In recent decades, extensive experimental research has been conducted on reinforced concrete (RC) beams to analyze their behavior under loading, with a particular focus on crack propagation. Cracks significantly compromise the structural integrity, stiffness, and ductility of RC beams, underscoring the need for precise and efficient damage assessment methods. This study proposes a robust framework for structural damage evaluation in longitudinally reinforced RC beams subjected to monotonic loading. The engineering application centers on quantifying two critical damage indicators—displacement (DID) and strength (DIL)—based on surface crack patterns. The artificial intelligence (AI) contribution includes the integration of advanced image processing for crack feature extraction, explainable AI (XAI) for interpretability, and probabilistic modeling to account for material uncertainties. A total of 621 crack images from 15 studies were analyzed using four machine learning models. The framework achieved an R-squared value of 0.85 for displacement-based damage prediction. The results demonstrate the synergy of AI techniques and structural engineering knowledge to deliver transparent, data-driven insights for structural health monitoring and post-damage evaluation of RC beams.

Current research on aluminum alloy composite structures primarily focuses on concrete-filled CFRP aluminum alloy tubular columns, and studies on aluminum alloy-steel-concrete composite columns have been scarce. The aim of this study was to investigate axial compressive behavior of concrete-filled aluminum alloy circular tubular stub columns with in-built H-steel through axial compression experiments and the finite element (FE) method. Eight stub columns were used in the experiments, with the internal core concrete type and H-steel content taken as variation parameters. Analysis revealed that the location of the bulging failure of the specimen was influenced by the internal core concrete type. The aluminum alloy tube and H-steel exhibited good deformation coordination under axial compression. Parametric analysis showed that the in-built H-steel can effectively inhibit the transverse deformation of concrete and delay the damage of stub column specimens. When the same type of concrete was used, the axial compression capacity, axial compressive stiffness, and energy absorption of the columns increased with the increase in steel content. When the steel content was maintained, the specimens with lightweight concrete exhibited higher bearing capacity and axial compressive stiffness. For energy absorption behaviors, the specimens with lightweight concrete and steel content of 8.76% exhibited superior performance. An expanded parameter analysis was conducted utilizing the FE method. Based on the results, a bearing capacity calculation formula for the composite columns was proposed using the superposition theory. The findings of this study can help address the durability issues of steel-concrete and aluminum alloy-concrete members and thus improve structural performance.

Accurately predicting the ultimate shear strength of RC beams, especially those with intermediate shear spans, presents a significant challenge. This complexity arises from the intricate interplay of flexural and shear behaviors within these structures. Although extensively researched, the contribution of arch action to the shear strength of these beams has frequently been underestimated. This study reexamines and quantifies the shear force contribution from arch action in intermediate reinforced concrete beams by introducing the arch action coefficient, alpha. This coefficient is determined using two analytical approaches: the compatibility method and the differential method. Both methods are subsequently validated through experimental testing on reinforced concrete beams. To simplify the calculation of ultimate shear strength, alpha charts are developed through finite-element analysis, incorporating the influences of yield strength, stirrup ratio, longitudinal reinforcement ratio, and span-to-depth ratio on alpha. Furthermore, theoretical models have been developed to consider the combined contributions of concrete and stirrup in beam action calculations. Generally, comparisons between these combined theoretical predictions (for beam action) and numerical simulations (for arch action) against experimental results show strong agreement.

Despite achieving high classification accuracy, current data-driven models often lack reliable confidence measures for their predictions. This limitation is particularly concerning in complex or out-of-distribution scenarios, where model reliability may be compromised due to unquantified uncertainties. Without robust uncertainty quantification, the practical deployment of these models in real-world structural assessments becomes questionable, especially in critical applications such as seismic risk evaluation, where erroneous predictions can trigger severe consequences. To address this gap and enhance the trustworthiness and interpretability of machine learning models, we developed a hybrid conformal prediction (HCP) framework for assessing seismic failure modes in reinforced concrete (RC) columns. This approach not only quantifies model uncertainties but also ensures proper calibration, thereby reducing the risk of consequential failures and improving decision-making in seismic risk evaluation. The proposed framework combines probability calibration to refine reliability scores with conformal prediction to generate statistically valid prediction sets. By integrating ensemble learners, this dual approach ensures accurate uncertainty quantification and guaranteed coverage across confidence level. Validation on 758 experimental RC column datasets confirms the framework’s effectiveness, delivering high coverage and efficiency at practical confidence levels. This advancement enhances the trustworthiness of failure mode assessments, providing engineers with quantifiable confidence metrics for improved decision-making in seismic safety evaluations.

This study investigates the axial compression behavior of circular Concrete-Filled Aluminum Tube (CFAT) short columns, emphasizing the influence of the confinement effect coefficient. Sixteen specimens fabricated from 6063-T5 and 6061-T6 aluminum alloys were tested, demonstrating effective composite action with concrete. Specimens failed primarily by buckling or shear: buckling failures occurred when the confinement effect coefficient exceeded 0.91, whereas low coefficients below 0.31 combined with diameter-to-thickness ratios (D/t) >= 50 led to large diagonal cracks and poor ductility. Finite element analysis showed minimal interaction between aluminum tube and concrete in the elastic stage, which increased gradually in the inelastic phase. The ultimate load capacity rose by up to 26.48% and 33.25% with increases in concrete strength and tube wall thickness, respectively. Due to aluminum's relatively low elastic modulus, introducing the confinement reduction factor (kal) and modulus reduction coefficient (ka) into existing Concrete-Filled Steel Tube (CFST) formulas significantly improved accuracy, aligning predictions with experiments. Theoretical analysis further indicates that when the aluminum content ratio (alpha) is below 0.1, ka approximates 1, allowing direct calculation using CFST theory without reduction. These results provide valuable theoretical guidance for the design and engineering application of CFAT short columns.

Concrete-filled double-skin steel tubes, or CFDSTs, increasingly adopt self-compacting concrete, or SCC, to improve casting quality. However, SCC-specific test data remain limited, and predictive models developed from normal concrete, or NC, specimens often suffer from systematic bias when transferred across material domains. To address this issue, this study establishes an axial-compression database of 189 circular CFDST stub columns, including 161 NC specimens and 28 SCC specimens, and confirms clear cross-domain discrepancies in key structural descriptors and load-carrying capacity. A Dual-Stream Residual Transfer Learning framework, termed DS-ResTL, is then developed for small-sample cross-material prediction. In this framework, a frozen base stream extracts transferable structural mechanics from the NC domain, while a lightweight residual stream learns SCCspecific deviations during target-domain adaptation. The learning objective integrates regression loss, latent-space maximum mean discrepancy alignment, and physics-consistency regularization to improve generalization while preserving mechanically reasonable monotonic relationships. Leave-one-out validation on SCC specimens shows that DS-ResTL achieves high and stable prediction accuracy with R2 = 0.964, outperforming conventional regressors while maintaining physically consistent sensitivity trends. The proposed framework provides an interpretable and robust solution for cross-domain capacity prediction of CFDST columns under limited target-domain data.

To investigate the flexural mechanical characteristics of aluminum-concrete-steel double-walled hollow composite members, four specimens were designed with the hollow ratio as the test parameter, and flexural performance tests were conducted on them, the bending moment-deflection at mid-span curve, bending moment-curvature curve, deformation patterns, and strain distribution characteristics were systematically analyzed. A refined finite element model was established and validated based on experimental data, through numerical simulation, the working mechanism throughout the component's loading process was revealed, simultaneously, the effects of parameters on the bending capacity and flexural stiffness. Finally, a calculation method for the bending capacity was proposed. The research results indicate that compared to concrete-filled Aluminum tube member, the design concept of double walled hollow composite member can effectively improve the collaborative mechanical performance between the various components of the member and enhance the overall deformation capacity of the member. Within the parameters investigated, the ultimate loading creases with the increase of hollow ratio. The deflection curve of the component under bending load is basically consistent with the sine half wave curve, and the longitudinal strain distribution of the mid span section basically satisfies the assumption of a plane section. As the bending load gradually increases, the neutral axis will also gradually move towards the compression zone. Through finite element simulation, it was found that the increase in wall thickness of aluminum alloy tube has a significant impact on the bending capacity and bending stiffness of components. The proposed bending capacity calculation method based on the limit equilibrium theory and unified theory demonstrates strong applicability to aluminum-concrete-steel double-wall hollow composite members.

This paper investigates the problem of quantitative seismic damage assessment in reinforced concrete beam-column joints using image-based damage assessment methods. The research question concerns how visual damage information can be integrated with mechanical uncertainty to predict seismic performance of exterior and interior joints. A physics-informed feature integration framework is developed by combining crack feature extraction, a validated probabilistic joint shear strength model with correlation-preserving Monte Carlo simulation, and machine learning regression. The proposed approach achieves accurate prediction of drift-, strength-, and stiffness-related performance indicators, with coefficients of determination up to 0.86 on unseen data. These results are important for engineers and infrastructure managers as they enable interpretable, risk-informed post-earthquake assessment without reliance on dense sensor networks. The study provides a foundation for future research on field-scale deployment and integration into structural health monitoring systems. It should be noted that the present study is based on laboratory-controlled experimental data, and the applicability of the proposed framework to real-world field conditions requires further investigation.

This study investigates the axial compression behavior of lightweight concrete-filled circular aluminum alloy tubular (LCFAT) stub columns for structural lightweighting applications. Nine groups of specimens with varying parameters were tested to examine the effects of the cross-sectional aluminum ratio and a confinement factor on the load-carrying capacity, axial stiffness, and failure modes. To induce ideal failure patterns under axial loading, the specimens were locally wrapped with carbon fiber reinforced polymer (CFRP) at both ends solely to provide end restraint and ensure uniform load transfer; the CFRP was not part of the load-carrying system. The experimental program is complemented by calibrated three-dimensional finite element analyses, which were further employed to conduct parametric studies on concrete strength, alloy yield strength, and tube wall thickness. Based on classical concrete-filled steel tubular (CFST) theory, a strength design model tailored to the aluminum–lightweight concrete interaction is formulated by introducing an elastic modulus reduction factor and a lateral confinement reduction factor. The proposed model captures the key characteristics of the axial response, and the numerical simulations were used to replicate the observed failure patterns and load–displacement curves, supporting an interpretation of the confinement mechanism and behavioral phases. These results provide a theoretical basis for engineering application and design optimization of lightweight composite columns.

Traditional aluminum alloy gusset (TR-AAG) joints exhibit limited mechanical efficiency because load transfer relies exclusively on the upper and lower cover plates, while the beam web does not effectively participate in force transmission. This study introduces an improved aluminum alloy gusset (IM-AAG) joint incorporating shear connectors and a hollow hexagonal prism to optimize the load-transfer path. A finite element model, validated against existing experimental data, was employed to compare the structural performance of IM-AAG and TR-AAG joints. Parametric analyses were conducted to quantify the effects of key design variables, including the number of web bolts, bolt-hole clearance, bolt pretension, and the thicknesses of the shear connector and cover plate. The mechanical response of the IM-AAG joint was further examined under axial loading, pure bending, pure shear, and combined bending-axial and bending-shear conditions. An analytical moment-rotation model was also developed to predict the global rotational behavior of the joint. The results show that the IM-AAG joint exhibits pronounced enhancements in initial bending stiffness and ultimate flexural capacity. The addition of the hollow prism and shear connectors effectively suppresses bolt slip and mitigates web buckling, promoting a more ductile and progressive failure mode. Increasing the cover plate thickness and bolt quantity further improves flexural performance. Under combined bending-axial loading, axial compression reduces flexural resistance; however, this adverse effect is alleviated by the shear connectors. Overall, the proposed joint configuration and analytical model offer a robust and practical basis for the design and optimization of aluminum alloy joints in spatial structural applications.

In Changqing Oilfield，ø60.3 mm （23/8 in） conventional tubing is replaced with ø50.8 mm （2 in） coiled tubing （CT） in newly commissioned gas wells， for purpose of high-efficiency well completion and low-cost development. In this study， the limitations of existing CT completion technologies and the need for main processes in the lifecycle of tight gas wells after they are put into production were analyzed. Then， according to the overall design concept of one-trip CT completion and multi-process seamlessly connected effective production， a multifunctional bottomhole assembly （BHA） for ø50.8 mm CT completion and production was designed and developed. The BHA comprises connector， upper sliding sleeve， soluble balls， fracturing fluid flowback cartridge， shear pins， lower sliding sleeve， flow nipple， plug， flapper bivalve of sliding sleeve， valve flapper， stop sleeve， and screen， etc. It is divided into six functional areas： connection area， fracturing fluid flowback area， ball-drop plugging area， choke area， temporary plugging area， and filtering area. The BHA can realize the operations such as landing of CT completion string under pressure， flowback or induced flow of fracturing fluid in the initial stage of production， bottomhole choke production in high-pressure period， liquid carrying of velocity string in low-pressure period， plunger gas lift in intermittent period， and tripping-out under pressure after ball-drop plugging in depletion period. Field applications in over 30 wells in the Sulige gas field in Changqing have demonstrated a 100% implementation success rate， verifying the feasibility of the CT completion and production process in newly commissioned gas wells and the reliability of the multifunctional BHA. The study results provide great support to the efficient and beneficial development of tight gas in China.

Dynamic analysis accurately captures progressive collapse behavior but is computationally expensive. To improve efficiency, the U.S. DoD guidelines use a Dynamic Increase Factor (DIF) to approximate inertial effects. However, its empirical provision for RC frames is ambiguous and ignores structural uncertainties. This study proposes a scenario-specific model with a boundary correction parameter to simplify the DoD procedure and improve applicability for both slab-incorporated and bare frames. A Bayesian framework is then introduced to incorporate structural uncertainties, enabling a probabilistic reassessment of the DIF. Results show that although slab participation has limited influence on the mean DIF, it substantially reduces the impact of uncertainties. Posterior parameter estimates derived from 400 slab-incorporated frame simulations, generated from uncertainty-informed samples. This Bayesian framework successfully addresses the deterministic limitations of the DoD guidelines, producing mean predictions consistent with observations and a 95 % credible interval that captures nearly all observed data. For practical application, two design-oriented models are developed: a highfidelity probabilistic analytical model that replicates Bayesian predictions at a lower computational cost, and a deterministic envelope model that conservatively bounds 90 % of observed DIF values. Both models provide efficient, reliable design tools that mitigate uncertainty-related risks using only deterministic data.

Reliable assessment of progressive collapse performance in reinforced concrete (RC) frame structures requires explicit consideration of material and geometric uncertainties. However, conventional sampling-based approaches typically requires large numbers of finite element simulations, which substantially increase computational costs. To alleviate it, this study develops an adaptive reinforcement learning-based ensemble (ARLE) framework for the reliability evaluation of RC frames under structural uncertainties. In this framework, a rewarddriven learning strategy is employed to adaptively update classifier ensemble weights to maximize model's predictive performance. In parallel, a tiered sampling scheme is introduced to identify the minimum training sample size required at various force levels, substantially reducing simulation demand without compromising accuracy. High-fidelity finite element simulations combined with correlation-controlled Latin hypercube sampling are used to construct an uncertainty-informed training dataset. The proposed framework is applied to a dynamic central-column removal scenario, capturing the transition of structural response from low to high collapse probability with increasing force levels. SHAP-based interpretability analysis further reveals that the concrete compressive strength and the yield strengths of slab and beam reinforcement are the dominant contributors to structural reliability. Overall, the proposed ARLE framework provides an efficient and interpretable alternative to conventional sampling-based methods for progressive collapse reliability assessment.

Predicting bearing capacity of reinforced concrete (RC) columns under specific failure modes plays a dominant role in the seismic safety of building structures. Traditional empirical expressions and existing machine learning (ML) methods struggle to balance model prediction accuracy and explainability. This paper presents an interpretable ML framework designed to accurately identify failure modes and predict bearing capacity of RC columns. A two-stage random forest algorithm is adopted to identify three typical failure modes. Gaussian process regression (GPR) models with various kernels are used to predict the probabilistic flexural and shear bearing capacity of RC columns. The Shapley additive explanations method is proposed to conduct a sensitivity analysis of key parameters for classification and regression learning. Results show that the proposed two-stage random forest method outperforms traditional one-stage classifiers in accuracy. The GPR-based bearing capacity model can output probabilistic distributions by Bayesian learning from high-dimensional inputs. The Matern3-based GPR model achieves better prediction results, superior to artificial neural network and support vector learners. The proposed flexural and shear capacity models also have better prediction accuracies than ACI, ASCE, and Eurocode design codes. Several RC columns attributes are identified as critical characteristic parameters in failure modes and bearing capacity prediction.

The metal-metal interactions between different metal atoms within dual-metal-supported catalyst have garnered significant attention in effectively regulating the chemical and electronic structure of active sites, thus providing enormous opportunities to optimize the catalytic performance. Herein, by loading Cu clusters and Pd nanoparticles on TiO2 (CuPd/TiO2), it is found that Cu clusters enable the metalize transformation of Pd as a result of the metal-metal interactions between Cu and Pd, which remarkably improves the surface catalytic reaction as well as the separation and transfer of photogenerated carriers. The resultant photocatalyst exhibits an excellent activity for H2 evolution coupled with benzyl alcohol oxidation, achieving yield rates of 43.7 mu mol h- 1 for H2 and 22.0 mu mol h-1 for benzaldehyde, and an apparent quantum efficiency (AQE) of 20.4 % at 313 nm for hydrogen evolution. This integration of metal-metal interactions within the dual-metal-supported catalyst open a new avenue for the design of advanced photocatalysts.

Accurately predicting shear deformation in reinforced concrete (RC) beams with intermediate shear spans remains a critical challenge in structural engineering. This difficulty arises from the intricate interplay between flexural and shear responses, as well as the significant reduction in shear stiffness after diagonal cracking. To address this issue, this study conducts a thorough investigation into the shear deformation mechanisms of RC beams with shear span-to-depth ratios ranging from 2 to 4. A new piecewise analytical model is developed to characterize the complete shear force-deformation response of structural elements, explicitly accounting for the influence of arch action on post-cracking shear stiffness. Through a comprehensive parametric study, the degrading effect of arch action on stiffness is quantified, with model predictions rigorously validated against available experimental results. Additionally, a decoupling methodology is introduced to isolate shear deformation from total measured deformations by systematically removing flexural contributions. This technique enables extraction of shear deformation components, providing a reliable benchmark for validating the proposed analytical model. Comparison between the proposed model and decoupled experimental results demonstrates good agreement, validating the model's accuracy and efficiency in predicting shear deformation behavior. These findings offer valuable insights for enhancing the structural design and health monitoring of the RC beams.

Accurate and efficient prediction of load-displacement behavior in reinforced concrete (RC) beams is essential for data-driven interventions and predictive maintenance in structural engineering. Such predictions are critical for ensuring the safety, reliability, and longevity of structures. Develop a comprehensive approach that incorporates advanced deep learning (DL) techniques, leveraging historical data and time series modeling to achieve highprecision predictions. This study introduces a hybrid feature-orientation Transformer-LSTM (HFT-LSTM) model for accurately predicting the load-displacement response of RC beams, from initial loading to failure. It leverages a Transformer architecture, embedded with a multi-headed attention mechanism, to achieve a deeper representation of constant features and dynamic fusion of variable features, enabling a more comprehensive understanding of the RC beam's behavior. It employs a bidirectional LSTM to capture dynamic and temporal relationships within the displacement series. To gain a deeper understanding of the history-dependent mechanical behavior of RC beams, we conducted a detailed analysis of the influence of input features on the loaddisplacement curve. The proposed model significantly outperforms existing DL models, achieving a 46 % reduction in Mean Absolute Error (MAE), a 58 % decrease in Root Mean Squared Error (RMSE), a 5 % increase in R-squared (R2) and Explained Variance (EV), and a 36 % reduction in Median Absolute Error (MedAE).

Despite advancements in machine learning (ML) that have boosted structural performance prediction, current ML models can still struggle to generalize to unseen situations, leading to performance degradation. This vulnerability arises from their overreliance on data, neglecting established engineering principles such as mechanical priors. Models trained on specific data distributions can suffer significant accuracy degradation when encountering inputs that fall outside those distributions. To overcome the limitations of data-driven models with unseen data, a mechanics-guided Gaussian process (MGGP)foraccurate prediction ofshear strength in reinforced concrete (RC) beams is proposed. The complex variation of shear strength in RC beams was captured using a Gaussian process (GP) model with a mean function derived from mechanical principles and a hybrid kernel to account for inherent prediction variability. This combination allows for accurate prediction of shear strength while considering the underlying physical mechanisms. This approach leverages domain knowledge from mechanics by incorporating a relevant design equation into the mean function of a GP model. This integration significantly enhances the models' ability to predict shear strength by capturing the underlying physical principles governing shear strength. Cross-validation studies have shown that the MGGP offers consistent performance compared to traditional GPs in predicting the shear strength of RC beams.

Corrosion of reinforcement is a crucial factor that significantly impairs the seismic performance of reinforced concrete (RC) columns in building structures. This paper examines three key parameters related to the seismic performance of RC columns: seismic failure modes, maximum bearing capacity, and effective stiffness. The design methodologies for RC columns vary significantly depending on the failure modes, as different failure modes involve distinct mechanical characteristics, thereby influencing the corresponding bearing capacity and effective stiffness. In this study, an intelligent approach is proposed, which first performs the classification of failure modes and subsequently predicts the associated bearing capacity and effective stiffness through regression analysis. Seven supervised learning algorithms and one deep learning algorithm were utilized in this study, with particular emphasis on the k-nearest neighbor (KNN) algorithm due to its simplicity and effectiveness. A comprehensive dataset comprising 221 corroded column specimens under cyclic loading tests was collected, and the findings indicate that the KNN-based model exhibits a high degree of accuracy in failure modes classification (accuracy=0.91), bearing capacity prediction (R2=0.99), and effective stiffness prediction (R2=0.95). Comparisons with other algorithms were conducted, and the results indicate that the KNN algorithm outperforms the others. Furthermore, the bearing capacity and effective stiffness predictions generated by the KNN algorithm were compared with empirical formulas from design codes, revealing the clear superiority of the proposed method. Thus, it can be concluded that machine learning techniques offer a promising alternative to traditional mechanics-driven models, particularly in the context of big data and the unique challenges posed by the complex mechanical behavior of RC columns in buildings.