414 - Enhancing the Efficiency of Drainage Network Partitioning for Parallel Computing

Parallel processing has applications in many real-world scenarios. Effective parallel processing for urban drainage networks (UDNs) can prove to be useful in reducing computational efforts and making hydraulic modelling a more viable tool for very large networks. Effective network partitioning is the basis of efficient parallel processing. Therefore, developing a partitioning strategy based on graph theory that caters for the topology of a UDN is important. Such an approach for partitioning drainage networks has been developed by University of Texas at Austin. The aim of this study is to make this drainage network partitioning approach more efficient by testing it for a number of diverse case studies with different spatial extents and network topologies in order to analyze its effectiveness in different cases. A cost objective function is defined to analyze the effectiveness of the partitioning algorithm for different case studies. The time for partitioning and the time for running the partitioned network in a hydraulic model is also studied. The results suggest that the cost function is higher for networks having complex topology. It is also concluded from the results that using other weights that are hydraulically more suitable than length can increase the efficiency of the algorithm.