Author(s): Alejandro Gonzalez Barbera; Sergio Chiva Vicent; Delia Trifi Rufino; Jaume Luis Gomez; Oscar Prades Mateu; Guillem Monros Andreu; Paloma Barreda Juan; Raul Martinez Cuenca; Rosario Arnau Notari; Javier Climent Agustina

Linked Author(s):

Keywords: Digital twin; ML; CFD; WWTP

Abstract: This research introduces a hybrid digital twin (DT) framework for Wastewater Treatment Plants (WWTPs), seamlessly integrating Computational Fluid Dynamics (CFD), Machine Learning (ML), and Industry 4.0 principles to support advanced decision-making tools. Leveraging historical inlet data, CFD simulations generate a comprehensive dataset of hydrodynamic fields within a WWTP biological tank, which then trains a high-fidelity ML surrogate model. This surrogate achieves strong predictive accuracy, with R² scores of 0.93 and 0.95 for |U| and nut, MSE of 1.3e-4 and 2.5e-4, and SSIM of 0.98 and 0.97, respectively. Following validation, the ML model couples with an OpenFOAM-based ASM1 solver to simulate the evolution of biochemical species, such as S_NO and S_NH, throughout the tank. The integrated workflow enables simulations up to 22 times faster than real-time scenarios. By bridging physics-based modelling with ML-driven efficiency, this DT advances real-time operational insights, fostering resilient and sustainable urban water systems.

DOI: https://doi.org/10.64697/iahr.proc.hic2026.148

Year: 2026