In recent decades, numerous innovative methods have been developed to improve the quality, speed, cost-effectiveness, and efficiency of structural damage detection. Among these, computer vision has emerged as a an auspicious approach, particularly due to recent advancements in machine learning. Many successful models have been developed to detect a wide range of structural damages such as cracks, spalling, corrosion, rusting, and bolt loosening. However, shear buckling damage has been almost entirely neglected in the literature. This type of damage commonly occurs in thin steel plates used in structural members such as steel plate shear walls, and its accurate detection and localization are essential for reliable post-event condition assessment, especially following seismic loading. This study investigates the application of computer vision for both detection and instance-level localization of shear buckling damage using an instance segmentation framework. A key challenge in this task is the limited availability of labeled real-world images. To address this issue, an innovative data augmentation strategy is proposed that combines synthetic images generated using finite element analysis (FEA) with visually enhanced synthetic images created using 3D modeling software. These synthetic datasets are then combined with real experimental images to form larger and more diverse training datasets. In total, five datasets were considered, including 208 real images, 343 synthetic images, and 551 combined real–synthetic images. All models were trained using the YOLO11 instance segmentation algorithm. Results demonstrate that a model trained solely on real images achieved strong segmentation performance, with a precision of 0.87 and a recall of 0.81. The best-performing model, trained using a combination of real and visually enhanced synthetic images, achieved a precision of 0.90 and a recall of 0.84, corresponding to an improvement of approximately 3% in mAP50 compared to the real-only model. These findings confirm that high-fidelity synthetic data can effectively mitigate data scarcity and significantly enhance shear buckling detection and localization performance.
Rebar corrosion critically affects the durability of concrete structures, necessitating accurate prediction of bond strength between the concrete and corroded reinforcement. This study presents a novel hybrid approach, combining Monte Carlo simulations for systematic selection of the optimal Levenberg–Marquardt-based Multi-Layer Perceptron (LM-MLP) architecture with Particle Swarm Optimization (PSO) for refining network weights and biases. Using 132 experimental data points, the optimized model achieved a maximum correlation coefficient (R) of 0.959, representing an improvement of up to 3.75%, and reduced the root-mean-square error (RMSE) by up to 21.42% compared to the conventional LM-MLP model. An empirical regression model is also developed for comparison, reaffirming the superior accuracy of the proposed approach. These results demonstrate the model’s robustness and effectiveness for rapid and reliable prediction of bond strength under varying corrosion conditions. This hybrid approach not only enhances the accuracy and stability of the model but also provides rapid and reliable predictions under varying corrosion conditions, outperforming classical methods.
In this study, the structural behavior of coupled shear panels with Easy-Yielding steel plates is thoroughly investigated. These advanced systems are designed to enhance energy dissipation while ensuring a controlled and predictable yielding mechanism in seismic-resistant structures. To this end, after introducing the design assumptions, three one-story, prototype-scale numerical models with Shear, Intermediate and Flexural link beam behaviors are designed using the Plate-Frame Interaction method, and the corresponding analytical formulations are derived. The selected configurations effectively capture diverse force transfer and deformation mechanisms typically encountered in practical engineering structures. Subsequently, the specimens are modeled in Abaqus software and subjected to nonlinear static analysis to verify the yielding sequence of structural members and evaluate the overall nonlinear response. Pushover curves are extracted afterwards and systematically analyzed to assess strength capacity and overall deformation characteristics. The results demonstrate that the proposed analytical formulations are fully consistent with the numerical simulations performed in Abaqus, as both approaches predict the same yielding sequence of structural members, which is a primary objective of the design.
The detection of delamination is important for structural engineering applications for assessing the reliability of fiber-reinforced polymer (FRP) laminates, which could provide appreciable effects on the behavior of composite structures. Numerical investigation of interlaminar shear fracture (Mode II) was studied in the present work by the end-notched flexure (ENF) testing configuration. 20 glass/epoxy composite beams were analyzed with two different stacking sequences, incorporating four distinct stacking sequences with different fiber orientation angles and five initial crack lengths (a = 35, 40, 45, 50, and 55 mm). Finite element analysis was performed using ABAQUS, and the virtual crack closure technique (VCCT) was used to determine the distribution of the critical strain energy release rate with respect to the width of the laminate. The results obtained were then analyzed using a subsequent finite element analysis of interlaminar stresses in the area ahead of the crack tip. The results from the numerical analysis showed that the dominant contribution to the total energy release rate was the mode II (GII > 99%). Furthermore, the variation in GII across the laminate width was almost unchanged. For the other nodes, however, near the free edges, the separation of the node pairs, combined with the need to meet the fracture criterion, resulted in significant shear effects, with mode III components (GIII) becoming comparable to GII.
This study presents a full-scale before–after evaluation of upgrading the Nikan Hospital (Tehran, Iran) wastewater treatment plant from an Extended Aeration (EA) activated-sludge process to a Moving Bed Biofilm Reactor (MBBR) by retrofitting the existing aeration tank with Kaldnes K3 carriers (40% fill) and introducing upstream screening/equalization and an anoxic zone. Over a six-month monitoring program (3 months under EA and 3 months under MBBR; composite sampling three times per week), conventional performance indicators (COD, BOD₅, NH₄⁺-N, TN, TP, and TSS) were measured and compared statistically. The results showed that the retrofit produced higher and more stable removals after start-up stabilization: COD and BOD₅ > 90%, NH₄⁺-N 70–85%, TN 65–75%, and TP 50–60%, with effluent TSS of 25–30 mg/L. Excess sludge yield decreased by 75% (0.50–0.60 to 0.12–0.15 kg VSS kg⁻¹ COD removed), lowering sludge handling frequency and contributing to an overall OPEX reduction of 35–40%. Specific aeration energy increased modestly (from 0.74 to 0.83 kWh m⁻³) due to carrier-mixing requirements, but energy intensity per kg COD removed decreased slightly because of improved removal. A first-order CSTR kinetic model was calculated from the full-scale data and was applied to estimate apparent rate constants and the HRT required to meet target effluent concentrations, supporting the observed capacity gain (effective HRT from 24 h to 9 h) within the same footprint.
This study uses advanced imaging techniques and deep learning algorithms to assess fatigue cracks through cyclic loading on asphalt specimens. Faster Region-Based Convolutional Neural Network (Faster R-CNN) and the You Only Look Once (YOLO) models were compared to detect fatigue cracks in Ground-Penetrating Radar (GPR) and Computed Tomography (CT) scan images and to detect concealed cracks in GPR field data. Crack detection was improved using transfer learning with pre-trained weights from the COCO dataset. Using the piecewise function model, the accumulative horizontal strain was accurately estimated. Based on the statistical analysis, the model's accuracy was verified, with no significant differences between experimental and predicted results. Moreover, a piecewise function applied to CT scan data resulted in a better understanding of fatigue behavior. The crack classification was improved after retraining pre-trained deep convolutional neural networks (PDCNNs). The YOLO models outperformed Faster-RCNN in terms of average precision. Models YOLOv7, YOLOv5s, and YOLOv8 performed well on the GPR dataset, while YOLOv5s, YOLOv5m, and YOLOv8 were the most effective models on the CT dataset.
In this study, a hybrid boundary-finite element approach is proposed for solving nonhomogeneous soil freezing models, including internal inclusions. In this regard, by dividing the nonhomogeneous model into homogeneous parts and applying the hybrid element approach to each part, the resultant equations are combined using the continuity and consistency conditions at the interface boundaries. After introducing the discretized form of the equations using hybrid element approach, the implementation is carried out in a computer code and verified by solving several classic examples. Finally, a parametric study is conducted in which a soil layer containing circular and square-shaped inclusions is modeled using the proposed method, and the effects of freezing pipes placement, inclusion geometry, and soil properties on the freezing growth performance are evaluated. The results showed that the freezing pipe configuration, combined with the inclusion geometry installed outside the inclusion, provided more uniform and optimal freezing results in comparison to other configurations make it suitable for stabilizing underground tunnel walls in a nonhomogeneous soil layer.
Determination of the ultimate pile bearing capacity is still one of the major concerns in geotechnical engineering due to the complex interaction between soil and structure. This study employs interpretable machine learning models to provide precise predictions of pile capacities while identifying the role of the key design variables. A detailed data set of 100 steel and concrete piles is evaluated by including eight important design variables: effective pile length, cross-sectional area, Flap number, drained cohesion, drained soil friction angle, effective unit weight of soil, pile–soil friction angle, and pile material. Prediction models for the pile capacities are established using the Random Forest, XGBoost, CatBoost, and Extra Trees algorithms, which are validated through a strict 5-fold cross-validation. The results show that the Extra Trees algorithm is the most stable and has the highest predictive capability, with a coefficient of determination (R2) of 0.95 ± 0.03 and RMSE of 1806 ± 999. Furthermore, SHapley Additive exPlanations (SHAP) analysis is performed to calculate the importance of the design parameters, indicating that effective pile length, cross-sectional area, and Flap number are the major contributing factors. This reveals that the proposed unique combination of Flap number with cutting-edge machine learning analysis is an accurate, clear, and viable process for pile capacities.
Electric vehicle technology is key to sustainable development, by reducing air pollution and dependence fossil fuel. However, adoption in developing markets remains slow. Shared electric vehicles help address high initial costs, improving accessibility. This study extends the Unified Theory of Acceptance and Use of Technology (UTAUT) by integrating perceived benefits, environmental concerns, and hedonic motivations. Perceived benefits play a central role, influenced by performance expectations and social impacts. A total of 303 data points were collected through an online survey and analyzed with structural equation modeling. The results indicated that all seven hypotheses were supported. The findings highlight that performance expectations and social influences significantly shape perceived benefits, while effort expectations, hedonic motivations, facilitating conditions, and environmental concerns drive adoption. Additionally, the study explored the heterogeneity of user's socio-economic variables, such as gender and age, in the acceptance of shared electric vehicles. These insights can inform policymakers and service providers to design effective interventions that encourage SEV use of in developing cities.
Groundwater level fluctuations and the lack of reliable methods for estimating them are major contributors to land subsidence. Data mining has increasingly applied artificial intelligence (AI) techniques in recent years to predict time series variations, including groundwater level changes. In this study, a temporal–spatial hybrid model was developed by integrating the Group Method of Data Handling (GMDH) with Empirical Bayesian Kriging (EBK) to predict monthly groundwater levels. The GMDH model was employed to extrapolate temporal variations one month ahead, while the EBK model interpolated spatial variations to generate regional groundwater level maps. The Silakhor Plain in Iran was chosen as the subject of the case study. The model was built using monthly data from 11 groundwater stations that were collected between 2003 and 2013. The hybrid model employed groundwater level observations and precipitation records as inputs. Results indicated that the GMDH-EBK model provided reliable and accurate predictions, with strong correlations in both training and testing phases. The model achieved coefficients of determination of 0.95, 0.91, 0.85, and 0.79 for the Hamyaneh, Chaghadon, Sugar Factory, and Valyan wells, respectively. Overall, the proposed methodology represents a significant advancement in regional groundwater modelling and offers a promising approach to supporting sustainable water resource management.
Early Warning (EW) is commonly understood as the forecasting of a potentially catastrophic event. The effective use of data for EW plays a vital role in failure management, encompassing tasks like vulnerability mapping, forecasting, warning systems, prevention, planning, and the execution of actions. This study introduces an Early Warning Protocol (EWP) to tackle the potential failure of the Shiadeh Earth Dam in Iran. The dam was commissioned in 1999 and, in 2004, experienced a landslide at the downstream left abutment and the downstream slope. All installed monitoring instruments failed, meaning no deformation data has been recorded. Consequently, the Radial Basis Function (RBF) algorithm was employed to predict the dam's settlement. Satellite imagery was utilized for monitoring and controlling the dam's settlement, and a 3D PLAXIS model was used to review and evaluate the dam's performance under various reservoir conditions during different seasons, including static and dynamic states. In this study, several comprehensive phases were conducted to ensure the reliability and validity of the proposed protocol. According to the analysis results, the action plan was recognized as the central element of the EWP. Furthermore, the study highlighted that the steps of observing, planning, decision-making, and acting are essential prerequisites for the proposed protocol. By integrating satellite monitoring, numerical modeling, radial basis function algorithms, and geodetic surveying of micro-geodetic points and the dam crest alignment, this study presents an innovative and reliable protocol for monitoring the Shiadah earth dam. This approach can play a significant role in enhancing safety and reducing the risk of dam failure. Moreover, the method is applicable to all earth and concrete dams across the country. The assessment results indicate that the dam is currently at alert level 2 (caution status).
The dynamic modulus |E*| and phase angle f are the most important properties for the viscoelastic characterization of asphalt mixtures. Their experimental determination requires time-consuming procedures and expensive laboratory equipment. Hence, different prediction procedures have been developed for the estimation of these rheological properties, with Witczak and Hirsch models being the most widely accepted. Nowadays machine learning (ML) techniques are applied to various engineering problems because of their abilities in data processing, optimization and estimation. This paper proposes the K-Nearest Neighbors algorithm as an ML method for the prediction of the viscoelastic properties of asphalt mixtures. The training and validation of the algorithm was based on a database containing the bitumen characteristics, volumetric properties and dynamic modulus and phase angle values at different frequencies and temperatures with more than 5500 data points. The obtained results indicate that the ML algorithm developed in this study is accurate and it could be an effective approach to predict the viscoelastic properties of asphalt mixtures.
The cyclic stress-strain behavior of FRP-confined concrete columns is a key aspect of their performance, particularly in the context of seismic retrofitting and design. Significant research has been conducted in recent years to study the cyclic response of FRP-confined concrete in both circular and rectangular columns subjected to unloading and reloading cycles. These studies have led to the development of various cyclic stress-strain models to predict the behavior of FRP-confined concrete columns under such loading conditions.However, critical gaps remain in the modeling of early-stage unloading strains, reloading slope reduction, and cross-section-dependent behavior. This paper begins by introducing the fundamental components of cyclic stress-strain behavior. It then reviews and evaluates the existing models proposed for each aspect of this behavior, focusing on comparing their predictions with experimental data. Furthermore, modifications are suggested for components of the cyclic models that exhibit discrepancies when validated against experimental observations. These include a proposed reloading slope factor and refined unloading path formulations. These improvements aim to enhance the accuracy and reliability of cyclic stress-strain models for FRP-confined concrete columns, contributing to their effective application in structural design and seismic resilience.
Moving least squares material point method (MLS-MPM) is a new method to allow for sharp separation of particles and to reduce the cell-crossing error in the material point method simulations. This paper uses MLS-MPM method to numerically simulate the trapdoor test and soil arching, a phenomenon which plays a vital role in the design of many geotechnical structures, such as pile-supported embankments, buried pipelines, tunnels, and trench excavations. Large deformation of soil is considered during the plastic phase of deformation. Non-associated Mohr-Coulomb plasticity is enhanced to capture strain hardening/softening typically seen in granular soils. The proposed method is validated against centrifuge trapdoor tests. The numerical stress fields agree with the formation of triangular shaped arch after medium relative displacements and mobilization of a prism of soil after large relative displacements of the trapdoor. It is shown that using MLS-MPM, the arching behavior of soil can be described in detail throughout the trapdoor test.
This paper presents a finite element investigation into the seismic performance and retrofitting of corner RC beam-column joints with an attached floor slab—a realistic detail often neglected in previous studies. Seven models, including code-compliant, non-ductile, and five advanced retrofitting configurations (external steel cages, bolted steel boxes, externally bonded CFRP, and hybrid steel–FRP systems), were analyzed under monotonic loading using a calibrated damage plasticity approach. Explicit modeling of the slab–joint interaction enabled a realistic assessment of strength, ductility, and damage control. Results indicate that hybrid retrofitting strategies, particularly CFRP-BSB and BSAC, significantly enhance both ductility and damage mitigation, matching or surpassing the performance of code-compliant specimens. CFRP-only and CFRP-NSM systems also provided substantial gains in energy dissipation and control of inelastic damage, while the BSB method, despite increased strength, showed greater localization of plastic strains. Analysis of the principal plastic strain index and force–displacement curves highlights the value of integrated composite and steel confinement in restoring seismic resilience to vulnerable joints. These findings provide practical recommendations for the seismic upgrade of existing RC beam-column joint with floor slabs, emphasizing the importance of considering slab effects, advanced hybrid retrofitting, and robust damage indices in retrofit design and assessment.
In this study, the effect of unsaturated conditions and curing time before exposure to an unsaturated environment on the properties of one-part alkali-activated slag concrete (O-AAS) and Ordinary Portland cement concrete (OPCC) was investigated. The samples were cured in water saturated with alkaline materials for 7, 28, 56, and 360 days. However, the specimens cured for 7, 28, and 56 days were stored in an unsaturated environment with 50% relative humidity until they reached 360 days of age. Compressive strength and rapid chloride migration tests (RCMT) were conducted at various ages up to 360 days, and corrosion initiation time was estimated by solving Fick's second law using the finite difference approach and results from RCMT. Results show that reducing the curing time before unsaturated exposure significantly decreased O-AAS's compressive strength and increased its chloride ion diffusion coefficient more significantly than OPCC. Chloride penetration modeling indicates that exposure to an unsaturated environment has a more pronounced effect on reducing the corrosion initiation time of O-AAS compared to OPCC. Also, increasing the curing time from 7 days to 56 days before exposure to an unsaturated environment caused a 186.7% increase in the estimated corrosion initiation time of O-AAS and a 20.4% increase in the estimated corrosion initiation time of OPCC. The required curing time for the O-AAS mixture to eliminate the effect of an unsaturated environment with a relative humidity of 50% on the compressive strength and chloride ion diffusion coefficient on standard specimens is 56 days, and for the OPCC mixture is 28 days.
Dampers are employed to improve the cyclic behavior of structures against wind- and earthquake-induced loads. Rotational friction dampers (RFDs) dissipate energy through reciprocal rotation. The present research investigates the behavior of steel frames with concretefilled columns under three different frames of identical configurations for beams and columns but different positions for the bracing and RFD systems to see the effect of eliminating a part of the bracing member. Two loading schemes were considered to study the behavior of the introduced frames. In the first stage, the frames were subjected to vertical loading coupled with a cyclic load at the beam level. In the second stage, however, the cyclic load at beam level was replaced with earthquake acceleration at the structure base. All sample frames were thenmodeled and analyzed in finite-element software. Applying the bracing system coupled with the RFD reduced the base shear compared to the structure without the bracing and RFD systems, although the reduction was not the same for all frames with different bracing schemes. Moreover, compared to the frame without the RFD, the damper could limit the structural drift in 6 models, though it boosted the drift in 1 model. Results showed that RFDs can dissipate seismic energy through rotational friction slippage. The dampers could dissipate up to 90% of the energy induced by lateral loads to the structure. Also, removing a portion of the bracing member was found to be useful for architectural purposes.
Buildings consume approximately one-third of the world's energy, with the commercial and housing sectors' Heating, Ventilation, and Air Conditioning (HVAC) systems being the largest contributors to energy. Energy wastage is significant as a result of system faults, which indicates the importance of efficient control of energy in HVAC in saving energy as well as providing comfort to the occupants. Techniques in Artificial Intelligence (AI), such as Machine Learning (ML) and Deep Learning (DL), are now used to optimize HVAC energy efficiency as well as facilitate predictive maintenance, which reduces downtime as well as costs. Past research has underestimated qualitative faults analysis in HVAC systems or suffered from inaccurate identification using AI. This paper proposes an innovative AI-based framework to manage energy in buildings. The framework uses Fault Tree Analysis (FTA) initially to perform qualitative analysis regarding the effect of HVAC system faults in energy consumption. Next, it applies AI models, namely Long Short-Term Memory (LSTM) networks as well as Gated Recurrent Unit (GRU) networks, trained using experimental data from real-building environments. The models are designed to detect faults accurately as well as in time. The main goal is to save energy from wastage as well as ensure occupant comfort through timely maintenance as well as replacement of faulty equipment. Most notably, the GRU approach showed higher accuracy in the identification of faults compared to LSTM. The framework's accurate identification of the occurrence as well as the nature of the faults is an improvement in the efficiency of the building.
Active transportation systems, such as bike-sharing systems, offer several advantages, notably their integration with public transportation networks, pollution reduction, congestion alleviation, and decreased fuel consumption. However, a major challenge for shared bicycle companies is the efficient redistribution of bikes to ensure balanced availability across stations. Predicting station demand at various times is crucial for achieving this balance. To address this, we propose a framework leveraging data from Chicago's shared bicycle system and employing modern machine-learning techniques to forecast station demand throughout the day. In this research, we utilize features such as weather, accessibility level for each station, and historical transactions for the accurate prediction of traffic flow. Specifically, we compare two parallel multilayer perceptron deep learning models, incorporating matrix factorization and gate recurrent unit (GRU) neural networks. Furthermore, this research compares two performance models for predicting traffic flow. This research not only aids in optimizing bicycle distribution but also lays the groundwork for predicting demand in other public transportation systems, such as subways and buses, utilizing smart card technology.
In this study, the finite element modeling of the Y-shape bracing system is initially developed and verified. Next, the seismic performance of steel frames with Y-shape braces under far-field and near-field earthquakes was evaluated. For this purpose, seismic analysis was performed using 12 seismic records including near-field and far-field on 4, 8, and 12-floor structures equipped with and without Y-shape bracing. A series of time history analyses were performed on the samples, which led to the precise review of the Y-shape bracing system. The results show that structures equipped with Y-shape braces have less lateral displacement and also less number and level of plastic hinges; and, the level of non-linear dissipated energy is higher in those structures. Also, the plastic hinges based on the failure indices of the structures indicate that - near-field earthquakes have a more destructive effect of about 20% higher than far earthquakes on structures equipped with Y-shape bracing. Y-shaped bracing has reduced the permanent displacement in the system by up to 30%.
