Structural Control & Health Monitoring最新文献

The deformation of concrete dams directly reflects their structural health and operational state, serving as a critical foundation for safety assessment and early risk warning. Accurately predicting dam deformation patterns is thus essential for ensuring long-term structural safety and enabling scientific operation management. However, existing models remain limited in addressing high-frequency noise in monitoring data, performing dynamic feature selection, and modeling complex spatiotemporal dependencies, which collectively constrain prediction accuracy. To overcome these challenges, this study proposes a dam deformation prediction model that integrates variational mode decomposition with wavelet thresholding (VMD–WT), a partial autocorrelation function (PACF)–based dynamic feature selection approach, and the ScaleGraph Block-Mamba-like linear attention (SGB–MLLA) –Transformer. The proposed model performs multiscale signal decomposition to suppress noise and extract dominant deformation trends, while dynamically selecting key influencing factors and incorporating spatial dependency modeling and lightweight attention mechanisms to enhance the representation of long sequence and multifactor coupled deformation features. To validate the model’s effectiveness, deformation data from monitoring points of a concrete dam in Jiangxi Province, China, were used for evaluation. Experimental results demonstrate that the proposed model achieves superior prediction performance across multiple monitoring points, achieving near-perfect accuracy (R² = 0.9993) with submillimeter error margins at GLD4, significantly outperforming existing models. These findings confirm that integrating frequency-domain decomposition with adaptive feature selection and employing linear attention for efficient long sequence modeling can substantially improve deformation prediction accuracy. This research provides a novel methodological framework for dam health diagnosis and safety management, offering both theoretical and practical value for the development of intelligent dam monitoring systems.

In this paper, the continuous deformation monitoring data of high arch dams during construction are obtained using the shape accel array (SAA) for the first time. First, the accuracy of the SAA measurement was tested in the laboratory. Then, the SAA was installed using the new method on a case dam section to obtain continuous deformation data during the construction period of the high arch dam. Finally, the self-developed finite element simulation software SAPTIS was used to conduct a simulation analysis of the case dam, considering the effects of concrete material creep, self-volume changes, water cooling, environmental temperature, and self-weight. The laboratory test results show that deformation measurement accuracy is significantly improved after noise reduction by wavelet analysis. The continuous deformation of the dam during construction can be monitored in real time by embedding SAA in the construction of the case dam section. Then, the finite element simulation results verify the accuracy of the measured results of the dam and quantify the impact of various factors on dam deformation. SAA provides an effective means for real-time monitoring and safety assessment of dam deformation.

In this study, an automated transmissibility-based procedure for damage detection is developed for output-only systems. The application of transmissibility has been previously investigated for damage detection. Despite the advancements, current techniques are not applicable in a general way, as vast experience or expert knowledge is needed to achieve accurate results, particularly for the frequency range selection. Moreover, the extent of noise influence still needs to be adequately addressed. A novel procedure has been developed to resolve these issues. First, the frequency range is determined by applying modal coherence using the first singular value of the cross-power spectral density (CPSD). Then, the transmissibility functions are calculated from the CPSD and smoothed using a moving mean approach to reduce the influence of noise. Afterward, the threshold is obtained from the transmissibility damage indicator values of the system’s healthy state. Finally, damage detection can be performed continuously for each subsequent dataset. The procedure is compared to damage detection based on eigenfrequencies and mode shapes using simulated data, demonstrating higher sensitivity to minor damages at low noise levels. Furthermore, the procedure is validated on experimental data from a steel cantilever beam, where various noise scenarios, damage severities, and damage positions are considered, and on field data from a lattice tower, showing high damage detection accuracy across three damage scenarios. The proposed procedure can be automated, demonstrating sensitivity to minor damages when high signal-to-noise ratio is available.

Pavement intersecting cracks expand outward under load, especially at intersections where stress leads to branching and a complex network. This study introduces Crack-DL, a deep learning framework for crack segmentation and feature extraction. We propose the YOLO-Segcrack model, which integrates the advanced FasterNet backbone with the SENet attention module. This combination leverages the computational efficiency of FasterNet for robust feature extraction and the discriminative ability of SENet to emphasize critical crack areas, and the model achieves significantly improved segmentation performance and precisely extracts pavement intersecting cracks. Additionally, a convolution kernel matching algorithm (CKMA) is developed based on morphological image processing for precise intersection point localization and to quantify crack lengths and intersection angles. Finally, the CrackX dataset containing pavement intersecting cracks is constructed to support this research. The proposed Crack-DL framework was tested on CrackX and public datasets, CrackTree260, demonstrating its accuracy and reliability. Experimental results show that using the YOLO-Segcrack model increases detection and segmentation precision by 11.1% and 4.8%, respectively. In addition, extensive experimental results on crack-seg, package-seg, and carparts-seg datasets further show that the improved YOLOv8s-seg model outperforms existing advanced methods in terms of performance. When applying the CKMA for detecting intersection points, the detection accuracy reached 73.19%. For the publicly available CrackTree260 dataset, the accuracy reached 91.5%. Furthermore, when the error is under 5 unit pixels (mm), the accuracy for calculating total crack length is 92.46% for ground truth images and 80.82% for the adaptively segmented binary images. These results demonstrate that the proposed model enhances the extraction of intersecting cracks area and the CKMA provides a reference value for the analysis of cracks propagation. The dataset and source code are available at https://github.com/Keeeram/Intersecting-Crack-Analysis.

Wind turbines with larger capacities face bending deformation due to taller towers and longer blades, necessitating mitigation against extreme seismic loads. A vertically installed inerter-based damper, referred to as the tuned viscous mass damper (TVMD), is proposed alongside a closed-form design approach. First, the mechanical model and simulation approach for the TVMD and wind turbines are introduced, followed by the derivation of governing equations and frequency response solutions, considering the parked state. Second, a nacelle-hub assembly displacement–oriented design principle is formulated, providing mathematical design expressions and closed-form solutions based on the generalized fixed-point principle. Finally, the effectiveness of the proposed framework is validated through design cases and comparative investigation of theoretical approaches, under parked conditions with negligible aerodynamics and thus low effective damping, highlighting the advantages of the closed-form design formulas. The results indicate that the vertically installed TVMD offers superior performance compared to traditional damping design approaches in wind turbines, enabling the simultaneous control of multiple seismic responses. Furthermore, the nacelle-hub assembly displacement–oriented design principle and closed-form design formulas provide a quantitative framework for optimizing key design parameters of vertical TVMDs, facilitating rapid design implementation and deeper theoretical understanding. In addition, the proposed closed-form design formulas ensure enhanced energy dissipation and specific modal tuning capacity, offering robustness against parameter variations.

J. Lee, C. Lee, I. Yeo, and S. Jeong, “Multivision System for High-Resolution Strain Measurement of Continuously Welded Rail,” Structural Control and Health Monitoring 2025, no. 1 (2025): 1–16, https://doi.org/10.1155/stc/2447466.

In the article titled “Multivision System for High-Resolution Strain Measurement of Continuously Welded Rail,” there was an error in the funding grant code.

The correct funding statement should be as follows:

This research was supported by a grant from R&D Program (PK2501D4) of the Korea Railroad Research Institute.

We apologize for this error.

Under ambient excitations, the vibration response data of structures exhibit significant time-varying characteristics as time progresses. This time-varying data causes domain shift, which greatly hinders the application of neural networks in structural health monitoring (SHM). This paper proposes a one-dimensional spatiotemporal convolution-based domain adversarial network (SDAN) to address the issue of decreased damage identification (DI) accuracy in neural networks caused by the domain shift. In SDAN, to effectively utilize the spatial information from different sensors, we designed a one-dimensional spatiotemporal convolution that integrates temporal and spatial characteristics of the vibration response data. The spatiotemporal convolution proposed was advantageous for extracting fine-grained features with spatiotemporal characteristics to enhance the performance of the domain adversarial network. Domain adversarial training is then employed to extract domain-invariant features from the data, enabling the identification of damage features in structural response data under ambient excitations and improving the applicability of the network in time-varying data. The effectiveness of the proposed network is validated using vibration response data collected from two real-world bridges, old ADA bridge and KW51 bridge, under ambient excitations. The results show that SDAN significantly reduces the impact caused by the domain shift, achieving F1 scores of 95.8% and 99.6% on the old ADA bridge and KW51 bridge datasets, respectively. This represents an improvement of 21.2% and 12.1% compared to a network without domain adaptation (NoDA). Furthermore, SDAN was compared with a domain adaptation network based on global feature alignment using deep adaptation network (DAN) and a domain adaptation network based on subfeature alignment using deep subdomain adaptation network (DSAN). SDAN achieved the highest F1 scores on both examples, illustrating the effectiveness of domain adversarial training in addressing domain shift issues caused by time-varying ambient excitations. This provides a promising approach for utilizing ambient excitations in real-time structural DI.

BIM is increasingly crucial for the building construction, yet deviations from actual construction limit its construction monitoring applications. Therefore, this paper proposes a BIM-construction interaction method that integrates parametric modeling, point cloud processing, and parameter optimization and fitting to enhance the construction monitoring. Proposed parametric modeling methods equip BIM elements with the ability of the pose adjustment and deformation modification, addressing potential deviations during construction. Developed component feature extraction algorithms efficiently capture pose and deformation features from the point cloud model that sufficiently and accurately reflect the actual state of the structural components. The proposed parameter optimization and fitting approach targets model parameters for optimization and aims to match pose and deformation features for fitting. By constructing objective functions that quantify the deviation between the BIM and point cloud models, the process is driven by the RBFOpt optimization algorithm and Opossum optimization solver. This approach enables the automatic updating of the design BIM into the as-built BIM and generates deviation data between the two models, providing a basis for comprehensive construction monitoring results. The BIM-construction interaction method was applied to the core area construction of the Shanghai Grand Opera House, where it reduced the root mean square error (RMSE) between the parametric BIM and point cloud models of the core column, concrete thick shells, and cantilever beams from 0.0352 m, 0.0411 m, and 0.0323 m to 0.0082 m, 0.0323 m, and 0.0053 m, respectively, significantly reducing deviations between BIM and the actual construction. Comprehensive and quantitative construction monitoring data, including pose deviations and structural deformations, were obtained to assess the precision and safety of the core area construction. The results demonstrate that the BIM-construction interaction method effectively supports the interaction between BIM and construction, enabling monitoring and evaluation based on point cloud data.

Regular monitoring of cable forces is critical to ensuring the long-term safety and performance of cable-stayed bridges. While vision-based methods offer noncontact, cost-effective alternatives to traditional vibration-based methods, most existing studies adopt an offline workflow in which videos are recorded and processed afterward. This study develops an embedded vision-based sensing system for cable force monitoring. Unlike offline vision approaches, the system performs on-site video acquisition, processing, and force estimation on-site, enabling real-time monitoring without external video transfer. First, an efficient and accurate visual object tracking (VOT) algorithm is proposed for real-time displacement extraction from video sequences. We benchmark the algorithm’s accuracy and computational efficiency on a Jetson Orin Nano using a public shaking table test dataset. The results show that the algorithm achieves a good balance between accuracy and computational efficiency, making it suitable for deployment on edge computing devices. Subsequently, the cable vibration experiment indicates that the embedded vision-based sensing system achieves maximum errors of 2.61% in cable frequency measurement and 5.68% in cable force estimation. In addition, the camera position did not materially affect system accuracy. Future work will enhance robustness under diverse field conditions and validate the system on full-scale bridges.

Industrial machinery plays a vital role as essential mechanical equipment across industries, such as aviation, transportation, and smart manufacturing. However, these machines are prone to various failures caused by complex and dynamic operating conditions, which can disrupt entire industrial systems, lead to significant financial losses, and pose serious safety hazards. This emphasizes the importance of fault diagnosis in these machines to improve system reliability and safety. Recently, artificial intelligence (AI)–based techniques have gained significant attention due to their reliability, superior performance, and adaptability in diagnosing faults. However, a comprehensive review of recent advancements in intelligent fault diagnosis (IFD) is still lacking, and clear future research paths for further advancement are not well-defined. In addition, choosing the appropriate fault diagnosis methods for specific fault types remains a challenge. To address these gaps, this paper provides an in-depth review of the latest advancements in AI techniques applied to fault diagnosis in industrial machinery. The review paper starts by introducing the basic concepts of AI methods and then delves into a detailed examination of their applications in IFD for industrial machinery. In addition, the review discusses the strengths and weaknesses of different variants of AI methods, including traditional machine learning, deep learning, and transfer learning, within the field. Based on the review results, existing research challenges and prospects are discussed to guide future directions, followed by conclusions. Thus, this review serves as an essential resource for professionals, researchers, and stakeholders involved in the research field.