Physics-Informed Criticality Analysis for Infrastructure Networks

Abstract: In this work a Machine Education (ME) based method is presented. A framework for Hydraulically-Informed (HI) criticality scoring of urban sewer network components was implemented by using SWMM as a physical educator. Capacity of each element is systematically reduced and hydraulic consequences are calculated. The framework generates consequence-based criticality indicators aggregated into a per-conduit HI score, validated across multiple real-world and benchmark networks. Building on the physically enriched dataset, surrogate Educated Machines are trained to predict HI criticality, substantially reducing the computational burden of criticality assessment at scale.