Structural health monitoring (SHM) is essential for ensuring the safety and durability of civil infrastructure. Traditional SHM approaches, based on manual inspections or threshold-based analyses, often fail to detect early or subtle structural changes. In this work, we propose a data-driven framework for detecting structural regime transitions using long short-term memory (LSTM) networks trained on power spectral density data. This method does not require prior knowledge of the excitation sources or structural dynamics, enabling robust and interpretable transition detection under real-world conditions. A key component of the framework is the empirical transition point, $T_a$, computed from prediction probabilities through persistence thresholding and entropy filtering. This allows for precise and automated detection of regime shifts, even in the absence of explicit ground truth. The model is validated on vibration data collected under ambient and train-induced excitations, achieving high accuracy in distinguishing pre- and post-retrofitting states. It demonstrates strong robustness across a range of operational and dynamic conditions. To enhance interpretability, we introduce the confidence variability index (CVI), which quantifies the temporal stability of the model's predictions and serves as an indicator of transition consistency. While the framework does not currently identify the physical causes of transitions, its sensitivity to dynamic changes makes it a valuable early-warning tool in SHM. Despite the inferential nature of transition detection due to the lack of ground truth, the approach offers a scalable, interpretable, and real-time solution for structural regime monitoring—contributing to the advancement of SHM systems through uncertainty quantification, causal inference, and intelligent infrastructure management.

We introduce a pointwise multiscale texture operator providing a mathematically grounded and interpretable description of how local image context evolves across observation scales. The operator is defined through the scale evolution of a contextual multiscale signature derived from Gaussian scale-space representations, thereby establishing direct connections with diffusion processes and multiresolution analysis. This formulation yields a stable descriptor of texture transitions that is robust to moderate noise, blurring, and acquisition variability, while remaining independent of any specific segmentation strategy. As a demonstration of its practical utility, we embed the proposed operator within a texture-coherence-driven region-growing framework for image segmentation. The resulting algorithm is dimension independent, computationally tractable, and does not rely on training data or complex architectural design. We illustrate its performance on 3D cone-beam computed tomography (CBCT) datasets, where it enables coherent segmentation and volumetric reconstruction of the dental pulp chamber under clinically relevant conditions involving low contrast, partial volume effects, and heterogeneous textures. Beyond this specific application, the proposed operator constitutes a general framework for multiscale texture analysis and structural characterization, particularly suited to imaging scenarios where interpretability, robustness to acquisition degradation, and limited annotated data are critical.

Structural health in civil engineering involved maintaining a structure's integrity and performance over time, resisting loads and environmental effects. Ensuring long-term functionality was vital to prevent accidents, economic losses, and service interruptions. Structural health monitoring (SHM) systems used sensors to detect damage indicators such as vibrations and cracks, which were crucial for predicting service life and planning maintenance. Machine learning (ML) enhanced SHM by analyzing sensor data to identify damage patterns often missed by human analysts. ML models captured complex relationships in data, leading to accurate predictions and early issue detection. This research aimed to develop a methodology for training an artificial intelligence (AI) system to predict the effects of retrofitting on civil structures, using data from the KW51 bridge (Leuven). Dimensionality reduction with the Welch transform identified the first seven modal frequencies as key predictors. Unsupervised principal component analysis (PCA) projections and a K-means algorithm achieved $70\mathrm{\%}$ accuracy in differentiating data before and after retrofitting. A random forest algorithm achieved $99.19\mathrm{\%}$ median accuracy with a nearly perfect receiver operating characteristic (ROC) curve. The final model, tested on the entire dataset, achieved $99.77\mathrm{\%}$ accuracy, demonstrating its effectiveness in predicting retrofitting effects for other civil structures.