Conference Agenda

Civil structures such as buildings, roads, railways, bridges, tunnels, and dams are present in every society, and the safest and most durable ones are those that are properly managed and maintained. Health monitoring plays an important role in ensuring safety, but engineers still depend on traditional inspection tools like crack gauges or comparator cards, which are slow, labour-intensive, and differ based on the inspector's judgment. There is also no proper tool that systematically analyses this data or suggests solutions based on standard codes using AI. Considering these limitations, this project aims to develop an AI model that processes images to classify whether a defect is structural or non-structural, identify the type of defect, assign a severity score where 0 is worst and 100 is best, suggest causes and remedies using an LLM, and estimate the remaining service life. Literature supports the need for such a system, as studies have focused on deep-learning-based detection of concrete damage, CNN-based comparison between traditional and AI inspection, and 3D scanning with machine learning for defect detection. Surveys on NDT tools like UPV and X-ray show improved efficiency over manual methods but still remain time-consuming and less precise.

Most research concentrates on a single defect such as cracks, and although some models can detect or localize them, there is still a major gap in systems that quantify severity or provide causes and repair suggestions. Many approaches stop at detection without offering guidance for maintenance, making it difficult to translate inspection data into decisions. To address this, datasets in this project are collected from public sources and field surveys, followed by preprocessing such as histogram equalization and median filtering. Images are then manually labelled into structural and non-structural categories and further classified by defect type. An image classification model is trained on this dataset to automatically detect and classify defects. Severity scoring is done using the formula Intensity × Extent, where intensity depends on measurable parameters like crack width or spall area, and extent depends on defect length or percentage affected. The model uses an LLM aligned with Indian Standard codes to suggest possible causes and remedies and predicts remaining service life based on severity. The novelty of this work lies in combining multiple functions that previous studies treat separately: multi-defect classification, quantitative severity scoring, LLM-based engineering interpretation, and service-life estimation. No existing system integrates all these components into a single workflow dedicated to structural inspection.

Buildings are long-term investments, and regular monitoring helps extend their lifespan and prevent failures. This project aligns with SDG 9, SDG 11, and SDG 12 by supporting safer, more resource-efficient, and sustainable infrastructure management. By bringing together defect detection, severity assessment, interpretation, and life prediction, the work addresses limitations of manual inspections and gaps in current automated tools. It supports better maintenance planning and contributes to a more comprehensive AI-based structural health-monitoring framework.

Accurate prediction of the structural temperature field is crucial for the static and dynamic monitoring of engineering structures, with particular significance for heritage buildings where material preservation is paramount. The complex, time-lagged, and non-linear relationship between external air temperature and structural thermal response poses a significant challenge for traditional empirical or statistical methods. This study proposes a framework utilizing statistical regression and advanced recurrent deep learning models to accurately predict structural temperature based on external data. A key focus is the efficient generation of input features to capture delayed and cumulative thermal effects. The methodology is applied in a case study to predict temperatures at various points within a historic basilica. The primary objectives are to remove temperature-induced effects from structural monitoring data for a more accurate assessment of the building's performance and to provide a robust method for imputing missing data. A comparative analysis demonstrates the performance of the models, evaluated using Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and the coefficient of determination (R²). The results offer a reliable and precise methodology for selecting the optimal approach for structural temperature field estimation and data correction in the assessment of heritage structures.

Urban seismology focuses on analysing subsurface structures and enhancing seismic risk management in the cities. Recent advancements, including digital seismic stations and ambient vibration analysis, have increased interests in non-natural seismic signals due to traffic, subways, and other human activities. On the other hand, historic monuments are cultural assets of each society, their importance stems from the variety of roles they play in urban landscape. Thus, their preservation is significant for development and sustainability of cities. This research focuses on the implementation of Structural Health Monitoring system at the Four Days Memorial located in Naples, which is positioned on an isolated podium supported by four elastomeric isolators and four sliding bearings. Firstly, modal analysis is performed to evaluate the dynamic characteristics of the monument before and after implementation of base isolation system. Furthermore, to assess dynamic behaviour of the monument during seismic activities and ambient vibrations generated by vehicular traffic and the adjacent Metro Line, triaxial velocimeters are installed on the foundation slab and the isolated podium. The collected data are analysed in accordance with the standards to determine the maximum response of the monument to various vibrations and to compare them with the thresholds established by the standards for short- and long-term monitoring. The results of data processing indicate that the vibration source significantly affects the dynamic response of the monument. Additionally, dominant frequency components of the vibrations due to earthquakes, subway transit and vehicular traffic are compared. Dynamic response at foundation level and isolated podium remains within the allowable limits for permanent and transient vibration caused by numerous earthquakes occurring recently due to the nearby Phlegraean Fields bradyseism activities. Finally, time sensitivity of the monument response to traffic-induced vibrations are presented.

Context / Content:
Civil infrastructure and heritage structures deteriorate with time due to environmental exposure, material aging, moisture ingress, pollution, and biological growth. Traditional inspection relies heavily on manual assessment, which is slow, risky, and subjective, especially in high‑rise or fragile heritage settings. The existing digital tools have concentrated mostly on 2D crack detection and have not provided depth estimation, biological segmentation, or integrated multi‑view analysis.
In order to overcome these limitations, the present work proposes an AI-driven system that can provide comprehensive structural health diagnostics from a single input structural image, using six parallel computer-vision pipelines with experimentally validated real-time inference performance.

Objectives:
- Automate crack detection with lightweight deep‑learning models
Identify and segment the biological growth that accelerates surface decay.
- Estimate depth variations to show severity and possible spalling.
- Provide integrated analysis across image processing and 3D heightmap generation
- Reduce unsafe manual inspections and support heritage preservation
- Scalable Structural Monitoring for Smart‑city and Cultural‑heritage Applications

Methods:
The system integrates several models and algorithms:
- Crack Detection → lightweight R-CNN optimized for 25 epochs
- Biological Growth Segmentation → U-Net–based pixel mask generation
- Depth Estimation → MiDaS monocular depth for surface profiling
- Material Analysis (Optional) → lightweight MobileNetV2 classifier
Edge Detection → Canny‑based structural contour extraction
- Integrated Analysis → Real-time processing across image analysis and 3D heightmap tabs

The pipeline runs on standard laptop CPUs without specialized hardware, sustaining ~14–23 FPS depending on the task, validating its suitability for real-time field inspection. It has three primary analysis tabs: Image Analysis for crack and vegetation detection, 3D Heightmap for depth visualization and surface profiling, and supporting analytics.

Results / Conclusions:
Tests run on 11,654 images, the system achieves real-time inference speeds of 0.0683 seconds per image for crack detection and 0.0424 seconds per image for segmentation on standard CPU hardware, enabling practical deployment without GPU dependency for concrete, brick, stone, and heritage materials demonstrate robust crack detection and strong segmentation performance for biological growth. Depth maps provide enhanced structural insight beyond traditional 2D inspection methods.
Principle integrated analysis allows for handling:
- Original Image
- Crack Detection
- Vegetation Segmentation
- Material/Surface Mask
- Depth Map
- Edge Detection

This significantly reduces the time taken for inspection and assists the engineers and conservation teams in the identification of defects even at their early stages. Although field deployment and full 3D reconstruction are beyond the scope of this phase, it forms a strong base toward automated, scalable, and data-driven structural health monitoring.

The Forum Women's Thermal Baths at Pompeii represent a unique case study for structural health monitoring of heritage masonry, combining exceptional archaeological significance with active seismic hazard (Seismic Zone 2, NTC2018). This paper presents a two-phase SHM strategy developed for this UNESCO World Heritage site. In the first phase, a finite element model of the thermal building was developed and validated, identifying fundamental frequencies of 6.00–9.00 Hz for the main structure and 80–120 Hz for the tegulae mammatae radiant wall system. These computational results informed the design of an ambient vibration testing campaign using high-sensitivity piezoelectric accelerometers, targeting both global structural modes and local facade dynamics across three sensor tiers (Fig. 1).

In the second phase, a permanent monitoring system based on triaxial MEMS accelerometers (noise floor < 3 μg/√Hz, Ethernet PoE, IP67) is being deployed to enable continuous structural surveillance. The paper discusses the methodological transition between the two phases, focusing on sensor selection criteria, frequency coverage compatibility, and the challenges imposed by heritage conservation constraints on spatial sensor placement. Preliminary OMA results are compared against FEM predictions, and the architecture of the permanent monitoring network is described. Finally, indications will be given in order to be able to subsequently use health monitoring techniques for future research on this archeological case study.