There are over 300,000 kilometres of sewers in England and Wales under the control of private operators. Review of the literature and current practices revealed that the
industry-wide management strategy which combines scheduled replacement of critical pipes with reactive maintenance is not sustainable and the Environment Agency is calling for improved management strategies to reduce the number of pollution incidents.
This study looks at the use of case based reasoning (CBR) approaches for appraising the asset condition and performance of water industry infrastructure. While it focuses on sewer pipes, it is proposed that the methods developed could be applied to other water industry infrastructure including pumps and water reticulation pipes.
Through a comprehensive literature review, CBR was recognised as a potential tool for use in the development of a more sustainable management strategy. Its ability to solve problems in complex domains without the need for generalised rules and reasoning make it an attractive tool. Further, it requires a relatively small number of historical cases (when compared to statistical models being trialled in Europe), meaning that extensive data covering a wider range of attributes can be collected for each pipe, providing a better representation of the complex deterioration processes.
Attributes to describe each case and its associated solution (management action) were chosen on the basis of known causal factors and available data. Preliminary discussions with a private water utility company indicated that availability of data would not be a limiting factor. However, the data collection process revealed a lack of connectivity between databases and inaccurate or missing data, limiting the data that could be collected for this study. It is proposed that if industry is to adopt a CBR approach, this could be overcome through a combination of local knowledge, CCTV inspections and site visits. Such data collection efforts can be justified on the basis that CBR methods require a relatively small number of cases. These efforts would not be possible for statistical models.
In developing a prototype CBR model, a nearest neighbour search algorithm with an attribute based similarity factor was found to be the most suitable method for retrieving similar cases from the case library. User defined weightings were used to represent the relative importance of each attribute. Preliminary trials indicated that the model is able to distinguish between cases within the case library and correctly identify the most similar pipes. Gaps identified in the coverage of the problem space can be addressed through improved case selection and the addition of new cases over time.
Indices were developed to infer information regarding the condition, performance and management decisions for unknown pipes from the most similar cases retrieved. A consequence index encompassing social, environmental and economic factors was developed to prioritise pipes for maintenance, minimising the total impacts across the
network. A qualitative ‘expert judgment’ index was also included to allow for the incorporation of experiential learning and tacit knowledge within the model.
The final result is the development of a management strategy with an emphasis on proactive maintenance, which will change the current economic model within the industry. Through the use of a consequence index that is not purely economic based, it represents not only potential economic savings but also social and environmental benefits, aiding managers in the development of a more sustainable management strategy.