Comparisons between a rule-based expert system and optimization models for sensor deployment in a small drinking water network

Abstract

In response to the needs for long-term water quality monitoring, one of the most significant challenges currently facing the water industry is to investigate the sensor placement strategies with modern concepts of and approaches to risk management. Most of the previous research mainly focuses on using optimization models to deal with small-scale drinking water networks. Yet the challenge of NP complete when handling large-scale networks can never be overcome. This study develops a rule-based expert system (RBES) to generate sensor deployment strategies with no computational burden as we oftentimes encountered via various types of optimization analyses. Two rules, including the accessibility and complexity rules, were derived to address the characteristics of effectiveness and efficiency required for sensor deployment in these networks. To retrieve the information of population exposure, the well-calibrated EPANET model was applied for the vulnerability assessment leading to the derivation of the accessibility rule whereas the graph theory was employed to retrieve the complexity rule eliminating the need to deal with temporal variability. Comparisons between this new expert system and 14 existing optimization and heuristic models confirm that the newly developed expert system in this paper can always compete with most of the optimization models. With no computational burden, the RBES that is designed to promote health risk management could also be applicable to deal with similar applications in large-scale drinking water distribution networks.