Sie sind auf Seite 1von 18

Building Serv. Eng. Res. Technol. 29,2 (2008) pp.

165181

Pressure transient identification of depleted appliance


trap seals: a pressure pulse technique
DA Kelly BEng, JA Swaffield BSc PhD FCIBSE, LB Jack BEng PhD MCIBSE, DP Campbell
BA PhD and M Gormley MSc MPhil PhD MCIBSE
Drainage Research Group, School of the Built Environment, Heriot-Watt University, Edinburgh, UK

The appliance trap seal plays a vital role in safeguarding occupied space from ingress of foul
sewer gases driven by the barrage of pressure transients generated within the system during
normal appliance discharge. The health risks related to depleted trap seals can be severe.
In 2003, the rapid spread of the SARS virus at the Amoy Gardens housing complex in
Hong Kong was attributed to depleted bathroom floor-drain traps.
This paper presents a technique whereby depleted trap seals can be located remotely by
monitoring the system response to an applied single pressure pulse. A Method of
Characteristic based numerical model allows the system pressure response to be predicted
while laboratory and site test results are shown to validate this proposed technique.
Practical application: Appliance trap seal depletion poses a serious health risk by
providing a route for cross-contamination and infection spread. Implementing a routine and
regular maintenance regime would help to ensure that the water level within the trap seal
remains above the critical level. However, current methods rely on visual inspections which
are highly impractical in large complex buildings. A technique allowing the status of all
connected trap seals to be quickly determined would be an invaluable tool for facility
managers by improving operational efficiency and by indicating persistent failures, thus,
highlighting areas requiring modification to ensure performance compliance.

List of symbols
A
c
CJ
CR
CT
C , 
D
f
F( ), f( )
K

Flow cross-sectional area (m2)


Wave propagation speed (m/s)
Compliance factor
Boundary reflection coefficient
Boundary transmission coefficient
Characteristics in xt plane
Pipe diameter (m)
Friction factor
Pressure waves
Loss coefficient

Address for correspondence: DA Kelly, School of the


Built Environment, Heriot-Watt University, Edinburgh,
EH14 4AS, UK.
E-mail: d.a.kelly@hw.ac.uk
Figures 1, 2, 57, 9 and 10 appear in colour online: http://bse.
sagepub.com
The Chartered Institution of Building Services Engineers 2008

L
mSp
N
p
Q
t
T
t
x

u


Subscripts
actual
appliance

Pipe length (m)


System response data
(mm water gauge)
Number of pipes, nodes or defect
Air pressure (mm water gauge, N/m2)
Flow rate (m3/s)
Time (seconds)
Pipe period (seconds)
Time step (seconds)
Internodal length (m)
Density (kg/m3)
Mean air flow (m/s)
Poissons ratio
Ratio of specific heats
Recorded data
Appliance side
10.1177/0143624408090202

166

Depleted appliance trap seal identification

atm
local
m
piston
P, R, S
System
Trap
D
J
t t, t
1, 2, N

Atmospheric conditions
Conditions at node
Monitoring location
Piston conditions
Nodes in MoC calculation
System side
Trap conditions
Defect
Defect number
Conditions at node at a time
Nodes

1 Introduction
The fundamental purpose of the building
drainage and vent system is to rapidly remove
appliance discharge while simultaneously
ensuring that foul air from the drainage network is prevented from entering occupied
space. The primary defence against crosscontamination is provided by the appliance
trap which, provided the water seal is retained, prevents the ingress of sewer gases into
the building. Appliance trap seal retention has
dominated the development of building drainage systems since the 1850s, dictating the
inclusion of ventilation pipes and active control devices, such as air admittance valves and
variable air volume containment devices,
which aim to reduce the effects of air pressure
transients created by normal system operation.1 Trap seal depletion, however, remains
a major issue and when coupled with additional causes of depletion such as evaporation, poor maintenance and bad design, the
potential risk of cross-contamination and
infection spread becomes a serious concern.
The consequence of trap seal depletion
was realised in 2003 following the outbreak
of the SARS virus which infected 8098
people worldwide, resulting in 774 deaths.2
An unusually high number of cases were
reported at the Amoy Gardens housing
complex in Hong Kong where a total of

321 residents were infected. The cause of this


large community outbreak has since been
attributed to oversized bathroom extract fans
and depleted trap seals.3 The operation of the
extract fan created a negative pressure within
the bathroom capable of drawing virus-rich
faecal droplets from the drainage network
into the habitable space via the open path
created by the dry trap seal. Further contamination occurred as the infected faecal droplets were then discharged to atmosphere via
the bathroom extract fan, exposing the virus
to adjacent apartments and neighbouring
buildings. This route of infection spread has
been verified by post-event forensic analysis
by the Hong Kong investigative team4 and
by numerical simulation of the system
operation.5
It is vital that a depleted trap seal is quickly
located and replenished to ensure that
cross-contamination is minimised. Current
methods rely solely on visual system inspections and are wholly impractical and effectively impossible to undertake in large
complex buildings. There is, therefore, an
urgent need to develop a technique whereby
the conditions of all the appliance trap seals
can be determined using a remote and noninvasive methodology.
The technique detailed in this paper uses
the transient response to an applied lowamplitude pressure pulse to detect and locate
an open trap seal. As a transient wave moves
along the system it will be reflected and
re-transmitted by each boundary encountered
whether a branch to stack junction, pressure
control device, stack termination or appliance
trap seal. The system response to such a transient wave can be analysed to decipher important system information conveyed by changes
in the pressure signal. The transmission and
reflection coefficients of each boundary
determine the shape of the resultant system
response. Any changes in boundary condition will result in a change to the system
response.

167

DA Kelly et al.
200
Pipe period

Closed end
Open end

150

Pressure (mm water gauge)

Note subsequent
alternating
positive and
negative peaks

Positive reflection
from closed end

100

50

0
0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.1

1.2

50
100
Negative reflection
from open end
150

Pressure D
transducer

Pressure
transient
generator

LD

Closed End
(CR=+1) or
Open End
(CR=1)

L
LT

200
Time (s)
Figure 1 Laboratory demonstration of the 1/1 reflection coefficients encountered at a closed or open end termination of a single
pipe system subjected to a pressure surge

Figure 1 shows the transient response of a


single pipe system to an applied positive pressure transient and demonstrates the change
in system response to an open and closed end.
The pressure traces are identical up to the
time when the reflected wave arrives at the
monitoring station, at which point the closed
end returns a 1 reflection and the open end
returns a 1 reflection. This normal mode of
transient propagation6,7 can be utilised to
locate a depleted appliance trap seal as the
open trap is analogous to an open end and
will return a1 reflection. Similar techniques
have been developed to identify leaks in large
water supply networks.8,9
1.1 Using pressure transients to identify
a depleted trap
The time, t, taken for the generated
transient to reach a system boundary is

given by:
t

L
c

where L is the pipe length from the monitoring point to the system boundary and c is the
wave propagation speed (around 320 m/s for
air in a pipe at ambient temperature).
The time at which the reflection arrives back
at the monitoring point is known as the pipe
period, T, and may be defined as:

2L
c

Taking the example shown in Figure 1 it


can be seen that the 1 and 1 reflections are
returned at t 0.31 s and, therefore, yields a
pipe length of 50 m. If a depleted trap exists
at point D Figure 1, then the time that the

168

Depleted appliance trap seal identification


50

Pressure (mm water gauge)

45

Defect free (Database)


Defective trap (Test set)

40
Cj=(DatabaseTest set)2

35
30
25
20
15
10
5
0
0

0.1

0.2

0.3

0.4

0.5
Time (s)

0.6

0.7

0.8

0.9

Figure 2 Basis of the compliance factor calculation. The relative divergence of the test data from that of the database is determined
for any particular trap seal failure

negative reflection will arrive at the monitoring point from the open trap will be:
TD

2  LD
c

where TD and LD are the pipe period of, and


the distance to, the depleted trap respectively.
Therefore, Equation (3) can be used to locate
a depleted appliance trap seal (or any change
in system boundary condition) provided the
arrival time of the reflection is known.
2 Identification of system conditions
using numerical analysis
A numerical method, similar to that used to
locate leaks in large scale water distribution
networks9 will be applied to the recorded
system response. The Compliance Factor, CJ,
will be used to measure the closeness of fit
between a system test response and that of a
previously measured or simulated database.
For a system with N appliance trap seals, the
database can be generated by firstly measuring the response of the system, at location m,
with no depleted traps (defect free system

denoted as 0) and then repeating for each


failure condition by sequentially depleting
each of the traps. This will provide a full set
of system responses, mSpJ1N, from which
to build the database. The closeness of fit
between any future test case, mSpactual, and
the database can be determined using a least
squares approach by summing the squares of
the differences between the two:
npoint
X
CJ
mSpJ1N  mSpactual 2 4
i1

where npoint is the number of data points to


be compared within each trace, see also
Figure 2. When the system test response corresponds to a database trace, then the value of
CJ tends to zero. Thus, the matrix created by
application of Equation (4) identifies the correct system condition, whether it be defect
free or one of the failure cases, by returning
a value of CJ which approaches zero.
3 Mathematical modelling
To model the propagation of low-amplitude
air pressure transients within the building

DA Kelly et al.

169

x +ve
Transient

Entry

Reflection

Exit

Boundary condition
solved with C
characteristic

C+

C+

Boundary condition
solved with C+
characteristic

Figure 3 Method of characteristics representation of air pressure transient propagation and reflection within a network branch

drainage and vent system, the accepted


Method of Characteristics (MoC) solution
of the equations of continuity and momentum
may be applied. The MoC solution was introduced by Lister,10 and was primarily applied
to large scale civil engineering waterhammer
studies. Since then, this method has been further developed and has been applied to such
areas as aircraft refuelling and free surface
drainage flows as well as to traditional
waterhammer.1113
The equations of continuity and momentum for the case of low-amplitude air pressure
transient propagation may be shown as
follows:14
Continuity: 

@u
@p @p
u 0
@x
@x @t

@p
@u
@u 4fujuj
u 
0
@x
@x
@t
2D
6
For low-amplitude air pressure transient
propagation, the density and pressure variables are linked via the wave propagation
speed, c:
 0:5
p
7
c

Momentum:

and, therefore, it is necessary to express the


equations in terms of wave speed and fluid
velocity:
c2



@u
2
@c
@c
0

c
u
@x   1 @t
@x



2
@c
@u
@u
4fujuj

c
u
0
  1 @x
@t
@x
2D

This pair of quasi-linear hyperbolic partial


differential equations may be solved via a
finite difference solution when transformed
via the MoC into the finite difference relationships represented by Equations (10) to (13)
that link conditions at a node one time step in
the future to current conditions at adjacent
upstream and downstream nodes (Figure 3).
For the C characteristics:
2
cP  cR
1
t
4 fR uR juR j
0
2D

uP  uR

when

dx
u c,
dt

10

11

170

Depleted appliance trap seal identification

And for the C  characteristics:


2
cP  cS
 1
t
4 fS uS juS j
0
2D

uP  uS

12

when
dx
u  c,
dt

13

These equations are set in terms of the air


mean flow velocity, u, at any section and the
local wave speed, c, at that location. The
choice of u and c, results from the interdependence of air pressure and density. As a
result it is necessary to determine pressure at
each node and each time step as:

plocal

patm
atm




c 2local

of a trap seal, are based on time or another


variable such as local pressure.
In the case of a junction of N pipes, some
terminating and some originating at the
junction, there is a need for 2N equations to
solve for airflow and pressure at the first and
last nodes of these pipes. Note that pressure is
taken as defined by wave speed through
Equation (14) and that initial flow velocity
is assumed positive in the direction of
increasing distance, x. In addition, continuity
of flow:
XN
1

Q0

15

and either the equivalence of pressure at the


junction, (N  1) equations:
p1 p2    pN

16

or a representation of the junction local loss:

 1=1
14

3.1 Simulating boundary conditions


Figure 3 shows the Courant Criterion,
namely t  x/(u c), that links internodal
distance to time step in terms of the airflow
velocity, u, and wave speed, c. It may also be
seen that only one characteristic equation will
exist at a pipe boundary, a C at a termination or a C  at an entry, introducing the
need for a boundary equation linking airflow
conditions to applied water flow or other
system parameters. Previous papers by
Swaffield and Campbell14,15 established the
concept of system boundaries and introduced
the categories of active or passive to differentiate boundary types. Passive boundaries,
e.g. pipe junctions or open and closed ends,
are independent of time and the local
unsteady flow regime while active boundaries,
e.g. the operating characteristics of an air
admittance valve (AAV) or the displacement

p1 p2 K12 Q21    pN K1N Q21

17

Together with the N characteristics provide


the 2N equations required.
At an open termination to atmosphere the
boundary condition to be solved with the
available C characteristic is represented by
constant pressure:
plocal patm

18

or some known external air pressure history, e.g. wind shear over roof terminations
or pressurisation of the habitable space.
Similarly, at a closed end the boundary
condition to be solved with either the available
C or C  characteristic is provided by
putting the local airflow mean velocity to zero:
ulocal 0

19

An AAV boundary depends upon the local


air pressure at the branch to AAV interface.16

DA Kelly et al.

The boundary can therefore be zero velocity


if the pressure is above the AAV opening
pressure level. If the suction in the pipe is
sufficient to open the AAV, then the boundary equation is provided by the pressure loss
through the valve at the flow rate and opening ratio, which in turn depends on local line
pressure. The AAV boundary condition may
thus be written as:


plocal patm  Klocalp Qlocal Qlocal 

20

The value of the AAV loss coefficient


Klocalp will vary decreasing as the diaphragm
lifts in response to greater suction pressures
until the valve is fully open and thereafter
having a constant value.14
The boundary condition applicable for a
trap seal is based on the application of the
equation of motion to the seal water column:
plocal, t t  patm proom
gHsystem  Happliance

curvature of the trap. This failure condition is


identified by monitoring the trap seal length
at each time step. Comparison with the
geometry of the trap allows the trap seal
boundary condition to be modified to represent a concentrated loss in either of these
cases. These cases are referred to as bubble
through trap seal failures and are difficult to
identify post-event as the water column will
revert to an equilibrium position as the
transient event abates.16
3.2 Transient reflection
Any change in operating system will result
in the transmission of a transient which
propagates throughout the network at the
appropriate acoustic velocity, c. Transients
are reflected at each boundary encountered
and reflections and re-transmissions occur.
Returning to the frictionless equations of
motion and continuity allow the propagation
of transients and their interaction with the
flow network to be explained:
Equation of Motion

 0:5ut t ut t  Ltrap 32=D2trap


 Ltrap Apt ut t  ut =t 0

@p
@v
 0
@x
@t

21

solved with the available C (Equation (10)).


This is inevitably an iterative solution as both
plocal,t t and ut t are unknown, the velocity term being defined by the C characteristic (Equation (10)) coupled with the wave
speed pressure relationship (Equation (14)).
Reference to Equation (21) illustrates the
importance of trap seal water retention. The
mass of the trap seal water, represented in
Equation (21) by the length of the water
column, Ltrap, is monitored at each time step.
Obviously, suction in the system will result in
water loss as the system-side height, Hsystem,
cannot exceed zero water flows into the
branch depleting the trap.
Either suction or positive pressure can displace the water column so that an air path is
formed through the trap as soon as the water
level on either side falls to the lowest point of

171

22

Equation of Continuity
@p
@v
c2 0
@t
@x

23

The general solution of these equations is


due to DAlembert and may be expressed as:


x
x
p  patm F t
f t
c
c

V  Vatm 

24


1 h 
x
xi
F t
f t
25
c
c
c

The F() and f() functions are entirely


arbitrary and may be selected to represent

172

Depleted appliance trap seal identification

transient propagation upstream and downstream from the site of the boundary condition change. The F( ) and f ( ) functions
conform to the principle of superposition of
pressure waves and may therefore be used to
develop and explain the more complex pressure time histories associated with transient
propagation in flow networks. The imposition
of a frictionless system excludes attenuation,
however, this approach allowed Allievi to
introduce the first graphical technique for
transient prediction.
Joukowsky17 dealt with two particular
boundaries that remain of fundamental
importance: The presence of either a dead
end or a fully open termination into a zone
held at constant or known pressure.
In the first case, the local velocity remains
zero provided that the pressure at that location remains above the fluid vapour or
dissolved gas release pressure. Thus setting
(V  Vatm) 0 in Equation (25) gives F( )
f ( ) and, therefore, the incoming pressure
wave is reflected with the same magnitude
and sign effectively doubling the local effect
of the incoming pressure wave. The reflection
coefficient may be expressed as:
CR 1
at a dead end.
In the case of a constant pressure at the
boundary, placing (p  patm) 0 in Equation
(24) implies that F( ) f ( ), and hence a
reflection coefficient at a constant pressure
zone boundary of:
CR 1
or a reflection of the incoming wave with a
change of sign.
The frictionless wave Equations, (24) and
(25), may be solved at any junction boundary
together with the equations of flow continuity
and pressure equivalence across the junction
to determine values for reflection and

transmission coefficients for the junction of


n pipes (Figure 4):
Continuity of flow
Q1 Q2 Q3    Qn

26

Commonality of pressure at the junction


p1 p2 p3    pn

27

Assuming that the incoming transient arrives along pipe 1, ongoing transmissions will
be generated along pipes 2 ! n and a reflected
transient will propagate back along pipe 1
based on pipe area and local wavespeed.
Solution of Equations (24) and (25) with
(26) and (27) for an incoming transient F1
in pipe 1, a transmitted wave F2 ! Fn in
pipe 2 ! pipe n and a reflected wave f1 in
pipe 1 yields a general expression for the
reflection and transmission coefficient respectively for the junction:
CR

CT

A1 =c  A2 =c  A3 =c      An =c
A1 =c A2 =c A3 =c    An =c
28
2A1 =c
A1 =c A2 =c A3 =c    An =c
29

Several points emerge from these expressions: The transmission into each of the
receiving pipes is identical and the transmission and reflection depend upon both the area
of the pipe and the wave speed within it,
which in turn involves the pipe material and
pipe wall thickness to diameter ratio. It will be
seen that it is the sum of the (area/wave speed)
that determines the coefficients, hence one
pipe with a low wave speed due to its wall
thickness or material can dominate the calculation. In the application discussed in this
paper, the wave speed in all pipes may be

DA Kelly et al.

173

F2

Pipe 2

F1
F3
Pipe 1

Pipe 3

f1

Fn
Pipe n
Figure 4 General pipe junction with n number pipes

assumed to be constant. As the pressure


transients are small, the system pipework
may be assumed rigid, simplifying the form
of Equations (28) and (29) to area ratios.
These expressions will become important
later in defining the transmission of a pressure
transient through a complex network.
Equation (29) will become important in
determining the effective range of any transient as it effectively determines the transient
strength after passing several junctions. The
transient will depend upon the relative crosssectional areas of the stack and branches.

4 Laboratory evaluation of the pressure


transient identification technique
A laboratory evaluation of the proposed
technique was undertaken using a simulated
77 m high vertical drainage stack consisting of

75 and 100 mm diameter pipework. Branch


connections leading to single appliance trap
seals were located at 3.2 m centres representing typical floor levels (Figure 5). A pneumatically operated piston, or pressure transient
generator (PTG), was located at the simulated
stack base which, on activation, would
impose a positive transient pulse into the
system, the response of which was recorded
by a pressure transducer located approximately 1 m from the PTG and connected to a
high scan rate data logging system.
Initially all traps were capped off to
represent a set of full appliance trap seals
providing the defect free system for comparison. The system response to the applied
single pressure pulse was then monitored for
each failure condition by sequentially removing the cap from trap 1 (T1) to trap 14 (T14),
thus exposing the trap to atmosphere and
simulating a depleted appliance trap seal.

174

Depleted appliance trap seal identification


Open
termination
19.7 m
100 mm

6.3 m
T14
T13
T12
T11
T10
T9
41.6 m
(13 3.2m)

T8
T7
T6

and are displayed as vertical lines along the


x-axis.
The generated transient arrives at the
transducer at t 1.0 s. The transient response
for the defect free system shows that the
pressure within the pipe continues to rise (to
around 100 mm Water Gauge) during the
motion of the PTG until a negative wave is
generated between t 1.35 and t 1.4 s by the
abrupt cessation of the piston motion. During
this period there are no negative reflections
returned by any of the traps, thus confirming
that all traps are full. The introduction of a
dry trap at T1, however, generates a negative
reflection with an arrival time of t 1.057 s
(pipe period of 0.057 s) which yields a trap
distance of 9 m (consistent with the location
of T1, see Figure 5). The arrival time of the
reflections returned from the depleted traps at
T3 and T12 also show good correlation with
the predicted pipe periods.

T5

5 Numerical modelling of the


system response

T4
T3
T2
T1

75 mm

2.8 m
100 mm
6.2 m
P Pressure transducer

The AIRNET computer simulation, based


upon the MoC solution of the St Venant
defining equations of unsteady flow, has been
used to model the system response of the
laboratory system. The action of the PTG at
the base of the stack is defined by the
boundary equation:

0.37 m
150 mm
Pressure transient
generator
Figure 5 Schematic of laboratory test rig

Figure 6 illustrates the measured system


response to the applied pressure transient
imposed by the PTG for the defect free system
and selected defective traps at T1, T3 and
T12. The pipe periods for each trap location
have been calculated using Equation (3)

xpiston, t t  xpiston, t
t

30

which may be solved with the available C 


characteristic to yield c and hence p at the
piston face. Figure 7 shows the simulated
transient response for the defect free system
and defective traps at T1, T3 and T12. As was
shown with the measured data, the return
times of the negative reflections generated by
the open traps correspond with those predicted theoretically. This high degree of agreement between the measured and simulated

Pressure (mm water gauge)

175

End

T14

T13

T12

T11

T10

T9

T8

T7

T6

T5

Defect at T3

80

T4

Defect at T1

100

T3

Defect free

120

T2

140

T1

DA Kelly et al.

Defect at T12

60
40
20
0
0.95
1
1.05
Negative
40 reflection
60 returned from
80 open trap T1

1.1

20

1.15

1.2

100

1.25

1.3

1.35

1.4

1.45

1.5

Time (s)

Pressure (mm water gauge)

100
80

End

T14

T13

T12

T11

T10

T9

T8

T7

T6

T5

T4

T3

Defect Free
AIRNET
Defect at T1
AIRNET
Defect at T3
AIRNET
Defect at T12
AIRNET

120

T2

140

T1

Figure 6 Pressure response of the defect free system compared with those with a depleted trap seal

60
40
20
0
20
40
60
80
100

0.95

1.05

1.1

1.15

1.2

1.25

1.3

1.35

1.4

1.45

1.5

Negative
reflection
returned
from open
trap T1
Time (s)

Figure 7 AIRNET simulated pressure response of the defect free system compared with those with a depleted trap seal

results provide confidence in the method of


characteristics simulation as a tool to predict
system response and as an appraisal of this
monitoring technique.
6 Site test validation
Validation of the proposed transient analysis
technique was provided under controlled site
test conditions using a standard single stack
drainage system in an unoccupied 17-storey

residential building in Dundee, Scotland.


The system consisted of a 150 mm cast iron
vertical stack, with 100, 35, 42, and 54 mm
diameter copper branch connections serving
the WC, WHB, bath and kitchen sink, respectively within each apartment. The PTG
was connected into the stack via existing
access panels at the three test locations as
indicated on Figure 8. These test locations
were chosen to investigate the practical
considerations of applying the transient
to the top, middle and bottom of the

176

Depleted appliance trap seal identification

drainage stack. The findings are discussed in


the following sections.
The system response was recorded using a
number of calibrated pressure transducers
connected to a high scan rate data logging
PC board, collected at a 500 Hz sampling
frequency. The results presented in this paper
are those recorded on the connecting branch
between the PTG and stack for Test Set 1 and
Test Set 2 only. The tests performed at the
bottom of the stack (Test Set 1) and the top of
the stack (Test Set 3) differ only by the order
in which the appliance trap seals are encountered by the incoming transient, T1 ! T17
(Test Set 1) and T17 ! T1 (Test Set 3).

Termination
open to atmosphere

0.9 m
WHB
Bath
WHB
Bath
WHB
Bath
WHB
Bath
WHB
Bath
WHB
Bath
WHB
Bath

6.1 Applied pressure transient to ground floor


level (Test Set 1)
Samples of the transient response recorded
at Test Set 1 are shown in Figure 9. It can be
seen from Figure 9(a) that the traces are
dominated by a large negative reflection,
which is returned before any of the predicted
trap pipe periods, at t 1.016 s; suggesting a
change of boundary condition at 2.61 m from
the transducer.
Once the transient arrives at the branchto-stack junction, ongoing transmissions will
be propagated along both the upper stack
section (toward the termination) and lower
stack section (toward the sewer) while a
reflected transient will be propagated back
along the incoming branch towards the PTG.
At a three-pipe junction such as this, with
pipes of matching diameter and wave speed,
the transmission and reflection coefficients will
be 2/3 and 1/3, respectively (see Equations (28)
and (29)). As the pressure transient propagates
throughout the upper and lower stack sections
it will be reflected and transmitted by each
boundary encountered. Therefore, the measured system response is a superposition of the
response from the three-pipe junction and the
reflections from both upper and lowers sections of the stack. The large negative reflection
from the sewer connection was removed by

WHB
Bath
WHB
Bath
41.6m
(16 2.6m)

WHB
Bath
WHB
Bath
WHB
Bath
WHB
Bath
WHB
Bath
WHB
Bath
WHB
Bath

Sink
WC (T17)

Test set 3
PTG
17F

Sink
WC (T16)

16F

Sink
WC (T15)

15F

Sink
WC (T14)

14F

Sink
WC (T13)

13F

Sink
WC (T12)

12F

Sink
WC (T11)

11F

Sink
WC (T10)
Sink
WC (T9)

10F
Test set 2
PTG
9F

Sink
WC (T8)

8F

Sink
WC (T7)

7F

Sink
WC (T6)

6F

Sink
WC (T5)

5F

Sink
WC (T4)

4F

Sink
WC (T3)

3F

Sink
WC (T2)
Pressure
Transducer

5.8 m

Roof

2F
Test set 1
PTG

0.8 m
GF

To sewer

Figure 8 The single-stack drainage system used for site testing.


Note: the ground floor level is double height with accommodation beginning at 2nd floor

Pressure (mm water gauge)

(a)

Defect free
Defect at T3
Defect at T7
Defect at T16

50
40

177

T2
T3
T4
T5
T6
T7
T8
T9
T10
T11
T12
T13
T14
T15
T16
T17
TOP

DA Kelly et al.

WHB
Bath

30
20

Sink

2F

WC (T2)

Pressure
transducer
PTG

10

GF

0.9

0.95

1.05

1.1

1.15

1.2

1.25

1.3

10

1.35

1.4

1.5

1.45

Sewer

20

Time (s)

(b)
50

T2
T3
T4
T5
T6
T7
T8
T9
T10
T11
T12
T13
T14
T15
T16
T17
TOP
Stack

Defect free

Pressure (mm water gauge)

Defect at T2
Defect at T6

40

WHB

Defect at T11

Sink

Bath

2F

WC (T2)

30
Pressure
transducer

20

PTG

10
0

Lower stack
blocked

0.9

0.95

1.05

1.1

1.15

1.2

1.25

1.3

1.35

1.4

1.45

GF

1.5

10
20
Time (s)
Figure 9 Site test pressure response of the defect free system compared with those with a depleted trap seal for Test Set 1 with
the: (a) sewer open, (b) sewer closed. Note: Test Set 3 results generally replicate those shown here

blocking the lower stack connection effectively transforming the three-pipe junction into
a two-pipe junction. Figure 9(b) demonstrates
the improvement made to the magnitude of
the propagating wave. It is now possible to
distinguish a negative reflection returned from
the open upper termination at t 1.3 s from
the defect free system trace, giving a pipe
period of 0.3 s and a subsequent stack height
of 48 m. On introduction of an open trap at
T2, the negative reflection is returned at
t 1.048 s giving a pipe period of 0.048 s and
a reduced transient travel distance of 7.7 m

(matching the pipe period of trap T2). The


remaining cases shown in Figure 9(b) confirm
this relationship.
The measured traces illustrated in
Figure 9(b) show a more rapid natural attenuation than those observed in the laboratory
tests. This can be attributed to an increase in
frictional loss due to the greater pipe wall
roughness of the older cast iron pipework
compared with that of the new uPVC pipework used in the laboratory together with
the larger number of appliance junctions on
each floor which reduces wave transmission.

Top of
stack
Sewer

T17

T16
T2

T15
T3

T14
T4

T12

T13
T5

Defect at T13

30

T6

Defect at T2

T11

Defect free

T7

40

T10

50

T8

(a)

T9

Depleted appliance trap seal identification

Pressure (mm water gauge)

178

Defect at T11
WHB

20

Sink

Bath

WC (T10)
Sink

WHB

10

Bath

0
10

WC (T9)

WHB

0.95

1.05

1.1

1.15

Bath

1.2

10F
PTG
9F

Sink

1.25

WC (T8)

1.3

8F

20

Defect at T11

Top of
stack

T17

T16

T15

T14

Defect at T10

T13

Defect free

T12

40

T11

50

Time (s)
T10

(b)
Pressure (mm water gauge)

30

Defect at T12

30

Defect at T13

WHB

Defect at T14

20

Defect at T16

10

Bath

0
1

1.05

1.1

1.15

10

WHB

WC (T10)
Sink

WHB

Defect at T17

0.95

Sink

Bath

Defect at T15

1.2

Bath

WC (T9)

10F
PTG
9F

Sink

1.25

WC (T8)

8F

1.3

20
30

Time (s)

Figure 10 Site test pressure response for the defect free system compared with those with a depleted trap seal for Test Set 2:
(a) both upper and lower stack sections open, (b) lower stack section blocked off

6.2 Imposed pressure transient applied at


mid-floor level (Test Set 2)
The initial transient response of introducing the incident transient at the middle of the
stack (Test Set 2) is shown in Figure 10(a).
Again, a large negative reflection is returned
earlier than any of the predicted trap pipe
periods at T 1.014 s. This reflection is generated by the branch-to-stack three-pipe junction, see Equations (28) and (29). To exclude
this unwanted reflection, each upper and
lower stack section may be treated sequentially by closing the relevant section.

A further issue requiring consideration


when introducing the incident transient at a
mid-floor level is the possibility of mirror
reflections being returned from defective traps
on equidistantly spaced floors. For example,
from Figure 10(a), it can be seen that the pipe
period for T7 in the lower stack section is
similar to that of T11 in the upper stack
section and, therefore, both will return a
negative reflection at a similar time. It is,
therefore, not possible to clearly identify which
of the two traps are defective from a single
monitoring point. Therefore, to remove this

DA Kelly et al.

179

Table 1 Compliance factor matrix for Test Set 1 with sewer blocked off (first 10 traps only). The correct system status is identified
when the test set matches the database trace giving Cj ! 0
Test
CF
Databank

DF
T2
T3
T4
T5
T6
T7
T8
T9
T10

DF

T2

T3

T4

T5

T6

T7

T8

T9

T10

13
45321
43434
41413
40262
41125
38074
32263
24559
24851

46077
24
14781
39460
53771
58996
58393
56374
49237
50448

42929
14740
33
13354
35197
49545
57808
62118
58791
61020

44135
41032
15134
69
11131
28549
44735
56077
62305
70317

41704
54749
37900
11607
39
6578
19992
31658
41079
51009

41150
59162
52529
29447
7537
91
4873
13910
23330
31556

42486
62614
63490
46005
21256
5986
218
4246
12867
19613

32767
56050
64233
55346
32074
15423
4011
59
3228
8439

26049
50555
61833
61663
41266
24670
11215
3570
92
2203

25613
50386
62338
67577
49772
32159
17225
8379
2348
52

unwanted reflection both the upper and lower


sections of the stack were tested separately by
blocking off the corresponding opposite section. Figure 10(b) shows the results after the
lower stack section was blocked off.
6.3 Application of the compliance factor
Analysis of the recorded system pressure
response has been shown to give good visual
confirmation of the location of depleted trap
seals by proving that the negative reflection
returned from an open trap corresponds to
the theoretical pipe period for that trap. It is
also important, however, to have a numerical
confirmation of the system condition to allow
automatic recognition of the system condition. The pressure compliance factor is used
to produce a quantitative value of closenessof-fit between a test trace and a pre-recorded
database trace (Equation (4)).
As an example of this numerical analysis
the pressure compliance factor has been
determined for the test data recorded during
Test Set 1 with the sewer blocked off. The
effectiveness of the pressure compliance
factor relies heavily on the requirement that
all pressure responses contained within the
database must have close correlation with the
recorded test traces. Any additional system
noise or change in operating conditions will
produce dissimilar traces causing the
pressure compliance factor to be inaccurate.

Initially, a single test trace of each defective


trap was compared with that of a single
database trace yielding a successful defective
trap identification rate of only 50%. This low
success rate can be attributed to a variance in
the measured transient generated by the PTG
due to occasional sticking of the piston on
activation which had the effect of increasing
the travel speed and, therefore, the pressure
generated by the piston. Thus, to reduce this
variation, an average pressure trace was
calculated for both the test trace and the
database trace, each from five single traces,
greatly improving the correlation.
Table 1 shows the pressure compliance
factor matrix calculated from these average
traces and for the first 0.3 s following the arrival of the pressure wave (the system pipe
period). Note that the compliance factor values
are dependant upon the test duration being
analysed which is itself dependant upon the
maximum pipe period of the connected traps.
In each case, the defective trap is correctly
identified with the pressure compliance factor
approaching zero when compared to the
matching system condition in the database.

7 Conclusion
Following the SARS epidemic at the Amoy
Gardens in 2003, the vital role of the

180

Depleted appliance trap seal identification

appliance trap seal in protecting building


inhabitants from the potentially virus-laden
air contained within the building drainage
system has gained much attention. The
remote and non-invasive technique detailed
in this paper allows depleted appliance trap
seals to be quickly identified to ensure that
the risk from cross-contamination is minimised. This not only has obvious health and
safety benefits but will also prove beneficial to
the buildings facility management team by
greatly improving the effectiveness of their
maintenance regime.
The proposed technique draws on the
principles of pressure transient theory and in
particular uses the change in reflection coefficient between an open and closed boundary
condition, and the change that this has on the
characteristic system response, to identify the
location of a depleted appliance trap seal.
It has been shown through laboratory and site
test results, and corroborated by numerical
simulation, that a depleted appliance trap seal
returns a negative reflection when subject to
an applied pressure transient. The time of this
returning reflection, obtained from the measured system response, has been shown to
accurately match that of the calculated trap
pipe period and when compared with that of a
defect free trace allows the location of the
depleted trap to be identified.
The pressure compliance factor method
used to identify the correct system condition
has been shown to be highly successful when
comparing a system test response with a database of previously measured baseline cases.
This numerical analysis requires that the
operating conditions used to collect the database traces closely match those used in subsequent system testing to ensure that a good
fit is achieved. Taking multiple test traces and
calculating an average pressure-time trace
improved the correlation between the test
and database traces.
The applicability of this technique requires
that the applied pressure transient be directed

to propagate in one direction along the


drainage stack to ensure that the superposition of reflected boundary condition information from equidistant pipe sections is
avoided. Methods of achieving this requirement without compromising the integrity of
the system are currently being investigated.
However, the main advantage of this technique is that the whole test period is very
short. For the 48 m high building used on the
site investigations, the required testing duration is within 0.3 s, as all of the physical
characteristics of the system can be defined
within the first 2L/c seconds of testing, clearly
a practical time limit for a technique which
is envisaged to take place during periods
of system non-use, e.g. during the night.
The operation of this technique could be
incorporated into the building management
system and programmed to perform daily
system checks.
The next stage in the development of this
technique will be to investigate the use of a
sinusoidal pressure wave as the input transient and also to undertake a full set of
operational trails within an occupied
building.

References
1 Swaffield JA, Galowin LS. The engineered
design of building drainage systems. 1992;
England, Ashgate Publishing Limited.
2 World Health Organisation. Summary of SARS
in the WHO European Region, CD News,
Communicable Disease Report, No. 31, Oct.
2003.
3 World Health Organisation. Inadequate
plumbing systems likely contributed to SARS
transmission. Press Release WHO/70, 26
September 2003, WHO, Geneva.
4 Hung HCK, Chan DWT, Law LKC, Chan
EHW, Wong ESW. Industrial experience and
research into the causes of SARS virus transmission in a high-rise residential housing estate

DA Kelly et al.

6
7
8
9

10

11

in Hong Kong. Building Services Engineering


Research and Technology 2006; 27(2): 91102.
Jack LB. Drainage design: factors contributing
to SARS transmission. Proceedings of the
Institute of Civil Engineers, ICE Municipal
Engineer 159, March 2006; Issue ME1: 4348.
Wylie EB, Streeter VL. Fluid transients.
New York, McGraw-Hill, 1979.
Swaffield JA, Boldy AP. Pressure surge in pipe
and duct systems. England, Avebury Technical
Press, 1993.
Brunone B. Transient test-based technique for
leak detection in outfall pipes. J Water Resour
Plan Manage 1999; 125(5): 30206.
Stephens M, Vitkovsky J, Lambert M,
Karney B, Nixon J. Transient analysis to assess
valve status and topology in pipe networks: 9th
Int. Conf. Pressure Surge, BHR Group,
Chester, 2426 March 2004; 21124.
Lister M. The numerical simulation of hyperbolic partial differential equations by the
Method of Characteristics. In: Ralston and
Wilf (eds), Mathematical methods for digital
computers. New York, John Wiley, 1960;
pp. 16579.
Doyle TJ, Swaffield JA. An evaluation of the
method of characteristics applied to a pressure
transient analysis of the Concord refuelling
system: Proceedings of the Institution of
Mechanical Engineers, Part E, 1972; 216:
13342.

181

12 McDougal JA, Swaffield JA. The influence of


water conservation on drain sizing for building
drainage systems. Building Services
Engineering Research and Technology 2003;
24(4): 22943.
13 Streeter VL. Waterhammer analysis including
fluid friction. J Hyd Div ASCE 1962; 39:
12731.
14 Swaffield JA, Campbell DP. Numerical
modelling of air pressure transient propagation in building drainage systems, including
the influence of mechanical boundary conditions. Building and Environment 1992; 27(4):
45567.
15 Swaffield JA, Campbell DP. Air pressure
transient propagation in building drainage
vent systems, an application of unsteady
flow analysis. Building and Environment 1992;
27(3): 35765.
16 Swaffield JA, Jack LB. Simulation and
analysis of airborne cross contamination
routes due to the operation of building
drainage and vent systems. Building
Research and Information 2004; 32(6):
45167.
17 Joukowsky N. Uber den hydraulisher Stoss in
Wasserlietungsrohen. Memoirs de lAcademie
Imperiale des Sciences de St Petersburg, 1900,
translated by Simin O. Proceedings
American Water Works Association, 1904; 24:
341424.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Das könnte Ihnen auch gefallen