Assessing the structural integrity of cultural heritage objects is of great importance for their structural conservation and long-term preservation. This paper focuses on the development of a non-destructive inspection (NDI) approach using 3D shearography to evaluate the structural integrity of wax-resin lined paintings, specifically for The Night Watch (1642), a large-format 17th-century canvas painting by Rembrandt van Rijn (1606–1669) that is on display in the Rijksmuseum, Amsterdam. The Night Watch has a complex treatment history that has many old repairs of structural defects and damages (holes, tears, etc.) and three wax-resin relinings. In 2021, before a new structural intervention involving retensioning of the canvas support, it was vital to evaluate the structural integrity of the painting, specifically the condition of the treatment carried out in 1975–76 when, among other actions, several long cuts in the area of Captain Frans Banninck Cocq's breeches were repaired and an old canvas insert in the drum was replaced. To assess the structural condition, we applied 3D shearography to quantitatively analyse the in- and out-of-plane surface strains with controlled thermal loading. First, a safe loading procedure was developed by inspecting a representative wax-resin lined test painting where reference delaminations and structural repairs to canvas supports were reliably identified with 3D shearography by raising the temperature with 1-2 °C. As part of Operation Night Watch, in November 2021 an in-situ investigation was carried out in the Rijksmuseum gallery. Two areas of interest in The Night Watch, the restored slashes in the Captain's breeches (0.5 × 1 m) and the canvas insert in the drum (0.2 × 0.5 m), were inspected from the reverse of the painting. Results revealed no critical structural problems associated with the repaired slashes, nor with adhesion of the lining. For the patched canvas in the drum, it showed higher in- and out-of-plane strain variations. Overall, 3D shearography provided valuable non-destructive inspection results for assurances regarding the structural integrity of the 1975 repairs and the adhesion of the lining canvas in The Night Watch.

Non-destructive inspection (NDI) of small-scale defects in fiber-reinforced composites is an urgent challenge to ensure the structural integrity of safety–critical components. Shearography is a non-contact and full-field optical NDI method that can be used to characterize surface strain components under loading. Thermal loading is widely used in shearography because of the advantages of being non-contact and convenient for in-service inspection. Shearography has received considerable industrial acceptance for the inspection of aerospace and marine composite structures, however its efficacies in detecting small mm and sub-mm defects have not been fully characterised. Besides, one major issue regarding shearography with thermal loading is fiber-related deformation or fiber noise, which can affect the efficacy of defect detection significantly, especially when detecting small and deep defects. In this study, the novel shearography pair method is proposed and developed to reduce fiber noise for reliable inspection of small mm and sub-mm defects in composites. The defect detection capabilities of the proposed method have been studied and compared with conventional shearography practice and with fast Fourier transform (FFT) and principal component analysis (PCA) based signal processing algorithms. The results demonstrate that the proposed shearography pair method has the advantages of less fiber noise, improved inspection results, and being faster with reduced number of datasets. It enables the detection of mm and sub-mm defects (down to 0.6–0.8 mm in diameter) in composites; these inspection results are one of the smallest defect sizes detected with shearography and reported in literature.

Twist and crimp values are of paramount importance to the textile industry in understanding the properties and performance of a textile, and their quantification has been a subject of study since the early 20th century. Twist and crimp are the result of how the fibers have been modified from the original bundle to shape the textile, so the industrial methods used to measure them are based on mechanically reversing such deformations. The same information is needed to study the mechanics of historical fabrics such as canvas paintings supports and historical textiles, but they are more difficult to obtain because these are often brittle and impregnated with foreign materials, less homogeneous and very limited in availability for sampling. Therefore, such fundamental parameters are usually unavailable for conservation studies.

This paper examines the protocols used in the textile industry and proposes new methods, developed from previous research, for the reliable measurement of twist and crimp in historical textiles. The twist measurement method is non-destructive as it is based on observing the textile and the fibers on the surface of the yarn. Crimp is the undulation of the interlaced yarns and its measurement is an invasive examination of the internal structure of the textile, as it requires the observation of individual yarns. Both methods, applied here to a group of historical textiles, provide data in accordance with the current parameters of the textile industry, and their use is relatively simple and inexpensive.

With the rapid development of artificial intelligence (AI) technologies, deep learning-based structural health monitoring (DeepSHM) methods have gained significant attention. However, their black box nature often limits interpretability and trust. The field of Explainable AI (XAI) aims to address this by enhancing model transparency and reliability through human-comprehensible explanations. This study investigates the use of XAI algorithms in interpreting a 1D convolutional neural network (1D CNN) developed for Lamb wave monitoring of bolt-loosening detection in multi-bolted double-layer aluminum plates under varying temperatures. Four existing XAI algorithms were employed, including Sensitivity Analysis, Deep Taylor, Gradient-weighted Class Activation Mapping (Grad CAM) and Guided Grad CAM. In addition, this paper introduces two new XAI methods, Smooth Simple Taylor and Deep Grad CAM as an enhancement of the Simple Taylor and Grad CAM methods, respectively. These six XAI algorithms were used to establish the relation between the 1D CNN model parameters and the input vector. The results were evaluated for their effectiveness in comparison to the physical insights of the input vector using two proposed methods, namely the Correlation Coefficient with Residual Signal and the Residual Signal Weighted Importance Score Ratio. The results of the evaluation methods, in conjunction with Infidelity, Sense sum, and Sanity check, were utilized to rank the performance of the six XAI algorithms. The rankings were consistent in both simulation and experiment data sets, and the newly proposed XAI algorithm, Smooth Simple Taylor, appeared to be the best in both data sets. Overall, this research establishes a novel approach to using XAI algorithms to enhance the explainability of AI in practical engineering applications.

Monitoring fatigue damage in mechanical connections is essential for maintaining the safety and structural integrity of offshore wind turbines (OWTs), particularly during the early stage of crack initiation. Recently, the C1 wedge connection (C1-WC) has emerged as a promising innovation for use in OWTs. Acoustic emission (AE) monitoring is a widely used real-time technique for detecting fatigue cracks. The space limitations of the lower segment holes in the C1-WC presents challenges for detecting surface cracks with conventional AE sensors. Thin Piezoelectric Wafer Active Sensors (PWAS), while small and lightweight, face limitations due to their poor signal-to-noise ratio. In this study, we propose a baseline-based approach to enhance the effectiveness of PWAS for accurate AE monitoring in confined spaces. A benchmark model correlating the damage state of specimens is created by breaking pencil leads. Multivariate feature vectors are extracted and then mapped to the Mahalanobis distance for damage identification. The proposed method is validated through testing on compact specimens and C1-WC specimens. To enhance the AE detection results, supplementary monitoring techniques, including digital image correlation, crack propagation gauges, and distributed optical fiber sensors, are employed. The experimental setup, signal acquisition, and detection efficiency of these techniques are briefly outlined. This study demonstrates that the proposed approach is highly effective in detecting early damage in C1-WC specimens using AE monitoring with PWAS.

Machine Learning (ML) has revolutionized various fields, enabling the development of intelligent systems capable of solving complex problems. However, the process of manually designing and optimizing ML models is often time-consuming, labor-intensive, and requires specialized expertise. To address these challenges, Automatic Machine Learning (AutoML) has emerged as a promising approach that automates the process of selecting and optimizing ML models. Within the realm of AutoML, Neural Architecture Search (NAS) has emerged as a powerful technique that automates the design of neural network architectures, the core components of ML models. It has recently gained significant attraction due to its capability to discover novel and efficient architectures that surpass human-designed counterparts. This manuscript aims to present a systematic review of the literature on this topic published between 2017 and 2023 to identify, analyze, and classify the different types of algorithms developed for NAS. The methodology follows the guidelines of Systematic Literature Review (SLR) methods. Consequently, this study identified 160 articles that provide a comprehensive overview of the field of NAS, encompassing discussion on current works, their purposes, conclusions, and predictions of the direction of this science branch in its main core pillars: Search Space (SSp), Search Strategy (SSt), and Validation Strategy (VSt). Subsequently, the key milestones and advancements that have shaped the field are highlighted. Moreover, we discuss the challenges and open issues that remain in the field. We envision that NAS will continue to play a pivotal role in the advancement of ML, enabling the development of more intelligent and efficient ML models for a wide range of applications.

With the increasing application of artificial intelligence (AI) techniques in the field of structural health monitoring (SHM), there is a growing interest in explaining the decision-making of the black-box models in deep learning-based SHM methods. In this work, we take explainability a step further by using it to improve the performance of AI models. In this work, the results of explainable artificial intelligence (XAI) algorithms are used to reduce the input size of a one-dimensional convolutional neural network (1D-CNN), hence simplifying the CNN structure. To select the most accurate XAI algorithm for this purpose, we propose a new evaluation method, feature sensitivity (FS). Utilizing XAI and FS, a reduced dimension 1D-CNN regression model (FS-X1D-CNN) is proposed to locate and predict the torque of loose bolts in a 16-bolt connected aluminum plate under varying temperature conditions. The results were compared with 1D CNN with raw input vector (RI-1D-CNN) and deep autoencoders-1D-CNN (DAE-1D-CNN). It is shown that FS-X1D-CNN achieves the highest prediction accuracy with 5.95 mm in localization and 0.54 Nm in torque prediction, and converges 10 times faster than RI-1D-CNN and 15 times faster than DAE-1D-CNN, while only using a single lamb wave signal path.

Hydrogen, which serves as a major driver of sustainable aviation, requires precise sensing methods. This study focuses on the development of a metal hydride-coated tilted fibre Bragg grating (TFBG) based sensor for hydrogen detection, with tantalum as a novel sensing material for fibre optic hydrogen sensing. Magnetron sputtering has been employed to deposit nanometer-scale metal films onto the optical fibre surface of the TFBG structure. In this proof of concept work, changes in both amplitude and the centre wavelength of cladding resonances of the TFBG transmission spectrum were observed in the hydrogen concentration range from 0.01% to 4% at room temperature.

Hydrogen, crucial in industrial and environmental realms, demands precise sensing methods. This study focuses on the design of a metal hydride-coated tilted fibre Bragg grating (TFBG) based sensor for hydrogen detection, introducing tantalum as a novel sensing material for fibre optic hydrogen sensor development. To facilitate the ellipsometry inspection of optical constants, magnetron sputtering technique has been employed to deposit nanometer-scale metal films onto a glass substrate. Numerical modeling results are presented for mode analysis of the proposed sensor design, analyzing transverse mode behavior for sensor optimization. The study also provides insights into other TFBG sensor design considerations, suggesting potential hydrogen sensing applications, such as hydrogen-powered aviation and storage solutions.

Acoustic emission (AE) is widely used for identifying source mechanisms and the deformation stage of steel material. The effectiveness of this non-destructive monitoring technique heavily depends on the quality of the measured AE signals. However, the AE signals from deformation are easily contaminated by the signals from noise in a noisy environment. This paper presents a hybrid model for deformation stage identification, which combines a self-adaptive denoising technique and an Artificial neural network (ANN). In pursuit of model generality, AE signals were collected from tensile coupon tests with various steel materials and loading speeds. First, a decomposition-based denoising method is applied based on the singular spectral analysis (SSA) and variational mode decomposition (VMD), which is defined as SSA-VMD. Its effectiveness is demonstrated by simulated signals and experimental results. Following the use of the denoising technique, an ANN is constructed to identify the deformation stage of steel materials with the input of features extracted from the filtered AE signals. The results indicate that the ANN achieves a high prediction accuracy of 0.93 in the test set and 0.87 in unseen data. By applying this denoising method, the ANN-based approach enables accurate correlation of the collected AE signals to deformation stages. The finding can be used as the basis for the creation of new methodologies for monitoring structural health status of in-service steel structures.

Palladium thin films have been studied as hydrogen sensing materials and applied to variety of optical hydrogen sensors. Recently, tantalum has emerged as an attractive option for hydrogen sensing materials due to its broad sensing range and flexibility in tuning the sensing range by modifying the alloying composition or elements. Following the demand for optical hydrogen sensors for aerospace applications, testing the performance of hydrogen sensing materials is of interest. This work examines the optical response in respect to changing hydrogen concentrations and thermal expansion of palladium-gold (Pd_0.65Au_0.35) and tantalum-ruthenium (Ta_0.97Ru_0.03 and Ta_0.91Ru_0.09) thin films at temperatures similar to a hydrogen combustion engine. Our results suggest that tantalum-ruthenium alloys are suitable for sensing hydrogen from ambient temperatures up to 270^◦C because its low detection limit (0.01% of hydrogen in the atmosphere) is well below the explosive limit of hydrogen (4% of hydrogen in the atmosphere).

Carbon fiber and glass fiber-reinforced polymer composites are extensively used in many industries such as aerospace, wind energy, and automobile. Nowadays, due to the high interest in manufacturing safe and cost-optimized composite structures, the need for nondestructive-evaluation (NDE) of these structures is rising significantly; either as a part of the manufacturing process or while the structure is in service. Among all NDE techniques, one of the most promising for composite materials is ultrasonic testing (UT). Ultrasonic NDE uses low amplitude acoustic/elastic waves with frequencies higher than about 20kHz to evaluate components without affecting the material properties and their integrity. This technique is extensively employed for the inspection of parts during in-service use, for quality control during the manufacturing process, as well as for material characterizations. The main advantages of the ultrasonic NDE technique are the possibility of high penetration in many commonly used materials, robustness against the conductivity of the material, accuracy in locating and measuring defects, and the ability to be performed with one-sided access. However, there are some drawbacks such as the need for a highly experienced operator and the necessity of maintaining a high degree of coupling between the transducer and the surface of the structure. While conventional ultrasonic NDE using single-element transducers has been commonly used in the past, it has some limitations for the evaluation of thick composite layers, such as the high attenuation and low signal-to-noise ratio. The phased array ultrasonic technique (PAUT) significantly improves the capability of UT for the inspection of composite materials as the ultrasonic waves can be generated and focused at desired energies, angles, and distances. This chapter is organized to provide the essential theoretical background and technical knowledge on ultrasonic NDE of thick composite laminates, starting from the physics of ultrasonics and phased array systems, going up to the signal-processing, considerations on how to deal with the collected data, and finally, the real-world applications.

With the improvements in computational power and advances in chip and sensor technology, the applications of machine learning (ML) technologies in structural health monitoring (SHM) are increasing rapidly. Compared with traditional methods, deep learning based SHM (Deep SHM) methods are more efficient and have a higher accuracy. However, due to the black box nature of deep learning, the trained models are usually difficult to interpret, which blocks their practical application. Therefore, it is of great importance to develop explainable artificial intelligence (XAI) methods to understand the internal decision-making mechanisms of damage classification in Deep SHM. In this paper, a novel XAI algorithm named Deep Gradient-weighted Class Activation Mapping (Deep Grad CAM) is proposed by combining the existing method Grad CAM with the convolutional neural network (CNN) deconvolution mechanism. In this paper, Deep Grad CAM is used to interpret a one-dimensional convolutional neural network trained to detect bolt loosening based on guided wave propagation. The interpretation performance of Deep Grad CAM is compared with Grad CAM, and their performances are quantified using Infidelity. The results show that the Infidelity of Deep Grad CAM is much smaller than that of Grad CAM, indicating significant improvements in explanation accuracy and reliability.

Metal hydrides have been widely studied as hydrogen sensing materials and applied to various optical sensor configurations. With the increasing interest in using hydrogen as an energy source across sectors involving combustion processes, there is a growing demand for reliable hydrogen sensors operating at temperatures above 100 °C. Therefore, it is necessary to evaluate the performance of potential hydrogen sensing materials at elevated temperatures. We conducted experiments to observe the optical response and structural characteristics of palladium, palladium–gold, tantalum, and tantalum-alloy thin films with respect to varying hydrogen concentrations from 28 °C to 270 °C. Our results demonstrate that the optical response of palladium and palladium–gold diminish at 270 °C. However, tantalum provides a remarkable optical response to hydrogen concentrations below 1% for all the observed temperatures and a stable response at 270 °C for 350 cycles. Our measurement results show that tantalum is the most suitable material for detecting hydrogen within the range of 0.01% to 100% at temperatures ranging from 28 °C to 270 °C.

Composite materials, e.g., carbon fibre-reinforced polymers (CFRPs), have been increasingly adopted in safety-critical structures across different industries. However, various defects including delaminations and fibre breakage can occur in the composite structures that may endanger the whole structure severely. Therefore non-destructive testing (NDT) of composites is critical to ensure structural integrity and safety. It is important to advance the capabilities of NDT of composite materials towards early-stage damage, e.g., small defects of millimetre scale, to avoid future failure. The objective of this work is to study the detection of small defects in CFRP laminates using shearography with thermal loading. In this paper, a thermal-mechanical FEM model was established in Abaqus to assist shearography inspection of the composite laminate. This FEM assistance is capable of evaluating different thermal loading schemes for defect detection. A rational selection of the reference and signal interferograms from the heating/cooling sequence is determined for reliable defect detection. We will present both experimental and numerical results on the detection of small defects in CFRP.

The acoustic emission (AE) technique is commonly utilized for identifying source mechanisms and material damage. In applications requiring numerous sensors and limited detection areas, achieving significant cost savings, weight reduction, and miniaturization of AE sensors is crucial. This prevents excessive weight burdens on structures while minimizing interference with structural integrity. Thin Piezoelectric Wafer Active Sensors (PWAS), compared to conventional commercially available sensors, offer a miniature, lightweight, and affordable alternative. The low signal-to-noise ratio (SNR) of PWAS sensors and their limited effectiveness in monitoring thick structures result in the decreased reliability of a single classical PWAS sensor for damage detection. This research aims to enhance the functionality of PWAS in AE applications by employing multiple thin PWAS and performing a data-level fusion of their outputs. To achieve this, as a first step, the selection of the optimal PWAS is performed and a configuration is designed for multiple sensors. Pencil break lead (PBL) tests were performed to investigate the compatibility between selected PWAS and traditional WSα and R15α sensors. The responses of all sensors from different AE sources were compared in both the time and frequency domains. After that, convolutional neural networks (CNNs) combined with principal component analysis (PCA) are proposed for signal processing and data fusion. The signals generated by the PBL tests were used for network training and evaluation. This approach, developed by the authors, fuses the signals from multiple PWAS and reconstructs the signals obtained from conventional bulky AE sensors for damage detection. Three CNNs with different architectures were built and tested to optimize the network. It is found that the proposed methodology can effectively reconstruct and identify the PBL signals with high precision. The results demonstrate the feasibility of using a deep-learning-based method for AE monitoring using PWAS for real engineering structures.

Nanoparticle- (NP-) doped optical fibres show the potential to increase the signal-to-noise ratio and thus the sensitivity of optical fibre strain detection for structural health monitoring. In this paper, our previous experimental/simulation study is extended to a design study for strain monitoring. 100 nm spherical gold NPs were randomly seeded in the optical fibre core to increase the intensity of backscattered light. Backscattered light spectra were obtained in different wavelength ranges around the infrared C-band and for different gauge lengths. Spectral shift values were obtained by cross-correlation of the spectra before and after strain change. The results showed that the strain accuracy has a positive correlation with the relative spectral sensitivity and that the strain precision decreases with increasing noise. Based on the simulated results, a formula for the sensitivity of the NP-doped optical fibre sensor was obtained using an aerospace case study to provide realistic strain values. An improved method is proposed to increase the accuracy of strain detection based on increasing the relative spectral sensitivity, and the results showed that the error was reduced by about 50%, but at the expense of a reduced strain measurement range and more sensitivity to noise. These results contribute to the better application of NP-doped optical fibres for strain monitoring.

The performance of defect detection in composite materials using digital shearography is important for correct decision-making in non-destructive testing. In this work, we compared a high-resolution 24-megapixel digital still camera (DSLR) and a conventional medium-resolution 5-megapixel camera to determine the detectability of blind holes in an aerospace-graded carbon-fiber reinforced polymer (CFRP) sample. The hole diameters ranged from 0.2 to 3 mm with a material thickness of 4 mm and the test sample dimensions of 200×200 mm. The sample was heated and observed from the front (defect-free side) by three halogen lamps for 5 minutes in pulsed heating mode. Speckle interferograms were acquired during the heating and cooling phases from both cameras simultaneously using identical shearing interferometers and shearing distances. Phase maps were calculated using the 4+4 temporal phase step algorithm and then unwrapped. Further, defect-induced deformation (DID) phase maps were obtained by polynomial curve fitting. The DID phase maps obtained from the two cameras were compared. Blind holes with diameters up to 1 mm were detected, which are one of the smallest defects detected with shearography and reported in literature. In addition, the DLSR camera was able to detect holes of 0.8 mm in diameter. We observed that nearly comparable detection capabilities were obtained from both cameras, even though the spatial resolution of the second camera (DLSR) was 5 times higher. Possible reasons of this limitation include effects such as fiber-related deformation in CFRP and speckle noise.

The colour of the ground layers of a painting has an influence on its visual appearance. In addition to the commonly used white ground layers, other colour ground layers have been used, for example, the grey ground layer used in Peter Paul Rubens’s painting Portrait of Clara Serena Rubens helps the colour transition of the skin tones. Understanding the effects caused by the colours of the ground layers is of significance for both technical art history and conservation. Optical non-destructive testing (NDT) techniques are useful tools for the investigation of paintings, for example, optical coherence tomography (OCT) can be used to study the surface and subsurface layers non-destructively. In this work, the interaction of light with paint and ground layers is modelled to supplement OCT measurements of paintings with ground layers. A previously described near-infrared light range OCT system provides high spatial and depth resolution measurements. A four-flux model has been developed for analysing the light interaction in the paint and ground layers. This model considers forwards-propagating collimated light, backwards-propagating collimated light, forwards-propagating diffuse light and backwards-propagating diffuse light. The model uses the optical material properties, including refractive index (RI), absorption and layer thickness, as input. This paper describes the construction of the model and an evaluation of its performance by comparison with OCT data.

The assessment of the structural condition of cultural heritage objects is important for conservation interventions and their long-term preservation. This investigation concerns The Night Watch (1642), a large-format 17^th-century canvas painting by Rembrandt van Rijn that is on display in the Rijksmuseum, Amsterdam. This painting, which has a complex treatment history, has various damaged areas and has undergone three wax-resin relinings. In 1975 the canvas was slashed twelve times with a serrated dinner knife, including several long slashes in the area of Captain Frans Banninck Cocq’s breeches. In 2021, prior to a proposed new structural intervention involving retensioning of the canvas, it was important to evaluate the structural condition of the repaired slashes and of another repair, specifically an old canvas insert in the drum. For this, an in-situ inspection was carried out in the Rijksmuseum as a part of Operation Nightwatch. 3D shearography instrument with thermal loading was used to inspect these two areas of interest on the reverse of The Night Watch. The results showed that the out-of-plane strain in the breeches does not show any large deviations, which alleviated conservators’ concerns about the adhesion of the lining canvas and stability of previous repairs in this region. The patch in the drum showed higher out-of-plane strain variations. This was explained by the lower quality of the patched canvas compared to the repaired slashes in the breeches of Banninck Cocq. Overall, 3D shearography provided valuable inspection results for assurances regarding the structural integrity of the 1975 repairs and the wax-resin lining in The Night Watch, reducing the risks and providing the confidence to proceed with the planned retensioning of the canvas.