This research investigates the application of machine learning, specifically Gradient Boosting, to predict the structural performance of main columns in high-rise buildings subjected to various vertical and lateral loading conditions. The study focuses on utilizing Gradient Boosting, an ensemble learning technique, to address the nonlinear behavior of structural components and predict key parameters such as vertical displacement (Dis(vert)), lateral displacements in the X and Y directions (Dis(X), Dis(Y)), maximum axial normal stress (Avr(sig-max)), and maximum shear stress (avr(tau-max)). A comprehensive GridSearchCV process, involving 3 folds for each of 972 candidates, totaling 2916 fits, was used for hyperparameter optimization to enhance model accuracy. The model demonstrated exceptional performance, with R² values consistently above 0.99 for all target variables, indicating its high reliability in predicting structural behavior. The results show that the model can accurately predict displacement and stress values, which are critical for evaluating the stability and integrity of high-rise buildings under dynamic loads. Additionally, the model exhibited strong potential for early damage detection and predictive maintenance, enabling the identification of structural anomalies before they lead to critical failure. The ability to predict future deterioration trends further supports the development of proactive maintenance strategies. Overall, this study highlights the significant potential of machine learning in Structural Health Monitoring (SHM), offering a powerful tool for real-time monitoring, predictive maintenance, and safety management in high-rise buildings. Future research could expand the model’s capabilities by integrating additional environmental data or exploring its applicability to other types of structures.

Smart composite materials with integrated sensing layers are gaining attention for their potential to improve structural health monitoring and damage detection in high-performance applications. This study experimentally evaluates validated advanced simulation techniques to investigate impact-induced damage in such composites, with a particular focus on barely visible impact damage (BVID). A refined finite element model is developed using user-defined cohesive materials to capture both matrix cracking and delamination, which are critical to understanding damage mechanisms associated with BVID. The model is applied to hybrid laminates incorporating surface-integrated sensing layers composed of ultra-high modulus carbon and S-glass fibres. These layers are designed to show visible signs of damage that can be correlated with internal failure mechanisms. The simulation results are compared against experimental data, including C-scan imaging and surface inspections, to assess accuracy in predicting damage initiation, growth, and patterns. Particular attention is given to the effects of through-thickness compression and the interaction between different failure modes. This work offers practical insights for reducing reliance on costly testing during the early stages of material and structural design for smart composites.

Mechanical metamaterials can amplify deformation through geometry-driven mechanisms, transforming small global strains into localized regions of enhanced strain without altering the constituent material. This capability is attractive for applications requiring controlled deformation and mechanical responsiveness, including sensing, actuation, energy harvesting, and adaptive structures. Although numerous auxetic lattice architectures have been proposed, their reported mechanical performances are difficult to compare because previous studies have employed different material models, geometric definitions, loading conditions, and evaluation methodologies. Consequently, quantitative design guidelines for selecting the most appropriate architecture remain limited. This study addresses this gap through a unified finite element framework that systematically compares four representative auxetic lattice architectures (re-entrant, star-shaped, arrowheaded, and chiral) under identical material properties, geometric scaling, boundary conditions, loading conditions, and performance metrics. A hyperelastic poly(butylene adipate-co-terephthalate) constitutive model was used to investigate 36 geometric configurations by varying cell angle and structural thickness. Mechanical performance was evaluated in terms of strain amplification factor (SAF), effective Poisson's ratio, effective stiffness, stress distribution, and relative density. The results reveal pronounced architecture-dependent trade-offs, with lattice topology producing nearly a six-fold variation in strain amplification (SAF ≈ 1.4–8). The star-shaped and re-entrant lattices achieved the highest strain amplification, whereas the arrowheaded lattice consistently exhibited the lowest strain amplification but the highest stiffness and most negative effective Poisson's ratio. The resulting Ashby-style performance maps are reported to establish a quantitative design framework for balancing strain amplification, auxeticity, stiffness, and lightweighting, providing practical guidance for selecting auxetic lattice architectures according to application-specific mechanical requirements.

This research investigates the application of machine learning, specifically Support Vector Machine (SVM) regression with a Radial Basis Function (RBF) kernel, to predict the structural performance of internal beams in high-rise buildings subjected to earthquake-induced lateral forces. The study focuses on utilizing SVM to address the nonlinear behavior of structural components and predict key parameters such as lateral displacements in the X and Y directions (Disp(X), Disp(Y)), maximum axial normal stress (Avr(sig-max)), and maximum shear stress (Avr(tau-max)). A comprehensive hyperparameter tuning process, using GridSearch optimization with five-fold cross-validation, was conducted to enhance model accuracy and prevent overfitting. The model demonstrated exceptional performance, with R2 values exceeding 0.91 across all target variables, indicating its high reliability in predicting structural behavior. The results show that the model can accurately forecast displacement and stress distributions, which are critical for evaluating the stability and integrity of internal beams under seismic loading conditions. Additionally, the model exhibited strong potential for early damage detection and predictive maintenance, enabling the identification of structural anomalies before they lead to critical failure. The ability to predict future deterioration trends further supports the development of proactive maintenance strategies. Overall, this study highlights the significant potential of machine learning in Structural Health Monitoring (SHM), offering a powerful tool for real-time monitoring, predictive maintenance, and safety management in high-rise buildings. Future research could expand the model's capabilities by incorporating real-time sensor data and exploring its applicability to diverse structural configurations.

Field measurement of structural displacement and inclination is hindered by drift in double integration, stringent filtering needs, and the limited compute/bandwidth of low-cost microcontrollers. This work presents node based on an ADXL345 tri-axial accelerometer with on-board processing that estimates dynamic tilt and symmetry in real time without double integration. Three real-time filters, Butterworth IIR (BWF), finite impulse response (FIR), and moving average (MAF, uniform-tap FIR), were implemented on the device and benchmarked against an offline Savitzky–Golay reference. A rigid-body rotation model links off-centre acceleration to inclination and defines a dimensionless rotation index for symmetry assessment. Calibration against analytical motion identified a linear-phase FIR as optimal, yielding the lowest RMSE over 0.5–8 Hz while preserving waveform shape. Computational profiling on an 11.0592 MHz microcontroller measured average per-sample execution of 2.5 µs (MAF), 6 µs (FIR), and 12 µs (BWF), enabling 800 Hz local sampling with ≥ 100 Hz wireless streaming and an 8 × reduction via on-device decimation (no compression). Under cyclic tension, lateral acceleration quantified asymmetry: non-cracked and two-sided-cracked specimens showed comparable lateral levels, whereas one-sided-cracked specimens exhibited markedly higher values consistent with crack-induced rotation. Vertical acceleration agreed with acceleration reconstructed from displacement, with peak deviations of ∼ 9–14%; accuracy decreased at the lowest acceleration levels, consistent with stronger low-frequency spectral content. Tilt from acceleration tracked displacement-based tilt with minor underestimation at the smallest amplitudes. Overall, the node delivers embedded, phase-faithful filtering and tilt/symmetry estimation with quantified computational cost, making acceleration-based monitoring practical on resource-constrained hardware while avoiding big-data burdens.

The construction sector faces a critical need to minimize its carbon footprint, which is currently stimulating the development of geopolymer concrete using recycled coarse aggregates as an eco-friendly material compared with Portland cement. Accurate prediction of the compressive strength of this eco-efficient concrete is complex, however, as a result of the complex, non-linear interactions between many of the mix-design and curing parameters. Although modern scientific literature and engineering practices have increasingly adopted machine learning (ML) for concrete strength prediction, a significant scientific gap remains. Most existing studies rely on “black-box” models that lack sufficient interpretability and frequently overlook the severe risk of data leakage during validation, limiting their practical engineering application. To address this gap, this study proposes a robust, data-leakage-aware framework driven by a rigorous nested GroupKFold cross-validation strategy. By grouping concrete samples by their unique Mix_ID, this approach ensures genuine generalization to entirely unseen mixtures. Within this reliable validation scheme, the CatBoost algorithm is utilized for compressive-strength prediction, with the Dung Beetle Optimizer (DBO) serving as an effective tool for hyperparameter tuning. The evaluation results across multiple random seeds show that the DBO–CatBoost model significantly outperforms the default CatBoost, rigorously tuned baseline models (Support Vector Regression and Random Forest), and a comparative metaheuristic benchmark (PSO–CatBoost). It achieves the most stable distribution of errors and excellent predictive accuracy (Test (Formula presented.), RMSE = (Formula presented.) ). In addition, the model predictions were demystified using the methods of SHapley Additive exPlanations (SHAP) and partial dependence plots (PDPs). The interpretability analysis revealed strong statistical associations, showing that Curing Time and Coarse Aggregate are the most prominent predictive features and the strongest pairwise interaction between each other; the NaOH molar concentration is the most important second-level influence on optimization of strength. Overall, the framework provides a robust data-driven screening tool that can assist in preliminary mix-design evaluation. By reducing the reliance on extensive empirical “trial and error” approaches, this predictive model supports more efficient material usage and facilitates preliminary optimization of low-carbon concrete formulations. Theoretically, this study advances the fundamental science of geopolymer materials by explicitly quantifying the complex, non-linear interactions between alkaline activators, curing conditions, and recycled aggregates. This provides a robust data-driven theoretical foundation for designing and optimizing next-generation eco-friendly concrete products and structures.

Auxetic cementitious cellular composites (ACCCs) exhibit high compressive deformation with recoverable behavior, demonstrating strong potential for self-sensing applications. In this study, compliant piezoresistive ACCCs were engineered to combine auxetic behavior with piezoresistive properties by reinforcing a cementitious matrix with polyvinyl alcohol (PVA) and carbon fibers. Uniaxial compression tests were conducted to evaluate their piezoresistive responses in damage sensing across distinct deformation stages, while Digital Image Correlation (DIC) and Acoustic Emission (AE) techniques were employed to investigate corresponding deformation and damage mechanisms. Cyclic loading tests further assessed the flexible sensing behavior of ACCCs within the recoverable deformation during the auxetic behavior range, with further insights obtained from microscopic analysis and X-ray computed tomography (CT). The piezoresistive sensing behavior of two typical ACCC specimens with different geometries was compared. The results demonstrate that the auxetic behavior promotes high compressive deformation, which can be delineated into four distinct mechanical stages. Each stage corresponds to specific damage mechanisms and exhibits characteristic piezoresistive responses, as reflected by stage-dependent variations in the fractional change in resistance (FCR). This staged behavior effectively extends the sensing range for damage detection. Under cyclic loading, the composites exhibit a reproducible piezoresistive response across an extended self-sensing range, attributable to their substantial recoverable deformation. This behavior highlights their superior strain and stress sensitivity, although residual plastic deformation progressively increases with greater loading amplitude. Given wide availability of cementitious materials, highly deformable ACCCs with piezoresistive strain/stress and damage sensing capabilities offer a cost-effective and sustainable solution for in-situ, real-time structural health monitoring.

Lightweight mechanical metamaterials have attracted increasing attention due to their ability to achieve unusual mechanical properties through architectural design rather than chemical composition. In this study, a novel cylindrical metamaterial based on a gyroid triply periodic minimal surface (TPMS) architecture is proposed, selected for its high strength-to-weight ratio, smooth stress distribution, and fully interconnected porosity that are advantageous for load-bearing biomedical applications. A combined experimental and numerical approach was employed, where the finite element model was validated against quasi-static compression tests conducted on 3D-printed specimens (1.9 mm thickness, eleven unit-cells). The validated model achieved excellent agreement with experiments (R2 = 0.965), capturing the two-phase deformation behavior: an initial elastic regime followed by plasticity-driven global buckling initiating at the core between neighboring unit-cells. A comprehensive parametric study was then performed to investigate the effects of unit-cell wall thickness (1.1–2.7 mm) and number of unit-cells per circumference (3, 7, and 11) on the mechanical response. Results show that increasing the unit-cell count enhances stiffness and Specific Energy Absorption (SEA) but reduces the displacement at which buckling initiates. In contrast, increasing wall thickness simultaneously improves stiffness, peak load, SEA, and instability displacement, with SEA increasing by up to 201% from the thinnest to the thickest configuration. The highest performance was achieved for the 2.7 mm thickness with eleven unit-cells, yielding a peak load of 3679.9 N, stiffness of 4495.2 N/mm, and SEA of 3331.0 mJ/g. The tunable mechanical behavior and lightweight porous architecture position the proposed TPMS metamaterial as a promising candidate for further investigation in load-bearing biomedical applications.

This study evaluates the performance of damaged concrete beams retrofitted with a purpose-designed mechanochromic composite, which provides structural reinforcement and visual feedback for structural health monitoring (SHM). The retrofitting process utilizes externally bonded reinforcement (EBR) on pre-damaged concrete prisms. The mechanochromic composite, a thin-ply hybrid material made of unidirectional ultra-high modulus (UHM) carbon/epoxy and S-glass/epoxy layers, changes color to indicate structural overload when the UHM carbon layer fractures due to excessive strain. Eighteen concrete specimens were prepared and subjected to four-point bending tests, assessing various combinations of damaged, undamaged, retrofitted, and non-retrofitted configurations. Results showed that the mechanochromic composite functions effectively as both a passive visual sensor and reinforcement. For instance, a 5 % crack depth reduced load-bearing capacity by 30 %, however, retrofitting with the mechanochromic composite improved load-bearing capacity by up to 208 % compared to undamaged beams. The study further discusses the effects of different damage levels on load-bearing capacity through flexural strength, load-displacement curves, and failure modes.

Modeling complex, non-stationary dynamics remains challenging for deterministic neural networks. We present the Chaos-Integrated Synaptic-Memory Network (CISMN), which embeds controlled chaos across four modules—Chaotic Memory Cells, Chaotic Plasticity Layers, Chaotic Synapse Layers, and a Chaotic Attention Mechanism—supplemented by a logistic-map learning-rate schedule. Rigorous stability analyses (Lyapunov exponents, boundedness proofs) and gradient-preservation guarantees underpin our design. In experiments, CISMN-1 on a synthetic acoustical regression dataset (541 samples, 22 features) achieved R ² = 0.791 and RMSE = 0.059, outpacing physics-informed and attention-augmented baselines. CISMN-4 on the PMLB sonar benchmark (208 samples, 60 bands) attained R ² = 0.424 and RMSE = 0.380, surpassing LSTM, memristive, and reservoir models. Across seven standard regression tasks with 5-fold cross-validation, CISMN led on diabetes (R ² = 0.483 ± 0.073) and excelled in high-dimensional, low-sample regimes. Ablations reveal a scalability–efficiency trade-off: lightweight variants train in <10 s with >95% peak accuracy, while deeper configurations yield marginal gains. CISMN sustains gradient norms (~2300) versus LSTM collapse (<3), and fixed-seed protocols ensure <1.2% MAE variation. Interpretability remains challenging (feature-attribution entropy ≈ 2.58 bits), motivating future hybrid explanation methods. CISMN recasts chaos as a computational asset for robust, generalizable modeling across scientific, financial, and engineering domains.

Purpose – This study aims to automate the visual inspection of piling sheets in water channel construction using artificial intelligence (AI). By employing image classification and object detection techniques, the research focuses on extracting and analysing geometric features to enhance the accuracy and efficiency of the inspection process. It also addresses key challenges associated with the unique characteristics of construction materials and the limited variability of available inspection datasets. Design/methodology/approach – Convolutional neural networks (CNNs) with varying complexities are employed for image classification, across four and six classes, and for object detection of piling sheets in water channel environments. A dataset provided by Witteveen + Bos is preprocessed to generate training sets, and the CNN architectures are optimized for enhanced performance. The accuracy and efficiency of the proposed models are evaluated and compared against traditional manual inspection methods. Findings – The AI-driven approach significantly reduces processing time, evaluating 40, 000 images in just 11.9 h, compared to approximately one month using manual assessment. The 4-class classification model achieves an accuracy of 96%, while the 6-class model attains 72%. The object detection model produces a mean average precision (mAP) of 79%. These results meet the performance standards set by the Dutch company Witteveen + Bos, which demonstrate the effectiveness of AI in automating the inspection of piling sheets. Originality/value – This study introduces a novel AI-based approach for assessing piling sheets, demonstrating substantial improvements over traditional inspection methods. It introduces a systematic evaluation of various CNN architectures and hyperparameters to optimize the models specifically for piling sheet inspection rather than relying on off-the-shelf solutions. The use of CNNs for both image classification and object detection adheres to relevant Dutch engineering standards. Notably, the reduction in processing time, from one month to around 12 h, represents a major advancement in the efficiency of civil engineering inspections.