Integrated fuzzy condition assessment and decision support for water pipe mains

This paper presents a novel Integrated Fuzzy Condition Assessment and Decision Support (IFCADS) method that has been applied to predicting the condition of large pre-stressed concrete cylinder water pipes. IFCADS encodes human expertise within fuzzy rules to emulate human reasoning and solve condition problems when data are scarce. It is based on a new, simple and intuitive elicitation method for: (i) converting both input variables and hierarchical concept memberships into fuzzy linguistic values and (ii) determining the relative influences or weights of each child node on the parent compared to their siblings. These tasks are scalable to high-dimensional and complex hierarchical knowledge domains because all the associated rules can be automatically generated without any further human input. Additional innovations of IFCADS involve improved representation and processing of fuzzy uncertainties. Fuzzy equality has been made more consistent with other fuzzy comparisons by ensuring fuzziness extends equally on both sides of the mid-point. The integrity of fuzzy values is maintained even when the combined fuzzy influence of children on their parent concept has an upper triangular extension beyond the parent’s value range. And there is no interim defuzzification during the inference process, which means full fuzziness is carried through from inputs to outputs. IFCADS uses this combination of fuzzy numbers, linguistic variables, and If-Then rules to facilitate elicitation of uncertainties associated with decision-making and convert expert knowledge into a mathematical formalism. It was applied to the challenges of assessing buried water pipe conditions using limited and imprecise data from the Libyan Man-Made River Project (MMRP), which manages thousands of kilometres of pipes carrying water from the desert to coastal conurbations. IFCADS performed better than the existing model used by the MMRP and, more pertinently, better than similar fuzzy approaches that lack the full-fuzzification innovations of IFCADS. The application demonstrates a method that has the flexibility and tractability to be applied in many different knowledge-rich and high-dimensional domains of human expertise.