Setback tapering and facade twisting are well-established aerodynamic optimization techniques for supertall buildings. This study reveals that these designs may shift the dominant wind load direction from crosswind to along-wind, significantly influencing vibration reduction. More importantly, this study demonstrates that combining setback tapering with rotation designs yields synergistic effects, achieving superior aerodynamic performance beyond individual strategies. Through synchronized pressure-measurement wind-tunnel tests, the study systematically investigates the effects of setback tapering layers and setback-rotation combinations on along-wind aerodynamic characteristics and wind-induced responses, analyzing variations in shape coefficient, aerodynamic force spectrum, base moment, and accelerations. Results indicate that setback designs reduce shape coefficient and base moment by decreasing windward area, with the setback-rotation combination—particularly the upper-lower rotation design—achieving optimal performance (54.4% reduction in shape coefficient). Mechanism analysis reveals that setback tapering creates layered terraces along the facade, inducing localized flow separation and reattachment, weakening vertical airflow coherence and reducing spectral energy in the upper region. Rotation design introduces three-dimensional helical flows and asymmetric wake interference, disrupting vortex shedding coherence and suppressing dominant-frequency energy. Their combined effects form a "layered interference and three-dimensional shear mechanism," reducing fundamental-frequency base moment PSD by over 50% and achieving up to 44.6% reduction in base moment response under a 50-year return period wind speed. Notably, setback tapering reduces static wind loads but may enhance crosswind vortex excitation, increasing overall acceleration. In contrast, the dual setback design with middle rotation achieves an optimal balance between along-wind load reduction and crosswind vibration suppression by simultaneously weakening dominant-frequency energy and generalized force input. This configuration effectively mitigates along-wind loads while suppressing crosswind vibrations, highlighting the novelty and engineering value of the proposed synergistic design strategy for aerodynamic optimization of complex supertall buildings.

Crack detection in underwater concrete structures relies heavily on optical imaging, where high-quality images are critical for accurate detection. However, due to optical absorption, scattering, and water turbidity, underwater crack images commonly suffer from color distortion, low contrast, and blurred textures, significantly affecting detection performance. To address these challenges, this paper proposes PG-UNIT (physics-guided unsupervised image-to-image translation), a framework for underwater crack image enhancement. A self-supervised pre-correction module, guided by an underwater optical model, adaptively estimates light attenuation parameters and performs initial corrections to reduce color cast and illumination imbalance, ensuring physically consistent inputs for downstream enhancement. The improved UNIT module integrates multi-scale convolution, selective kernel attention, and a hybrid multi-attention mechanism to enhance structural feature extraction and detail restoration. Furthermore, color consistency loss and underwater light reflection loss are introduced to improve visual realism and physical plausibility. Experimental results show that PG-UNIT outperforms state-of-the-art methods, achieving improvements of 48.2 %, 58.0 % and 70.0 % in BRISQUE, NIQE and PIQE, respectively. The proposed method demonstrates strong generalization and robustness under diverse illumination and turbidity conditions, providing reliable visual inputs for automated crack detection and quantification in underwater concrete structures.

Designing hydrogel-based composites for internal curing in cementitious systems remains challenging due to the persistence of polymer phases as discrete particles, which leads to hydrogel-related pore formation and reduces mechanical properties. In this study, a basalt-polyacrylamide hydrogel composite (BCPA) was engineered by introducing micron-sized basalt particles as physical crosslinking centers to construct a polymer-particle interface whose stability is responsive to the surrounding ionic environment. Upon exposure to the ion-rich cement pore solution, cation adsorption on basalt surfaces is proposed to weaken the original polymer-basalt association, leading to a proposed interfacial self-debonding process and inferred redistribution of polyacrylamide segments within the cement matrix. This interfacial evolution induces a transition of the polymer phase from a discrete particulate state to a dispersed segmental configuration, thereby altering the structural response of the hydrogel composite after incorporation. As a consequence, hydrogel-related marcopores are markedly suppressed. The inferred redistribution of polymer segments allows internal curing functionality to be retained while enabling a refined pore structure that can be readily accommodated by hydration products. At a dosage of 1 wt%, BCPA reduces 7-day autogenous shrinkage by approximately 75% and decreases cracking area by nearly one order of magnitude, while limiting compressive strength loss within 20%, substantially lower than the approximately 44% loss observed for the chemically crosslinked polyacrylamide reference (a typical internal curing agent). These findings suggest that regulating interfacial interactions to modulate polymer state evolution offers a promising strategy for designing hydrogel composites that reconcile shrinkage mitigation with mechanical integrity in cementitious systems.

Conventional variable frequency pendulum isolators (VFPIs) and variable friction pendulum bearings (VFPBs) require curvature radius and surface roughness to vary continuously along the radial direction, making precise control difficult in practice. This study proposes a segmented friction pendulum bearing (SFPB) with two independent discs, each allowing individual design of curvature and roughness to achieve diverse nonlinear hysteretic behaviors. Based on the dynamic equations for a tall building isolated with the SFPB, the dynamic responses including displacement, acceleration and shear force under seismic excitation and wind loads were investigated separately. Furthermore, multi-objective optimization analyses were conducted for the isolated tall building considering seismic excitation alone, strong wind loads alone, and both earthquake and strong wind. The optimal design parameters of SFPB obtained under each loading scenario were compared. It indicates that the parameters obtained by considering both seismic and wind actions simultaneously achieve satisfactory performance in both seismic isolation and wind resistance. For instance, the building top displacement is reduced by 56 % under rare earthquakes, and the peak acceleration at the building top is reduced by 41 % under high wind speeds.

Current wind pressure prediction methods for low-rise buildings primarily address the determination of wind pressure when a single geometric parameter of building changes, with limited research on scenarios where multiple geometric parameters vary simultaneously. This paper establishes a multi-parameter deep learning model for predicting wind pressure statistical values of low-rise buildings. The model is trained and tested for generalization capability using the TPU aerodynamic database. The influence of whether the eave height and roof slope of the tested buildings are included in the training dataset on prediction accuracy, as well as the accuracy of extrapolation predictions, are discussed. Furthermore, the impact of training set size on prediction accuracy is investigated, and the model's predictive capability for low-rise buildings with arbitrary eave heights and slopes but the same planar dimensions, as well as its adaptability to buildings with different planar dimensions, is demonstrated. The study by 15 scenarios shows that the proposed model exhibits strong generalization capability for low-rise buildings with multiple varying geometric parameters. Whether the eave height and roof slope of the tested buildings are included in the training dataset has no significant impact on prediction accuracy, and the accuracy of extrapolated predictions remains relatively stable.

Although the problems of wind pressure prediction have been extensively studied, the prediction on real complicated large-span roofs has received limited attention, in which previous studies focused on error analysis at local taps rather than across the entire roof. This paper addresses this research gap by introducing a Bayesian optimized deep neural network (BO-DNN) method for predicting the mean, root mean square (RMS), and minimum wind pressure coefficients across the entire surface of a real large-span circular roof of Guangzhou International Sports and Performing Arts Center (GISA) under untested wind directions. The effectiveness of BO-DNN for the real large-span roof is thoroughly assessed by various error metrics for the entire roof. The correlation coefficients achieve 0.995 with mean square error (MSE) less than 0.003, and the predicted contours are very close to the truth contours even in the area with abrupt geometric changes and steep pressure gradients. More than 90% of predicted samples of minimum wind pressure coefficients dominating the wind-resistant design have relative errors (REs) less than 10%, and the taps with REs > 20% are located outside the area with high pressure gradients. The machine learning interpretability tool Shapley additive explanations (SHAP) demonstrates that the vertical geometric variation of the roof has the most impact on the prediction results, and the algorithm comparison indicates that BO-DNN outperforms random forest (RF) and XGBoost (XGB) for wind pressure prediction.

Structural health monitoring systems provide critical information for assessing the safety of high-rise buildings, yet their effectiveness can be compromised when data become incomplete or corrupted during extreme events such as earthquakes and strong winds. Reconstructing unavailable responses from the remaining channels is thus essential to maintain reliable condition assessment and timely decision-making. This study introduces a PyramidLong Short-Term Memory-Attention (P-LSTMA) architecture, which combines a hierarchical pyramid for multiscale feature fusion, LSTM-based temporal memory, and an attention mechanism to capture rapid transients and long-term temporal dependencies in structural responses. The proposed model was evaluated on numerically simulated wind- and earthquake-induced response datasets under both single-domain and cross-domain transfer settings, consistently outperforming baseline approaches in reconstruction accuracy. Finally, the robustness and practical efficacy of P-LSTMA were validated through two real-world case studies: a 32-story residential building in Los Angeles subjected to seismic events, and Guangzhou's 303 m Leatop Plaza during a typhoon.

Nonlinear dynamic problems are ubiquitous in engineering applications, and accurately solving their governing equations is essential for understanding system behavior. Physics-informed neural networks (PINNs) have emerged as a new computing paradigm for solving partial differential equations. However, conventional PINNs struggle to predict accurate solutions for dynamic systems affected by strong nonlinearity, damping, and spatiotemporal coupling. To address this challenge, this work proposes a nonlinear vibration stepping PINN (NVS-PINN) approach for analyzing the complex nonlinear behavior of dynamic systems. This approach introduces a trainable scaling parameter within each time segment to adaptively adjust the network output. Furthermore, the hyperbolic tangent function is adopted as hard constraints to ensure that the network output consistently satisfies the initial and/or Dirichlet boundary conditions of nonlinear dynamic systems. Three illustrative examples, including a single-degree-of-freedom parametric Duffing oscillator, a two-degree-offreedom nonlinear damped vibratory system, and a nonlinear elastic circular arch under wind load, are considered for validation. Numerical results demonstrate that the NVS-PINN approach can accurately predict long-duration nonlinear vibration responses, including complex multi-stable limit cycles, escape motions, and wind-induced vibrations. For the Duffing oscillator, the NVS-PINN approach achieves highly accurate results, with average relative errors of 5.3444 x 10-3 for the strongly nonlinear system and 4.9573 x 10-4 for the strongly nonlinear system with damping. For the elastic arch under wind load, incorporating a trainable scaling parameter in NVS-PINN reduces the maximum relative error by about 82.6 %. Moreover, using smaller time intervals can improve network accuracy.

The structural behaviour and load-carrying capacity of recycled aggregate concrete-filled stainless steel tubular (RACFSST) columns under eccentric loading are investigated through laboratory tests and numerical simulations in this paper. The experimental programme included material tests on the concrete and stainless steel tubes, measurements of initial global geometric imperfections and eccentric compression tests on twelve RACFSST column specimens, which were designed with four levels of initial loading eccentricities and three types of concrete. The test results, including failure loads, failure modes and full load-deformation histories, were reported and analysed in detail. It was found that failure load decreases with initial loading eccentricity, with the maximum reduction of 33.2 % observed, while ductility increases with recycled coarse aggregate replacement ratio, with the maximum increase of 40 % obtained. Finite element models were developed and validated based on the test data, and subsequently used for parametric studies to generate supplementary numerical data. Based on both test and numerical results, the applicability of relevant design rules, as specified in the Chinese technical code, European code and American specification, was evaluated. The assessment results indicated that the European code provided relatively accurate and consistent failure load predictions for eccentrically loaded RACFSST columns, while the Chinese technical code and American specification tended to yield slightly conservative and scattered predictions. In response, a revised design method was proposed based on the Chinese technical code, resulting in an increase of 9.6 % in accuracy and an improvement of 46 % in consistency of failure load predictions for RACFSST columns under eccentric loading.

In structural optimization, data-driven surrogate models are often explored as alternatives to finite element analysis to reduce computational cost. However, conventional neural networks usually fail to capture key structural characteristics and are limited to predicting global responses (e.g., top displacement), but usually fail to achieve accurate internal force predictions with conventional training data volumes. As a result, most existing studies involving surrogate models did not concern internal force constraints. To address this issue, this study proposes a structural optimization framework based on a pre-trained Physics-Informed Neural Network (PINN) surrogate model. By embedding static equilibrium equation into the loss function, the model achieves higher predictive accuracy, particularly for internal forces, while pre-training accelerates convergence and enhances stability. Combined with an improved multi-swarm particle swarm optimization (MPSO) algorithm, the framework enables efficient optimization of multi-story frame structures under internal force and multiple other constraints. The application to a six-story frame structure validates its effectiveness: compared with a DNN-based model, the PINN-based model improves the coefficient of determination for internal force prediction from 0.8874 to 0.9937. These results demonstrate that the proposed method offers a promising approach for efficient optimization of multi-story frame structures.

In April 2024, Guangzhou, China, was struck by a tornado, resulting in extensive damage to infrastructure and buildings. To assess the impact of this event, a research team conducted a rapid forensic investigation focusing on areas with clearly defined and uniform building types. Forensic analysis indicated that the tornado was rated as an EF3 on the Enhanced Fujita (EF) Scale, with numerous low-rise buildings exhibiting diverse damage states, including roof uplift, facade destruction and overall collapse. The airtightness or ventilation status of buildings during the tornado had a significant impact on their resistance to wind-induced loads. To further investigate the impact of this condition on wind-induced building damage, numerical simulations focusing on low-rise structures were conducted. The results reveal that for sealed buildings, the pressure drop caused by the tornado was a critical factor leading to roof failure. Conversely, for buildings with single-sided openings, the rapid influx of air significantly increased internal pressure. Failure of peripheral walls alters the internal flow characteristics of low-rise buildings, leading to an increase in roof loading by more than 60% compared to sealed structures. These findings highlight the importance of maintaining the integrity of wall systems during tornado events.

The good performance of vibration reduction by friction pendulum system-tuned mass damper (FPS-TMD) installed on buildings has been confirmed. To prevent excessive displacement of the slider under extreme conditions, a restraining rim is commonly set at the edge of the sliding plate. Consequently, the collision between the slider and the restraining rim becomes inevitable. This study establishes a bidirectional coupled dynamic model for an FPS-TMD with a restraining rim mounted on a single-story frame or a tall building, using the Euler-Lagrange method. The collision between slider and restraining rim is modeled as a non-linear viscoelastic force, and the friction is related to the state variables and external forces of the slider. To demonstrate the proposed theoretical model, the unidirectional and synchronous bidirectional motions of a given parameter system under harmonic loads are analyzed. The influences of parameters associated with FPS-TMD and the effects of bidirectional motion on structural displacement and energy dissipation are investigated. Finally, based on the genetic algorithm, the vibration reduction effectiveness of the FPS-TMD with optimal parameters on a tall building under recorded earthquake waves is evaluated. This study provides a theoretical foundation for better understanding the friction and collision effects of FPS-TMD with restraining rim.

Underwater crack segmentation is crucial to assessing the integrity of submerged concrete structures, yet optical degradations such as scattering, turbidity, and uneven illumination impede accuracy. This study proposes a novel entropy-guided and adaptive foundation model-based framework for high-fidelity underwater crack segmentation. The framework incorporates a multiscale entropy-guided prompting mechanism that automatically generates structure-aware cues for unsupervised crack localization, followed by a dynamically gated mixture of low-rank adaptation experts for efficient fine-tuning, and a frequency-aware enhancement strategy based on directional wavelet compensation to recover high-frequency details and preserve structural integrity. Extensive experiments on in-house underwater datasets demonstrate that the proposed framework achieves an intersection over union of 0.813 and an F1-score of 0.895, yielding 7.7 % and 3.1 % F1-score improvements over the baseline model and other state-of-the-art models, respectively. The results confirm that the framework delivers robust and interpretable crack segmentation, providing a scalable foundation for intelligent inspection and maintenance of submerged infrastructure.

In the wind-resistant design of tall buildings, structural safety and occupant comfort need to be considered simultaneously, while a single aerodynamic measure may improve one aspect of performance but may not achieve a balanced improvement in both. To investigate the synergistic effects of tapering and chamfer modification on the along-wind aerodynamic performance of tall buildings, this study systematically examines their individual and combined effects using rigid pressure-model wind tunnel tests. The results show that increasing the taper ratio can significantly reduce the along-wind base bending moment, with a maximum reduction of 43.1% under the 100-year return period. However, it also increases the top acceleration response, with a maximum increase of 48.8%, indicating a potential adverse effect on occupant comfort. In contrast, chamfer modification reduces both the aerodynamic load power spectral density and the acceleration response by improving the local flow separation characteristics around the building corners and redistributing fluctuating wind pressures. For the reference model, chamfer modification reduces the aerodynamic load power spectral density and top acceleration response by 51.8% and 14.2%, respectively, although its effect on reducing the base bending moment is relatively limited. Tapering and chamfer modification therefore play different but complementary roles in static load reduction and dynamic response control. Based on this complementary behavior, a combined configuration integrating a smaller taper ratio, β = 2.2%, with chamfer modification is further evaluated. Compared with the prismatic square reference model, this configuration reduces the top-section shape coefficient by 0.42 and the base bending moment by 33.7%, while increasing the maximum top acceleration by only 2.28 milli-g. Compared with the model using only a taper ratio of β = 6.6%, it reduces the top acceleration by 20.1%. Overall, the combination of a smaller taper ratio and chamfer modification can effectively control the dynamic acceleration response while maintaining favorable static load-reduction performance, thereby achieving a more balanced along-wind aerodynamic performance among structural safety, occupant comfort, and useable building area. These findings provide a practical reference for the aerodynamic design of tall buildings.

This paper presents experimental and numerical studies on the behaviour of stainless steel-recycled aggregate concrete-carbon steel double-skin tubular (SRCDST) columns subjected to eccentric compression. The experimental study included material tests of concrete and steel tubes, measurements of initial global geometric imperfections and eccentric compression tests of twelve SRCDST column specimens, which were designed with varying initial eccentricities, recycled coarse aggregate replacement levels and hollow ratios. The test results of failure loads, deformation responses and failure modes were reported, and the ductility performance, lateral deformation development and longitudinal strain distributions were analysed. Finite element models were developed, validated against the experimental data and subsequently employed to conduct parametric studies covering a wide range of geometric dimensions and loading combinations. Three design codes, including Chinese technical specification, European code and American specification, were evaluated for their applicability to eccentrically loaded SRCDST columns against the test and numerical results. The evaluation results generally revealed that the Chinese technical specification offered conservative and slightly scattered failure load predictions for eccentrically loaded SRCDST columns, while the European code resulted in accurate and consistent failure load predictions and the American specification provided accurate predictions but with slight scatter. Finally, a revised design method was proposed based on the Chinese technical specification, demonstrating improved design accuracy and consistency.

Composite structures are often subjected not only to single impacts but also to repeated low-velocity impacts (RLVIs), which in this study refer to two successive impacts. This work investigates, both experimentally and numerically, the mechanical response and damage mechanisms of CFRP laminates under single impacts and RLVIs across various energy levels, with particular focus on non-uniform energies between the first and second impacts. The experiments focused on RLVIs under constant impact energy conditions, conducted at three energy levels of 10 J, 15 J, and 20 J. A complementary finite element model (FEM) incorporating a composite damage model was developed in ABAQUS/Explicit to simulate RLVIs using the restart technique. Based on this FEM, three impact energy scenarios were analyzed: (1) RLVIs with the same energy levels as the experiments; (2) highenergy (20 J and 30 J) second impacts following an initial low-energy (10 J) impact; and (3) low-energy (10 J) second impacts following high-energy (20 J and 30 J) initial impacts. The impact energies were classified as low or high depending on whether they were below or above the critical threshold (approximately 15 J) for fiber damage initiation. The qualitative effects of impact energy and sequence on the mechanical response were evaluated in terms of variations in impact force, displacement, absorbed energy, and response duration. Furthermore, the associated damage mechanisms, including fiber and matrix damage as well as delamination, were analyzed in detail.

This study investigates the wind-induced vibration of a 299.1 m linked twin-tower super-tall building - the Shenzhen Bay Innovation and Technology Centre-based on field measurements during Super Typhoons Saola (2023) and Ragasa (2025). Using an FE-assisted coordinate modal decomposition method together with established modal identification methods, this study analyzes the structural responses and modal evolution characteristics during the two typhoon events. The results reveal a counter-intuitive phenomenon where the maximum acceleration of the lower tower (T2) consistently exceeded that of the taller tower (T1). This is attributed to the sky-bridge coupling effects, which render T2 significantly more sensitive to torsional modes, while T1 remains dominated by translational motion. Furthermore, the structural natural frequencies exhibited a nonlinear "decrease-then-recovery" trend, fully returning to baseline levels approximately 18 h post-peak, while damping ratios showed an overall increasing tendency with vibration amplitude. Notably, a distinct temporal lag was observed between the peak environmental wind speed and the maximum structural response. This suggests that, for the investigated linked super-tall structure, the critical wind direction may play a more dominant role than the absolute wind speed magnitude in determining the peak structural response.

An accurate inlet boundary condition that captures the characteristics of the atmospheric boundary layer (ABL) is critical for large eddy simulations (LES) of high-rise buildings. This study proposes an optimized inflow turbulence generator-termed the Modified Improved Narrowband Synthesis Random Flow Generation (Modified Improved NSRFG, MINSRFG)-which combines the strengths of the Improved NSRFG (INSRFG) and the Modified NSRFG (MNSRFG) methods. The MINSRFG introduces a spectral energy correction coefficient alpha i and a time parameter tau 0, along with an analytical expression for the spatial coherence adjustment coefficients gamma j, thereby avoiding extensive trial-and-error in parameter selection. It enhances spatial coherence, improves temporal correlation, and compensates for the spectral energy loss in INSRFG, while addressing the difficulty of selecting spatial coherence parameters in MNSRFG. First, the method was numerically validated in a representative turbulent ABL flow field. The results show that the turbulence generated by MINSRFG accurately reproduces key turbulence statistics, including the fluctuating wind speed power spectrum, spatial coherence, and temporal correlation. Subsequently, MINSRFG, INSRFG, and MNSRFG were applied to three-dimensional LES of wind effects on the Commonwealth Advisory Aeronautical Research Council (CAARC) standard high-rise building model, and the results were compared with wind tunnel tests. Analysis of building surface pressure coefficients, base moment coefficients, and maximum top displacement indicates that MINSRFG outperforms the other two methods in both accuracy and error control for mean and fluctuating wind pressures as well as base moments. Under multiple wind directions, it also performs well in predicting building top displacement. In summary, MINSRFG offers clear parameter definitions and high computational accuracy, providing a high-precision inflow turbulence generation approach for LES of high-rise buildings.

This study introduces a lightweight cementitious composite for 3D printing applications, incorporating hollow glass microspheres (HGM) and polyvinyl alcohol (PVA) fibers to optimize both mechanical strength and printability. The buildability, extrudability, mechanical performance, and microstructural characteristics of the composites with varying microsphere contents were systematically investigated. The results showed that the mortar exhibited a buildability value above 65 mm, while a continuous filament could be extruded during the test. When the microsphere content was below 20%, the composites met the ACI 213 requirements for the density and compressive strength of lightweight concrete. Further increasing the content to 40% reduced the bulk density to 737 kg/m³ , while still exhibiting considerable mechanical properties. The 3D-printed samples exhibited anisotropic mechanical behavior. The alignment of PVA fibers along the printing direction significantly enhanced the flexural strength, which increased by 157% - 200% compared to that of the cast samples. Moreover, the microspheres demonstrated good interfacial bonding with the cement matrix. Scanning electron microscopy (SEM) image analysis showed that no significant segregation of microspheres occurred in either the printed or cast samples, which contributed to the overall quality of the composite. The combination of HGMs for reduced density and PVA fibers for enhanced flexural strength provides a solution to the challenges of producing high-strength, low-density materials for prefabricated concrete components.

This paper proposes an innovative frequency degradation model, termed the NTF (Number of load cycles, Temperature, real-time Frequency) model, for predicting the real-time frequencies, fatigue life (for the design stage), and residual fatigue life (during the service stage) of fiber reinforced polymer (FRP) laminates at various ambient temperatures (below the glass transition temperature). Built upon the traditional stiffness degradation model, the NTF model integrates the time-temperature superposition principle (TTSP) for frequency degradation along with a novel threshold frequency-based fatigue failure criterion. To validate the effectiveness of the proposed NTF model, tensile fatigue tests and modal tests on GFRP laminates at different ambient temperatures were conducted. The resultant experimental results indicate that higher-mode frequencies exhibit more stable degradation trends during temperature-influenced fatigue loading, establishing themselves as robust indicators for fatigue life prediction. Importantly, the validation results confirm that the proposed NTF model achieves satisfactory accuracy in predicting real-time frequencies (the highest average prediction error is below 1.40 %), fatigue life (about 90 % of the predicted data fall within the 2 times error band), and residual fatigue life (over 90 % of the predicted data fall within the 2 times error band). Notably, this study is the first to integrate the frequency degradation mechanism and the TTSP into a unified model for predicting the temperature-dependent fatigue life of FRP laminates.