Chen y CFD Study of An Ufad System Ncembt-091012

FINAL REPORT NCEMBT-091012
A CO
OMPUTA ATIONALL FLUID DYNAMICS STU
UDY OF
CONNVENTIO
ONAL AN ND UNDEERFLOOR
R AIR
DISTTRIBUTIION SYS
STEMS
OCTO
OBER 2009
9
Yitung Chen,
C Ph.D.
Huajunn Chen, Ph.D.
Universsity Of Nevada, Las Vegas
Davor Novosel
N
Nationaal Center for Ennergy Managem
ment and Buildding Technologgies
FINAL REPORT NCEMBT-091012
NATIONAL CENTER FOR ENERGY MANAGEMENT

AND BUILDING TECHNOLOGIES TASK 05-09:
APPLICATION OF COMPUTATIONAL FLUID
DYNAMICS FOR HIGH ENERGY EFFICIENCY DESIGN
WITH HUMAN COMFORT OF CADVAV AND UFAD
SYSTEMS
OCTOBER 2009
Prepared By:
Yitung Chen, Ph.D.

Huajun Chen, Ph.D.
University Of Nevada, Las Vegas
Davor Novosel
National Center for Energy Management and Building Technologies
Prepared For:
U.S. Department of Energy

William Haslebacher
Project Officer / Manager
This report was prepared for the U.S. Department of Energy

Under Cooperative Agreement DE-FC36-03GO13072
NOTICE
This report was prepared as an account of work sponsored by an agency of the United States government. Neither the
United States government nor any agency thereof, nor any of their employees, makes any warranty, express or
implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any
information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned
rights. Reference herein to any specific commercial product, process, or service by trade name, trademark,
manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by
the United States government or any agency thereof. The views and opinions of authors expressed herein do not
necessarily state or reflect those of the United States government or any agency thereof.
NATIONAL CENTER FOR ENERGY MANAGEMENT AND BUILDING TECHNOLOGIES CONTACT

Davor Novosel
Chief Technology Officer
National Center for Energy Management and Building Technologies
601 North Fairfax Street, Suite 250
Alexandria VA 22314
703-299-5633
dnovosel@ncembt.org
www.ncembt.org
ii NCEMBT-091012
TABLE OF CONTENTS
EXECUTIVE SUMMARY............................................................................................................................................. 1
1. PROJECT OBJECTIVE............................................................................................................................................ 2
2. BACKGROUND .................................................................................................................................................... 3
3. LITERATURE REVIEW ........................................................................................................................................... 4
4. PROBLEM DESCRIPTION AND EXPERIMENTAL SETUP .......................................................................................... 7
4.1 Problem Description ..................................................................................................................................... 7
4.2 Experimental Setup ...................................................................................................................................... 8
5. METHODOLOGY ................................................................................................................................................14
5.1 Governing Equations ...................................................................................................................................14
5.2 Numerical Procedure ..................................................................................................................................17
5.3 Parallel Processing .....................................................................................................................................19
6. NUMERICAL ANALYSIS ON THE CAD SYSTEM OF THE BTLAB ...............................................................................20
6.1 CFD Analysis of a Single Four-Way Diffuser ..................................................................................................20
6.2 CFD Analysis of the Btlab Laboratory Space.................................................................................................24
7. CFD SIMULATION OF A UFAD SYSTEM ................................................................................................................32
7.1 Preprocessing ............................................................................................................................................32
7.1.1 Creating a 3-D Geometry .....................................................................................................................32
7.1.2 Mesh Generation .................................................................................................................................33
7.2 Simulation of a Single Swirl Diffuser ...........................................................................................................35
7.3 Impact of the Spray Angle on the Performance of the Swirl Diffuser ..............................................................38
7.4 CFD Simulation of the BTLab UFAD System Without Thermal Load ................................................................48
7.5 CFD Simulation of the BTLab UFAD System With Thermal Load.....................................................................53
8. COMPARISON OF SIMULATION RESULTS WITH EXPERIMENTAL DATA .................................................................57
8.1 Comparison of Simulation Results With Experimental Data for Zero theraml Load .........................................58
8.2 Comparison of Simulation Results With Experimental Data for Half Thermal Load .........................................71
9. CONCLUSIONS .................................................................................................................................................84
10. REFERENCES ..................................................................................................................................................85
NCEMBT-091012 iii
LIST OF FIGURES
Figure 1. The conventional design process of an HVAC system ................................................................................. 4
Figure 2. The design process of HVAC system with CFD simulation ........................................................................... 5
Figure 3. Computational domain of the BTLab......................................................................................................... 8
Figure 4. Ceiling Supply Air Diffusers And Return Air Grills In The BTlab .................................................................... 9
Figure 5. Floor Supply Air Diffusers In The BTlab ...................................................................................................... 9
Figure 6. Plan View Of Diffusers, Floor Heaters And Test Grid In The BTlab ..............................................................10
Figure 7. Section View ..........................................................................................................................................10
Figure 8. HVAC System Overview ...........................................................................................................................11
Figure 9. Top View Of The UFAD Access Floor .........................................................................................................12
Figure 10. Distribution Duct In The Underfloor Plenum ..........................................................................................12
Figure 11. Square cone diffuser and swirl diffuser in BTLab ...................................................................................13
Figure 12. A typical control volume and the notation used for a Cartesian 2D grid ..................................................17
Figure 13. Parallel FLUENT architecture (Fluent manual 6.3 ..................................................................................19
Figure 14. Modeling of a four-way square cone diffuser; (a) geometry, (b) mesh system ..........................................20
Figure 15. Grid system and computational domain for single diffuser characteristics study ....................................21
Figure 16. Velocity vector graph at the selected slice (Y= 5 ft) ................................................................................22
Figure 17. Variation of velocity magnitude as a function of X at the selected slice (Y= 5 ft) ......................................22
Figure 18. Path lines released from the top inlet of the diffuser. The path line colors are based on their respective
temperature. A total of 80 path lines was tracked..................................................................................................23
Figure 19. The temperature distribution at the selected slice (Y = 5 ft)....................................................................23
Figure 20. Variation of air temperature as a function of X at the selected slice (Y= 5 ft) ...........................................24
Figure 21. Grid system and computational domain for test space of BTLab ............................................................25
Figure 22. Path lines released from the top inlet of the diffuser; the path lines are colored by their respective
temperature .........................................................................................................................................................26
Figure 23. Velocity vector graph at the selected slice (Y= 5 ft) ................................................................................27
Figure 24. The temperature distribution across the selected slice (Y = 5 ft) .............................................................28
Figure 25. Variation of average surface air temperature and velocity magnitude as a function of height. .................28
Figure 26. Locations of temperature and velocity sensors in the BTlab ...................................................................29
Figure 27. The air velocity magnitude at test planes 1 and 2 ..................................................................................30
Figure 28. The air temperature at the test planes 1 and 2 ......................................................................................31
Figure 29. Geometry of the Nailor NFD flow swirl diffuser .......................................................................................32
Figure 30. 3-D model of the swirl diffuser; (a) exploded view; (b) assembled view ...................................................33
Figure 31. 3-D modeling (a) exploded view (b) assembled view ..............................................................................33
iv NCEMBT-091012
Figure 32. Meshes for a single swirl diffuser ..........................................................................................................34
Figure 33. Grid system and computational domain for modeling of a single swirl diffuser .......................................34
Figure 34. Velocity contour distribution (in m/s) at slice Y=0 m..............................................................................35
Figure 35. Path lines, which are colored depending of their respective temperature, as they were released from the
outlet of the single diffuser ....................................................................................................................................35
Figure 36. Iso-surfaces at different temperatures ..................................................................................................36
Figure 37. Temperature (K) distributions at different heights .................................................................................37
Figure 38. Temperature (K) distribution at the slice (Y=0 m) ...................................................................................37
Figure 39. Modeling different spray angles of the swirl diffuser ..............................................................................38
Figure 40. Velocity contours (m/s) at slice X=0 ft ..................................................................................................40
Figure 41. Path lines released from the outlet of the swirl diffuser (colored by path line ID) .....................................42
Figure 42. Temperature iso-surfaces at 299K (25.9C) for spray angles of 3, 4, 5 and 7 ................................43
Figure 43. Temperature (K) distributions at different heights for spray angles of 3, 4, 5 and 7 ........................47
Figure 44. Path lines released from the outlet of the swirl diffuser. The path lines are colored according to their
respective temperature (K) ....................................................................................................................................49
Figure 45. Velocity (m/s) distribution at the slice Y=0 m ........................................................................................50
Figure 46. Velocity vector graph (m/s) at the slice Y=0 m.......................................................................................50
Figure 47. Temperature (K) distribution at the slice Y=0 m .....................................................................................51
Figure 48. Velocity (m/s) as a function of x for different heights at the slice Y=0 .....................................................51
Figure 49. Temperature (K) as a function of x for different heights at the slice Y=0 ..................................................52
Figure 50. Average temperature and velocity as a function of height ......................................................................52
Figure 51. Schematic design of UFAD system of BTLab with floor heater ................................................................53
Figure 52. Schematic of the BTLab with distribution of the thermal loads (Tan 2008) .............................................54
Figure 53. Path lines as released from the outlet of the swirl diffuser; the path lines are colored according to their
respective path line ID ..........................................................................................................................................55
Figure 54. Temperature (K) distribution at the slice Y =5ft......................................................................................55
Figure 55. Schematic of test grid in the BTLab; the data obtained from grids in the region with the green box were
used for comparison of the modeled values versus the experimental ones ..............................................................58
Figure 56. Comparison of temperature and velocity in test grid A1 for zero thermal load .........................................59
Figure 57. . Comparison of temperature and velocity in test grid B1 for zero thermal load ......................................59
Figure 58. Comparison of temperature and velocity in test grid C1 for zero thermal load.........................................60
Figure 59. Comparison of temperature and velocity in test grid D1 for zero thermal load ........................................60
Figure 61. Comparison of temperature and velocity in test grid B2 for zero thermal load.........................................61
NCEMBT-091012 v
Figure 80. Comparison of temperature and velocity in test grid A1 for half thermal load .........................................71
Figure 81. Comparison of temperature and velocity in test grid B1 for half thermal load .........................................72
Figure 82. Comparison of temperature and velocity in test grid C1 for half thermal load .........................................72
Figure 83. Comparison of temperature and velocity in test grid D1 for half thermal load .........................................73
vi NCEMBT-091012
Figure 100. Comparison of temperature and velocity in test grid A6 for half thermal load .......................................81
Figure 101. Comparison of temperature and velocity in test grid B6 for half thermal load .......................................82
Figure 102. Comparison of temperature and velocity in test grid C6 for half thermal load .......................................82
Figure 103. Comparison of temperature and velocity in test grid D6 for half thermal load .......................................83
NCEMBT-091012 vii
LIST OF TABLES
Table 1. Boundary conditions for modeling the UFAD system of the BTLab with a thermal load ...............................54
viii NCEMBT-091012
EXECUTIVE SUMMARY
Underfloor air distribution (UFAD) systems have been applied as an alternative to conventional air
distribution (CAD) systems in recent years. This system was first introduced in the 1950s to cool
computer rooms. A major paradigm of the UFAD system is that its air distribution characteristics are
different from conventional overhead air distribution (CAD) systems. It has been suggested that new load
calculation and design methods need to be developed to properly design and implement this air
distribution technology. Empirical rules have been used to design UFAD systems. Limited 3-D
computational fluid dynamics (CFD) studies of UFAD system as primary means of distributing
ventilation air through a building have been conducted abroad. There is a lack of standardized and
verified design guidelines for UFAD systems.
Another National Center for Energy Management and Building Technologies project (Tan 2008)
compared the laboratory performance of an UFAD system versus a conventional overhead design. The
laboratory test data form that project was used here to benchmark the developed models.
This project applied a 3-D computational fluid dynamics (CFD) analysis to numerically model both a
conventional overhead air distribution system (CAD) with variable air volume (VAV) and an underfloor
UFAD system. CFD with structured or unstructured meshes evaluated the impact of different operating
conditions, including different supply air flow rate and different thermal loads for CAD and UFAD
systems, on the air flow pattern and temperature distribution. Various geometry model and mesh system
of the CAD and UFAD system respectively were generated. The final numerical model was
benchmarked against the experimental data from the Building Technologies Laboratory (BTLab) at
University of Nevada Las Vegas.
The objectives of the projects were only partially met. Models of the CAD and UFAD system as installed
in the BTLab were developed and initial simulations indicated good agreement with previous work.
Detailed comparisons between the temperature and velocity values from numerical simulations and
experiments, however, revealed significant differences for both zero and half thermal load. In case of the
half thermal load the differences exceeded the limits for comfort conditions as set forth in
ANSI/ASHRAE Standard 55-2004.
Possible reasons for the differences between the numerical and experimental data are:
Size of the computational domain. Although the BTLab was modeled using about 5,000,000
cells, this may be insufficient to model the air distribution in the mixing zone of each individual
swirl diffuser. Running simulations with significantly higher cell count, i.e., tighter mesh, was
not feasible to hardware limitations and budgetary constraints.
Air leakage. According to the experimental reports, the UFAD floor of the BTLab experienced
about 12 percent category 1 leakage, i.e., uncontrolled leakage from the UFAD plenum into the
conditioned space. His air leakage was not accounted for in the numerical model.
Non-uniformity for the inlet mass flow rate for the swirl diffuser. Our model assumed that the
flow rate in each swirl diffuser was equal.
Experimental blind regions. The experimental data was recorded for data points that were 12
inches apart on the x and y axes and 6 inches apart on the z-axis. This leaves large amounts of
the experimental space were no data were recorded. The computational grid was significantly
finer than the experimental one.
NCEMBT-091012 1
1. PROJECT OBJECTIVE
1. PROJECT OBJECTIVE
The objectives of this task were:
1. To develop CFD models of CAD and UFAD systems, particularly of the type of diffusers used in
the Buildings Technology Laboratory (BTLab) experiments, conducted under NCEMBT Task 2
2. To verify the developed models using the experimental data from Task 2
3. To provide detailed information regarding velocity and temperature concentrations and
turbulence intensity throughout the domain of interest.
This task was organized into four subtasks:
1. Collecting and analyzing existing experimental and numerical data on the CAD-VAV and UFAD
systems.
2. Designing geometry and generating meshes for different types of CAD-VAV and UFAD systems.
3. Establishing CFD models
4. Benchmarking the developed models against the numerical data from NCEMBT Task 2.
2 NCEMBT-091012
2. BACKGROUND
2. BACKGROUND
In recent years HVAC system design has been strongly influenced by increasing emphasis on indoor air
quality (IAQ), energy conservation, environmental effects, safety, and economics. The relative placement
of system components can significantly affect the thermal comfort and energy performance of the air
handling system. To design high-energy efficiency HVAC systems, it is necessary to gather the detailed
information about the behavior of the airflow in both the spaces and the rooms of the building. The
fundamental information concerning the flow comprises air velocity, temperature, relative humidity, and
species concentrations. All these parameters are important in assessing thermal comfort and indoor air
quality.
The conventional design of ventilation systems normally relies on valuable know-how, empirical
formulas and past experience. Although practical knowledge and basic methods provide successful
solutions, this type of engineering cannot take into account specific air flow patterns which are affected
not only by the positioning of openings and exhausts in a room, but by the distribution of objects and
energy sources as well. Consequences related to the absence of these elements include over-design and
unnecessary cost.
In this project, computational fluid dynamics techniques were used to predict optimal locations for
different types of diffusers with variable air volume due to various thermal loads within a commercial
building. A numerical solution of the HVAC design gives detailed and complete information, providing
relevant information regarding velocity, pressure, temperature, concentration, and turbulence intensity
throughout the domain of interest. The benchmarked CFD models and design data can be saved into a
database and used for general policy decision making, training undergraduate and graduate engineering
students, providing technical and scientific support for development of approaches to minimize impact of
acts of chemical, biological, or radiological (CBR) terrorism in buildings, and assisting new and
renovation construction with regard to building security protections, energy efficiency and indoor
environmental quality.
NCEMBT-091012 3
3. LITERATURE REVIEW
Over the past few decades, there were many studies on the CFD analysis and numerical simulations on
the indoor environment. According to the previous literature survey, it can be found that the numerical
simulation of the HVAC systems mainly focus on the following topics:
Air buoyancy flow in the room
The turbulence model
Typical laminar flow in clean rooms
Diffuser
Air flow simulations in a ventilated room
Studies on distribution of contaminants within different HVAC systems
General issues on simulations of internal fluid flow and heat transfer inside a room
All of the previous research achievements also show that the increasing developments of computational
fluid dynamics (CFD) in the recent years have opened the possibilities of low-cost yet effective methods
for improving HVAC systems in design phase, with less experiment required. The advantage and
disadvantage on conventional designs of HVAC systems, with or without using CFD techniques, are
shown in Figure 3.1 and Figure 3.2.
Figure 1. The conventional design process of an HVAC system
4 NCEMBT-091012
Figure 2. The design process of HVAC system with CFD simulation
CFD models have been used to study indoor air quality (IAQ) problems, pollutant distributions, and
performance of HVAC systems (Chow and Fung (1996), Emmerich (1997), Gadgil et al. (1999)). Juan et
al. (2004) showed a computational fluid dynamics model on the real environment of computer room. In
their study, the geometric model was created using the parametric features of the pre-processor Gambit, in
combination with elements created with the Rhinoceros NURBS modeling tool. Joseph et al. (2001)
numerically investigated the temperature distribution and air movement within an air-conditioned
gymnasium with four different, but commonly found, exhaust positions in Hong Kong. Hirnikel et al.
(2002) investigated contaminant removal effectiveness of three air distribution systems for a
bar/restaurant by using CFD modeling. They showed that directional airflow systems could reduce
peoples exposure to contaminants. Thermal comfort can be predicted based on Fangers PMV model
(1970), which assumes a uniform thermal environment. Based on Rohles and Nevins work (1971), a
thermal sensation index is also widely used for assessing thermal comfort. Relative humidity can be
computed by using the procedure recommended in ASHRAE (1997). Son et al. (2005) gives a detailed
numerical simulation of thermal comfort and contaminant transport in air conditioned rooms. However,
their results are limited to two-dimensional geometry, and the real environment is simplified into regular
geometry.
Recently, underfloor air distribution (UFAD) systems have become popular design alternatives to
conventional air distribution (CAD), such as overhead air distribution systems for thermal and ventilation
control (Woods 2004, Webster et al. 2002). This system was first introduced in the 1950s to cool a
computer room, and is emerging as a leading ventilation system design in modern commercial buildings.
According to Loudermilk (1999), there exist two major advantages of this system. One is that ventilation
cool air is certain to reach the occupants (as it is introduced within the occupied zone). Another is that
convection heat gains that occur above the occupied zone are isolated from the calculation of the required
space supply airflow. The potential advantages of a well-designed UFAD system include (1) improved
thermal comfort; (2) improved ventilation efficiency and indoor air quality; (3) reduced energy use; and
(4) reduced floor-to-floor height in new constructions.
According to Bauman and Webster (2001), there is a higher risk to designers and building owners due to
a lack of objective information and standardized design guidelines. Woods (2004) did a literature review,
searches, and field investigations to assess the actual performance of UFAD system in the real world. He
showed that there are gaps in available data: valid and reliable field data are not from a sufficient
population of existing facilities to conclude that an underfloor systems performance is superior to an
NCEMBT-091012 5
overhead system; and that designers must be made aware that underfloor as well as overhead systems
require more care in design, installation, and operations. He also recommended that objective analysis
should be made before choosing an HVAC system. Webster et al. (2002) presented a series of full-scale
laboratory experiments to determine room air stratification for a variety of design and operating
parameters. Fukao et al. (2002) carried out comparative field measurements for both systems in an actual
large-scale office building. Webster et al. (2002) presented a study about a building that operated with an
UFAD system. They showed little troubleshooting with the system operation, pointing out the positive
aspects of using well-designed UFAD systems. Bauman (1999) offered a work presenting a discussion
about several advantages shown by the UFAD systems. In the design stage, CFD simulation can play an
important role in improving the understanding of any particular system.
Based on the literature review, the following conclusions were drawn for the modifications to the Throw
Room and planned experiments:
CFD models have been used to study indoor air quality (IAQ) problems, pollutant
distributions, and performance of HVAC systems.
The increasing developments of computational fluid dynamics (CFD) in the recent years have
opened the possibilities of a low-cost, yet effective, method for improving HVAC systems in
design phase, with less experiment required.
Lack of data on CFD model on the UFAD system.
Lack of understanding how UFAD systems work in the real world.
6 NCEMBT-091012
4. PROBLEM DESCRIPTION AND EXPERIMENTAL SETUP

4.1 PROBLEM DESCRIPTION
The numerical simulation of the ceiling air distribution (CAD) system and numerical simulation of the
underfloor air distribution (UFAD) system, based on the dimensions of the BTLab at UNLV, was studied
using a computational fluid dynamics (CFD) technique. The CAD system of the BTLab included four
square cone diffusers with four exhaust fans. The UFAD system had eight swirl diffusers and sixteen
heaters and four exhaust fans. The fundamental information about the UFAD experimental system in
UNLVs BTLab was collected. An analysis of the experimental system was made and a simplified
geometry of the BTLab for UFAD system simulation was generated.
The initial computational domain data for simulating the CAD and UFAD systems of the BTLab is shown
in Figure 3.
Parameters for test space of the BTLab were:
Space length = 9.1 m
Space width = 6.1 m
Ceiling height = 2.7 m
Space floor area = 55.7 m2
Air diffusion performance index >80
South wall surface temperature, Tsw = 23.90C
North wall surface temperature, Tnw = 23.90C
East wall surface temperature, Tew = 23.90C
West wall surface temperature, Tww = 23.90C
NCEMBT-091012 7
Figure 3. Computational domain of the BTLab
4.2 EXPERIMENTAL SETUP

The University of Nevada Las Vegas (UNLV) was conducting the laboratory phase of this task. The
principal investigator was Dr. Liangcai (Tom) Tan. This task compared the performance of UFAD and
CAD systems.
The conventional air distribution system has a VAV box, an end duct, four ceiling diffusers, and four
return air grills in the four corners of the hung ceiling, as shown in Figure 4. The underfloor air
distribution system includes a vertical shaft, an end duct, an underfloor plenum, and eight floor diffusers,
as shown in Figure 5 . The return air grills are the same as the CAD system. Figure 6 and Figure 7 show
the plan and section views of diffusers respectively. A three-dimensional drawing of the HVAC system is
shown in Figure 8. A view of the UFAD floor is depicted in Figure 9 and the distribution of air in the
UFAD plenum in Figure 10. The two types of the diffusers employed in this experiment are shown
in Figure 11. A detail description is given in (Tan 2008).
8 NCEMBT-091012
Figure 4. Ceiling Supply Air Diffusers And Return Air Grills In The BTlab
Figure 5. Floor Supply Air Diffusers In The BTlab
NCEMBT-091012 9
EAST
0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 30 ft
0 ft
PE3: the third group of floor heaters, service area 200 ft2
2
1 2 3 4 5 6 7 8 9 10 1 1 12
A 4
B 6
8
C
SOUTH
2
PE2: the second group of floor heaters, service area 200 ft
NORTH
10
D
E 12
14
F
16
G
T
18
PE1: the first group of floor heaters, service area 200 ft2
20
WEST
Ceiling Diffuser Floor Swirl Diffuser Baseboard Heater T Reference Point Test Position
Figure 6. Plan View Of Diffusers, Floor Heaters And Test Grid In The BTlab
CEILING
0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 30 ft
17 (104)
Zone C: 16 (98)
Upper zone in the test space
6 to 9 ft above floor 15 (92)
Induction zone in CAD systems 14 (86)
Stagnant zone in UFAD systems
13 (78)
12 (72)
11 (66)
9 ft
SOUTH
10 (60)
NORTH
Zone B: 9 (54)
Middle zone in the test space T
Uniform mixed zone in CAD systems 7 (42)
Uniform mixed zone in UFAD systems
6 (36)
Zone A:
Lower zone in the test space 5 (30)
Uniform mixed zone in CAD systems 3 (18)
Mixing zone in UFAD systems
2 (12)
1 2 3 4 5 6 7 8 9 10 1 1 12 1 (4)
FLOOR
Ceiling Diffuser Floor Swirl Diffuser T Reference Point Test Point
Figure 7. Section View

OOf Diffusers, Floor Heaters And Test Grid In The BTlab
10 NCEMBT-091012
UFAD system supply duct

CAD system supply duct
Smoke dampers
Connect to flexible duct and
ceiling diffusers Main return duct
Main supply duct

Nozzle chamber
Fan room, w/, return

Vertical duct DX coil, fan and
supply DX coil
Heat pumps
outdoor units
Floor diffusers
Figure 8. HVAC System Overview
NCEMBT-091012 11
4. PROBLEEM DESCRIPTIO
ON AND EXPERIMENTAL SETU
UP
F
Figure 9. Top Vieew Of The UFAD Access
A Floor
Recttangular
Solid Duct Pedestal
w 1
with
In
nline
Insulation Floor
Diffuser
West
Eastt
Ro
ound Solid Ductt
with 1 Inline
Insulation Damper Vertical Duct
V
O
Outlet along
North Wall
Figuree 10. Distributioon Duct In The Unnderfloor Plenum

m
12 N
NCEMBT-091012
Figure 11. Square cone diffuser and swirl diffuser in BTLab
Stratification has not been observed in several installed UFAD systems; rather the systems appear to
perform similarly to a CAD system. Therefore, the UFAD Task Force of the National Center of Energy
Management and Building Technology recommended that the initial experiments of this task compare
UFAD and CAD systems and specifically determine if stratification is present or not. The comparison
was made with both systems having been designed with an air distribution performance index (ADPI) of
greater than 80%. The design of experiment (DOE) was based on the (null) hypothesis that There are no
significant differences in air velocity and thermal distribution in interior office zones between CAD and
UFAD systems that are designed to perform at an ADPI greater than 80%. The null hypothesis implied
that there is no thermal stratification in a UFAD system. Thus, the existence of stratification was verified
or disproved by this experiment. The experiment followed the guidelines of ASHRAE Standard 113-
1990 - Method of Testing for Room Air Diffusion.
NCEMBT-091012 13
5. METHODOLOGY
5. METHODOLOGY
5.1 GOVERNING EQUATIONS
Based on the design supply air flow rate, the air flow inside the test space of BTLab is turbulent. For the
turbulent gas flow, the velocities and temperature can be divided into a mean and a fluctuating value,
u j = U j + u 'j and t = T + T ' . Then the governing time-averaged partial differential equation for
conservation of gas phase mass momentum, and energy are:
Continuity equation:
u j
=0 (5.1)
x j
Momentum equation:
U i U j

(U iU j )= P + + u 'i u ' j g i
(5.2)
x j xi x j j x xi
Energy equation:

(u j C pT )= T u ' j T ' (5.3)
x j x j x j
where is the density, is the viscosity, p is the pressure, C p is the specific heat capacity, is the
thermal conductivity. The turbulent stress and heat flux are determined by
U i U j 2
u 'i u ' j = t + ij k (5.4)
x j xi 3
t T
u' j T = (5.5)
Prt x j
where ij is the Kronecker delta function, ij = 1 when i = j and zero when i j , k is the turbulent
kinetic energy, Prt is the turbulent Prandtl number and taken as 0.9, and t is the turbulent viscosity,
t = C k 2 / , where C = 0.09 and is the turbulence dissipation.
Various turbulence models have been proposed in the past twenty years. The k model is the most
popular of the two equation models and produced qualitatively satisfactory results for a number of
complex flows. This technique uses wall functions to treat the near-wall sub-layers. Recently, the
realizable k model (Shih et al. 1995) is used in the study for closing the turbulent NavierStokes
equation system. This model constitutes a revised standard k model (Jones and Launder 1972). A
bibliographical study concerning the validation of the CFD model, including the turbulence model (the
realizable k model), showed that this approach has been extensively validated for a wide range of
flows, including the channel and layer flows (Shih et al. 1995; Kim et al. 1997; Fluent 2005). Comparing
14 NCEMBT-091012
5. METHODOLOGY
the standard the k model, this model is likely to provide superior performance for flows involving
rotation, boundary layers under strong adverse pressure gradients, separation and recirculation (including
flows over obstacles). For all of these cases, the performance of the model has been found to be
substantially better than that of the other turbulent models. Finally, the model used satisfies certain
mathematical constraints on the Reynolds stresses, consistent with the physics of turbulent flows.
The realizable k model differs from the standard k model in two important ways. One is that the
realizable k model contains a new formulation for the turbulent viscosity. And another is that a new
transport equation for the dissipation rate has been derived from an exact equation for the transport of the
mean-square vorticity fluctuation. In realizable k model, the eddy viscosity is no longer constant and its
determination depends on the mean strain and rotation rates, the angular velocity of the system rotation,
and the turbulence fields. The realizable eddy viscosity formulation is based on the realizability
constraints, the positivity of normal Reynolds stresses and the Schwarz inequality for turbulent shear
stresses (Shih et al. 1995). The addition equations in the realizable k model can be written as
k

(kU j )= + t + Gk (5.6)
x j x j k
j
x

(U j )= + t + C1 S C 2
2
(5.7)
x j x j k x j k +
where C1 = max[0.43; + 5], C 2 = 1.9 , k = 1.0 , = 1.0 , = S

k
and S is the modulus of

the mean rate of strain tensor defined by
S = 2 S ij S ij (5.8)
Gk is the production of turbulence kinetic energy and is the cinematic viscosity. The inlet boundary
condition of k and , which depend on the turbulent intensity I, the turbulence length scale l (depending
on the hydraulic diameter Dh) and the inlet velocity, can be written as
3 C 3 / 4 k 3 / 2
k= (u I ) 2 and = (5.9)
2 l
1 / 8 l = 0.07 D
where I = 0.16 Re and h , Re is the Reynolds number. For our configuration, the turbulent
intensity is taken equal to 2%. The set of equations for air is completed by using the ideal gas equation of
state.

In this project, it is assumed that the mass diffusivities of species in air are scalars ( k ) and the thermal
diffusion is negligible. Since there is neither source nor chemical reaction, the equations for the mass
conservation of water vapor and contaminant gas are

(u jk )= Dk T
k = water or contaminant gas (5.10)
x j x j x j
where Dk is the diffusion coefficient for k scalar.
NCEMBT-091012 15
5. METHODOLOGY
Once the temperature, water vapor concentration and pressure are obtained, the relative humidity can be
computed by
pw
= (5.11)
p ws
where
(101325 + p) water
pw = (5.12)
0.62198 + 0.37802 water
5800
p ws = 1000 exp 5.516 0.04864T + 4.176 10 5 T 2 1.445 10 8 T 3 + 6.546 ln T (5.13)
T
To evaluate the thermal comfort, the Fanger model (Guan et al 2003) was used in this project. This model
was originally developed to predict human thermal comfort in office-like environments, and has gained
wide usage in HVAC industry because of its simplicity. In this model, predicted meant vote is the
parameter for assessing thermal comfort in the occupied zone based on the conditions of metabolic rate,
clothing, air speed besides temperature and humidity. It can be expressed as
PMV = {0.303 exp[0.0036( M W )] + 0.028}{( M W )
3.05 10 3 [5733 6.99( M W ) p w ] 0.42[( M W ) 58.15]
4 4
(5.14)
1.7 10 5 M (5867 p w ) 0.0014 M (34 Ta ) 3.96 10 8 f cl [Tcl + Ta ]
f cl hc (Tcl Ta )}
where
{ 4 4
Tcl = 35.7 0.028( M W ) I cl 3.96 10 8 f cl [Tcl + Ta ] + f cl hc (Tcl Ta )} (5.15)
hc = 2.38(Tcl Ta ) 0.25 (5.16)
1.00 + 1.29 I cl I cl 0.078 m 2 K/W

f cl = (5.17)
1.05 + 0.645I cl I cl > 0.078 m 2 K/W
The scale of PMV values refer to the follows
3=hot, 2=warm, 1=slightly warm, 0 = neutral, -1=slightly cool, -2=cool, -3= cold
For assessing the effectiveness of an occupied zone, the contaminant removal effectiveness(CRE) was
used . Here, the CRE was determined based on the mean contaminant concentration in the supply inlet,
the exhaust and the occupied zone (Emmerich 1997). In our case, it assumed that the supply air flow is
contaminant free and the CRE can be computed as
CRE = E mean (5.18)
where E is the mean contaminant concentration in exhaust and mean is the mean contaminant
concentration in the occupied zone.
16 NCEMBT-091012
5. METHODOLOGY
5.2 NUMERICAL PROCEDURE

After selecting the mathematical model, a suitable discreetization method had to be selected. The
commercial FLUENT 6.3 software is in our numerical simulation, in which the discreetization method is
the finite volume method (FVM).
The FVM is a numerical method for solving partial differential equations (PDE) that calculates the values
of the conserved variables averaged across the volume. The FVM uses the integral form of the
conservation equations as its starting point. Assume is a generic conserved intensive property (for mass
r
conservation, =1; for momentum conservation, = u ; for conservation of a scalar, represents the
conserved property per unit mass), then the integral form of the generic conservation equation can be
expressed as:
r r r
t

d + u ndS = grad ndS + q d
S S
(5.19)
For simplicity, steady-state is considered. Assuming the velocity field and all fluid properties are known,
then the generic conservation equation becomes:
r r r

S
u ndS = grad ndS + q d
S
(5.20)
The solution domain is subdivided into a finite number of contiguous control volumes (CVs) by a grid
which defines the control volume boundaries, and the conservation equations are applied to each CV. At
the centroid of each CV lies a computational node at which the variable values are to be calculated.
Interpolation is used to express variables at the CV surface in terms of the CV center values. Typical 2D
Cartesian control volumes with the notations are shown in Figure 12. For maintenance of conservation, it
is important that CVs do not overlap; each CV face is unique to the two CVs which lie on either side of it.
Figure 12. A typical control volume and the notation used for a Cartesian 2D grid
A typical 2D CV is taken as an example. Its surface is subdivided into four plane faces ( S e , S w , S s and
S n ) on which e, w, s and n represent the center of the surface; ne, nw, sw and se represent
the corner nodes and P is the centroid. Thus, the net flux through the CV boundary is the sum of integrals
over the four CV surfaces:
NCEMBT-091012 17
5. METHODOLOGY
fdS = fdS (5.21)

S k
k
r r r
where, f is the component of the convective ( v n ) or diffusive ( grad n ) vector in the direction
normal to CV surface. As the velocity field and all fluid properties are known, thus quantity is the only
unknown in f.
A typical CV face ( S e ) is taken as an example to calculate the surface Eq. (5.20), the analogous
expressions may be derived for all faces by making appropriate index substitutions. To calculate the
surface integral on S e , an approximation must be introduced because only the CV center values of f are
known. Approximation is best done using two levels: a) The integral is approximated in terms of the
variable values at one or more locations on the cell face; b) The cell face values are approximated in
terms of the CV center values.
The simplest approximation to the integral is the midpoint rule that can be expressed as:
Fe = f dS f e S e (5.22)
Se
Where f e is the value of f at the center of face S e .
Another second order approximation of the surface integral in 2D is the trapezoid rule, which leads to:
Se
Fe = f dS ( f ne + f se ) (5.23)
Se 2
In this case the integrand at the CV corners are needed. For higher order approximation, the integrand is
needed at more than two locations. For example, using the Simpsons rule, a fourth-order approximation
of the integral over S e is estimated as:
Se
Fe = f dS ( f ne + 4 f e + f se ) (5.24)
Se 6
Here the values of f are needed at three locations: the cell face center and the two corners.
As the values of f is unknown on S e , the values of f e , f ne and f se have to be expressed in terms of the
values of CV centers by interpolation ( if velocity field and other properties are known, then only the
values of quantity need to be approximated).
Some terms in the transport equations require integration over the volume of a CV. The simplest second-
order accurate approximation is to replace the volume integral by the product of the mean value and the
CV volume:
Q p = qd q p (5.25)

Where qpis for the value of q at the CV center. Since all variables are available at node P, this quantity is
easily calculated, no interpolation is necessary.
An approximation of higher order requires the values of q at more locations than just the center. These
values have to be obtained by interpolating nodal values, or equivalently, by using shape function.
Fluxes through CV faces coinciding with the domain boundary require special treatment, which must be
either known or be expressed as a combination of interior values and boundary data. Since there are no
18 NCEMBT-091012
5. METHODO
OLOGY
nodes ouutside the boundary, theese approxim

mations must be based on one-sideed differencees or
extrapolattions.
One advaantage of the FVM over finnite differencce methods (F FDM) is that it does not reequire a strucctured
mesh (grids are based on some reguular distributiion of the noddes), althoughh a structuredd mesh can allso be
used. Furtthermore, thee FVM is preferable to othher methods because
b bounndary conditioons can be appplied
non-invassively, and beecause the vallues of the coonserved variaables are locaated within thhe volume eleement,
and not ata nodes or surfaces.
s Thee FVM is esspecially pow werful on coaarse non-unifform grids and a in
calculations where the mesh moves to track interrfaces or shoccks.
5.3 PARRALLEL PROCCESSING

About fivve million cellls were used ini the numerical model as shown in Figgure 13. It reqquired more thant 3
GB of meemory. Howeever, a 32-bit processor proovides addresssing only up to 4 GB of memory.m Therre is a
2 GB mem mory limitatioon per processs. Hence, wew used parallel processingg with this moodel. In FLU UENT,
the paralleel solver allow
wed us to com mpute a solutiion by using multiple
m proccesses that couuld be executted on
the same computer, orr on differentt computers inn a network. Figure 13 illustrates the parallel FLU UENT
architectuures. Parallel processing inn FLUENT innvolves an innteraction betw ween FLUEN NT, a host proocess,
and a sett of compute-node processses. FLUEN NT interacts with
w the hostt process andd the collectiion of
compute nodes
n using a utility calleed cortex, thaat manages FLLUENT's useer interface and
a basic grapphical
functions.. In our calcu ulations, we used
u four CP PUs running at
a 3.4 GHz annd a total of 121 GB of mem mory.
Four dayss of computaational time were w requiredd to complete one simulaation run for the whole UFAD U
system.
Figure 13
3. Parallel FLUEENT architecture (Fluent manual 6.3
NCEM
MBT-091012 19
6. NUMERIICAL ANALYSIS
S ON THE CAD SYSTEM
S OF THEE BTLAB
6. NUMERICA
AL ANALY
LYSIS ON
N THE CA
AD SYSTEM OF THE
T BTLA
AB
6.1 CFD
D ANALYSIS OF A SINGLEE FOUR-WAYY DIFFUSER
(a) (b)
Figuree 14. Modeling of
o a four-way squuare cone diffuseer; (a) geometry,, (b) mesh system
m
To study the effects off the diffuserr characteristiics on the disscharge air floow patterns inn the near diffuser
region, a diffuser
d modeel was configured. Figure 14 shows the geometry annd mesh system m for the fouur way
T width of bottom for thhe four way diiffuser was 2 ft. The neck of the diffuseer was
square conne diffuser. The
4 ft. To siimulate the sm
mall scale airr flow inside the diffuser, 22,782
2 tetrahhedron cells were
w used to model
m
the diffuseer.
Figure 144 illustrates the
t grid systeem and compputational doomain for thee single diffuuser. The doomain
extented as
a follows:
x-coordinate: minn (m) = 0.000000e+000, max
m (m) = 4.2667200e+000
y-coordinate: minn (m) = -1.2199200e+000, max
m (m) = 4.2267200e+000
z-coordinate: minn (m) = 0.0000000e+000, max
m (m) = 3.5337192e+000
V
Volume statisttics:
miniimum volumee (m3): 8.312303e-008
maxiimum volumee (m3): 2.0388813e-004
totall volume (m3)): 6.434677e+
+001
F
Face area statiistics:
miniimum face areea (m2): 2.616056e-005
maxiimum face arrea (m2): 3.7116122e-003
20 N
NCEMBT-091012
6. NUMERICAL ANALYSIS ON THE CAD SYSTEM OF THE BTLAB
In the computation domain, the top surface, which connects the diffuser, was considered a wall and the
other surfaces of computational domain were treated as pressure-outlet. For whole domain, 540,684 cells
were used. The supply airflow for the single diffuser was 0.098 m3/s and temperature of the supply
airflow was 12.80C while the environment temperature was 26.90C. Figure 16 shows the velocity vector
graph at the selected slice (Y= 5 ft). As the temperature of supply airflow was lower than the
environment temperature, air had a strong vertical downward flow due to the buoyancy effect. Figure 17
shows the variation of velocity magnitude as a function of X at the selected slice (Y= 5 ft). From that
graph we determined that the diffuser only affected the air flow in the range 2 ft < X< 12 ft. For any
point beyond that range, the airflow almost stayed stagnant. As the height increased, the velocity
magnitude of the points below the diffuser obviously decreased. Strong airflow was found near the
region X=4 and X =10. Figure 18 shows the path lines which were released from the top inlet of the
diffuser. A total of 80 path lines were tracked and colored based on their respective temperature. As
shown in Figure 18, it was found that the airflow was mainly concentrated on the sides with very little
airflow found at the corners. This phenomenon was identical to the actual airflow for the diffuser in the
BTlab. The temperature distribution at the selected slice is illustrated in Figure 19. As the cold supply
airflow was discharged from the diffuser, the temperature of air near the diffuser region was cooled.
Figure 20 shows the variation of air temperature as a function of X at the selected slice. Similar to
variation of the velocity magnitude with X, the air temperature of points in the range 2 ft < X< 12 ft was
obviously influenced by the diffuser. The figure also shows that when the height is smaller than 4 ft, the
gas temperature has almost the same value as that of the environment.
Figure 15. Grid system and computational domain for single diffuser characteristics study
NCEMBT-091012 21
Figure 16. Velocity vector graph at the selected slice (Y= 5 ft)
0.5
0.4 Z= 2 ft
Z= 0 ft
Velocity magnitude (m/s)
0.3
Z= 4 ft
0.2
Z= 6 ft
0.1
Z= 8 ft
0
0 2 4 6 8 10 12 14
x (ft)
Figure 17. Variation of velocity magnitude as a function of X at the selected slice (Y= 5 ft)
22 NCEMBT-091012
Figure 18. Path lines released from the top inlet of the diffuser. The path line colors are based on their respective temperature.
A total of 80 path lines was tracked
Figure 19. The temperature distribution at the selected slice (Y = 5 ft)
NCEMBT-091012 23
305
Z = 0 ft
300
Air temperature (K)

Z= 2 ft
Z= 6 ft
Z= 4 ft
295
Z= 8 ft
290
Z= 10 ft
285
0 2 4 6 8 10 12 14
X (ft)
Figure 20. Variation of air temperature as a function of X at the selected slice (Y= 5 ft)
6.2 CFD ANALYSIS OF THE BTLAB LABORATORY SPACE

The grid system and computational domain for laboratory space of the BTLab is shown in Figure 21.
Meshes with 438,472 cells, 569,722 cells, and 876,482 cells were used to verify that the computational
results were independent of the number of cells. As there was no sizable difference between the loose
and the fine mesh, the results shown here were generated with the mesh with 438,472 cells. For the
computational domain, the bottom and top walls were treated as having no heat flux. A constant
temperature of 23.9C was assumed for the side walls.
To check the flow behavior of the diffuser path lines as released from the top inlet of the diffuser were
plotted as shown in Figure 22. Here the path lines are colored by their respective air temperature. A total
of 320 path lines was tracked. Compared to the case of the single diffuser the path lines are complicated
as the diffusers interact with each other. The path lines are also affected by buoyancy force due to the
higher temperature of the side walls. The figures also indicate that the air inside the test space is well
mixed and the temperature in the region below the diffusers is quite uniform. Figure 23 plots the velocity
vectors at the selected slice (Y= 5 ft). The close-up view at section A-A discloses two vortexes in the
region below the diffuser. Because the side wall temperature is higher than that of the supply air flow,
there is an upward flow near the wall. Figure 24 depicts the temperature distribution across the selected
slice (Y=5 ft) revealing a quite uniform temperature distribution as the cold supply air is well mixed.
Figure 25 illustrates the variation of average surface air temperature and velocity magnitude as a function
of height. As these figures indicate the variation of the average surface air temperature with height is less
than 1 K. The average temperature for the whole test space is about 289.05 K (15.9C) and almost 8 K
less than the wall temperature. The curves for the average velocity magnitude show that the average air
velocity magnitude is less than 0.1 m/s and increases as Z increases.
24 NCEMBT-091012
Figure 21. Grid system and computational domain for test space of BTLab
NCEMBT-091012 25
(a) 3D view
(b) ZY plane view
(c) ZX plane view

Figure 22. Path lines released from the top inlet of the diffuser; the path lines are colored by their respective temperature
26 NCEMBT-091012
1 m/s A-A'
Y X
(a) Y= 5 ft
1 m/s
Y X
(b) Close-up view of section A-A

Figure 23. Velocity vector graph at the selected slice (Y= 5 ft)
NCEMBT-091012 27
Y X
Temperature (K):
285 287 289 290 292 294 296
Figure 24. The temperature distribution across the selected slice (Y = 5 ft)
0.1
289.4
Average surface velocity magnitude (m/s)

Average surface air temperature (K)
0.08
289.2
0.06
289
0.04
288.8 Average surface air temperature
Average surface velocity magnitude 0.02
288.6
0
0 2 4 6 8
Z (ft)
Figure 25. Variation of average surface air temperature and velocity magnitude as a function of height.
28 NCEMBT-091012
Figure 26 shows the locations of the temperature and velocity sensors in the BTlab. In the experiment,
three test planes with 13 test columns were used. Considering the symmetry of the geometry, we only
investigated the first six test columns in test plane 1 and test plane 2. Figure 27 shows the variation of air
velocity magnitude at the test positions as a function of height. In test plane 1 the velocity was much
larger than at any the other points for Z > 8 ft because these points were near the diffuser. For the points
Z< 8ft, the velocity was less than 0.1 m/s. As test plane 2 was further away from the diffusers, the
velocity was more uniform than in the test plane 1. Figure 28 shows the variation of air temperature at
the test planes as a function of height. In the test plane 1, the air temperature changed for Z >6ft, while
the air temperature for Z <6ft almost did not vary. Similar variations were found in the test plane 2.
Figure 26. Locations of temperature and velocity sensors in the BTlab
NCEMBT-091012 29
6
Z (ft)
x = 4 ft
4 x = 6 ft
x = 8 ft
x = 10 ft
2 x = 12 ft
x = 14 ft
0
-0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
Test point velocity magnitude (m/s)
(a) Test plane 1
6
Z (ft)
x = 4 ft
4 x = 6 ft
x = 8 ft
x = 10 ft
2 x = 12 ft
x = 14 ft
0
-0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
Test point velocity magnitude (m/s)
(b) Test plane 2
Figure 27. The air velocity magnitude at test planes 1 and 2
30 NCEMBT-091012
6
x = 4 ft
x = 6 ft
Z (ft)
x = 8 ft
4
x = 10 ft
x = 12 ft
2 x = 14 ft
0
286 287 288 289 290
Test point air temperature
(a) Test plane 1
6
x = 4 ft
Z (ft)
x = 6 ft
4 x = 8 ft
x = 10 ft
x = 12 ft
2 x = 14 ft
0
286 287 288 289 290
Test point air temperature
(b) Test plane 2
Figure 28. The air temperature at the test planes 1 and 2
NCEMBT-091012 31
7. CFD SIMULATION OF A UFAD SYSTEM

7.1 PREPROCESSING
A review of existing literature revealed only a very limited number of studies of UFAD systems. To the
best of our knowledge, no one has ever performed a detailed numerical simulation of UFAD with swirl
diffusers. This may be due to the complex geometry and flow features involved.
In our research, the 3-D modeling of the swirl diffuser was done by Pro/E which created the , IGES file
necessary for the meshing. The 3-D meshing for swirl diffuser was done by GAMBIT.
7.1.1 Creating a 3-D Geometry

There were no detailed dimensions available for the swirl diffusers used in the BTLab, which were
manufactured by Nailor. All the dimensions used in Pro/E modeling were measured by us. The structure
for the Nailor NFD floor swirl diffuser is shown in Figure 29, the 3-D geometry modeled by Pro/E is
shown in Figure 30.
Figure 29. Geometry of the Nailor NFD flow swirl diffuser
32 NCEMBT-091012
(a) (b)
Figure 30. 3-D model of the swirl diffuser; (a) exploded view; (b) assembled view
7.1.2 Mesh Generation

The bottom part of the swirl diffuser was simplified when input to GAMBIT for meshing. This included
the top part computational domain and a short cylinder. Figure 31 (a) shows the top part in modeling,
while Figure 31 (b) shows the corresponding computational domain.
(a) (b)
Figure 31. 3-D modeling (a) exploded view (b) assembled view
Figure 32 shows a close view of the computational mesh generated in GAMBIT for the swirl diffuser in
which tetrahedral cells were employed.
NCEMBT-091012 33
Figure 32. Meshes for a single swirl diffuser
The numerical simulation of the eight swirl diffusers as they were present in the BTLab would have
incurred computational costs that were beyond the scope of this project. The developed alternative was to
obtain the flow and thermal pattern of one swirl diffuser first and then use the simulation results as the
input boundary conditions for the eight swirl diffusers. For meshing the UFAD system with a single swirl
diffuser, 1,053,841 tetrahedral cells were used. The modeling configuration is shown in Figure 33.
Figure 33. Grid system and computational domain for modeling of a single swirl diffuser
The grid parameters, i.e., the extent of the computational domain, as shown in Figure 33 were:
X-coordinate: min (m) = -1.35, max (m) = 1.35
Y-coordinat: min (m) = -1.35, max (m) = 1.35
Z-coordinate: min (m) = -0.1, max (m) = 2.01
34 NCEMBT-091012
7.2 SIMULATION OF A SINGLE SWIRL DIFFUSER

Simulation results for a single diffuser are shown in Figures 34 through 38. Figure 34 shows a 2-D view
of velocity contours across the center of the swirl diffuser. Figure 35 shows the path lines, which are
colored depending of their respective temperature, as they were released from the diffuser. Figure 36
shows iso-surfaces at different temperatures. Figures 37 and 38 show the temperature distribution at the
center plane at different heights. These figures show that the injection of air from the swirl diffuser was
helical and the clear zone could easily be found. Determining the clear zone was a major objective of this
numerical study. The clear zone is defined as the area of the space around the diffuser outlet where the
velocity of air exceeds 50fpm (0.254 m/s).
Figure 34. Velocity contour distribution (in m/s) at slice Y=0 m
Figure 35. Path lines, which are colored depending of their respective temperature, as they were released from the outlet of the
single diffuser
NCEMBT-091012 35
T = 296 K
(22.9C) T = 297 K
(23.9C)
Iso-surfaces at different temperature values
T = 298 K T = 299 K
(24.9C) (25.9C)
Figure 36. Iso-surfaces at different temperatures
36 NCEMBT-091012
Z=0.05m Z=0.25m
Z=0. 5m Z=1m
Figure 37. Temperature (K) distributions at different heights
Figure 38. Temperature (K) distribution at the slice (Y=0 m)
NCEMBT-091012 37
7.3 IMPACT OF THE SPRAY ANGLE ON THE PERFORMANCE OF THE SWIRL DIFFUSER
Figure 39. Modeling different spray angles of the swirl diffuser
To investigate the effect of different spray angles on the air flow pattern, a numerical study of the swirl
diffuser with different spray angles was conducted. As shown in Figure 39, swirl diffusers with angles
from 3o to 7o were studied. The geometry model was been created and meshed in GAMBIT. In our
simulation, we only considered the spray flow from a single swirl diffuser. The inlet flow rate for the
swirl diffuser was 104 cfm (0.049 m3/s). The supply air temperature was set at 291.3 K (18.1C) and the
ambient air temperature was 300K (26.9C). These conditions corresponded to the one encountered in
the experimental study (Tan 2008).
Figures 40 through 43 show the flow patterns and temperature distributions from the simulation results.
The flow patterns strongly depend on the spray angle as shown in Figure 40. For a small spray angle, the
mixing zone seemed to concentrate in a very narrow column above the diffuser. However, the column is
not significantly higher for a small angle than that of the optimal angle which seemed to be about 5.
Given the very narrow mixing zone the temperature distribution in the room would most likely be very
become non-uniform.
When we increase the spray angle of the swirl diffuser, the mixing zone becomes gradually larger and a
clear zone forms. According to a previous publication (Loudermilk 1999), the clear zone for a classical
swirl diffuser is approximately 1.2 m (4 ft) high and 0.6 m (2 ft) in diameter directly above the diffuser.
Outside of the clear zone, room air velocities will be less than 0.25 m/s (50 fpm). When the spray angle
was about 5o, the clear zone from the present model was approximately within the range of the previous
study. Based on the path lines as shown in Figure 41 we postulated that the optimal spray angle would be
5o. When the spray angle was increased beyond 5 the mixing zone grew flatter until it completely
collapsed at an angle of 7 as seen in Figures 40 through 43.
Figures 42 and 43 show the temperature distributions and temperature iso-surfaces for spray angles of 3,
4, 5 and 7. As the heat transfer from the diffuser air to the space environment is controlled by the
force of the convection, all of the phenomena in the temperature distribution with a different spray angle
can be explained by the respective air flow pattern.
38 NCEMBT-091012
(a) = 3
(b) = 4
NCEMBT-091012 39
(c) = 5
(d) = 7
Figure 40. Velocity contours (m/s) at slice X=0 ft
40 NCEMBT-091012
(a) = 3
(b) = 4
NCEMBT-091012 41
(c) = 5
(d) = 7
Figure 41. Path lines released from the outlet of the swirl diffuser (colored by path line ID)
42 NCEMBT-091012
(a) = 3 (b) = 4
o o
(c) = 5 (d) = 7
o o
Figure 42. Temperature iso-surfaces at 299K (25.9C) for spray angles of 3, 4, 5 and 7
NCEMBT-091012 43
(a) = 3
44 NCEMBT-091012
(b) = 4
NCEMBT-091012 45
(c) = 5
46 NCEMBT-091012
(d) = 7
Figure 43. Temperature (K) distributions at different heights for spray angles of 3, 4, 5 and 7
NCEMBT-091012 47
7.4 CFD SIMULATION OF THE BTLAB UFAD SYSTEM WITHOUT THERMAL LOAD
One of the major objectives of this project was to develop a CFD model of the UFAD system of the
BTLab and compare the simulation results with experimental data. Having developed a model of a single
swirl diffuser as installed in the BTLab, a model of the UFAD system of the BTLab was developed.
Previous work (Loudermilk 1999) defined a clear zone around a swirl diffuser. The clear zone is the
mixing zone above the diffuser outlet where the air velocity exceeds 50 fpm. To check if our developed
model produces results that are in the range of the data reported by Loudermilk, we generated flow
patterns as seen in Figure 44. Here, the path lines were colored according to their respective air
temperature. Unlike in the case of a single swirl diffuser, the interactions between the diffusers make the
fluid flow more complicated. To determine the clear zone of each swirl diffuser, velocity contours across
the slice through the center line of a diffuser row (Y=0) are shown in Figure 45. As illustrated in Figure
45, the clear zone from the present model was within the range reported by previous work. Here, the
height of the clear zone for each diffuser was around 1 m and 0.6 m in diameter. Figure 46 shows the
velocity vector graph in the slice (Y=0). Since the slot of the swirl diffuser is very small, high air velocity
can be found near the slot, and the flow pattern is very complicated.
Figure 47 shows the temperature distribution at the slice Y=0. Here, we observed similar behavior as in
the case of the single diffuser, i.e., the mixing occurs outside the clear zone. We also observed some
stratification along the Z-axis, i.e., the height above the floor. The degree of stratification seemed to be
within the empirical values that were previously reported.
To analyze the velocity and temperature distributions, the velocity and temperature as functions of x for
different heights at the slice Y=0 are plotted in Figures 48 and 49. Both the temperature and velocity
distribution displayed patterns that seemed to be within those reported by previous studies. Figure 50
illustrates the average temperature and velocity as a function of height.
Although the temperature and velocity distributions seemed to be well within the values reported by
previous work, only a comparison with the actual experimental data could provide verification of the
developed model. A detail comparison is provided in Chapter 8. Comparison of Simulation Results With
Experimental Data.
48 NCEMBT-091012
(a) Full view
(b) View of a single diffuser

Figure 44. Path lines released from the outlet of the swirl diffuser. The path lines are colored according to their respective
temperature (K)
NCEMBT-091012 49
Figure 45. Velocity (m/s) distribution at the slice Y=0 m
(a) Full view
(b) Zoom View

Figure 46. Velocity vector graph (m/s) at the slice Y=0 m
50 NCEMBT-091012
Figure 47. Temperature (K) distribution at the slice Y=0 m
z = 0.5 m
z=1m
z = 1.5 m
0.8
z=2m
0.6
0.4
0.2
0
0 2 4 6 8
X (m)
Figure 48. Velocity (m/s) as a function of x for different heights at the slice Y=0
NCEMBT-091012 51
297
z = 0.5 m
z=1m
z = 1.5 m
296
z=2m
z = 2.5 m
Temperature (K) z=3m
295
294
293
292
0 2 4 6 8
X (m)
Figure 49. Temperature (K) as a function of x for different heights at the slice Y=0
295 0.12
Average temperature
Average velocity magnitude (m/s)

Average velocity magnitude
Average temperature (K)
294.5
0.08
294
0.04
293.5
293 0
0.5 1 1.5 2 2.5 3
Z (m)
Figure 50. Average temperature and velocity as a function of height
52 NCEMBT-091012
7.5 CFD SIMULATION OF THE BTLAB UFAD SYSTEM WITH THERMAL LOAD
The UFAD system in the BTLab was tested with different thermal loads. The thermal load was created
by twelve electric resistance heaters which were located on the floor. The total power draw for all the
heaters was 4192W (Tan 2008).
A CFD model which incorporated the floor heaters was developed, as shown in Figure 51. To simulate
the heaters, two kinds of boundary conditions could have been used. The boundary condition of the
heaters could have been taken as a constant temperature, while it also could have been applied as a
constant heat flux. Here, the constant temperature condition was adopted in our model.
The boundary conditions for the simulation were derived from the experimental data and are listed in
Table 1. Figure 52 shows the schematic of the BTlab and the distribution of the thermal loads. About
five million mesh cells were used to simulate this problem. This was determined to be sufficient to obtain
results which would be independent of the number of mesh cells. The size of the mesh exceeded the
capability of a single CPU, the computations were executed with four CPUs running at 3.4 GHz featuring
12 GB of memory configured as a single PC. It took four days to complete one simulation run.
Figure 51. Schematic design of UFAD system of BTLab with floor heater
NCEMBT-091012 53
Table 1. Boundary conditions for modeling the UFAD system of the BTLab with a thermal load
Boundary Name Value

East Wall Temperature 76.1F (297.7 K)
South Wall Temperature 76.0F (297.6)
West Wall Temperature 76.1F (297.7 K)
North Wall Temperature 75.4F (297.3 K)
Four Wall Temperature 76.0F (297.6 K)
Ceiling Temperature 76.3F (297.8 K)
Floor Temperature 71.4F (295.0 K)
Floor Heater Temperature 130.0F (327.6 K)
Floor Supply Air Temperature 64.3F (291.1 K)
Floor Supply Diffuser Airflow Rate 1064 cfm (30.13 m/min)
Figure 52. Schematic of the BTLab with distribution of the thermal loads (Tan 2008)
The airflow pattern in this case was expected to differ from the one without a thermal load. The floor
heaters induce buoyancy forces which should move the air higher within the laboratory space. To check
the airflow pattern, we plotted the path lines as they were released from the outlet of the swirl diffusers
are shown in Figure 53. The path lines were colored according to their respective path line ID. Unlike
the single swirl diffuser case, both of the interactions among the diffusers and the effect of heaters were
expected to further complicate the fluid flow.
54 NCEMBT-091012
Figure 53. Path lines as released from the outlet of the swirl diffuser; the path lines are colored according to their respective path
line ID
Figure 54. Temperature (K) distribution at the slice Y =5ft
NCEMBT-091012 55
Figure 54 shows the temperature distribution in the slice Y=5 ft. Comparing the case with thermal loads
to the one without, i.e., as shown in Figure 54 compared to Figure 46, there does not seem to be much
difference. The vertical air movements were different but the actual temperature distributions were
within very similar ranges. Both simulations showed relative uniform temperature mixing across the
whole space. The modeled values were within the boundaries of vertical temperature differences as set
forth in ANSI/ASHRAE Standard 55-2004 (ASHRAE 2004).
56 NCEMBT-091012
8. COMPARISON OF SIMULATION RESULTS WITH EXPERIMENTAL DATA
8. COMPARISON OF SIMULATION RESULTS WITH

EXPERIMENTAL DATA
One of the main goals of this project was to verify the computational models using data from National
Center for Energy Management and Building Technologies (NCEMBT) Task 2. In this project, we
compared the simulation results with the experimental data at zero thermal loads (Test 10) and half load
(Test 11). The experimental data were obtained from the BTLab, which is located at the University of
Nevada, Las Vegas.
Figure 55 shows the schematic of the test grid in the BTLab. From the experiments, we obtained velocity
and temperature records for each test point. Due to the symmetry structure of the BTLab, we just
compared our numerical results with experimental data in the region inside the green box as indicated in
Figure 55.
NCEMBT-091012 57
Figure 55. Schematic of test grid in the BTLab; the data obtained from grids in the region with the green box were used for
comparison of the modeled values versus the experimental ones
8.1 COMPARISON OF SIMULATION RESULTS WITH EXPERIMENTAL DATA FOR ZERO THERAML
LOAD
Figure 56 through Figure 79 plot the simulation results and the experimental data side by side for zero
thermal load.
A reasonable good agreement was achieved for temperature between numerical results and experimental
data. The simulated data generally followed the experimental trends. The overall difference between the
numerical results and the experimental ones is about 10 percent. In absolute terms this represented about
1K or 2F.
Unfortunately, the same cannot be said for velocity. Although the overall difference between the
simulated values and the experimental ones was about 25 percent, the simulated data did not follow the
experimental trends. For most grid points the simulation predicted velocities between 0 and 1m of height
which generally were significantly higher than the experimental results. The numerical velocity data
showed trends, i.e., modulations in velocity, which were not found experimentally. Interestingly, a good
agreement was achieved for grid points A6 through D6.
58 NCEMBT-091012
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)
0.2 Experimental data
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
Figure 56. Comparison of temperature and velocity in test grid A1 for zero thermal load
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
Figure 57. . Comparison of temperature and velocity in test grid B1 for zero thermal load
NCEMBT-091012 59
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
Figure 58. Comparison of temperature and velocity in test grid C1 for zero thermal load
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
Figure 59. Comparison of temperature and velocity in test grid D1 for zero thermal load
60 NCEMBT-091012
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
Figure 61. Comparison of temperature and velocity in test grid B2 for zero thermal load
NCEMBT-091012 61
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)

Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)

Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
62 NCEMBT-091012
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)

Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
NCEMBT-091012 63
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)

Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)

Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
64 NCEMBT-091012
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)

Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
NCEMBT-091012 65
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)

Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
66 NCEMBT-091012
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)

Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
NCEMBT-091012 67
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
68 NCEMBT-091012
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290 0.5 1 1.5 2 2.5

Height (m)
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)

Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
NCEMBT-091012 69
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)

Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)

Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
70 NCEMBT-091012
8.2 COMPARISON OF SIMULATION RESULTS WITH EXPERIMENTAL DATA FOR HALF THERMAL
LOAD
Figure 80 through Figure 103 plot the simulation results and the experimental data side by side for half
thermal load.
A good agreement for both temperature and velocity was not achieved between numerical results and
experimental data. The simulated data for both parameters generally did not follow the experimental
trends. The difference between the numerical results and the experimental ones for temperature in
absolute terms was as much as 3K or 5.4F. That difference was higher than the vertical temperature
gradient allowed by ANSI/ASHRAE Standard 55-2004.
The difference in velocity between numerical and experimental results was even greater than for
temperature. The difference was as much as 0.2 m/s or 39.4 fpm. This difference alone exceed the draft
limit as set forth in ANSI/ASHRAE Standard 55-2004.
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
Figure 80. Comparison of temperature and velocity in test grid A1 for half thermal load
NCEMBT-091012 71
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
Figure 81. Comparison of temperature and velocity in test grid B1 for half thermal load
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)

Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
Figure 82. Comparison of temperature and velocity in test grid C1 for half thermal load
72 NCEMBT-091012
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290 0.5 1 1.5 2 2.5

Height (m)
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
Figure 83. Comparison of temperature and velocity in test grid D1 for half thermal load
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
NCEMBT-091012 73
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290 0.5 1 1.5 2 2.5

Height (m)

Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
74 NCEMBT-091012
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290 0.5 1 1.5 2 2.5

Height (m)
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290 0.5 1 1.5 2 2.5

Height (m)

Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
NCEMBT-091012 75
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)

Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
76 NCEMBT-091012
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290 0.5 1 1.5 2 2.5

Height (m)

Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
NCEMBT-091012 77
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290 0.5 1 1.5 2 2.5

Height (m)
0.3
Experimental data
Numerical data
0.2
0.1
0
0.5 1 1.5 2 2.5
Height (m)
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290 0.5 1 1.5 2 2.5

Height (m)

Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
78 NCEMBT-091012
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290 0.5 1 1.5 2 2.5

Height (m)
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290 0.5 1 1.5 2 2.5

Height (m)
0.3
Experimental data
Numerical data
0.2
0.1
0
0.5 1 1.5 2 2.5
Height (m)
NCEMBT-091012 79
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290 0.5 1 1.5 2 2.5

Height (m)
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290 0.5 1 1.5 2 2.5

Height (m)

Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
80 NCEMBT-091012
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290 0.5 1 1.5 2 2.5

Height (m)
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290 0.5 1 1.5 2 2.5

Height (m)

Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
NCEMBT-091012 81
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290 0.5 1 1.5 2 2.5

Height (m)
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290
0.5 1 1.5 2 2.5
Height (m)
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
82 NCEMBT-091012
300
298
Temperature (K)
296
294
Experimental data
292 Numerical data
290 0.5 1 1.5 2 2.5

Height (m)
Numerical data
0.1
0
0.5 1 1.5 2 2.5
Height (m)
NCEMBT-091012 83
9. CONCLUSIONS
9. CONCLUSIONS
The major objectives of this project were to a) develop CFD models of CAD and UFAD systems,
particularly of the type of diffusers used in the Buildings Technology Laboratory (BTLab) experiments,
conducted under NCEMBT Task 2; b) verify the developed models using the experimental data from
Task 2; and c) provide detailed information regarding velocity and temperature concentrations and
turbulence intensity throughout the domain of interest.
It was hoped that by developing a verified CFD model of a UFAD system, numerical simulations could
supplement experimental results and eventually a model could be develop to design UFAD system for
real world applications.
The objectives of the projects were only partially met. Models of the CAD and UFAD system as installed
in the BTLab were developed and initial simulations indicated good agreement with previous work.
Detailed comparisons between the temperature and velocity values from numerical simulations and
experiments, however, revealed significant differences for both zero and half thermal load. In case of the
half thermal load the differences exceeded the limits for comfort conditions as set forth in
ANSI/ASHRAE Standard 55-2004.
Possible reasons for the differences between the numerical and experimental data are:
Size of the computational domain. Although the BTLab was modeled using about 5,000,000 cells,
this may be insufficient to model the air distribution in the mixing zone of each individual swirl
diffuser. Running simulations with significantly higher cell count, i.e., tighter mesh, was not feasible
to hardware limitations and budgetary constraints.
Air leakage. According to the experimental reports, the UFAD floor of the BTLab experienced about
12 percent category 1 leakage, i.e., uncontrolled leakage from the UFAD plenum into the conditioned
space. His air leakage was not accounted for in the numerical model.
Non-uniformity for the inlet mass flow rate for the swirl diffuser. Our model assumed that the flow
rate in each swirl diffuser was equal.
Experimental blind regions. The experimental data was recorded for data points that were 12 inches
apart on the x and y axes and 6 inches apart on the z-axis. This leaves large amounts of the
experimental space were no data were recorded. The computational grid was significantly finer than
the experimental one.
84 NCEMBT-091012
10. REFERENCES
10. REFERENCES
ASHRAE Inc. 1997. ASHRAE Handbook of Fundamentals, American Society of Heating, Refrigerating
and Air Conditioning Engineers, Inc., Atlanta, Georgia.
ASHRAE Inc. 2004. Thermal Environmental Conditions for Human Occupancy. ANSI/ASHRAE
Standard 55-2004.
Bauman F. 1999. Giving occupants what they want: guidelines for implementing personal environmental
control in your building. Proceedings at World Workplace 99, Los Angeles, CA.
Bauman F and Webster T W. 2001. Outlook for underfloor air distribution. ASHRAE Journal, 43(6):18
27.
Bjerg B, Svidt K, Zhang G, Morsing S, and Johnsen J. 2002. Modeling of air inlets in CFD prediction of
airflow in ventilated animal houses, Computers and Electronics in Agriculture, 34: 223-235.
Chen Q, and Xu W. 1998. A zero-equation turbulence model for indoor airflow simulation, Energy and
Building. 28:137-144.
Chow W.F., and Fung W.F. 1996. Numerical studies on indoor air flow in the occupied zone of ventilated
and air-conditioned space, Building and Environment, 31: 319-344.
Dorer V, Weber A. 1999. Air, contaminant and heat transport models: integration and application, Energy
and Buildings, 30: 97-104.
Emmerich S. 1997. Use of computational fluid dynamics to analyze indoor air quality issues. NISTIR
5997, Building and Fire Research Laboratory, National Institute of Standards and Technology,
Gaithersburg, MD.
Fan Y. 1995. CFD modeling of the air and contaminant distribution in rooms, Energy and Buildings,
23:33-39.
Fanger P. 1970. Thermal Comfort analysis and applications in environmental engineering, McGraw-Hill,
New York.
FLUENT, Inc. 2005. FLUENT 6.2 Users Guide, Lebanon, NH.
Fukao H, Oguro M, Ichihara M, and Tanabe S. 2002. Comparison of underfloor vs. overhead air
distribution systems in an office building. ASHRAE Transactions, 104(1): 64-76.
Gadgil A, Finlayson E, Hong K, and Sextro R. 1999. Commercial CFD software capabilities for modeling
a pulse release of pollutant in a large indoor space. Proceedings of Indoor Air 99, Edinburgh.
Gan G, and Awbi H. 1994. Numerical Simulation of the indoor environment, Building and Environment,
29(4): 449-459.
NCEMBT-091012 85
10. REFERENCES
Hirnikel D, Lipowicz P, and Lau W. 2002. Predicting contaminant removal effectiveness of three air
distribution systems by CFD modeling. ASHRAE Transactions, 108(1), pp. 350-359.
Joseph C and Apple L. 2001. Analysis and energy simulation of a gymnasium, Building and
Environmental, 36: 351-358.
Jones W, and Launder B. 1972. The prediction of laminarization with a two-equation model of
turbulence. International Journal of Heat and Mass Transfer, 15: 214301.
Juan A, Daniel B, Marcelo R, and Benoit O. 2004. Air flow modeling in a computer room. Building and
Environmental, 39:1393-1402
Kim S, Choudhury D, and Patel B. 1997. Computations of complex turbulent flows using the commercial
code FLUENT. In: Proceedings of the ICASE/LaRC/AFOSR Symposium on Modeling Complex Turbulent
Flows, Hampton, VA.
Lim J, Jo J, Kim Y, Yeo M and Kim K. 2006. Application of the control methods for radiant floor cooling
system in residential buildings, Building and Environment, 41: 60-73.
Lin Y, Linden P. 2005. A model for an under floor air distribution system, Energy and Buildings 37: 399-
409.
Loudermilk K. 1999. Underfloor air distribution solutions for open office applications, ASHRAE
Transactions, 105(2): 605613.
Niu J, and Van D. 1992. Two-dimensional simulation of airflow and thermal comfort in a room with
open-window and indoor cooling systems, Energy and Buildings, 18: 65-75.
Mazzei P, Minichiello P, and Plma D. 2005. HVAC dehumidification systems for thermal comfort: a
critical review, Applied Thermal Engineering, 25: 677-707.
Posner J, Buchanan R and Dunn-Rankin D. 2003. Measurement and prediction of indoor air flow in a
model room, Energy and Buildings, 35:515-526.
Rouaud O and Havet M. 2002. Computation of the airflow in a pilot scale clean room using k-
turbulence models, International Journal of Refrigeration, 25: 351361.
Rohles F and Nevins R. 1971. The nature of thermal comfort for sedentary man. ASHRAE Transactions,
77(1): 239-244.
Shih T, Liou, W, Shabbir, A, Yang, Z, and Zhu J. 1995. A new ke eddy viscosity model for high
Reynolds number turbulent flows. Computers & Fluids, 24(3): 227238.
Son H, Luis R, and Muhammad M. 2005. Analysis of thermal comfort and contaminant removal in a
office room with underfloor air distribution system. Proceeding of HT2005, 2005 ASME Summer Heat
Transfer Conference, July17-22, San Francisco, California, USA.
86 NCEMBT-091012
10. REFERENCES
Tan L, Landsberger BJ and Novosel D. 2008. Underfloor Air Distribution - An Experimental

Comparison Of Air Diffusion Performance Between Underfloor Air Distribution And Conventional
Overhead Air Distribution Systems. Final Report NCEMBT-080801. Alexandria, Virginia: National
Center for Energy Management and Building Technologies
Webster T, Bauman F, and Reese J. 2002. Underfloor air distribution: thermal stratification. ASHRAE
Journal, 44(5): 28-33.
Webster T, Bannon R, and Lehrer D. 2002. Teledesic broadband center field study. Center for the Built
Environment (CBE), Summary Report April 2002.
Woods J. 2004. What real-world experience says about UFAD alternatives. ASHRAE Journal, 46(2): 3-15.
Zhao B, Li X, and Yan Q. 2003. A simplified system for indoor airflow simulation, Building and
Environment, 38: 543 552.
NCEMBT-091012 87
NATIONAL CENTER FOR ENERGY MANAGEMENT AND BUILDING TECHNOLOGIES
601 NORTH FAIRFAX STREET, SUITE 250
ALEXANDRIA, VA 22314
WWW.NCEMBT.ORG

Chen y CFD Study of An Ufad System Ncembt-091012

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Chen y CFD Study of An Ufad System Ncembt-091012

Încărcat de

Drepturi de autor:

Formate disponibile

FINAL REPORT NCEMBT-091012

NATIONAL CENTER FOR ENERGY MANAGEMENT

Yitung Chen, Ph.D.

U.S. Department of Energy

This report was prepared for the U.S. Department of Energy

NATIONAL CENTER FOR ENERGY MANAGEMENT AND BUILDING TECHNOLOGIES CONTACT

Figure 1. The conventional design process of an HVAC system

Figure 2. The design process of HVAC system with CFD simulation

4. PROBLEM DESCRIPTION AND EXPERIMENTAL SETUP

Figure 3. Computational domain of the BTLab

4.2 EXPERIMENTAL SETUP

Figure 5. Floor Supply Air Diffusers In The BTlab

Figure 7. Section View

UFAD system supply duct

Main supply duct

Fan room, w/, return

Figure 8. HVAC System Overview

Figuree 10. Distributioon Duct In The Unnderfloor Plenum

Figure 11. Square cone diffuser and swirl diffuser in BTLab

where C1 = max[0.43; + 5], C 2 = 1.9 , k = 1.0 , = 1.0 , = S

hc = 2.38(Tcl Ta ) 0.25 (5.16)

1.00 + 1.29 I cl I cl 0.078 m 2 K/W

5.2 NUMERICAL PROCEDURE

fdS = fdS (5.21)

Where f e is the value of f at the center of face S e .

nodes ouutside the boundary, theese approxim

5.3 PARRALLEL PROCCESSING

Figure 19. The temperature distribution at the selected slice (Y = 5 ft)

Air temperature (K)

6.2 CFD ANALYSIS OF THE BTLAB LABORATORY SPACE

(b) ZY plane view

(c) ZX plane view

(b) Close-up view of section A-A

Average surface velocity magnitude (m/s)

288.8 Average surface air temperature

Average surface velocity magnitude 0.02

Figure 26. Locations of temperature and velocity sensors in the BTlab

Figure 27. The air velocity magnitude at test planes 1 and 2

7. CFD SIMULATION OF A UFAD SYSTEM

7.1.1 Creating a 3-D Geometry

Figure 29. Geometry of the Nailor NFD flow swirl diffuser

7.1.2 Mesh Generation

Figure 32. Meshes for a single swirl diffuser

7.2 SIMULATION OF A SINGLE SWIRL DIFFUSER

Figure 34. Velocity contour distribution (in m/s) at slice Y=0 m

Iso-surfaces at different temperature values

Figure 36. Iso-surfaces at different temperatures

Figure 38. Temperature (K) distribution at the slice (Y=0 m)

Figure 39. Modeling different spray angles of the swirl diffuser

(a) Full view

(b) View of a single diffuser

Figure 45. Velocity (m/s) distribution at the slice Y=0 m

(a) Full view

(b) Zoom View

Figure 47. Temperature (K) distribution at the slice Y=0 m

Average velocity magnitude (m/s)

Figure 50. Average temperature and velocity as a function of height

Boundary Name Value

Figure 54. Temperature (K) distribution at the slice Y =5ft

8. COMPARISON OF SIMULATION RESULTS WITH

292 Numerical data

0.2 Experimental data

292 Numerical data