Prognostic Health Management (PHM) aims to predict the Remaining Useful Life (RUL) of degrading components and structural systems using monitoring data, forming the basis for maintenance optimization within a Predictive Maintenance (PdM) paradigm. In this context, this study presents an integrated Structural Health Monitoring (SHM) framework focused on Serviceability Limit State (SLS) assessment, with emphasis on real-time deflection control in reinforced concrete beams. The system employs a mid-span LVDT displacement sensor installed on a 3.0 m simply supported beam and connected to an IoT architecture with time-series database storage for continuous processing. To mitigate acquisition noise, a linear Kalman filter with a displacement–velocity state formulation is implemented, where the process covariance is heuristically adapted when the residual between model prediction and measured response exceeds a statistical threshold, allowing transient discrepancies to be absorbed without compromising physical consistency. Preliminary results indicate an approximate 50% reduction in signal standard deviation relative to raw measurements, yielding a more stable real-time displacement estimate. For short-term prognostics, a Temporal Convolutional Network (TCN) is employed to forecast displacement with a 24-hour horizon based on the filtered series, achieving a mean absolute error on the order of 0.4 mm and a normalized root mean square error below 6% under time-shifted validation. The filtered and projected states are integrated into a high-fidelity Digital Twin engine that reconstructs the full beam deflection profile using an Inverse Finite Element Method (iFEM) based on Euler-Bernoulli beam theory. Unlike static models, this engine implements a continuous model-updating scheme where the effective flexural rigidity (EI) is dynamically adjusted via an exponential degradation law to reflect accumulated structural damage. This allows for the evaluation of serviceability limit states (SLS) against the threshold for both current and 24-hour forecasted scenarios. The hybrid IoT–Kalman–TCN–Digital Twin architecture enables simultaneous monitoring of real-time and projected structural health, reducing false alarms by approximately 40% compared to raw signal thresholding, and providing a quantitative, physics-informed foundation for uncertainty-aware structural prognostics.

Monitoring of civil engineering structures and buildings is seeing increasingly widespread practical application.

Depending on the particular case, both long-term and short-term situative condition monitoring solutions are available.

Despite their prevalence, especially in structural steelwork, bolt connections are rarely monitored in practice. Still, in many cases periodic inspection is carried out, either through re-torquing or other methods of assessing the preload, such as a sounding hammer.

These methods have in common, that they require good physical access and skilled personnel to perform but are limited in their accuracy.

Combined with the fact that bolt preload can vary significantly upon installation due to uncertainties in friction conditions, digital force measurement is an attractive and reliable method to quickly and comprehensively assess the condition and performance of bolt connections.

Up until recently, digital bolt force measurement usually required an expert setup with costly hardware, complex data handling and in many cases a degradation or change in bolt performance due to the sensor integration.

With the fischer SensorBolt, an off-the-shelf bolt monitoring system is now available. By presenting several practical applications, the authors show how this system can help operators and owners of bolted installations to easily and reliably monitor the condition of their fixings, both immediately after installation and after an extended period of time.

To increase practical feasibility, a software solution complements the physical products and offers unified data and device management for projects of any scale. Alternatively, integration into existing systems is possible, to interpret the sensor data from bolts together with other SHM data.

The growing demand for sustainable energy solutions has driven research on converting wasted mechanical energy into usable electrical power. Among various renewable approaches, mechanical energy harvesting has shown significant potential for civil infrastructure, where continuous vibrations, loads, and dynamic movements are abundant. This study developed a carbon nanotube (CNT)-based polymer composite energy harvester specifically designed for large-scale civil structures such as bridges, pavements, and tunnels. The CNT/polymer composite was engineered to efficiently convert mechanical stress and strain energy into electrical signals through piezoelectric-like mechanisms enabled by the conductive CNT network. Unlike traditional piezoelectric ceramics, which are brittle and unsuitable for curved or irregular surfaces, the CNT-based composite exhibited high flexibility, strong adhesion, and easy processability into thin films. These characteristics made it suitable for structural applications requiring large-area coverage and conformal installation.
The research was conducted in three stages: (1) investigation of the fundamental mechanisms of mechanical-to-electrical conversion in CNT/polymer composites, (2) fabrication and optimization of CNT composite materials with controlled dispersion and interfacial bonding, and (3) experimental evaluation of energy-harvesting performance under cyclic mechanical loading. The composite was prepared using ultrasonic dispersion and casting techniques to ensure homogeneous CNT distribution within the polymer matrix. Electrical output was measured under compressive and vibrational loads using a precision electrometer to evaluate voltage, current, and power density. The developed composite achieved an energy generation density exceeding 1 mW/m², which was sufficient to power small electronic devices or low-energy sensors. A systematic parametric analysis was also performed to examine the effects of CNT content, polymer type, and structural geometry on electrical output efficiency. The results provided valuable insights into the balance between mechanical compliance, electrical conductivity, and durability of CNT-based composites under dynamic loading conditions. Overall, the CNT-based energy harvester demonstrated a novel and practical approach to recovering mechanical energy in civil engineering applications. By integrating lightweight, flexible, and conductive nanomaterials, the developed technology successfully utilized mechanical energy from ambient vibrations or traffic loads to generate usable electrical power. This study contributed to the advancement of sustainable civil engineering materials and provided a foundation for future developments in energy-harvesting infrastructure systems.

This paper presents an online adaptive framework for infrastructure health monitoring, leveraging an operational maglev vehicle as a mobile sensor. The objective is to develop a methodology capable of continuously tracking the physical parameters of the guideway in real time, particularly in the case of non-stationary dynamic conditions. The proposed methodology is based upon a state-space representation of the vehicle-track interaction. A recursive filtering approach, specifically an Extended Kalman Filter, is employed for the joint estimation of the dynamic states of the system and key parameters describing its physical health. The framework operates in two distinct phases: calibration, and online monitoring. In the calibration phase, the filter learns the baseline characteristics of a healthy system by fusing multi-source sensor data from instrumented guideway sections. Subsequently, in the online monitoring phase the calibrated filter operates using only vehicle-borne sensor data. The adaptive capability of the framework is validated on an experimental dataset that features two consecutive, physically identical low-stiffness sections: while the first induces an abrupt stiffness transition that leads to a severe transient impact, the second does not. The validation focuses on the ability of the framework to distinguish the aforementioned dynamic regimes and rapidly converge to the correct parameter estimate, despite the different dynamic responses captured by the measurements. This work therefore establishes a method to describe the dynamics of adaptive systems for infrastructure management, enabling a shift from periodic inspections to continuous, in-operation condition awareness.

This paper presents the main results obtained from field measurements carried out on a curved railway span composed of four large spans (45.0 m, 47.5 m, 95.0 m, and 57.5 m) between supports T2-133 and T2-137. The structure consists of steel I-girders, and several experimental tests were conducted, including Ambient Vibration Tests (AVT), Dynamic Load Tests (DLT), and Static or Pseudo-static Load Tests (SLT), using real train wagons loaded with sandbags. The objective was to assess and verify the structural safety of the original design.

To analyze the dynamic structural behavior of the selected span, vertical displacements at the decks, end beams, and bearings were recorded during train passages on both tracks at different speeds under design loading. Seventeen triaxial accelerometers were strategically positioned along and across the structure. The AVT provided the natural vibration periods (Ti) in the longitudinal, transverse, and vertical directions, as well as the predominant ground-motion period (Ts) at the site. The obtained fundamental vibration periods were consistent with the structural configuration.

Complementary topographic measurements were performed before, during, and after the DLT to determine deflection patterns under service loads. Analytical predictions from a three-dimensional finite-element model were compared with experimental data, showing good agreement. Both experimental and analytical results confirmed that the maximum vertical deflection occurred at mid-span of the 95.0 m main span.

Finally, the vertical deformation limits specified by national and international design codes (SICT–AASHTO) were verified and satisfied. Based on these results, the structural behavior of the analyzed curved railway span is considered adequate, and the relative vertical displacements in the superstructure are not significant.

The transport, energy and defense industries are currently facing the challenge of reducing maintenance costs while increasing structures availability. Today, structural inspections are carried out through periodic, systematic procedures that may lead to unnecessary ground immobilisation when no damage is ultimately detected. A key lever to address this limitation is to shift towards individualized maintenance, in which inspections are triggered only when required. Such an evolution presupposes that each structure is capable of reporting its structural health throughout its service life. Embedded Structural Health Monitoring (SHM) systems constitute a promising pathway to achieve this goal. However, despite extensive research and numerous prototypes, SHM solutions remain insufficiently mature for large-scale industrial deployment. One major obstacle lies in the difficulty of integrating sensors and associated systems (power supply, data transfer, optical or electrical linkages) within metallic and composite structures. In many current developments, sensors are conceived independently from their integration constraints, which are addressed only a posteriori. As a result, integration steps remain largely manual, intrusive, and highly dependent on operator expertise, preventing the repeatability, robustness, and scalability expected in industrial manufacturing chains.

The present work focuses on the development of non-intrusive integration strategies for SHM-related power and connectivity systems directly within manufacturing processes. The objective of the demonstrator initiated at IRT Jules Verne is to build technological and methodological competences enabling the functionalisation of structures during fabrication, as well as after assembly or during maintenance operations. The proposed approach aims to design and validate versatile, non-intrusive interconnection solutions, electrical, optical, or hybrid, that can be adapted to various sensors and manufacturing routes for both elementary parts and assembled components.

Insights gathered from several SHM workshops highlight strong industrial expectations toward solutions that are: non-intrusive; compatible with both in-process integration and end-of-line or repair operations; mature in terms of reliability, accuracy, durability, and measurement repeatability; easy to implement manually with robust procedures; and capable of being interrogated non-continuously in order to limit system complexity and operational costs.

Positioned at the interface between advanced manufacturing and structural monitoring technologies, IRT Jules Verne is developing processes and integration routes that seek to industrialise SHM embedding steps, transform current manual practices into controlled and repeatable operations, and address the persistent challenges associated with connectivity within structures. By reinforcing the manufacturability and durability of SHM integration, this work contributes to paving the way toward deployable, certifiable, and economically viable monitoring solutions for next-generation transport, energy and defense structures.

Wind turbines, being a source of renewable energy, are getting increased attention to meet the growing demand of green energy. The performance of a wind turbines can be significantly compromised due to the occurrence of various forms of damage within its components. Moreover, an undetected damage can grow rapidly over a period and may cause a failure of the entire wind turbine, resulting in a significant decrease in the output of a wind energy farm that may incur a sizable economic loss. However, the material anisotropy and curved surface of a composite wind turbine blade pose a significant challenge to apply a conventional time domain method for damage localization. To address this issue, a methodology has been developed that involves Hilbert Transform and correlation to analyse ultrasonic guided wave signals to localize and categorize damage in a composite wind turbine blade. The probing signal used is a Hanning window modulated 5- cycles sine signal. In t wind turbine his method, the targeted zone is discretized into several grid points which are assumed to be damage locations. The actual damage location is then identified through a correlation index of Hilbert transformed signals. The correlation index obtains the maximum value when the assumed damage location is a true damage location. Moreover, the results also show that the pattern in the Hilbert transform signal can be used for identifying the type of damage. An extensive numerical investigation has been performed considering various damage locations/types in the wind turbine blades to realize the effectiveness of the method. In the experiment, a few damage locations/types are considered. The method has localized and characterize damages with a high degree of accuracy in both the cases. Therefore, indicating a large-scale application of the method in the real wind energy systems for inspection of structural health and maintenance.

Structural Health Monitoring (SHM) systems strongly rely on informative experimental data to enable accurate identification of structural parameters, loads, and damage-related quantities under uncertainty. In this context, Bayesian Optimal Experimental Design (OED) provides a rigorous framework to determine sensor configurations that maximize the expected information gained from measurements. However, the practical application of Bayesian OED in SHM is often limited by the high computational cost associated with repeated evaluations of complex forward models and the estimation of the Shannon Expected Information Gain (SEIG).

This work aims to develop and demonstrate an efficient Bayesian OED strategy tailored for SHM applications, capable of significantly reducing computational costs while preserving accuracy. The main objective is to identify optimal sensor placements that maximize SEIG in problems involving uncertain mechanical properties and external loads, which are typical in SHM of civil structures.

The proposed methodology is based on a two-stage Kriging framework. In the first stage, a Kriging surrogate model is constructed to approximate the expensive forward structural model for fixed experimental designs. This surrogate is then embedded within a Double Loop Monte Carlo (DLMC) scheme to efficiently estimate the SEIG without repeated calls to the original model. In the second stage, Efficient Global Optimization (EGO) is employed to maximize the SEIG over the design space. This requires building a second Kriging surrogate, now of the SEIG itself, and using the expected improvement criterion to balance exploration and exploitation during the optimization process. By decoupling the surrogate construction in the parameter space and the design space, the proposed approach alleviates the curse of dimensionality commonly encountered in surrogate-based OED.

The methodology is evaluated through two numerical examples with increasing complexity, including nonlinear benchmark problems and SHM-inspired beam models based on Timoshenko theory. The SHM case studies address optimal strain gauge placement for identifying elastic properties and for characterizing multiple unknown loads acting on a structure. Results show that the proposed two-stage Kriging approach achieves SEIG values comparable to reference methods based on direct DLMC, while reducing the number of forward model evaluations by several orders of magnitude. In more complex SHM scenarios, the method outperforms state-of-the-art surrogate-based approaches, providing higher information gain with substantially lower computational effort.

These results demonstrate that the proposed framework is a robust and efficient tool for Bayesian OED in SHM, enabling optimal sensor placement in realistic structural problems where computational cost is a major constraint.

Guided waves are widely used for damage detection in composite structures due to their high sensitivity to defects and long propagation distances. However, strong material attenuation in composites limits the effective inspection range. This study employs multi point excitation over large and complex structures and introduces a time varying spatial filter to track S₀ mode and its convertion to A₀ mode, and mitigate mutual interference among sources. The filter achieves single mode isolation under multi point actuation and performs attenuation compensation to extend the inspection domain. Clean damage signatures are then extracted and used for imaging. The proposed method is validated through experimental tests on aircraft CFRP stiffened panel with barely visible impact damage, demonstrating wide area coverage, and improved detection reliability.

Detecting concealed detachments and localized adhesion failures in historical plasters—especially in stratified or layered architectural surfaces—remains a central challenge in the non-invasive diagnostics of cultural heritage. Conventional auscultation, relying on manual tapping and auditory assessment, is fast and intuitive but inherently subjective, leading to inconsistencies in acoustic classification and, consequently, diagnostic reproducibility.
The advent of instrumented acoustic probes has facilitated the objective collection of impact response data, capturing both time-domain and frequency-domain features suitable for algorithmic interpretation. Yet, applying supervised learning in situ is limited by the persistent absence of richly labeled and spatially resolved ground truth data, a common constraint in real-world conservation contexts.

To address this, the present study proposes a novel methodology that combines minimal supervised input with temporal and spectral feature analysis to allow reliable and repeatable adhesion classification. The central innovation is a two-point training approach, wherein the neural network is trained using data from just two manually selected reference sites within the target area—one firmly adhered (label 0), the other clearly detached (label 1). Each site provides repeated impact responses using the PICUS probe, a handheld probe that can combine accelerometer and microphone sampled signals, establishing the acoustic boundary conditions for classification. The trained model then processes new impact signals acquired through a systematic surface scan, classifying each according to its proximity to the known states. The outputs are spatially interpolated to generate adhesion maps delineating intact and detached zones.

This localized, two-point strategy introduces a flexible AI-driven diagnostic model that adapts to site-specific conditions, bridging the gap between controlled laboratory datasets and variable field environments. It enables each survey area to function as its own calibration space, ensuring context-sensitive reliability without the need for large-scale labeling or external reference models.

To enhance interpretability and improve model generalization, each impact record from the reference datasets is analyzed through the extraction of a compact suite of spectral and temporal features that characterize the mechanical and acoustic behavior of the material.

These physical descriptors serve as input to a convolutional neural network (CNN). The network effectively discriminates between the two acoustic states, demonstrating robust performance across different materials such as lime and pozzolanic based mortars like in heritage structures or nowadays synthetic materials. The model is further extendable via transfer learning to unlabeled datasets, allowing autonomous classification and mapping where ground truth is unattainable—a critical capability for practical conservation work.

The approach proves robust against sensor and environmental variability, as the learned features are grounded in spectral signatures rather than absolute amplitude values.

This approach offers a practical and scalable solution for on-site, data-driven diagnostics, reducing dependence on expert subjectivity while maintaining physical interpretation through feature-based learning. Future developments will focus on extending the method toward multi-class classification, integrating transitional adhesion states, refining spatial interpolation algorithms for high-resolution adhesion mapping and lightweight deployment compatible with TinyML environments.

This paper investigates processing methods for Lamb wave signals intended for wireless structural health monitoring systems. In such systems, data acquisition is limited by reduced sampling rate compared to standard wired configurations, which introduces challenges in accurately extracting damage-sensitive features. Conventional approaches based on time-of-flight (ToF) delay estimation may become unreliable under these conditions due to waveform distortion and insufficient temporal resolution. Therefore, alternative processing strategies are required to ensure accurate damage detection and localization. In this study, the feasibility of reconstructing the original high-resolution Lamb wave signal from low-sampling rate wireless measurements is evaluated, enabling the continued use of ToF-based damage detection. Different signal-reconstruction techniques are examined and compared in terms of accuracy, robustness and computational efficiency. The results demonstrate that signal reconstruction significantly improves ToF estimation reliability, making it a promising solution for future wireless Lamb-wave-based monitoring platforms.

Filament-wound glass fiber reinforced polymer (GFRP) vessels are widely deployed in water management systems for the collection, separation, and treatment of stormwater and industrial effluents. During installation as well as long-term operation, these composite structures may be exposed to complex loading scenarios, including localized stress concentrations that can initiate damage and subsequently compromise structural integrity. Reliable structural health monitoring (SHM) methods are therefore essential for detecting strain concentrations and early-stage damage in order to ensure long-term operational safety. Carbon Fiber Sensors (CFS), which utilize changes in electrical resistance to quantify mechanical strain, represent a promising low-cost sensing technology suitable for direct integration into composite structures, see (1). This contribution extends previously published work (2) focused on the strain measurement of GFRP plate-type specimens with a central hole subjected to four-point bending. In the earlier study, CFS and distributed fiber optic sensing (DFOS) were used to evaluate strain distribution and validate sensor performance, with the results showing good agreement between both measurement techniques and finite element method (FEM) simulations. Building on these findings, the present study investigates the sensitivity of CFS sensors to artificially introduced damage in the form of controlled notches of increasing depth representing crack-like defects in the GFRP plates.

New experiments are conducted on filament-wound GFRP specimens instrumented with surface-mounted CFS and DFOS sensors. Each specimen is subjected to repeated four-point bending cycles for multiple predefined damage levels. The progressive increase in notch depth enables a systematic evaluation of the relationship between structural response and sensor output. The resulting data shows the determination of functional dependencies between specimen deflection, local strain, CFS resistance change, and crack depth. Particular attention is given to the repeatability of measurements, the detectability threshold of damage, and the influence of sensor placement relative to the defect.

Numerical simulations of the experiment are performed using a detailed shell-based FE model consistent with the geometry and material properties of the tested specimens. Strain values are extracted along the paths corresponding to sensor locations and compared with both DFOS and CFS measurements. This comparison provides insight into the accuracy, limitations, and potential application of CFS technology for detecting and quantifying damage in GFRP composites.

The experimental and numerical findings confirm that carbon fiber sensors possess adequate sensitivity not only for monitoring global deformation but also for detecting localized damage, such as crack-like defects. The clear correlation between notch depth and electrical resistance response shows their potential for tracking damage evolution in filament-wound GFRP structures.

[1] Horoschenkoff, A., & Christner, C. (2012). Carbon Fibre Sensor: Theory and Application. In N. Hu (Ed.), Composites and Their Applications. InTech. https://doi.org/10.5772/50504

[2] Schmidová, N., Doubrava, K., Padovec, Z., Blaha, D., Novotný, C., & Růžička, M. (2024). Bending Analysis of GFRP Composite Plate with a Hole Using Carbon Fiber Sensors. In 62nd Conference on Experimental Stress analysis - Book of Extended Abstracts. CTU.

Inflatable structures are promising for application in large spacecraft due to their lightweight nature, low cost, and high deployment reliability. However, the flexible textile composites used in such structures are vulnerable to hyper-velocity impacts from micrometeoroids and orbital debris (MMOD) during service, which poses a serious threat to both spacecraft safety and astronaut lives. To address this issue, this paper proposes an integrated sensing approach in which a carbon nanotube (CNT) sensing layer is embedded into the flexible textile composite, combined with electrical impedance tomography (EIT), to enable continuous online monitoring of MMOD hyper-velocity impacts. This allows for timely detection of impact events and identification of damage location and approximate size. Firstly, CNT films synthesized via floating catalytic chemical vapor deposition (FCCVD) and associated circuitry are integrated into the textile composite using hot-press lamination, endowing the composite with self-sensing functionality. Then, low-amplitude currents are periodically injected into the CNT sensing layer, while boundary voltage data are recorded. The occurrence of a hyper-velocity impact event is detected by monitoring changes in the boundary voltage. Finally, the EIT algorithm is then employed to reconstruct the conductivity distribution within the sensing layer, visualizing variations caused by impact damage. This process yields image-based information on the location and approximate size of the damage. To validate the method, flexible self-sensing textile composite specimens are fabricated and subjected to high-velocity projectile tests to produce damage similar to that caused by MMOD impacts. Experimental results are used to confirm the feasibility and effectiveness of the proposed technique.

Structural Health Monitoring (SHM) relies on the deployment of permanently installed sensors over long periods (e.g., several years). However, while these devices enable the monitoring of a structure, a key question remains: how can the sensors themselves be monitored over time? Existing studies addressing this issue often focus on a specific subsystem of the sensor (e.g., piezoelectric element, battery, etc.), leading to a fragmented view of sensor reliability.

Before considering self-monitoring or self-diagnostic strategies, it is necessary to provide a conceptual clarification of what constitutes a wireless SHM sensor. It is also essential to identify the main drift and failure mechanisms affecting its various features, including electronics, data acquisition, communication, and power management.

This work provides a state-of-the-art overview based on a review of the scientific literature, complemented by industrial feedback related to failure mechanisms and metrological drift. The objective is to establish the current state of knowledge, identify the approaches already explored, whether in SHM or in related fields, and highlight the remaining gaps.

The analysis reveals the major scientific and technological challenges that must be addressed to design smarter sensors able of assessing their own health status and improving the long-term reliability of SHM instrumentations.

Spaceborne Synthetic Aperture Radar (SAR) is a non-contact remote sensing technology that detects surface deformation by analyzing the phase differences between radar images acquired over the same area at different times. Due to its extensive coverage, high spatial resolution, and all-weather operational capability, spaceborne SAR has become an established technique for large-scale, continuous monitoring of civil infrastructure. Transportation networks constitute a fundamental component of urban infrastructure, playing a pivotal role in enabling efficient mobility and fostering regional economic development. Extreme weather events severely threaten the durability and operational safety of transportation networks. However, limited funding restricts the deployment of traditional sensors for detailed and comprehensive monitoring of the entire transportation network system. In this research, a stack of Sentinel SAR images acquired over a two-years period is collected from the Copernicus Data Space Ecosystem, and subsequently processed with Persistent Scatterer Interferometric Synthetic Aperture Radar (PS-InSAR) technology. A dedicated post-processing procedure, consisting of PS points refinement and clustering analysis, is applied to the displacement time series derived from the PS-InSAR processing. Then, statistical control limits method is employed to evaluate the risk levels across transportation network. Finally, the reliability and effectiveness of the proposed risk assessment framework are validated through a specific bridge case study. These findings demonstrate the potential of the proposed framework for large-scale, risk-informed assessment of transportation networks, thereby contributing to more proactive and data-driven transportation network management strategies.

A wide variety of metallic structures are extensively used in aircraft. However, these metallic structures inevitably develop fatigue cracks during long-term service, which seriously threatens structural safety. With the advancement of structural health monitoring and digital twin technologies, promptly detecting the initiation and growth of cracks and making informed decisions are of great significance for ensuring flight safety. This paper presents an experimental study on the online monitoring of crack growth in metallic structures under fatigue loading using a piezoelectric sensor array and ultrasonic guided wave technology. A center-cracked aluminium alloy plate is prepared, and a network of piezoelectric wafer sensors is deployed on its surface. During the fatigue test, while observing crack growth, the piezoelectric wafers were used to excite and receive ultrasonic guided wave signals respectively, monitoring the fatigue crack growth behaviour between the actuator and sensor pairs. Through time-domain signal processing, original guided wave signals and energy characteristics of wave packets post-Hilbert transform are extracted. An energy damage index (EDI) was defined to evaluate its variation with crack growth. Finally, by integrating EDIs obtained from all actuator-sensor paths in the piezoelectric sensor network using a damage probability imaging method, the change in crack length is visually characterized in the form of images. Experimental results have verified the feasibility and effectiveness of the proposed method.

Structural health monitoring (SHM) systems play a crucial role in ensuring the safety of civil structures under natural haz ards and ambient conditions. In general, SHM can be conducted using modal-based or data-driven approaches. The former estimates physical parameters of a structure from its vibrational response but struggles with computational challenges and model uncertainties [1]. In contrast, data-driven approaches have gained popularity for their independence from structural physics. While machine learning and statistical methods have been widely applied, deep learning algorithms demonstrate superior feature extraction, particularly in complex systems. However, most SHM systems rely on comparison-based algo rithms with an intact baseline, often unavailable in real-world applications. To address this, we propose a baseline-free SHM framework based on a deep learning–driven unsupervised anomaly detection algorithm that utilizes ambient vibration signals. We propose a generative model, known as a variational autoencoder (VAE), a probabilistic version of a deterministic autoencoder (AE). The VAE belongs to Bayesian machine learning, and is based on the concept of variational inference [2]. It consists of an encoder and a decoder that work jointly to learn the data distribution through an encoding–decoding strategy. The encoder approximates a posterior distribution qϕ(z|x), while the decoder approximates a likelihood distribution pθ(x|z). The encoder compresses high-dimensional input data into a low-dimensional latent representation by enforcing the posterior distribution to approach a prior distribution p(z), typically chosen as a standard Gaussian. In contrast, the decoder reconstructs input data from these latent representations [3]. VAEs are trained by maximizing the Evidence Lower Bound (ELBO), which consists of a reconstruction term and a regularization term. Accordingly, the VAE loss function is written as [4]: L(θ,ϕ|x) = Eq(z|x,ϕ)[logp(x|z,θ)] − DKL(q(z|x,ϕ)||p(z)) (1) where the former term is the log-likelihood of the data, equivalent to the Mean Square Error (MSE) in practice. The latter term is the Kullback-Leibler (KL) regularizer which penalizes the posterior distribution from diverging from the prior distri bution. To this end, we propose a baseline-free VAE for SHM systems, coupled with a damage index algorithm based on latent representations. The anomaly score is computed as the Euclidean distance in the latent space between extracted features and a healthy reference. In the absence of baseline data, features with the lowest 5–10% reconstruction errors are assumed to represent healthy conditions. The resulting latent-based damage index forms a time-series signal, where distinct spikes indicate potential structural damage. A schematic of the proposed SHM framework is shown in Figure 1.The proposed algorithm is evaluated using acceleration signals from the IASC–ASCE Structural Health Monitoring Task Group Phase II benchmark experiment [5]. The experiment involves a four-story, two-bay-by-two-bay steel frame structure at the University of British Columbia subjected to various excitations. In this study, only the ambient vibration case is considered, with structural damage simulated by removing diagonal braces. Several damage patterns are analyzed to demonstrate the algorithm’s effectiveness under different scenarios. The results confirm the approach’s capability to detect structural damage through distinct spikes in the latent-based damage index, with multiple spikes observed in more severe damage cases, highlighting the approach’s sensitivity.

In 1461, Sultan Mehmed the Conqueror laid the foundations of the Tersane-i Amire in Istanbul's Golden Horn. The shipyard was continuously expanded by successive Ottoman sultans over the centuries with the addition of slipways, warehouses, a torpedo facility, and a foundry for propellers. It also developed an ecosystem with a range of administrative and public buildings, including a hospital, hammam, school, and mosque. Tersane-i Amire remained in operation until the 1970s. Afterwards, it was partially moved and partially closed, with its historic slipways and buildings protected as a part of the city’s industrial heritage. In 2019, a project was initiated to transform the whole complex into a tourism and lifestyle destination by restoring the historical slipways, workshops, and buildings while preserving their original form.

This study presents the results of two sets of ambient vibration surveys (AVS) applied to one of the magnificent historic buildings re-functioned as an event hall in the complex (Figure 1). The masonry building has been retrofitted by the addition of steel frames located within the building. The conducted AVSs aimed to reveal the effects of the applied retrofitting. Thus, the first AVS was applied before the connection of the steel frames to the original masonry walls, thereby determining the dynamic properties of the original building. The second AVS was performed after connecting the steel frames to the original system and determined the dynamic properties of the retrofitted building. The comparison of the obtained dynamic properties for the two forms of the building proved the efficiency of the applied retrofitting on the dynamic properties of the building.

This paper presents a field-based comparison between conventional accelerometers and distributed acoustic sensing (DAS) for vibration monitoring of a shotcrete-lined railway tunnel located in northern Brazil (Pará State). Approximately 40 m of single-mode optical fiber were installed along the tunnel lining, while three triaxial piezoelectric accelerometers were placed at locations corresponding to selected virtual channels of the fiber and used as reference sensors, with acceleration components aligned with the fiber direction considered for comparison purposes. Vibration data were acquired simultaneously by the DAS interrogator and a data acquisition system connected to the accelerometers during multiple train passages. Following data acquisition, the recorded signals were organized into individual train-passage events and preprocessed to reduce noise and remove outliers, ensuring a fair basis for comparison. The DAS measurements captured the tunnel wall dynamic response, exhibiting time-domain patterns consistent with those recorded by the accelerometers; however, differences in signal duration were observed due to the distinct sensing principles of the two technologies and data acquisition specifics of each system. To enable a quantitative comparison, representative frequency-domain features were extracted from both datasets, including spectral entropy, spectral centroid, and relative power within specified frequency bands, and were used to evaluate the consistency and stability of the representation of significant vibration characteristics across both sensing technologies. The results indicate that DAS provides a stable and spatially continuous representation of the tunnel’s dynamic behavior, highlighting its potential to complement or, in some applications, replace conventional accelerometers in structural health monitoring (SHM) of underground infrastructure.

This paper introduces a integrates compressive sensing with a physics-guided reconstruction method for high-speed full-field laser scanning–based structural damage detection. Conventional full-field laser scanning requires high-resolution spatial sampling, which leads to excessive measurement time and computational cost in large-scale inspections. To address this limitation, the proposed approach employs pattern-based compressive ultrasonic scanning, in which the inspection area is partially sampled using two-dimensional random patterns, and steady-state responses are efficiently acquired using a Laser Doppler vibrometer (LDV) under single-frequency excitation. The sparsely measured data are then reconstructed into full-field wave responses using a neural network–based method that incorporates physical constraints governing wave propagation. By simultaneously leveraging the physical wave model and the measured data, the proposed reconstruction scheme is designed to recover spatially coherent wavefields even under high compression ratios. The reconstructed wavefield is subsequently analyzed using wavenumber-based damage visualization techniques to estimate the location and size of structural damage. The proposed framework was validated through experiments on aluminum and composite plates with various types of damage, including corrosion and delamination. The results demonstrate that the combination of compressive sensing and physics-guided reconstruction yields higher accuracy and improved physical consistency compared to conventional methods, significantly enhancing the efficiency of full-field ultrasonic damage detection and visualization.

Out-of-Autoclave (OoA) techniques are emerging as sustainable alternatives to the traditional autoclave-based production of high-value composite structures for safety-critical applications. Although they provide the controlled environment required to minimise porosity defects and ensure structural soundness, the operation of autoclaves is expensive, energy-intensive, and limits the size of the final component. On the other hand, OoA methods, particularly resin infusion of dry fibres, followed by cure using mould-integrated heating systems, offer substantial reductions in operating costs and more flexibility in terms of component sizing. However, in the absence of the high consolidation pressures associated with autoclaves, the likelihood of porosity defects increases, restricting the adoption of OoA in high-performance sectors. This limitation requires reliable in-situ monitoring of the resin flow to identify deviations in the process at an early stage.

This study proposes an Ultrasonic Guided-Wave (UGW) sensing array integrated into the upper lid of an infusion mould as a non-intrusive method for mapping the progression of the resin in real time. The main objective is to establish a robust and scalable approach that can capture the fluid distribution over large areas. To achieve this, piezoelectric ultrasonic transducers are used to excite guided waves through the lid and interact with the advancing resin. Through combined theoretical, simulation-based, and experimental observations, the fundamental antisymmetric mode is identified as optimal for liquid propagation monitoring due to its amplitude variation with respect to the proportion of the solid medium covered by fluid.

To evaluate and validate the method, a custom experimental setup is developed to allow the collection of repeatable fluid infusions within a mould, under controlled flow conditions. Time stamped ground-truth data for resin-front location are obtained using a machine-vision system, enabling direct correlation with the collected UGW data. Sensor positioning is optimised to balance measurement sensitivity with increased coverage of the monitoring region, ensuring that the results remain accurate with a sparse sensor network.

The collected dataset is used to assess several predictive modelling strategies. Firstly, an analytical model based on functional approximations of the evolution of the UGW energy provides a baseline prediction accuracy with a best error of approximately 10% of the distance between the sensors. Subsequently, a machine learning model reduces the best error to 5%, demonstrating the potential of the method, even when the training set is limited in size. Lastly, to address data scarcity for industrial deployment, a high-fidelity COMSOL digital twin is used to generate synthetic UGW data. This data is used to augment the training experimental datasets and domain-adaptation techniques are explored.

This research demonstrates that UGW could be a tool for reliable resin infusion front tracking in OoA processes, offering a practical method that could prevent the formation of void defects during manufacturing. Further validation on larger parts, with more complex geometries and different infusion conditions, is required to ensure robust industrial deployment.

Accurate condition assessment is a major challenge in Bridge Health Monitoring (BHM). Traditional vibration-based BHM systems based on accelerometers are relatively costly to install and hard to maintain in long-term condition monitoring projects. Meanwhile, those discrete signals are hard to interpret and assess in terms of the spatial dynamic response. Recently, the Distributed Acoustic Sensing (DAS) technology has offered new possibilities for BHM. DAS offers direct and continuous measurement of vibration data to capture tiny perturbations, which can be used to turn an optical fiber into a distributed vibration sensor. Such a high-accuracy and resolution acquisition system enables early damage detection and damage localization. In this study, we conduct an in-situ test on a two-span concrete bridge. During the test, the dynamic responses are extracted along the optical fiber and measured with accelerometers. We utilize Empirical Mode Decomposition (EMD) for accelerometers and Stochastic Subspace Identification (SSI) algorithms for DAS signals to determine the natural frequencies, damping ratio, and modal shapes for the bridge. We also investigate the possibility of fusing multisensory data to assess the spatial and temporal variability of bridge dynamic parameters.

Although significant progress has been made in ultrasonic guided wavefield damage imaging, its engineering application remains constrained by the requirement for high-density sampling. Compress sensing technology can break through the limitation of Nyquist sampling law and reconstruct data from a small amount of data samples. Based on compress sensing technology, this paper proposes a guided wavefield damage imaging method via two-dimensional multi-frequency sparse wavenumber spectrum reconstruction. Firstly, the sparse sampling method is selected to ensure the uniformity and randomness of the sparse sampling process, the pulse laser is used for point-by-point excitation and the sparse wavefield is obtained by the piezoelectric ceramic sensor. Then, the dictionary matrix is constructed based on the two-dimensional ultrasonic guided waves propagation physical model, and the two-dimensional orthogonal matching pursuit (2D-OMP) algorithm is used to reconstruct the complete multi-frequency wavefield signal from a small number of known wavefield data. Based on the reconstructed multi-frequency wavefield data, the multi-frequency two-dimensional wavenumber spectrum is calculated. Finally, the frequency wavenumber filtering method is used to extract the damage information in the multi-frequency spectrum, and the filtered multi-frequency spectrum is subjected to root mean square processing to achieve defect imaging. The proposed method is verified in simulation and experiment respectively. The experimental and simulation results show that the method can better realize the reconstruction of frequency wavenumber spectrum and damage imaging in aluminum plate.

Long-span cable-supported bridges are highly susceptible to windstorm-induced excitations due to their flexible structures, low damping ratios, and large aerodynamic surfaces. Extreme wind events can trigger excessive vibrations, loss of serviceability, and even structural instability, posing serious risks to both safety and traffic operations. Timely bridge closure during severe windstorms is therefore critical to prevent accidents and ensure operational resilience. However, manual decision-making may suffer from delays and inaccuracies under rapidly evolving windstorm conditions. To address this challenge, this study proposes an automated AI-aided alert framework for bridge closure during windstorms based on the concept of unsupervised anomaly detection. Building upon this, an unsupervised anomaly detector is trained by solely using acceleration time histories recorded under normal wind conditions. In this case, the developed anomaly detector can autonomously learn the baseline dynamic behaviour of the bridge being monitored. When windstorm-induced vibrations occur, the trained anomaly detector identifies deviations from the learned normal patterns in real time based on a decision threshold and triggers automated closure alerts. The proposed method is validated using measured acceleration responses measured from a long-span cable-stayed bridge under both normal and windstorm conditions. Results demonstrate that the proposed model can effectively detects windstorm-induced anomalies with high sensitivity and low false-alarm rates, enabling timely and reliable bridge closure decisions.

Light-frame timber structures have become increasingly popular due to their low cost and ease of construction, but their susceptibility to damage demands efficient vibration-based structural health monitoring (SHM) strategies. However, monitoring these lightweight elements is challenging, as capturing localized damage and distinguishing it from environmental or operational variability remain difficult. This study proposes a data-driven SHM methodology within the Statistical Pattern Recognition (SPR) framework that integrates feature optimization and machine learning to classify and quantify damage in timber beams. In an initial experimental phase, ambient and impact vibration tests have been conducted on a full-scale laboratory timber beam in its undamaged state and under multiple simulated damage scenarios introduced by attaching small masses at predefined locations. System identification was used to extract modal parameters (natural frequencies and mode shapes) from these tests, providing a basis for numerous damage-sensitive features. In the next phase, both global features (e.g., shifts in natural frequency and modal assurance criterion) and local features (e.g., mode shape curvature, modal strain energy, and coordinate modal assurance criterion) will be derived and evaluated for their sensitivity to damage. A wrapper-based feature selection technique will then be employed to identify an optimal subset of these features. Using this optimized feature set, a neural network-based surrogate model will be trained to perform multi-output damage identification – classifying the presence of damage, localizing it along the beam, and estimating its severity. It is expected that local modal features will exhibit higher sensitivity to the mass-induced damage than global indicators, and that the surrogate model will provide robust and interpretable predictions under varying conditions. Fig. 1 shows the overall workflow of the study. By combining conventional modal analysis with advanced feature selection and machine learning, the proposed approach is anticipated to deliver a practical, low-cost, and interpretable workflow for early damage detection in timber members, ultimately supporting the broader implementation of SHM in light-frame structural systems.

Vibrational monitoring of structures traditionally relies on contact sensors such as accelerometers, displacement transducers, and strain gauges, which provide reliable physical information for structural health assessment. However, these sensors require manual installation and direct access to the structure, resulting in practical limitations in terms of installation time, safety, and accessibility, while long-term maintenance may compromise the overall Structural Health Monitoring (SHM) system reliability. These constraints often hinder the systematic implementation of SHM, particularly for large-scale infrastructures such as bridges. Alternative approaches present complementary limitations: vision-based techniques, such as digital image correlation, are sensitive to lighting conditions, camera calibration, and line-of-sight occlusions, whereas fibre-optic sensing systems, despite their high accuracy and distributed capabilities, require permanent installation and physical integration within the structure, limiting their suitability for rapid or temporary monitoring campaigns.

To address these challenges, Ground-Based Interferometric RADAR (GB-InRA) technology has emerged as a promising non-contact alternative for vibration monitoring, enabling safe measurements where visual inspection or contact sensor deployment are impractical. Interferometric radar enables the detection of sub-millimetric displacements by measuring the phase difference between transmitted and received signals. In this context, Real Aperture Radar (RAR), which provides one-dimensional line-of-sight measurements, is typically preferred for vibration monitoring over Synthetic Aperture Radar (SAR), which enables two-dimensional imaging. Nevertheless, radar-based measurements remain sensitive to instrument-to-target distance and antenna tilt, which must be carefully calibrated to ensure reliable results.

In the perspective of bridge monitoring, this work presents an experimental study conducted on a simply supported flexible steel beam (Figure 1). Vibrations were measured using an IBIS-FS microwave interferometric radar and compared against conventional displacement transducers adopted as ground truth. Three corner reflectors were installed at distinct beam locations to enable the estimation of the first three vibration mode frequencies and shapes. Impulsive hammer excitations induced sub-millimetric displacements, allowing a quantitative assessment of radar sensitivity and measurement accuracy. Preliminary results indicate displacement amplitude errors ranging between 0.02 mm and 0.2 mm for reference amplitudes between 0.5 mm and 12 mm, when compared to displacement transducers. Different configurations of vertical tilt and radar-to-target distance were also investigated to evaluate their influence on displacement estimation and modal identification. Overall, the findings demonstrate the effectiveness and practical applicability of GB-InRA as a robust, non-invasive tool for vibration monitoring and modal identification of civil structures.

Accurate reconstruction of complex structural deformation is essential for understanding transient mechanical responses in hybrid and multi-material systems. This study presents a vision-based approach for three-dimensional deformation reconstruction of a composite–wood hybrid structure subjected to transient load release. The investigated specimen consists of a U-shaped carbon-fiber-reinforced composite beam connected perpendicularly to a wooden cantilever beam, forming a configuration that exhibits strong coupling between bending and torsion during load removal.

A single-camera imaging system was employed to capture the structural response throughout the release process. Surface motion was extracted from image sequences using feature tracking and temporal filtering techniques. Based on the known structural geometry and camera calibration parameters, the three-dimensional deformation field was recovered from single-view data. This method enables non-contact and full-field reconstruction without requiring embedded or attached sensors, making it well suited for hybrid or geometrically irregular structures.

Preliminary experiments demonstrate that the proposed approach can effectively capture the evolution of deformation and reveal coupled twisting and bending behaviors during the transient release event. The reconstructed deformation patterns qualitatively agree with visual observations, highlighting the capability of single-view vision methods to characterize dynamic structural responses.

The proposed framework provides a practical and low-cost solution for dynamic deformation monitoring of complex and multi-material structures. It underscores the potential of vision-based sensing as a valuable complement to conventional techniques in structural health monitoring applications where sensor deployment or accessibility is limited.

Composite structures are widely used in the aerospace industry as important load-carrying components due to their excellent performance. However, composite structures are susceptible to impact damage and manufacturing flaws, compelling the designers and airline operators to ensure continuous safety. Lamb wave-based techniques are gaining immense popularity for structural health monitoring (SHM) applications to detect internal and surface damage in thin-walled structures, such as aerospace structures, using embedded and surface-bonded transducers. The damage detection using Lamb waves conventionally relies on a comparison of the damage features of the current response with the baseline response previously obtained from the same structure in its pristine condition. However, the same Lamb wave features vary by a change in the operating temperature of the structure, leading to possible false alarms for damage. To overcome this problem, baseline-free damage detection methods have been pursued in the recent past. Among various techniques, the one based on the time-reversal process (TRP) of Lamb waves has emerged as the most promising candidate for damage detection in thin-walled structures. For a system of sensor arrays to be used for SHM following TRP of Lamb waves, it is essential that a single frequency should be used to probe the structure along all the paths of the sensor array and the time reversibility in an undamaged structure should be the same for all the paths. However, to the best of the author's knowledge, no study has reported how the layup configuration of a composite structure affects the time reversibility of Lamb waves in different paths of a sensor array.

The present work aims to numerically study the effect of the layup configuration of the composite structure on the time reversibility of the Lamb waves. The numerical study is conducted using three-dimensional (3D) finite element (FE) simulation in CIVA, which uses a transient spectral finite element. The simulation is conducted on two configurations of composites, namely, quasi-isotropic and cross-ply. A composite plate of 2 mm thickness with 16 laminae of 0.125 mm thickness each is considered. Five transducer locations are considered on each plate in such a way that one transducer (T1) is located at the center and others (T2, T3, T4, and T5) are located on a quarter circle of radius 200 mm. The transducer system constitutes four paths T1-T2, T1-T3, T1-T4, T1-T5 which makes 0°, 22.5°, 45°, and 90° to the global x-axis of the system. The simulation is conducted by first actuating the transducer T1 with a 5-cycle tone burst signal and emitting back the time-reversed version of the response obtained at T2. Then, the normalized version of the reconstructed signal obtained at T1 is compared with the normalized version of the input signal to quantify the degree of the Lamb wave reconstruction. The simulation is being repeated for all the paths and for both plate configurations. The result shows that while in the case of a quasi-isotropic configuration, the variation of the quality of time reversibility obtained for different paths is minimal. Its variation in the case of cross-ply laminate is significant.

Lamb waves are widely used in structural health monitoring (SHM) of composite materials due to their sensitivity to defects and boundaries. In carbon fiber reinforced polymers (CFRP), complex anisotropic properties pose challenges for accurate wave interpretation and modeling.

This work aims to investigate the propagation and reflection behavior of Lamb waves in CFRP using steady-state excitation as an alternative to conventional burst signals. The central question is how phase and frequency shifts in the reflected waveforms can be analyzed to enhance boundary interaction modeling.

Signal-based methods are employed, using continuous sinusoidal excitation to generate steady-state Lamb waves. Phase and frequency variations in the reflected signals were measured using a low-cost modular Field Programmable Gate Array measurement system. The measurement data are analyzed using frequency and time-frequency representations and transfer function models.

The results reveal that steady-state excitation enables high-resolution characterization of phase shifts at CFRP boundaries, which are difficult to resolve with transient signals. Frequency-dependent reflection behavior is observed, demonstrating the sensitivity of steady-state Lamb waves to local boundary conditions.

From the findings it can be concluded that steady-state excitation offers a robust alternative for studying wave–boundary interactions in anisotropic materials, with the potential to improve SHM techniques and modeling approaches. The method opens new pathways for high-precision wave analysis and encourages further research into frequency-selective boundary characterization in composites.

Monitoring the dynamic behavior and long-term stability of railway tunnels is essential for ensuring structural safety and efficient maintenance. This study presents the testing of a cost-effective prototype based on the ADXL355 accelerometer for measuring both vibration and inclination in tunnel structures. The proposed system aims to provide a compact solution suitable for long-term Structural Health Monitoring (SHM) applications. Laboratory experiments were conducted, and field tests were carried out in a real tunnel environment to assess the prototype’s performance under controlled and operational conditions, including train-induced vibrations and ambient noise. The field data confirmed the system’s capability to capture both transient and quasi-static behaviors relevant to tunnel performance assessment. This work supports the feasibility of implementing cost-effective, sensor-based monitoring systems for continuous assessment of tunnel stability, contributing to the development of IoT-enabled SHM solutions for critical underground infrastructure.

In the context of corrosion detection and related material degradation follow-up over time, there is growing interest in replacing classical inspection solutions with an SHM system, reducing maintenance costs, human factors, and risk of accidents. This publication presents the results of development initially focused on the detection and the developmental measurement of corrosion in reinforced concrete structures. In addition, it has shown capabilities in the field of material degradation follow-up due to liquid ingress into composite or any other complex materials.

We present an SHM solution and results with high sensitivity enabling reliable diagnostics. For the selected use case, the principle is based on electrochemical potential difference between a reference electrode and the steel reinforcement bar within the concrete structure, enabling continuous or on-demand monitoring of corrosion and humidity levels.

The corrosion process in reinforced concrete is fundamentally electrochemical. Steel reinforcement in alkaline concrete (pH > 12.5) forms a passive oxide layer that prevents corrosion. However, chloride ingress from marine environments or carbonation can destroy this protective layer, initiating an electrochemical corrosion cell. We call this the Wenner method, which allows us to determine the resistivity using the measured voltage. We can then estimate the corrosion rate through its correlation with resistivity value.

The electronic design development of the SHM system was oriented to perform measurement actions and enable data transmission to the owner/operator operational base. The achieved measurement precision is in the range of millivolts (± 2 mV). The sensor integrates an analog-to-digital converter for high-resolution voltage measurements, a precision current source for resistivity testing, and onboard signal processing to filter noise and compensate for temperature variations. Data is processed locally using embedded algorithms and transmitted wirelessly via LoRa protocol to a central database for remote monitoring.

For material degradation analysis and comparison with predictions, a specific current sequence is injected into the material, making the solution suitable also for design optimization and laboratory research on complex reinforced concrete systems, see example in attached figure.

When current is first injected into reinforced concrete, the material exhibits polarization, meaning its electrical resistance temporarily increases as mobile ions within the concrete pore solution align with the applied electric field. This polarization process typically completes within 100-200 seconds. Once the current injection stops, the concrete undergoes depolarization or relaxation, gradually returning to its original resistivity state over approximately 1000-1100 seconds. This behavior reflects the electrochemical nature of concrete, where ions are temporarily trapped by electrostatic interactions with fixed charges in the material but eventually return to their equilibrium distribution. Understanding this relaxation phenomenon is crucial for accurate measurements, as repeated testing must account for the time needed for the concrete to fully recover its baseline state between measurements.

It has been demonstrated that humidity level in concrete is also measurable and offers predictive maintenance capabilities through early detection of corrosion initiation.

Several examples of corrosion, humidity detection and rating will be illustrated.

A critical challenge in implementing sensors for Structural Health Monitoring (SHM) is minimising the intrusiveness of the sensors and cabling implied. Low-profile and lightweight sensors can be manufactured using printed electronics technologies, which are able to print scalable and flexible devices. By printing the connection wires all-together, heavy extensive cabling is also avoided. This solution would allow integrating larger number of sensors over large area without reducing the mechanical performances of the structure.

In this context, printed strain gauges were developed using conductive inks (silver-based and carbon-based) and evaluated as an alternative to conventional strain sensors. The developed silver-based printed sensors were first attached to metal and composite plates using thermoplastic films. Commercially available strain gauges were also attached on the other side of the plates for comparison. Under fatigue testing, the silver-based printed strain gauges showed good sensitivity to strain with resistance change similar to that of commercial strain gauges, despite a very low untrained resistance of 10 Ω. They also showed good repeatability, and fatigue resistance under 1000 cycles of tensile loading. The resulting strain gauges were estimated about 30 times thinner and 30 times cheaper than commercial strain gauges.

In addition to improved integration to the structure, using printing technology allows to in-situ fabricate a cheap network of sensors more accurately and more efficiently than manually bonding commercial sensors. To demonstrate adaptability, tailored designs of strain gauges in different directions, i.e. strain rosette, along with connection wires were printed which demonstrated their feasibility to achieve full-field multi-axial strain map. In the attached figure, silver-based printed strain gauges are compared to carbon-based printed strain gauges attached to composite plates. The developed carbon-based printed strain gauges have higher untrained resistance of about 1 kΩ. Under manual bending of the composite plate, they showed considerably higher sensitivity. They were also able to detect strain direction, as the higher strain was for the strain gauge aligned with strain direction; half-strain was detected by the strain gauge at 45°; almost no strain was detected by the strain gauge at 90°.

On-going work is further evaluating the monitoring performances of the carbon-based printed strain gauges. These sensors show good potential for embedding within composite structures to create smart structures. The development of other types of printed sensors is also considered to diversify the monitoring performances.

Structural Health Monitoring involves identifying and assessing structural changes or damage with or without relying on predefined numerical models. Model-free methods exploit only damage-sensitive dynamic characteristics obtained directly from measured responses. Some of these are based on hypotheses regarding expected damage, which turn into recognizable footprints. Alternatively, model-free methods require the availability of a baseline that refers to an intact condition. The proposed damage identification methods fall into the latter category because they enable efficient and reliable monitoring of structural integrity under operational conditions using an array of accelerometers. The scope of this study is to validate both methods experimentally by applying them to scale-model tests simulating the onset of damage on a real ship.

The first method is based on a macro index, mapping the probability of damage over the investigated structure. It incorporates several damage indices that process the sampled modal curvatures. These modal curvatures, linearly averaged or squared over the sensor mesh (geometric strain energy), are computed in both the reference-intact and current-damage conditions and fed into the different functions characterizing the various indices. If damage is effectively present in one element of the sensor mesh, damage severity can be inferred from the index value. Nonetheless, damage localization can be improved if the indices are Z-score normalized and combined into the macro-index using ensembling strategies. Thus, thresholds for damage existence can be properly set low on the index average to have sufficient sensitivity, while agreement conditions among indices can be exploited to reduce false warnings.

The second method is a novelty detection approach based on a histogram score. It shares with the previous method also the use of one of the indices considered above (the Cornwell’s formulation of the Modal Strain Energy Index) though in a different way. The operational modal analysis provides the vibration modes from the tests, which are characterized by noise from different sources. To train the method with a sufficiently large ‘intact’ population, the mode shapes are first averaged and then contaminated with Gausisan noise, experimentally modelled. The statistical distribution of the Cornwell’s damage index provides the baseline to evaluate whether the same quantitity, computed directly from the experimental data, can be attributed to an underlying structural modification. The threshold to separate the intact and damaged classes plays again a crucial role.

The damage identification techniques are validated using data collected in scale-model tests of a navy vessel within the “Digital Ship Structural Health Monitoring project” (dTHOR), granted by the European Defence Fund, aimed at developing a system based on innovative utilization of extensive on-board measurements, a comprehensive digital framework, and hybrid analysis and modelling. The ship longitudinal bending stiffness is reproduced by an elastic backbone connecting the hull portions. The damage is artificially generated by removing the plate elements on top and side faces of the aluminium backbone and affects the operational modes identified while the physical model is towed at the CNR-INM wave basin. Results, though quite promising, are critically reviewed in the perspective of further increasing their accuracy.

This paper explores guided wave propagation in a square aluminum beam through both numerical and experimental methods. The damaged and healthy conditions of the aluminum beam were considered for the analysis. Damage was modeled experimentally and numerically in the form of an open crack. The guided waves are excited by a piezoelectric disc bonded to the beam at the front face. The aim is to determine the optimal placement of fibre optic strain sensors to enable the separation of propagating guided wave modes. Mode separation is crucial because it simplifies the signal processing required for damage detection. To achieve high sensitivity to guided waves with amplitudes in the nanometre range, fibre Bragg grating (FBG) sensors combined with an edge filtering method were utilised. The advantage of FBG sensors in comparison to piezoelectric transducers (PZTs) is that FBGs can be embedded into composite laminates without a detrimental effect on their strength. Complementary measurements were taken using scanning laser Doppler vibrometry, allowing full wavefield analysis. Additionally, numerical simulations of elastic wave generation and sensing were conducted using COMSOL software. A 3D model is applied with electromechanical coupling, which includes asymmetry caused by the shape of PZT (simulating wrap-around electrode). The focus of this research is on the potential for sensing symmetric and antisymmetric guided wave modes with mode separation capabilities. For this purpose, a series of FBG sensors positioned on the perimeter of the specimen are employed.

Methodology

An aluminium beam measuring 2 m × 0.01 m × 0.01 m was used for both experimental and numerical investigations. A crack with a depth of 2.5 mm was introduced at a location 1.2 m from the left end of the beam. In the experimental setup, an actuator was mounted on the front face of the beam to excite guided waves using a PZT transducer. It should be noted that the wrap-around electrode and solder material used in the PZT actuator introduce asymmetries. Wave responses along the beam were captured using scanning laser Doppler vibrometry. The acquired signals were band-pass filtered around the respective carrier frequencies. A five-cycle Hann-windowed sinusoidal tone burst with a carrier frequency of 50 kHz was generated using a waveform generator, and the excitation signal was amplified tenfold using a voltage amplifier. The sampling rate was set to 1.28 MHz.

Results

The experimental results were compared with the numerical simulations. The waterfall plot for guided wave propagation from experimental measurements and numerical simulations show two guided wave modes, A0 and S0. A very good match of velocities between experiment and numerical simulation is observed.

Acknowledgment

The authors would like to gratefully acknowledge the support given by the National Science Centre, Poland, under grant agreement no. 2020/39/B/ST8/01753 in the frame of the OPUS project entitled: "Study of elastic wave mode sensing and separation using FBG sensors for structural health monitoring".

The Structural Health Monitoring (SHM) of aerospace vehicles faces extreme
challenges due to harsh operating conditions, stringent latency requirements, high reliability
constraints, and limited energy resources. While conventional wired SHM architectures remain
reliable, they significantly increase the total weight and complexity of the vehicle’s electrical
system. This weight penalty is particularly pronounced in commercial and military aircraft, where
the extensive wiring infrastructure has become a critical design constraint. According to the
Wireless Avionics Intra-Communications (WAIC) initiative, modern aircraft can have
approximately 100,000 wires totaling 470 km in length and weighing around 5,700 kg, with about
30% of these wires being potential candidates for wireless substitution to reduce weight and
simplify installation. Consequently, there is a growing trend toward adopting wireless sensor
networks (WSNs) to maintain reliability while reducing wiring mass. While WSNs offer
promising weight reduction benefits, current implementations lack the adaptive intelligence
required to satisfy the complex, multi-dimensional QoS constraints of aerospace systems,
particularly when facing simultaneous demands for minimal latency, maximum reliability, and
optimal energy consumption. These multi-objective constraints necessitate more intelligent
network management, that can autonomously adapt to changing mission requirements and
environmental conditions.
This work presents an embedded Artificial Intelligence (AI) framework for autonomous WSNs
dedicated to aerospace vehicle SHM. The proposed system leverages Time Slotted Channel
Hopping (TSCH) and RPL routing protocols, enhanced by an adaptive AI-based controller
(CEREBRO), that dynamically optimizes communication schedules based on real-time network
conditions and environmental variations. The proposed AI framework combines convolutional
layers for capturing local spatial structures with specialized temporal processing blocks designed
to model sequential dependencies and temporal dynamics in the data. The combined model
outputs adaptive routing and scheduling vectors, which CEREBRO uses to adjust slot allocations
and routing paths based on network operations and conditions, ensuring that critical
measurements are delivered reliably and efficiently while extending network lifetime in complex
aerospace environments.
The proposed approach has been experimentally validated on a real testbed using the FIT IoT-
LAB platform, showing significant improvements in reliability, latency, and energy consumption
compared to static TSCH configurations. The resulting architecture provides a scalable,
lightweight, and energy-efficient foundation for embedded SHM in aerospace structures. Overall,
this work highlights the potential of embedded intelligence to transform SHM systems into fully
autonomous and adaptive diagnostic networks, capable of operating reliably and efficiently in the
extreme environments of aerospace systems. To accommodate different deployment scenarios,
CEREBRO is available in both a CPU implementation for resource-constrained nodes and a GPU-
accelerated version on NVIDIA Jetson platforms for high-performance edge computing.
The framework advances the paradigm of edge-based diagnostics by bringing intelligence directly
to the acquisition system, enabling real-time analytics and adaptive decision-making without
relying on centralized cloud processing.

High-density heat exchangers play a pivotal role in advanced thermal systems, yet precise internal monitoring remains challenging due to the invasiveness and limited access of conventional sensing methods. Accurate characterization of local fluid temperature and heat transfer coefficients (HTC) is essential for evaluating thermal performance and ensuring structural integrity in complex micro-channels. In this study, we introduce a new design variable, 'Fiber Drag Length (LD)', to resolve the inherent cross-sensitivity between temperature and strain in Fiber Bragg Grating (FBG) sensors. By employing a self-compensating differential measurement technique—combining a strain-free reference sensor with active measurement sensors—the system effectively decouples mechanical flow drag from thermal loads without the need for bulky packaging. Validated in a shell-and-tube heat exchanger environment under turbulent flow conditions (Re 5889~20769), the sensor array successfully captured real-time internal flow behaviors. Additionally, an integrated analysis using the LMTD and Wilson Plot methods allowed for the precise separation of the internal convective heat transfer coefficient (hi), yielding a correlation of hi=3370*V^0.8 that aligns well with theoretical predictions. To overcome the flow disturbances caused by existing invasive sensors and the limitations of surface thermography, the proposed FBG system provides a scalable and minimally invasive solution for the real-time structural health monitoring and design optimization of next-generation heat exchangers. Furthermore, it presents a new paradigm of 'thermo-hydraulic integrity monitoring' to ensure both optimal heat transfer efficiency and structural safety.

In recent years, composite materials have assumed a crucial role in advanced engineering sectors such as automotive, aerospace, and wind energy, owing to their outstanding specific mechanical properties. However, composite components subjected to cyclic loads experience a progressive degradation of their mechanical performance due to the accumulation and interaction of multiple damage mechanisms, including off-axis cracking, delamination, and fibre breakage. Consequently, continuous monitoring of the structural integrity of such components is essential in order to enable timely intervention when the functional requirements defined at the design stage are no longer satisfied. In the case of electrically conductive laminates, Structural Health Monitoring (SHM) can be effectively achieved through electrical methods, as damage phenomena induce detectable changes in electrical conductivity. This work presents an analytical framework capable of accurately describing damage evolution and the associated stiffness degradation in multidirectional conductive laminates affected by off-axis cracks and delamination in multiple layers, using the increase in electrical resistance as input parameter. The accuracy of the proposed framework is first assessed through comparison with finite element analyses, demonstrating excellent agreement. Furthermore, fatigue tests are conducted on CNT-modified quasi-isotropic GFRP laminates, with continuous monitoring of damage evolution, stiffness degradation, and electrical resistance variation. The experimental results closely match the analytical predictions, thereby confirming the accuracy of the framework and demonstrating its effectiveness as a reliable tool for the Structural Health Monitoring of composite materials.

The development of an accessible electronic circuit for analyzing Structural Health Monitoring (SHM) algorithms using piezoelectric elements is a relevant research topic in current structural health monitoring projects [1]. This is because the development of SHM requires harmony between its mathematical methods and low-cost technological design [1]. Moreover, removing the analog-to-digital converter (ADC) in SHM systems can address key technical hurdles, particularly in reducing overall power consumption for practical applications, among other important issues [1]. In this presentation, we introduce a novel electronic circuit for SHM that employs analog reference signals, building upon the suggestion in [1]. We also integrated an electrically manipulable boost DC-DC converter into our electronic system to manage the main 12V voltage source. This voltage management was possible because our DC-DC converter is a controllable power source that utilizes a hysteresis loop circuit via a microcontroller device of 8-bits from Microchip manufacturer [2]. See Figure 1. A photograph of the experimental setup is also provided. Our experimental setup comprises three aluminum bars representing different structural states: Case A (healthy), Case B (moderate damage via a single perforation), and Case C (severe damage via two perforations). Three piezoelectric transducers (PZTs) mounted on each bar base serve as actuators; these are connected in parallel and driven by a single power exciter. Three additional PZTs are positioned at each bar top to serve as receiving elements. Figure 1 also displays the experimental residual signal, which was acquired via a powerful digital oscilloscope (Picoscope technology) and displayed on a computer. This device was configured to ensure the residual signal was clearly visible. For each measurement scenario, the boost converter was deactivated to allow for the establishment of a new sensor connection point (the sensor selector switch on Figure 1) before being restarted. The resulting residual signal clearly shows that a stable average voltage can effectively distinguish among each testing case (A, B or C case). From a SHM data processing perspective, our closed-loop design relies on analog circuitry, with the exception of a PIC microcontroller. This microcontroller functions as the primary interface among analog control signals and the gate drive of the boost DC-DC converter. The converter requires a PWM (Pulse Width Modulation) signal to regulate its operation. The PIC microcontroller is configured to generate a PWM signal with a nominal frequency of about 15 kHz. Consequently, our system processes signals in the analog domain, eliminating the sampling rate requirements typically imposed by digital-to-analog conversion. Additionally, each piezoelectric actuators is driven by an analog signal ranging from 10V to 60V (these levels adjustable via the P₁ and P₂ trimmers) with a 30 Hz harmonic content, a direct result of the NE555N timer’s operational frequency. Therefore, our SHM approach, which employs analog signals for damage detection, aligns with established procedures in residual signal analysis [3].

References

[1] Owen, R., et al. (2011). (No. NASA/CR-2011-217153).

[2] Acho, L. (2025). Cybernetics and Physics, 14(4), 315-319.

[3] Vidal, Y., et al. (2012). Mechanical Systems and Signal Processing, 29, 447-456.

To identify the cracking mechanisms that occur when trucks and cars pass over a structure, this study introduces a new approach to analyzing acoustic emission (AE) signals from a prestressed concrete viaduct. This method utilizes the information contained within the narrow partial power (PP) bands of the acquired AE signals. These bands have proven to be extremely sensitive to changes in concrete deterioration modes in both field tests and laboratory experiments. The results demonstrate the ability to classify trucks based on the cracking information contained within the narrow PP bands of the AE signals acquired when trucks pass by. Additionally, laboratory bending tests with AE monitoring were conducted on reinforced concrete (RC) T-beams. The narrow PP strips' effectiveness in distinguishing between modes of degradation in the beams was remarkable. The proposed method uniquely offers the advantage of visualizing the frequency composition of a large number of AE signals using a heat map based on narrow PP bands. This study involves simultaneously measuring deformation and AE signals as trucks and cars pass by. The correlation between these two types of data is also presented. This research demonstrates that frequency information from AE signals can be used to monitor prestressed and reinforced concrete structures.

Traditional structural health monitoring (SHM) systems relying on conventional piezoelectric transducers (PZTs), such as phased arrays, often face significant hurdles due to hardware complexity, high power consumption, and prohibitive integration costs. To address these challenges, this study explores the development of a Steerable Acoustic Transducer (SAT) designed with inherent directional capabilities for generating and sensing elastic waves.

Unlike omnidirectional sensors, the SAT leverages a frequency-dependent spatial filtering effect. The device’s response is governed by the intersection of the medium’s dispersion relation and the wavenumber representation of the transducer’s electrode geometry (Fig. 1). By strategically distributing electrode material, a unique relationship is established between the signal frequency and the direction of wave propagation. This allows for precise steering of guided waves by simply tuning the frequency, drastically simplifying the necessary excitation hardware.

While initial studies have demonstrated the potential of SATs in isotropic structures, alternative technologies for their development have received limited attention. This proposal investigates the feasibility of SATs for advanced SHM applications. Specifically, a novel manufacturing approach utilizing high-powered, pulsed laser beam is proposed to achieve the required electrode precision. This method is demonstrated and verified through experimental wavefield characterization.