WDNDB provides a space where all the researchers and companies share with public the results of their optimization algorithms in solving these 12 benchmark problems.
What is the WDNDB challenge?
The multiobjective design of real-world, large-scale water distribution systems (WDS) are challenging problems for multiobjective optimization (MOO) algorithms.
The design cost of the network is often among the multiple objectives in this context and millions of dollars could be saved taking design decisions on the basis of a superior approximate Pareto front (PF).
Efforts toward identification and development of MOO algorithms that are capable of good performance on WDS design problems have a long history now.
Performance comparisons of MOO algorithms in dealing with WDSs of various sizes is a critical component of research in optimization.
Wang et al. (2015) set up an archive of 12 benchmark bi-objective WDS design problems.
The design objectives are to minimize total pipe material costs and maximize the system resilience calculated according to Wang et al. (2015).
The benchmark problems range in network size from 8 to 3032 pipes and 8 to 567 integer decision variables associated with defining the pipe diameters of some or all pipes in the network.
Wang et al. (2015) aimed to obtain the best-known PFs for those benchmark problems given extensive computational budgets and applied five different MOEAs.
The algorithm used in Wang et al. (2015) were
The well-organized set of WDS benchmark design problems presented by Wang et al. (2015), and the corresponding reported best-known PFs for those problems
provides the opportunity to implement and investigate the performance of other multi-objective optimization algorithms in solving the set of benchmark problems.
presents the best-known Pareto fronts of twelve benchmark design problems of Water Distribution Systems. These problems were collected from the literature and the Pareto fronts were generated using five state-of-the-art multi-objective evolutionary algorithms (MOEAs).
These benchmark problems are categorised into four groups (small, medium, intermediate and large) according to the size of search space. Each problem is formulated to minimise the total cost and to maximise the network resilience. Five MOEAs including two hybrid algorithms (AMALGAM and Borg) were employed to identify the currently best-known Pareto front of each problem given extensive computational budget. The EPANET input file, the associated source code and the data of best-known Pareto front for each benchmark problem are provided in this page.
The aim of this public data is to allow other researchers in the community to have multiple reference points to test new techniques and optimisation algorithms.