Fiber Bragg grating (FBG) sensors have attracted growing interest in road health monitoring due to their high sensitivity, accuracy, and resilience to harsh environmental conditions. Continuous monitoring is essential for identifying patterns in the collected data and FBG sensors meet this need by continuously measuring strain across multiple pavement layers at high sampling frequencies, creating an extensive, high-resolution dataset. However, such large data volumes present substantial challenges for transmission, storage, processing, and analysis in structural health monitoring (SHM). To address these challenges, this study introduces a data reduction approach grounded in probability theory. The proposed method utilizes a relative damage assessment framework to minimize the need for full data storage and processing. Instead of analyzing each strain measurement, this approach leverages the distribution of strain events to estimate potential structural changes. By focusing on cumulative strain event counts at specific threshold levels, it identifies shifts in strain distribution patterns that can be an indication of structural changes. Then, the effectiveness of this approach was validated through real-world data from an in-situ (field) monitoring campaign. This streamlined data interpretation process significantly reduces the volume of data for storage and processing, thereby enabling efficient real-time damage assessment of complex infrastructure systems. Overall, the probability-based data reduction method proposed provides a scalable and responsive solution for SHM systems utilizing FBG sensors, particularly in applications requiring dense sensor networks and continuous monitoring. This approach holds promise for enhancing the scalability and efficiency of SHM systems, especially in large-scale infrastructure projects.

Managing and extracting insights from the large volumes of data generated by optical fiber sensor networks is a major challenge. This paper presents an intelligent, scalable framework for real-time road health monitoring using fiber Bragg grating (FBG) sensor data. The proposed framework reduces reliance on manual data handling and cuts storage needs by over 99 % by constructing a compact health indicator (HI). Data preprocessing and fusion reduce volume and variability, while a wavelet scattering network (WSN) extracts damage-sensitive features that are encoded via a long short-term memory (LSTM) autoencoder to represent the health state. Temperature data is integrated to distinguish structural damage from environmental effects. The approach is evaluated through laboratory fatigue tests and synthetic damage data generated from healthy-state field measurements. Results demonstrate accurate, efficient monitoring with potential for edge deployment, enabling low-cost, real-time, long-term structural health management and representing a significant step toward automated, resource-efficient infrastructure maintenance.

Structural health monitoring (SHM) of infrastructure using sensor networks presents significant challenges, particularly for linear structures that require extensive coverage of critical hotspots. Among the various sensing technologies, optical fiber sensors have recently gained attention as a promising solution for distributed and long-span monitoring due to their ability to provide a high density of sensing points. However, the vast amounts of data generated by these sensors create substantial challenges in data handling, processing, management, and analysis. These challenges are further intensified under random loading and unknown conditions, where discerning patterns becomes particularly difficult. To address these issues, this study proposes a probabilistic-based framework for generating a health indicator (HI) through cumulative loading-time analysis of sensor data. The method reduces data dimensionality by calculating cumulative loading time within predefined windows and strain levels, thereby extracting meaningful features by fitting a cumulative distribution function. These features are then used to construct sensor-specific distributions, and Kullback–Leibler divergence is employed to monitor shifts between a trained baseline distribution and the current distribution. This produces the HI, enabling quantitative tracking of distribution shifts caused by structural changes or long-term anomalies. The proposed approach was validated through experimental fatigue tests, in which strain sensors monitored responses during fatigue progression. Results demonstrated the method’s effectiveness in detecting and localizing damage in two scenarios: when damage occurred directly at sensor locations and when it occurred nearby. Furthermore, the method was evaluated using both healthy-state field data and synthetic damage data generated from a fiber Bragg grating sensor network embedded in a roadway. This real-world scenario, characterized by random and unknown-magnitude loading, further validated the method’s robustness and applicability. Overall, the results demonstrate the potential of the proposed framework for practical deployment in SHM systems, offering efficient monitoring using large-scale sensor networks.

Abstract: New manufacturing techniques like 3D printing are under development, and they need monitoring methods to ensure the quality of the manufactured parts. Artificial Intelligence has outperformed traditional methods in the monitoring process and has shown high potential in recent years. New approaches in Artificial Intelligence, particularly Neural Architecture Search (NAS), have unlocked the potential for automated design of high-performance and resource-efficient deep learning models. In this work, we propose a training-based, low-fidelity NAS framework to systematically discover optimal architectures for regression tasks. Leveraging 8,610 candidate topologies, we trained models on only of the data for 10 epochs, enabling faster execution and selection of the architecture using low-fidelity information. The dataset belongs to Laser Powder Bed Fusion (LPBF), which is a manufacturing method that is still not well mastered and requires many trials before obtaining a satisfactory result. To cope with this issue, we developed a NAS algorithm to design a lightweight AI model (an architecture with a low number of parameters) to predict the process parameters from video information to ensure having the same printing parameters in action. The ultimate goal is then to embed the AI model in a low-latency feedback control loop that enables on-the-fly supervision of the printing process. The final designed architecture is based on 3-dimensional convolutional neural networks. The final AI models are 3–30 times lighter than off-the-shelf ones, while maintaining almost the same accuracy. This shows the potential of our methods when dealing with regression tasks in an industrial case study.

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.

Hyperspectral imaging has a variety of commercially important applications in Earth observation because of its advanced functionalities that enable precise material identification. This manuscript presents the structural design and analysis of a 3U CubeSat carrying a hyperspectral imaging payload, as part of the Earth Moon Education CubeSats (EMEC) mission by EuroMoonMars. Due to its high resource demands, hyperspectral imaging requires an efficient design of the structural subsystem to accommodate the payload and other subsystem components, but publicly available research on this is extremely limited. A CAD model of the complete CubeSat was designed in Autodesk Fusion based on the mass, protrusion, and CG requirements from the dispenser. The design satisfied these requirements, and the total mass of the CubeSat was 6.482 kg after accounting for margins. Structural FE analyses were performed using ANSYS Mechanical to ensure that the CubeSat withstands the vibrational loads (quasi-static loads, random vibrations, and shock loads) from the launch vehicle. After simplifying the geometry of the CAD model to reduce the computational effort, and applying appropriate loads and constraints, the analyses were run on a multi-core processor. The first natural frequency of the CubeSat was 822.69 Hz, which was above the minimum requirement of 115 Hz, indicating that resonance was unlikely. The margins of safety against yielding for the quasi-static loads, random vibrations, and shock loads were 9.95, 3.09, and 0.13, respectively. Based on these preliminary results that showed positive margins of safety, it was concluded that the CubeSat withstands these loads. A sensitivity analysis on uncertain parameters was conducted in optiSLang, and the results of this will be used for validating and updating the FE analyses, to be carried out by testing a physical model of the CubeSat in dedicated setups.

Nowadays, employing deep learning for Structural Health Monitoring is a common practice. However, one of the main challenges here is the lack of data. Several methods have been developed to address this issue. Quantum machine learning is known to be trained faster and with less data, therefore, it could be a suitable option to be used for this purpose. However, since at the current stage limited numbers of qubits can remain stable at the same time, hybrid quantum-classical deep learning approaches can be a replacement. In this study, the benefit of incorporating a quantum layer into a classical deep learner for detecting damage is investigated. For this purpose, a deep learning model with and without a quantum layer is used to predict damage in a wind turbine blade by using ultrasonic inspection data. The results indicate the benefit of employing hybrid quantum-classical ML in detecting damage.

Over the past 30 years, the polymer composite industry has flourished, producing advanced structural materials for the aviation, energy, and transportation sectors. However, the use of crosslinked thermoset matrices has been linked to significant end-of-life challenges, presenting a critical issue for the industry. Moreover, the industry is characterized by numerous labor-intensive processes. In alignment with Industry 4.0 principles, two major routes have been identified to enhance sustainability: the utilization of high-performance thermoplastic matrices and the integration of artificial intelligence in manufacturing. Nevertheless, there are substantial concerns regarding the life cycle assessment of these technologies, which are not accounted for in the initial calculations, including the environmental footprint of polymer synthesis and energy requirements for training AI. This perspective aims to address potential and significant CO₂ emissions from chemical feedstocks and the high computing requirements of these new technologies.

Bone encompasses a complex arrangement of materials at different length scales, which endows it with a range of mechanical, chemical, and biological capabilities. Changes in the microstructure and characteristics of the material, as well as the accumulation of microcracks, affect the bone fracture properties. In this study, two-dimensional finite element models of the microstructure of cortical bone were considered. The eXtended Finite Element Method (XFEM) developed by Abaqus software was used for the analysis of the microcrack propagation in the model as well as for local sensitivity analysis. The stress–strain behavior obtained for the different introduced models was substantially different, confirming the importance of bone tissue microstructure for its failure behavior. Considering the role of interfaces, the results highlighted the effect of cement lines on the crack deflection path and global fracture behavior of the bone microstructure. Furthermore, bone micromorphology and areal fraction of cortical bone tissue components such as osteons, cement lines, and pores affected the bone fracture behavior; specifically, pores altered the crack propagation path since increasing porosity reduced the maximum stress needed to start crack propagation. Therefore, cement line structure, mineralization, and areal fraction are important parameters in bone fracture. The parameter-wise sensitivity analysis demonstrated that areal fraction and strain energy release rate had the greatest and the lowest effect on ultimate strength, respectively. Furthermore, the component-wise sensitivity analysis revealed that for the areal fraction parameter, pores had the greatest effect on ultimate strength, whereas for the other parameters such as elastic modulus and strain energy release rate, cement lines had the most important effect on the ultimate strength. In conclusion, the finding of the current study can help to predict the fracture mechanisms in bone by taking the morphological and material properties of its microstructure into account.

To design a more efficient energy absorber, it is critical to evaluate how changing the design parameters affects its performance, and also determine each one’s order of significance. In this paper, using a new approach, the behavior and response of straight, double-tapered, and triple-tapered thin-walled tubes with rectangular cross sections under axial and dynamic loading are investigated by performing a sensitivity analysis on a support vector machine (SVM) as a surrogate machine learning model. First, a finite element model of the energy absorber is constructed and validated with available experimental and theoretical studies. Next, a design of experiments was developed using the Sobol series sampling method and an appropriate dataset was created. This information is then used to develop an SVM model to predict the initial peak load and mean load of tubes. The accuracy of the machine learning created in this study is then assessed, and it is demonstrated that the developed model can precisely predict the performance of the absorber. The machine learning model is then subjected to a Sobol sensitivity analysis, and the outcomes are compared to those of the parametric study. The results suggest that the thickness of the tube has a stronger effect on the absorber performance than other geometric parameters. Comparing the effects of different material parameters on the behavior of tubes, the results show that yield strength has the greatest impact on the response of the energy absorber. It is also observed that the tapered tubes have a much lower initial peak load compared to straight ones.

Nowadays, using vibration data in conjunction with pattern recognition methods is one of the most common fault detection strategies for structures. However, their performances depend on the features extracted from vibration data, the features selected to train the classifier, and the classifier used for pattern recognition. Deep learning facilitates the fault detection procedure by automating the feature extraction and selection, and classification procedure. Though, deep learning approaches have challenges in designing its structure and tuning its hyperparameters, which may result in a low generalization capability. Therefore, this study proposes an ensemble deep learning framework based on a convolutional neural network (CNN) and Dempster–Shafer theory (DST), called CNN-DST. In this framework, several CNNs with the proposed structure are first trained, and then, the outputs of the CNNs selected by the proposed technique are combined by using an improved DST-based method. To validate the proposed CNN-DST framework, it is applied to an experimental dataset created by the broadband vibrational responses of polycrystalline nickel alloy first-stage turbine blades with different types and severities of damage. Through statistical analysis, it is shown that the proposed CNN-DST framework classifies the turbine blades with an average prediction accuracy of 97.19%. The proposed CNN-DST framework is benchmarked with other state-of-the-art classification methods, demonstrating its high performance. The robustness of the proposed CNN-DST framework with respect to measurement noise is investigated, showing its high noise-resistance. Further, bandwidth analysis reveals that most of the required information for detecting faulty samples is available in a small frequency range.

Measurements are not exactly accurate, and measurement errors could lead to a biased trained classifier, and finally to a wrong classification of the parts. This paper extends the recently proposed (Integrated) Mahalanobis Classification System with the concept of Interval Mahalanobis distance (IMD) in order to account for measurement uncertainty. This novel Integrated Interval Mahalanobis Classification System (IIMCS) is applied to an experimental case study of complex shaped metallic turbine blades with various damage types. The turbine blades have been vibrationally tested in a wide frequency range. The IIMCS selects a subset of optimal features that contribute the most to the system under the framework of Binary Particle Swarm Optimization, and determines the optimal decision threshold based on Particle Swarm Optimizer. A Monte Carlo method (MCM) is implemented to account for measurement uncertainty, and as such yields an indicator of reliability, implying the confidence level of the classification results. The obtained results illustrate a high performance of the IIMCS for classifying turbine blades based on vibrational response data with measurement uncertainty.

IntraOcular Pressure (IOP) is one of the most informative factors for monitoring the eye-health. This is usually measured by tonometers. However, the outputs of the tonometers depend on the physical and geometrical properties of the cornea. Therefore, the common practice is to develop a numerical model to generate some correction factors. The main challenge here is the accuracy and efficiency of a numerical model in predicting the IOP and Dynamic Corneal Response (DCR) of each patient. This study addresses this issue by developing a two-step surrogate model based on adaptive sparse Polynomial Chaos Expansion (PCE) for fast and accurate prediction of the IOP. In this regard, first, an FE model of the cornea has been developed to predict the DCR parameters. This FE model has been replaced with a PCE-based surrogate model to speed up the simulation step. The uncertainties in the geometry and material model of the cornea have been propagated through the surrogate model to estimate the distributions of the DCR parameters. In the second step, the combination of DCR parameters and the input parameters provide a proper parameter space for developing an efficient data-driven PCE model to predict the IOP. Moreover, sensitivity analysis by using PCE-based Sobol indices has been performed. The results demonstrate the accuracy and efficiency of the proposed method in predicting the IOP. Sensitivity analysis revealed that IOP measurement was influenced mostly by deflection amplitude and applanation time. The analysis indicates the importance of the interactions between the parameters.

This paper studies the behavior and response of triple thin-walled tubes with rectangular cross-sections under axial and dynamic loading. First, a finite element model of the energy absorber is prepared, and the results are validated with available theoretical and experimental studies. Then the effect of different input parameters such as tube-thickness, cross-sectional ratio, slope angle, and material parameters on the performance of thin-walled energy absorbers is studied.The simulation results show that the thickness of the tube has a more significant effect on the absorber’s performance than other geometric parameters. The results also show that changing the cross-sectional ratio and the inclination angle of the tube changes the initial peak load more than the average load of the absorber. Comparing the effects of different materials on the performance of absorbers, the results show that steel alloys record the highest average loads and initial peak loads, followed by titanium alloys and then aluminum alloys. This information is then used to develop a machine learning model to predict the performance of the absorbers. Then, the performance of the machine learning developed in this study is evaluated, and it is shown that the developed machine learning can accurately predict the absorber’s performance. Finally, a Sobol sensitivity analysis is performed on the machine-learned model and the results are compared with those of parametric study.

Vibration-based quality monitoring of manufactured components often employs pattern recognition methods. Albeit developing several classification methods, they usually provide high accuracy for specific types of datasets, but not for general cases. In this paper, this issue has been addressed by developing a novel ensemble classifier based on the Dempster-Shafer theory of evidence. In the proposed procedure, prior to DST combination, three steps should be taken: (i) selection of proper classifiers by maximizing the joint mutual information between predicted and target outputs, (ii) optimal redistribution of the classifiers’ outputs by considering the distance between the predicted and target outputs, (iii) utilizing five different weighting factors to enhance the fusion performance. The effectiveness of the proposed framework is validated by its application to 13 UCI and KEEL machine learning datasets. It is then applied to two vibration-based datasets to detect defected samples: one synthetic dataset generated from the finite element model of a dogbone cylinder, and one real experimental dataset generated by collecting broadband vibrational response of polycrystalline Nickel alloy first-stage turbine blades. The investigation is made through statistical analysis in presence of different noise levels. Comparing the results with those of five state-of-the-art fusion techniques reveals the good performance of the proposed ensemble method.