Structural Control & Health Monitoring最新文献

Timely mastering the preload state of transmission tower bolts is crucial for maintaining the structural performance of transmission towers and preventing the occurrence of tower collapse accidents. A preload state detection method for transmission tower bolt joints based on sound recognition is proposed. Based on the Mel spectrogram, the bolt sound signal features are extracted, and the mapping relationship between the sound signal and the preload state is established after feature fusion; extreme gradient boosting (XGBoost) is introduced as an independent predictor in the deep forest model, and an improved deep forest model is obtained; finally, experiments are conducted on single-bolt and multibolt joints. The results show that the fusion feature proposed in this paper can effectively support multiple machine learning models. The improved deep forest model has an accuracy of 97.6% and 96.9% in identifying the preload state of single-bolt and multiple-bolt joints, respectively, and has a fast detection speed. The model can still achieve accurate classification even in the case of sample imbalance. The recognition accuracy of the method remains above 95% at −2, 0, and 2 dB signal-to-noise ratio (SNR) levels, demonstrating excellent noise immunity.

The durability and stability of cold regional tunnels are extensively threatened by cyclic freeze–thaw environments, highlighting the necessity of accurate monitoring and assessing the accumulative damage within lining concrete. This study proposed an innovative integrated sensing strategy that concurrently incorporated the piezoelectric-based electromechanical impedance (EMI) and wave propagation (WP) techniques for the detection of cyclic freeze–thaw damage in bended concrete. First of all, the visual inspection and four-point flexural testing were performed to study the characteristics of damage evolution and strength deterioration of cyclically freeze–thawed concrete under different bending loads. In addition, the coupling effect of freeze–thaw cycling and bending on damage progression was interpreted. The degrading flexural strength was employed to quantify the overall damage degree. Moreover, the performances of EMI and WP methods in monitoring the concrete freeze–thaw damage were explored by analyzing the varying conductance and stress wave signatures based on conventional approaches. It was found that the statistic metrics of conductance and energy of stress wave signals served as reliable indicators of damage. The key novelty of this study was that a dual-channel feature fusion network (DCFF-Net) was developed to extract and coordinate damage-relevant features from cross-dimensional conductance and continuous wavelet transform spectral data, enabling automated and intelligent damage assessment. The proposed framework demonstrated a substantial improvement in damage classification performance, achieving accuracies exceeding 0.98 on two distinct datasets. This performance notably surpassed that of single-sensing methods. These findings manifested the feasibility and effectiveness of leveraging dual-channel piezoelectric data fusion for improved detection and characterization of complex damage evolution within concrete infrastructures.

In long-period structures, rate-independent linear damping (RILD) offers superior control of acceleration responses compared with conventional viscous damping. However, the noncausality of RILD limits its application in buildings. Further, practical implementation of methods to approximate RILD in specific situations remains challenging owing to the difficulty of mechanical realization. In this study, the inerter, Maxwell–Wiechert model, and negative stiffness were employed to construct novel RILD-approximating devices: the negative-stiffness inerter Maxwell–Wiechert (NSIMW) model and the NSI-spring Maxwell–Wiechert (NSISMW) model. These devices were designed to replicate the storage and loss stiffness of RILD at the target frequencies. This study proposes causal devices for approximating RILD using inerters, negative stiffness, and Maxwell–Wiechert models. The initial design—the negative-stiffness viscous-inerter Maxwell–Wiechert (NSVIMW) model—was unrealizable owing to negative design parameters. To resolve this, the viscous damper was removed, resulting in a simplified NSI Maxwell–Wiechert (NSIMW) model. An enhanced version of the NSISMW model was further developed by adding a spring element. Both models were designed to match the storage and loss stiffness of RILD at target frequencies, and direct design methods were developed to determine their parameters. Real-time hybrid simulations and analytical analyses were conducted to evaluate seismic control performance. The findings demonstrate that while NSIMW provides a causal approximation of RILD with moderate effectiveness, the NSISMW model achieves a more satisfactory seismic performance than the RILD, negative-stiffness Maxwell–Wiechert model, and traditional tuned viscous mass dampers. These results clarify the relative merits of the two designs and suggest that the NSISMW model offers a promising direction for the practical implementation of RILD-inspired damping systems.

Engineering structures usually have rare damage scenarios, which provide insufficient labels to identify the structural damage. Therefore, the effectiveness and accuracy of traditional data-driven methods with limited training samples is an urgent task. This study develops a semisupervised transfer learning (TL) method for structural damage identification. The effectiveness is validated through experiments on a scale cable-stayed bridge specimen. Initially, a convolutional neural network (CNN) was trained based on a numerical dataset simulated by a finite element model, where a source domain is modeled with strong recognition performance and generalizability. Subsequently, the network structure and hyperparameters in the source domain were transferred to the corresponding positions in the experimental training model to create a pretrained model in the target domain; this pretrained model was updated by using site-specific measured data from the engineering structure in the target domain. Finally, a dynamic threshold self-training algorithm was employed to further optimize the model: in the initial stage, a high-confidence threshold of 90% was set to filter reliable pseudolabels, and the threshold decreases to 80% gradually, under the condition that the model cannot provide predictions with sufficiently high confidence. Experimental study on a scale cable-stayed bridge specimen was conducted to demonstrate the effectiveness of the proposed method. Four networks were established based on (1) experimental samples; (2) both labeled and unlabeled experimental samples with the self-training algorithm; (3) TL from the finite element model and experimental samples; and (4) applying self-training with unlabeled samples to the target model derived from TL. The results indicate that the damage identification accuracy of the aforementioned models is 73.6%, 79.4%, 84.6%, and 89.8%, respectively. The TL improves the reliability of generating pseudolabels by utilizing the self-training process and unlabeled data and decreases errors in pseudolabels. Both finite element simulation data and practical unlabeled sample data were successfully combined by the TL and self-training semisupervised learning method for damage identification in cable-stayed bridges effectively.

Optimizing the mesoscopic structure of concrete based on energy dissipation theory is an effective approach to enhancing its performance. However, related works are limited, and the mathematical relationship between the mesoscopic structure and the energy dissipation mechanism remains unclear. In this study, image processing technology and MATLAB were employed to quantitatively characterize the mesoscopic structure of concrete. A mesoscopic model of concrete was constructed using PFC2D to study the influence of mesoscopic parameters on energy dissipation, and a mathematical model of energy dissipation incorporating mesoscopic parameters was fitted using a deep neural network. In addition, the genetic algorithm with associated individuals (GA-AIs) was applied to optimize the mesoscopic parameters. The main findings are as follows: (1) Based on the entropy-containing random aggregate method, Fourier shape reconstruction analysis, and density-damping random field method, the aggregate spatial arrangement, shape characteristics, and mortar matrix characteristic property were quantified separately. (2) Under uniaxial compression, the three types of mesoscopic parameters showed positive correlations to varying degrees with the energy dissipation capacity of concrete. By adjusting these parameters, the energy dissipation capacity can be effectively modulated. (3) The GA-AIs efficiently optimized the mesoscopic parameters, enabling effective control of the energy dissipation capacity of concrete. Based on the optimization results under a specific working condition, the algorithm can infer mesoscopic parameter values to meet different performance requirements for this condition.

As aging bridge infrastructure poses increasing safety risks, there is a critical need for reliable and scalable Structural Health Monitoring (SHM) systems. Traditional SHM methods, which rely on fixed sensor networks and assessments of individual bridges, face significant challenges in scalability, cost, and efficiency—particularly in complex urban environments. To address these limitations, this study introduces the Multibridge Inference SHM (MISHM) framework. MISHM leverages drive-by monitoring and crowdsensing to observe multiple bridges simultaneously. It employs a feature-based analysis using Mel-frequency cepstral coefficients (MFCCs) and Kullback–Leibler (KL) Divergence to identify structural changes. Here, “inference” refers to drawing conclusions about the health of each individual bridge by comparing patterns and features gleaned from the entire network, rather than relying on isolated measurements. By making multiple comparisons across all monitored structures, MISHM enhances fault tolerance, reduces missed detections, and offers a scalable solution for smart city infrastructure monitoring. This framework represents a vital advancement in SHM systems, addressing the evolving needs of large-scale urban infrastructure management.

The impact of pit excavation on the surrounding environment is closely related to the deformation characteristics of the surrounding enclosure structure. However, most existing methods rely on calculating pit unloading stress based on the Mindlin solution, which does not adequately account for the dynamic deformation characteristics of the enclosure structure at different excavation stages and is difficult to apply for real-time assessment. This paper presents a new calculation method based on the mobilizable strength design (MSD) approach to dynamically predict the horizontal displacement of the shield tunnel adjacent to the excavation pit. By introducing dynamic evaluation of the horizontal displacement of the enclosure structure, the applicability of the traditional MSD method is enhanced. The paper compares and analyzes the differences between this method, the modified MSD (MMSD) method, the MSD method, and measured data from actual pit excavation cases. The results demonstrate that the proposed method more accurately reflects the deformation characteristics of the enclosure structure at different excavation stages and its dynamic impact on the horizontal displacement of the shield tunnel. The spatial distribution of horizontal displacement in the enclosure structure under zoned excavation is analyzed, revealing the coupling relationship between the deformation characteristics of the enclosure structure and the tunnel’s deformation response. The findings of this study provide valuable references for the safety assessment and protective measures of shield tunnels during pit excavation.

Accurate prediction of deformation under thermal influences is critical for the safety assessment and long-term performance of high dams. This study proposes a novel two-stage prediction framework that integrates statistical modeling with deep learning to enhance the interpretability and accuracy of dam deformation forecasting. In the first stage, sparse principal component analysis (sPCA) is employed to extract dominant features from high-dimensional thermometer data. These features are then used to construct an interpretable dam deformation monitoring model using multiple linear regression (MLR), referred to as the HT_sPCAT-MLR model. In the second stage, the multilayer bidirectional gated recurrent unit (multi-Bi-GRU) network is developed to model the residuals of the HT_sPCAT-MLR framework, leveraging advanced gating mechanisms and bidirectional temporal learning to improve long-term prediction accuracy. Furthermore, the adaptive genetic algorithm (AGA) is utilized to optimize the hyperparameters of the multi-Bi-GRU model, enhancing the robustness and generalization of the residual correction module. The proposed methodology is validated using real-world monitoring data from an ultra-high arch dam. Quantitative evaluation at four representative measurement points shows that the proposed model consistently outperforms baseline methods across all key metrics. Specifically, it achieves R² values above 0.99, mean absolute error reductions of over 80% compared to traditional models, and the lowest sMAPE across all cases. The experimental results demonstrate model’s superior prediction accuracy, robustness, and practical applicability for dam deformation. The integrated framework offers a reliable and interpretable solution for thermal deformation forecasting in high dam structures.

The increasing complexity of transportation infrastructure demands advanced, data-driven approaches for early pavement distress detection and maintenance decision-making. Traditional assessment methods often fail to provide reliable, interpretable, and proactive insights into pavement degradation. This study introduces an Explainable Artificial Intelligence (XAI) framework that integrates clustering algorithms with principal component analysis (PCA) to improve early-stage pavement distress analysis. The proposed framework leverages K-means, Gaussian mixture models (GMMs), and hierarchical clustering, applied to a customized dataset encompassing pavement performance metrics, geospatial information, and aggregate properties. By incorporating ground-truth validation, our approach not only differentiates between high-quality and deteriorating pavement sections but also reveals underlying factors contributing to distress, overcoming the opacity of traditional machine learning (ML) models. Results demonstrate that this transparent, interpretable AI-driven framework enhances infrastructure resilience by enabling data-informed decision-making for predictive maintenance. Beyond transportation engineering, the methodology establishes a scalable paradigm for explainable AI applications in civil infrastructure, advancing the intersection of ML, geospatial analysis, and material science.

This article proposes a new approach to interface both numerical and experimental substructures of a real-time hybrid simulation experiment running at different sampling rates. A regularized Wiener statistical finite impulse response filter is applied to the slow-rate sequence of interface target displacements at the numerical substructure to predict the next data point. Then, an interpolation rule using monomials is applied to obtain the sequence of interface target displacements at the experimental substructure running at a fast sampling rate. The Wiener filter is trained using offline simulations of the partitioned reference structure before the main experiment. The proposed scheme achieves good results for virtual simulations with linear and nonlinear structures, and it separates the task of determining simulation rates between substructures, ensuring both accuracy and stability in the experimental test.