Manufacturing is the foundation of a country. Made in China 2025 has identified ten key areas as the core engine for the transformation and upgrading of China's manufacturing industry. This study reviews the background against which the ten key areas were proposed, elaborates on the criteria for selecting these ten areas, and sorts out international evaluations on the research outcomes concerning the ten key areas. The study adopts an organizational model of "academician-led, expert-coordinated, and field-based in-depth research" and employs a comprehensive evaluation system that "integrates quantitative and qualitative methods, and complements horizontal and vertical analyses". It summarizes and evaluates the breakthrough development effectiveness of 16 subsectors across the ten key areas from three dimensions: market influence, technological competitiveness, and security and controllability. The findings show that remarkable progress has been achieved in the ten key areas of China's manufacturing industry: five sectors have exceeded expectations and reached world-leading levels; two sectors have fully accomplished their tasks and attained world-advanced levels; eight sectors have basically completed their objectives but still lag behind the world's advanced level to a certain extent; and one sector has failed to meet its expected targets with a considerable gap from the world's advanced level. Looking ahead, efforts should be made to dynamically optimize development priorities to accelerate the cultivation of new quality productive forces in light of local conditions, focus on key core technologies to break bottleneck constraints in the industry, optimize and upgrade traditional industries to consolidate the foundation for developing new quality productive forces, and deploy future industries in advance to build new advantages for China's manufacturing industry, thereby providing solid support for enhancing the competitiveness of China's manufacturing industry, securely seizing the initiative in global industrial development, and achieving the goal of upgrading China from a manufacturer of quantity to one of quality.

In this study, the optimal design of longitudinal stiffeners in steel web plates is investigated based on inelastic numerical analysis. Unlike previous research that was predominantly based on elastic eigenvalue buckling analysis, a nonlinear finite element analysis using ABAQUS is integrated with an optimisation framework implemented in MATLAB to determine the optimal locations and minimum required flexural rigidities of longitudinal stiffeners. Parametric studies are conducted on web plates reinforced with one to three longitudinal stiffeners under pure bending and pure shear, across a wide range of aspect ratios. Through comparisons with existing studies and design codes based on elastic eigenvalue buckling analysis, new insights into the design of longitudinal stiffeners are revealed. For instance, the obtained optimal stiffener locations of singly- and doubly-stiffened plates under pure bending are generally closer to the compressed edges than those recommended by elastic eigenvalue buckling analysis. Moreover, the influence of the aspect ratio and the number of longitudinal stiffeners on the bending resistance and shear resistance of optimally stiffened plates is examined. The bending resistance of the optimally stiffened plates is only slightly influenced by the aspect ratio, and can be substantially improved by adding longitudinal stiffeners, with the most pronounced benefits observed when increasing from one to two stiffeners. The shear resistance of the optimally stiffened plate generally decreases with increasing the aspect ratio, and can be enhanced by adding longitudinal stiffeners. Design recommendations are proposed based on these findings, offering substantial practical value for the stiffening of large cross-section steel box girders.

Accurate service performance evaluation of reinforced concrete (RC) beam structures is crucial for ensuring structural safety and guiding maintenance decisions. However, current practice primarily relies on qualitative visual inspections that fail to quantitatively link apparent defects to internal mechanical behavior. To address this, a novel evaluation framework fusing apparent crack features with static and dynamic responses is proposed. A context-aware grid-based deep learning model (CGDL-Crack) is developed that combines transfer learning with skeleton extraction, achieving crack localization with a maximum validation AP of 96.4% under complex backgrounds. Based on large-scale parametric finite element simulations and Sobol global sensitivity analysis, key state indicators—including static reaction forces, modal frequencies, and crack widths—are identified, and an artificial neural network (ANN) surrogate model is constructed to map multi-source monitoring data to material constitutive parameters. Full-process failure tests on 17 RC beams demonstrate that crack width follows bilinear growth and remains sensitive after stiffness indices saturate. The updated FE model accurately predicts ultimate bearing capacity, demonstrating the effectiveness of the proposed framework and its application potential for RC beam-type components in bridge and building engineering.

The vast inventory of self-built housing in China, exceeding 580 million units, poses a significant challenge for structural safety management, as traditional manual inspection methods are inefficient for large-scale assessment. To address this problem, this paper proposes a rapid, automated multimodal data fusion framework for community-scale inspection of safety hazards in self-built houses. The framework integrates 3D point clouds and high-resolution images to automatically identify, locate, and quantify surface cracks, while simultaneously detecting unauthorized structural addition or demolition at the community level. Data is acquired using a custom mobile inspection device equipped with LiDAR, an IMU, and multi-view cameras, enabling efficient data collection via SLAM. Experimental results demonstrate the framework's high precision: it achieves a relative distance error of less than 1.4% in 3D modeling and a relative error of 11.4% in estimating crack widths as fine as 2 mm. A field deployment in a real community confirmed its practicality and efficiency, reducing total inspection time by 51.7% and cutting on-site manual operation time by 84% compared to traditional manual methods. By transforming discrete multimodal sensor data into structured, actionable spatial information, this framework provides a scalable and efficient solution for the proactive safety inspection of large-scale houses, advancing the digital transformation of community-scale infrastructure management.

This study investigates the optimum position and minimum flexural rigidity of longitudinal stiffeners for steel web plates under coupled bending and shear stresses. Given that existing codes and literature mostly focus on pure bending or shear, a systematic stiffening design framework for bending-shear interaction is established by integrating finite element eigenvalue buckling analysis with mathematical optimisation algorithms. Parametric analyses are carried out on steel web plates with 1–3 longitudinal stiffeners, covering a wide range of geometric and stress parameters. The optimum stiffener position is determined by maximising the buckling coefficient, and the minimum flexural rigidity is defined to ensure effective restraint of web subpanels. Comparisons with existing codes show that current stiffness provisions may be unconservative under bending-shear interaction, indicating the need for refined design methods. The optimisation results are used to train artificial neural networks, which accurately capture the nonlinear mapping between input parameters and optimum stiffener designs. A graphical user interface tool is further developed based on the trained models for practical engineering application. The proposed framework and tool provide valuable guidance for the design of stiffened steel web plates under complex stress states.

Concrete girder bridges worldwide experience structural deterioration and damage over time, necessitating damage identification. However, temperature variations often adversely affect the accuracy of the identification. To address this issue, this study investigated the impact of temperature on structural dynamic characteristics. Dynamic tests on multiple concrete beams under varying damage and temperature conditions were conducted, revealing that the effects of temperature variation on model frequency can mask those of damage. Additionally, recognizing that structural damage is often characterized by stiffness, the research explored a method based on model updating to infer the bending stiffness and boundary stiffness using dynamic characteristics. Subsequently, a damage identification strategy that distinguishes local and overall structural damage from global stiffness variations was proposed, while also considering the impact of temperature. This approach integrates a comprehensive statistical methodology, employing the iterative Grubbs test to identify local damage units and Monte Carlo simulation for non-destructive baseline generation. The method was validated through experiments, which demonstrated promising results in damage identification.

Steel box girders are commonly used in long-span bridges. The orthotropic steel deck (OSD), which directly supports traffic loads on steel box girders, is susceptible to fatigue failure in areas like U-ribs, welds, and openings. This reduces structural performance and service life while increasing demands for operation and maintenance. This study presents a novel framework for fatigue crack detection in steel box girders using computer vision and SLAM, operable without Global Navigation Satellite System (GNSS) signals (e.g., GPS) and under low-light conditions. A handheld device was developed to capture synchronized LiDAR point clouds and high-resolution crack images, followed by deep learning-based crack identification and projection-based localization. The localization error of laboratory tests and field tests are 21.6 mm and 33.5 mm, respectively. SLAM demonstrated strong robustness, with only one degradation observed in ten scans. These quantitative results confirm the effectiveness and practical feasibility of the proposed framework, providing a foundation for future automated inspections in steel box girder.

This study systematically analyzes temperature effects during construction of steel-concrete-steel composite bridge towers, proposing a multi-physical field coupled simulation process integrating dynamic hydration heat release model, zonal solar radiation loading algorithm, and convective boundary parameter correction method. A 3D transient temperature field model is established to develop structural deformation prediction, considering combined effects of concrete hydration heat, solar radiation and ambient temperature change, with ANSYS incorporating non-uniform temperature field and early-age concrete hardening shrinkage. Verified via full-scale tests on Shiziyang Bridge's 342-meter tower, the numerical model shows concrete core hydration heat peaks at 70°C (1.8 days) with ±2°C deviation from field data. Solar radiation significantly affects concrete within 15 cm depth (max 13°C fluctuation) but is spatially separated from hydration heat. Steel plate deformation from liquid concrete lateral pressure accounts for 14% of instantaneous max displacement and 25% of final residual deformation. This method supports thermal deformation control in large composite structures, revealing material thermal expansion coefficient differences' critical impact on synergistic deformation.

Reticulated shells are highly sensitive to initial geometric imperfections (IGIs), and may fail or collapse due to IGIs in extreme cases. However, existing 3D reconstruction studies predominantly focus on straight members and rarely address multi-construction-phase detection, necessitating further research for IGI detection considering curved members and throughout production, transportation, and installation processes. A manufacturing BIM-based method for detecting IGIs was carried out in the Beijing Sub-Center Jingfan Roof Project. By converting manufacturing BIM into synthetic point clouds and developing shape and pose coordinate system (SPCS) generation algorithm, the shape and pose of steel members were accurately recognized across pre-, mid-, and post-construction stages. Engineering applications demonstrated the method’s capability to track IGI evolution, identify key contributing factors, and support construction control.

Reality capture technologies, including terrestrial laser scanning, mobile laser scanning, and UAV-based sensing, offer non-contact means to document civil infrastructure. The ultimate goal of Scan-to-Engineering is to convert dense point clouds into engineering-usable models for measurement, information management, simulation, and decision support. Drawing on 123 core studies, this review examines recent advances in acquisition, point cloud processing, model generation, and downstream applications across civil infrastructure. The reviewed work is organized into three task-specific outputs: Scan-to-Geometry, Scan-to-BIM, and Scan-to-FEM. Engineering Model Readiness Levels and Validation Maturity are used to relate model content to evidence strength. Current studies show substantial progress in geometry-ready, information-ready, and selected analysis-ready workflows, but decision-ready applications remain limited. The review concludes by identifying challenges in observability, automation, uncertainty traceability, interoperability, and updatable digital twins.

This study proposes a deep learning surrogate-based differentiable optimisation inversion method for structural condition assessment of assembled slab girder bridges, addressing the limitations of traditional finite element model updating in handling heterogeneous hinge joint deterioration and girder stiffness degradation. An improved hinge-jointed slab-girder theory is first developed to efficiently compute bridge deflection responses and generate large-scale damage condition-bridge response datasets. A deep learning surrogate model is then constructed to replace finite element analysis, providing a fully differentiable forward mapping from damage condition parameters to structural responses. Leveraging this differentiable nature, the assessment task is formulated as an optimisation-based inversion problem solved via gradient descent. The method is validated on an existing bridge through a field stacked heavy-weight loading test. Results show that the proposed approach accurately identifies the most deteriorated hinge joints and correctly ranks girder stiffness degradation. The reproduced deflections achieve an RMSE of 0.048 mm, outperforming gradient-based and genetic-algorithmbased finite element model updating. In addition, the method exhibits substantial efficiency gains, as the trained surrogate model can be reused for repeated assessments. Overall, the proposed framework provides a rational, accurate and highly efficient solution for structural condition assessment of assembled slab girder bridges, and holds significant potential for applications involving multiple condition assessments, such as wholelife-cycle bridge maintenance and maintenance of bridge groups with identical configurations along highways.

As large language models (LLMs) continue to demonstrate their potential in handling complex tasks, their value in knowledge-intensive industrial scenarios is becoming increasingly evident. Fault diagnosis, a critical domain in the industrial sector, has long faced the dual challenges of managing vast amounts of experiential knowledge and improving human–machine collaboration efficiency. Traditional fault diagnosis systems, which are primarily based on expert systems, suffer from three major limitations: ① ineffective organization of fault diagnosis knowledge, ② lack of adaptability between static knowledge frameworks and dynamic engineering environments, and ③ difficulties in integrating expert knowledge with real-time data streams. These systemic shortcomings restrict the ability of conventional approaches to handle uncertainty. In this study, we proposed an intelligent computer numerical control (CNC) fault diagnosis system, integrating LLMs with knowledge graph (KG). First, we constructed a comprehensive KG that consolidated multi-source data for structured representation. Second, we designed a retrieval-augmented generation (RAG) framework leveraging the KG to support multi-turn interactive fault diagnosis while incorporating real-time engineering data into the decision-making process. Finally, we introduced a learning mechanism to facilitate dynamic knowledge updates. The experimental results demonstrated that our system significantly improved fault diagnosis accuracy, outperforming engineers with two years of professional experience on our constructed benchmark datasets. By integrating LLMs and KG, our framework surpassed the limitations of traditional expert systems rooted in symbolic reasoning, offering a novel approach to addressing the cognitive paradox of unstructured knowledge modeling and dynamic environment adaptation in industrial settings.

The normal service of reinforced concrete hollow-slab beams, which are widely used in small- and medium-span bridges, is a topic of interest. This paper presents an experimental study on four existing hollow-slab beams removed from a real bridge. The mechanical properties, including failure mode, load-deflection response, crack development and load-strain response, were investigated. Moreover, a fibre beam element-based model updating method was applied to these hollow-slab beams for material property parameter inversion and quantitative structural performance evaluation. The two-step model updating achieved accurate inversion of the material property parameters. The updated fibre beam element-based model can reproduce the load-deflection response and crack width development well, i.e., achieve quantitative structural performance evaluation, and can assist with further maintenance decisions such as designing a strengthening scheme. The results indicate that evaluating the ultimate capacity on the basis of stiffness is not very reliable because a member with greater stiffness may not always possess a greater ultimate capacity. The yield strength of steel bars is the predominant parameter of the ultimate capacity; thus, to evaluate the ultimate capacity, more attention should be given to assessing the yield strength of steel bars instead of testing the stiffness under a normal service load. Although the initial stiffness values of these existing hollow-slab beams decreased by 17 similar to 33 %, the beams exhibited high ductility, and their actual residual ultimate capacity values exceeded the design expectations by 7 similar to 10 %. In future bridge maintenance projects, the fibre beam element-based model updating method can be implemented with deflection and crack data from field static load tests to predict material property parameters and to quantitatively evaluate performance.

Construction control of large-span spatial steel structures faces significant challenges, particularly during the transition from partition assembly to structural closure. Factors such as component cutting errors, splicing inaccuracies, and temperature-induced deformations further complicate this process. Traditional deformation control relies on forward calculations based on design models and cannot establish a closed loop connecting analytical models with as-built structures. To address these limitations, a mechanics-based digital twin model is proposed using 3D scanning and finite element (FE) model updating. Real-to-virtual model establishment is achieved through 3D point cloud data and self-weight compensation; virtual-to-real construction prediction is performed by considering temperature effects and boundary condition transformations. The proposed method is applied to a high-speed railway station in Fujian Province, China. Accurate predictions of the structural deformations and cutting lengths of embedded members in construction are obtained, with an average relative error of 0.25 %.

Ensuring the safety and reliability of common reinforced concrete hollow slab beam bridges through structural performance evaluations is crucial for maintaining highway system functionality, social stability, and economic prosperity. This paper presents field tests and a structural performance evaluation of an existing hollow slab beam bridge. First, a lateral load distribution test was conducted to reveal the lateral load distribution performance of this bridge. Second, a single-beam destructive test was conducted to investigate the structural performance of the hollow slab beams. A modified hinge-connected slab/beam method was presented to estimate the actual stiffnesses of hollow slab beams and hinge joints, revealing the lateral load distribution performance and damage distribution of the bridge. On the basis of the estimated actual stiffnesses and experimental data, the flexural stiffnesses and deflections of hollow slab beams under the serviceability limit state were evaluated. In addition, the applicability of the flexural stiffness formulas in JTG 3362-2018, AASHTO LRFD BDS and EN 1992-2 to the structural performance evaluation of existing bridges was assessed. Moreover, the formulas in these standards were adopted to estimate the bending capacities of hollow slab beams, and their applicability was also discussed. This study offers guidance for evaluating the structural performance of existing concrete bridges.

To address the limitations of current inspection technologies and equipment for internal defects in steel box girders, this study proposes a tracked inspection robot–based defect detection system. A four-swing-arm tracked robot with multiple locomotion gaits was designed to adapt to the complex internal environment of steel box girders. Equipped with a camera, LiDAR, and lighting system, the robot enables rapid and low-cost acquisition of images and point clouds in its surrounding area. Its control system adjusts motor parameters based on external sensory information and preset strategies. A defect identification and localization system was developed, which employs image recognition, SLAM-based mapping, and projection localization to restore defect positions to the global coordinate system of the bridge, thereby supporting data sharing and visualization of defect distribution inside the girder. To validate the system, a prototype was tested in field experiments on the Huangmaohai Sea-Crossing Passage. The results demonstrate that the tracked inspection robot system exhibits strong environmental adaptability, achieves small errors in both robot localization and defect three-dimensional positioning, and preliminarily realizes intelligent detection of surface defects inside steel box girders.

Objective The Shenzhen-Zhongshan Link, an expressway that connects the cities of Shenzhen and Zhongshan, has a total length of 6845 m, of which the immersed tube section is 5035 m long and uses a steel shell-concrete structure. The design and construction of the final joint posed notable challenges, provided the complex sea conditions and site-selection constraints in the construction area. As a solution, the Shenzhen-Zhongshan Link innovatively adopted the "prefabricated push-type construction method" for the construction of the final joint, considerably increasing the construction efficiency. This paper conducts a detailed mechanical analysis of the procedures involved in the "prefabricated push-type construction method" employed in the Shenzhen-Zhongshan Link. Methods First, the detailed construction process of the underwater push-out final joint in the link is discribed. The underwater construction of the final joint in the link is split into six main processes, covering key construction steps such as steel-shell transportation, water pumping and pressure fitting, and steel tie rod welding. Subsequently, this paper conducts a model verification of the final joint during the construction phase. Monolithic finite element models are established for the push-out part and expanded part, and finite element calculations are conducted on the basis of the loads of each working condition to confirm structural safety. Finally, a detailed mechanical analysis of the underwater push-out process is conducted; this paper observes that the process involves changes in the internal forces of the steel rods and the deformation of the GINA waterstop. This paper observes potential structural safety risks in the relevant process. Therefore, theoretical calculations and finite element model verifications are conducted for this special stress condition. A theoretical analysis model is established, and the effect of rail friction on the rebound amount of the GINA waterstop is studied via formula derivation. A refined finite element model is established to analyze changes in the internal forces of the steel rods during the underwater push-out process. Results The results of the model verification during the construction phase indicated that under all working conditions, the maximum stress and floor deformation of the push-out part and expanded part were within the design safety range. This suggested that the structural design of the "prefabricated push-type construction method" is relatively reliable with a considerable safety margin. The results of the mechanical analysis of the underwater push-out process showed that rail friction caused greater rebound on the upper side than on the lower side, hence generating greater tensile forces in the upper steel rods. Furthermore, the underwater push-out process may lead to uneven spatial distribution of internal forces in the steel tie rods. Conclusions The "prefabricated push-type construction method" adopted for the final joint in the Shenzhen-Zhongshan Link exhibits relatively structural-stress characteristics during the construction phase. This paper verifies the most unfavorable conditions in each construction process, and the results show that the relevant structural design is reasonable with a sufficient safety margin. During the underwater push-out process, uneven spatial forces can be generated in the rods because of the influence of rail friction and the spatial distribution of the steel tie rods on the cross section. This study suggests that similar construction processes should monitor tie rod stress data and flexibly employ anti-backward devices to ensure structural safety.

Reticulated shells are highly sensitive to initial geometric imperfections (IGIs), and may fail or collapse due to IGIs in extreme cases. However, the regular buckling mode method is limited in handling measured IGIs or predicting effects of IGIs during early construction phases. A new methodology integrates finite element model (FEM) update with IGI simulation techniques, including IGI addition, node classification strategies, 3D normal distribution model, and an asynchronous method to decouple joint and member imperfections through forward/inverse transformations. A field test was carried out in Beijing Sub-Center Comprehensive Transportation Hub Jingfan Roof Project. Validation is achieved via field static loading tests, in which FEM incorporating measured IGIs (via 3D scan data) demonstrates sufficient accuracy. Structural safety assessment shows that measured IGIs change the pattern of structural stress ratio or overall stability. Stochastic simulations establish construction control standard of IGIs, providing reliable and efficient guidance for the construction process. Furthermore, the analysis reveals that measured or stochastic IGIs may marginally increase the structural stability factor, which is an unexpected finding against convention.

Wind energy is a sustainable and renewable energy source, valued and invested in by national governments. Wind power currently has significant installed capacity and growth, and wind turbines will face substantial maintenance demands in the future. However, inspection and assessment of wind turbine towers lack sufficient research attention and effective automated methods. Fine cracks in large-scale images of wind turbine towers are often obscured by noise, making accurate identification difficult. The projection of the same crack in adjacent images is susceptible to double counting, leading to incorrect evaluations. This paper proposes a crack assessment method for concrete wind turbine towers based on unmanned aerial vehicle (UAV) imaging, utilizing computer vision and artificial intelligence (AI). The flight trajectory and photographic strategy for high-rise structures are designed for data acquisition. An attention-enhanced grid-based convolutional neural network (CNN) for crack identification is developed, along with an incremental crack projection algorithm to address overlapping regions in adjacent images. Field tests demonstrate that the proposed method achieves excellent accuracy and efficiency in crack identification, localization, and quantification. The image acquisition strategy ensures reliable crack identification and three-dimensional reconstruction. The classification model, optimized by the global receptive field and attention mechanism, combined with digital image processing (DIP) adapted to grid-based classification, effectively filters noise and extracts fine cracks. Three-dimensional reconstruction and incremental projection based on structure from motion (SfM) provide accurate and efficient crack localization, allowing precise crack parameters to be obtained from the surface model for structural assessment. The proposed method integrates drone imaging, AI-based crack recognition algorithms, and three-dimensional reconstruction and projection techniques, effectively enabling automatic crack assessment for concrete wind turbine towers, thereby assisting more informed maintenance decisions.

The detection of bridge cracks is a crucial task in infrastructure health monitoring, as it plays a significant role in ensuring bridge safety and extending their lifespan. Current methods for segmenting bridge cracks often face challenges such as discontinuous segmentation results, excessive noise, and substantial quantization errors. To address these issues, this paper introduces a Fine Crack Refinement Segmentation Network (FCR-Net). This network innovatively integrates the Transformer architecture with rendering concepts from computer graphics and incorporates Full Attention (FA) modules, Feature Pyramid modules for multi-scale feature fusion, and rendering-based boundary refinement modules. These components collectively enhance the precision and continuity of fine crack segmentation. Furthermore, the paper explores transfer learning and fine-tuning strategies during the model training process and conducts a comprehensive evaluation of the model's performance through interpretability analysis and ablation experiments. Experimental results demonstrate that FCR-Net significantly outperforms traditional convolutional neural networks and baseline methods in terms of segmentation accuracy, boundary refinement, and quantization precision for fine cracks. The proposed method effectively enhances the automation and accuracy of bridge crack detection and offers valuable insights for the automated detection of other infrastructure health monitoring tasks.