Within distributed infrastructure systems like water supply networks, some assets are individually much more critical than others. So-called ‘reactive assets’ constitute the majority of assets.
Reactive assets have a low consequence of failure and thus can be left to operate until failures start to occur. A decision must then be made on whether to maintain or replace them. Such decisions must consider budget constraints, the economics of continuing to operate the existing asset, impacts on customer service, and the availability of other risk management strategies such as pressure reduction, shut-off block reduction, etc.
Even when a reactive asset is identified as being a potential candidate for renewal, risk-based techniques still need to be applied to determine its priority in comparison to other candidates.
For distributed network assets, a clustering exercise will also help to develop economic work packages. This involves assessing the economic and practical benefits of replacing near-by assets at the same time. In PARMS this is achieved using a seed-grow-expand model:
Since reactive assets are allowed to fail, deterioration models can be developed via analysis of failure data. These describe how assets are expected to fail into the future based on patterns of past failures, as explained by a range of factors.
In PARMS, a series of cohort-specific deterioration curves are developed via statistical analysis of historical failure records and corresponding pipe attributes, including length, material, age, diameter, soil type, pressure, prior failure history, etc.
These cohort level models are then adjusted using asset-level failure histories via the application of so-called Best Linear Unbiased Predictors (BLUPS). BLUPs allow meaningful predictions to be made at the asset level. This is important when considering risk and integrating new failure data into the modelling as it arises, which is needed for effective targeting of expenditure.
PARMS Retic