You are on page 1of 11

Building and Environment 38 (2003) 521 – 531

www.elsevier.com/locate/buildenv

A new duct design software tool


E.H. Mathews∗ , D.T. Claassen
Centre for Research and Commercialization, University of Potchefstroom, PO Box 2156, Faerie Glen 4, 0043, South Africa

Received 30 October 2001; received in revised form 4 December 2001; accepted 25 July 2002

Abstract

This article describes the features and operation of a new duct design software tool. The tool was used in a case study to establish its
practicality and performance. This tool is unique in the sense that it employs an optimisation method called the T-method. As far as the
authors are aware no commercial duct design software package incorporates optimisation techniques. To enhance user friendliness the
program also features a CAD interface for data input and output.
By using the new tool an existing duct system was redesigned with the same constraints and speci3cations set by the original designer.
The design program managed to optimise the 30-section system in less than 10 s on a 50 MHz 486 personal computer. In comparison
to the existing system the redesigned system yielded savings of 8% on the duct material cost while the energy cost was lowered by 3%.
The life-cycle cost decreased by 5%.
The redesigned system has nine transition 3ttings compared to the 3ve of the original system, which will reduce the 3nal savings.
Stability problems were encountered in a few sections that had relatively large C-coe9cients, which were referenced to upstream velocities.
This led to grossly oversized ducts. The problem was overcome by 3xing the diameters of the few problem sections at realistic values.
This problem was analysed in a previous article by these authors (Building Environ. 33(4) (1998) 173). It must also be said that for this
case study the T-method was outperformed marginally by the equal friction method, a traditional duct design method.
Notwithstanding the problem with the optimisation method itself the new duct design tool speeds up the design process and provides
a user-friendly way of designing duct systems.
? 2003 Elsevier Science Ltd. All rights reserved.

1. Introduction program that makes use of them. Already in 1991 a need for
such a program was stated by the inBuential ASHRAE or-
Most commercial duct design software packages use con- ganization in their Applications Handbook [3]. A literature
ventional design methods like equal friction and static re- study suggested that the new T-method is the most promis-
gain. The duct design program that holds the biggest market ing in terms of robustness and speed and was therefore
share in the world is the Carrier e20-ii program [2]. It o@ers chosen for the new duct design computer tool.
the user both the static regain and the equal friction methods, Section 2 gives a brief overview of the theory of the
which can be used in combination. It also has an optional T-method. A previous article by these authors discusses
CAD interface that simpli3es the data input and produces a the theory of the T-method in detail [1]. The duct design
CAD drawing of the 3nal design. program follows a simple 3ve-step process to design an
Conventional methods cannot produce optimum designs. air-distribution network. Section 3 explains this methodol-
Furthermore, many conventional procedures, including the ogy by applying the computer program to simple example.
popular static regain method, are unable to deliver a duct Section 4 reports on a duct design case study. It is unique
system design that balances at the speci3ed airBows. Op- because it involves a practical design problem—most
timisation procedures do not share these shortcomings, examples one 3nds in literature are textbook problems with
yet a literature survey showed that there is no commercial little or no bearing on reality. The air handling system for
the Health & Racquet Club located in Die Strand, South
∗ Corresponding author. Tel.: +27-12-991-5568; fax: +27-12-991- Africa was used as basis for the investigation. The aim of
5716. the study was to redesign the system using the new computer
E-mail address: emathews@researchtoolbox.com (E.H. Mathews). design tool. In doing so the practicality and e9ciency of

0360-1323/03/$ - see front matter ? 2003 Elsevier Science Ltd. All rights reserved.
PII: S 0 3 6 0 - 1 3 2 3 ( 0 2 ) 0 0 1 3 4 - 8
522 E.H. Mathews, D.T. Claassen / Building and Environment 38 (2003) 521 – 531

has the same hydraulic and 3rst cost characteristics as the


Nomenclature whole system.
Cc C-coe9cient of a “branch” or “straight” Optimum fan pressure calculation: The optimum pres-
section referenced to the velocity of the sure drop is established for the root section.
“common” section (dimensionless) System expansion: The optimum pressure is distributed
Ci C-coe9cient of a “branch” or “straight” between the various sections in the network. Duct sizes are
section referenced to its own velocity calculated in compliance with the distributed pressure drops.
(dimensionless)
Abbreviations 3. A simple ve-step duct design methodology
TB Label that refers to a terminal base object
The design process is performed in 3ve steps by using
B Label that refers to an ordinary base object
this new duct design tool. For the :rst step the user prepares
a drawing (in dxf format) of the layout of the system. The
the design program and the optimisation algorithm were user draws a centre-line sketch of the duct network, placing
established. markers along the lines to indicate air terminal positions.
This drawing is placed on a separate drawing layer called
Input Layer. The centre-line sketch can therefore be super-
2. Theory of the T-method imposed on the drawing of the building plan. The 3rst step
is completed when the user has imported the dxf input 3le
This section gives a qualitative overview of the theory into the design tool. In the second step the program uses the
of the T-method as it was presented by authors Tsal et al. information in the dxf 3le to identify duct sections in the
[4,5]. For a mathematical understanding of this method it is drawing and to set up a computational structure necessary
imperative that the reader study the relevant papers of Tsal for the design calculations. It also labels the duct sections for
et al. [4,5]. the third step where the user speci3es certain information
like the terminal airBows and constraints for the individual
2.1. System costs duct sections. Duct design calculations are performed in the
fourth step and :nally an output dxf 3le is written where
The 3rst step in discussing the theory is to determine the the duct sections are drawn to scale.
objective function. In the case of duct design problems the To illustrate this process the program was applied to the
objective function to be minimised is the life-cycle cost, design problem used by Tsal et al. [5]. Fig. 1 portrays the
which is the total cost of the duct system over its life span dxf input 3le prepared by the user. Note the markers and
expressed in current value terms. the circle, indicating the positions of the air terminals and
The life-cycle cost of a duct system consists chieBy of the root node, respectively. The user then turns to the main
two components namely the initial and the running costs. screen of the duct design tool where he/she speci3es the
The former includes the cost of space occupied by ducting input and output dxf 3le paths, selects the correct design
and equipment and the installed cost of the air distribution and drawing options and sets a number of relevant defaults
system itself including the ducting, fans, sound-absorbing (Fig. 2). The user presses the Start button to commence the
equipment, control systems, supply and return grilles, dif- 3ve steps listed in the Project status box. Each item in the
fusers, etc. The running cost is the sum of the energy, main-
tenance and operational costs. It is obvious that some of
these costs would be di9cult to determine. However, for
the sake of 3nding the optimum duct sizes only those com-
ponents, which are a function of the duct sizes, need to be
included in the objective function. For a complete analysis
of the cost components in the objective function the reader
is referred to the paper of Tsal et al. on the theory of the
T-method [4]. The only components considered in the new
duct design tool are the installed cost of the ducting and the
energy cost over the lifetime of the system.

2.2. General optimisation procedure

The T-method follows three major steps.


System condensing: The entire network is condensed into
a single imaginary section, called the root section, which Fig. 1. Input drawing for duct design tool.
E.H. Mathews, D.T. Claassen / Building and Environment 38 (2003) 521 – 531 523

Fig. 2. Main screen of duct design tool. Fig. 4. Base object data screen.

Fig. 5. Output drawing generated by duct design tool.


Fig. 3. Duct network data screen.

list is checked after the relevant step has been completed.


After the computational data structure has been created objects, that is why the airBow for B5 is greyed, which pre-
(step 2) the program displays the Duct network data screen vents any editing of the Bow. A terminal base object is de-
(Fig. 3). It shows all the lines that were drawn on In- 3ned as equipment that is situated at the terminating points
put layer. In the right-hand corner of the screen the user of the duct system e.g. air di@users.
can turn on the labels of the objects he wants to see. If one At the bottom half of the screen the user can add, edit or
of the 3ve buttons in the Object data box is pressed a list of delete element objects from the base object. An element ob-
all the objects in that category appears next to the buttons. ject is de3ned as a constituent of a base object. The left-hand
The user can then double-click with his mouse on any one list box shows the current element objects of B5. The two
of the items in order to enter certain information for that straight objects were created from the information in the dxf
speci3c object. input 3le while the user added a generic element object. A
Fig. 4 shows the screen that would appear if the user generic element has a 3xed C-coe9cient, which can be set
has double-clicked on the B5 base object. A base object is by the user.
de3ned as a duct run which has no change in shape, area, The user presses the OK button of the Duct network data
Bow, and friction coe9cient. screen (Fig. 3) after all the design data has been entered. The
After the user has speci3ed the shape of the base ob- program then continues to execute the T-method calculations
ject, the maximum and minimum dimension constraints are after which it prompts the user to open the output dxf 3le in
selected. A dimension can also be pre-selected or 3xed at a the drawing program. A portion of the 3nal output drawing
certain value. AirBows are only speci3ed for terminal base created by the design tool is shown in Fig. 5.
524 E.H. Mathews, D.T. Claassen / Building and Environment 38 (2003) 521 – 531

Fig. 6. Health and Racquet Club air conditioning layout.


E.H. Mathews, D.T. Claassen / Building and Environment 38 (2003) 521 – 531 525

Table 1
Economic data

Parameter Value Source of information and/or assumptions

Energy demand cost, Ed (R/kVA) 382.8 Die Strand Municipality


Power factor, pf 0.8 Typical value
Unit energy cost, Ec (R/kWh) 0.05628 Die Strand Municipality
Operational time, Y (h) 6300 Design speci3cations set by air conditioning consulting
engineer: 18 h a day, 7 days a week, 50 weeks a year.
Expected life span, a (yr) 15 Good maintenance procedures are assumed.
Annual interest rate, air (%) (the rate at which the 17 This 3gure should ideally be obtained directly from the
building owner is able to grow his money.) building owner. The author got it from the consultant
engineer responsible for the project.
Annual electricity escalation rate, aer (%) 5 Die Strand Municipality
Fan e9ciency, f (%) 77.5 Air handling unit manufacturer data
Motor and drive e9ciency, m (%) 77.5 Air handling unit manufacturer data
Installed cost per unit area, Sd (R=m2 ) 112.4 Installed cost of existing system divided by surface area
of existing design.

Table 2
Miscellaneous data

Parameter Value Source of information/assumptions

T , Temperature (K) 295 Typical temperature in Die Strand


v Kinematic viscosity (m2 =s) 1.5215e-5 Heat Transfer by Mills [6]
, Density (kg=m3 ) 1.1985 Heat Transfer by Mills [6].
, Duct inside surface roughness (m) (Ductwork has spiral seams) 0.09e-3 ASHRAE Fundamentals Handbook (1993), [7, p. 32.5, Table 1].

4. Introduction to case study for the airBow at each terminal point. This will be discussed
in the next sections.
The air distribution system, as designed by the air condi-
tioning consultants, is shown in Fig. 6. (The small system
that serves the aerobic hall is not considered in this study.) 5.1. Using the new duct design tool to analyse the
Tables 1 and 2 list the design parameters. existing system
The goal of this case study is to improve this design in
terms of reducing the life-cycle cost while maintaining the
With the new duct design tool the design process of
same layout, constraints and design speci3cations including
any duct system is performed in 3ve steps. However, when
the type of 3ttings, air quantities and the maximum air speed.
analysing a system, step 5, which is to produce a CAD draw-
The only variables in this optimisation problem are the duct
ing of the 3nal results, becomes unnecessary.
diameters. (The ducts are all round and not insulated.) By
Fig. 6 portrays the existing air conditioning system and
adjusting them in an optimal manner it is hoped that the
is a printout of the dxf 3le received from the consultant
life-cycle cost of the system will be reduced.
engineer. In step 1 a centre-line sketch is drawn on top
of this system but on a di@erent drawing layer. Fig. 7
shows the duct centre lines along with only a few selected
5. Evaluating the existing design layers to avoid crowding the 3gure. Note the markers and
the circle indicating the positions of the air terminals and
In order to judge the results of the system the capital, the root node, respectively. The 3rst step is completed when
running and life-cycle costs have to be established. For the the user has imported the appended dxf input 3le into the
existing system (shown in Fig. 6) the installed cost of the design tool. In step 2 the program uses the information in
ducting is known to be R 30751. From this 3gure we can the dxf 3le to identify duct sections in the drawing and to
calculate Sd, the installed cost per square metre ducting. set up a computational structure needed for the design cal-
We will assume that this 3gure also holds for the optimised culations. Fig. 8 shows which duct sections are represented
system. To calculate the operating cost of the existing system by which base objects. Note that some base objects include
over the expected system life the fan total pressure, PFan, more than one line segment for, example B7. This is con-
is required. To calculate Pfan we will use the design values sistent with the de3nition of a base object, namely that it
526 E.H. Mathews, D.T. Claassen / Building and Environment 38 (2003) 521 – 531

Fig. 7. Input 3le.


E.H. Mathews, D.T. Claassen / Building and Environment 38 (2003) 521 – 531 527

Fig. 8. Labels of base and terminal objects.


528 E.H. Mathews, D.T. Claassen / Building and Environment 38 (2003) 521 – 531

Fig. 9. Output 3le.


E.H. Mathews, D.T. Claassen / Building and Environment 38 (2003) 521 – 531 529

represents a duct run which has no change in shape, area, Table 3


Bow or friction coe9cient. Step 3 involves specifying the
Base Flow Length Diameter
design airBows for the terminal base objects and the 3xed label [m3 =s] [m] [m]
dimensions for the ordinary base objects. Table 3a and b
contain the duct dimensions and design Bows. In the case (a) Input data
B1 9.28 0.92 1.2
of the existing system step 4 involves no sizing since all the
B2 3.08 1.8 1.2
base objects have 3xed dimensions; however the pressure B3 2.8 4.27 0.8
losses are calculated. Fig. 9 shows the 3nal drawing. B4 2.52 4.7 0.8
B5 2.24 4.7 0.8
B6 1.96 4.49 0.8
5.2. Results
B7 1.68 7.28 0.6
B8 1.12 5.69 0.6
The only real purpose of the calculations in this case, B9 0.56 5.68 0.6
where all the duct diameters are 3xed, is to determine Pfan, B10 6.2 3.67 1
the required fan pressure di@erence. To calculate the pres- B11 5.92 3.4 1
B12 5.56 2.31 1
sure losses for the base objects, the C-coe9cients (dynamic
B13 5.28 5.71 1
pressure loss coe9cients) of the various 3ttings are needed B14 4.72 3.47 1
(Table 4a). The 3tting data are obtained from the ASHRAE B15 4.22 3.93 0.85
Fundamentals Handbook [7,8]. The C-coe9cients of junc- B16 3.72 4.33 0.85
tion 3ttings as calculated from the ASHRAE 3tting tables, B17 3.44 4.57 0.85
B18 3.16 4.79 0.85
Cc, are referenced to the velocity of the upstream “com-
B19 2.88 4.57 0.85
mon” section. However, since these C-coe9cients are taken B20 2.32 5.67 0.7
into account at the downstream “straight” and “branch” sec- B21 1.76 5.61 0.7
tions, Cc must be adjusted to give a C-coe9cient which is B22 1.2 5.84 0.5
referenced to the base object’s own velocity [7,8]. B23 0.64 5.65 0.5
B24 0.32 3.81 0.35
Certain pressure imbalances were found during the pres-
B25 0.32 3.82 0.35
sure calculations (Table 4b). These imbalances will have to B26 0.36 2.09 0.35
be recti3ed with dampers to ensure that the design airBows B27 0.1 1.47 0.25
are achieved. B28 0.26 2.68 0.35
Assuming that the pressure imbalances are recti3ed with B29 0.16 5.42 0.35
B30 0.1 1.47 0.25
dampers it was calculated that the total pressure loss be-
tween the fan discharge and the air di@user exits are 75 Pa.
Other pressure losses for the existing system under consid- (b) Input data
eration include the return system loss (30 Pa), the coil loss Terminal Flow
(100 Pa) and the 3lter loss (200 Pa). In addition a 3xed base label [m3 =s]
30 Pa loss is assumed for each terminal base object (air dif- TB1 0.56
fuser). The required total pressure di@erence by the fan is TB2 0.56
therefore 435 Pa. TB3 0.56
TB4 0.28
TB5 0.28
6. Redesigning the system TB6 0.28
TB7 0.28
TB8 0.28
6.1. General remarks TB9 0.32
TB10 0.32
Assuming that the duct cost per square metre, Sd, remains TB11 0.56
constant the design tool is used to improve the existing de- TB12 0.56
TB13 0.56
sign by minimising the life-cycle cost of the system. Es-
TB14 0.56
sentially the same 3ve-step methodology as is described in TB15 0.28
Section 5 is used, therefore, Figs. 8–9 apply to Section 6 as TB16 0.28
well. TB17 0.28
TB18 0.5
TB19 0.5
6.2. Using the new duct design tool to optimise the TB20 0.56
existing system TB21 0.28
TB22 0.1
Table 5 lists the 3nal results produced by the new duct TB23 0.16
design tool. The diameters of those objects with an aster- TB24 0.1
TB25 0.28
isk have been 3xed either by the duct design tool or by the
530 E.H. Mathews, D.T. Claassen / Building and Environment 38 (2003) 521 – 531

Table 4

Base Label Type Ashrae Fitting Parameter Cc Ci

(a) Fitting resistance calculation


B2 Junction 5-10 Vs=Vc = 0:33 0.18 1.64
B3 Junction 5-10 Vs=Vc = 2:05 0 0
B4 Junction 5-10 Vs=Vc = 0:90 0.01 0.01
B5 Junction 5-10 Vs=Vc = 0:89 0.01 0.01
B6 Junction 5-10 Vs=Vc = 0:88 0.01 0.02
B7 Elbow 3-2 r=D = 1:0 No: of Pieces = 5 0.33
B7 Junction 5-10 Vs=Vc = 1:52 0 0
B8 Junction 5-10 Vs=Vc = 0:67 0.05 0.11
B9 Junction 5-10 Vs=Vc = 0:50 0.1 0.4
B10 Junction 5-10 Vb=Vc = 0:96 0.44 0.47
B11 Junction 5-10 Vs=Vc = 0:95 0 0.01
B12 Junction 5-10 Vs=Vc = 0:94 0.01 0.01
B13 Junction 5-10 Vs=Vc = 0:95 0.01 0.01
B14 Junction 5-10 Vs=Vc = 0:89 0.01 0.01
B15 Elbow 3-2 r=D = 1:0 No: of Pieces = 5 0.33
B15 Junction 5-10 Vs=Vc = 1:24 0 0
B16 Junction 5-10 Vs=Vc = 0:88 0.01 0.02
B17 Junction 5-10 Vs=Vc = 0:92 0.01 0.01
B18 Junction 5-10 Vs=Vc = 0:92 0.01 0.01
B19 Junction 5-10 Vs=Vc = 0:91 0.01 0.01
B20 Junction 5-10 Vs=Vc = 1:19 0 0
B21 Junction 5-10 Vs=Vc = 0:76 0.03 0.05
B22 Junction 5-10 Vs=Vc = 1:34 0 0
B23 Junction 5-10 Vs=Vc = 0:53 0.09 0.3
B24 Junction 5-33 Vb=Vc = 1:02 1 0.96
B25 Junction 5-33 Vb=Vc = 1:02 1 0.96
B26 Junction 5-10 Vb=Vc = 0:50 0.68 2.77
B27 Junction 5-10 Vb=Vc = 0:54 0.65 2.2
B28 Junction 5-10 Vs=Vc = 0:72 0.04 0.07
B29 Elbow 3-2 r=D = 1:0 No: of Pieces = 5 0.33
B29 Junction 5-10 Vs=Vc = 0:62 0.06 0.15
B30 Junction 5-10 Vb=Vc = 0:75 0.54 0.96

(b) Pressure imbalances

Base label Pb [Pa]

B2 44.3
B24 0
B26 21.4
B28 2.2
B29 1.4
TB2 1.4
TB3 4
TB4 15.9
TB5 17
TB6 18.5
TB7 20.4
TB8 22.3
TB11 11
TB12 16
TB13 18.5
TB14 21.9
TB15 23.6
TB16 25.8
TB17 28.3
TB18 31.1
TB19 45
TB20 46.7
TB21 49.8
TB25 53.7
E.H. Mathews, D.T. Claassen / Building and Environment 38 (2003) 521 – 531 531

Table 5 7. Results
Final results

Object Velocity Diameter Pressure The table below summarizes the results for the existing
label [m/s] [m] drop [Pa] and the redesigned systems.
B1* 6.99 1.3 0.3 Existing Redesigned Percentage
B2* 6.13 0.8 1.1 system system change (%)
B3* 7.28 0.7 3
B4* 6.55 0.7 3 PFan (Pa) 435 421 −3:2
B5* 7.92 0.6 4.6 QFan (m3 =s) 9.28 9.28
B6* 6.93 0.6 3.9 Duct surface area (m2 ) 273 252 −7:7
B7* 5.94 0.6 11.6 Capital cost (R) 30735 28368 −7:7
B8* 5.7 0.5 3.9 Operating cost (R) 39291 38032 −3:2
B9* 7.92 0.3 12.9 Total (R) 70026 66400 −5:2
B10* 7.89 1 13.1
B11* 7.54 1 1.8
B12* 7.08 1 1.2
B13* 6.72 1 2.4
8. Conclusion
B14* 7.42 0.9 1.9
B15* 6.63 0.9 10.8
B16* 5.85 0.9 1.8 An 8% saving in material cost has been achieved by using
B17 7.34 0.772 2.9 the new duct design computer tool. This 3gure can result in
B18 7.01 0.758 3 signi3cant savings for big systems. On the negative side, the
B19 6.78 0.735 2.7
number of transition 3ttings was increased from 5 to 9 in
B20 6.18 0.691 3.2
B21 5.56 0.635 2.9 the process. A factor that needs consideration is that while
B22 4.79 0.565 2.7 the same maximum velocity constraint (8 m=s) was applied
B23 3.72 0.468 2.2 when the system was redesigned the average duct velocity
B24* 3.33 0.35 9.7 rose from 4.9 to 5:7 m=s which will lead to higher noise
B25* 3.33 0.35 9.7
levels. However, this does not pose a big problem since the
B26* 3.74 0.35 24.2
B28* 2.7 0.35 1 system is situated in a gymnasium where other sounds may
B29* 1.66 0.35 1.4 still dominate.
B30* 2.04 0.25 2.7 We can not draw general conclusions from this case study.
B27* 2.04 0.25 5.8 To determine if current and past system designs in gen-
eral depart signi3cantly from the optimum situation a wide
authors themselves. Base objects B24 to B30 were 3xed at variety of systems needs to be designed in this fashion. How-
sizes that correspond to that of the existing system for rea- ever, we can say that for this case study the duct design tool
sons of compatibility with the ceiling di@users or because performed its task and that it speeds up the design process.
they are situated in regions where high velocity noise can-
not be tolerated. Objects B10, B15 and B16 had to be 3xed
because it was found that they tended to become extremely References
oversized. For example, the diameter of B15 would other-
wise exceed 1:5 m. In this case, study, it was observed that [1] Mathews EH, Claassen DT. Problems with the T-method. Building
if a base object has relatively large C-coe9cients that are and Environment 1998;33(4):173–9.
referenced to an upstream base object’s velocity, the base [2] Malan AG. The feasibility of developing an improved HVAC system
design tool: an entrepreneurial investigation. Masters Degree thesis,
object tended to become oversized. This is a rather serious University of Pretoria, 1995.
problem because it casts doubts on the T-method’s overall [3] ASHRAE Applications Handbook, SI Ed. American Society of
robustness. Nevertheless, after the troublesome bases had Heating, Refrigerating and Air Conditioning Engineers, Inc., 1791
been 3xed, the T-method proceeded to produce a design that Tullie Circle, N.E., Atlanta, GA 30329, 1991.
did in fact lower the life-cycle cost. [4] Tsal RJ, Behls HF, Mangel R. T-method duct design, Part I:
optimization theory. ASHRAE Transactions 1988;94(2):90–111.
It must be noted that many of the base objects have be-
[5] Tsal RJ, Behls HF, Mangel R. T-method duct design, Part II:
come 3xed during the calculations because they violated calculation procedure and economic analysis. ASHRAE Transactions
the velocity constraint of 8 m=s. This constraint was put in 1988;94(2):112–49.
place because the highest velocity in the existing system is [6] Mills AF. Heat transfer, international student edition. Homewood, IL:
8:2 m=s. The size of those sections that violated the veloc- Irwin, 1992.
ity constraint would be assigned to the smallest multiple of [7] ASHRAE Fundamentals Handbook, SI Ed. American Society of
Heating, Refrigerating and Air Conditioning Engineers, Inc., 1791
100 mm that does not violate the velocity constraint. The
Tullie Circle, N.E., Atlanta, GA 30329, 1993.
3xed base objects led to pressure imbalances, which will [8] ASHRAE Fundamentals Handbook, SI Ed. American Society of
have to be taken up by air dampers. The extra cost due to Heating, Refrigerating and Air Conditioning Engineers, Inc., 1791
the air dampers were not taken into account. Tullie Circle, N.E., Atlanta, GA 30329, 1989.

You might also like