Computer-Aided Civil and Infrastructure Engineering最新文献

The cover image is based on the article An effective ship detection approach combining lightweight networks with supervised simulation-to-reality domain adaptation by Ruixuan Liao et al., https://doi.org/10.1111/mice.13501.

Detection of subsurface road targets is a crucial task in road engineering. This study focuses on detecting three types of subsurface targets: looseness, pipeline, and voids. Ground-penetrating radar (GPR) was employed to acquire real-world data. gprMax was utilized to generate additional data to address the scarcity of the original dataset. Recognizing the substantial disparity between directly simulated gprMax data and actual GPR images, this paper introduces a novel method for synthesizing gprMax-generated data with real measurements, thereby achieving effective GPR image augmentation. Furthermore, a generative adversarial network (GAN) was employed to rapidly produce large volumes of GPR images. Deep learning models were implemented to detect subsurface road targets using datasets of varying scales. Experimental results indicate that data augmentation utilizing gprMax and GAN can substantially improve the detection accuracy for subsurface road targets, achieving a rate of 0.767. This represents a 21.2% enhancement, compared to the results obtained from training on the original dataset. The findings of this research hold practical significance for supporting road maintenance operations.

A critical issue for data-driven and machine learning-based damage detection of engineering infrastructures is associated with unlabeled datasets and distribution shifts in cross-domains. To overcome this challenge, this study develops an unsupervised cross-domain method for bridge damage detection based on interclass alignment of time-frequency features extracted from multichannel sensor data. The computational framework was developed based on a deep subdomain adaptation network integrating digital and physical information. Initially, a multichannel symmetric dot pattern was utilized to transform the structural acceleration signals into a comprehensive image. Subsequently, a convolutional block attention module-enhanced ResNet34 (CBAM-ResNet34) was constructed to extract discriminative time-frequency features, where a local maximum mean discrepancy principle was introduced to perform class-conditional alignment across subdomains. Compared with traditional global domain alignment methods, the proposed approach focuses on aligning class-conditional distributions within subdomains to improve the generalization performance with unlabeled datasets. The proposed method was validated on both simulated and experimental datasets collected from a laboratory-scaled steel truss bridge. Furthermore, a case study on the Old ADA Bridge in Japan was presented to demonstrate the robustness and practical applicability of the proposed approach, serving as a benchmark against classic unsupervised methods. The results show that the proposed framework has a substantial improvement in source-to-target transfer recognition performance. Discussions were conducted on the application prospects of the proposed framework for more in-service infrastructures in complex conditions.

The cover image is based on the article Domain-adaptive self-supervised learning for corrosion detection and 3D building information model mapping in steel tunnels by Shreejan Maharjan et al., https://doi.org/10.1111/mice.70077.

The cover image is based on the article Excavator 3D pose estimation from point cloud with self-supervised deep learning by Mingyu Zhang et al., https://doi.org/10.1111/mice.13500.

This study presents a diffusion-assisted particle reconstruction model (DPRM), a novel framework for reconstructing high-fidelity 3D particle morphology from a single 2D image of granular assemblies. DPRM leverages cascaded large vision models in three stages: (1) segmentation of individual grains via a U-Net-enhanced segment anything model, (2) multi-view synthesis for each particle using denoising diffusion probabilistic models (DDPMs), and (3) 3D geometry approximation via a DDPM-assisted large reconstruction model, generating simulation-ready mesh representations. After mesh decimation and size correction, the outputs are compatible with discrete element modeling and other physics-based simulations. Quantitative validations confirm DPRM's accuracy in predicting particle size and shape distributions. Crucially, the developed method enables zero-shot generation to novel scenarios without extensive retraining, overcoming limitations of prior methods. This work establishes the first end-to-end pipeline for particle-level 3D reconstruction from monocular scene images, enabling the generation of statistically realistic particle shape for physics-based granular simulations in engineering and industry.

Settlement monitoring of high-speed railway bridge piers (HSR-BPs) is critical for ensuring construction and operational safety. However, existing methods for BP settlement prediction face three challenges: limited dataset size, inconsistent observation periods across measuring points, and difficulty in synchronizing spatiotemporal correlations between points. To address these issues, this paper proposes a multi-point prediction method for HSR-BPs based on the Inverted Hermite Kolmogorov–Arnold Network (KAN) Transformer. First, construct a dataset with consistent observation days through interpolation and measuring points screening; Second, integrate a HermiteKANLinear Module into the Key-Query-Value attention mechanism to capture spatial–temporal dependencies; Then, replace the traditional multilayer perceptron with a HermiteKAN to improve prediction accuracy. Experimental validation using 34 sets of augmented settlement data demonstrates the proposed method's superiority over 14 baseline methods. Experiments of both single and multiple measurement points show significant performance gains: Compared to the single measurement point, the multiple measurement points using the proposed method reduce mean absolute error, root mean square error, and MAPE by 21.16%, 20.57%, and 21.11%, respectively. Furthermore, the proposed method outperforms eight deep learning models across varying prediction lengths. Ablation studies confirm that each proposed component contributes to the overall optimal performance in all evaluation metrics, validating the proposed method's effectiveness and precision. The dataset and codes are available at https://github.com/RSIDEA-ECUT/IHKTransformer.

Accurate and rapid prediction of deep excavation deformation is crucial for construction safety and environmental protection. Traditional finite element analysis is time-consuming, while single-objective optimization (SOO) tends to cause parameter compensation effects. This paper proposes a deep learning and multi-objective optimization (MOO) approach for excavation deformation prediction. A triple-attention convolutional network (TACN) is constructed to capture the complex interactions among soil parameters, deformation locations, and excavation stages. Integrating the proposed TACN, a TACN-MOO optimization framework is established to perform rapid parameter identification and deformation prediction by simultaneously considering wall deflection and ground settlement. Validation through a Shanghai excavation project shows: (1) TACN effectively captures nonlinear soil-deformation relationships with higher accuracy than convolutional neural network models; (2) the MOO framework effectively mitigates parameter compensation effects while reducing computation time from 8+ h to 1–2 min; (3) engineering applications demonstrate that the method achieves high accuracy in wall deflection prediction and good agreement in settlement estimation with excellent transferability. This research provides an efficient and reliable technical framework for intelligent prediction and dynamic control of deep excavation deformation.

Dam safety assessment systems play a pivotal role in evaluating the structural integrity of critical hydraulic infrastructures. Current implementations frequently exhibit limitations in critical functionalities including multi-source data integration and automated anomaly detection. This study proposes an ontology-enhanced intelligent assessment system featuring three technical innovations. A multi-level semantic representation framework is proposed to formally model structural components, sensor networks, and their spatiotemporal relationships through domain-specific ontology engineering. A hybrid anomaly detection architecture employs spatiotemporal variational autoencoder to enable unsupervised identification of abnormal signals. A knowledge-informed reasoning framework integrates empirical safety rules and detection results through Semantic Web Rule Language and SPARQL Protocol and Resource Description Framework Query Language query. Experimental validation on a double-curvature arch dam demonstrated superior performance. The proposed system achieves 89.1% anomaly detection accuracy, simplifies the semantic query through ontology-driven knowledge indexing, and enables automated diagnostic reasoning that identifies the causal relationships between abnormal signals and environmental triggers.

Suspended ceiling (SC) systems constitute a critical nonstructural building component. Excessive deformation of the ceiling surface can cause life-threatening falling debris during earthquakes and create voids that may expose occupants to hazardous materials concealed above the ceiling. To address limitations of the in-service detection of SC deformation, this paper presents a point cloud–based full-scene SC detection method, integrating region growing, Hough Transform, a customized Set2Seq network, and robust principal component analysis to achieve a complete workflow from ceiling segmentation, panel extraction to deformation quantification. Point cloud data with color information acquired from two precision-differentiated devices are used in substage tests and holistic evaluation. The substage tests demonstrate that the local panel deformation quantitative accuracy of the proposed method is generally over 80%, and the holistic experiments show the feasibility of full-scenario practical application.