Dataset for: Accurate and Efficient Junction Mixing Prediction in Water Distribution Networks via Surrogate Modelling

The quality of the water flowing through Water Distribution Networks (WDNs) is a fundamental parameter for the operation of these systems, as it affects public health, local economies and the networks themselves. Hence, the use of models predicting water quality for, among others, scenario analysis, monitoring, and control is widely understood to be of high importance. However, by virtue of the extensive nature of most WDNs, these models are simplified, typically being one-dimensional and assuming perfect mixing in the network’s junctions. This assumption is often incorrect and can lead to severe misrepresentations of the expected propagation of solutes or the possible consequences of contamination events. The growing importance of contaminants of emerging concern has further increased the significance of these considerations. Accordingly, in this work, a high-resolution screening of a wide range of possible operational conditions for three junction geometries commonly found in WDNs was performed with an experimentally validated Computational Fluid Dynamics (CFD) model. The resulting extensive dataset is publicly available and was used as the basis for a surrogate model based on artificial neural networks. The proposed approach was found to be capable of accurately recovering the original CFD predictions at a fraction of the computational cost. Moreover, satisfactory interpolation along the studied operational space was observed as well. Finally, the surrogate model was coupled with the EPANET hydraulic simulator, which allowed for the efficient analysis of the influence of imperfect mixing under different conditions in a realistic urban WDN.

Jaar van publicatie:2025

Toegankelijkheid:open

Original language:UN

Uitgever:Zenodo

Licentie:CC-BY-SA-4.0

Formaat:text/csv

Trefwoorden: Dataset