Abstract:

Generative artificial intelligence （AI） large models face challenges in the engineering construction domain， including insufficient domain-specific cognition and unclear application scenarios and undefined workflows. To explore the potential applications and engineering value of generative AI large models in the vertical field of construction ， this paper systematically analyzes and reviews their potential applications and challenges in China’s construction sector. Through literature review， expert consultations， and value-chain analysis， the current state of AI model applications in construction is clarified. A “Construction-3L” large-model application framework is established， consisting of three dimensions： the general model layer （L1）， the industry model layer （L2）， and the scenario model layer （L3）. Based on this framework， application opportunities and value-empowerment pathways of large-model technologies in the stages of project initiation， design， and construction are explored. A development workflow for construction-scenario large models is proposed. Furthermore， common challenges related to algorithms， data （computing evidence）， and evaluation mechanisms in engineering applications of large models are analyzed， and future prospects are discussed with respect to application scenarios， technological innovation， and policy directions in the engineering construction field.

Abstract:

To address the issues of hallucinated generation and high deployment costs encountered by large language model （LLM）–based question answering systems in construction scenarios， this paper proposes a construction-oriented question answering system based on a collaborative expert mechanism. The system integrates shared experts and routing experts in a coordinated manner， which significantly improves the accuracy of answer generation and inference efficiency while preserving the model’s expressive capacity and reducing computational overhead. In addition， a domain knowledge base–injected fine-tuning strategy is introduced to guide the model to deeply learn professional semantics in the construction domain during training， thereby enhancing its understanding of engineering-related texts and ensuring that the generated responses better align with practical engineering requirements. Experimental results demonstrate that， with only approximately one-third of the model parameters activated， the proposed system achieves a generation semantic similarity of 81.1%， effectively balancing efficiency and performance and providing an efficient， reliable， and construction-specific intelligent decision-support tool for construction management.

Abstract:

With the application and development of digital technology， the concept of digital lean construction has been proposed and increasingly emphasized. To support the application and promotion of digital lean construction， this paper explains the core technologies of the “Evaluation Standard for Digital Lean Construction Maturity of Construction Engineering Projects”， namely the evaluation system and methodology for digital lean construction. First， based on a digital lean construction management model， a maturity assessment index system is established. Then， the weights of the evaluation indicators and the evaluation methods are determined. Finally， its practical application of the standard in engineering projects is introduced and analyzed. The results show that this standard can be used to quantitatively assess the level of digital lean construction in engineering projects.

Abstract:

To address the modeling and quantitative assessment of cascading failures in emergency events induced by the coupling between transportation networks and water supply networks， this study takes the central urban area of Shanghai as a case study. Using Taxi GPS data， along with GIS （geographic information system） data for road and water supply networks， it constructs a dynamically coupled network through spatial matching and temporal attribute integration. Based on percolation theory， it proposes a network influence index to quantify the impacts of cascading failures. The results indicate that there are significant spatiotemporal coupling effects between the transportation network and the water supply network. In particular， during peak hours， failures in the water supply network have a direct impact on the connectivity and operational efficiency of the transportation network. Furthermore， by developing decoupling strategies， it proposes optimization schemes for network planning and maintenance， which can reduce cross-network cascading failure effects between the two networks. These findings provide a useful reference for enhancing the resilience， safety， and operational reliability of critical urban infrastructure systems.

Abstract:

The behavior of multiplanar KK′X gap joints in circular hollow section （CHS） trusses was investigated experimentally and numerically. First， failure tests on two KK′X joints with different geometric parameters were conducted with different geometric parameters to examine their deformation feature， failure mode， load-deformation responses， ultimate bearing capacities， and strain intensity distributions. Next， a finite element （FE） model of KK′X joints was established and validated against the experimental results. The plastic zone and its development were analyzed using this model. The results indicate that the failure mode of the all joints belongs to chord plasticization. The depression deformation of chord wall at K compressive brace came to maximum. Increasing the brace diameter could enhance the ultimate bearing capacity of the joint. The stress distribution near the intersection between brace and chord is more complicated， with the maximum strain intensity occurring at the chord saddle adjacent to the K compressive brace， where yielding initiates first. The plastic zone then propagates outward， eventually encompassing the most region of the joint region.

Abstract:

Parametric analysis for the ultimate bearing capacity （UBC） of multiplanar KK′X gap joints in circular hollow section （CHS） trusses is conducted using the finite element （FE） model validated by tests. The study reveals that τ_K and τ_X have minimal impact on the UBC of the joints， whereas other geometric parameters exert a greater influence. UBC of K-brace significantly decreases with the increase of axial force in the X-brace. The axial compressive force in the chord member adversely affects the UBC， while axial tension has little impact. Based on the calculation formulas for UBC of uniplanar K-joints in design standard of steel structures， the formula for UBC of multiplanar KK′X joints subjected to axial force on K-brace with chord plasticization failure is proposed by means of geometric adjustment coefficient μ_KKX and influence coefficient ψ_m considering the ratio of axial force between braces. The accuracy of this formula has been validated by FE analysis and experimental data. Using the calibration method prescribed in design standard of steel structures， a design formula for UBC of multiplanar CHS KK′X joints is finally recommended， providing a practical design guideline for the engineering application of this type of complicated tubular joints.

Abstract:

To investigate the influence of socket connections on the punching shear behavior of pile caps. Experimental tests were conducted on 1 cast-in-place （CIP） pile cap and 3 socket connection （SC） pile caps with varying embedment depths and reinforcement configurations. The punching shear performance of SC pile caps was compared with that of the CIP counterpart， and the effects of embedment depth and U-shaped reinforcement on punching shear capacity were examined. Additionally， the existing code formulas for estimating the punching shear capacity of SC pile caps were evaluated and modified based on the test results. The results indicate that SC pile caps exhibit a damage development similar to that of the CIP， albeit with a 6.75% to 7.85% reduction in punching shear capacity. Increasing the embedment depth and incorporating U-shaped reinforcement have minimal impact on vertical load-displacement behavior and punching shear strength of SC pile caps. The study highlights the inadequacy of existing code formulas in capturing the variation of punching shear strength with respect to different embedment depths and reinforcement arrangements. A modified strut-and-tie model is proposed， demonstrating its ability to provide reasonable predictions of the punching shear strength for SC pile caps.

Abstract:

This paper provides a detailed review of research on isothermal coupled solute transport （ICST） theory and， and compares solute flux predictions from different theoretical models using COMSOL numerical software. The results show that ICST models include both the counter-diffusion （CD） model and the salt-diffusion （SD） model. Although the flux predictions of the CD and SD models are basically consistent， the SD model is theoretically more accurate. The advection-diffusion （AD） model overestimates solute transport. However， when the restricted diffusion is considered， the flux predictions obtained by the AD model substantially diminish the discrepancies observed with the ICST models. In practical applications， it is recommended to appropriately use the SD model or ADC model considering site-specific conditions. Given the complexity of the near-field environments in engineering barriers， future theoretical research should focus on elucidating coupled solute transport mechanisms under non-isothermal， unsaturated， and soil deformation conditions， and on developing a fully coupled thermo-hydro-mechanical-chemical solute transport model that accurately reflects the effects of concentration and temperature on model parameters.

Abstract:

To overcome the limitations of traditional intelligent optimization algorithms， such as low accuracy， slow convergence， and susceptibility to local optima， this paper proposes a multilevel learning adaptive particle swarm optimization （MLAPSO） algorithm. The algorithm incorporates a best-point set strategy and multiple search mechanisms， including global search， the FDB mechanism， and the Levy flight strategy. Tests on the CEC-2022 benchmark functions demonstrate that MLAPSO significantly outperforms traditional optimization algorithms in terms of search accuracy and stability. Furthermore， combined with the load-structure model of foundation pit excavation， it proposes a method for soil parameter inversion and staged deformation prediction of foundation pits based on MLAPSO. The method is validated using monitoring data from a metro station. Results show that the approach can accurately invert soil parameters， and the predicted deformation of the retaining structures based on these parameters aligns closely with measured deformations， confirming the effectiveness and reliability of the method.

Abstract:

The site selection of large temporary facilities plays a vital role in the construction of high-speed railways. They are mainly used to produce and provide prefabricated beams for the construction of high-speed railways. To address the limitations inherent in traditional site selection methods—specifically the high costs of preliminary surveys， excessive subjectivity， and a lack of quantifiable assessment criteria—this paper systematically identifies the logical framework of influencing factors， constructs a multi-level indicator system for site evaluation and， in combination with ArcGIS spatial analysis， develops a site selection model for large temporary facilities of high-speed railways construction. The model is verified using the locations of large temporary facilities that have been built in Shandong Province， and the model accuracy is found to be 90%， which confirms the feasibility and effectiveness of the proposed site selection model. This paper can provide a theoretical basis for the site selection of similar facilities in the future， and also help form a new working paradigm for the site selection of large temporary facilities for high-speed railways.

Abstract:

To establish a comprehensive and scientific evaluation system for the operational performance of precise measurement and precise tamping （PMPT） on ballasted tracks， taking into account four aspects： track quality improvement， track quality tracking， track wavelength energy improvement， and vehicle dynamic response， a comprehensive evaluation system for the operational performance of PMPT is established. Considering current inspection methods and the characteristics of PMPT operation， nine evaluation indicators are proposed. Based on extensive statistical analysis of track geometry inspection data and vehicle response data， and in accordance with existing management standards， these indicators were classified into six grades. Using the analytic hierarchy process （AHP） combined with expert scoring and evaluation， the weights for different evaluation modules and indicators are determined. A comprehensive evaluation method based on AHP is proposed for assessing the operational performance of PMPT. Through a case study， the operational performance of different PMPT modes from three track maintenance divisions is compared. The analysis results indicate that the operational performance of track maintenance division 1 is the best， followed by track maintenance divisions 3 and 2. This method can be applied to evaluate the performance of different PMPT sections， compare and analyze the performance of different maintenance operation modes， scientifically guide the selection of on-site operation modes， and provide a foundation for the scientific and systematic evaluation of the operational performance of PMPT.

Abstract:

Using a miniature split Hopkinson pressure bar （m-SHPB） system， the dynamic compression properties and severe plastic deformation microstructure of a high-manganese austenitic steel （NPR） were investigated under ultra-high strain rates. The NPR steel demonstrated excellent impact resistance and strain-hardening capability under ultra-high impact loading， with a yield strength of approximately 550 MPa， an ultimate compressive strength of approximately 1 500 MPa， and a fracture strain of 0.7. Analysis of the strain-rate sensitivity coefficient （λ） under quasi-static and dynamic conditions revealed two distinct strain-rate-sensitive regions for NPR steel. At a strain rate of 1×10⁴ s^-1， NPR steel did not fracture， but a significant number of aggregated microvoids formed along grain boundaries on the impact cross section， with the pore initiation direction oriented at an angle of 65° relative to the impact direction. As the strain rate increased to 2×10⁴ s^-1， NPR steel fractured， with adiabatic shear bands forming on the impact cross section due to the coalescence and growth of these microvoids， oriented at an angle of 70° to the impact direction. Additionally， a large number of heterogeneous structures was formed along the impact direction， including kink bands， deformation twins， and rotational nanocrystals generated in twin-twin interaction regions， indicating that NPR steel exhibits complex damage mechanisms under ultra-high strain rates.

Abstract:

To address the issue of flash evaporation caused by insufficient pressure at the inlet of liquid hydrogen pumps in hydrogen refueling stations—a phenomenon that threatens operational stability—it is imperative to develop a dynamic simulation of the onboard transfer process and establish an effective pressure management strategy for liquid hydrogen storage tanks. Using the MatLab software platform， this paper develops a thermodynamic model of the liquid hydrogen transfer process that integrates boil-off gas recirculation and tank self-pressurization. It systematically analyzes the variations in pressure and temperature during transfer， leading to the proposal of a pressure management strategy designed to prevent flash evaporation. Furthermore， it optimizes the target self-pressurization pressure within this strategy， yielding a definitive optimal value. This paper provides theoretical guidance for programming self-pressurization procedures during liquid hydrogen transfer in hydrogen refueling stations.

Abstract:

To address the two key issues of head effect and low recommendation accuracy in traditional collaborative filtering algorithms， this paper proposes a collaborative filtering algorithm based on an adaptive penalty factor and user rating behaviors. The penalty factor adjusts similarity calculations by penalizing popular items， while the adaptive parameter identifies the optimal penalty intensity for different datasets through iterative searching， effectively mitigating the long-tail effect of traditional algorithms. User rating behavior further refines similarity calculations by considering time differences between ratings and variations in rating distribution characteristics， thereby improving algorithm accuracy. The proposed algorithm was validated on four public datasets. For the well-performing MovieLens 1M dataset， the algorithm achieved an average F₁ score improvement of approximately 13.9% over traditional algorithms while maintaining the same recommendation count， significantly enhancing recommendation quality. Additionally， the inverted index was employed to construct the item-user interaction matrix， improving computational efficiency. On the MovieLens 1M dataset， the runtime of the proposed algorithm was reduced by approximately 72.1% compared to traditional methods.

Abstract:

To address the issue of driver fatigue detection， this paper develops a deep HM-LSTM network to achieve fatigue recognition based on facial features. The LOTR model is used to extract facial spatial features. Considering that driver fatigue is a gradual process， it defines eight temporal fatigue features and uses stacked HM-LSTM to capture fatigue semantic features over both short and long time series， in order to accurately identify signs of driver fatigue at an early stage. This approach provides reliable monitoring of abnormal driver behavior through low-cost visual-based methods. Model training and system integration are conducted by using publicly available datasets and self-constructed D1-DBB datasets.

Abstract:

In recent years， the momentum behind the expansion total social financing has shown signs of stagnation. To underlying causes and explore effective pathways to stimulate its growth， this paper employs the XGBoost algorithm along with two interpretability tools to analyze the key factors influencing total social financing growth， yielding consistent and robust results. The findings indicate that consumption and investment are the primary driving forces behind total social financing growth. Time-varying impulse response analysis reveals that since 2012， the ability of investment to drive total social financing has gradually weakened， while the driving role of consumption has steadily increased. The sluggish consumer market has emerged as the main factor restraining recent total social financing expansion. Further regional analysis suggests that the driving effect of consumption on financing growth exhibits spatial heterogeneity， with the eastern region experiencing the strongest impact. Accordingly， this paper recommends to better leverage the fundamental role of consumption to foster a virtuous cycle of “consumption-led investment，” ultimately achieving steady growth in total social financing through a dual engine of consumption and investment.