Physics-informed intelligence in structural health monitoring: a state-of-the-art review and roadmap towards digital twins

Built-environment assets such as bridges require monitoring methods that support maintenance prioritisation, repair planning, and risk-informed intervention under sparse, noisy data. Physics-informed intelligence, which embeds governing equations, material behaviour, and engineering constraints into deep learning frameworks, has emerged as a promising approach for structurally consistent structural health monitoring and digital twin-enabled bridge asset management. This review synthesises 50 studies published between 2021 and 2026, including 42 primary studies and eight review papers, to provide an evidence-based synthesis of physics-informed intelligence for bridge health monitoring. The synthesis is organised through three components: a taxonomy of architectures, training strategies, applications, and structural types; a seven-criterion quality appraisal with comparison of representative models; and a dependency-gated roadmap for bridge digital twin deployment. This contribution connects method classification, evidence appraisal, and deployment guidance, showing that deployment readiness depends on the engineering prior, validation tier, uncertainty quantification, benchmark comparability, field calibration, and decision-support integration.