Multi-Criterion Optimization of Partition Devices in Pumping Station Pressure Boxes
Abstract
: This study introduces a multi-criterion optimization approach for the partition device installed in the pressure box of pumping stations for cross-district water delivery systems. The device must distribute the flow effectively among independent branches while limiting hydraulic losses and structural cost, yet its design is still largely guided by empirical rules. To address this limitation, a three-dimensional Reynolds-averaged Navier–Stokes (RANS) computational framework is coupled with Kriging surrogate models and multi-objective particle swarm optimization (MOPSO). Three geometric parameters, the channel length (1200–1800 mm), channel radius (400–700 mm), and edge seal plate height (300–600 mm), are varied over 64 configurations in a full-factorial design. The resulting flow fields are characterized in terms of maximum velocity ratio, local head loss, mean turbulence intensity, manhole pressure difference, and steel consumption. Kriging models constructed for these five responses reproduce the computational fluid dynamics (CFD) results with a coefficient of determination R 2 > 0.99 on independent validation data, providing an efficient basis for the subsequent optimization. The 100 solutions identified by MOPSO reveal three distinct regions of trade-off between hydraulic performance and material consumption, from which a practical three-zone decision framework is established. Within this framework, the design L = 1310 mm, R = 605 mm, and H = 300 mm provides a balanced solution and is independently confirmed by high-fidelity CFD simulations under representative operating conditions. Its implementation is estimated to reduce annual costs by approximately $650,000.