One of the key words being used in the water and wastewater industry globally is ‘risk' . Asset managers now plan the most cost-effective approaches to managing their network by determining the risk grade for each asset.
This article first discusses in detail how network
|modeling can automate the process of measuring the impact of the failure of a link, and then examines automated methods for calculating the probability of failure.|
For the first time, network models can now be used routinely as the basis for intelligent asset management in the water industry. The reasons for these developments are:
- Network models, once the preserve of the specialist and previously used solely for capital planning, are now being used in operational environments
- Corporate data systems (GIS, SCADA) readily provide the up-to-date data for network models
- Advances in hardware and software mean models and results can now be delivered orders of magnitude faster than previously
- Modeling is moving from a PC environment to a central server, allowing the rapid sharing of the latest information
- The solving capability of network modeling software engines, such as InfoWorks WS, provides the scope to run simulations in batch mode for several scenarios
- Modeling software can now provide audit trails, giving backup to decision making if required.
A suggested definition of a critical link is ‘a link whose failure causes a reduction in level of service below acceptable thresholds or causes other assets to fail'. A failure of a link can have the following results:
- Supply problems - loss of supply, reduction in water pressure, deterioration in water quality
- Financial loss - revenue loss, cost of water losses, customer rebates, repair
- Compensation payments - flooding of the local community, road and property damage
- Environmental damage
- Health and safety issues
- Traffic disruption
- Damage to company reputation.
The output of a critical link analysis is a worst first main-by-main consequence based ranking.
Critical link analysis (CLA) will help water authorities meet their regulatory obligations as the water industry moves into an intelligent asset management environment, underpinned by risk considerations.
The outcome of the analysis should be a prioritized list of water mains by criticality score, along with alternative ways of operating systems to mitigate such failures. Criticality grading is carried out on a quantitative basis although unquantifiables such as traffic disruption and public image / perception should not be overlooked.
A quantitative approach to CLA
The simplest form of automated CLA is carried out by modeling the removal of one main and measuring the resulting impact based on a number of criteria such as:
- Number of key clients isolated without water – large revenue clients or ‘at-risk' clients such as dialysis patients
- Total number of customers isolated or receiving below a minimum pressure threshold
- Percentage demand reduction
- Maximum network pressure. Mains failure can have a knock-on effect causing other assets in the network such as pressure transducers and pipes to fail
- Incidence of negative network pressures. The impact of local high points on the network hydraulics must not be overlooked. When pressure suddenly becomes sub-atmospheric at such a point, arising from a main's burst or outage, either at that point or somewhere else in the network, the possible ingress of contaminants at that point, and their migration throughout the network, has to be considered
- Number of customers affected by sediment re-suspension
- Number of customers receiving water over a specific age
- Number of customers receiving water outside water quality parameters e.g. minimum residual chlorine
- Maximum volume / cost of water lost due to the burst main.
To be able to carry out accurate critical link analysis specific information is required:
- A calibrated hydraulic model of the network
- The average response time/ outage period required to repair a burst main
- The location and status of valves to enable both the current status of the network and the operational alternatives to be assessed
- The flow rate of a typical burst at a nominal pressure
- Geo-referenced customers identifying key customers (although some software packages allow calculation to be based around nodes and weightings).
Hydraulic analysis requirements
In its simplest terms critical link analysis models a pipe failure as the closure of a pipe. But often this is not the case. In reality a pipe burst may prevent any flow continuing downstream (such as if the failure is at a bridge crossing) or may reduce pressure-related consumption (underground failure). The outage area (isolated network) also changes over time as valves are changed during the managing and repair of the burst. The final consideration is that the outage may not be the direct cause of any reductions in level of service; therefore it must be possible to extend the analysis to such issues as reservoirs emptying.
To fully address the real complexities of a pipe burst, the software should have the capability to model the following accurately:
All assets, and not just pipes - CLA to date has concentrated on pipes. But for full analysis all pumping stations, valves, meters and other assets should be considered as capable of failing and need to have a criticality grade assigned to them.
Pressure related demand (PRD) - Removing pipes from service frequently has a major impact on pressures and hence the demand. Therefore, a realistic pressure related demand capability is essential for obtaining realistic results of automated CLA.
Volume of burst flows - Often CLA analyses are simplified to simply removing the asset from service and not considering the consequence of the burst, which in reality may have a major impact on demand and pressure, especially it is a large diameter main. In ‘true' CLA the loss through burst flow should also be considered. For example, comparing a pipe that has collapsed over a bridge in comparison to a pipe leaking underground, the former has far greater loss and prevents downstream supply, whereas the latter may still allow demand downstream to be supplied.
Changes to burst flows over the duration of the outage - The outage period needs to be considered in order to allow the time between the occurrence of the burst/pipe isolation and the completion of the repair to be analyzed. Often the increased demand caused by the losses of the burst will not extend for the full length of the outage period, due to the failure being isolated by closing bounding valves.
Seasonal variations in demand - As all hydraulic modelers should be aware the seasonal and diurnal variations of demand can drastically effect pressures within the network. It is therefore essential that any analysis undertaken can account for this.
Base results comparison - Often networks can contain known areas that are outside the thresholds e.g. high pressures caused by pumping. To ensure these areas are excluded from the subsequent results analysis it is desirable for the software to automatically create and compare the results produced from the critical link analysis with a base simulation; this will allow only the increases in threshold failures to be identified. The base simulation contains the results for the ‘normal' operation of the network.
Likelihood of failure analysis
Risk assessment involves combining the estimated impact of a failure with the probability of that failure occurring. Probability of failure estimates can be based on such data as:
Asset failure records
Soil / bedding environment
Water quality failures
Hydraulic results – friction factor
Pipe leakage rate
Pipe sampling information.
Since the data required for this type of analysis is very varied in type and possibly very large, it is suggested that this type of analysis should not be carried out in a network-modeling package but instead carried out in an asset management tool such as Wallingford Software's InfoNet.
Future enhancements and research
Further enhancement to the tools in the software could include:
- What happens during the outage period? If the failed pipe is isolated by valve shut-offs, the analysis must reflect the extended ‘zone of influence' of the isolated pipe during the outage period. Greater user defined pressure related demand functionality, based on for example user-defined curves or coefficients.
- Option to weight burst link repair time / outage based on pipe diameter and depth of cover.
- Water quality (WQ) and sedimentation (SED) impacts being included in the simulations in conjunction with criticality analysis. Currently all software available that provides this functionality only allows it as a subsequent and separate analysis to the critical link analysis.
- Automated identification of the immediate responses that enable temporary supply of water into the area impacted by the outage.
The automation of Critical Link Analysis and probability of failure analysis will provide a resource for operational modeling using ‘real time' SCADA and telemetry data to provide predictive models, shut down manuals and contingency plans. It should provide the technical applications that are needed by water utilities to meet the ‘risk based consequence of failure' approach that is expected of them.
Automated CLA will provide a test bed for any main where a contingency plan/shutdown manual is required in the event of an outage. Caution is advised for very large trunk mains where other factors other than re-zoning are relevant, such as upping the output from treatment works. The aim of such contingency plans is to provide information on:
Which valves to operate in an emergency
Where the valves are located
Whether the valve is operational.
CLA and automated risk assessments will never completely replace the knowledge of an engineer, but it can drastically reduce investigation time, and direct management towards the best value for money solutions, both operational and capital.
Wallingford Software's InfoWorks WS for water distribution modeling supports automated Critical Link Analysis. Specific features of Info Works WS that are advantageous for CLA are:
Standalone engine - CLA can often be very resource intensive, leading to long run times. The InfoWorks WS software engine is independent of other applications, ensuring that other areas work can be continued without interference, such as GIS updates.
Batch running of simulations – To further address the issue of long run times, CLA works in batch mode, which is supported by InfoWorks WS. It is also recommended that results are kept only for links that fail the threshold criteria, to guard against very large results files.