4.3.1 Non revenue water (NRW) trend determination
Historical data collected was analysed for the levels of non revenue water (NRW) by water balance method, where the volume of water sold was deducted from the volume of water produced. Farley and Trow (2003) and Lambert (2003) indicated that NRW in a water distribution network is a term which has been recently introduced by the International Water Association (IWA) instead of unaccounted for water (UFW). This is because of the widely interpretations of the term ‘UFW’ worldwide according to Lambert (2001b).
The recommendation by IWA as reported by Liemberger (2002) also suggests that percentage use should only be as a financial performance indicator and that it is unsuitable for assessing the efficiency of management of distribution systems. Lambert (2001b) however indicated that NRW expressed as a percentage of system input is easily calculated and frequently quoted and is certainly the most common indicator. Based on both recommendations, the study analysed NRW in both percentages and quantities’ performance indicators (PIs). The quantities PI included m3/connection/year and m3/km/year, choice of which was based on density of connections as indicated by Charalambous (2005). He reported that when the density of connections for a network is more than 20 connections/km, the recommended PI is m3/connection/year and vice versa.
Increase in NRW levels on an annual basis was determined by determining the average slope of the graph plotted for NRW against time. The results were also confirmed by calculating all yearly increases using excel spreadsheets and average of the increases.
4.3.2 Analysis of minimum night flows
MNF values were estimated from flow measurements by determining the minimum flow value that occurred between 12 am to 4 am as recommended by Farley and Trow (2003). The determined MNF data was analysed using the SANFLOW model (analysis software) which was developed by the South African Water Research Commission (McKenzie, 1999) in order to get unexplained leakage (excess night flows) as well estimated service pipe bursts.
Additional data required to run SANFLOW model included population, number of connections, length of mains, average zone pressure and details of major water users.
Furthermore, other data required such as the various leakage coefficients and pressure correction coefficients, which are not readily available in southern Africa as reported by McKenzie (1999) were assumed as shown in Table 4.1. The assumed parameters were based on default values as suggested by McKenzie (1999) and condition of infrastructure in the specific study areas. Details of the model’s structure and formulation are provided in Appendix A2.
Table 4.1 Assumed parameters for analysis of MNF using SANFLOW model
DESCRIPTION UNITS CHINYONGA BCA
Background losses from mains Litres/km/h 50 30 Background losses from connections Litres/con/h 3 3 Background losses from properties Litres/con/h 1 0.25
% of population active during night flow exercise 6% 1%
Quantity of water used in toilet cistern Litres 10 10 Number of small non-domestic users - - Average use for small non-domestic users - -
Use by large non-domestic users - -
Background losses leakage exponent 1.5 1.5 Burst/leaks leakage exponent 0.5 0.5 Note: BCA main pipeline was replaced two years ago and as such expected to have low background losses. As
regards to properties, all houses in Chinyonga are directly connected whilst in BCA, most of the houses are provided with stand pipes.
4.3.3 Investigation of pressure relationships with leakage, minimum night flow, and burst frequency
Pressure – leakage relationship
WHO (2001) indicated that average daily leakage is obtained from the equation, pf
ENU MNF
Ld =( − )× …...(4.1a) Where, Ld is leakage for the whole day, MNF is minimum night flow, ENU is total expected night use obtained from estimated domestic night and non domestic night use, and pf is pressure adjustment factor for a day. However, leakage at a specific hour of minimum night flow measurement is given by:
ENU MNF
L= − ………...………..(4.1b)
Excel analysis:
Excel spreadsheet charts of pressure against leakage were plotted using all the available chart types such as: power, exponential, and polynomial (best fit) so as to establish a chart type with a good correlation coefficient (R2). The chart with the highest correlation coefficient (R2) at least close to 1.0 was adopted as the one indicating a good relationship.
Theoretical analysis:
Lambert (2001a) as well as Fanner and Thornton (2005) indicated that there is a relationship between pressure and leakage as follows:
1
0 1 0 1
N
P P L
L ⎟⎟
⎠
⎜⎜ ⎞
⎝
=⎛ ………...………(4.2)
Where, P0 = initial pressure; P1 = reduced or increased pressure; L0 = leakage at initial pressure (P0); L1 = leakage at the changed pressure; and whereas N1 is a standard IWA symbol for leakage exponent, which is a coefficient relating pressure and leakage.
Based on principles of the hour day factor (HDF) model as explained by McKenzie et al.
(2002b), the value of N1 was computed using the following rearranged equation:
⎟⎟⎠
⎜⎜ ⎞
⎝
⎛
⎟⎟⎠
⎜⎜ ⎞
⎝
⎛
=
0 1 0 1
ln ln 1
P P L L
N ……….…………(4.3)
The computed values of N1 were then used to determine pressure management opportunities as recommended by Lambert (2001a), Thornton and Lambert (2005), and Fantozzi et al.
(2006) by computing leakage reduction from possible pressure reduction using the following equation:
% 100 1
1
0
1 ⎟⎟×
⎠
⎞
⎜⎜
⎝
⎛
⎟⎟⎠
⎜⎜ ⎞
⎝
−⎛
=
∆
N
P
L P ...……….…...(4.4)
or 1 100%
1
0
0 ⎟⎟×
⎠
⎞
⎜⎜
⎝
⎛
⎟⎟⎠
⎜⎜ ⎞
⎝
⎛ −∆
−
N
P P
P = 1 1 100%
1
0
⎟×
⎟
⎠
⎞
⎜⎜
⎝
⎛
⎟⎟⎠
⎜⎜ ⎞
⎝
⎛ ∆
−
−
N
P
P =
(
1−(
1−%∆P)
N1)
×100%…..….(4.5) Where, ∆L = leakage reduction in percentage, ∆P = reduction or increase in pressure and %∆P= percentage of pressure reduction or increase. A positive result arising from the computation indicated that the pressure was reduced and as such a corresponding reduction in leakage.
Pressure – minimum night flow relationship
The analysis to confirm the relationship between pressure and MNF was done following a statement by McKenzie et al. (2002b) who indicated that UK expressed that there is no relationship between pressure and MNF in 1994, yet did not offer an alternative method. Two approaches of analysis were done to confirm the relationship.
Theoretical analysis:
An equation for confirmation of a relationship between pressure and MNF was developed based on the theoretical relationship of pressure and leakage as indicated by Lambert (2001a).
Since,L∝ PN1 (Lambert, 2001a) and that L= MNF−ENU
Where, L = leakage; P = pressure; ENU = total expected night use; and N1 as defined already.
Then, MNF ∝ENU +PN1 Therefore,
1
KPN
ENU
MNF = + ………..…...…(4.6)
Where, K is a constant and
⎥⎦
⎢ ⎤
⎣
⎡
⎥⎦
⎢ ⎤
⎣
⎡
−
−
=
0 1 0 1
ln ln 1
P P
ENU MNF
ENU MNF
N , in a similar manner as for leakage equation 4.3 ……….….(4.7)
Equation 4.6 was then used to compute MNF for any given average zone night pressure (AZNP). The computed MNF values were compared with the values of measured MNF at the same AZNP by computing deviations so as to confirm the relationship between pressure and MNF.
Excel analysis
Excel spreadsheet charts of pressure against MNF were plotted using all the available chart types such as: power, exponential, and polynomial (best fit). The chart with the highest correlation coefficient (R2) at least close to 1.0 was adopted as the one indicating a good relationship.
Pressure – burst frequency relationship
Thornton and Lambert (2005) suggested that investigations of the relationship between pressure and burst frequency should be done using the following provisional relationship:
Burst frequency (or repairs cost) varies with pressure to the power N2
The relationship was rearranged in similar way as equation 4.2 as suggested by Thornton and Lambert (2005) to give,
2
0 1 0 1
N
P P BF
BF ⎟⎟
⎠
⎜⎜ ⎞
⎝
=⎛ or
⎟⎟⎠
⎜⎜ ⎞
⎝
⎛
⎟⎟⎠
⎜⎜ ⎞
⎝
⎛
=
0 1 2 1
ln ln 2
P P BF BF
N ………..(4.8)
Where, BF0 = burst frequency at initial pressure, P0, BF1 = burst frequency at the changed pressure (P1), and N2 is burst frequency exponent (coefficient relating pressure and burst frequency).
The determined values were then used to determine pressure management opportunities as put forward by Lambert (2001a), Thornton and Lambert (2005), and Fantozzi et al. (2006) by computing frequency reduction from possible pressure reduction using the following equation:
% 100 1
2
0
1 ⎟⎟×
⎠
⎞
⎜⎜
⎝
⎛
⎟⎟⎠
⎜⎜ ⎞
⎝
−⎛
=
∆
N
P
BF P = 1 100%
2
0
0 ⎟⎟×
⎠
⎞
⎜⎜
⎝
⎛
⎟⎟⎠
⎜⎜ ⎞
⎝
⎛ −∆
−
N
P P
P ………...……….(4.9)
Where ∆BF is burst frequency reduction realized upon pressure reduction in percentage.
4.3.4 Examination of pipe failure modes
Collected samples of broken pipes were investigated on their mode of failure in order to relate a pipe material to a specific failure mode within the study area. Upon establishing the most predominant failure mode for a pipe material, the likely leakage exponent was predicted for that pipe material which could be used to make quick estimates of leakage rates. The analysis involved recording number and type of pipe failure modes for each pipe material. The categories for pipe failure modes that were used in the analysis are: longitudinal cracks, circular cracks, hole or pitting, and joint failure. Excel spreadsheets were used to carry out the calculations.
4.3.5 Investigation of prediction tools for pipeline replacement
Basic information collected was used to run an ECONOLEAK model in order to verify it’s suitability in making predictions as regards to pipeline replacement in the study area. The ECONOLEAK model is basically a Microsoft Excel based model designed by the South African Water Research Commission (WRC) to determine an economic level of leakage (McKenzie, 2002). The model uses the bursts and background estimates (BABE) concepts to determine current annual real losses (CARL) and unavoidable annual real losses (UARL). On the other hand, infrastructure leakage index (ILI) is also calculated by:
UARL
ILI = CARL……….………(4.10)
The value of ILI provides an indication of infrastructure condition. Liemberger (2003) and Fantozzi et al. (2006) indicated that for developing nations, desirable values of ILI should be less than 8. Values of ILI which are above 8.0 indicate that the infrastructure is in poor condition and options of replacement could be suggested. Additionally, the model gives the economic leakage level of a system, which is a point where active leakage control (ALC) intervention is cost effective and necessary to be undertaken according to McKenzie (2002).
From the model, the study based its decisions of pipeline replacement whenever the current annual real losses exceed the economic level of leakage point in the model because ALC and repairs only would not be cost effective according to WHO (2001).
An alternative method for making pipeline replacement decisions was investigated to use burst frequency exponent (N2). N2 values computed for a specific pipeline segment would be compared to the values found elsewhere by other researchers, which are reported to be in the range of 0.5 to 6.5. When the value of N2 found exceeds the determined limits it indicates that the infrastructure condition is poor, and as such an indication that replacement is necessary.