Assessment of Shear Strength and Failure Mechanisms in Exterior Reinforced Concrete Beam–Column Joints Using Machine Learning and Explainable Artificial Intelligence
Aug 2026· Buildings· Vol 16, pp. 3203· 0 citations· 53 references
TL;DR
The results indicate that the governing parameters for failure mechanisms differ from those controlling shear strength, and highlight the importance of simultaneously assessing shear strength and failure mode in RC beam–column joints.
Abstract
The seismic performance of reinforced concrete (RC) beam–column joints depends on both shear strength and failure mechanisms, the assessment of which remains challenging because of complex interactions among geometric, material, loading, and reinforcement parameters. This study presents a data-driven framework for assessing the shear strength and failure mechanisms of exterior RC beam–column joints. A database comprising 210 experimental specimens was systematically compiled from published studies. Seventeen input variables were selected based on structural mechanics, seismic design provisions, and previous experimental investigations. Machine learning models were developed for shear strength prediction and failure mode classification. SHAP was employed to interpret the trained models, while symbolic regression derived an interpretable design-oriented equation. On the independent test set, XGBoost achieved the highest shear strength prediction (R2 = 0.973, RMSE = 40.09 kN), whereas the Support Vector Machine achieved 80.5% classification accuracy. The results indicate that the governing parameters for failure mechanisms differ from those controlling shear strength. Joint shear capacity was primarily influenced by geometric dimensions and longitudinal reinforcement ratios, whereas axial load ratio and joint transverse reinforcement had a greater influence on failure mechanisms. These findings highlight the importance of simultaneously assessing shear strength and failure mode in RC beam–column joints.
Accurate prediction of the shear strength of reinforced concrete (RC) beams remains a challenging problem due to the complex nonlinear interactions among material properties, reinforcement characteristics, and beam geometry. This study presents a data-driven artificial neural network (ANN) framework for predicting the shear strength of RC beams using a systematically curated experimental database comprising 1,977 specimens collected from published literature. The database was preprocessed to remove incomplete and duplicate records, and the optimal ANN architecture was selected using the total goodness function. Model performance was evaluated using 10-fold cross-validation together with multiple statistical metrics, including the coefficient of determination (R²), mean absolute error, root mean squared error, bias, and prediction interval. The ANN achieved R² values of 0.9968 and 0.9686 for the representative training and testing datasets, respectively, and an overall R² of 0.993, with 1,668 predictions falling within the ± 30% error criterion. Comparative evaluation with the Canadian Standards Association (CSA), American Concrete Institute 318 (ACI 318), and Eurocode 2 (EC2) design-code models demonstrated that the proposed ANN consistently achieved superior predictive accuracy and reliability. Parametric analyses further confirmed that the predicted trends agree with established reinforced concrete shear mechanics, highlighting the dominant influence of the shear span-to-depth ratio, beam geometry, and reinforcement ratio on shear resistance. The proposed framework provides an accurate and robust decision-support tool that complements conventional design-code methods for predicting the shear strength of RC beams.
F. Khalid, Milad Razbin, S. Fareed et al.· Scientific Reports· 0 citations
The bond behavior between steel reinforcement and geopolymer concrete under cyclic loading is a critical factor for the performance of reinforced concrete structures in seismic regions. Traditional empirical models often fail to capture the complex, nonlinear interactions at the steel concrete interface, particularly under repeated loading conditions. This systematic literature review aims to synthesize and critically analyze existing research on machine learning-driven approaches for predicting and interpreting bond strength in this context. We systematically identified and evaluated studies that develop, validate, or apply machine learning algorithms including artificial neural networks, support vector machines, and ensemble methods to model bond-slip relationships, failure modes, and degradation mechanisms. The review methodology involved a structured search and thematic analysis of peer-reviewed articles, focusing on how these models incorporate key variables such as concrete compressive strength, fiber reinforcement type, confinement conditions, and loading history. Our analysis reveals that Several reviewed studies reported improved predictive performance of machine learning models compared with selected empirical equations. However, differences in datasets, validation strategies, and performance metrics limit direct comparison among studies and prevent definitive conclusions regarding consistent superiority and generalizability., achieving higher accuracy in predicting bond strength under both monotonic and cyclic regimes. Furthermore, we found that feature importance analyses from these models provide new insights into the relative influence of material properties for instance, the critical role of fiber volumetric ratio and lateral confinement in mitigating bond degradation under reversed cyclic loads. The review also identifies significant gaps, including the scarcity of experimental datasets for high-magnitude seismic loading and the limited generalizability of models across different geopolymer mix designs. We conclude that machine learning offers a powerful framework for advancing bond strength prediction in geopolymer concrete systems, but future work must prioritize the development of robust, transferable models trained on more diverse, large-scale cyclic test data. These findings provide a foundation for more reliable seismic design guidelines and inform the selection of machine learning strategies for structural performance assessment.
Qaim Shah, Waheed Ali Khoso, Mussa Umali· Journal of Infrastructure Pr...· 0 citations
This work presents a rational kinematics‐based model for assessing the shear capacity of rectangular reinforced concrete short columns. Due to their high stiffness and brittleness, short columns are particularly vulnerable to seismic forces, often exhibiting critical failures. The proposed model is based on three degrees of freedom and integrates equilibrium equations and constitutive relationships to accurately predict shear strength while maintaining simplicity and a clear physical basis. The model is validated with 47 test specimens and compared to the shear provisions of ASCE41‐17 and Eurocode 8‐3. The results show that the proposed approach reduces the scatter of code‐based strength predictions, providing a refined strategy for evaluating the shear resistance of critical short members. It is also shown that the model captures the effect of governing column properties by explicitly modeling the variation of shear‐resisting mechanisms associated with concrete and reinforcement.
D. Arcuri, L. Petrini, B. Mihaylov· Structural Concrete· 0 citations
Many existing reinforced-concrete (RC) frames do not have joints which are detailed according to modern seismic
detailing provisions. Consequently, the joint core of many existing frames is at risk of experiencing diagonal tension, bond
deterioration, bar-slip, loss of stiffness and brittle shear failure. An insightful synthesis of 50 studies from the years 2021 to 2026
involving the numerical investigation and retrofitting of RC beam-column joints, especially methods useful for existing
structures. An audit was informed by PRISMA on the corpus supplied: all 50 deduplicated records were screened; four complete
open-access papers were audited in full text, and 46 were evaluated through verified publisher, DOI, and abstract-level records.
The literature was divided into FRP, steel and mechanically anchored devices, advanced jackets, hybrid and precast connection
strategies, SHM, and latest finite-element/explained-machine-learning approaches. Non-linear finite element analysis was
applied across the direct evidence – most often based on concrete damage plasticity, non-linear reinforcement constitutive laws
and explicit interface or bond assumptions. This was deemed effective when calibrated against full-scale cyclic tests. The results
obtained through tests indicate that the performance of retrofit is less dependent on the nominal strength of the materials used,
but more so on the modification of the force-path, anchorage, interface behaviour, and confinement, and prevention of
premature joint-core shear. According to one numerical study, CFRP improves the ultimate-load by 103.93%, U-shaped bars by
43.55% and pre-tensioned steel-angle strengthening by 148.50%. Furthermore, a post-tensioned metal-strap programme claims
a core recasting gave up to 42%, and further strapping up to 25% increase in shear-capacity recovery. According to the account,
joint shear deformation was more than 40% less for the combined C-FRP rope-and-sheet strengthening. Further, the peak load
was 65% higher as well as 55% higher equivalent viscous damping. Modelling degraded bond, corrosion, bidirectional loading,
slab participation, field installation variability, long-term durability, and uncertainty-aware validation has important gaps. The
conclusion of the review points out that hybrid, mechanical anchoring and damage-controlled retrofits offer the most reliable
route once the design is supported by experimentally justified nonlinear models and performance-based acceptance criteria.
Yash D. Wani, U. Deshpande, Rais A. Chougale· International Journal for Re...· 0 citations
In this paper, an ANN-based model is suggested for the prediction of shear strength of steel fiber-reinforced concrete (SFRC) beams without transverse reinforcements, and the model is developed directly using Excel software to ensure computational transparency and statistical interpretability, while remaining an empirical data-driven tool. In this study, a comprehensive database including 923 experimental data is collected, prepared and utilized to develop the neural network. Following the completion of the learning process, the ultimate weight and bias for the neural network are obtained. These parameters are employed in the Excel worksheet. With this process, the learned model is converted into a standalone and operational design tool. Sensitivity analysis, which has been performed through a step-by-step and perturbation method, revealed that the beam width (b) is the most influential parameter in the determination of shear strength (33.4%), followed by concrete compressive strength (fc, 17.1%) and effective depth (d, 15.3%). Then, a parametric study is conducted to study the effect of the shear span-to-depth ratio (a/d), longitudinal reinforcement ratio (ρ) and fiber aspect ratio (Lf/df) on the shear behavior of the members.
Majid Al‐Gburi, Asaad Almssad, A. A. Alhayani· Buildings· 0 citations
This paper aims to investigate the effects of steel fiber position, inclination angle and length on the fracture behavior of steel fiber reinforced concrete (SFRC), while reducing the computational burden associated with high-fidelity phase-field fracture simulations.
A computational framework integrating phase-field modeling (PFM) and machine learning (ML) is developed. A strain-orthogonal phase-field formulation is employed to simulate interfacial damage and crack propagation in SFRC across different concrete grades and fiber configurations. Based on these simulations, a dataset of 357 samples is generated, comprising peak load (Pmax), critical displacement (U) and mechanical work (W). Gradient-boosting algorithms, including CatBoost, LightGBM and XGBoost, are trained and optimized using Bayesian optimization with five-fold cross-validation to construct efficient surrogate models.
The results show that CatBoost consistently provides the highest prediction accuracy, achieving test R2 values of 0.999 for peak load, 0.986 for critical displacement, and 0.942 for mechanical work. The developed ML surrogates enable near-instantaneous prediction of fracture responses, offering a substantial reduction in computational cost compared with standalone phase-field simulations. Model interpretation based on SHAP reveals that matrix stiffness and initial crack length dominate peak load, while fiber inclination plays a more significant role in post-peak mechanical work and energy dissipation.
This study proposes a hybrid phase-field–machine learning framework for efficient fracture analysis of SFRC. By exploiting high-fidelity numerical simulations as a data source for surrogate modeling, the proposed approach enables rapid parametric studies and optimization of fiber-reinforced concrete systems within a computational engineering context.
B. Vu, V. Hoang· Engineering computations· 0 citations
