Probabilistic live load modeling is fundamental to structural design, but a significant gap exists between theoretical models and their application to continuous structural systems. Current live load design values rely on probabilistic models calibrated for simply-supported members and address spatial load effects in real multi-span systems using deterministic patterns, such as the checkerboard layout. This methodology conflates stochastic load modeling with deterministic design scenarios, creating a theoretical inconsistency. To bridge this gap, this study develops a novel probabilistic live load model specifically for multi-span continuous structures, encompassing the formulation of the theoretical model, parameter identification and calibration, and load effect analysis. Case studies based on conventional code methods demonstrate that the proposed model effectively mitigates parameter sensitivity, a critical issue associated with the current model. The live load value derived from the multi-span probabilistic model can be up to 1.12 times that specified in current design codes based on mid-span bending moment equivalence, and up to 1.25 times based on beam-end bending moment equivalence. For residential and office buildings, the checkerboard loading pattern represents an extreme and conservative simplification that overestimates the actual structural load effects by a margin of approximately 16%. These findings underscore the importance of incorporating spatial load randomness for more rational and reliable design of continuous structures.

Catenary icing may induce aerodynamic galloping and thereby threaten the operational safety of high-speed railways. Estimating the distributed icing mass from sparse structural responses is thus an important yet severely ill-posed indirect measurement problem. Physics-informed neural networks (PINNs) are well suited to this task in principle, but their direct application to fourth-order catenary dynamics is hindered by strong loss imbalance, and uniform collocation becomes inefficient when only a few sensors are available. To address these difficulties, this work develops the Uncertainty-Guided Active Sampling Bayesian PINN (UGAS-BPINN), which combines neural tangent kernel (NTK) trace-based gradient balancing with uncertainty-guided active sampling. The proposed framework adaptively reweights the data and physics terms and uses an annealed acquisition rule that couples residual information with Monte Carlo dropout spread. Results on a reduced-order finite element benchmark and an artificial-climate-chamber experiment show clear improvements over conventional PINN baselines and uncertainty-oriented references. Relative ℓ2 errors of 3.5% in simulation at a signal-to-noise ratio of 14 dB and 3.8% in the physical experiment are obtained with only Ns ≤ 5 sensors, indicating that accurate physics-constrained icing-field identification is feasible under severe data scarcity.

Seismic response simulation of mega-scale urban building clusters is crucial for assessing urban resilience, but is hindered by data scarcity, prohibitive computational costs, and the limited scalability of conventional analytical methods. This paper presents an intelligent computational framework designed to overcome these limitations, enabling efficient and accurate seismic simulation for urban areas with hundreds of thousands to millions of buildings. The framework is composed of three integrated modules: (i) a parameter calibration module that infers key structural dynamic properties from readily available building information; (ii) an adaptive dynamic K-means clustering module to drastically reduce computational workload; (iii) a response prediction module that employs a novel graph neural network to learn from inter-entity relationships, enabling accurate regression from a minute fraction of known data. The framework's efficacy is validated through a mega-scale case study, yielding 4.416 billion seismic response predictions for 1.15 million buildings in Shanghai under 3840 ground motions. The results demonstrate exceptional performance. The framework's predictions of engineering demand parameters achieved a response error of just 4.19%. Significantly, by learning from just 0.0087% of the total response space, the proposed framework reduces runtime from 12.2 years (estimated) with time history analysis to just 59.8 h, offering a robust and scalable solution for quantitative seismic resilience assessment and disaster risk management in modern mega-cities.

This study proposes a novel energy-harvesting rotational inertial dual-tuned mass damper (EH-RIDTMD) integrated with a semi-active damping control strategy, for the simultaneous human-induced vibration mitigation and energy harvesting in large-span floors. The system features a dual-tuned mechanical design combined with an electromagnetic damper that employs a ball-screw transmission mechanism and a buck-boost converter circuit. Based on theoretical modeling and Simulink co-simulation, this study analyzes the electromechanical coupling characteristics and uncovers the frequency-dependent damping response behavior governed by four fixed points in the acceleration spectrum. Subsequently, it proposes a frequency-driven semi-active control protocol that achieves real-time damping adaptation through duty cycle modulation, enabling effective vibration suppression while preserving energy harvesting capability. Comparative analysis demonstrates the semi-active EH-RIDTMD outperforms passive designs in sustained vibration control, particularly outside the frequency fixed-point region, with the harvested energy capable of powering its control circuitry. The proposed solution addresses critical challenges of conventional TMD systems by integrating mass amplification, adaptive damping, and sustainable energy recovery in space-constrained applications.

This study investigates hunting-type shimmy oscillations of the Automated People Mover (APM) running gear induced by contact clearance between guide wheels and rails. A nonlinear dynamic model with five degrees of freedom is established, where clearance-induced contact forces are characterized using third- and fifth-order polynomial fits. Numerical simulations conducted via the generalized-α integration method demonstrate stable limit cycle oscillations under specific conditions, exhibiting typical shimmy behavior. The simulation results indicate vibrations dominated by yaw and lateral dynamics, prompting the construction of a simplified model preserving the primary dynamic characteristics. Stability mechanisms influenced by restoring stiffness and damping are systematically revealed through eigenvalue analyses, and an analytical expression for critical speed is derived using Hopf bifurcation theory. Additionally, the First and Second Lyapunov coefficients under nonlinear contact conditions are computed by the projection method to determine the Hopf bifurcation type (subcritical or supercritical). Bifurcation diagrams under various parameter conditions are plotted using the shooting method. Furthermore, a Generalized Hopf bifurcation is identified and analyzed, and a two-parameter bifurcation diagram is constructed to characterize its dynamic regions. To validate the theoretical model, a comparative simulation model is constructed using the UM multibody dynamics platform. Results show that restoring stiffness effectively extends the stable operating region, while restoring damping significantly increases the critical speed; both parameters distinctly regulate the evolution of limit cycle amplitudes and present two main bifurcation forms. The UM simulation results show good agreement with the theoretical and numerical analyses in terms of both vibration trends and dominant yaw frequencies, indicating that the proposed model and method are accurate and practically applicable.

The whole process of structural progressive collapse exhibits highly nonlinear behavior and inherent randomness, making the design optimization for structures against progressive collapse still a significant challenge. Therefore, this study proposes an efficient probability-based optimization framework for structures against progressive collapse considering various column loss scenarios, by incorporating physics guided deep and cross network (PhyDCN), probability density evolution method (PDEM) and particle swarm optimization (PSO) algorithm. The PhyDCN models for predicting the system resistance curve before the first failure under different column loss scenarios are developed based on the alternative load paths method, and comprehensively validated through four single-point capacity-based and two resistance curve-based metrics. The PDEM combined with equivalent extreme value event method is introduced to assess reliability and risk of structures against progressive collapse. With reliability or risk as constraints and the structural cost increase rate as the objective function, the PSO algorithm is employed for structural design optimization. Finally, the proposed design optimization framework is applied to a prototype RC frame structure, and the effectiveness of different optimization schemes for enhancing structural robustness is investigated. The results show that for the investigated structure, under both reliability- and risk-based design optimization, strengthening only the top-floor beam reinforcement is a more effective strategy than strengthening beams on all floors.

Extraordinary loads in building structures feature short durations, low occurrence frequencies, and large amplitude variability. These characteristics introduce significant uncertainty into their stochastic evolution. Traditional survey methods rely on subjective human memory, which inevitably introduces epistemic uncertainties. To address these limitations, this study proposes a big data-driven stochastic simulation framework for the probabilistic modeling of extraordinary loads. By integrating virtual reality (VR) property systems with physical furniture databases from e-commerce platforms, the framework executes Monte Carlo simulations (MCS) for two primary transient events, namely temporary furniture stacking and crowd gatherings, to generate representative load amplitude samples. Furthermore, a global sensitivity analysis based on Kullback-Leibler (KL) divergence indicates that the extreme value distribution of the maximum combined load is predominantly governed by the variability of load amplitudes. The influence of occurrence intervals is minor. Based on this finding, the compound Poisson point process is simplified into a stationary binomial process, providing an efficient approach for structural reliability assessment.

Vibrations in railway infrastructure constructed on soft soil require efficient strengthening techniques. This research examines the dynamic behavior of a geogrid-reinforced railway embankment supported by pile (GRSP) subjected to high-speed train loads with a validated 3D finite element model. The model includes a moving train load via a Hertzian contact model and is calibrated using field data from the Harbin-Dalian line. A full fractional factorial (FFF) L27 design of trials is used to investigate the effects of pile stiffness, diameter, and embankment height on vibration acceleration. The results show that embankment height has the greatest influence, with taller embankments reducing ground vibrations while increasing track-level accelerations due to the soil arching effect. A multiple-linear regression model is developed to forecast peak accelerations, which helps with the design of GRSP systems for optimal vibration control. The study sheds new light on the coupled dynamics of train-track-GRSP systems and offers practical design advice.

Seismic damage predictions of building portfolios are now recognized as a critical issue in the construction of resilient cities. Existing approaches suffer from insufficient realism and detail, limiting the development of urban retrofit programs and seismic planning. To this end, a multi-level approach for building portfolios from the material scale to portfolio scale is intended to be put forward to improve the realism and level of detail of seismic damage prediction. In this study, the framework of the multi-level approach established on the finite element model is introduced first. Fiber beam elements and multi-layer shell elements are used to describe the mechanical behavior of all components in building portfolios. Then, a parametric modeling procedure is proposed to realize the automatic generation of computational models. The constitutive plastic-damage model applicable to quasi-brittle materials is adopted to forecast the damage and collapse performance of building portfolios. Combining the explicit parallel algorithm and collapse analysis method, an efficient realization of the multi-level approach is provided. In addition, considering the spatial variability of ground motions, a random field model is introduced to generate the seismic scenario of local sites, which reflects wave passage, frequency dispersion and attenuation effects of ground motions. Finally, the applicability of the multi-level approach and some key results are explored using the Tongji campus which is subjected to the ground motion field as a case study. The results demonstrate that the multi-level approach can comprehensively predict the damage from the material scale to portfolio scale, and can well reproduce the collision, collapse and falling debris of building portfolios in an earthquake.

The vibration serviceability of lightweight timber floors under crowd-induced excitation presents significant challenges due to their high stiffness/weight ratio and susceptibility to human-induced vibrations. Existing design codes lack specific methods for predicting dynamic responses under crowd bouncing loads, particularly when considering crowd-structure interaction (CSI). This study proposes a response spectrum model that considers CSI and crowd synchronization effects to predict timber floor vibrations. First, stochastic analysis reveals that CSI significantly changes the mean dynamic properties of the system and introduces significant uncertainty. A response reduction factor (RRF) is introduced to quantify crowd synchronization, showing an exponential decrease with increasing crowd size and demonstrating insensitivity to structural parameters. Integrating these effects, a design-oriented response spectrum model is proposed, incorporating stochastic human parameters and validated through field tests. Predictions at the 75% confidence level conservatively exceed measured data, confirming the reliability of the model for vibration serviceability design of timber floors under crowd bouncing loads. The proposed model provides a practical tool for vibration serviceability design of timber floors.

The surface inspection of precast concrete (PC) components is a critical quality control step, yet current deep learning methods are severely limited by the scarcity of annotated defect data. To address this, a few-shot generation method is proposed to generate high-quality, mask-aligned defect images. This three-stage method first learns pure defect appearance features through a novel training strategy combining ControlNet-fused Textual Inversion with cross-attention fine-tuning. Second, it generates diverse defect masks by employing a two-phase LoRA-based fine-tuning strategy. Finally, it conducts a generation pipeline integrating region enhancement, a dual-control mechanism, and background preservation for precise defect rendering. Extensive experiments on a PC component surface dataset show the proposed method outperforms existing approaches in realism and diversity. Furthermore, data augmentation with the generated images significantly enhances defect detection, achieving a leading 99.30 % Average Precision (AP) in image-level classification and surpassing the state-of-the-art in pixel-level localization with relative improvements of 11.96 % in mean AP and 9.95 % in F1-max. Overall, the proposed method balances superior enhancement capability with efficient training and generation, offering a practical solution for automated inspection in the PC industry.

Random field models are essential for characterizing the spatial variability of floor live load. The traditional homogeneous random field (HRF) assumes a spatially constant mean. This assumption fails to capture real loads with location-dependent mean intensity. To overcome this limitation, this paper proposes a non-homogeneous random field (NHRF) model. The NHRF is defined as the pointwise product of an HRF and a deterministic intensity modulation function. This function scales the local mean while preserving the HRF's correlation structure. The model is applied to determine the equivalent uniformly distributed load (EUDL), a key factor for designers, and its accuracy is validated through stochastic simulations. Theoretical analysis reveals that the impact of load non-homogeneity on EUDL depends on the correlation between the intensity modulation function and the structural influence surface. Simulation results demonstrate that EUDL for slab midspan bending moments is significantly greater than that for column axial forces, a gap often overlooked in existing research. Compared to the HRF, the NHRF improves the accuracy of the EUDL mean, variance, and characteristic value by up to 84.93%, 253.00%, and 13.06%, respectively. Code comparisons reveal that GB 55001-2021 overestimates large-area design loads by 30%, while ASCE/SEI 7-22 underestimates by 40%. Current area reduction factors thus require revaluation based on probabilistic distributions to ensure consistent reliability.

Earthquake ground motions demonstrate inherent non-stationarity and significant intensity variability due to the complex interaction of source mechanisms, propagation paths, and site conditions. This presents dual challenges for existing modeling approaches, including data scarcity in extreme scenarios and coupling of time-frequency-amplitude features. This paper proposes an innovative decoupled generative framework by hierarchically separating three stages: time-frequency generation, amplitude prediction, and time history reconstruction. First, leveraging a conditional generative model, peak-normalized acceleration time-frequency spectra are generated randomly based on key physical parameters as constraints, revealing temporal variations in frequency and amplitude. Then, a cross-attention mechanism is employed to fuse parameters and spectral representations, enabling precise prediction of peak ground acceleration (PGA). Lastly, a phase recovery algorithm is applied to reconstruct the time-frequency spectra into time-domain signals, with amplitude scaled by PGA to produce realistic waveforms. Additionally, data augmentation strategies are proposed to improve physical consistency, alongside a gradient field analysis method to quantify the impact of physical constraints and improve interpretability. To validate its extrapolation capabilities, comparative experiments with the end-to-end simulation method were conducted in cross-regional earthquake scenarios. The results demonstrate that the proposed framework achieves better performance in both the accuracy of response spectra prediction and the fidelity of time-frequency detail restoration. This study provides a novel approach for stochastic simulation of strong earthquakes in data-scarce scenarios.

The weight of a building is a fundamental parameter that plays a critical role in structural hazard assessment, finite element model updating, foundation settlement prediction, and other related fields. At present, two primary approaches are commonly used to estimate building weight: calculation based on design drawings and estimation from the total floor area. The first approach is often hindered by incomplete or missing design drawings, particularly for older buildings, while the second suffers from low accuracy and does not reliably represent the actual weight. To date, no direct method has been established for measuring building weight in engineering practice. To address this limitation, this study introduces a novel method for estimating building weight through modal testing using human-induced excitation. In this approach, the excitation generated by human activity is either directly measured or reconstructed, while the corresponding dynamic response of the building is recorded simultaneously. By integrating established modal testing techniques and measurement devices with the particle swarm optimization algorithm, the modal mass of the structure is identified. The building's weight is then obtained through the conversion relationship between modal mass and total weight. The proposed method was validated through three case studies: a 24-story reinforced concrete building, an 8-story steel frame building, and a 12-story reinforced concrete building. The measured weights were compared with the actual values, yielding errors of 7.37%, 8.66%, and 2.70%, respectively. These results demonstrate the feasibility and the accuracy of the proposed method.

The generation of realistic wind speed time histories is essential for wind engineering analysis but remains challenging due to the scarcity of high-quality measured data and the non-stationary, stochastic nature of atmospheric turbulence. While existing artificial intelligence-based data-driven methods in wind engineering mainly focus on conditional forecasting tasks, the unconditional generation of independent wind speed time histories has received limited attention. To address this gap, a novel spectrogram-based generative framework, called DiffWind, is proposed for wind speed time history generation based on denoising diffusion probabilistic models (DDPM). Wind speed time histories are transformed into magnitude spectrograms using the short-time Fourier transform (STFT), modeled in the spectral domain using a U-Net-based diffusion model, and reconstructed through the Griffin–Lim algorithm (GLA). Field-measured wind speed records were employed to tune the STFT-GLA hyperparameters and train the DDPM. The effectiveness and accuracy of the STFT-GLA combination were validated through numerical experiments, while the diffusion-based spectrogram generation was evaluated using quantitative metrics. The results indicate that the proposed framework can generate high-fidelity wind speed time histories that reproduce key statistical, temporal, and spectral characteristics of measured wind data while demonstrating the capability to generate longer-duration records, highlighting its potential for wind engineering applications.

Floor live load exhibits significant spatial variability, which conventional homogeneous random field (HRF) models fail to capture. This study introduces a novel conditional inhomogeneous random field (CIRF) model, motivated by amplitude-modulated processes, to overcome this limitation. The CIRF model decouples load intensity variations into two multiplicative components: (1) a spatially homogeneous random field, representing random fluctuations with consistent statistical properties across space, and (2) deterministic spatial-dependence intensity modulation functions, representing dimensionless shape functions that describe the spatial distribution of loads. A parameterized analytical expression for the modulation function is proposed, where parameters map to the geometric characteristics of the load distribution pattern, and the randomness of the pattern is captured through conditional probabilities. Parameters and associated conditional probabilities are derived from a large-scale online load survey, proving the usefulness of the proposed formulation. A probabilistic framework based on the CIRF model is developed to determine design loads, incorporating load effects. Theoretical analysis reveals that load inhomogeneity has a significant impact on design loads, notably when it is strongly correlated with structural responses, thereby amplifying load requirements. Further analysis integrating survey results confirms that design loads are highly sensitive to the randomness of distribution patterns, with traditional HRF models significantly underestimating upper quantile loads.

Regional seismic risk assessment (RSRA) is crucial for enhancing urban seismic resilience and guiding disaster reduction strategies. However, correlations in ground motion intensity measures (IMs) and structural damage measures (DMs) across buildings can significantly affect assessment outcomes, and adequately capturing these correlations remains a major challenge. This study presents an enhanced system reliability-based RSRA framework, which reformulates the problem as a dependent k-out-of-n system reliability model, and evaluates the exceedance probability of the number of damaged buildings within a building cluster. While analytical solutions are provided for structurally independent cases, a numerical scheme integrating Gaussian copula sampling with the probability density evolution method (PDEM) is developed for correlated cases. The numerical scheme is validated through comparisons with both analytical solutions and Monte Carlo simulations. The framework is demonstrated by a case study of 29,461 buildings in Shanghai, China, subjected to a magnitude 7.0 earthquake scenario. The probabilities of exceeding four damage states of the building cluster are evaluated to quantify the effect of inter-structural IM and DM correlations on regional seismic risk.

Building material management is fundamental to engineering projects. Currently, the counting of building materials is primarily conducted manually, a process that is time-consuming and prone to errors. Recent advances in computer vision and deep learning have significantly facilitated automation in the construction industry. However, the effectiveness of deep learning approaches hinges heavily on large-scale data sets with accurate, manually annotated data. Existing construction-domain image data sets are ill-suited for material counting tasks, and this lack of large-scale, publicly available data sets has become a major barrier to progress in this area. To address this gap, this study introduces and publicly releases a new large-scale image data set, called Material Counting in Construction (MACO), collected directly from construction sites. The MACO data set comprises 4,426 images and 563,595 annotated objects, covering seven common building materials and diverse real-world construction scenarios. To enhance the utility of MACO and to support more downstream tasks, three popular annotation techniques were employed: horizontal bounding boxes (HBBs), oriented bounding boxes (OBBs), and segmentation masks. This data set is the first large-scale, multimaterial, multiannotation, and high-density open data set (with 127.3 annotations per image) among existing construction data sets. The validity of MACO is confirmed through benchmarking with two state-of-the-art one-stage object detection algorithms, achieving a maximum mean average precision at an intersection over union (IoU) threshold of 0.5 (mAP50) of 91.6%. This provides a robust benchmark for method selection in similar tasks. In addition, to compare the one-material-one-model and multimaterial-one-model paradigms, experiments were conducted on models trained on single versus multiple materials. Results indicated that the multimaterial counting model exhibits performance degradation due to cross-material feature interference, suggesting that developing a generalized detection and counting model requires further research. Overall, MACO is designed to advance intelligent construction site management, including material detection and counting, specification measurement, and robotic grasping and assembly.

Engineering structures are subjected to multiple loads throughout their service life, making the evaluation of combined load effects critical for structural design and assessment. These loads inherently exhibit uncertainty in both spatial and temporal dimensions. When combined, these uncertainties propagate and interact, amplifying the complexity of the structural response. Despite this, existing studies often oversimplify or neglect spatiotemporal uncertainties to reduce computational costs. To address this, a framework for load effect combination analysis considering the spatiotemporal uncertainty of multiple loads is proposed in this paper. The framework first utilizes stochastic harmonic functions (SHF) to reduce the dimensionality of uncertainties in both space and time. It then applies the coincidence principle of random variables to determine the combined load effects. Finally, the probability density evolution method (PDEM) is employed to quantify the probabilistic characteristics of the combined effects. The proposed approach is validated using a reinforced concrete (RC) planar frame. Comparisons with Monte Carlo simulations (MCS) demonstrate the feasibility and accuracy of the framework.