremast platform

11. CFD CODES available GOSUD/SAMOS Workshop 2-4 May 2006 VECTIS - www.ricardo.com –Cost: $11k academic license ($5.4k per additional processor) –Accuracy: 2% for low flow distortion sites (up to 10% bias) –Best validated –11 ships modeled - (Yelland et al. 2002) –Fast mesh generation process (staff time = 3 hours/run) –Computation times of 3-5 days per relative wind direction

12. CFD CODES available GOSUD/SAMOS Workshop 2-4 May 2006 VECTIS - www.ricardo.com –Cost: $11k academic license ($5.4k per additional processor) –Accuracy: 2% for low flow distortion sites (up to 10% bias) –Best validated –11 ships modeled - (Yelland et al. 2002) –Fast mesh generation process (staff time = 3 hours/run) –Computation times of 3-5 days per relative wind direction FLUENT - www.fluent.com –Cost: $?k depends upon application (equivalent to VECTIS ) –Accuracy: ? % –1 ships modeled - Dupuis et al. (2003) –Mesh generation process Initial Staff time = 2 weeks Less for other wind directions –Computation times of 1 day

13. CFD CODES available GOSUD/SAMOS Workshop 2-4 May 2006 GERRIS - gfs.sourceforge.net –Cost: FREE (GNU software license) –Accuracy: 4% (1 ships modeled - Popinet et al. 2004 ) University of Tokyo are currently modeling 2 ships (part of a PhD. Project) –Mesh generation fully automatic –Mesh adapts to the flow where specified –Computation times vary 3 to 5 days (depends on geometry detail required)

14. GOSUD/SAMOS Workshop 2-4 May 2006 New ‘parametric method’ proposed by Ship Science Assumption: hulls are similar and geometries don’t need to be accurate everywhere Script to transform ship shape into representations of other ships (each deck xyz +position rel. to bow) Investigate the level of detail required

15. GOSUD/SAMOS Workshop 2-4 May 2006 New ‘parametric method’ proposed by Ship Science Automatic generation of the large scale geometry Detailed foremast model added for each ship Automatic mesh generation for each wind direction Flow then simulated in FLUENT or CFX

16. Possible parametric method projects? GOSUD/SAMOS Workshop 2-4 May 2006 –Project based at Ship Science, Southampton University, UK with NOC advising –Initial feasibility study to determine accuracy of method - needs $27k funds –Full scale project aimed at modelling a significant fraction of the 50 ships - needs at least $36k/year funds

17. Comparison of approaches GOSUD/SAMOS Workshop 2-4 May 2006 Present detailed geometry approach –GERRIS/VECTIS/FLUENT for 1 ship at 10 wind direction: 8 weeks staff plus analysis time 5 to 10 weeks computation –OK for a low number of ships –Of known accuracy –Time scale: could start now (given staff time and software) –NOC would make avaialble existing geometries of the RV Ron Brown, RV Knorr and FS Polarstern

18. GOSUD/SAMOS Workshop 2-4 May 2006 Parametric geometry and mesh creation –Once developed, for 1 ship at 10 wind directions: about 2 days staff plus analysis time 1 day computation (parallel processing) –Feasible for studying the airflow over 50 ships –Accuracy not known yet –Validate by comparison to existing detailed models –Lead time: active research area which requires funding Comparison of approaches

19. SAMOS aims GOSUD/SAMOS Workshop 2-4 May 2006 Accuracy required? What relative wind directions to study? Number of ships to study? When ? Funding ?

20. CONTACTS GOSUD/SAMOS Workshop 2-4 May 2006 VECTIS - ben.moat@noc.soton.ac.uk www.noc.soton.ac.uk/JRD/MET/cfd_shipflow.php GERRIS - s.popinet@niwa.co.nz http://gfs.sourceforge.net NEW PARAMETRIC METHOD - S.R.Turnock@ship.soton.ac.uk