A simulation-based digital-twin prototype for pressure reconstruction and burst response in water distribution networks

Abstract

This dataset contains the complete code, random seeds, and result arrays for a simulation-based digital-twin prototype that reconstructs the full nodal pressure field of a water distribution network from sparse pressure sensors. It includes a global-gradient Hazen-Williams solver verified against the EPANET engine to sub-millimetre agreement, a learned surrogate of the hydraulic model, and a streaming ensemble Kalman filter, together with weighted-least-squares and extended-Kalman baselines. Experiments cover a 100-node grid, the EPANET Net3 benchmark under an independently aged truth network, and span solver verification, surrogate training and timing, filter diagnostics, representation-error schemes, sensor placement and multi-asset studies, and pressure-dependent emitter bursts with calibration separated from testing. Each figure and reported value in the associated manuscript is reproduced by a named script, and all results are stored as JSON with the seeds needed to regenerate them.