Capacity of urban drainage systems (UDSs) can substantially influence flooding properties of urban catchments. This motivates many studies to optimally design UDSs often using multiobjective evolutionary algorithms (MOEAs) as they can explore trade‐offs between conflicting objectives (e.g., cost vs. system reliability). However, MOEA‐based approaches are typically computationally demanding and their solutions are often practically unacceptable as engineering domain knowledge is often not explicitly considered. To address these two issues, this paper proposes an efficient optimization framework for UDS design, where an engineering‐based design method (EBDM) is developed to generate approximate solutions to initialize the MOEA's search, thereby greatly enhancing the optimization efficiency. To improve the solution practicality, two ideas have been implemented in the proposed optimization method (PM): (i) the variability of peak depths across pipes is minimized and (ii) a constraint is introduced to ensure that sizes of pipes in the downstream direction are no smaller than their corresponding upstream diameters. Two real‐world UDSs of different size are used to demonstrate the effectiveness of the PM. Results show that (i) the proposed EBDM is effective in producing initial solutions that are very close to the final solutions identified by the optimization methods, (ii) the minimization of the variability of peak depths in pipes is practically meaningful as it can facilitate to identify solutions with great ability in handling future uncertainties (e.g., rainfall variability), and (iii) the PM significantly improves optimization efficiency and solution practicality compared to the traditional optimization approach, with benefits being more prominent for larger UDSs.

No abstract

No abstract