Modal tracking plays a vital role in structural health monitoring since changes in modal parameters help us understand a structure’s dynamic characteristics and may reflect the potential deterioration of structural performance. Although numerous modal parameter estimation (MPE) methods exist, it is not guaranteed that an MPE process will exclude all spurious modes and not lose any physical modes every time over a long-term monitoring period. Relatively large damping of a structure, poor data quality, and significant changes in structural modal parameters may make the estimated modal parameters spurious, missing, or misclassified. It makes long-term modal tracking semiautomated or manual, which constrains timely downstream applications such as anomaly detection, condition assessment, and decision making. This research aims to propose a long-term continuous automatic modal tracking algorithm based on Bayesian inference even when the modal parameters, damping, and data quality change significantly. Bayesian inference is used to determine the physical modes from the results of existing MPE methods. Both the modes identified from the most recent response set and the modal probability model from multiple previous response sets are considered in the Bayesian model to better determine the physical modes from the results of MPE. Moreover, the proposed algorithm requires only three extra hyperparameters compared to general modal tracking algorithms, and they can be quickly determined by a grid search method. The performance of the proposed algorithm is verified by a numerical example and a real-world civil structure Z24 Bridge benchmark.

Efficient level-of-detail (LOD) management is crucial for handling large-scale 3D meshes in BIM, GIS, and digital twin applications. In practice, both individual models and complex multi-mesh scenes require multi-resolution representations. Yet two practical issues persist: (i) simplification rates are often fixed a priori, lacking principled guidance and yielding suboptimal fidelity–cost trade-offs; and (ii) after a scene-level target is set, workflows commonly impose a uniform rate on all models, which is ill-suited to heterogeneous geometry and produces uneven visual quality. This paper presents an automatic approach that constructs a cumulative edge collapse loss curve using a QEM (Quadric Error Metrics)-based process. Shape analysis of this curve defines four representative LOD targets, and an automated procedure then determines their corresponding simplification rates. The method is first developed for individual meshes and then extended to multi-mesh scenes, assigning model-specific rates that satisfy a prescribed scene-level reduction while maintaining visual consistency. Experiments on complex engineering datasets show higher fidelity than uniform-rate baselines, especially at high reductions. The approach provides a practical, automated framework for object- and scene-level LOD generation.

Large-scale engineering of the 3D scenes used in BIM, GIS, digital twins, and geospatial web delivery frequently suffer from significant geometric redundancy after export to mesh-based delivery formats, arising in part from the inconsistent reuse of geometry, where many repetitive components are stored as independent meshes rather than being fully instantiated. This paper proposes an affine–covariant mesh instancing framework designed to achieve a lightweight representation of watertight triangular solids. The core of the method lies in a canonicalization pipeline: each mesh is normalized via volume-centroid translation, principal-axis alignment derived from volume covariance, and anisotropic covariance whitening. This process effectively decouples the influence of translation, rotation, and non-uniform scaling, projecting diverse geometries into a unified canonical space. Within this space, geometric similarity is quantified by evaluating compact descriptors against user-defined tolerances. A greedy clustering strategy is then employed to group affine–similar models based on these descriptors. Finally, the scene is efficiently reconstructed by applying inverse affine transformations to the representative instance of each cluster. The output stores one shared geometry per cluster alongside per-instance $4\times 4$ transform matrices, preserving the original spatial layout while reducing redundant geometry storage. Experiments on four real-world engineering scenes demonstrate varying compression benefits. The results prove particularly effective for scenes containing unlinked repetitive parts and affine–similar parametric components, while also revealing a controllable trade-off between fidelity and compression rate. The method is therefore suitable as a post-export geometry-lightweighting step in mesh-based BIM/GIS integration, infrastructure digital twins, and large-scale 3D mapping workflows.

Shanghai Tower is currently the tallest building in China, equipped with a tuned mass damper (TMD), and situated in a typhoon-prone area. This study presents the findings of a vibration-based structural health monitoring system for continuously tracking the dynamic characteristics during a major typhoon event (Ampil Jul 2018). While previous studies manually categorized time history into three distinct periods (before, during, and after the typhoon), this study utilizes an Ordering-Points-To-Identify-Clustering-Structure clustering algorithm to divide the time history into four finer-grained time periods for more objective and automated decision. The dynamic characteristics of the building-TMD system in the four time periods are analyzed in both the time and frequency domains. In view of tracking continuity, and potential presence of spurious modes or omission of physical modes, a Bayesian approach that considers both reference modes and reference probability model is investigated.

Structural health monitoring (SHM) plays a critical role in ensuring the safety of major infrastructure, yet monitoring data inevitably encounters various anomalies, including structural damage, adverse environmental effects, and sensor faults. However, existing methods are typically designed for single anomaly types and either require prelabeled training data or depend heavily on expert knowledge to interpret and categorize detected anomalies. To address these limitations, this study proposes a spatiotemporal clustering‐based anomaly detection algorithm to identify and interpret normal and anomalous structural states. The algorithm identifies multiple structural states without requiring prelabeled data or predefined anomaly metrics, given a user‐specified number of clusters k . A method similar to expectation maximization (EM) is applied to compute the Gaussian inverse covariance matrix, capturing the spatiotemporal relationships of multisource monitored physical quantities and providing physically interpretable information to assist engineers in the preliminary identification of each structural state. The effectiveness of the algorithm is validated through a numerical example and further applied to the Shanghai Tower project. The results demonstrate that the algorithm not only detects anomalies effectively but also automatically discriminates between different structural states, with an inference time of less than one second enabling near‐real‐time anomaly detection.