Bridges and viaducts play a critical role in urban infrastructure, ensuring mobility and supporting essential logistical chains. Their exposure to weather, dynamic loads, and aging requires continuous monitoring, as undetected damage can eventually develop into costly structural failures. Traditionally, visual inspections are used for evaluation, but they face limitations: difficulty accessing critical areas and subjectivity in defect interpretation, which depends on the inspector s expertise. Structural Health Monitoring (SHM) systems have emerged as alternatives, using dynamic properties (natural frequencies and mode shapes) to identify changes in structural stiffness. However, environmental factors such as thermal variations interfere with these properties, necessitating methods to distinguish external effects from actual damage. In this context, integrating Building Information Modeling (BIM) and digital twins represents a strategic advance in asset management, enabling the centralization of multidisciplinary data (design, intervention history, inspections) and their association with updatable numerical models to create a dynamic, intelligent infrastructure representation. This work proposes an innovative framework combining SHM, machine learning, and a BIM-based framework to optimize bridge management, focusing on early damage detection and reducing human uncertainty. The methodology was validated through a case study on the Rio Claro overpass (SP-340), which underwent environmental two vibration tests performed 10 years apart. A finite element model (FEM), was developed and updated using the experimental data with the objective of replicating the structure s dynamic behavior. The calibrated model was used to simulate artificial damage at various positions and intensities. Modal curvatures, damage sensitive features derived from vibration mode shapes, were extracted from simulated cases to train artificial neural networks (ANNs) for damage localization and severity estimation. The ANNs achieved classification accuracy exceeding 90 percent, thus, validating their effectiveness in filtering environmental interference. The pre-digital twin, developed from a BIM model enriched with historical, technical, and inspection data, was integrated with the ANNs via a Python script, automating integrity reports based on field data and enabling interactive 3D visualization of structural conditions. The results demonstrate that the proposed approach not only improves diagnostic accuracy but also consolidates fragmented information into a single platform, facilitating predictive maintenance decisions, which can help extend the asset s service life. It is intended to provide an intuitive user interface so that asset managers can make data-driven decisions, overcoming the limitations of traditional visual inspections. This research attempts to bridge the gap between BIM and SHM applications by offering a replicable, efficient solution for infrastructure management.