Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Journal of Building Engineering 41 (2021) 102802

Contents lists available at ScienceDirect

Journal of Building Engineering


journal homepage: www.elsevier.com/locate/jobe

Flow topology and loss analysis of a square-to-square sudden expansion


relevant to HVAC systems: A case study
Eszter Lukács *, János Vad
Department of Fluid Mechanics, Faculty of Mechanical Engineering, Budapest University of Technology and Economics, Bertalan Lajos U. 4-6, H-1111, Budapest,
Hungary

A R T I C L E I N F O A B S T R A C T

Keywords: Mechanical ventilation of buildings contributes significantly to the world’s overall energy consumption, there­
Ventilation fore, it is important to consider and control the fluid mechanics losses of air duct components. The current
Square-to-square sudden expansion investigation aims to investigate the Borda-Carnot sudden expansion, which is often used in ventilation systems
Laser Doppler anemometry
in place of diffusers when the flow cross-section needs to be increased. While a significant amount of research has
Reattachment length
Borda-Carnot loss
been carried out on axisymmetric sudden expansions, expansions of square or rectangular cross-sections are
seldom discussed in the literature. As rectangular air ducts are widely used in ventilation systems, a claim on the
research of the flow topology and the loss characteristics of rectangular expansions is raised and is assessed in the
present study. In a first step, the Borda-Carnot loss was examined for uniform and fully developed inlet velocity
profiles by means of semi-empirical methods. It was concluded that the classic loss formula derived for uniform
inlet underestimates the losses and may cause a 2–3% error for each sudden expansion in the calculation of losses
in a ventilation system. In a second step, the flow details and the loss characteristics of a square-to-square sudden
expansion of an area ratio of 2.78 were examined by experimental means. The investigated Reynolds number
range was (0.36–1.8)⋅105, being representative of ventilation systems. Detailed velocity and turbulence-related
measurements were carried out with a laser Doppler anemometer. The downstream effects of the square-to-
square expansion were found to extend significantly farther downstream than that of an axisymmetric expan­
sion, resulting in a 1.5–2.5 times larger flow reattachment length and a 1.5–2 times longer path to reach a fully
developed state. The elongated downstream effect gains importance when there is a need to place control or
measurement devices downstream of the expansion. As revealed by wall static pressure measurements, the semi-
empirical formula derived for the Borda-Carnot loss for fully developed upstream flow approximates the losses
more accurately than the classic formula derived for a uniform inlet. Therefore, a methodology has been pro­
posed herein for calculating the Borda-Carnot loss in ventilation systems, incorporating the loss formula for fully
developed inflow.

1. Introduction In buildings, the fresh air is mostly to be artificially supplied for


health and comfort purposes by ventilation systems, representing a
Due to the rapidly growing population and humankind’s increasing significant factor in the overall energy consumption [1]. An important
need of comfort, energy consumption assumes considerable proportions, aim is to consider and control the fluid mechanics losses occurring in the
displaying a dynamic growth, especially in the building sector. The components of heating, ventilation and air conditioning (HVAC) sys­
contribution of residential and industrial buildings to the world’s total tems, among others, in air ducts. The specified element dimensions and
energy consumption is claimed to be around 20–40% [1,2], which is a auxiliary guide vanes prescribed by the related standards [4,5] can
remarkably high figure. Due to the increasing energetic awareness of the effectively moderate the pressure loss of elbows, joints and contraction
engineering society, strict regulations came into effect in 2002 in the elements incorporated in ventilation ductworks. Expansion elements,
framework of the European Energy Performance of Buildings Directive however, deserve special attention. They are found in ventilation sys­
[3], targeting a general reduction in the energy use of buildings. tems where the reduction of the average velocity or the increase of the

* Corresponding author.
E-mail address: lukacs@ara.bme.hu (E. Lukács).

https://doi.org/10.1016/j.jobe.2021.102802
Received 5 February 2021; Received in revised form 27 May 2021; Accepted 27 May 2021
Available online 1 June 2021
2352-7102/© 2021 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
E. Lukács and J. Vad Journal of Building Engineering 41 (2021) 102802

static pressure is needed. The energetically most effective way of Objective 3: carry out wall static pressure measurements in a
increasing the flow cross-section is to apply a diffuser of an opening square-to-square sudden expansion for its loss analysis and for assessing
angle of ~6–12◦ – depending on the area ratio [6] –, which has the the validity of the semi-empirical relationships set up for loss calcula­
drawbacks of being relatively long and costly. In order to avoid such tions, for uniform and fully developed turbulent upstream flow.
drawbacks, shorter and more wide-angled diffusers are also available on To the authors’ best knowledge, this paper is the first one presenting
the market, ranging from 13◦ to 67◦ [7], while the related standards [4, concerted wall pressure and LDA experiments and their evaluation on a
5] maximize the opening angle at 60◦ . One has to be aware that a square duct with a sudden expansion being relevant to HVAC applica­
diffuser of an opening angle larger than 40◦ –60◦ has already worse en­ tions in terms of geometry and operational conditions.
ergetic characteristics than a simple sudden expansion, i.e. a 180◦
diffuser [6,8,9]. This energetic feature and the fact that a sudden 2. Ranges of characteristics
expansion is the shortest, most inexpensive and most easily manufac­
tured diffuser ratifies its use whenever the flow cross-section is to be Table 1 contains representative ranges of characteristics of ventila­
increased. tion ducts and sudden expansions influencing the related loss. The
An extensive amount of literature is available from various fields of ranges valid for the present study are also indicated in the table. It is to
fluids engineering on sudden expansions, mostly focusing on axisym­ be emphasized that these values are not strictly regulated but are rather
metric geometries, e.g. Refs. [10–23]. Such reports form a basis for rules of thumb representing the engineering practice [4,5,7,29,31]. The
adaption to ventilation ducts of circular or elliptical profiles. Although Reynolds number (Re) is defined as shown in Eq. (2):
such profiles are widespread due to their airtightness and
u1 ⋅dh1
easy-to-produce geometry, it is to be emphasized that rectangular air Re = (2)
ν
duct profiles are used in the vast majority of the cases due to their easier
adaptation in restricted ceiling voids [24]. However, the literature is where u1 is the area-averaged streamwise velocity upstream of the
seldom for three-dimensional flows in square or rectangular expansions sudden expansion, dh1 is the hydraulic diameter of the upstream pipe
[25–28], and even these reports focus on non-Newtonian and/or section, and ν is the kinematic viscosity of the fluid.
laminar flows. In ventilation systems, however, Newtonian turbulent Fig. 1 gives a visual guide to the notations used in Table 1.
flow features are valid, making a claim on their research. The data in Table 1 are commented as follows. Cross-section. An
The loss coefficients of various elements in HVAC systems are aspect ratio (i.e. width/height) of a rectangular cross-section of less than
available in building service engineering handbooks and standards [9, 3:1 is advised [29] unless constrained otherwise by the available space.
29] or manufacturers’ datasheets. The loss coefficients are experimen­ Square cross-section is chosen for the present studies as it provides a link
tally determined for fully developed flows, except for one element, the between the research fields related to circular and rectangular
Borda-Carnot sudden expansion, for which the semi-empirical formula cross-sections in terms of 90-degree periodicity of wall geometry along
presented in Eq. (1) is widespread. The equation suggests the loss co­ the circumference. Hydraulic diameter. Recommended sizes for rect­
efficient to be dependent only on the area ratio (nAR) of the sudden angular air ducts vary between 100 mm × 150 mm and 1200 mm ×
expansion: 2000 mm, resulting in a hydraulic diameter range of 120 mm–1500 mm
( )2 [4], covering the application areas of runouts, branch ducts, and main
ΔpBC,uni 1
ζBC, uni = ρ 2 = 1 − (1) ducts [29]. Average velocity. A maximum of 30 m/s is allowed in
u nAR
2 1 Ref. [29]. However, a trend toward lower velocities is appreciated to
moderate pressure drop and noise. For low-, normal-, and high-pressure
where ζBC, uni is the loss coefficient and ΔpBC,uni is the pressure loss of the
ventilation systems, the recommended maximum velocities - usually
Borda-Carnot element, both for uniform upstream flow, ρ is the density
occurring in the main ducts - are 10 m/s, 15 m/s, and 20 m/s, respec­
of the fluid, u1 is the average streamwise velocity upstream of the sud­
tively [29]. The lowest velocities, generally still above 2 m/s, occur in
den expansion and nAR is the area ratio of the sudden expansion.
the runouts. Reynolds number. The ranges of duct size and average
The formula in Eq. (1) is to be critically revised since it has been
velocity correspond to the orders of magnitude of 104 < Re < 106 in
obtained from the momentum equation using some assumptions, e.g.
HVAC systems, an intermediate range of which, i.e. 3.6⋅104 < Re <
idealistically – and, for a ventilation duct, doubtfully – presuming uni­
1.8⋅105, is performed in the experiments presented herein. Velocity
form flow profiles at both the inlet and the outlet. Therefore, the effect of
profile. The uniform velocity profile is an idealistic, non-existing feature
uniform versus non-uniform upstream flow will be a primary subject to
in HVAC ducts. For its approximation, the technical literature discusses
scrutiny in the present paper.
plug flow, i.e. a mostly uniform flow with a very thin boundary layer
Preliminary experimental studies for a square-to-square sudden
expansion have been carried out by the lead author of the present paper
[30], focusing on the reduction of pressure loss of the expansion. The
Table 1
experimental setup was restricted to short upstream and downstream
Ranges of characteristics of ventilation ducts and sudden expansions.
duct sections. These restrictions have been overcome by a new experi­
mental setup presented in this paper. The new setup enables the study of name notation ventilation systems present study

the effect of fully developed turbulent inlet flow. Furthermore, it pro­ VENTILATION DUCT
vides a means for studying the evolution of the flow field downstream of cross-section A circular, elliptical, rectangular:
rectangular square
the expansion. The objectives of the paper are as follows:
hydraulic dh 120 mm–1500 mm [4,5] 150 mm → 250
Objective 1: assess the discrepancy between the loss coefficients of a diameter mm
sudden expansion in case of uniform and fully developed turbulent average velocity u 2 m/s – 20 m/s – (30 m/s) 3.8 m/s – 18.5 m/s
inflow by theoretical means. Evaluate the significance of this discrep­ [29,31]
ancy in the context of losses developing in an entire HVAC air distri­ Reynolds Re 104–106 (orders of 3.6 ⋅ 104–1.8 ⋅ 105
number magnitude)
bution system.
velocity profile u(y) fully developed; disturbed fully developed
Objective 2: carry out detailed laser Doppler anemometer (LDA)
roughness ε hydraulically smooth [6] hydraulically
studies on the velocity field and turbulence intensity distribution in a smooth
square-to-square sudden expansion with a fully developed turbulent SUDDEN EXPANSION
upstream flow, enabling flow structure analysis and comparison with geometry concentric, eccentric concentric
area ratio nAR 1.1–4–6.4[7] 2.78
axisymmetric sudden expansions studied by other researchers.

2
E. Lukács and J. Vad Journal of Building Engineering 41 (2021) 102802

Fig. 1. Basic parameters in ventilation ducts and sudden expansions: a visual guide to Table 1.

[10,13,15]. For HVAC elements, a more realistic approach is the fully with the present research, carried out for a square-to-square sudden
developed velocity profile [12,14,16,21,22] for both the inlet and expansion. Finally, a detailed description of the physical model, i.e. the
outlet, serving as a basis in pressure drop calculations in engineering experimental setup used in the present research, is given.
practice [32] and also in the present studies. Roughness. The frequently
used material for air ducts is galvanized steel with a nominal surface 3.1. The Borda-Carnot loss for fully developed turbulent inlet flow
roughness of ε = 0.15 mm [9]. In the above described Reynolds number
range, air ducts can be considered hydraulically smooth [6]. For the Objectives 1 and 3 of the present research, formulated at the end of
studies presented herein, hydraulic smoothness has been confirmed by the Introduction (section 1), are to study the importance of the upstream
preliminary pressure measurements. Geometry. Eccentric expansions flow profile on the Borda-Carnot loss. Whereas Eq. (1) has been ob­
have the advantage of avoiding a dead region between the duct wall and tained with the idealistic assumption of uniform upstream flow, a gen­
the ceiling. Still, concentric expansions are widespread and are expected eral formula for non-uniform inflow has been described in Ref. [6] and is
to exhibit a symmetric flow, offering a simplification in their investi­ discussed herein. Eq. (3) shows the loss coefficient of a sudden expan­
gation – as utilized in the present paper. Area ratio. Since size range is sion in case of a fully developed upstream flow, containing the effect of
available in the literature only for circular expansions [7], the authors the non-uniform shape of the velocity profile through the upstream
were compelled to consider such range also for rectangular expansions. momentum (M) and energy coefficients (N), defined by Eq. (4a) and
Area ratios vary between 1.1 < nAR < 6.4, but the range of 1.1 < nAR < 4 (4b) [6,32]. In the case of uniform flow, M and N are equal to unity, so
is more representative. An area ratio of nAR = 2.78 has been chosen as Eq. (3) becomes identical with Eq. (1).
the subject of the present study, approximating the mean value of the
2M 2M − N
representative range. ζBC, dev = N − + (3)
nAR n2AR

3. Materials and methods ∫


u2 dA
M= (4a)
u2 A
This section firstly discusses the semi-empirical model of the Borda-

Carnot loss for fully developed turbulent upstream flow, derived from u3 dA
the momentum equation. The obtained formula will serve as a basis of N= (4b)
u3 A
comparison of losses for uniform and fully developed turbulent up­
stream flow. Next, the fluid mechanics features developing in an where ζBC, dev is the loss coefficient of the Borda-Carnot sudden expan­
axisymmetric sudden expansion are described briefly. These features sion in case of a fully developed upstream flow, M is the momentum
were obtained from the literature, and serve as a basis of comparison coefficient, N is the energy coefficient, nAR is the area ratio of the sudden

3
E. Lukács and J. Vad Journal of Building Engineering 41 (2021) 102802

expansion, u is the streamwise velocity component given at a point, u is the duct, ucl is the centreline velocity and m is the power profile
the area-averaged streamwise velocity component and A is the cross- parameter.
section of the duct. The integration in Eqs. (4a) and (4b) is consid­
ered over representative duct cross-sections. 3.2. Fluid mechanics features in an axisymmetric sudden expansion
The velocity profile is either determined from measurements or, in
the case of fully developed turbulent flow, is approximated using a Objective 2 of the present study is a fluid mechanics comparison
power-law function [8,33]. The power-law function is shown in Eq. (5), between circular and square sudden expansions. To this end, the basic
while the resultant momentum and energy coefficients are presented in features of a circular sudden expansion are outlined first, as illustrated
Eq. (6a) and (6b): qualitatively in Fig. 2. A fully developed turbulent flow is approaching
( )1 the sudden expansion. Separation of the boundary layer takes place at
u
= 1−
2y m
(5) the step edge, where the evolving free shear layer provides a border
ucl d between the main flow and the recirculation zone. The time-averaged
location of this border is frequently referred to as the dividing stream­
where u is the streamwise velocity at a given point, ucl is the centreline line. The net volume flow rate between the dividing streamline and the
velocity, being equal to the maximum value of the axial velocity over the wall is zero. Based on the shape of the dividing streamline, the upstream
cross-section, y is the distance between the duct axis and the given point, jet is observed to remain unexpanded up to 3–4 step heights (h) [10,14].
d is the pipe diameter, and m is the power profile parameter. Similarly to a free jet, a potential core is known to exist up to 3–4h along
(2m + 1)2 (m + 1) the duct axis [18,23], maintaining a nearly constant centreline velocity.
M= (6a) The separation region is divided into a higher-velocity primary and a
4m2 (m + 2)
lower-velocity secondary recirculation zone – the latter is a stagnant
(2m + 1)3 (m + 1)3 zone right at the base of the step. The backflow velocity maximum in the
N= (6b) primary separation bubble can reach 20% of the average upstream flow
4m4 (m + 3)(2m + 3)
and is located at 4hdownstream of the step [13]. The primary separation
The power law has originally been derived for two-dimensional flows ends where the flow is reattached. Since the reattachment of the shear
between two planes, but it has been found to be valid also for circular layer is oscillating back and forth, a reattachment zone occurs. The
cross-sections [33]. Its applicability is to be surveyed for the square length of the secondary recirculation region is approximately 1h [15,
cross-section presented herein. The momentum and energy coefficients 34], whereas reattachment takes place between 6 and 9h [12–15,18,21].
in Eq. (6a) and (6b) have been derived analytically for circular The reattachment length in terms of step height tends to be nearly in­
cross-sections. Nevertheless, these formulae are reported in Ref. [6] to dependent of the area ratio and the Reynolds number [21,23]. Further
be representative also for square cross-sections. downstream, in the relaxation region, a new boundary layer develops.
The power profile parameter m has been determined experimentally Finally, a fully developed flow is reached, at an approximate location
by Nikuradse for the Reynolds-number range of 4⋅103 < Re < 3.2⋅106 between 4d2 [14] and 5d2 [18] downstream of the expansion.
[33]. A correlation between the power profile parameter and the pipe The flow-symmetry-braking behaviour of sudden expansions de­
friction coefficient, based on the measurement results of Nikuradse, is serves a comment herein. Should the flow be expanded only in one
given in Eq. (7) [8], where the pipe friction coefficient (λ) for fully dimension, i.e. in two-dimensional double backward-facing steps (BFSs),
developed duct flows can be determined iteratively in function of the the two distinct separation bubbles are unable to communicate with
Reynolds number using the formula of Prandtl, given in Eq. (8). each other, which can lead to flow asymmetry downstream of the
1 expansion. As a consequence, the reattachment length on one side can
m ≈ √̅̅̅ (7) be even five times larger than that on the other side [34]. Contrarily, the
λ
flow in axisymmetric sudden expansions remains symmetrical [35] due
1 ( √̅̅̅ ) to the quick pressure redistribution in the circumferentially coherent
√̅̅̅ = 2 ⋅ log Re λ − 0.8 (8) separation bubble.
λ
At the bottom part of Fig. 2, a qualitatively correct wall static pres­
where m is the power profile parameter, λ is the pipe friction coefficient sure distribution is depicted. Static pressure linearly decreases upstream
and Re is the Reynolds number. of the step due to duct wall friction in the fully developed flow, and it
As the Reynolds number, thus the pipe friction coefficient does not continues to drop further past the expansion, reaching a minimum be­
change significantly between the upstream and downstream pipe sec­ tween 2 and 3h [13,21,23]. Advancing further downstream, static
tions, the power profile parameter m is approximated to be constant, and pressure recovery takes place, and the wall static pressure increases up
is calculated based on the upstream flow characterstics [6]. The to 15–20h [10]. Then it starts to decrease again as it smoothly reaches
Borda-Carnot loss in case of a fully developed inlet flow can principally the linear pressure drop region belonging to the fully developed duct
be determined from the formerly described equations, using them in the flow. The following observations can be explained only with knowledge
following order: Eq. (8) → Eq. (7) → Eqs. (6a,b) → Eq. (3). The of flow fine structure exceeding the qualitative trends discussed on the
appropriateness of this workflow is checked for the present case study in basis of Fig. 2: (1) Further decrease of wall static pressure close down­
the Results and discussion section (section 4). stream of the expansion – even in lack of any contraction and acceler­
The power profile parameter can be further used to estimate the ation of the flow in the potential core. (2) Location of wall static pressure
relationship between the average and the centreline flow velocity in the minimum – not coinciding [13] with the location of backflow velocity
case of a fully developed turbulent flow, as shown by Eq. (9) and (10) maximum. (3) Further decrease of the wall static pressure – even past
∫ the reattachment zone. This latter observation holds for BFS and for
udA axisymmetric flows, as well and has been explained for BFSs in Refs. [36,
u= (9)
A 37] considering the wall-normal variation of the shear stress.
u 2m2
= (10) 3.3. Experimental setup
ucl (m + 1)(2m + 1)

where u is the area-averaged streamwise velocity component, u is the 3.3.1. Test rig and basic instrumentation
streamwise velocity component given at a point, A is the cross-section of For the present study characterized in Table 1, an experimental setup
has been built. The experimental uncertainty is reported herein for each

4
E. Lukács and J. Vad Journal of Building Engineering 41 (2021) 102802

Fig. 2. Basic flow features and wall static pressure distribution in axisymmetric sudden expansions (not drawn to scale).

measured quantity for a confidence level of 95%, based on [38]. The principle of the LDA is presented herein shortly. For each velocity
sketch of the experimental apparatus with the indication of the most component to be simultaneously measured, a pair of coherent, mono­
important dimensions in meters is shown at the top, while photos of the chromatic laser beams intersect at their waist, forming an ellipsoid-
real experimental setup are shown at the bottom of Fig. 3. The air enters shaped measurement volume inside the fluid flow to be investigated.
the measurement facility through an inlet orifice plate of an inner At the intersection of the laser beams, a system of interference fringes,
diameter of 120 mm for flow rate measurements (1), designed and being parallel with the beam bisector, are formed. The fluid flow is
installed with an adequate free upstream space, in accordance with seeded with light-scattering particles that are selected to follow the flow
Ref. [39]. The pressure difference related to the orifice plate was at high fidelity. As these particles pass the interference fringe system,
measured using a Setra 293 pressure transducer. The uncertainty of the they scatter the light in a series of flashes. The frequency of these flashes
volume flow rate was estimated to be 3%, based on Ref. [39], consid­ – the so called Doppler frequency – depends on the distance of the
ering the uncertainty of pressure difference, temperature, and atmo­ fringes and the fringe-normal velocity of the particles – the latter rep­
spheric pressure, following the root sum square error propagation rule resents the flow velocity component under measurement. The fringe
[40]. Moving further downstream, the turbulence and the lateral ve­ spacing is known a priori from the laser wavelength and the angle of the
locity components are reduced by a combination of a honeycomb (2) laser beams, thus the flow velocity can be directly calculated from the
and three screens (3), as specified by Scheiman and Brooks in Ref. [41]. Doppler frequency. In practice, thousands of velocity realizations are
Followed by a circle-to-square transition element (4), the air enters in a gathered in each measurement point, allowing for ensemble-based sta­
square duct, with a hydraulic diameter (i.e. side length) of dh1 = 0.15 m tistics to calculate the time-averaged velocity and Reynolds stresses at
(5). In accordance with Ref. [6], this duct section has been designed to the selected locations. Furthermore, if one of the laser beams is phase-
be sufficiently long (8 m ≈ 53dh1), with suitable reserve, to have a shifted by a Bragg-cell, a moving interference pattern forms, enabling
naturally developed turbulent velocity profile when the flow enters into the determination of the sign of the velocity component, e.g. for
the Borda-Carnot sudden expansion (6). After the cross-section is rapidly surveying reverse flow in separation bubbles. The LDA was preferred
increased, the airflow is allowed to recover in a square duct with a hy­ over other velocity-measuring devices due to its favourable character­
draulic diameter (i.e. side length) of dh2 = 0.25 m (7), resulting in a step istics, such as non-intrusiveness, high spatial and temporal resolution
height of h = 0.05 m. In order to avoid the upstream effect of the cen­ and directional sensitivity when equipped with a Bragg-cell.
trifugal fan and to have a sufficiently long duct section for flow reat­ In the present study, a two-component TSI LDA was used in back­
tachment and pressure recovery, an 18dh2 = 4.5 m long relaxation scatter mode, fixed on an ISEL positioning system. The side and bottom
section is ensured downstream of the sudden expansion. Finally, a walls of the channel were made of 18 mm thick plywood, which were
centrifugal fan with a frequency converter (9), responsible for air painted in black to reduce excessive light scattering during the LDA
transmission at the desired flow rate, is attached to the test rig through a measurements. The top of the channel was covered with a 4 mm thick
square-to-circle transition element (8). SCHOTT Amiran anti-reflective glass to provide optical access. The three
linear translation stages of the positioning system were mounted
3.3.2. LDA velocity and turbulence measurements orthogonally. The null-point error, i.e. the accuracy of the positioning
As sketched in Fig. 3, velocity and turbulence measurements were system, was 0.05 mm in the x and y directions, and 0.3 mm in the z
carried out by a laser Doppler anemometer (LDA). The operation direction (conf. Fig. 3). The precision of the traverse was 20 μm. The

5
E. Lukács and J. Vad Journal of Building Engineering 41 (2021) 102802

Fig. 3. Experimental setup. Top: top view of the experimental setup, not drawn to scale. Middle: side view of the experimental setup and measurement devices;
operation principle of the LDA. Bottom: photos of the real experimental setup.

light produced by a Melles Griot 350 mW air-cooled Argon ion gas laser
Table 2
was separated into green and blue pairs of laser beams in the fiber­
Characteristics of the LDA system.
lightTM Multicolor Beam Generator. The green and blue laser beams
were used for measurement of the x- and y-components of the velocity, Green (x- Blue (y-
component) component)
respectively, and for obtaining the related normal Reynolds stresses.
One of the beams in both pairs was phase-shifted by a Bragg cell. Laser Angle of laser beams (◦ ) 3.94 3.94
Outlet diameter of laser beams (mm) 2.65 2.65
light was then transferred via optical cables to a TR260 fiberoptic probe,
Focal length (mm) 363 363
responsible for focusing the laser beams to form the measurement vol­ Wavelength (nm) 514.5 488
ume. The flow was seeded with Safex Blitz fog fluid atomized into Fringe spacing (μm) 3.729 3.547
micron-sized droplets by a Safex F2005 fog generator. The volume flow Measurement volume diameter (μm) 89.73 85.11
rate of the seeding fluid was set to allow for single-realization mode [42] Measurement volume length (mm) 1.3 1.2

and to avoid altering the turbulence intensity of the airflow [43]. The
scattered light was then collected by the fiberoptic probe, turned into thermal expansion of the optical probe, non-zero extent of the mea­
electric signals with the help of a PDM photodetector module, and surement volume, positioning of the measurement volume, and statis­
processed in an FSA 3500 signal processor unit. Analysis of the incoming tical uncertainty due to the finite duration of data collection. Velocity
data was carried out with the FLOWSIZERTM 64 software. The data bias was minimized by applying gate-time weighing. The absolute un­
collection in one point lasted for 60 s in the main stream and 240 s in the certainty of the LDA velocity measurements was in the order of 10− 2 m/s
separation bubble. During that period, several tens of thousands of ve­ in the whole cross-section, which resulted in a relative error of a
locity realizations were recorded. Further characteristics of the LDA maximum of 0.5% in the main flow region and around 2% in the shear
system are given in Table 2. layer. The relative error in the separation bubble was dominated by the
In the estimation of the uncertainty of the time-averaged velocity range of 10–20%.
data based on [42,44,45], the following sources of error were taken into
consideration: accuracy of the frequency tracking and fringe spacing,

6
E. Lukács and J. Vad Journal of Building Engineering 41 (2021) 102802

3.3.3. Wall static pressure measurements discrepancy is introduced as follows:


Static pressure taps of an internal diameter of 0.5 mm were manu­
ζBC,dev − ζBC,uni
factured according to Ref. [46] and were installed on both sides of the δζBC = (11)
ζBC,uni,MAX
channel in the horizontal symmetry plane. The distance between the
taps was 0.25 m farther away from the sudden expansion and 0.05 m in
, where ζBC,dev and ζBC,uni is the Borda-Carnot loss coefficient for fully
the vicinity of the sudden expansion, for better spatial resolution. The
developed [Eq. (3)] and uniform [Eq. (1)] velocity inlet, respectively,
pressure taps were connected to a 48-channel Scanivalve pressure
and ζBC,uni,MAX is the largest possible value of the loss coefficient in case
multiplexer by silicone tubes, then the pressure was forwarded to a Setra
of a uniform inlet, being equal to unity as nAR → ∞ [conf. Eq. (1)], i.e. it
293 capacitive pressure transducer. For each location, pressure mea­
represents the inlet dynamic pressure as loss. Fig. 5a shows the relative
surements with a sampling frequency of 1000 Hz were carried out for 10
discrepancy as function of the area ratio and the Reynolds number. The
s. Measured values were stored and processed by a personal computer
black plane presents the 5% error, being a representative limit of
with the use of the LabVIEW software. The pressure transducer was
acceptance in engineering practice. In Fig. 5b, the domain for which the
calibrated to a Betz manometer at the beginning of each measurement
relative discrepancy is higher than 5% is shown with enchanced visi­
session. Thus, the absolute uncertainty of the wall pressure measure­
bility. From Fig. 5, the following conclusions are drawn.
ments was considered to be 0.5 Pa, which resulted in a relative uncer­
tainty of less than 0.5% in the measured range.
• Since δζBC is always positive, Eq. (1) underestimates the loss of the
sudden expansion in comparison with Eq. (3).
3.3.4. Auxiliary measurements
• The relative discrepancy between the two loss-calculating methods
Temperature and barometric pressure in the laboratory were moni­
increases with nAR and decreases with Re.
tored using a GREISINGER electronic GMH 3710 thermometer and a
• The relative discrepancy δζBC between the two loss-calculating
Setra 470 barometric pressure transducer, respectively, for air density
methods may exceed 5% within the domain of Re < 7⋅104 and nAR
calculations. The uncertainty of the temperature measurements was
> 2.
0.3 ◦ C, and that of the barometric pressure measurements was 100 Pa,
which data contain the instrumental error and the variance in atmo­
Although the relative discrepancy δζBC tends to increase towards low
spheric conditions during the measurement scenarios.
Re values, the absolute discrepancy of calculated pressure losses,
As a summary of section 3, the research methodology described
expressed in Pascals, is expected to increase with the velocity, i.e. with
above is shown in a comprehensive flowchart in Fig. 4.
Re. As an illustration, Fig. 6 presents the difference between the pressure
losses calculated for developed as well as for uniform inflow in a square-
4. Results and discussion
to-square sudden expansion in Pascals (Δ(ΔpBC) = ΔpBC,dev - ΔpBC,uni), as
function of the average velocity in the inlet air duct, for various dh1 and
4.1. Theoretical results – calculations
nAR parameters. One must note that a pipe with dh1 = 0.1 m is generally
used as a runout, where high velocities are unlikely to happen. However,
In order to quantify the difference between the two versions of the
a pipe with dh1 = 0.5 m can be already considered a main duct, where
Borda-Carnot loss coefficients in Eq. (3) and Eq. (1), a relative

Fig. 4. Flowchart of research methodology, in terms of the Objectives in the Introduction (section 1).

7
E. Lukács and J. Vad Journal of Building Engineering 41 (2021) 102802

• The higher the nAR and the smaller the dh1 is, de larger the difference
is.
• The effect of nAR is more significant than the effect of dh1, especially
for larger dh1 values of 0.5–1 m.

Δ(ΔpBC) is to be evaluated also in the context of losses developing in


an entire HVAC air distribution system. For systems of low, normal and
high pressure classes, the maximum permissible pressure drop is Δpsys,
max = 100, 200, 300 Pa [47], and the maximum velocity – typically
occurring in the main duct – is 10, 15 and 20 (30) m/s [29], respectively.
A relative discrepancy between the two calculation methods from a
system point of view is defined as δ(ΔpBC) = Δ(ΔpBC)/Δpsys,max and its
magnitude is shown as function of nAR in Fig. 7. The curves have been
calculated for dh1 = 0.5 m, as a representative size for main ducts. The
parameter appearing in Fig. 7 is the maximum allowed velocity in the air
ducts [29] for the different pressure classes [47]. Table 3 summarizes
the results of Fig. 7, where the value of the maximum relative discrep­
ancy is given in the last row. It can be concluded that the relative
discrepancy for all three categories of extraction systems is 2–3%, except
for the extreme case of an airflow of 30 m/s, where it exceeds 5%. It is to
be emphasized that the data in Fig. 7 and Table 3 have been calculated
for a single sudden expansion element. Since there are usually multiple
sudden expansions in a system, the relative discrepancy may reach
considerable proportions.

4.2. Experimental results

4.2.1. LDA velocity and turbulence measurements


LDA measurements were carried out for Re = 3.6⋅104 exclusively,
due to seeding constraints occurring at higher Reynolds numbers. The
volume flow rate was qv = 0.085 ± 0.001 m3/s = 306 ± 4 m3/h,
resulting in an average upstream velocity of u1 = 3.8 ± 0.1 m/s and an
average downstream velocity of u2 = 1.36 ± 0.04 m/s. The turbulence
Fig. 5. (a) Relative discrepancy of the pressure loss coefficient; (b) Critical intensity in the core of the inlet flow was measured as 4%, being in
pairing of the Reynolds-number and the area ratio for which the relative accordance with the value predicted semi-empirically for fully devel­
discrepancy exceeds 5%. oped turbulent flows as Tu = 0.16⋅ Re− 1/8 = 4.3% [48].
As a first step, velocity data were gathered upstream of the sudden
expansion at x = − 2.8h in the wall-normal symmetry planes (horizontal
plane: z = 0, vertical plane: y = 0; conf Fig. 3). The upstream velocity
profile has been checked in terms of symmetry as well as correspondence
to the power-law function in Eq. (5), as shown in Fig. 8. Pressure
gradient measurements upstream of the expansion resulted in a pipe
friction coefficient of λ1 = 0.023 ± 0.002, from which the power profile
parameter was estimated as m = 6.6 ± 0.3 using Eq. (7). Fig. 8 illustrates
a fair coincidence of the measurement data in the four quadrants of the
duct, demonstrating symmetrical upstream flow. A fair agreement be­
tween the measurements and the power-law theory is also observed,

Fig. 6. Absolute discrepancy between the pressure loss in a sudden expansion


for fully developed flow and for uniform flow in function of the average inlet
flow velocity.

high velocities may occur. Therefore, the representative range in HVAC


systems is demonstrated by the grey area. Based on Fig. 6, the following
observations are made.

• Δ(ΔpBC) is increasing with increasing velocity and may reach 15–20


Fig. 7. Relative discrepancy between the pressure loss in a sudden expansion
Pa.
for fully developed flow and for uniform flow in function of the area ratio. dh1
= 0.5 m. Parameters (upper left): inlet mean velocity.

8
E. Lukács and J. Vad Journal of Building Engineering 41 (2021) 102802

Table 3 downstream of the step edge at x/h = 1, 3, 5, 7, 9, 11. Transversal


Maximum expected pressure discrepancies for different extraction system measurement coordinates for the wall-normal plane are shown in terms
pressure categories. of the hydraulic diameter (i.e. side length) of the upstream duct, while
System pressure category those for the diagonal planeare shown in terms of the diagonal of the
low normal high
upstream duct; thus, such dimensionless coordinates collapse, and the
measured results can be presented together. Fig. 10 shows the dimen­
maximum pressure loss, Pa 100 200 300
sionless axial velocity profiles (u/u1 ) together with the calculated
maximum velocity, m/s 10 15 20 30
maximum relative discrepancy, % 2.2 2.3 2.7 5.6 dividing streamlines, indicated by a solid line in the wall-normal plane
and a dashed line along the diagonal. The transversal position of the
dividing streamline was approximated by such means that the axial
velocity profile, incorporating forward and reverse flow regions, gives a
zero mean value along the transversal line of LDA traverse between the
dividing streamline and the wall. Fig. 11 depicts the turbulence in­
tensities, calculated as shown in Eq. (12):
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
1
(u′ u′ + v′ v′ + w′ w′ )
(12)
3
Tu =
u1

where Tu is the turbulence intensity, u u , v v and w w are the normal


′ ′ ′ ′ ′ ′

component of the Reynolds stress tensor in the x, y and z directions,


respectively, and u1 is the average streamwise velocity upstream of the
sudden expansion. Concerning the Reynolds stresses, only two compo­
nents (u u and v’v’) were measured, while the third component was
′ ′

estimated as the average of the other two: w’w’ = 1/2⋅(u’u’ +v’v’) [49].
Further velocity field analysis is shown in Fig. 12, where a more
detailed measurement of the axial velocity component is depicted along
the duct axis between 0 and 11h. In Fig. 13, the maximum backflow
Fig. 8. Velocity measurement data 2.8h upstream of the sudden expansion at velocities in the recirculation region appear, both for the wall-normal
the wall-normal symmetry planes – comparison with the power-law theory. (filled marker) and the diagonal (empty marker) symmetry planes.
Observations on the macroscopic behaviour of the flow are made
suggesting a fully developed flow. The fully developed nature of flow using Fig. 10, in the view of the features sketched in Fig. 2. The jet,
was confirmed by auxiliary velocity measurement along the axis, where entering from the small to the large duct, is expanding gradually. After
∂u/∂x was found to be practically negligible. reattachment, a relaxation zone is known to occur to reach a fully
As a second step, velocity and turbulence related measurements have developed velocity profile. Recirculation is present in the separation
been carried out downstream of the sudden expansion up to x/h = 11. bubble. Results over the wall-normal and diagonal planes are fairly
Time-averaged velocity data in the wall-normal symmetry planes were overlapping in the core flow region; increased departure is observable
gathered across the entire cross-section, indicating flow symmetry. As an only near the wall. Reattachment tends to occur at 10–11h in the wall-
example, the axial velocity component at 5h downstream of the step is normal symmetry plane and further downstream in the corners. As no
shown in Fig. 9. velocity measurement data are available downstream of x/h = 11,
In order to gain further insight into the flow characteristics, LDA reattachment in the corners has been estimated by extrapolating the
measurements along the diagonal of the downstream duct were also related dividing streamline. A second-order polynomial fits the points of
carried out. As flow symmetry have been established earlier, and the the dividing streamline with a coefficient of determination of 0,997.
diagonal is also expected to be a symmetry plane, data were gathered Extrapolating this polynomial, the reattachment length was estimated to
only along one half of the diagonal. Figs. 10 and 11 show time-averaged be 15h in the corners. Right after the step, the dividing streamline de­
LDA measurement profiles in the wall-normal symmetry plane (y > 0, z flects outward and inward along the wall-normal and diagonal sym­
= 0; filled markers) and along the diagonal (y = z; empty markers) metry plane, respectively. Then, they both remain straight and nearly
parallel to the axis, i.e. the upstream jet remains unexpanded up to about
3h. Afterwards, the dividing streamlines are diverted outward but less
intensely over the diagonal symmetry plane. The aforementioned ob­
servations suggest that the edges of the original square-shaped jet
entering the expansion tend to be gradually smoothened along the axis.
Flow visualization using wool tufts and titanium-oxide powder –
kerosene mixture confirmed the existence of the secondary recirculation
bubble at the base of the step. The axial extension of the secondary
recirculation bubble was found to be 1h, which is in accordance with
observations in the case of axisymmetric expansions. The main conclu­
sion drawn from Fig. 10 is that flow reattachment in the examined ge­
ometry occurs between 10 and 15h, contrary to the reattachment length
of 6–9h in axisymmetric sudden expansions. Knowing the boundaries of
the separation bubble is of practical importance, e.g. in avoiding the
installation of HVAC measurement devices within the separation zone.
The calculated turbulence intensities and the location of their
maximum values are shown in Fig. 11 for the wall-normal and diagonal
symmetry planes. Up to 5h, the maximum turbulence intensity line tends
Fig. 9. Velocity measurement data 5h downstream of the sudden expansion at to deflect slightly outward and inward along the wall-normal and
the wall-normal symmetry planes – checking flow symmetry.

9
E. Lukács and J. Vad Journal of Building Engineering 41 (2021) 102802

Fig. 10. Axial velocity profiles with dividing streamlines.

Fig. 11. Turbulence intensity profiles with maximum turbulence intensity lines.

Fig. 12. Axial velocity along the duct axis. Fig. 13. Maximum backflow in the recirculation region.

diagonal symmetry plane, respectively. These trends are then expansions [14,18]. A potential core lasting up to 2–3h can also be noted
commuted, and eventually, both lines tend to deflect inward. The in Fig. 12, which is slightly shorter than the 3–4h predicted for
maxima of the turbulence intensities are between 16 and 21%, axisymmetric sudden expansions [18,23].
increasing in the downstream direction, in accordance with the obser­ For further analysis of the primary separation zone, values of
vation regarding axisymmetric sudden expansions [13]. maximum backflow velocities are given in the examined measurement
Fig. 12 suggests that the flow has not yet reached a completely planes in Fig. 13, and a comparison is made with the theoretical results
developed state, as the centreline velocity is still decreasing at 11h. predicted in Ref. [13]. The maximum backflow in the wall-normal plane
Though measurement results are not available further downstream, an is 21% of the average upstream velocity, being in good accordance with
expected location of a completely developed velocity profile can be the prediction in Ref. [13], while the maximum backflow in the diagonal
given by linear extrapolation of the available data. The centreline ve­ plane reaches a higher value, namely 28%. The maximum backflow
locity can be calculated using Eq. (10) by substituting m = 6.6, on the location in both planes lies between 4 and 5h, being close to the value of
basis of which a completely developed turbulent velocity profile is 4h in axisymmetric cases. The measurement data in the wall-normal
predicted at 40h = 8dh2, contrary to 4-5d2 expected for axisymmetric symmetry plane are in good accordance with the theory in Ref. [13]

10
E. Lukács and J. Vad Journal of Building Engineering 41 (2021) 102802

up to 4h downstream, but the discrepancy increases as moving further


downstream, suggesting a delayed flow recovery, similarly to Figs. 10
and 12..

4.2.2. Wall static pressure measurements


Wall static pressure measurements were carried out for five different
Reynolds numbers. Fig. 14 presents the distributions of the pressure
coefficient, defined on the basis of [50] as follows:
px − px=− h
cp = (13)
pmax − px=− h

where cp is the pressure coefficient, px is the wall static pressure at


location x, px=− h is the wall static pressure at x = − h and pmax is the
maximum value of the wall static pressure downstream of the expansion.
In Fig. 15, the measured pressure loss coefficient of the square-to-square
sudden expansion is shown together with theoretical values for uniform
(ζBC,uni) and fully developed turbulent (ζBC,dev) inlet flows. The measured
loss coefficients were obtained as described in Eq. (14) and (15): Fig. 15. Loss coefficient of the sudden expansion in function of the Reynolds
number – comparison of measurement and theory.
Δpid − Δpmeas
ζBC,meas = ρ 2 (14)
u
2 1 to occur at the same location in the examined Reynolds number range. In
( ) accordance with the flow field analysis, the pressure distribution curves
ρ ρ also confirm that for square-to-square sudden expansions, a completely
Δpid = N⋅ u21 − u22 (15)
2 2 developed velocity profile is reached further downstream than for
axisymmetric sudden expansions. Lengths of the typical flow regions are
where ζBC,meas is the measured loss coefficient, Δpid is the ideal pressure
summarized in Table 4.
rise calculated for fully developed turbulent inlet [6], Δpmeas is the real
Finally, the energetic behaviour of the square-to-square sudden
pressure increase, measured between the step edge and the maximum
expansion in the case of fully developed upstream flow is evaluated
pressure downstream of the expansion, as done in Refs. [10,11], ρ is the
using Fig. 15. There is a better agreement between the measurements
density of the fluid, u1 and u2 are the area-averaged upstream and
and theory for a fully developed inlet than for a uniform inflow. The
downstream flow velocities, respectively, and N is the energy
measured values are consistently higher than the loss coefficient pre­
coefficient.
dicted for uniform inlet flow and decrease with Re. The slope of the
Wall static pressure is evaluated in conjunction with the flow fea­
measurement-based decrease tends to be somewhat higher than that
tures in Fig. 14, also depicting data from Ref. [10] for axisymmetric
expected from the theory for a fully developed inlet. This is dedicated to
cases. The wall static pressure distribution is qualitatively similar to that
the measurement uncertainty as well as to the approximate nature of the
evolving in an axisymmetric expansion (conf. Fig. 2). The static pressure
loss formula.
is decreasing downstream of the expansion, it hits a minimum, then
starts to increase beyond reattachment, and finally reaches a maximum.
5. Conclusions and future objectives
Similarly to the axisymmetric case, the pressure minimum is located at
2–3h in the square-to square sudden expansion, which does not coincide
Turbulent flow in a square-to-square sudden expansion of an area
with the maximum backflow location. The location of the maximum
ratio of 2.78 – representing a frequent element in HVAC systems – has
pressure in the present case is between 25 and 30h, contrary to 15–20h
been analyzed by theoretical and experimental means. The two essential
valid for the axisymmetric case. The pressure coefficients for the
aspects of the research were the prediction of the Borda-Carnot loss and
square-to-square expansion for various Reynolds numbers overlap,
the analysis of the flow topology downstream of the sudden expansion.
suggesting that flow similarity holds, i.e. flow reattachment is expected
The Borda-Carnot loss was examined for uniform as well as for fully
developed upstream flow, with the help of the semi-empirical formulae
derived for axisymmetric expansions from the momentum equation. It
was shown that the loss in the case of a fully developed upstream flow is
always greater – and reaching occasionally 15–20 Pa – than that for
uniform inflow. As the loss formula derived for uniform flow is generally
used by practicing engineers, but the flow in a ventilation duct system is
never uniform, the loss calculated by such means is usually under­
estimated. The magnitude of this underestimation may reach 2–3% of

Table 4
Lengths of typical flow regions.
Zone name Zone length

Square Axisymmetric

secondary recirculation 1h 1h
pressure minimum 2–3h 2–3h
core region 2–3h 3–4h
maximum backflow 4–5h 4h
reattachment 10–15h 6–9h
pressure maximum 25–30h 15–20h
Fig. 14. Wall static pressure coefficient distribution for various Rey­
total flow recovery 8d2 4–5d2
nolds numbers.

11
E. Lukács and J. Vad Journal of Building Engineering 41 (2021) 102802

the maximum allowed loss in ventilation extraction systems per each and numerically in order to give a simple and robust way for the ener­
Borda-Carnot element. Therefore, the use of the semi-empirical formula getic improvement of square-to-square sudden expansions in HVAC
derived for a fully developed turbulent upstream flow is suggested, systems.
necessitating additional calculations but being yet mathematically easy-
to-use for practicing engineers. Furthermore, the steps of the proposed Funding
loss calculation methodology have been confirmed for square-to-square
expansions via detailed fluid mechanics measurements. The research reported in this paper and carried out at the Budapest
The flow topology and the downstream disturbing effect of the University of Technology and Economics has been supported by the
square-to-square sudden expansion were investigated by detailed LDA National Research Development and Innovation Fund under contracts
velocity and wall static pressure distribution measurements. The flow TKP2020 Institution Excellence Subprogram, Grant No. BME-IE-WAT
features were compared to those forming in axisymmetric sudden ex­ and TKP2020 National Challenges Subprogram, Grant No. BME-NCS
pansions, and were found to be qualitatively similar. A major quanti­ based on the charter of bolster issued by the National Research Devel­
tative difference, shown by the current research, was that the opment and Innovation Office under the auspices of the Ministry for
downstream disturbing effects of the investigated square-to-square Innovation and Technology, furthermore under contract No. K 129023.
expansion were remarkably longer than those of axisymmetric sudden
expansions. The reattachment length for the square expansion was CRediT authorship contribution statement
found to be between 10 and 11 step heights for the wall-normal sym­
metry planes and 15 step heights in the corners, being 1.5–2.5 longer Eszter Lukács: Conceptualization, Methodology, Validation, Formal
than in axisymmetric expansions. The length being necessary to reach a analysis, Investigation, Resources, Data curation, Writing – original
fully developed turbulent velocity distribution in the square expansion draft, Writing – review & editing, Visualization, Supervision, Project
was estimated to be 8 downstream hydraulic duct diameters, being administration. János Vad: Conceptualization, Methodology, Re­
1.5–2 times longer than in axisymmetric expansions. Based on the cur­ sources, Writing – original draft, Writing – review & editing, Supervi­
rent case study, it can be advised to avoid installing flow controlling and sion, Funding acquisition.
measurement devices closer than 8 downstream duct hydraulic di­
ameters to the expansion.
As a future objective, the authors plan to carry out an exhaustive Declaration of competing interest
Computational Fluid Dynamics (CFD) study in the relevant Reynolds
number range for various area ratios to have a more comprehensive The authors declare that they have no known competing financial
view of the flow behaviour. The eventual goal is to investigate the effect interests or personal relationships that could have appeared to influence
of the loss reducing appendices described in Ref. [30] experimentally the work reported in this paper.

Nomenclature

A cross-section, m2
cp pressure coefficient
d diameter, m
dh hydraulic diameter, m
h step height, m
m power profile parameter
M momentum coefficient
N energy coefficient
nAR area ratio
p wall static pressure, Pa
px=− h wall static pressure at location x = -h, Pa
pmax maximum wall static pressure, Pa
pmin minimum wall static pressure, Pa
px wall static pressure at location x, Pa
ΔpBC Borda-Carnot loss
Δpid ideal pressure increase in a Borda-Carnot sudden expansion, Pa
Δpmeas measured pressure increase in a Borda-Carnot sudden expansion, Pa
Δpsys,max maximum allowed pressure drop in a ventilation extraction system, Pa
Δ(ΔpBC ) absolute discrepancy between the Borda-Carnot loss for fully developed and uniform inlet
δ(ΔpBC ) relative discrepancy between the Borda-Carnot loss for fully developed and uniform inlet
Re Reynolds number
Tu turbulence intensity
u streamwise mean velocity, m/s
u area-average streamwise mean velocity, m/s
ucl centreline streamwise mean velocity, m/s
umax, backflow maximum streamwise mean backflow velocity, m/s
uu normal component of the Reynolds stress in the x-direction, m2/s2
′ ′

vv normal component of the Reynolds stress in the y-direction, m2/s2


′ ′

ww normal component of the Reynolds stress in the z-direction, m2/s2


′ ′

x streamwise coordinate, m
y transverse, horizontal coordinate, m

12
E. Lukács and J. Vad Journal of Building Engineering 41 (2021) 102802

z transverse, vertical coordinate, m


ε surface roughness, m
ζBC Borda-Carnot loss coefficient
δζBC relative discrepancy between the loss coefficients for fully developed and uniform inlet
λ pipe friction coefficient
ν kinematic viscosity of the fluid, m2/s
ρ density of the fluid, kg/m3

Subscripts
1 upstream of the sudden expansion
2 downstream of the sudden expansion
uni uniform flow
dev fully developed flow

References [25] A. Sau, Three-dimensional simulation of flows through a rectangular sudden


expansion, Phys. Fluids 11 (1999) 3003–3016.
[26] A.N. Alexandrou, T.M. McGilvreay, G. Burgos, Steady Herschel–Bulkley fluid flow
[1] L. Pérez-Lombard, J. Ortiz, C. Pout, A review on buildings energy consumption
in three-dimensional expansions, J. Non-Newtonian Fluid Mech. 100 (2001)
information, Energy Build. 40 (2008) 394–398.
77–96.
[2] A. Costa, M.M. Keane, J.I. Torrens, E. Corry, Building operation and energy
[27] C. Mistrangelo, L. Bühler, Numerical investigation of liquid metal flows in
performance: monitoring, analysis and optimisation toolkit, Appl. Energy 101
rectangular sudden expansions, Fusion Eng. Des. 82 (2007) 2176–2182.
(2013) 310–316.
[28] P.C. Sousa, P.M. Coelho, M.S.N. Oliveira, M.A. Alves, Laminar flow in three-
[3] Directive 2002/91/CE of the European Parliament and of the Council of 16
dimensional square–square expansions, J. Non-Newtonian Fluid Mech. 166 (2011)
December 2002 on the Energy Performance of Buildings, 2002.
1033–1048.
[4] En 1505, Ventilation for Buildings. Sheet Metal Air Ducts and Fittings with
[29] CIBSE, CIBSE Guide B2:2016 Ventilation and Ductwork, CIBSE, London, 2016.
Rectangular Cross Section, Dimensions, 1997.
[30] E. Lukács, T. Régert, Pressure loss reduction of a Borda-Carnot sudden expansion
[5] En 1506, Ventilation for Buildings. Sheet Metal Air Ducts and Fittings with Circular
applying passive flow control method, Proc. IME J. Power Energy 226 (2011)
Cross Section, Dimensions, 2007.
182–191.
[6] I.E. Idelchik, Handbook of Hydraulic Resistance, third ed., Jaico Publishing House,
[31] ASHRAE, ASHRAE Handbook-Fundamentals, first ed., ASHRAE Inc., Atlanta, USA,
Mumbai, India, 2008.
2009.
[7] Lindab, Stainless Duct System, Technical Information, 2021. https://itsolution.li
[32] P.J. Oliveira, F.T. Pinho, Pressure drop coefficient of laminar Newtonian flow in
ndab.com/LindabWebProductsDoc/PDF/Documentation/ADS/Lindab/Technical/
axisymmetric sudden expansions, Int. J. Heat Fluid Flow 18 (1997) 518–529.
StainlessSteel-Technical.pdf.
[33] H. Schlichting, Boundary Layer Theory, 1st English edition, Pergamon Press Ltd,
[8] P. M Gerhart, R.J. Gross, Fundamentals of Fluid Mechanics, Addison-Wesley
London, 1955.
Publishing Company, USA, 1985.
[34] D.E. Abbott, S. J Kline, Experimental investigation of subsonic turbulent flow over
[9] Recknagel, Sprenger, Schrameck, Taschenbuch für Heizung+Klimatechnik 2000
single and double backward facing steps, Journal of Basic Engineering 84 (1962)
(in Hungarian), Dialóg Campus Kiadó, Budapest-Pécs, 2000.
317–325.
[10] G. Heskestad, A suction scheme applied to flow through sudden enlargement,
[35] F. Durst, A. Melling, J.H. Whitelaw, Low Reynolds number flow over a plane
Journal of Basic Engineering 90 (1968) 541–552.
symmetric sudden expansion, J. Fluid Mech. 64 (1974) 111–128.
[11] R.P. Benedict, N.A. Carlucci, S.D. Swetz, Flow losses in abrupt enlargements and
[36] W.D. Moss, S. Baker, Re-circulating flows associated with two-dimensional steps,
contractions, J. Eng. Gas Turbines Power 88 (1966) 79–81.
Aeronautical Quartetly 131 (1980) 151–172.
[12] P.P. Zemanick, R.S. Dougall, Local heat transfer downstream of abrupt circular
[37] J.Y. Yoo, S.J. Baik, Redeveloping turbulent boundary layer in the backward-facing
channel expansion, J. Heat Tran. 92 (1970) 53–60.
step flow, J. Fluid Eng. 114 (1992) 522–529.
[13] R.G. Teyssandier, M.P. Wilson, An analysis of flow through sudden enlargements in
[38] ISO 5168-2005, Measurement of Fluid Flow. Procedures for the Evaluation of
pipes, J. Fluid Mech. 64 (1974) 85–95.
Uncertainties.
[14] L.F. Moon, G. Rudinger, Velocity distribution in an abruptly expanding circular
[39] ISO 5167-2:2003, Measurement of Fluid Flow by Means of Pressure Differential
duct, J. Fluid Eng. 99 (1977) 226–230.
Devices Inserted in Circular Cross-Section Conduits Running Full — Part 2: Orifice
[15] R.P. Durrett, W.H. Stevenson, H.D. Thompson, Radial and axial turbulent flow
Plates.
measurements with an LDV in an axisymmetric sudden expansion air flow, J. Fluid
[40] S. Kline, F. Mclintock, Describing uncertainties in single-sample experiments,
Eng. 110 (1988) 367–372.
Mech. Eng. 75 (1953) 3–8.
[16] M. Stieglmeier, C. Tropea, N. Weiser, W. Nitsche, Experimental investigation of the
[41] J. Scheiman, J.D. Brookes, Comparison of experimental and theoretical turbulence
flow through axisymmetric expansions, J. Fluid Eng. 111 (1989) 464–471.
reduction from screens, honeycomb and honeycomb-screen combinations,
[17] W.J. Devenport, E.P. Sutton, An experimental study of two flows through an
J. Aircraft 18 (1981) 638–643.
axisymmetric sudden expansion, Exp. Fluid 14 (1993) 423–432.
[42] H.E. Albrecht, M. Borys, N. Damaschke, C. Tropea, Laser Doppler and Phase
[18] D.R. Cole, M.N. Glauser, Flying hot-wire mesurements in an axisymmetric sudden
Doppler Measurements Techniques, third ed., Springer, Germany, 2003.
expansion, Exp. Therm. Fluid Sci. 18 (1998) 150–167.
[43] S. Elghobashi, On predicting particle-laden turbulent flows, Appl. Sci. Res. 52
[19] B. Guo, T.A.G. Langrish, D.F. Fletcher, Numerical simulation of unsteady turbulent
(1994) 309–329.
flow in axisymmetric sudden expansions, J. Fluid Eng. 123 (2001) 574–587.
[44] T.T. Yeh, J.M. Hall, NIST Special Publication 250-79 Airspeed Calibration Service,
[20] C. Duwig, M. Salewski, L. Fuchs, Simulations of a turbulent flow past a sudden
Special Publication (NIST SP), 2007, pp. 79–250.
expansion: a sensitivity analysis, AIAA J. 46 (2008) 408–419.
[45] V.A. Iyer, M.A. Woodmansee, Uncertainty analysis of LDV measurements in a
[21] Y. Bae, Y.I. Kim, Prediction of local pressure drop for turbulent flow in
swirling flowfield, AIAA J. 43 (2005) 512–519.
axisymmetric sudden expansions with chamfered edge, Chem. Eng. Res. Des. 92
[46] ISO 5801:2007, Industrial Fans — Performance Testing Using Standardized
(2014) 229–239.
Airways.
[22] Y. Bae, Y.I. Kim, Prediction of local loss coefficient for turbulent flow in
[47] EN 13779:2007, Ventilation for Non-residential Buildings. Performance
axisymmetric sudden expansions with a chamfer: effect of Reynolds number, Ann.
Requirements for Ventilation and Room-Conditioning Systems.
Nucl. Energy 73 (2014) 33–38.
[48] ANSYS, Inc, ANSYS fluent user’s guide, Release 19.0 (2018). Equation (6.68).
[23] Ali Nouri-Borujerdi, Ardalan Shafiei Ghazani, Simulation of compressible and
[49] W.D. Moss, S. Baker, Re-circulating Flows associated with two-dimensional steps,
incompressible flows through planar and axisymmetric abrupt expansions, J. Fluid
Aeronaut. Q. 31 (1980) 151–172.
Eng. 141 (11) (2019) 111107.
[50] M. Narayanan, Y.N. Khadgi, P. Viswanath, Similarities in pressure distribution in
[24] Lindab, Ductwork Turns Full Circle, 2002. Retrieved January 29, 2021, http://
separated flows behind backward-facing steps, Aeronaut. J. 25 (1974) 305–312.
www.lindab.com/be/Documents/Catalogues_Manuals/Lindab_Waarom%20rond%
20gebruiken.pdf.

13

You might also like