Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

ARTICLE IN PRESS

Building and Environment 43 (2008) 17191733 www.elsevier.com/locate/buildenv

The inuence of stacks on ow patterns and stratication associated with natural ventilation
Shaun D. Fitzgerald, Andrew W. Woods
BP Institute for Multiphase Flow, Madingley Rise, Madingley Road, Cambridge CB3 OEZ, UK Received 16 September 2007; received in revised form 24 October 2007; accepted 26 October 2007

Abstract We investigate the steady state natural ventilation of an enclosed space in which vent A, located at height hA above the oor, is connected to a vertical stack with a termination at height H, while the second vent, B, at height hB above the oor, connects directly to the exterior. We rst examine the ow regimes which develop with a distributed source of heating at the base of the space. If hB ohA , then the unique ow solution involves inow through vent B and outow through vent A up the stack. If H 4hB 4hA , then two different ow regimes may develop. Either (i) there is inow through vent B and outow through vent A, or (ii) the ow reverses, with inow down the stack into vent A and outow through vent B. With inow through vent A, the internal temperature and ventilation rate depend on the relative height of the two vents, A and B, while with inow through vent B, they depend on the height of vent B relative to the height of the termination of the stack H . With a point source of heating, a similar transition occurs, with a unique ow regime when vent B is lower than vent A, and two possible regimes with vent B higher than vent A. In general, with a point source of buoyancy, each steady state is characterised by a two-layer density stratication. Depending on the relative heights of the two vents, in the case of outow through vent A connected to the stack, the interface between these layers may lie above, at the same level as or below vent A, leading to discharge of either pure upper layer, a mixture of upper and lower layer, or pure lower layer uid. In the case of inow through vent A connected to the stack, the interface always lies below the outow vent B. Also, in this case, if the inow vent A lies above the interface, then the lower layer becomes of intermediate density between the upper layer and the external uid, whereas if the interface lies above the inow vent A, then the lower layer is composed purely of external uid. We develop expressions to predict the transitions between these ow regimes, in terms of the heights and areas of the two vents and the stack, and we successfully test these with new laboratory experiments. We conclude with a discussion of the implications of our results for real buildings. r 2007 Elsevier Ltd. All rights reserved.
Keywords: Natural ventilation; Stacks; Stratication; Flow regimes

1. Introduction The growing interest in reducing energy demand from buildings has stimulated much research in natural ventilation. Many of the key principles of natural ventilation have been identied using simplied analogue laboratory experiments, and supporting theoretical models [15]. The majority of these studies focus on buoyancy driven displacement ventilation in which relatively cool air enters the base of the building, is heated, and then discharges from a vent at the top of the building.
Corresponding author. Tel.: +44 1223 765714.

E-mail address: shaun@bpi.cam.ac.uk (S.D. Fitzgerald). 0360-1323/$ - see front matter r 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.buildenv.2007.10.021

As an approximation, the vent and stack geometry are often simplied as being openings at the base and top of the building. However, in many real naturally ventilated buildings, the vents (e.g. windows) may be located at intermediate levels in a room and there may be substantial stack structures which draw air from the building and channel this upwards prior to venting to the exterior. In buildings with vaulted ceilings and sloping roofs, there may be vents or stacks connected to both the lower and the upper part of the roof (Fig. 1a). The present study was inspired by a new classroom block at the Hagley School in Worcestershire in which the top oor classrooms are ventilated using a stackvent conguration analogous to Fig. 1a. A critical question in the design phase of the

ARTICLE IN PRESS
1720 S.D. Fitzgerald, A.W. Woods / Building and Environment 43 (2008) 17191733

stack Sloping roof A hA Internal heat load

B window hB

H B A

A hA

B hB

hB

Fig. 1. Schematic of the steady ventilation regimes in a room heated by a distributed source at the base and ventilated by two intermediate level openings, one of which is connected to a stack which extends to the top of the room. In (a) we show a schematic of the conguration of the vent and stack at Hagley School Worcestershire, in which there is a sloping roof, a high level window on the vertical wall adjacent to the highest point of the roof, and a stack rising above the vent which connects to the lower side of the roof. In (b)(d) we show a generic building, used for the modelling, which can accommodate the stackvent conguration of (a) by suitable choice the vent and stack elevations, but which also allows for the opening of a low level vent/window (Fig. b). In (b)(c) the room ventilates in simple upward displacement mode, while in (d), which has the same geometrical conguration as (c), the ow reverses, now entering the room through the stack.

building concerned the height of the stack which would optimise the outow from the classroom. As we identify in this work, with the geometry of Fig. 1a, it is possible that the stack may involve inow to the room or outow from the room. We will show that a rich spectrum of natural ventilation ows may arise from such a conguration of openings, depending on the relative heights of the vents and termination of the stack. We also show that with a localised source of buoyancy, the nature of the temperature stratication which develops also depends on the locations of the vents and stack. In order to develop a systematic understanding of the different ow regimes, we explore the natural ventilation in a simplied model building with two vents, and in which one vent is connected to a stack. We allow the level of both vents to vary from the oor to the level of the top of the stack (Fig. 1bd). By comparison with Fig. 1a, one can recognise that the design of the stack and vents at Hagley School, with a sloping roof, is captured by this model geometry. In this paper, we explore the ow regimes and patterns of internal stratication which develop with such a conguration of the vents. To our knowledge, many of the ow regimes which we identify have not been described in the literature on natural ventilation (eg. [3]). Earlier work on the inclusion of intermediate level vents [5] has been restricted to the case in which vents are also placed at both the top and bottom of a building. In this paper we focus on a simplied open-plan type building, in which there is a low level heat load and only two vents, one at the oor or at an intermediate level, and one which accesses a stack venting from the side of the building. The case of a distributed heat load, in which the air becomes well-mixed within the space, builds on the work of

Gladstone and Woods [4] who considered the case of a room with vents at high and low level in the space. In that case, there is a unique upward displacement ventilation ow. If in contrast, there are two vents, A and B, at intermediate level in the space, with vent A connected to a stack which extends to the top of the building, then the ow regime depends on the relative height of vents A and B. If vent A lies above vent B, then we expect an upward displacement ow in the stack (Fig. 1b). However, if vent B lies above vent A, then there may be two different regimes. Firstly, the ow may continue to enter the space through vent B and exit through vent A rising up the stack (Fig. 1c). In this case, the ow is driven by the buoyancy force associated with the column of buoyant air in the stack above the level of the inow, vent B. In the second regime, the external air enters the stack and descends into the room through vent A, while air vents from the space through vent B (Fig. 1d). In this case, the ow is driven by the buoyancy force associated with the column of air between the level of vent A and vent B. The two different ow regimes arise from the non-linearity associated with upow and downow in the stack. In Section 2, we develop a model of these multiple states, and, in Section 3, we successfully compare our predictions of the different ow regimes with some new laboratory experiments. Such multiple ow regimes which arise from the non-linearity of ow in a stack have been recognised in the different context of mixing ventilation through roof mounted stacks [6]. The case in which the room is heated by a localised source of buoyancy is more complex and forms the subject of the remainder of this paper. In broad-brush terms, as with a distributed heat source, there is a transition from a

ARTICLE IN PRESS
S.D. Fitzgerald, A.W. Woods / Building and Environment 43 (2008) 17191733 1721

unique upow displacement regime, in the case that vent A connected to the stack lies above the other vent B, to two complementary ow regimes when vent A which is connected to the stack lies below the other vent B. However, with a localised source of heat, a two-layer stratication typically becomes established in the space [1,2]. In the original work of Linden et al. [2], in which there were vents at the top and base of the space, it was shown that the interface lies between the inow and outow vents, that the lower layer is composed purely of external uid and that the outow uid is derived purely from the upper layer. We show here that with a stack connected to one of the vents, there are a number of different ow regimes which can develop depending on whether the vent connected to the stack lies above or below the vent connected directly to the exterior. In each of these ow regimes, the interior uid develops a two-layer stratication, but depending on the vent sizes and elevations, the outow may issue from either the lower or the upper layer. Similarly, the inow may enter either the upper or lower layer, and in the case in which the uid enters the upper layer, the lower layer becomes of intermediate density between the exterior and upper layer. In Section 4 we describe a theoretical model which categorises these different ow regimes, and we present some new analogue laboratory experiments in Section 5 in which we demonstrate each of the regimes. We also test our predictions, by developing an analogue theoretical model for the experiments and comparing the transitions in ow regime with that model. In Section 6, we discuss the implications of these results for the design of naturally ventilated buildings, and consider some avenues for further research. 2. Distributed heat loads We consider a room in which there is a distributed heat load QH at the base of the room which leads to vigorous convection and a well-mixed interior [4]. It is assumed that there are openings A and B on the sides of the room, of area aA and aB , at heights hA and hB above the oor, with opening A connected to a stack which rises to a termination at elevation H above the oor, where H XmaxhA ; hB . It is important to recognise that in this model, H does not correspond to the height of the building, but the height of the top of the stack. However, with a sloping roof conguration, one can imagine that the elevation of vent B, on the upper end of the roof, could coincide with the elevation of a stack above vent A, located on the lower end of the roof. We investigate rst the ow regime which develops when the outow is through vent A and rises up the stack while the inow is through vent B (Fig. 1b and c). The buoyancy driving the ow in this case is associated with a column of buoyant room air extending from the level of vent B to the top of the stack, g0 H hB , where g0 is the reduced gravity of the air in the room, dened as g0 gre rr =re , where

re and rr are the density of the exterior and interior uid, and g is the acceleration due to gravity. If the effective opening area of the two vents is A (cf. [2]; also see Section 4 herein), then the ow rate V is given by V A g0 H hB 1=2 , while the heat ux QH is given by the balance QH rC p DTV , (2.2) where DT is the temperature elevation in the room, r is the density of air and C p is the specic heat capacity of air. For small changes in temperature, g0 $gaDT , (2.3) where the coefcient of expansion for air a 1=T , T is absolute temperature expressed in Kelvin, and so the temperature elevation of the room is !1=3 Q2 H DT (2.4) 2 ar2 C 2 p A g H hB and the ventilation rate is given by combining (2.2) and (2.4),  2 1=3 A H hB gaQH V . (2.5) rC p If the height of vent A, which connects to the stack, lies below vent B, then it is possible that the reverse ow regime is established in which there is inow through the stack, and outow through vent B (Fig. 1d). In this case, the buoyancy driving the ow is associated with the buoyancy of the air in a column between vent A and vent B, given by g0 hB hA , and the temperature elevation of the room is !1=3 Q2 H DT , (2.6) 2 r2 C 2 p A ga hB hA while the ventilation is now given by  2 1=3 A hB hA gaQH V . rC p (2.1)

(2.7)

Fig. 2 illustrates the variation of dimensionless internal 2 2 1=3 as a function temperature Dy DT r2 C 2 p A gaH =QH of the height of vent B hB hB =H in the case that vent A, which connects to the stack, is located at the points hA 1 3 and 2 3 above the oor of the space, where hA hA =H . The two complementary ow regimes, which lead to two different temperatures, develop when hA 4hB , as expected. 3. Experimental investigation of multiple ow regimes with distributed heating Following a similar approach to Gladstone and Woods [4], we carried out a series of analogue laboratory experiments to test whether the multiple ow regimes predicted in Section 2 do indeed develop in an analogue experimental system. Using water as the working uid, we

ARTICLE IN PRESS
1722 S.D. Fitzgerald, A.W. Woods / Building and Environment 43 (2008) 17191733

2.5

1.5

0.5 0 0.2 0.4 hB


Fig. 2. Non-dimensional temperature Dy as a function of the nondimensional height of vent B. The thick solid line corresponds to outow from vent A and through the stack and the thin solid and dotted lines to 2 inow through the stack for hA 1 3 and 3, respectively. Also, shown is comparison of experimental data (symbols) with theory for the case hA 0:58. The dot-dashed line and open squares correspond to inow and black squares to outow through vent A.

0.6

0.8

immersed a perspex tank of height 28.6 cm and area 17:8 cm 17:8 cm, with walls of thickness 0.8 cm, in a large reservoir of water. The experimental tank included a number of openings at different levels on the side of the tank, and there was also a central vent, which connected the room to the exterior through a stack. To model a distributed source of heating, we used a coiled high resistance wire placed just above the oor of the tank. The heat ux produced could then be calculated from measurement of the current through and voltage loss across the wire [6]. Note that this is somewhat different to the technique used by Gladstone and Woods [4] in which the base plate was maintained at a constant temperature and the associated heat ux was inferred from empirical laws for turbulent thermal convection. The temperature distribution in the tank was recorded by type K thermocouples, of accuracy 0:1  C, which were spaced at regular vertical intervals of 3 cm within the tank. With sufcient heat load, in the range 200500 W, this experimental system leads to turbulent thermal convection in the experimental room, with Rayleigh numbers of order 108 109 . At these values of the Rayleigh number the uid within the tank is well-mixed and isothermal, as measured directly from thermocouples in the tank [4], thereby providing a good analogue to the well-mixed ventilation regime within a room heated by a distributed source of buoyancy. The ow through the openings, of diameter 12 cm, with speed 510 cm/s, has Reynolds number of order 10002000. In order to apply the model of Section 2 to such ows, we require an estimate of the loss coefcient

for the openings. The effective combined loss coefcients for the vents and stacks were measured using a calibration experiment in which the tank was lled with a saline solution and then immersed in the reservoir of fresh water. Rubber plugs were removed from two of the ventilation openings, at high and low level, and the rate at which the solution then drained from the tank was recorded (cf. [4]). As described in the Appendix, the effective loss coefcient in the calibration experiment was found to be $0:6, with typical Reynolds numbers in the stack of order 1000. We therefore use this value in comparing our experimental data with the predictions of the model of Section 2. In principle, one can model the frictional losses through both the pipe, which serves as the stack, and the opening from the tank into the pipe [7], but for the range of Reynolds numbers in the present experiments, the empirical value for the loss coefcient of 0:6 is sufciently accurate. By making these measurements, and demonstrating that the constant value for the loss coefcient describes this experimental data, we have established that our model of the ow through the experimental stack system is analogous to that for a real building; in applying the model to a real building one would however need to determine the loss coefcient which pertains to the actual building. Note that in this experimental system, the heat losses from the walls of the experimental tank are small compared to the heat ux convected with the ventilation ow (cf. [4,7]). We infer that the ow regimes which develop within the experimental model are dynamically analogous to those in a real building (cf. [3,4]). We then conducted a systematic series of experiments to explore the steady state ow regimes as a function of the elevation of vent B, with a xed height of vent A which connects to the stack of 12 cm. Experiments were conducted for a range of heat uxes, and both the inow and outow regimes in the stack were realised. In Fig. 2, we compare the experimental measurements of the dimensionless temperature excess in the space, !1=3 A2 r2 C 2 p gH aw Dy DT , (3.1) Q2 H where aw is the coefcient of expansion for water, as a function of the dimensionless height of the vent B, hB . According to the model of Section 2, Dy 1 1 hB 1=3 1 hB hA 1=3 , (3.2)

for the stack outow mode, while Dy , (3.3)

for the stack inow mode. It is seen in Fig. 2 that there is very good agreement between the predictions and the experimental results. Note, however, that as the vertical separation between the mid-points of the two vents becomes small relative to the vertical extent of the

ARTICLE IN PRESS
S.D. Fitzgerald, A.W. Woods / Building and Environment 43 (2008) 17191733 1723

openings, the nature of the ow through each vent changes from uni-directional to bi-directional and the model is then no longer valid. 4. Localised source of buoyancy We now turn to the more complex case in which the heat source is localised rather than distributed over the oor of the space. We explore the effect of the height of the two vents, and the presence of a stack, on the ventilation regime. A localised source of heating at the base of the room typically generates a turbulent buoyant plume and if the space is ventilated through vents at the top and base of the room, then the interaction of the plume with the ventilation leads to the formation of a two-layer stratication within the space [2]. In this ow regime, the interface lies between the two vents, the lower layer is typically composed of pure external uid and only upper layer uid vents from the space. We now show how variations in the height of the two vents, and the presence of a stack connected to one of the vents, can change the internal stratication and ow regime substantially. For our analysis, we assume there is a stack connected to a vent A, located at an intermediate height, hA , while the other vent, B, connected directly to the exterior, has height hB (cf. Fig. 1). We assume there is a localised heat source of

magnitude QH at the base of the room. We explore the different ow regimes which may develop as the elevation of vent B rises from the base of the room to the top of the room, and we focus on the case hB oH , where H is the height of the termination of the stack, as is typically the case in practice. Again, we emphasise that the height of the top of the stack H does not need to correspond to the height of the top of the room; indeed, even in the case that vent B has the same elevation as the top of the stack above vent A, one can imagine a building with a sloping roof, with the stack connected to vent A on the lower part of the roof, and a simple vent B connected to the upper part of the roof (cf. Fig. 1a). The results for the case A 0:1 are shown in the regime diagram of Fig. 3(a), where A A% =l3=2 H 2 with l % 0:12 for a fully developed turbulent buoyant plume [8], A% 2 2 2 2 1=2 cA aA cB aB =1 and where c is the loss 2 cA aA cB aB coefcient for the vent. When vent B is located at the base of the room, two regimes, which we denote as I and V, may develop depending on the elevation of vent A above the base of the room relative to the theoretical height of the uid interface, hL , as predicted by the theory of Linden et al. [2]. If vent A is located sufciently high in the room, then the interface is predicted to lie below vent A, hL ohA , and there is pure outow of the upper layer uid, while the lower layer is composed of external uid (regime I,

1 0.8 0.6 hA 0.4 V 0.2 B 0 0 III IV B A I

1 0.8 A hA 0.6 0.4 0.2 V 0.2 0.4 hB 1 0.8 0.6 hA 0.4 0.2 0 0 0.2 0.4 hB 0.6 0.8 A=1 1 V IV I II III 0.6 0.8 A=0.1 1 0 0 III 0.2 0.4 hB 0.6 0.8 A=0.001 1

II

II

IV

Fig. 3. Regime diagram illustrating the different ow regimes which develop in the case of a localised heat source for the cases in which (a) A 0:1, (b) A 0:001, (c) A 1. In each case, the stack provides the conduit for the outow.

ARTICLE IN PRESS
1724 S.D. Fitzgerald, A.W. Woods / Building and Environment 43 (2008) 17191733

Fig. 3(a)). This situation is essentially the same as described by Linden et al. [2]. However, if the interface hL is predicted to lie above vent A, hL 4hA , then the ow regime changes. Now the interface becomes xed at the level of the vent, so that the relatively buoyant upper layer uid can vent from the room and a steady state becomes established in which there is outow from both layers through the common stack (regime V, Fig. 3a). In the following analysis, we rst consider the case in which hA has a xed value, hA 4hL , and explore how the structure of the ow changes as vent B is located progressively higher above the oor, for example following the line AA0 on Fig. 3a. We then consider the situation in which hA ohL when vent B lies at the oor, and explore how the ow regime evolves as vent B is located progressively higher up the wall, for example following line BB0 on Fig. 3a. 4.1. Interface below the outow vent A, hA 4hL , when hB 0 We start by considering the case in which the inow vent is located at the base of the room, hB 0, and determine the conditions under which the interface lies below the outow vent A, leading to regime I. If the interface lies below the height of the outow vent A, then all the uid venting from vent A and through the stack originates from the upper layer of buoyant uid. The volume ux in the plume as it rises through the density interface therefore equals the volume ux entering through the lower vent B and exiting through the outow stack. These observations can be combined with the classical theory of turbulent buoyant plumes to determine the height of the interface hL (cf. [2])  5 1=2 hL for hL ohA . (4.1) A 1 hL Here the dimensionless height of the interface hL hL =H , the dimensionless height of the access point to the stack hA hA =H . In order that hL ohA and there is single layer outow, we require  5 1=2 hA . (4.2) Ap 1 hA As A increases or hA decreases, the interface height hL eventually coincides with the height of the outow vent, hL hA . The limiting case
1=2 , A h5 A =1 hA

height of vent A, and regime V becomes established. We consider regime V in more detail in Section 4.2. To understand how ow regime I evolves as the inow vent B rises from the base of the room, it is helpful to follow line AA0 marked on the regime diagram (Fig. 3a, corresponding to the case A 0:1). As the height of the inow vent B increases, the interface height remains the same as long as hB ohL . The elevation of the inow vent hB eventually coincides with the height of the interface. Any further increase in hB leads to the inow vent B being located above the height of the interface, while also lying below the elevation of the outow vent A. This is denoted by regime II in Fig. 3a. As a result of the inowing uid passing through a part of the upper layer, where it entrains upper layer uid, the uid supplied to the lower layer now has intermediate composition. As the height of the inow vent B continues to rise, it eventually coincides with the height of the outow vent A. Any further increases in hB lead to the situation in which the inow vent B lies above the outow vent A, which in turn lies above the interface. It is then possible that on further increase of hB , the elevation of the inow vent B, the interface between the two layers will rise to the height hA , corresponding to the outow vent, A. If the inow vent B still lies below the top of the room, then with a further small increase in the elevation of the inow vent B, the interface remains xed at the level of the outow vent A, but now a mixture of upper and lower layer uid vents from the room. This is denoted by regime III in Fig. 3a, and is considered in more detail in Section 4.2 below. We also show in Section 4.2 that if the inow vent B is able to rise even further, then eventually the interface will move above the level of the outow vent A. Now pure lower layer uid vents from the space, and this is denoted by regime IV in Fig. 3a. 4.2. Interface xed to the outow vent A when hb 0 In the case that the inow vent B is located at the oor, and the outow vent A is located sufciently close to the oor, then the interface becomes xed at the level of the outow vent, and regime V develops. To our knowledge, this ow regime, with outow of both upper and lower layer uid, has not been described in detail before, and so we develop a model for this herein. We then explore how this ow regime evolves as the inow vent B rises above the oor of the space. 4.2.1. Regime V If the access point to the stack hA decreases, or the effective area, A, increases such that  5 1=2 hA A4 , (4.4) 1 hA then the above model for regime I (Section 4.1) predicts that the interface lies above the outlet vent A. In steady

(4.3)

corresponds to the case in which all the uid owing through the stack originates from the upper layer, but the interface is located at the level of vent A. If vent A is located above the critical height hA given by (4.3), then the interface lies below the vent and ow regime I is established; if the height of vent A moves below this critical height, then the interface becomes xed to the

ARTICLE IN PRESS
S.D. Fitzgerald, A.W. Woods / Building and Environment 43 (2008) 17191733 1725

state, this is not possible since there would be no loss of buoyant uid from the room. Instead, the interface depth becomes xed at hA in order that buoyant uid can vent from the space. The density of the upper layer then equals that of the plume as it reaches the lower interface of this layer. However, now the volume ux supplied to the upper layer by the plume is smaller than the buoyancy driven outow through the stack associated with uid of that buoyancy. As a result, to maintain a steady state, lower layer uid is also convected up the stack. This decreases the buoyancy of the uid in the stack. This regime is denoted as V in the regime diagram (Fig. 3a). In this equilibrium regime, the volume ux in the stack V is given by the sum of the ux in the plume V p at height hA , V p lb1=3 hA ,
5=3

0.8

Vl / ( Vl + Vp)

0.6

0.4

0.2

(4.5)
0 0 0.2 0.4 hA
Fig. 4. Variation of the proportion of lower layer uid which vents from the stack as a function of the height of access to the stack hA when a room is heated by a point source at the base and ventilated by an opening in the base and an opening on the side wall which connects into a stack. Curves are given for dimensionless vent areas A 0:1, 0.5, 1 and 2 by the solid, dashed, dotted and dot-dashed lines, respectively.

and the ux from the lower layer, V l into the stack, where b is the buoyancy ux at the base and l 0:12 as before. The buoyancy of the uid in the stack g0s , which is assumed to be homogenous, is given by the buoyancy of the plume at height hA , g0p b=V p , multiplied by the factor V p =V p V l owing to the dilution by the lower layer (ambient) uid which enters the stack. By matching the volume ux associated with the outow of uid from the stack p A% g0s H hA with the sum of the volume ux supplied to the upper layer by the plume and V l , we deduce that the ratio of the ux from the lower layer to the total uid entering the stack, V l =V l V p , is given by the relation, h Vl 1 2=3 A . V l V p A 1 hA 1=3
5=3

0.6

0.8

(4.6)

The variation of V l =V l V p with hA is shown in Fig. 4 for several values of A. As the ratio V l =V l V p increases, the buoyancy of the outow decreases, since the uid is being diluted with progressively more lower layer ambient uid. As expected, it is seen that V l =V l V p increases with the effective opening area, A, and with decreasing height of the point of access to the stack hA . Note that in practice, when there is outow from both layers, the interface elevation far from the outow vent may be a little different from that at the vent. This is because near the outow the uid is moving much more rapidly than far from the vent and this can lead to a reduction in the uid pressure. Indeed, in the experiments reported in the next section, in which H 20230 cm, we nd that in this mixed ow regime there is a small difference, o0:5 cm, between the far-eld height of the interface and the elevation of vent A, hA . 4.2.2. Transition VIII We now consider how the internal stratication evolves as the height of the inow vent B rises from the base of the room in the situation in which the ow regime commences in regime V. The discussion corresponds to moving along line BB0 in Fig. 3a. As hB rises, the interface remains xed to the level of the outow vent, hA . Eventually, hB reaches

this same level and then rises above the interface and outow vent A, so that the ow adjusts to regime III as described above and shown in Fig. 3a. Now, the incoming uid will enter the upper layer and, being relatively dense, will descend through this layer. As it descends, it will entrain some of the upper layer uid before reaching the interface. As a result, it supplies the lower layer with uid of intermediate density. While the interface remains xed at the level of the outow vent A, the uid which vents from the room is composed of a mixture of upper and lower layer uid. As the height of the inow vent B increases, the lower layer becomes progressively more buoyant, and the fraction of lower layer uid venting from the space increases. Eventually, a critical point is reached at which all the uid which vents from vent A and through the stack originates from the lower layer. If hB , the elevation of the inow vent B increases any further then this will cause the interface to rise above the level hA corresponding to the height of the outow vent A. At this critical point, the ow regime evolves to regime IV, as shown in Fig. 3a. 4.2.3. Regime transition IIIIV We now develop a simple model to predict the point of transition between regime III and IV, which occurs when the interface is located at the level of the outow stack, hA , but all the outow is derived from the lower layer. To model this transition, we note that the conservation of buoyancy in the room may be written as b Vg0l (4.7)

ARTICLE IN PRESS
1726 S.D. Fitzgerald, A.W. Woods / Building and Environment 43 (2008) 17191733

where V is the net ventilation ow through the room, b is the buoyancy ux at the base as before and g0l is the reduced gravity of the lower layer. The volume ow is given by Bernoullis equation and hydrostatic balance V A g0l H hA g0u hB hA 1=2 , g0u (4.8)

for which the interface height coincides with hA while all the uid venting from vent A and through the stack originates in the lower layer. After some algebra, the system of equations may be reduced to two coupled implicit algebraic equations relating zo , hA and hB :
2 5=3 z5 o hA A hA 1 hB hB hA zo 5=3 5=3

is the reduced gravity of the upper layer. It now where remains to calculate g0l and g0u , by modelling the entrainment in the plume which descends from the inow vent B through the upper layer, and g0u by modelling the entrainment of lower layer uid into the plume which ascends from the source of buoyancy to supply the upper layer. The air entering the room mixes with the ambient uid in the room as it descends through the upper layer; the detailed parameterisation of this mixing depends on the geometry of the inow opening and the initial momentum, buoyancy and direction of the inowing stream of uid [9]. In general, the ow may require several vent diameters to evolve into a fully turbulent plume, and this transition depends on the local circumstances; therefore, as a simplied leading order model, we assume that the air entering the room descends as a simple vertical turbulent plume as it passes through the upper layer: comparison with our new experimental data in Section 5 suggests this is a reasonable initial model of the ow. In this case, the ow may be approximated as originating from a virtual origin point source of pure buoyancy issuing from a source a distance zo above the vent B where zo is given by Turner [9] zo V 2=5 l3=5 g0u 1=5 . (4.9)

(4.15)

and
2 3=5 2 1=5 hA 1 hA z 5 hB hA 7=5 1 hA z5 o =A o =A

zo hB hA 3=5 , where z o z o =H .

4:16

(4.17)

The volume ux entering the lower layer V 1 with the descending plume can then be written as V 1 lb1 hB hA zo 5=3 ,
1=3

Note that Eqs. (4.15) and (4.16) may be recast in a number of forms, but owing to the non-linearity, there does not appear to be a simple analytic expression relating hA and hB . The solution of these coupled implicit equations determines zo and hA as functions of hB and A, and is shown in the regime diagram (Fig. 3a). It is seen that while hB 51, then hA increases with hB . However, as hB approaches 1, the ow begins to stall and the buoyancy of the upper and lower layers increase rapidly, while the distance from the virtual origin to opening B, zo , decreases rapidly. Owing to the non-linear parameterisation of the plume mixing in the expression for zo (cf. [9], (4.9)), we nd that in this limit, hB ! 1, the height of the interface which coincides with the stack inow, hA , is predicted to decrease; however, it is difcult to test this prediction experimentally using the approach described in Section 5 without much larger experimental apparatus. 4.2.4. Regime transition IIIII Finally, there is a further transition for the case in which the stack rising from vent A is the source of outow, as discussed at the end of Section 4.1. This transition corresponds to the point at which the stack changes from being completely lled with upper layer uid to being lled with a mixture of upper and lower layer uid, and where the lower layer is comprised of warm uid. The analysis is similar to that discussed in Section 4.2.3 for the transition from III to IV. However, in this case conservation of buoyancy in the room is given by b Vg0u and the volume ux through the room is given by V A g0u H hB 1=2 . Conservation of buoyancy is given by g0u g0l b . V2 (4.20) (4.19) (4.18)

(4.10)

where b1 is the strength of the buoyancy source associated with the descending plume b1 Vg0u . (4.11) In steady state the reduced gravity of the air entering the lower layer from the descending plume matches the reduced gravity of the lower layer g0l   V 0 0 gl gu 1 . (4.12) V1 The volume ux entering the upper layer from the ascending plume V 2 is described by the theory of Morton et al. [8] V 2 lb1=3 hA .
5=3

(4.13)

Finally, in steady state the volume uxes entering and leaving the lower layer are equal so that V 1 V V 2. (4.14) Eqs. (4.7)(4.14), may be combined to determine the relationship between hA and hB as a function of A and the plume entrainment coefcient l, at this critical condition

Combining (4.18)(4.20) and (4.9)(4.13) we nd that the transition point at which the interface is located at the outow vent A but where the stack is completely lled with

ARTICLE IN PRESS
S.D. Fitzgerald, A.W. Woods / Building and Environment 43 (2008) 17191733 1727

hA REGIME i

hA REGIME ii

hB

Fig. 5. Schematic diagram showing the two layer stratication in the room when inow through the stack occurs. Regime i occurs when the interface in the room lies above the base of the stack and regime ii when the interface lies below the base of the stack.

upper layer uid is given by hA


1 2 hB

2=5

1 hB

1=5

(4.21)

The solution of this relation is also shown on the regime diagram (Fig. 3a). Again, we nd that in the limit hB ! 1 then Eq. (4.21) requires that hA again decreases, owing to the non-linear parameterisation of the mixing in the uid which ows in through the inow opening. Indeed, in the limit hB ! 1 we nd that hA ! 0:5. Again, although the overall predictions are consistent with our experiments, we have not been able to test this detailed prediction using our experimental system owing to limitations of the size of our apparatus (Section 5). The regime diagram shown in Fig. 3a corresponds to the case A 0:1. In Figs. 3b and c we show the regime diagram for the cases in which A 0:001 and 1, respectively. It can be seen that in the case A51, for which the room is effectively very tall relative to the typical dimension of the openings, then regimes II and IV dominate, whereas when A is larger, the other, intermediate regimes can be observed for a wider range of values of hA and hB . Note that in the case of very small A, the assumption that there is no heat loss through the fabric of the ventilated space becomes less accurate, and, in the case of outow through the stack (Fig. 3b) this will lead to a reduction in the temperature of the upper layer and hence the buoyancy force and overall ventilation ow rate. Modelling such effects in detail is beyond the scope of the present study, but some of the effects of such heat losses have recently been described by Livermore and Woods [13]. 4.3. Reverse ow regime with inow through the stack and vent A and outow through vent B As mentioned in the introduction, once the elevation of vent A, connected to the stack, lies below that of vent B which is connected directly to the exterior, it is possible for the ow to reverse in direction, with inow through the stack and vent A and outow through vent B. In this ow regime, the interface either lies above (regime i, Fig. 5) or

below (regime ii, Fig. 5) the inow vent A at the base of the stack, where the cold ambient uid enters the room. In regime i, the ventilation of the room is similar to the classic case described by Linden et al. [2], but with a modied height hB hL over which the buoyancy acts to drive the ow, where hL is the height of the interface within the room. Consequently, hL is given by a modied form of (4.1)  1=2 h5 L A . (4.22) hB hL In order that the interface lies above the inow vent, we require hA ohL where hL is given by (4.22). If the interface lies below the point of access to the stack then regime ii develops whereby the incoming uid will enter the relatively hot upper layer and descend. As the relatively dense ambient air descends, it will form a plume and entrain some of the hot upper layer. Consequently, the lower layer will be warmer than the exterior uid. The delineation between this regime ii and regime i is shown in the regime diagram (Fig. 6), with the transition from regime i to ii being given by setting hA hL in Eq. (4.22). 5. Experimental observations We have developed a series of new analogue laboratory experiments to investigate whether the different ow regimes that we described in Section 4 actually develop in practice. In the experiments of Section 3, we modelled a distributed source of heating using a heated wire. However, in order to model the convective ow associated with a localised heat source, it is easier to use the analogue system of fresh and saline water as described by Baines and Turner [10]. The focus of the experiments is to demonstrate that the multiplicity of different ow regimes predicted in Section 4 can indeed develop in an analogue experimental system. We also develop the model of Section 4 for application to the experiments, now using properties of a saline plume in water, and we compare the experimental observations of the conditions for transitions in the ow

ARTICLE IN PRESS
1728 S.D. Fitzgerald, A.W. Woods / Building and Environment 43 (2008) 17191733

regime with the predictions of this model. Similar experimental models exploring the dynamics of saline turbulent plumes within a ventilated, enclosed chamber have been applied as an analogue model for natural ventilation ows from localised sources of heating in a number of earlier studies (cf. [3,11]). The turbulent buoyant plumes generated from the localised source of saline water within the model room represent an analogue for the natural convective ows generated from localised heat sources within a building, provided that the source mass ux is much smaller than the ventilation ow, and that the plume is of sufcient Reynolds number to be fully turbulent [11]. The experimental system consisted of a small perspex tank (the room) of height 28.6 cm and area 17:8 cm 17:8 cm. This was partially immersed in a large reservoir of height 48 cm and area 88 cm 43 cm, which acted as the exterior environment. We used fresh water in

1 0.8 0.6 hA 0.4 0.2 i 0 0 0.2 0.4 hB


Fig. 6. Regime diagram illustrating the different ow regimes which develop, in the case of a localised heat source, for the case in which A 0:1 and when the stack is the source of inow.

ii ii

0.6

0.8

the environment and saline solution as the source of (negatively) buoyant uid. The system was therefore run upside down, with the plume source located below the free surface of the water in the tank, and the stack termination being located above the base of the tank (Fig. 7). The dense saline solution injected through the plume source, which moves downwards, corresponds to a source of hot air rising from a localised heat source on the base of a real room. One side wall of the experimental tank had a large number of circular ventilation holes, of diameter 5, 6 and 15 mm. These ventilation holes were sealed by rubber stoppers. These could be removed to model a wide range of effective vent areas and heights between the model room and the exterior reservoir. A hole in the base of the tank also enabled stacks of internal diameter d, and with the entry point being located at a height below the roof of the tank, ha , taking values d ; hA given by (13 mm, 189 mm), (13 mm, 245 mm), (10 mm, 208 mm), (10 mm, 230 mm). The stack entrance for the uid was the horizontal plane of the top of the cylindrical pipe used for the stack. The room and exterior tank were initially lled with fresh water. Then, for convenience a 5% salt solution was supplied from above at a constant rate 0.3 cc/s via a twinfeed peristaltic pump, as the source for a descending (negatively) buoyant plume. The nozzle used for the plume source was described by Woods et al. [11]. The room was elevated 20 cm from the base of the exterior tank so that the dense uid exiting from the room could sink to the base of the reservoir, away from the room. A systematic series of experiments were conducted with the plume source 7.1, 10.8, 13 and 14.5 cm from the base of the tank, and with the number of upper vents ranging from one 6 mm diameter vent to, at most, 4 mm 6 mm and 3 mm 15 mm diameter vents all being open simultaneously. As in Section 3, the effective combined loss coefcients for the vents and stacks were found to be $0:6 with typical Reynolds

Peristaltic Pump Plume source Free surface

Reservoir tank Source of saline fluid to pump into tank Experimental model room

Ventilation opening in side of model room

Pipe connected to the opening in base represents analogue of a stack

Supports to raise model room off floor of reservoir

Fig. 7. Schematic of the experimental apparatus, illustrating how the model room is immersed within the external reservoir tank, and the geometry of the inow nozzle and the experimental stack.

ARTICLE IN PRESS
S.D. Fitzgerald, A.W. Woods / Building and Environment 43 (2008) 17191733 1729

0.5 0.4

hA=0.48 hA=0.36

0.3 hL 0.2 0.1 0 0 0.1 A 1 0.8 Vl / (Vl + Vp) 0.6 0.4 0.2 0 0 0.2 0.4 A 0.6 0.8 hA=0.36 0.2 0.3

hA=0.3 hA=0.22 hA=0.16

hA=0.16

Fig. 8. (a) Variation of interface height hL as a function of vent area A for the case in which the room is ventilated by a stack and a vent at the height of the plume source. The thick solid grey line corresponds to Eq. (4.1) cf. Linden et al. [2]. The other lines indicate the stack heights hA used in the experiments, results of which are represented by symbols. Open squares correspond to hA 0:16, lled squares to hA 0:22, open triangles to hA 0:3, lled triangles to hA 0:36 open diamonds to hA 0:48 and lled diamonds to hA 0:57. (b) Variation of proportion of ambient uid in the stack as a function of A for mixed outow mode. The solid and dashed lines denote the predictions from (4.6) for hA 0:16 and 0.36, respectively, and the open squares and lled triangles represent the corresponding experimental results. Experimental errors due to measurement of interface height are estimated to be 0.5 cm, which is at most 5%.

numbers of order 1000. This is discussed further in the Appendix. In comparing the experimental observations of the transition in ow regime with the predictions of the theoretical model, we adopt the model of Section 4, but now use physical properties of water and aqueous saline solution. The main comparison of the experimental results with the theoretical predictions is in testing the ow regimes which develop with different plume source heights and with different net sizes of ventilation opening; however, we also compare the interface height observed in the experiments with the theoretical predictions (Fig. 8a) and the density of the outowing uid with the theoretical predictions (Fig. 8b). 5.1. Regimes I and V In the rst series of experiments a stack was located in the base of the tank whilst the number of ventilation holes

at the top of the tank was systematically increased. For each conguration, we xed (a) the distance between the plume source nozzle and external termination of the stack, H, which in our experiments also corresponds to the depth of the room, and (b) the stack height hA , and we determined the critical value of A% at which there was a transition in the ow regime, as may be seen in the measurements of the steady state height of the interface (Fig. 8a). The transition point corresponds to a change in regime from single layer outow (regime I, Fig. 3) to mixed outow with the interface xed at the level of vent A connected to the stack (regime V, Fig. 3). The photographs in Fig. 9 illustrate the two different regimes, with Fig. 9a showing the case with both upper and lower layer uid supplying the stack and Fig. 9b the case of only uid from the buoyant layer in the stack. Note that in Fig. 8a we have combined the experimental results for a range of different stack heights by plotting hL as a function of A. This allows the theoretical prediction of interface height for the single layer outow mode (regime I, Eq. (4.1)) to be compared with experimental observations whilst clearly separating the transition points for different stack heights. The proportion of upper and lower layer uid in the outow from the stack can be deduced from measurements of salt concentration in the stack and the lower saline layer. In Fig. 8b we show how this varies as a function of effective vent area in the mixing outow regime. Note that in each case, for small vent area the outow is derived purely from the lower layer of saline uid and the interface lies above the outow stack in the experiment (regime I, Fig. 3). However, with larger area, the interface is xed at the level of the access point to the outow stack and the outow is a mixture of uid from the upper ambient layer and lower saline layer. To compare the ow behaviour directly with the predictions of our model we need to account for the effect of a virtual origin since the plume source has nite mass ux (cf. [11,12]). The correction method suggested by Hunt and Kaye [12] has been used, and it may be seen in Figs. 8a and b that the model predictions are in good accord with the experimental observations. In particular the agreement of interface height in the cases for which the interface lies below the level of the outow vent, as shown in Fig. 8a, suggests that the experimental plumes are well modelled by the theory of turbulent plumes (cf. [2]). 5.2. Intermediate level vent and intermediate access point to stack In the main series of experiments, the plume source was located just below the free surface of the water in the room, 1317 cm above the oor of the room (Fig. 7), whilst a series of stacks of different heights and a series of vents connecting directly to the exterior were used in order to map out all the different ow regimes predicted in Fig. 3. Five percent of saline solution was injected from the plume source at a rate of 0.3 cc/s and each experiment allowed to

ARTICLE IN PRESS
1730 S.D. Fitzgerald, A.W. Woods / Building and Environment 43 (2008) 17191733

Fig. 9. Photographs from two experiments for the case in which the room is ventilated by a stack and a vent at the height of the plume source to illustrate (a) the mixing outow regime (V) and (b) the single layer outow regime (I). Dotted lines indicate the position of the interfaces, arrows the direction of ow through the vents.

run for over an hour until a steady state was established. Note, during each experiment, care was taken that the external reservoir uid at the level of the experimental tank did not become contaminated with buoyant uid issuing from the experimental tank. This was achieved by adding a small amount of food dye to the saline solution. Using the shadowgraph technique to detect the level of the interface of the lling box in the exterior, the experiment was stopped when the interface reached the bottom of the experimental tank. Each experiment was conducted a number of times, with different initial conditions; experiments were conducted in which the experimental tank was initially lled with ambient uid or a saline solution. All seven of the ow regimes predicted in Section 4 were observed, as illustrated in the photographs of Fig. 10. The full range of experimental results is shown in Fig. 11, where the experimental observations are compared with the theoretical predictions for the various modes. It is seen that there is general agreement between the theory and experimental results. However, it was found that in the present experiments, the regime IV mode was only stable for a limited range of intermediate level vent heights and access points to the stack. When relatively large values of the vent height hB or relatively small values of the access point to the stack hA were used, the stack tended to revert to an inow rather than an outow. 6. Conclusions In this paper we have analysed both theoretically and experimentally the ventilation of an open-plan building in which there are two vents on the side of the room, one of which is connected to a stack. By allowing the height of the two vents to vary, we have modelled natural ventilation in buildings which may include sloping roofs (cf. Fig. 1a), and in which the top of the stack has different elevation from the two vents. For both the cases of a uniform and a distributed heat load at the base of the building, we have found that if the vent connected to the stack lies above the other vent then

there is a unique ow regime, with outow through the stack. However, if the vent which connects to the stack lies below the other vent, then two different ow regimes may develop, with either inow or outow through the stack. In the case of a distributed heat load, the results form a natural continuation of the model presented by Gladstone and Woods [4], with the room remaining effectively wellmixed: the novelty herein lies in the possibility of two ow regimes with different ventilation rates and internal temperatures for the same heat load. This is a result of the nite change in the buoyancy force driving the ow when the stack is in inow or outow mode. With a point source of buoyancy the picture is more complex, with seven different detailed ow regimes possible under different conditions (Fig. 3). For example, in the case that the stack provides the outow path for the ventilation, and when the inow vent is at the base of the room, then the ow may be analogous to the results of Linden et al. [2] if the vent connected to the stack is sufciently high. A two-layer stratication is established in the room, and the interface is located between the two vents, with the lower layer being composed of external uid, while the upper layer vents through the stack. However, as the elevation of the inow vent increases, it eventually rises above the interface. Now cold external uid entering the space forms a descending turbulent plume which mixes with the hot upper layer to form a lower layer of intermediate density. A further increase in the height of the inow vent, may then cause the interface to rise to the level of the outow stack, and a mixture of the upper and lower layers ventilates from the space. If this inow vent is located even higher, near the top of the space, then there may be a further transition across which the interface rises above the vent connected to the stack. Now, it is the lower layer of uid which issues from the space. As well as the above ow regimes, in which the outow passes through the stack, a second ow regime, with inow through the stack, is also possible if the vent connected to the stack lies at a lower elevation than the other vent. In

ARTICLE IN PRESS
S.D. Fitzgerald, A.W. Woods / Building and Environment 43 (2008) 17191733 1731

Fig. 10. Photographs from seven experiments in which the room was ventilated by a stack accessed at an intermediate level and an intermediate level vent. Cases shown for internal stack diameter 13.5 mm, intermediate level vent 15mm diameter. (a) Regime Ionly upper layer uid vents from the stack, cold lower layer (hA 0:46, hB 0:15); (b) Regime Vuid from both the lower and upper layers vents from the stack, cold lower layer (hA 0:25, hB 0:1); (c) Regime IIonly upper layer uid vents from the stack, warm lower layer (hA 0:46, hB 0:77); (d) Regime IVonly lower layer uid vents from the stack, warm lower layer (hA 0:16, hB 0:5); (e) Regime IIIuid from both the lower and upper layers vents from the stack, warm lower layer hA 0:33; hB 0:38; (f) Regime iinow through stack, cold lower layer (hA 0:21; hB 0:94); (g) Regime iiinow through stack, warm lower layer (hA 0:46; hB 0:77). All experiments are conducted upside down c.f. a heated room and in the descriptions above cold refers to fresh water in the actual experiment and warm to saline water. Dotted lines indicate the position of the interfaces, arrows the direction of ow through the vents.

this case, the interface always lies below the outow vent. However, depending on the location of the inow vent supplied by the stack, the interface may lie above or below

the inow vent, and therefore the lower layer is composed of either pure external uid or uid of intermediate composition.

ARTICLE IN PRESS
1732 S.D. Fitzgerald, A.W. Woods / Building and Environment 43 (2008) 17191733

0.8

II

II, ii distance (cm) IV, ii

35 30 25 20 15 10 5 0 0 50 100 150 200 250 300 350 time (s)

0.6

hA

0.4
V

0.2
IV, i

Fig. A1. Experimental measurements of height of interface as a function of time (diamonds) compared with the theoretical prediction (solid line) for the case in which the loss coefcient for the stack and vent 0:6.

0 0 0.2 III, i 0.4 hB 0.6 0.8 III, ii 1

Fig. 11. Regime diagram for the laboratory experiments in which the room is ventilated by a stack accessed at an intermediate level and an intermediate level vent. The case shown is for A 0:13. Experimental observations of the ow regime are indicated by symbols: Diamonds correspond to regime I, open circles to V, lled squares to II, open squares to IV, lled circles to III, triangles to regime i and crosses to ii.

height of the top of the stack. In this case, all the regimes may be accessible. In addition to the importance of recognising these regimes for developing control systems for the ow, and in particular for locating sensors to detect the different ow regimes, these ndings may also be of importance for modelling the dispersal of smoke or chemicals from a point source release within a building, and for predicting the concentration of contaminants issuing from the building. Appendix

The ow regimes predicted by the models have been demonstrated in a series of new analogue laboratory experiments, for both the distributed and the point sources of buoyancy. Furthermore, we have developed the theoretical models for application to the laboratory experiments, and found reasonable agreement between the experimental observations of transitions between regimes and the prediction of the model. The work is signicant in that it exposes the dominant control that the elevation of the ventilation openings, and the presence of stacks, may have on natural displacement ventilation ows. The situation is of relevance, since many naturally ventilated buildings are designed to have outow through stacks. However, we have shown that a range of different ow patterns can develop with the stacks providing a pathway for either inow or outow, and that in some situations, both of these different ow regimes may develop. As well as stacks and vents designed for natural ventilation, the work has implications for the natural ows which may develop when a room exchanges air through a combination of a chimney and window. In some buildings with at roofs, stacks extend above the height of the building, to height H above the base of the room. In this case, the dimensionless height of the vent (e.g. a window) which opens directly to the environment, scaled relative to H, hB is always smaller than unity. As a result, only a sub-set of the ow regimes described herein (Figs. 3 and 6) are accessible. However, in other buildings, for example, those with sloping roofs, as at the Hagley School in Worcestershire, the height of the opening which connects directly to the exterior may be intermediate to the height of the opening to the stack from the interior space and the

The effective loss coefcient for a vent connected to a stack was determined experimentally. The tank with an open top and a stack connected to the base was lled with 5% saline solution. The external reservoir was then carefully lled such that the water level was 7 cm above the top of the experimental tank. A loosely placed rubber bung which was used to seal the stack whilst the apparatus was lled was then dislodged and the descending interface recorded by video using the shadowgraph technique. The effective area A was determined by comparing the data with the theoretical prediction [3] and tting the loss coefcient. The best t was obtained for a loss coefcient of 0.6 as shown in Fig. A1. This example corresponded to a stack of vertical extent 9.5 cm and diameter 13 mm. The deviation from the theoretical prediction at late time time 130 s corresponded to the time at which the interface approached the top of the stack. At this time the uid exiting from the stack became a mixture of fresh and saline uid and the draining box model [3] was therefore no longer appropriate to describe the ventilation ow. References
[1] Sandberg M, Lindstrom S. Stratied ow in ventilated roomsa model study. In: Proceedings of roomvent 2nd international conference on air distribution in rooms, Oslo, Norway; 1990. [2] Linden PF, Lane-Serff GF, Smeed DA. Emptying lling boxes: the uid mechanics of natural ventilation. Journal of Fluid Mechanics 1990;212:30935. [3] Linden PF. The uid mechanics of natural ventilation. Annual Review of Fluid Mechanics 1999;31:20138.

ARTICLE IN PRESS
S.D. Fitzgerald, A.W. Woods / Building and Environment 43 (2008) 17191733 [4] Gladstone C, Woods AW. On buoyancy-driven natural ventilation of a room with a heated oor. Journal of Fluid Mechanics 2001;441:293314. [5] Fitzgerald SD, Woods AW. Natural ventilation of a room with vents at multiple levels. Building and Environment 2004;39:50521. [6] Chenvidyakarn T, Woods A. Multiple steady states in stack ventilation. Building and Environment 2005;40:399410. [7] Livermore S, Woods AW. Natural ventilation of a building with heating at multiple levels. Building and Environment 2007;42:141730. [8] Morton BR, Taylor GI, Turner JS. Turbulent gravitational convection from maintained and instantaneous sources. Proceedings of Royal Society of London Series A 1956;234:123. 1733 [9] Turner JS. Buoyancy effects in uids. Cambridge: Cambridge University Press; 1979. [10] Baines WD, Turner JS. Turbulent buoyant convection from a source in a conned region. Journal of Fluid Mechanics 1969;37: 5180. [11] Woods AW, Cauleld CP, Phillips JC. Blocked natural ventilation: the effect of a source mass ux. Journal of Fluid Mechanics 2003;495:11933. [12] Hunt GR, Kaye NG. Virtual origin correction for lazy turbulent plumes. Journal of Fluid Mechanics 2001;435:37796. [13] Livermore S, Woods AW. On the effect of distributed cooling in natural ventilation. Journal of Fluid Mechanics 2007, in press.

You might also like