ABSTRACT The calibration of drinking water distribution network (DWDN) models is essential to ensure accurate simulation, efficient operation, and informed decision-making. As DWDNs evolve due to seasonal changes, shifting demands, or infrastructure updates, maintaining model accuracy over time becomes increasingly important. However, limited measurement availability and high model complexity make calibration a persistent challenge. To address this, ES-NEAT is introduced, an automatic calibration methodology that combines expert systems (ES) with neuro-evolution of augmenting topologies (NEAT). The method integrates expert knowledge with neural network evolution to efficiently solve high-dimensional calibration problems. ES-NEAT achieves high accuracy under sparse data conditions while keeping computational costs moderate. It also stores calibration knowledge in a structured format, enabling faster and more consistent recalibration over time. This adaptability supports long-term model reliability. The methodology was validated on a benchmark network and a real DWDN in Flanders, Belgium, demonstrating robust performance, efficient convergence, and generalizability across calibration scenarios.
Gómez et al. (Wed,) studied this question.