Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

ARTICLE IN PRESS

Building and Environment 41 (2006) 12791290


www.elsevier.com/locate/buildenv

Numerical simulation of the dynamic operation of


multi-outlet siphonic roof drainage systems
G.B. Wrighta,, S. Arthurb, J.A. Swafelda
a
Drainage and Water Supply Research Group, School of the Built Environment, Heriot-Watt University, Edinburgh, EH14 4AS, Scotland
b
Sustainable Water Management Group, School of the Built Environment, Heriot-Watt University, Edinburgh, EH14 4AS, Scotland
Received 7 July 2003; received in revised form 18 March 2005; accepted 4 May 2005

Abstract

Although siphonic roof drainage systems have been installed in Europe since the 1970s, the design of such infrastructure is still
based on steady state theory. Such approaches are only truly applicable at the design condition, under specic rainfall conditions,
and cannot be used to assess the response of siphonic systems to alternative rainfall conditions and/or operational scenarios.
Previous research at Heriot-Watt University led to the development of a numerical model capable of accurately simulating the
performance of single outlet systems under a range of different conditions. The work reported herein details the logical extension to
this work; i.e., the development of a numerical model to simulate the performance of multi-outlet siphonic roof drainage systems.
The experimental work undertaken to assist in the development of the new model boundary conditions is briey described, and
relevant results are illustrated. The development of the numerical model is then detailed, and model output is presented. Finally,
conclusions are drawn regarding the developed model, and plans for future work are outlined.
r 2005 Elsevier Ltd. All rights reserved.

Keywords: Siphonic roof drainage; Multi-outlet; Numerical modelling

1. Background gutter outlets and the point of discharge being equated


to the head losses in the system. Although this approach
Siphonic roof drainage systems have been in existence neglects the small quantities of entrained air that always
for approximately 30 years, and are becoming an enter a siphonic roof drainage system, it has been
increasingly common element of urban drainage infra- reported to yield operational characteristics similar to
structure. In that time, the construction industry in most those observed in laboratory test rigs at the fully primed
developed countries has been gradually persuaded of the state [1,2]. However, existing design methods cannot be
benets that these systems offer when compared to used to analytically assess the ability of a system to
conventional roof drainage technologies. Current design prime. Furthermore, steady state design methods cannot
practice assumes that, for a specied design storm, a predict performance when a system is exposed to rainfall
siphonic system lls and primes rapidly with 100% events below the design criteria, when the ow may
water. This assumption allows siphonic roof drainage contain substantial quantities of entrained air, or events
systems to be designed utilising steady state hydraulic with time varying rainfall intensity. As such events are
theory. The steady ow energy equation is normally the norm, current design methods may not be suitable
employed [1], with the elevation difference between the for determining the day-to-day performance character-
istics of siphonic roof drainage systems. This is a major
Corresponding author. Tel.: +44 131 4518261; disadvantage, as it is during these events that the
fax: +44 131 4513161. majority of operational problems tend to occur, e.g.
E-mail address: g.b.wright@hw.ac.uk (G.B. Wright). noise and vibration. In addition to these types of

0360-1323/$ - see front matter r 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.buildenv.2005.05.003
ARTICLE IN PRESS
1280 G.B. Wright et al. / Building and Environment 41 (2006) 12791290

Nomenclature Q ow rate
Sf friction slope
A ow cross-sectional area S0 pipe slope
c wave celerity t time
f friction factor V ow velocity
g acceleration due to gravity x distance along pipe
H pressure head e difference director parameter
I1 rst moment of the area A about the free r ow density
surface i,P,R,S nodal subscripts
m hydraulic radius n time superscript

everyday operational issues, a number of more serious numerically model the effect of different types of
problems are known to have occurred with siphonic connection between a siphonic system and the below
roof drainage systems; experience has shown that such ground sewer network. In order to achieve these aims,
failures are normally due to blockages of an outlet or as the primary objectives of the research were as follows.
a consequence of delays and/or difculties in achieving
steady-state operating conditions. It should be noted 1. Experimentally investigate the performance of multi-
that thousands of siphonic systems have been installed outlet siphonic systems, including the effects of
in Europe over the past decade, and there is no data to system imbalance, different types of sewer connection
suggest that they are more prone to failure than and outlet blockages.
conventional systems. However, although work is now 2. Develop an enhanced numerical model to simulate
ongoing, there is as yet no specic British or European the operation of multi-outlet siphonic drainage
standard for the design of siphonic drainage systems. systems under a variety of realistic scenarios.
A siphonic roof drainage research programme,
initiated at Heriot-Watt University in 1996, has led to
a better understanding of the performance character- 3. Laboratory investigations
istics of single outlet siphonic systems, with particular
reference to the priming of such systems [2] (the purging Laboratory investigations were undertaken to gain a
of air from the system). In turn, this work resulted in the better understanding of the conditions occurring within
development of a numerical model capable of accurately multi-outlet siphonic systems, and hence to assist in the
simulating the priming phase of single outlet siphonic development of the boundary conditions and solution
roof drainage systems. However, the majority of techniques to be used in the enhanced numerical model.
installed systems incorporate multiple outlets, and the Whilst a detailed description of the laboratory work is
interaction between such outlets was not well under- given elsewhere [4], the following sections summarise the
stood. It was therefore recognised that further research salient results.
was required to extend the applicability of the existing
model to multi-outlet applications. This work is 3.1. Overview
reported herein.
The experimental test rig used in this study (Fig. 1) was
designed to be representative of a simple multi-outlet
2. Aims and objectives siphonic roof drainage system. To ensure realistic ow
conditions, each gutter was fed via a rear supply trough
The aim of the reported work was to further develop and a simulated sloping roof. A PC based data acquisition
the unsteady ow model that simulates conditions system was used in conjunction with pressure transducers
within an idealised, single outlet siphonic roof drainage and magnetic induction ow meters to record system
system [3]. It was envisaged that the enhanced numerical conditions. As the pipework was transparent, direct
model would be capable of simulating the conditions observations and high-speed video footage were used to
within multi-outlet siphonic roof drainage networks, assist in the identication of relevant ow conditions.
where local gutter water levels may mean that one or
more of the outlets acts as an air inlet to the system, 3.2. Experimental results
causing re-pressurisation, transient propagation and
cessation of siphonic action. In addition to this dynamic It should be noted that, unless otherwise stated, the
balancing, it was also intended to investigate and gutter inow rates were constant during the testing
ARTICLE IN PRESS
G.B. Wright et al. / Building and Environment 41 (2006) 12791290 1281

Pumped flow to simulated gutters


F
F

Flow direction
0.37m

12.1m

4.47m

T2 T1
gutter 2 gutter 1

Collection tank and


recirculation pump
T4 0.85m
Legend 1.65m
53mm diameter siphonic outlet T5 0.65m
T3
69mm diameter pipework
57mm diameter pipework
T1 transducer location 2.37m
F flow meter Plan view of section above

Fig. 1. Schematic of the experimental test rig.


Gutter depth (mm)

periods; that is the rainfall events were assumed to


150
instantaneously reach a constant intensity.
100
Depth in gutter 1 (T1)
3.2.1. Design criteria rainfall event (fully primed system) 50 Depth in gutter 2 (T2)

The priming procedure of the test rig was charac- 0


Pressure in branch 1 (T3)
0.3
Pressure in branch 2 (T4)
terised by the formation of full bore ow conditions Pressure in common pipe (T5) 0.0

Pressure (mH2O)
upstream of the branch junction, and the subsequent
-0.3
propagation of these conditions throughout the system.
During the priming process, trapped air pockets formed -0.6
time
and moved downstream through the system, exiting lag -0.9
either as a single entity or becoming mixed with water at -1.2
the branch junction to form a bubbly ow. Once all of 32.5 33.0
Enlarged view of -1.5
the initial air pockets had exited the downpipe, system re-pressurisation phase
pressures decreased and remained relatively constant. -1.8
0 10 20 30 40 50 60
The system was then fully primed, although small Time since start of simulated rainfall event (s)
quantities of entrained air continued to enter with the
Fig. 2. Measured conditions for design criteria rainfall eventgutter 1
water inows.
inow 5.85 l/s, gutter 2 inow 7.78 l/s (refer to Fig. 1 for
Fig. 2 illustrates typical data collected during the transducer layout).
priming of the test rig. As shown, the different head
losses associated with the upstream branches meant that
the system was only in a balanced state (primed with
constant gutter water depths) when the inow into siphonic system is similar to that associated with a single
gutter 2 was signicantly higher than that into gutter 1. outlet system [3]; the only signicant difference is that
The partial re-pressurisation phase occurring at approxi- the inter-relationships inherent in multi-outlet systems
mately 33 s illustrates the effect of the trapped air result in more complex ow conditions, particularly
pockets travelling through the downpipe and exiting the with respect to the formation and movement of trapped
system; the time lag between pressure peaks clearly air pockets within the system.
illustrates that the re-pressurisation wave was generated
at the downstream end of the common pipe, and 3.2.2. Rainfall events below the design criteria
propagated upstream. At ow rates up to approximately 15% of the design
In general, the recorded data and visual observations criteria inows, the laboratory test rig behaved as a
conrm that the priming process for a multi-outlet conventional roof drainage system. Above these inows,
ARTICLE IN PRESS
1282 G.B. Wright et al. / Building and Environment 41 (2006) 12791290

220 mm 380mm

1.25m

system system system system


termination 1 termination 2 termination 3 termination 4

Fig. 3. Siphonic terminal connection types investigated.

the test rig was found to exhibit one of three different Table 1
sets of unsteady characteristics. Observations also high- Variation in design criteria gutter inows with type of terminal
connection
lighted that, if the inow into one of the branches of a
fully primed system was reduced sufciently to induce Terminal Approximate Fully primed depth Fully primed
free surface ow conditions within that branch, the connection time to (mm) capacity (l/s)
remaining branch would continue to ow full bore up type priming (s)
Gutter 1 Gutter 2 Gutter 1 Gutter 2
until the junction, after which the combined down-
stream ow would contain large quantities of air, and 1 33 144 100 7.46 5.71
exhibit pulsing characteristics. 2 34 153 99 7.35 5.57
3 43 211 138 7.28 5.54
4 52 272 164 7.11 5.45
3.2.3. Rainfall events above the design criteria
With rainfall events above the design criteria, the
pressures measured in the test rig were almost identical
to those obtained at the design condition; the only and variations in effective system head1. As shown in
minor differences being due to the slight variations in Table 1, the use of any conguration other than a freely
driving head associated with higher gutter depths. discharging vertical downpipe resulted in a lower system
However, the additional system inows (above the capacity, even though the gutter depths (and hence
design criteria levels) resulted in continuously increasing driving head) may have actually been marginally higher.
gutter depths, which would have eventually lead to Furthermore, this data illustrates that submerged
overtopping of the gutter(s). terminal connections resulted in signicantly longer
priming times and higher gutter ow depths than those
3.2.4. Gradually increasing and staggering gutter inflows discharging directly to the atmosphere; a consequence of
Gradually increasing the gutter inows up to design the greater (positive) pressures required to purge the
criteria levels, and hence simulating a progressive initial air pockets from the system.
increase in rainfall intensity, was found to simply
increase the priming time of the test rig. Experimental
data obtained with staggered gutter inow start times, 3.2.6. Blockage of an outlet
which may be thought of as representing systems Theory suggests that the sudden blockage of a
incorporating widely varying roof geometries or branch siphonic outlet (e.g. by leaves) can lead to the generation
lengths, yielded very complex ow conditions; exhibiting of very large pressure transients; the sudden cessation of
two or three of the unsteady ow regimes identied ow producing a negative pressure wave that propagates
previously. However, in both of these scenarios, it was and reects throughout the system. Although it is
apparent that after a short period at the design criteria hazardous, and virtually impossible to replicate such
inows, the gutter ow depths and system pressures severe events in the laboratory, a number of less extreme
mirrored those obtained with constant inows and blockage events were simulated. Fig. 4 illustrates that
synchronised start times. signicant pressure transients were generated even when
the blockage was only partial and relatively slow. With
3.2.5. Different terminal connection types reference to Fig. 4, it should be noted that the initial,
Although siphonic systems are normally designed to high positive transient was a result of the method used
terminate at a point of free discharge, site layout and/or to simulate the blockage event (pushing an upturned
surcharging of the surface water sewer network may 1
Effective head is dened as the elevation difference between the
preclude this. Experimental work on four different gutter outlets and the point of discharge (free discharge case) or the
terminal connection types (Fig. 3) highlighted the effect point at which the downpipe becomes submerged (submerged
of head losses associated with different connections discharge case).
ARTICLE IN PRESS
G.B. Wright et al. / Building and Environment 41 (2006) 12791290 1283

0.0 written in conservative form as [6]


Pressure (mH2O)

-0.4 qA qQ
0, (1)
qt qx
-0.8  
qQ q Q2
-1.2 gI 1 gAS 0  S f . (2)
qt qx A
-1.6 Whilst the continuity and momentum equations of one-
120 121 122 123 124
dimensional, unsteady full bore ow may be written
Time since start of simulated rainfall event (s)
as [7]
Fig. 4. Variation in branch 2 pressure (T4) with a partial, slow and  
2 qV qH qH
temporary blockage of the outlet in gutter 2design criteria rainfall rc V rg S 0 rg 0, (3)
event. qx qx qt

qV qV qH
V g gS 0  S f . (4)
Table 2 qx qt qx
Measured system conditions with outlet 1 unblocked and blocked Eqs. (1)(4) form the basis of SIPHONET2. As they are
hyperbolic partial differential equations, they cannot be
Outlet blocked Fully primed Capacity of Minimum
capacity (l/s) outlet in gutter measured solved directly, and recourse must be made to some
2 (l/s) system pressure form of numerical solution technique.
(m H2O)
4.2. Numerical solution techniques
None 13.63 7.78 1.388
Outlet 1 11.30 11.30 2.147
The existing numerical model developed at Heriot-
Watt University utilises a method of characteristics
based solution technique [8], which has previously been
container over the outlet), rather than a result of the
employed successfully at Heriot-Watt University in the
blockage itself.
simulation of both free surface and full bore ow
Where an outlet was blocked prior to the initiation of
conditions. However, during the development of SIPH-
a simulated rainfall event, the system acted as a single
ONET2, it became clear that the method of character-
outlet siphonic system, with a lower, fully primed
istics technique was not particularly suited to the
capacity. Table 2 compares the conditions occurring in
simulation of moving hydraulic jumps in networks as
the unblocked test rig, and in the test rig with the outlet
complex as multi-outlet siphonic systems. As the wide
in gutter 1 permanently blocked. As shown, although
ranging movement of hydraulic jumps is an essential
the total system capacity was lower with the outlet
element of system priming, the MacCormack method [9]
blocked, the capacity of the open outlet was actually
was employed to simulate the initial free surface ow
higher than was the case in an unblocked system. This
conditions, whilst retaining the method of character-
data also highlights that system pressures were con-
istics technique for the simulation of full bore ow
siderably lower when the outlet was blocked; indicating
conditions.
that if a system were designed to operate at very
low pressures (below approximately 7 m H2O), a
4.2.1. MacCormack method
complete blockage of an outlet may result in the onset
The MacCormack solution technique relies on the
of cavitation and/or failure of the system by pipe
hyperbolic nature of the governing equations, which
implosion [5].
leads to spontaneous discontinuities that have real
physical meanings, e.g. hydraulic jumps. The technique
is a non-centered, two step nite difference scheme
4. Development of an enhanced numerical model
which is second order, accurate in time and space.
Starting from the initial time level, when steady state
The observations made and the data collected during
theory is used to calculate conditions at regular points
the laboratory experimental work were used in the
throughout the system (nodal points), the solution at the
development of an enhanced numerical model (SIPH-
next time level is computed in a two-step predictor
ONET2).
corrector process. This technique is illustrated schema-
tically in Fig. 5, and may be described by rst expressing
4.1. Governing equations
the governing Eqs. (1) and (2) in vector form, thus
The continuity and momentum equations of one qU qF
S, (5)
dimensional, unsteady ow in open channels may be qt qx
ARTICLE IN PRESS
1284 G.B. Wright et al. / Building and Environment 41 (2006) 12791290

Pipe length chosen such that it is in the same direction as that of any
wave moving through the system [6].
The second-order nature of the MacCormack method
means that it generates spurious oscillations in the
New time level (n + 1)
vicinity of hydraulic jumps; oscillations which have no
Corrector step real meaning, and are purely a numerical anomaly. One
Intermediate level
method of alleviating this problem is to add articial
t
(predictor values) viscosity to the solution scheme [6]. In SIPHONET2, the
Predictor step Jameson [10] articial viscosity technique has been
employed. Furthermore, although the governing equa-
Current time level (n)
x x x x
tions have been derived specically for free surface ow
node i - 2 node i -1 node i node i +1 node i +2 cases, utilisation of the Preissmann slot technique [11]
Legend
means that they may also be applied to the short periods
pipe node of pressurised, full bore ow conditions that occur just
information route (data from one node passed to another node)
before the vertical downpipe starts to ll. In common
Fig. 5. Schematic representation of one version of the MacCormack with many such numerical techniques, it is necessary to
method. supply additional information at system boundaries to
allow the solution to proceed.
It should be noted that, as the MacCormack method
where is not readily applicable to vertical pipework, the free
" # " # " # surface ow conditions in the downpipe are assumed to
A Q 0 be annular [12], whilst the lling of this pipework (just
U ; F Q2 ; S .
Q A gI 1 gAS0  S f before the system primes) is simulated using a volu-
metric based technique.
Application of the MacCormack technique now enables
the solution to proceed as follows: 4.2.2. Method of characteristics
As the vertical downpipe within a siphonic system
starts to ll, the system will start to depressurise, thus
1. From known ow variables at time n, the values of rendering the MacCormack method unsuitable. At this
U ni , F ni and Sni can be calculated for all nodes. point, it is necessary to switch over to the classical
2. Predictor values U pi can then be calculated for each method of characteristics solution technique of the
node from governing equations of full bore ow. This technique is
Dt   illustrated schematically in Fig. 6, and may be described
U pi U ni  1  F ni1  1  2F ni  F ni1 DtS ni .
Dx by rst manipulating the governing Eqs. (3) and (4) to
(6) yield two total derivative equations (characteristics), thus
C characteristic:
3. The predictor values and F pi
can then be S pi
calculated for each node from the predictor values dV g dH g f jV j dx
V Sin S0 0 when V  c,
U pi . dt c dt c 2m dt
4. Using the predictor values (again with known U ni (8)
values at time n), the U n1
i values and hence ow C+ characteristic:
variables at time n+1 can then be calculated from
dV g dH g f jV j dx
Dt   V Sin S0 0 when V c.
U n1
i 0:5U ni U pi  dt c dt c 2m dt
2Dx (9)
  Dt
F pi1 1  2F pi   1F pi1 S pi . These equations provide the basis for a nite
2
7 difference solution, yielding two expressions which
may be solved simultaneously for V and H. With
reference to the points R, S and P shown in Fig. 6, these
In the above expressions, the e term may be set equal expressions may be written as
to either zero or one, and is used to vary the direction of  
differencing, e.g. if e is set to 1 forward differencing is g
VP VR  H P  H R  gS f R  S 0 Dt, (11)
used during the predictor stage, whilst backward cR
differencing is used during the corrector step (this is  
the approach shown in Fig. 5). Alternatively, the g
VP VS H P  H S  gS f S  S0 Dt. (12)
direction of differencing in the predictor step could be cS
ARTICLE IN PRESS
G.B. Wright et al. / Building and Environment 41 (2006) 12791290 1285

Pipe length

New time level (n + 1) P

t

Current time level (n)


R S
x x x x

Legend
pipe node
C+ characteristic (downstream information route from one time level to the next time level)
C- characteristic (upstream information route from one time level to the next time level)

Fig. 6. Schematic representation of the method of characteristics applied to full bore ow conditions.

When the system starts to depressurise, the model critical conditions exist at entry, no characteristics are
checks for regions where free surface conditions exist; it available, and it is necessary to assume normal ow
is these regions that correspond to the trapped air conditions.
pockets observed during the experimental work. Once  Two-pipe junction, e.g. 901 bends or changes in pipe
the system is primed, these air pockets are then tracked diameter. Both pipes are treated as a single pipe
as they leave the system, their velocities being set equal length, with varying depth/area characteristics if
to that of the local water ows and their volumes/ necessary, and the conditions are determined by
pressures being dependent on the universal gas laws and solution of the C+ characteristic (upstream pipe) and
local water pressure conditions. Depending on system the C characteristic (subcritical conditionsdown-
layout, certain types of air pockets are assumed to stream pipe, supercritical conditionsupstream
become mixed with the water ow before exiting the pipe). This approach is necessary to ensure the
system, hence forming the type of bubbly ow smooth progress of any upstream moving hydraulic
discussed previously. As with the MacCormack method, jumps, and is considered a valid technique given that
the method of characteristics solution technique requires the degree of ow disruption at such elements is
boundary conditions to enable the solution to proceed. relatively minor under free surface conditions.
 Three-pipe junction (Fig. 8b). As ow is always
4.3. Boundary conditions subcritical at this type of boundary, there are three
characteristic equations available. These character-
When using either of the solution techniques detailed istics are solved with the junction continuity equation
above, the conditions occurring at system boundaries and an empirical expression relating the common
(internal and external) are normally calculated by junction depth to the junction through ow and the
solving the available characteristics with a relationship diameter of the downstream pipe.
relating depth or pressure head to ow rate (Fig. 7).  Horizontal to vertical pipework connection. As ow
These relationships are termed boundary conditions within the system is free surface, there will be a free
(BCs), and can take the form of empirical formulae discharge from the last horizontal pipe section. In
derived from experimental data. Boundaries within free the case of subcritical ow, the available C+
surface ow are dealt with as follows: characteristic is solved with the critical depth equa-
tion, whilst supercritical conditions are calculated
 System entry (Fig. 8a). In subcritical ow, the using the two available characteristics at the down-
available C characteristic is solved with a weir type stream end of the last horizontal pipe (as shown in
expression relating system inow to gutter depth. A Fig. 7, the C characteristic slopes downstream in
volumetric balance of gutter inows (rainfall) and supercritical ow).
gutter outows (entering system) during the previous  System exit. As conditions within the vertical down-
timestep yields the necessary gutter depths. If super- pipe are annular, the conditions at the system exit are
ARTICLE IN PRESS
1286 G.B. Wright et al. / Building and Environment 41 (2006) 12791290

Pipe length

Qentry = f (Hentry) Qexit = f (Hexit)


New time level (n + 1)

t

Current time level (n)


x x x x
node 1 node 2 node n -1 node n

Legend
pipe node
subcritical and supercritical C+ characteristic
subcritical C- characteristic
supercritical C- characteristic

Fig. 7. Schematic representation of boundary solution using either the MacCormack method or the method of characteristics.

Q = function (gutter depth)


C+ equation available at C- equation available at
downstream end of pipe 1 upstream end of pipe 3

Q1 Q3

C+ equation available
at downstream end continuity equation and
of pipe 2 common depth equation at
junction

(a) C- equation available at (b) Q2


upstream end of first pipe

Fig. 8. System entry and three pipe BCsfree surface ow.

determined using a simple annular ow formulation, Q = function (Hg, AC, Hh, Hle)

and no specic BC is required. gutter depth (Hg)= f (inflows, outflows)


air content (AC)= f (gutter depth)

Boundaries within full bore ow are dealt with as hydrostatic head head losses through
follows: (Hh) entry section (HLe)

 System entry (Fig. 9). The available C characteristic


is solved with expressions for gutter depth, air C- equation available at
content of the ow (itself related to gutter depth), upstream end of first pipe

losses through the entry section (outlet, short vertical Fig. 9. System entry BCfull bore ow.
pipe, bend) and hydrostatic head at the base of the
entry section.
 Blocked system entry. This boundary is dealt with in blockage). The method of characteristics solution
a similar way to that detailed above; the only technique then automatically calculates the effect of
difference being that, at a specied time, the inow this event on system conditions.
is set to zero (total, instantaneous blockage) or some  Two-pipe junction, e.g. 901 bends or changes in pipe
percentage of the pre-blockage ow (partial/slow diameter. The available C+ characteristic (upstream
ARTICLE IN PRESS
G.B. Wright et al. / Building and Environment 41 (2006) 12791290 1287

pipe) and C characteristic (downstream pipe) are  The model utilises sharp crested weir theory [6] to
solved with an expression relating the junction head calculate the onset and volumetric extent of ooding
loss to the junction through ow. due to gutter overtopping.
 Three-pipe junction. The two available C+ charac-  Both spatial and timeline interpolation techniques
teristics (upstream pipes) and the C characteristic [11] are employed to determine the base conditions
(downstream pipe) are solved with an expression necessary for the method of characteristics solution at
relating the junction head loss to the junction through each time step.
ow. This BC can also account for the conuence of  The allowable timestep using both solution techni-
ows with different air contents, and the possibility ques is limited by the Courant criteria [6].
that one of the junction branches reverts back to free  Frictional representation is via the ColebrookWhite
surface ow conditions, i.e. drains due to insufcient equation, with due allowance being made for air
inow. content of the ow [11].
 Horizontal to vertical pipework connection. The C+
characteristic available at the exit of the last
horizontal pipe is solved with the C characteristic 4.5. Model output
available at the entry to the downpipe; the method of
characteristics solution automatically accounting for Figs. 1014 show a selection of measured data (from
the change in pipe slope. the test rig) and predicted data (from SIPHONET2)
 System exit. The C+ characteristic available at the obtained under a number of different rainfall/system
system exit is solved with an expression relating exit scenarios. In general, it can be seen that the numerical
head loss to ow. Inclusion of a hydrostatic exit model can accurately simulate a number of important
pressure term also accounts for the possibility of phenomenon, including:
discharge to a submerged or sealed manhole.

It should be noted that the full bore BCs are applied  System priming, including the partial re-pressurisa-
whenever surcharged conditions occur, irrespective of tion that occurs when the main air pocket leaves the
the solution technique being employed. system (Fig. 10).
 Rainfall events below the design condition. Fig. 11
4.4. Additional numerical considerations illustrates that the model accurately simulates the
decrease in gutter depth that occurs as a result of
The following additional points concerning the insufcient gutter inows. As shown, this results in
computational principles underlying the numerical increased air entrainment, decreased ow densities
model are worth noting: and, hence, an increase in the steady state system
operating pressures.
 Once a system entry ows full bore, it is assumed  Variable rainfall conditions. Fig. 12 shows that the
that any ow entering the system is a homogeneous model can accurately simulate the draining and
air/water mixture; the air content, and hence relling of an individual system branch as variations
wave celerity, being a function of the gutter water in gutter inows, and hence gutter depths, results in
depth. the cyclical formation and cessation of siphonic
 As there is no satisfactory, theoretical method of conditions.
simulating the type of pulsing ow conditions that  Outlet blockages, which may result in the generation
were observed to occur downstream of the three-pipe of pressure transients and a decrease in the prevailing
junction under certain circumstances, the model steady state system pressures (Fig. 13).
represents these conditions by assuming full bore  Submergence of the system exit, which will result in a
ow conditions with a high air content (based on the decrease in the available driving head, an increase in
relative ow rates converging at the junction). To system pressures and a resulting decrease in system
simulate the additional head losses incurred due to ow rates (Fig. 14).
the pulsing nature of the ow, an increased pipe
roughness value is utilised downstream of any such The only signicant discrepancy in these gures are
junctions. the oscillations that occur when the system starts to de-
 As well as tracking the progress of the full bore front pressurise, and the model switches from using the
in the vertical downpipe, the model also tracks the MacCormack solution technique to the method of
progress of any free surface fronts, and can hence characteristics (see Fig. 10). These may be considered
simulate the draining of individual system temporary adjustment errors; a fact borne out by the
branches or complete systems as rainfall intensities accuracy of the model immediately prior to and
decrease. immediately after this switch has occurred.
ARTICLE IN PRESS
1288 G.B. Wright et al. / Building and Environment 41 (2006) 12791290

150

Gutter depth (mm)


100 Measured gutter 1 inflow = 5.85l/s
Predicted gutter 1 inflow = 5.91l/s

50 Measured gutter 2 inflow = 7.78l/s


Predicted gutter 2 inflow = 7.80l/s

0 0.3
Numerical oscillations as solution
method switches
0.0

-0.3

Pressure (mH2O)
-0.6

-0.9
Depth in gutter 2 (measured)
-1.2
Depth in gutter 2 (predicted)

Pressure in common pipe (measured) -1.5


Pressure in common pipe (predicted)
-1.8
0 10 20 30 40 50
Time since start of simulated rainfall event (s)

Fig. 10. Measured and predicted conditions for design criteria rainfall.

120
Gutter depth (mm)

80

40

0
Depth in gutter 2 (measured)
0.3
Depth in gutter 2 (predicted)

Pressure in common pipe (measured) 0.0

Pressure in common pipe (predicted) -0.3


Pressure (mH2O)

-0.6

-0.9

-1.2

-1.5

-1.8
0 20 40 60 80
Time since start of simulated rainfall event (s)

Fig. 11. Measured and predicted conditions for a rainfall event below the design criteria (gutter 1 inow 5.10 l/s; gutter 2 inow 6.10 l/s).

5. Conclusions and future work  The priming of a multi-outlet siphonic roof drainage
system is similar to, although more complex, than
The conclusions of the work reported herein may be
that of a single outlet system.
briey summarised as follows:
 A numerical model has been developed which can
 Current design methods may over simplify ow condi- accurately simulate the operational characteristics of
tions within siphonic roof drainage systems, which could multi-outlet siphonic roof drainage systems under all
result in operational problems and/or system failure. realistic scenarios.
ARTICLE IN PRESS
G.B. Wright et al. / Building and Environment 41 (2006) 12791290 1289

160

120

Flow rate (l/s)


80

40
Depth in gutter 1
0 (measured) 0.4
Depth in gutter 1
(predicted) 0.0
Pressure in common
pipe (measured)

Pressure (mH2O)
-0.4
Pressure in common
pipe (predicted)
-0.8

-1.2

-1.6

-2.0
0 25 50 75 100
Time since start of simulated rainfall event (s)

Fig. 12. Measured and predicted conditions for design criteria rainfall event, with no ow into gutter 1 between 55 and 86 s.

0.4 2.5
Measured Predicted

Predicted
0.0 0.0
Pressure (mH2O)
Pressure (mH2O)

-0.4 -2.5

-0.8 -5.0

-1.2 -7.5

-1.6 -10.0
115 125 110 120 130 140
Time since start of simulated Time since start of simulated
(a) rainfall event (s) (b) rainfall event (s)

Fig. 13. Measured and predicted common pipe pressures for design criteria rainfall event, when the outlet in gutter 2 was blocked and reopened. (a)
Partial, slow and brief blockage, (b) total, instant and prolonged blockage.

Although the model performance is good, numerical near future. In addition, modications are being made
stability problems and extended computational run to the existing model to extend its applicability; for
times do restrict its general applicability. However, the example, a simplied version has been developed as
model continues to evolve, and it is envisaged that it part of a collaborative EPSRC research programme
will form the basis of a toolkit to be used for diagno- on the impact of climate change on urban drainage
stic design purposes and code development in the systems [13].
ARTICLE IN PRESS
1290 G.B. Wright et al. / Building and Environment 41 (2006) 12791290

Flow rate (l/s)


time = 49s : depth of submergence = 220mm
4 predicted inflow to branch 1 = 5.51l/s (measured = 5.54l/s)
predicted inflow to branch 2 = 7.27l/s (measured = 7.28l/s)
2 time = 57s : depth of submergence = 380mm
predicted inflow to branch 1 = 5.41l/s (measured = 5.45l/s)

Pressure / Depth of submergence (mH2O)


predcited inflow to branch 2 = 7.11l/s (measured = 7.11l/s)
0 1.0

0.5

Inflow to branch 1 0.0


Inflow to branch 2
Depth of submergence of exit . 5
-0
Pressure in common pipe

. 0
-1

. 5
-1
0 20 40 60 80
Time since start of simulated rainfall event (s)

Fig. 14. Predicted conditions for design criteria rainfall event, with gradually submerging system exit.

Acknowledgements [5] Bowler R, Arthur S. Siphonic roof rainwater drainagedesign


considerations. Proceedings of water supply and drainage for
The researchers remain grateful for the assistance buildings: CIB W62 1999. Edinburgh; 1999.
[6] Chaudhry MH. Open channel hydraulics. New York: Prentice-
given by Dales Fabrications Ltd. (UK), EPSRC (UK), Hall; 1993.
Fullow Ltd. (UK), Geberit AB (Switzerland), HR [7] Chaudhry MH. Applied hydraulic transients. New York: Van
Wallingford Ltd. (UK), Pick Everard (UK), Simona Nostrand Reinhold; 1979.
Ltd. (Germany), Sommerhein AB (Sweden), The Scot- [8] Lister M. Numerical solutions of hyperbolic partial differential
tish Executive (UK), Rattenbury Engineering LLC equations by the method of characteristics. In: Ralston A, Wilf
HS, editors. Numerical methods for digital computers. New
(USA). York: Wiley; 1960.
[9] MacCormack RW. Numerical solution of the interaction of a
shock wave with a laminar boundary layer. Proceedings of second
international conference on numerical methods in uid dynamics,
References Berlin; 1971.
[10] Jameson A, Schmidt E, Turkel E. Numerical solutions of the
[1] May RWP, Escarameia M. Performance of siphonic drainage Euler equations by nite volume methods using Runga-Kutta
systems for roof gutters. Report No SR 463. HR Wallingford; time stepping schemes. Proceedings of AIAA 14th uid and
1996. plasma dynamics conference, Palo Alto, USA; 1981.
[2] Arthur S, Swafeld JA. Siphonic roof drainage: the state of the [11] Wylie EB, Streeter VL. Fluid transients in systems. New Jersey:
art. Urban Water 2001;3(1):4352. Prentice-Hall; 1993.
[3] Arthur S, Swafeld JA. Numerical modelling of the priming of a [12] Swafeld JA, Galowin LS. The engineered design of building
siphonic roof drainage system. Building Services Engineering drainage systems. UK: Ashgate; 1992.
Research and Technology 1999;20(2). [13] Blanksby J, Ashley R, Saul AJ, Cashman A, Packman J,
[4] Wright GB, Swafeld JA, Arthur S. The performance character- Maksimovic C, Jack L, Wright G, Kay D. Adaptable urban
istics of multi-outlet siphonic rainwater systems. Building drainage (AUDACIOUS). Proceedings of NOVATECH 2004:
Services Engineering Research and Technology 2002;23(3) fth international conference on sustainable techniques and
127141(15). strategies in urban water management, Lyon; 2004.

You might also like