Understanding Blockage Formation in Sewer Systems By-Case Approach.

Transcription

1 Understanding Blockage Formation in Sewer Systems By-Case Approach. A Case- Dr S. Arthur*, Dr H. Crow, & Mr L. Pedezert** * Sustainable Water Management Research Group, School of the Built Environment, Heriot-Watt University, Edinburgh EH14 4AS, Scotland UK. ** Nopac-Ingénieur Travaux Génie Civil, Rue John Hadley - BP 29, Villeneuve d'ascq, France Abstract This paper documents a key part of the development of a risk analysis process which may be used to manage maintenance expenditure within the context of serviceability. Against the background of poor availability of reliable data, the text outlines the basis of a methodology which can be used to model blockage likelihood with limited information and without the need for additional data gathering. Specifically, the work reported herein relates primarily to the research undertaken to better understand what network and catchment specific factors influence the formation of blockages in combined and foul drainage systems. The role of each of these factors is considered in turn and the significance of their impact is estimated. Keywords Sewerage Asset Management, Proactive Maintenance, Risk Analysis, Blockages. INTRODUCTION Internationally, there has been a concerted move to use serviceability as a key performance indicator in sewerage provision (e.g. Shepherd et al, 2005). Serviceability covers not only general service provision, but also socio-economic factors and environmental impacts such as: Sewer blockages which lead to flooding Plant failures which lead to flooding Properties affected by restricted discharge to sewer Traffic disruption Health and safety issues CSO spills In the UK, sewerage provision regulators (OFWAT and WICS) categorise all sewerage derived flooding as being one of two types: those due to hydraulic overloading and those due to all other causes. Although the media tends to focus on hydraulic overloading problems, 84% of sewerage derived flooding incidents (>26,000 per year) in England and Wales fall into the latter of these categories and >90% of these are due to blockages (Table 1). Table 1: Annual blockage rates in England and Wales based on data submitted to OFWAT in June Note: It is not meaningful to make direct comparisons between different Water Services Providers performances due to the unique challenges they face. Water Services Provider Domestic Properties Connected to the Sewer Total Length of Sewer (km) Total Flooding Incidents Total Flooding Incidents due to Overloading Total Flooding Incidents due to Blockages Blockages which Result in Internal Flooding Blockages per Million Households Blockages per km Anglian Water % 81.7% 13.8% Northumbrian % 57.3% 18.4% 2, Severn Trent % 81.5% 14.3% South West % 75.4% 9.2% 1, Southern Water % 84.5% 5.4% 2, Thames Water % 84.1% 13.7% United Utilities % 85.8% 12.7% 1, Welsh Water % 45.7% 16.3% Wessex Water % 84.6% 4.0% 1, Yorkshire Water % 63.2% 20.3% Total Wt Average % 78%

2 This text reports a small part of a larger project which aims to create a technique to enable asset managers to determine where proactive maintenance of sub-critical sewers is most viable, from a serviceability perspective, and therefore enable them to proactively manage capital maintenance expenditure (e.g. Arthur & Crow, 2007). The focus is to produce a technique which can be instigated with little or no additional data collection beyond that which is already available to the majority of water companies. The process is based on risk analysis as a function of consequence and likelihood of failure occurring Failure Modes, Effects and Criticality Analysis (FMECA) (e.g. Moss & Woodhouse, 1998). Values for both consequence and likelihood are developed by considering a number of pipe flow and catchment characteristics which are readily available from a combination of local government and wastewater company reports. These values are then used to produce a rank of pipes which are at greatest risk from flooding. To ensure that the application of the methodology is as flexible as possible. It can be divided into four key modules, namely, 1. Screening to improve computational efficiency, this phase aims to remove any low risk elements from the analysis. A key research outcome in this module was successfully profiling problems pipes in a real network in a manner which transferable and scaleable. 2. Consequence scoring this module focuses on assessing the consequence of a system failure. A substantial research challenge to be overcome in this phase was quantifying potential failure consequences, and then assigning validated relative weights to them. 3. Likelihood scoring this key module focuses on assessing the perceived likelihood of a failure; in this case blockages. 4. Score combination and ranking. Specifically, this text reports research which was undertaken to investigate which factors lead to the creation of blockages so that their likelihood of occurrence may be better estimated within the wider methodology (Module 3). Considering the role of blockages is key, as across the world they are probably the number one cause of loss in sewer serviceability in either dry or wet weather flow conditions (Ashley et al, 2004). When this happens, flooding may occur. As illustrated in Table 1, data collected in 2005/6 indicates that there were ~25,000 sewer blockage related flooding incidents reported in that year; 11.9% of which resulted the internal flooding of a property. At best this can be inconvenient. At worst, blockages can lead to damage to property, stress, illness or possibly even death in rare occasions. Recently reported research in this area has advocated the use of powerful data analysis tools to identify which pipes are most likely to fail (Baur et al, 2003, Savic et al, 2006 & Fenner et al, 2006). Savic et al, 2006 successfully used evolutionary computing to interrogate and analyse a blockage dataset with 10,300 elements to generate select key parameter groups which are used to construct catchment specific empirical equations to predict assets which are liable to fail. Other ongoing work in this area has produced good results by statistically profiling large, asset databases and linked failure data (Fenner et al, 2006) To augment the important results produced by these studies, the work reported herein takes a somewhat different approach by focusing on a relatively small catchment and considering each failure individually. STUDY CATCHMENT The catchment selected serves a small town of approximately 10,000 inhabitants located in the Southeast of England. That catchment was selected as it is served by a typical UK system and because a verified hydraulic model was available which included Section 24 pipes. A previously verified Infowork CS model was provided by Southern Water to support this study. The sewer system is hybrid (combined and separate), with the separate system predominately serving more recent developments. Unless otherwise stated, this text will focus on the combined and foul network which comprises 733 modelled pipe lengths with a total length of 30.8km (See

3 Table 2). More than 86% of the pipes (96% of the foul pipes and 71% of the combined pipes) have a diameter ranging from 150mm to 300mm. These pipes carry the wastewater to the wastewater treatment plant mostly by gravity. Table 2: Pipe Repartition by Sewer Type Sewer type Number of pipes Total length (m) combined 438 (29.9%) 18,955 (30.8%) foul 295 (20.1%) 11,844 (19.1%) total 733 (50.0%) 30,799 (49.7%) Complaints Data To understand where problems may exist within the catchment, a database of customer complaints obtained from the water services provider was interrogated. The customer complaint data spanned the period from July 2001 to April 2004 and comprised 95 complaints. Connected with each customer complaint was a diagnosis and a note of any action taken (e.g. blockage cleared by jetting etc). This data also included a tag which described the type of pipe encountered (e.g. main sewer, interceptor, Section 24 etc.). Analysis of the data indicated that 51 of these complaints were a result of individual blockage incidents in 37 pipes. Research undertaken by the US Environmental Protection Agency in 1999 (EPA, 1999) reported that the average expected number of complaints per mile of pipework per year as being 4. In the catchment studied in this text, there was found to be an average of 1.85 complaints per mile. Whilst there may be cultural or management difference between the two populations, these performance data were found to be statistically similar (at the 5% level) and were used as a basis to conclude that the catchment was not atypical. The conclusion was supported by data presented within the UK SRM (WRc, 2000); it estimates the UK average blockage rate as being 0.1 to 2.0 per km per year. This compares favourably with the study catchment s blockage rate of 0.62 per km per year. Other UK studies have noted blockage rates of 0.3 per km per year and per km per year; both of which correspond well with the UK SRM range. UNDERSTANDING WHAT CAUSES BLOCKAGES Key to research reported herein was the philosophy that with a relatively small catchment it should be possible to assess each blockage incident in isolation and reach a conclusion about its cause. Based on this, each blockage was assessed in turn together with related network characteristics (physical and hydraulic) using MapInfo and Infoworks CS. This analysis resulted in several features being highlighted as being possible factors which may indicate that an individual pipe has an increased propensity to block. These factors included: 1. Network type foul or combined. 2. Surcharge state whether or not the pipe surcharges during a 1-Year design storm. Surcharging is where flows in the sewer become pressurised (i.e. flow depth is greater than the pipe diameter). In most cases this will mean that flow will begin to rise up any manhole shafts. 3. Lack of capacity whether or not flooding occurred at the upstream end of the pipe as a result of the 5-Year design storm. 4. Downstream control whether or not the flow is less that expected for a given depth. 5. Proximity to flow confluence it was noted that many blockages occurred just upstream of a junction. 6. Whether or not flow velocities fail to meet the self cleansing criteria.

4 7. Self-cleansing gradient - whether or not the pipe has been laid at a less than 1:Pipe Diameter.(mm). 8. Large direct inputs it was noted that pipes which had blocked had nodes which received disproportionately large direct inputs (in terms of contributing population and/or area). 9. Pipe size the majority of the pipes which blocked were 225mm in diameter or less. To validate each of these hypotheses, the entire database was checked for blockages which could be considered to have similar causes. The remainder of this section reports this process. Network Type foul or combined. Most large UK sewer networks comprise a mixture of separate and combined elements. Combined sewers experience at the macro temporal level extreme variations in flow, whilst foul sewers experience fluctuations in inflow at a much smaller scale. The design of both of these types of system takes advantage of peaks in flow patterns to ensure there is no long-term build up material which could lead to a blockage occurring. Combined and separate systems also convey very different types of solids inputs. In particular, surface inputs to combined and surface drains can include large amounts of decaying flora, grit/sand/silt, litter and construction waste. These very different ambient hydraulic conditions and inputs mean that combined and foul systems may have very different blockage patterns. Within the study catchment it was noted that 295 of the 733 pipes were classified as being part of a foul system and that the foul and combined systems had suffered 11 and 40 blockages respectively. These data indicate that the combined pipes in the network were 2.5 times more likely to block. To test the significance of this result, a Chi-Squared ( 2 ) Distribution test (e.g. Clarke & Cook, 2005) was used. Using this approach, a null hypothesis is established; which in this case is that there is no pattern to the data (i.e. the blockages are spread randomly across the catchment). The value for Chi-Squared is then calculated using the failure data and compared to significance tables, which confirm whether or not any deviation from the null hypothesis in the observed data is by chance or can be concluded to be statistically significant. This approach showed that, at the 5% confidence level (i.e. the chance of the result being wrong is less than 1:20), the result was statistically significant. Modelled Surcharge State Surcharging flow conditions within gravity pipelines occur when the flow depth is higher than the crown of the pipe (i.e. flow rises up the manhole shaft). Although sewerage systems in the UK have traditionally been designed to avoid surcharging, it is not necessarily a problem in a well maintained system. Indeed, if there is no downstream flow restriction, surcharging may result in the flow having increased energy available to convey solids. Based on a 1-year return period event, modelling showed that within the study catchment 85 of the 438 pipes in the combined network were classified liable to surcharge. A total of nine of these pipes were pipes which were found to have previously blocked. These data indicate that the pipes which surcharge during rainfall events may be at increased risk of blocking. To test the significance of this result, a Chi-Squared Distribution test was again used. This showed that, at the 5% confidence level the result was not statistically significant. Flooding Nodes Areas with low hydraulic capacity should not necessarily be most susceptible to blockage. However, it was hypothesised that some nodes may have flooded due to a downstream hydraulic feature causing a backwater effect. In an extension to the analysis undertaken to understand if surcharge state can be used as an indicator of increased blockage risk, nodes immediately

5 upstream which were liable to flood were also examined. As there were too few flooding nodes (a single occurrence) at the 1-year return period, a 5-year return period was adopted as an alternative. As the data in Table 3 indicates, the flooded nodes had a markedly increased probability of being linked to pipes which blocked. The Chi-Squared Distribution test confirmed that this result was statistically significant. Table 3: The use of flood risk to indicate an increased blockage likelihood. No Flood Flood Understanding the Role of Any Backwater Effect Many urban drainage systems have been designed using the Modified Rational Method; this is particularly true of systems serving small residential, commercial and industrial developments. Whilst this approach is perfectly valid, it often cannot fully account for the consequences of the dynamic hydraulic processes which can occur within sewerage systems. Key amongst such phenomena is the backwater effect which can lead to reduced flow rates (increased flood risk) and diminished solids transport capacities (increased risk of sedimentation). Where this project was concerned, pipes which were considered to be suffering from a backwater effect were those where the flow rate was less than free surface flow theory predicted. It was observed that 5 of the 40 blocked pipes (13.5%) and 32 of the remaining pipes (8.7%) experienced a backwater effect. To test the significance of this result a Chi-Squared Distribution test was again used. This showed that, at the 5% confidence level the result was not statistically significant. The Role of Flow Confluences A common cause of backwater effects in open channel flow is flow confluences. In sewerage systems these take the form of pipe junctions and lateral inputs from individual properties or road gullies. Considering only those confluences formed where two branches of the network meet, the study catchment data set (Table 4) indicated that pipes upstream of flow confluences in the network showed an increased propensity to block. To test the significance of this result a Chi- Squared Distribution test was again used. This showed that, at the 5% confidence level, the result was not statistically significant. Table 4: The role of a flow confluence immediately downstream in increased blockage likelihood. DS Junction No DS Junction Design for Self-Cleansing At the design stage the risk of blockage is reduced by ensuring bed shear stress is above a predefined level. In practice, this is often undertaken by ensuring that the peak flow velocity exceeds a predefined threshold or minimum gradient. Although this approach is well understood and was recently updated (Ackers et al, 1996 & Arthur et al, 1999) networks do not always perform as expected and adequate storms may not occur at the expected frequency. The initial phase of this research noted that low velocities were associated with a number of blockages. To explore this, the theoretical full-pipe velocity during a 2-Year return period event was compared to a self-cleansing velocity of 1.0m/s. The data from this analysis are presented in Table 5.

6 Table 5: The use of ambient flow velocity to indicate an increased blockage likelihood. Velocity Pass Velocity Fail A similar check was undertaken on pipes with relatively slack gradients. In this case, the minimum gradient criteria specified in EN752-4 (BSI, 1998) (gradient no less then 1:N, where N is the pipe diameter) were used as a benchmark. The results of this analysis are illustrated in Table 6. Table 6: The use of a slack gradient to indicate an increased blockage likelihood. Gradient Pass Gradient Fail As the data presented in Table 5 and 6 indicates, it was observed that areas which failed to meet self cleansing were found to be at increased risk of experiencing a blockage. To test the significance of these results a Chi-Squared Distribution test was again used. This showed that, at the 5% confidence level the results were statistically significant. Population Density Analysis of the data indicated that pipes with disproportionately high direct inputs also demonstrated an increased likelihood of suffering a blockage. It was hypothesized that these pipes were more likely to be blocked by something inappropriate entering the network. Table 7 illustrates that across the catchment pipes which had a contributing population five times the average had increased likelihood of blocking. To test the significance of this result a Chi-Squared Distribution test was used - at the 5% confidence level the results were statistically significant. Table 7: The use of high direct flow inputs to indicate an increased blockage likelihood. Normal Population High Population The Role of Pipe Size At the upstream end of systems pipework tends to be of much smaller diameter and the flows can be much more intermittent; often, other than any infiltration, there may be no flow for a number of hours. This means that solids transport may be a somewhat discontinuous process (Ashley et al, 2004). This combined with the fact that solids are relatively large compared to the pipe diameter means that blockages in smaller pipes are commonplace. However, as these pipes service smaller catchments the impact of any failure often not significant when considered in isolation. In larger sewers the problem is somewhat different. Extensive research in the UK and elsewhere (Ashley et al, 2004) has shown that larger sewers tend to suffer from the long term accumulation of inorganic solids derived from the catchment surface. These deposits tend to result in a longterm degradation in capacity and/or an increased likelihood of a blockage. Although the reasons for the deposits occurring are often catchment specific, a major contributing factor is the practice of using a single self-cleansing velocity to design all sizes of pipes. The use of self-cleansing

7 criteria for the design of sewers does not take into account the pipe size and the concentration of sediment. The critical velocity approach was shown by Yao (1974) to under-design larger diameter sewers and over-design those with smaller diameters, with respect to sedimentation. This was confirmed by others in field and laboratory studies (Arthur et al, 1999) and as a result the design of UK sewers to manage sedimentation was updated (Ackers et al, 1996). As illustrated in Table 8, the pipes in the network range in size from 150mm to 900mm in diameter. This data indicated that the smaller pipes in the network were much more likely to block (at the 5% confidence level) and result in a customer complaint. This finding is collaborated by respected laboratory based research focusing on (artificial) solids transport in straight small diameter pipes of constant size and gradient (Memon et al, 2007 & McDougall and Swaffield, 2003). This work concluded that although solids were transported further in smaller diameter pipes, these pipes were also more liable to block. Table 8: The use of pipe size to indicate an increased blockage likelihood. Pipe Diameter (mm) Total Number of pipes. Number of pipes which have blocked Proportion which have blocked (%) DISCUSSION Analysis of a blockage dataset on an event by event basis highlighted a number of factors which may contribute to the likelihood of a blockage occurring. Testing each hypothesis by searching for similar incidents across the catchment highlighted other similar cases. However, testing the statistical significance of the results indicated that in some cases the increased blockage likelihood was not sufficient for firm conclusions to be drawn. These results are summarised in Table 9. Table 9: Factors which may indicate increased blockage likelihood. Factor Increased Likelihood Combined pipes 146% Yes Pipes which surcharge during a 5-Year event 44% No Pipes which flood during a 5-Year return event 839% Yes Where flows are influenced by a backwater effect. 55% No Proximity to flow confluence 38% No Pipes which fail to achieve a self-cleansing velocity 80% Yes Pipes laid at a gradient less than 1:Pipe Diameter. 102% Yes Pipes with large direct inputs 167% Yes Pipe size the majority of the pipes which blocked were 225mm in diameter or less Statistically Significant? 592% Yes CONCLUSION The research reported herein has demonstrated an alternative approach to identifying where blockages may occur in sewer networks. The approach has highlighted key readily available catchment, hydraulic and network parameters which appear to influence the likelihood of a

8 blockage occurring. These results should help sewerage asset managers to move beyond the traditional practice of targeting blockage hotspots and focus their proactive maintenance activities (e.g. jetting) on identifying where additional future blockages may occur. The approach may also be used to identify where the blockage risk is high in networks with no recorded failure history (e.g. new networks). Although considering each failure on a case-by-case basis was key to the philosophy adopted to meet the aims of the reported project, this has affected the results in two key ways. Firstly, the resources available meant only a small catchment could be considered in detail. This meant that the significance of some results, principally those related to proximity to confluences and the influence of any backwater effect could not be fully tested. Secondly, the limited size of the data set has also made it difficult to consider analysing in detail how these influencing factors interact to create the conditions which allow blockages to form. Research is therefore required to evaluate any interaction between the factors which influence the likelihood of a blockage occurring. Furthermore, like other methods which use blockage databases and network characteristics, the approach cannot account for factors which are known to contribute significantly to blockage statistics (such as poor jointing the systematic dumping of fats by restaurants). The authors are currently seeking suitable catchments to test and further develop the reported methodology by meeting these research needs. ACKNOWLEDGMENTS The authors are grateful for the support given to the project by Southern Water, particularly John Challoner. The research was funded by EPSRC. REFERENCES Ackers, J.C., Butler, D and May, R.W.P., (1996), Design of sewers to control sediment problems (R141), CIRIA, London, ISBN, Arthur, S. & Crow, H., (2007), Prioritising sewerage maintenance using serviceability criteria, Water Management, The Proceedings of the Institution of Civil Engineers, Vol 160, No 3, September 2007, Pages Arthur, S., Ashley, R.M., Tait, S. & Nalluri, C. (1999), Sediment transport in sewers - towards a design approach for real and existing sewers, Water Maatine & Energy, The Proc of the Institution of Civil Engineers, Vol 136, No. 1. ISSN Ashley, R.M., Bertrand-Krajewski, J.L., Hvitved-Jacobsen, T. and Verbanck, M (2004), Solids in sewers - characteristics, effects and control of sewer solids and associated pollutants, IWA Publishing, ISBN Baur R, Herz R, Kropp I, (2003), Procedure for choosing the right rehabilitation technology, CARE-S report D16. BSI (1998) BS EN 752-4: 1998 NA Drain and sewer systems outside buildings. Hydraulic design and environmental considerations, British Standards Institution. Clarke, G.M. and Cook, D., (2005), A basic course in statistics, Arnold, London, ISBN EPA, (1999), Optimization of collection system maintenance frequencies and system performance, EPA Cooperative Agreement #CX , February Fenner, R., McFarland, G., & Thorne, O., (2006), A case based reasoning approach for managing sewerage assets to be published in Water Management, The Proceedings of the Institution of Civil Engineers, Vol 160, No 1, McDougall, J.A. and Swaffield, J.A. (2003). The influence of water conservation on drain sizing for building drainage systems. Building Serv. Eng. Res. Technol. Vol 24 No. 4, Memon, F. A., Fidar, K. Littlewood, D. Butler, C. Makropoulos and S. Liu (2007). A performance investigation of small-bore sewers. Water Science & Technology, Vol 55 No 4 pp Moss, B. & Woodhouse, J. (1998), Criticality Analysis Revisited, Jou of Asset Management, Vol 12, No. 3, pp Savic, D., Giustolisi, O., Berardi, L., Shepherd, W. Djordjevic, S. and Saul, A. (2006), Modelling sewer failure by evolutionary computing, Water Management, The Proc of the Institution of Civil Engineers, Vol 159, issue 2. Shepherd, W., Cashman, A., Djordjevic, S., Dorini, G., Saul, A., Savic, D., Lewis, L., (2005), Investigation of blockage relationships and the cost implications for sewerage network management, Proc of the Int Conf of Urban Drainage, Denmark. WRc, (2000), Sewerage rehabilitation manual (SRM) 4th Ed, Water Research Centre, Swindon, ISBN Yao, K.M., (1974), Sewer line design based on critical shear stress, ASCE, Journal of the Environmental Engineering Division, Vol 100,