1. CHAM Seminar PHOENICS/FLAIR for the Wind Environmental Simulation 2008 by Jeremy Wu Computer Simulation of Fluid Flow, Heat Flow, Chemical Reactions and Stress in Solids.

2. CHAM Seminar PHOENICS Overview P arabolic H yperbolic O r E lliptic N umerical I ntegration C ode S eries PHOENICS is a general-purpose CFD code The name PHOENICS is an acronym standing for:

3. CHAM Seminar PHOENICS Overview PHOENICS is based on the finite volume method. Domain is discretized into a finite set of control volumes or cells. General conservation (transport) equation for mass, momentum, energy, etc., are discretized into algebraic equations as follows: unsteadyconvectiondiffusionsource

4. CHAM Seminar Discretize the equations in conservation (integral) form Finally For the conservation equation for variable , the following steps are taken: o Integration of conservation equation in each cell. o Calculation of face values in terms of cell-centered values. o Collection of like terms.

5. CHAM Seminar Main Features of PHOENICS 1-,2- and 3-D geometries Relational input capability Cartesian, Polar and Body-Fitted Coordinates Local multi-level fine-grid embedding Cut-cell technique for complex geometry Conjugate Heat Transfer Single or Multi-Phase Flow Particle Tracking Chemical reaction Radiation Non-Newtonian Flow Choice of equation solvers and differencing schemes Automatic generation of user code Automatic convergence control

6. CHAM Seminar Main Features of PHOENICS PHOENICS predicts quantitatively:- how fluids (air, water, steam, oil, blood, etc) flow in and around: – engines, – process equipment, – buildings, – lakes, river and oceans, – and so on; the associated changes of chemical and physical composition;

7. CHAM Seminar PHOENICS Structure PHOENICS consists of several modules: – Pre-processorfor setting up problems, – Solver for solving the problem, – Post-processor for visualising results; and – POLIS for providing information.

8. CHAM Seminar PHOENICS Structure POLIS Information

9. CHAM Seminar How the problem is defined Problem definition normally involves making statements about: ogeometry, ie shapes, sizes and positions of objects and intervening spaces; and grid, ie the manner and fineness of the sub-division of space and time; oprocesses, for example:- whether the heat transfer is to be calculated; whether materials are inert or reactive; whether turbulence is to be simulated and if so by what model;

10. CHAM Seminar omaterials, ie thermodynamic, transport and other properties of the fluids and solids involved; oEnvironmental or boundary conditions; and oother numerical (ie non-physical) parameters affecting the speed, accuracy and economy of the simulation. How the problem is defined

11. CHAM Seminar The Virtual-Reality Interface Model setup – VR Editor Clicking on an object brings a dialogue box onto the screen. This enables the information about the object to be edited.

12. CHAM Seminar Model setup – VR Editor The object geometry can be taken from a library of shapes, or loaded from a CAD file. CAD geometries are read using the STL format The Virtual-Reality Interface Import a CAD file

13. CHAM Seminar Setting Up Problems: PHOENICS-VR main menu The PHOENICS-VR Main menu allows you to make all the settings required for a problem, including: Geometry Variables to be solved (models) Fluid properties Initial values Boundary conditions Monitoring options

14. CHAM Seminar A Typical EARTH Convergence Monitor Plot EARTH is the program that performs the simulation. The graphical monitor shows the converging solution.

15. CHAM Seminar The EARTH run can be interrupted to change several parameters:  Monitoring position  Relaxation factors  Graphical monitor settings  Intermediate result files can also be dumped Solver-Graphics Monitoring

16. CHAM Seminar Analysis of Results - VR Viewer The VR Viewer allows users to see their results in a number of different ways:

17. CHAM Seminar Iso-surfaces Streamlines, static or animated Vectors Contours Analysis of Results - VR Viewer

18. CHAM Seminar AUTOPLOT X-Y graphs plot This allows for easy comparison of PHOENICS solutions with experimental or analytical data.

19. CHAM Seminar Programmability oUsers' subroutine, GROUND oPLANT converts user-defined formulae into error-free Fortran coding and places it correctly in GROUND oIn-Form carries the PLANT idea one stage further, by: oeliminating the Fortran writing, oeliminating the compilation, oeliminating the executable-building, osimplifying the syntax, and oenlarging the functionality.

20. CHAM Seminar What is FLAIR? FLAIR is a Special-Purpose version of the general CFD code PHOENICS. It has been created by removing many unneeded generic features, and adding several specific features. FLAIR is designed to provide an air-flow and thermal- simulation facility for the HVAC community, and for those concerned with the performance of airflow systems for the built environment, and with fire, smoke and pollutant hazards for both internal and external flow scenarios.

21. CHAM Seminar What does FLAIR do ? FLAIR predicts air-flow patterns, temperature distributions and smoke movement in buildings and other enclosed spaces. FLAIR can be used during the design process to detect and avoid uncomfortable air speeds or temperature. FLAIR also predicts effects of smoke movement, or any other gaseous pollutant, helping to achieve safe design of buildings, underground systems, aircraft or train cabins etc.

22. CHAM Seminar FLAIR Features FLAIR uses the PHOENICS VR-Editor to set the problem up, with the following additional items: –ISO 7730 Comfort index calculations: PMV, PPD. –ISO 7730 Draught rating. –CIBSE dry resultant temperature. –Humidity calculations, with output of humidity ratio and relative humidity. –Smoke movement calculation, with output of PPM, smoke density and visibility. –Mean age of air calculation. –Fan operating point calculation for single and multiple fans. –System-curve calculations.

23. CHAM Seminar In addition, the following object types have been added: –Diffuser Round FLAIR Features

24. CHAM Seminar Round diffuser

25. CHAM Seminar Vortex FLAIR Features In addition, the following object types have been added: –Diffuser

26. CHAM Seminar Vortex diffuser

27. CHAM Seminar Rectangular FLAIR Features In addition, the following object types have been added: –Diffuser

28. CHAM Seminar Rectangular diffuser

29. CHAM Seminar Directional FLAIR Features In addition, the following object types have been added: –Diffuser

30. CHAM Seminar Directional diffuser

31. CHAM Seminar Grille FLAIR Features In addition, the following object types have been added: –Diffuser

32. CHAM Seminar Grille/Nozzle

33. CHAM Seminar Displacement FLAIR Features In addition, the following object types have been added: –Diffuser

34. CHAM Seminar Displacement diffuser

35. CHAM Seminar FLAIR Features In addition, the following object types have been added: –Diffuser –Fire

36. CHAM Seminar FLAIR Features In addition, the following object types have been added: –Diffuser –Fire –Person (standing or sitting facing any Direction)

37. CHAM Seminar

38. CHAM Seminar FLAIR Features In addition, the following object types have been added: –Diffuser –Fire –Person –Crowd –Sunlight Created in Shapemaker

39. CHAM Seminar FLAIR Features In addition, the following object types have been added: –Diffuser –Fire –Person –Crowd –Sunlight –Spray head

40. CHAM Seminar Spray-head represents sprinklers user for fire- suppression. It uses GENTRA to model the droplet paths. Evaporation is considered, and is linked to the FLAIR humidity model. The GENTRA inlet table is written automatically. FLAIR Features

41. CHAM Seminar FLAIR Features In addition, the following object types have been added: –Diffuser –Fire –Person –Crowd –Sunlight –Spray head –Jet fan

42. CHAM Seminar Jet fan

43. CHAM Seminar FLAIR applications FLAIR applications include:- Clean rooms Operating theatres Sports arenas Car parks Road and rail tunnels Industrial environments Residential developments Fire safety scenarios Pollution control

44. CHAM Seminar Environmental Simulation The questions to be addressed: Geometry and its representation Domain Size Grid Size What Turbulence models Conservation equations Boundary conditions Numerical scheme Convergence

45. CHAM Seminar Geometry creation DWG---AutoCad  STL Facetfix might be needed if there are defects in the STL file.

46. CHAM Seminar Repairing holes with Facetfix

47. CHAM Seminar Geometry creation from a PDF file The numbers represent the height of the building

48. CHAM Seminar Geometry creation from a PDF file The solution was to save the image as a GIF file, then use it as a backdrop in AC3D We could trace round each part, then extrude

49. CHAM Seminar Geometry creation from a PDF file The resulting geometry was then exported to PHOENICS

50. CHAM Seminar Domain Size Generally, the boundary distance from the built area should be large enough in order that the physical boundary conditions can be reasonably specified as follows: The upper boundary should provide zero axial pressure gradient; The upstream boundary should be sufficiently far from the built area to ensure that the latter’s influence on the pressure field is negligible. Location of the downstream and lateral boundaries should be well beyond any separation regions in the wake.

51. CHAM Seminar Domain size

52. CHAM Seminar Domain Size Some literature recommend: Upstream > 8H Downstream >15H Upper > 6H where H is the height of the object Some recommend: Any side of the boundary > 4H So there is no agreed domain size. Advice: to use coarse grid to check the domain size by inspecting the pressure and velocity fields

53. CHAM Seminar The boundaries are too close to the object

54. CHAM Seminar Acceptable boundary distances from the object

55. CHAM Seminar Grid distribution A sufficient high mesh resolution in the vertical direction close to the bottom of the computational domain; The minimum grid number between buildings; A distance Yp from the centre point P of the wall-adjacent cell to the wall that is larger than the physical roughness height Ks of the terrain (Yp>Ks); A horizontally homogeneous ABL flow in the upstream, and downstream region of the domain; Away from the object, the grid could be expanded for the purpose of economy; More grids should assigned in the region when    x is great; Generally, the grid should be fine enough to catch the interested details, but coarse enough for the economy.

56. CHAM Seminar PARSOL Cut-cell technique Its advantage Its disadvantage How to check the grid distribution ? One sweep and PRPS contours

57. CHAM Seminar The first grid size

58. CHAM Seminar Inlet boundary condition Either a logarithmic or power-law velocity profile can be specified, as follows: Log law velocity profile: U/U* = ln(z/zo) / κ The turbulence quantities are set to: k = U*2/0.3 ; ε= U*3/(κ*z) where U is the total velocity at the height z from the ground (assumed to be at the origin of the object); κ is von Karman’s constant (=0.41), zo is the effective roughness height of the ground terrain. typically Zo~ 0.1hc

59. CHAM Seminar Log law wind profile It postulates that near given surface, the wind profile should be determined by the height z above the surface (which scales the eddy size) and the surface fluxes) Surface mom flux (u’w’) or shear stress u* = (u’w’) 1/2 (friction velocity) Km ~ (velocity) (length) =(u*) (k z) (eddy viscosity) u’w’ = - Km (du/dz) = (u*) 1/2 =ku*z (du/dz) u(z)/u* = 1/k lnz+C u(z)/u* = 1/k ln(z/zo) zo=height where u=0

60. CHAM Seminar Inlet boundary conditions Power-law wind profile Power law profile: U/Ur= (z/zr)** α The turbulence quantities are set to: k = U*2/0.3 ; ε= U*3/(κ*z) where U is the total velocity at the height z from the ground (assumed to be at the origin of the object); κ is von Karman’s constant (=0.41); Ur is the reference velocity at the reference height zr, and α is the power-law exponent. The total friction velocity U* is given by: U*= Ur κ/ln(zr/zo)

61. CHAM Seminar Inlet wind profile Typical values of the roughness height zo and the power-law exponent α are provided in PHOENICS

62. CHAM Seminar Downstream, Lateral and upper boundary conditions Downstream boundary condition: atmospheric pressure condition Upper boundary condition: Atmospheric pressure condition with the velocity appropriate at that height Lateral boundary condition: atmospheric pressure condition

63. CHAM Seminar What turbulence model ? What is turbulent flow ? Prediction methods RANS turbulence model Comparison of RANS turbulence models Recommendations

64. CHAM Seminar What is turbulent flow It has a number of characteristic features: 1. Irregularity, random, chaotic; consists of a spectrum of different eddy sizes. 2. Diffusivity increases. 3. Large Reynolds Number. 4. Three-dimensional. 5. Dissipative; kinetic energy in the small eddies are transformed into internal energy. 6. Continuum. The small turbulent scales are much larger than the molecular scale so we can treat the flow as a continuum.

65. CHAM Seminar Turbulent length scale and energy distribution

66. CHAM Seminar Prediction Methods DNS, LES & RANS l  = l/Re L 3/4 Direct numerical simulation (DNS) Large eddy simulation (LES) Reynolds averaged Navier-Stokes equations (RANS)

67. CHAM Seminar RANS turbulence models oA turbulence model is a computational procedure to close the system of mean flow equations. oFor most engineering applications it is unnecessary to resolve the details of the turbulent fluctuations, but only need to know how turbulence affects the mean flow. oTurbulence models allow the calculation of the mean flow, with expressions for the Reynolds stresses, without first calculating the full time-dependent flow field, oFor a turbulence model to be useful it:  should have wide applicability,  be accurate,  simple,  and economical to run.

68. CHAM Seminar Turbulent viscosity A new quantity appears: the turbulent (or eddy) viscosity  t The turbulent viscosity is not homogeneous, i.e. it varies in space. It is, however, assumed to be isotropic. It is the same in all directions. This assumption is valid for many flows, but not for all (e.g. flows with strong separation or swirl).

69. CHAM Seminar Turbulence models The following models are available in PHOENICS and can be used to predict the turbulent viscosity: oMixing length model oLVEL model oStandard k-  model oKato-Launder k-  model oK-  Chen-Kim model ok-  RNG model ok-  model oRMS oand many more

70. CHAM Seminar Standard k-  model Advantages: –Economical run –Leads to stable calculations that converge relatively easily. –Reasonable predictions for many flows. Disadvantages: –Poor predictions for: swirling and rotating flows, flows with strong separation, –Simplistic ε equation. Many attempts have been made to develop two- equation models that improve on the standard k-ε model. For example: K-  RNG, k-  KL and k-  realizable models.

71. CHAM Seminar Reynolds stress model RSM closes the Reynolds-Averaged Navier-Stokes equations by solving additional transport equations for the six independent Reynolds stresses. Resulting equations contain terms that need to be modeled. RMS avoids the isotropic eddy viscosity assumption. RSM is good for flows with streamline curvature, swirl, rotation and high strain rates:  Cyclone flows, swirling combustor flows.  Rotating flow passages, secondary flows.  Flows involving separation.

72. CHAM Seminar Comparison of RANS turbulence models Model Strength weakness STD K-   Robust, economical,  reasonably  poor results for flows with accurate, long accumulated severe pressure gradients, performance data strong streamline curvatures, swirl and rotation. K  RNG Good for moderately complex subject to limitations due to K-  KL flows like jet impingement, isotropic eddy viscosity separating flows, swirling flows assumptions and secondary flows RMS Physically account for transport, more CPU efforts and anisotropy of turbulent stresses less easy to achieve converged results

73. CHAM Seminar Recommendation For wind-engineering applications, usually very large mesh numbers, excessive iterations are required, this prevent most of the CFD users from the adequate use of RMS model. K-  KL, K-  RNG and K-  Chen-Kim models should provide reasonably accurate results for the purpose of most wind-engineering simulations. LES model is not recommended.

74. CHAM Seminar Conservation equations

75. CHAM Seminar Numerical convection scheme Numerical scheme where is the source term for the control volume, P and Jf f and D f represent, respectively, the convective and diffusive fluxes of  across the control-volume face. !! The value of  is not available at the control volume faces and is determined by interpolation.

76. CHAM Seminar Numerical scheme Central Differencing Scheme (CDS) or P e E  (x) PP ee EE   e = 0.5 (   E +    P )

77. CHAM Seminar Upwind Differencing Scheme (UDS)  e =  P if (v. n) > 0  e =  E if (v. n ) < 0 P e E  (x) PP ee EE Flow direction

78. CHAM Seminar Hybrid Differencing Scheme  e = 0.5 (  E +  P ) for Pe < 2  e =  P if (v. n) > 0 for Pe > 2  e =  E if (v. n ) < 0 This is the default option in PHOENICS

79. CHAM Seminar Other schemes Other schemes are available (see POLIS) For example, QUICK (Quadratic Upwind Interpolation for Convective Kinetics) P e E  (x) PP ee EE W WW Flow direction  e = 3/8  E + 3/4  P - 3/4  w U >0

80. CHAM Seminar Convergence Automatic convergence control Check the followings: oMass balance oSpot values oResiduals oWhole-domain maximum absolute corrections oAny extreme values beyond the physical reality in the domain

81. CHAM Seminar Transfer objects for too large a terrain Smoke movement Iterations may be needed

82. CHAM Seminar Use of In-Form ACH = (Cubic Feet per Min x 60) / Vol of Room >4 Wind profiles Derived variables

83. CHAM Seminar Recommended CFD modelling practices CFD limitations oHuman oScience oComputing power Requirements oQuality - Physically plausible - numerically converged - validation (analytical solutions; experiments or well-established numerical predictions) oEconomy

84. CHAM Seminar Recommended CFD Modelling practices Collect Information o Geometry o Properties o Environmental conditions Simplification of physical problem CFD modelling o Simple geometry to complex one o Coarse grid to fine grid o 1D -> 2D -> 3D o Steady -> Transient o Step-by-step adding and testing physical models

85. CHAM Seminar Recommended CFD Modelling practices o Flow -> Heat Transfer -> Reactions o Single phase -> two-phase o Grid refinement test o Convergence o Interpreting results and Validation