Pressure Sensor Placement for Pipe Roughness Calibration Based on Graph-Based Surrogate Model Coupled with Genetic Algorithm

In this study, a graph-based method is implemented for sensor placement in a water distribution network (WDN) instead of using a hydraulic model. The proposed methodology determines the pressure sensors’ location based on the node betweenness centrality of nodes from their source, considering the WDN topology and assigning hydraulic-inspired edge weights. Furthermore, the Non-dominated Sorting Genetic Algorithm (NSGA-II) determines the end node of the WDN’s critical paths for sensor placement to maximize monitoring network efficiency to calibrate the model and avoid additional data collection. For different numbers of sensors, the NSGA-II algorithm is implemented 10 times and the final Pareto front is determined. The graph-based approach reduces the sensor placement problem complexity to an acceptable level and can be implemented as a surrogate approach for hydraulic-based sensor placement.