Statistical analysis and definition of blockages-prediction
formulae for the wastewater network of Oslo by evolutionary
computing
Rita Ugarelli1*, Stig Morten Kristensen2, Jon Røstum 3, Sveinung Sægrov3, Vittorio Di Federico1
(1) Bologna University, D.I.S.T.A.R.T., Viale Risorgimento 2, 40126 Bologna, Italy. (2) NTNU, Norwegian University of Science and Technology, Dept. of Hydraulic and
Environmental Engineering, NO-7491 Trondheim, Norway.
(3)SINTEF, Building and Infrastructure Department, N7465, Trondheim, Norway.
*Corresponding author, e-mail [email protected]
ABSTRACT
Oslo Vann og Avløpsetaten (VAV) – the water/wastewater utility in the Norwegian capital city of Oslo – is assessing future strategies for selection of most reliable materials for wastewater networks, taking into account not only material technical performance but also material performance, regarding operational condition of the system.
The research project undertaken by SINTEF, NTNU and Oslo VAV adopts several approaches to understand reasons for failures that may impact flow capacity, by analysing historical data for blockages in Oslo.
The aim of the study was to understand whether there is a relationship between the performance of the pipeline and a number of specific attributes such as age, material, diameter, to name a few. This paper presents the characteristics of the data set available and discusses the results obtained by performing two different approaches: a traditional statistical analysis by segregating the pipes into cohorts, each of which with the same explanatory variables, and a Evolutionary Polynomial Regression model (EPR), developed by Technical University of Bari and University of Exeter, to identify possible influence of pipe’s attributes on the total amount of predicted blockages in a period of time.
KEYWORDS
Sewer, EPR (Evolutionary Polynomial Regression), Blockages, Oslo
INTRODUCTION
Blockages in the sewerage system represent an expensive nuisance to householders and businesses through loss of service and inconvenience. Blockages can also be a cause of pollution and health risk if discharge of raw sewage follows the wrong path.
In this study the collection of data related to blockages occurring in the wastewater network of Oslo was performed with the aim to link data on the past events with the characteristics of the pipeline.
Two approaches have been applied within this study: the EPR tool and statistical analysis; both are briefly described below.
Berardi et al., 2006, presents the use of a hybrid data-driven modelling technique called the Evolutionary Polynomial Regression (EPR) for correlating the number of sewer blockages
Orazio Giustolisi (Technical University of Bari) and Prof. Dragan Savic (University of Exeter).
The methodology is based on both numerical and symbolic regression. EPR uses a genetic algorithm to find the form of polynomial expressions, and least squares optimisation to find the values for the constants in the expressions. The incorporation of least squares optimisation within symbolic regression enables fast and effective model building (Giustolisi et al., 2006). EPR is a technique based on observed data but the mathematical structure it returns is symbolic and usually very parsimonious (Laucelli at al., 2005).
In order to analyze the evolution in time of the blockage rates of the wastewater network of Oslo and identify critical cohorts of pipe regarding blockages, it was decided to perform also a statistical analysis. The blockages rates have been computed in units of number of blockages per km and per year [bl/km.y] for different classes of pipes.
PRESENTATION OF DATA AVAILABLE
Data related to blockages on the wastewater network of Oslo were collected trying to link data on the past events with the characteristics of the pipeline. The data source is Gemini VA, as for almost every municipality in Norway. Gemini VA is used for keeping control on the water and wastewater assets. In Gemini it is possible to extract data on blockages for the Oslo’s network from 1984, but it is only with 1991 that the recording becomes a practice done with a higher level of accuracy.
With the aim of linking the blockage’s events to pipes attributes, it was not possible to use all the available data, because of missing information about the pipe affected by blockage: i.e. lack of information about pipe’s material or installation year forced to link the frequency of blockages to the age of the asset or the material, or the age and material together. Thus the amount of data to be analysed was reduced.
The complete dataset on blockages that it is possible to extract from Gemini covers the period of analysis 1991-2006. A total number of 2232 blockages have occurred during the 16 years of recordings. The distribution of blockage events over the year does not give any clear statistical indications of possible causes due to seasonal factors, but there is a tendency towards a higher blockage frequency during the months of March, April and May. The reason for this may be the cleaning of sand from icy roads during springtime, in combination with poor attendance of the sand traps. This can in turn lead to sand traps filling up and sand flushing into the sewer and storm water network, clogging the pipes. In the same way, a slightly lower blockage frequency during the coldest winter months can lead to the assumption that there is a link between a low dry and wet weather flow (precipitation being stored as snow on the surface in combination with ground frost) and blockages. This assumption seems reasonable, but it would be natural to expect a clearer correlation between these factors than can be found from the tables.
Regarding the location of blockages, these are also distributed well all over the city. It is, however, possible to locate clusters with a higher frequency, and possible causes. Figure 1 shows a cluster of blockages in the city centre. This is the area containing the oldest parts of the network, and the area with the poorest incline. Pipes holding poor quality may be a cause for blockages in themselves. Combined with poor sloping conditions, sediments will be able to clog the pipes over time. It is fair to state that old age and poor sloping conditions are two obvious causes for many of the blockages found in Oslo.
As previously introduced, it was not possible to use all the blockage events registered, because the variables required by statistical analysis are missing for several pipes. In the following paragraphs, first the statistics on blockages for the cohorts are described and discussed; then the results obtained by running EPR are illustrated and discussed.
Figure 1: Location of recorded blockages in Oslo (1991-2006), Oslo VAV
BLOCKAGES VERSUS PIPE’S COHORTS
By dividing pipes into cohorts, each of with the same explanatory variables, the number of recorded blockages could be fitted to an equation. This would allow the expected failure rate to be extrapolated over time. This is a direct measure of performance and can be used in conjunction with costs of sewer maintenance related to blockages. The success of this approach depends on the amount of data available.
On the basis of the technology (and material of construction) adopted to construct and install, the pipelines could be categorized into seven distinct groups (Table 1). On the basis of size, they could be classified as Small, Medium and Large. On the basis of the function they serve, they could be designated as Storm water, Sewage or Combined Flow. That gives us 63 (7 * 3 * 3) distinct classes. One could thus adopt a functional, dimensional or material wise categorization to obtain an insight into the factors influencing blockages.
Four different classes have been created for concrete material and two for plastic pipes as a function of installation year, in order to take into account the different construction practices over time and standards for pipes production and material tests. Different installation periods show different failure characteristics. These characteristics are more dependent on the construction practice for each era than on time since installation (age).
Before 1948 there was not much use of digging trenches with mechanical equipment such as excavator. Usually ditches were dug by hands and the result was narrow trenches. After 1948, machines and excavators were introduced for digging trenches that became wider; the weight on the pipe is more than the pillar of soil above the pipe. To reduce this weight, it is possible to compress the masses around and above the pipe, but until 1960, this was not done. Only around 1964 the importance of control of execution of the trenches was felt: engineers gained understanding for the demand of strength for pipes, the use of rubber gasket and the compression of masses to reduce the weight on pipes laid in wide ditches (Sægrov, 1992). Jointing techniques have improved over the years, allowing greater deflections at joints. In addition, during the post-war housing boom of the 1950s and 1960s, quantity was more important than quality of the construction. Poor bedding conditions and quality of workmanship are reported for this era (Mosevoll, 1994; Sundahl, 1997).
The important years of change in terms of standard for pipes in concrete material are 1948, 1966 and 1970. Between 1964 and 1980, a greater understanding of the importance of backfill material, and of the relationship between weight on pipes and pipe’s thickness, was achieved. Plastic pipes for sewer and pressurized systems had already taken over a large part of the market at the beginning of 1960. However, very little was known about performance of plastic pipes; internationally it was agreed to assume a conventional lifetime of 50 years, through ISO certification. A security factor was defined for pipes. Improvement of the material quality for PVC-U, PE and PP has led in several European Countries to the definition of higher maximum allowed stress and lower safety factors. As a consequence, in Europe the safety factor for PVC was reduced from 2.5 to 2.0; that for PE from 1.6 to 1.25. However, many Norwegian municipalities have kept the original safety factors. In the seventies, several methods for testing plastic material were introduced, which resulted in a better quality of the plastic pipes produced after 1980.
The “history” of material and construction practice evolution in Norway guided the decision to group the pipes in the main classes listed in Table 1. For concrete pipes, it was agreed to include in the first group the pipes laid before 1948; the second group includes the pipes laid between 1948 and 1964; the difference is that in the latter period, engineers were learning from experience that something should change in the way of digging trenches and creating bedding conditions. Since the best performance can be found in pipes produced after 1980, the fourth group of pipes includes those laid after that year, while the third group includes those made during the evolution of the standards, that is between 1964 and 1970.
For plastic pipes, which started to be largely applied in Norway during the seventies, the border between pipes with different performances is identified in year 1977.
For each cohort, descriptive statistics on blockages distribution over the period of time 1991-2006 were obtained.
Figure 2 depicts the distribution of the 625 blockages among the seven classes of pipelines. The oldest of the lot, C1 pipes, by virtue of their age, account for nearly 32 per cent of the total considered for the study. It should be noted however that only 30 per cent of the 1548 distinct pipelines are being considered for the study. So, the predominance of C1 is only for this subset. One may however wish to extend this distribution to the entire set of 1548 pipes, but that may introduce uncertainties. Unfortunately, we have to proceed with the analysis within these constraints. Classes C2 and C3 account for 44 per cent between themselves, with P2 coming in fourth with over 18 per cent.
C4 14; 2% F 17; 3% P1 1; 0.16% P2 116;19% C1 197; 31% C2 185; 30% C3 95;15%
Figure 2 Oslo wastewater pipelines experiencing blockages in 1991-2006 categorised on a
material and technology basis. (category, number, Percentage of the total of 625 considered for this study)
While Figure 2 depicts the total number of pipes blocked in the said period for each of the seven classes, what one would need for analysis to a greater depth of detail is the blockage rate or in other words, the blockages in a particular class of pipelines per unit length of pipelines of that class in the stock in the year in which the rate is measured. Here, it must be remarked that the total length of pipelines in a given class may not be the same during the 16-year period under consideration. Pipelines could be added to extend the network, as in the case of P2 pipelines, or pipelines belonging to any class may be rehabilitated using Cured-in-Place-Pipe (CIPP) technology. After rehabilitation using CIPP technology, the rehabilitated pipeline no longer belongs to the class to which its earlier avtar belonged. Hence, it ceases to be bracketed under any of the seven classes referred to. This means that the denominator term in the ’blockage rate ratio’ may not be a constant.
It is important to clarify that it was not possible to compute the length of each sub-class in every year of analysis, but we could only refer to the length of the sub-classes at the end of year 2006.
However it was possible to estimate the lengths for every year for the seven main classes by analyzing the databases about rehabilitation and pipes added every year. The data available in those database did not always allow to address the rehabilitation works to a specific subclass, but only to the main class. So a more precise computation of blockage rates was done at level of main classes..
Table 2 lists the total number of blockages and blockage rates computed every year for the main class and for the whole network using the variation of length over time.
However, from the analysis made to analyse the effect of the length assumed for computation, it can be concluded that, for the Oslo case study, using the 2006 value for lengths of pipelines in different classes as constants for all the 16 years, does not affect the blockage rate much. The highest difference on the average blockage rate is provided by results for class C1 and class C2 when we move from the analysis at subclass level to main class level. The difference is the consequence of the different lengths of the class available for analysis.
Table 2 Number of blockages and blockage rates over time for the 7 main classes with estimated length over time
Figure 3 plots the blockage rates of the seven classes. The lengths of ferrous (F) and the newest PVC pipes (P2) have increased over the years owing to additions to expand the network. P1 remains constant during the 16 year period, while the concrete pipes of all four classes have registered drops in their total lengths, owing to conversion of some of them by rehabilitation to CIPP pipelines. The average blockage rate is the highest for the P2 pipes (0.039 per kilometre per year), the small size perhaps justifying itself as a determining factor in this respect. P1 pipes have performed admirably well (a good number of these are intermediate in dimension). In terms of total number of blockages, the C1 pipes come first on the list (Figure 2), but because of the fact that they account for the largest share of the network (with C2 and C3 following not very far behind), the average blockage rate tends to be a bit lower than that for C2 pipelines (0.022 for C1 and 0.024 for C2). Ferrous pipes are usually the pressurized pipes connected to pumping stations and that can also explain the very low amount of blockages events.
Pipes in class C4 includes concrete pipes laid in the last 27 years, that means better material and better construction practice in addition to short age as parameters influencing the low degree of blockages.
566,23 585,86 0,01 0,048 555 560 565 570 575 580 585 590 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 Calendar year
Total length of C1 pipelines in kilometres
0 0,01 0,02 0,03 0,04 0,05 0,06
Blockage rate measured as
Blockages per kilometre of C1 pipelines per year
Length in kms
Blockage rate in blockages/km/year
465 470 475 480 485 490 495 500 505 510 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 Calendar year
Length of C2 pipelines in kilometres
0 0,01 0,02 0,03 0,04 0,05 0,06
Blockage rate expressed as
Blockages per kilometre of C2 pipelines in the network per
year
Length in kms Blockage rate in blockages/km/year
464 466 468 470 472 474 476 478 480 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 Calendar year
Length of C3 pipelines in kilometres
0 0,005 0,01 0,015 0,02 0,025 0,03
Blockage rate expressed as
Blockages per kilometres of C3 pipelines in
network/year
Length in kms Block age rate in blockages/km/year 72,6 72,7 72,8 72,9 73 73,1 73,2 73,3 73,4 73,5 73,6 73,7 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 Calendar year
Length of F pipelines in kilometres
0 0,01 0,02 0,03 0,04 0,05 0,06
Blockage rate expressed as
Blockages per kilometre of F pipelines in the
network per year
Length in kms Bloc kage rate in blockages/km/year 0 10 20 30 40 50 60 1991 1992 1993 1994 19951996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 Calendar year
Length of PI pipes in kilometres
0 0,005 0,01 0,015 0,02 0,025 0,03 0,035 0,04 0,045 0,05
Blockage rate expressed as
Blockages per kilometre of PI pipeline in
network per year
Length in kms Blockage rate in blockages/km/year
0 50 100 150 200 250 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 Calendar year
Length of P2 pipelines in kilometres
0 0,01 0,02 0,03 0,04 0,05 0,06 0,07 0,08 0,09 0,1
Blockage rate expressed as
Blockages per kilometre of P2 pipelines in the
network, per year
Length in kms
Block age rate in blockages/km/ year
Figure 3Blockage rates of Oslo wastewater pipelines experiencing in 1991-2006 (Clockwise
from top left: C1, C2, C4, P1, P2, F, C3)
FUNCTIONWISE AND DIMENSIOWISE COMPARISON OF
BLOCKAGE RATES
The blockages could also be divided amongst sewage, stormwater and combined flow pipelines. Table 3 depicts this distribution. Sewage carriers account for a lion’s share of the blocked pipes considered for the study. Stormwater carriers quite expectedly account for a meagre 0 per cent, with combined flow pipes taking up 33 per cent. Likewise, if the categorisation is done on a dimensional basis, it is at once clear that small diameter pipes are (have been) more susceptible to blockages, commanding an 81 per cent share of the total. Considering that the proportions of sewage, stormwater and combined flow pipelines in the network have been fairly equal from 1991 onwards, follows that the average blockage rate of sewage pipes is nearly 2.5 times that of the combined flow pipes. The total length of small
190 192 194 196 198 200 202 204 206 208 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 Ca len da r yea r
Length of C4 pipeline network in kilometres
0 0,01 0,02 0,03 0,04 0,05 0,06
Blockage rate expressed as
Blockages per kilometre of C4 pipelines
per year
Length in km s B locka ge rate in block ages /km /yea r
the large-diameter pipes from 1991 to 2006. Referring to Table 4, it is clear that the average blockage rate of the small pipelines is about thrice as that of the medium-size ones.
Sewer Stormwater Combined
year blockages λ [block/km.y] blockages λ [block/km.y] blockages λ [block/km.y]
1991 26 0.043 0 0.000 22 0.032 1992 44 0.073 0 0.000 26 0.038 1993 24 0.040 0 0.000 13 0.019 1994 19 0.032 0 0.000 14 0.020 1995 26 0.043 1 0.002 13 0.019 1996 29 0.048 1 0.002 14 0.020 1997 33 0.055 0 0.000 13 0.019 1998 20 0.033 0 0.000 10 0.015 1999 26 0.043 0 0.000 18 0.026 2000 40 0.067 0 0.000 14 0.020 2001 16 0.027 1 0.002 6 0.009 2002 32 0.053 0 0.000 13 0.019 2003 23 0.038 0 0.000 6 0.009 2004 11 0.018 0 0.000 11 0.016 2005 27 0.045 0 0.000 6 0.009 2006 20 0.033 0 0.000 7 0.010 total 416 0.043 3 0.000 206 0.019
Table 3Blockage rates over time for pipe’s type
S M L
year blockages λ [block/km] blockages λ [block/km] blockages λ [block/km]
1991 38 0.0377 10 0.0099 0 0.0000 1992 64 0.0636 6 0.0060 0 0.0000 1993 33 0.0328 3 0.0030 1 0.0010 1994 27 0.0268 6 0.0060 0 0.0000 1995 30 0.0298 10 0.0099 0 0.0000 1996 35 0.0348 9 0.0089 0 0.0000 1997 40 0.0397 6 0.0060 0 0.0000 1998 24 0.0238 6 0.0060 0 0.0000 1999 33 0.0328 11 0.0109 0 0.0000 2000 47 0.0467 7 0.0070 0 0.0000 2001 19 0.0189 4 0.0040 0 0.0000 2002 37 0.0367 8 0.0079 0 0.0000 2003 21 0.0209 8 0.0079 0 0.0000 2004 19 0.0189 3 0.0030 0 0.0000 2005 26 0.0258 7 0.0070 0 0.0000 2006 21 0.0209 6 0.0060 0 0.0000 total 514 0.0319 110 0.0068 1 0.0001
Table 4Blockages rates over time size-wise: ‘S’ = diameter lower than 249mm,
‘M’=diameter between 250 and 499mm, ‘L’=diameter greater than 500mm.
BLOCKAGES ANALYSIS WITH THE EVOLUTIONARY
POLYNOMIAL REGRESSION (EPR)
EPR is used in this study to correlate sewer blockages to possibly influencing pipe’s attributes. Collaboration between NTNU and the University of Exeter was established, in
order to test EPR using the data base of Oslo, applying the model at pipe level. The analysis was conducted using the available database, including the data from 1984. The data processing started with a database of 53198 pipes, that was reduced to 6979 after having deleted from the register the pipes missing the required variables to run EPR.
The first step was to choose two variables as basis for the grouping. Dimension, slope and age were selected. Two different combinations were chosen: dimension-age and dimension-slope. Before applying EPR, data have been grouped using age (Ae), diameter (De) and slope (Se) classes. The following attributes were computed according to class: (1) the sum of pipe lengths (Len); (2) the sum of Blockage events (Block); (3) the number of pipes (Np). Ae is the equivalent age for the pipes in a particular class. Equivalent age is calculated as
p p class t L A Ae L ⋅ =
∑
where Ap and Lp are age and length at pipe level. In this way the longer pipes have a bigger influence on the mean age, which gives a more realistic picture because the longer pipes constitute an equivalent bigger share of the total sewer system. The same calculation is performed for dimension and slope to compute the variables equivalent dimension (De) and equivalent slope (Se).
The EPR task was to discover a symbolic relationship among five inputs (Ae, Len, Np, De, Se) and one output (Block), that describes the total number of blockages in the last 22 years. The selection between the different alternative relationships produced by the model was made by looking at the value of the coefficient of determination (CoD), by choosing the model having a parsimonious structure. If EPR returns polynomials with more than one term, the formulae having common terms are preferred.
The simulations have been performed with dimension and age first, then with dimension and slope as basis variables, on seven selected class of pipes for the input files: all pipe, concrete pipe, plastic pipes, combined/concrete pipes, combined/plastic pipes, Sewer / Concrete pipes and Sewer / Plastic for a total of 14 simulations. Table 3 shows the two models selected for each simulation using the different basis variables. Inside each group of pipes the more parsimonious model is preferred as blockage-prediction model.
Basis Group CoD Model
Dim-Age All 0.8909 2 192.6113 Ae Len Block De Se ⋅ = ⋅ ⋅ Dim-Slope All 0.9687 2 3 43428303376.7506 Np 340894.0839 Ae Len Block De Len Se De Np ⋅ = ⋅ + ⋅ ⋅ ⋅ ⋅ Dim-Age Concrete 0.9802 3 2383314.4011 Ae Np Block De Se ⋅ = ⋅ ⋅ Dim-Slope Concrete 0.9631 3 3 2 1280932.8509 Np 0.022086 Ae Len Block De De Se ⋅ = ⋅ + ⋅ ⋅ Dim-Age Plastic 0.9351 3 5 2 1.2059 10 Ae Len Block De Np − ⋅ = ⋅ ⋅ ⋅ Dim-Slope Plastic 0.9241 2 2 33.1996 Ae Np Block De Len Se ⋅ = ⋅ ⋅ ⋅
Dim-Age Combined /Concrete 0.9993 3 3 3 3 2 592320.5848 Ae Np 17649.4864 Ae Np Block De Se De Len ⋅ ⋅ = ⋅ + ⋅ ⋅ ⋅ Dim-Slope Combined /Concrete 0.9846 2 3 3 2 1076260775.8467 Np 0.028043 Ae Np Block De Len Se De ⋅ = ⋅ + ⋅ ⋅ ⋅ Dim-Age Combined/ Plastic 0.4278 2 3 3 2 2 3644535336.1823 Ae Np Block De Len Se ⋅ = ⋅ ⋅ ⋅ Dim-Slope Combined/ Plastic 0.6943 2 3 3 11073.0854 Ae Np Block Len Se ⋅ = ⋅ ⋅ Dim-Age sewer/ Concrete 0.9978 3 3 2 3 3 461571.2906 Ae Np 781213.8576 Ae Np Block De Len Se De Len Se ⋅ ⋅ = ⋅ + ⋅ ⋅ ⋅ ⋅ ⋅ Dim-Slope sewer/ Concrete 0.8951 3 2 3 17.8768 Ae Np Block De Se ⋅ = ⋅ ⋅ Dim-Age sewer/ Plastic 0.9186 3 2 3 2 0.00028978 Ae Len Se Block De Np ⋅ ⋅ = ⋅ ⋅ Dim-Slope sewer/ Plastic 0.9120 3 2 0.20821 Ae Len Block De Se ⋅ = ⋅ ⋅
Table 5 Optimal models selected for the different input pipes with different variables as basis
To summarize the results, the general schematization of the blockage prediction model is:
( )
Block class = ⋅k Aeα⋅Npβ ⋅De Se Lenγ ⋅ δ ⋅ ε
Where the total number of blockage (Block) predicted can be expressed as a function of the relevant parameters Ae, Np, De, Se and Len and parameter k(class), being α, β, γ, δ and ε the exponents that are assumed to be integer and within the range [-3: 3], including zero.
Without considering the example run for the combined plastic pipes that gave low values of CoD, certainly due to the small size of the stock in the network, the results obtained highlight a direct dependency of the expected blockage number from the variable average age, and an inverse relationship with the average diameter and slope, both with changing values of the exponents. The relation with the length of the class and the size of the class stock is unclear. Often the prediction model identifies an inverse relationship with the variable length that is the opposite of what we would expect. Looking at the class of plastic pipes, it is difficult to choose within the two models having similar value of the CoD, but having an opposite relationship between blockages and the variables Len and Np.
With the aim of using the results to predict expected blockages for each pipe, the EPR derived expression were subsequently used for the prediction of cumulative number of failures (blockages) over 22 years time interval. It is assumed that the average failure rate (number of failures per year) can be predicted by simply dividing the aforementioned expression by the number of years analysed. The same procedure applies to determine the number of failures per year per km of mains, for example. This approach allows assessment of the probability of a blockage for each pipe falling into the i-th pipe class as follows:
( ) ( i)
i
Lp Block pipe Block class
Lt
The extension of time period of data available, 22 years, and the reduced sample of data left after filtering the database, 621 events, clearly influences the prediction of blockages at pipe level by using EPR.
CONCLUSIONS
Blockages data have been extracted from GEMINI and examined using:
• Statistical analysis for the period 1991-2006
• EPR, with data referred to the period 1984-2006
The aim of the analysis performed was to identify critical components of the network and if any relation between the concurrency of blockages and pipe’s attributes exists.
The level of detail required by the analysis, brought to reduce the available data from about 2000 events to 600, the events connected with pipes missing information as material, age, diameter and slope were deleted.
From the statistical analysis, it can be stated that age, size and function do seem to have a marked influence on the proneness of a pipeline to blockages, but, for the reduce sample available it is difficult to say which variable it is more influencing. If we look at total number of blockages the class C1 seems to be the most prone to blockages, but looking at blockage rates, then it is the youngest class, class P2 showing the highest blockage rate.
EPR allowed to identify the relation between attitude to block and pipe’s attributes in order to understand what affect the possibility to have a blockage in the pipe. EPR provides formulae to compute the accumulated number of blockages for a pipe class at the end of a given period of time.
The analysis with EPR highlighted a clear direct dependency of the accumulated expectable number of blockages (EPR) with class age.
The analysis showed that sewer pipes have a much higher frequency and attitude to block than combined pipes, while storm water pipes do not block easily.
The statistical analysis elected the pipes with diameter lower than 250mm, and EPR confirms an inverse relation between blockage and pipe’s diameter.
ACKNOWLEDGEMENT
This work used software developed by Orazio Giustolisi (Technical University of Bari) & Dragan Savic (University of Exeter).
REFERENCES
Berardi, L., Savic, D.A., Giustolisi, O., (2005), “Investigation of burst-prediction formulas for water distribution systems by
evolutionary computing”, in Computer and Control in Water Industry (CCWI), Vo.2, pp. 275-280, ISBN 09 5391 403 8.
Giustolisi, O., and Savic D., A symbolic Data-driven Technique Based on Evolutionary Polynomial Regression”, J. of Hydroinformatics, 2005.
Giustolisi, O., Kapelan, Z., Savic, D.A., (2006), “Multi-objective Evolutionary Polynomial Regression”, 7th International
Conference on Hydroinformatics, HIC 2006, Nice, France, P. Gourbesville, J. Cunge, V. Guinot & S.Y. Liong (Eds), Research Publishing, Vol. 1, pp 725-732, ISBN 81-903170-1-6.
Laucelli, D., Berardi, L., Dogliosi, A., 2005, “EPR Toolbox, Version 1.SA”. Technical University of Bari, Dept. of Civil and
Environmental Engineering.
Mosevoll, G. (1994). Vedlikehold og fornyelse av VA- ledninger: Modeller for tilstands-prognose / Funksjonskrav til informasjonsystemer. Dr.ing avhandling, Institutt for Vassbygging, Norges Tekniske Høgskole, Universitetet i Trondheim.
Sægrov S (1992) Condition analysis concrete pipes for wastewater; PhD thesis at Norwegian University of Science and technology, 1992.
