Asian Journal of Civil Engineering最新文献

The affordable housing community in Jordan is finding it increasingly difficult to reconcile indoor thermal comfort and energy efficiency considerations, particularly within an arid climate. The existing models of thermal comfort, such as PMV/PPD, fail to represent dynamic interactions within the environment as well as interpersonal variations. The objectives of this work include creating a predictive model that uses machine learning and metaheuristic algorithms to predict and inform the design of the sustainable envelope of buildings. Using a set of 1336 hourly data patterns that represent actual housing envelope designs as well as environmental conditions found indoors, the Random Forest model of machine learning algorithm was used to prove the robust nature of the algorithmic approach and prove that it can create linear patterns. Five metaheuristic algorithms of PSO, GWO, Genetic Algorithm, BWO, and FA were used in combination with Random Forest for the process of feature selection and hyperparameter tuning. The algorithms were validated on the basis of R-Squared, Root Squared Error, and Mean Absolute Error on 10-fold cross-validation. The algorithm that worked best within Random Forest with PSO as the PSO algorithm contributed an overall value of R-Squared of 0.867 and an error of 0.321. The result of the research includes the design of an important model that helps influence the implementation of ML as part of an appropriate housing design within the context of climate change in that country. The findings highlight the potential of hybrid AI-driven tools to enhance energy efficiency and indoor comfort in low-income residential environments.

In this paper, experimental studies were carried out for enhancement the torsional behavior of reinforced concrete beams using steel fibers (SF), polyvinyl alcohol fibers (PVA) and carbon fibers (CF). Nine reinforced concrete beams were tested under pure torsion and loaded till failure. The first specimen was control which fabricated without any fibers. Steel fibers and PVA fibers were used separately with varying fiber ratio such as 1% and 1.5% by volume of concrete. Hybrid fibers of ( SF and PVA) and ( PVA and CF) were incorporated with varying fiber content of 0.5% and 0.75% in term of volume to cast the hybrid fiber reinforced concrete (HFRC). The results of torsional moment, angle of twist, crack patterns, ductility, energy absorption and torsional stiffness were presented in this paper. Experimental studies clarified that the use of different types of fiber such as SF, PVA and CF improved torsional moment and angle of twist. The achieved enhancements in the maximum torsional strength were 111%, 64%, 91% and 36% when a fiber of (1.5% SF), (1.5% PVA), (0.75% SF and 0.75% PVA) and (0.75% PVA and 0.75% CF) added, respectively. Non-linear finite element analysis (NLFEA) was performed using ANSYS to carry out a validation with the experimental studies. A good agreement was noticed from the comparison between the experimental results and the finite element predictions. Finally, this paper developed a proposed model to take the effect of fiber on the maximum torsional moment of hybrid fiber reinforced concrete (HFRC) beams. The ultimate torsion predictions of the proposed model were validated with 74 experimental test results in the current study and from previous studies in the literature. The comparison showed that the proposed model performs well in predicting the ultimate torsional moment of hybrid fiber reinforced concrete (HFRC) beams.

In the modern era of rapid infrastructure development, the depletion of natural resources and the rise in carbon dioxide (CO₂) emissions have emerged as pressing environmental concerns, driving global warming and ocean acidification. Carbon capture and storage (CCS), also referred to as carbon capture and isolation, has been recognized as an effective strategy to mitigate these challenges by capturing CO₂ emissions from major industrial sources such as cement plants and biomass power facilities. Traditionally, CCS involves the storage of CO₂ in underground geological formations; however, the long-term integration of CO₂ into building materials presents an innovative and sustainable pathway for reducing industrial emissions. In this context, sequester-based carbon mortar offers a promising alternative, enabling the dual benefit of material performance enhancement and carbon mitigation. The present study investigates the mechanical behaviour of sequester-based carbon mortar exposed to acidic conditions, with a focus on compressive, split tensile, and flexural strengths. To complement the experimental program, Artificial Neural Networks (ANNs) were employed to mix composition of the mortar as inputs. The ANN models achieved high correlation coefficients and low mean squared error values, demonstrating their effectiveness in mapping nonlinear relationships between mix composition and mechanical performance. This integration highlights ANN as a robust predictive tool for advancing sustainable construction materials.

The deterioration of civil infrastructure assets presents a serious global concern, affecting safety, functionality, and economic sustainability. Traditional statistical models, which often rely on linear assumptions and fixed deterioration rules, struggle to capture the complex patterns of asset degradation. This study conducts a comprehensive comparison of six machine learning algorithms i.e., multiple linear regression, decision tree regression, random forest regression, artificial neural network, extreme gradient boosting, and extra trees for predicting structural deterioration rates using real-world data from Indian bridges, roads, and pipelines. The dataset incorporates structural, environmental, operational, and maintenance-related variables. Models were rigorously trained using leakage-free cross-validation and evaluated using metrics such as coefficient of determination (R²), root mean squared error (RMSE), index of agreement (IOA), prediction interval (PI) coverage, and percentage of predictions within ± 20% of actual values (a20). Among all models, XGBoost demonstrated the highest predictive performance (R² = 0.87, RMSE = 0.93, PI coverage = 93.3%). Feature importance and interpretability were assessed using SHAP (SHapley Additive exPlanations), identifying age, chloride concentration, and traffic volume as the most influential predictors. The study provides a generalizable, interpretable, and uncertainty-aware framework for infrastructure asset management, offering practical guidance for data-driven maintenance planning and future extensions involving hybrid models and real-time sensor integration.

Efficient project scheduling requires balancing conflicting objectives such as time and cost. This study proposes a Modified Adaptive Weight Multi-Objective Mountain Gazelle Optimizer (MAWA-MGO) to address the Construction Time–Cost Trade-off (TCT) problem. The algorithm enhances the original MGO by using an adaptive weight adjustment strategy that dynamically balances exploration and exploitation, preventing premature convergence and improving Pareto-optimal solution quality. A benchmark 9-activity project was used to test the model, minimizing both project duration and total cost under precedence and resource constraints. Comparative results with non-dominated sorting GA (NSGA-II), multi-objective particle swarm optimization (MOPSO), and standard MGO show that MAWA-MGO achieves a more diverse and convergent Pareto front, a significant reduction in duration and cost, respectively. The findings confirm MAWA-MGO’s robustness and practicality as a decision-support tool for optimizing construction schedules. Future studies may extend it to include quality and environmental objectives for sustainable planning.

Structural damage detection (SDD) is essential for the safety and operational reliability of civil structures. This paper proposes an optimized machine learning (ML) technique for damage detection of the ASCE benchmark building, which is based on simulated structural data provided by an ANSYS numerical model. The building model is tested under several damage scenarios, and time-domain acceleration data are gathered under impact excitation. Relevant statistical features are retrieved from the simulation results and used as inputs for the model. The K-Nearest Neighbours (KNN) technique is used as the base classifier. Hyperparameter optimization is performed using particle swarm optimisation (PSO) and grid searching (GS) techniques. The results show that optimisation considerably enhances the technique’s damage classification accuracy, with the PSO-KNN technique achieving high accuracy and computational efficiency. Moreover, the results are compared by applying principal component analysis (PCA) by selecting important features. The results are also compared with traditional KNN, which yielded lower accuracy, thereby highlighting the necessity of employing optimization techniques. Furthermore, the outcomes are then compared with those obtained using the ANN technique. Additionally, the robustness of the technique is verified under a noisy dataset. The study indicates that combining ANSYS-based numerical modelling with optimized ML techniques creates a strong foundation for reliable structural state evaluations in SDD applications.

This study develops a multi-layer seismic risk assessment framework integrating fragility analysis, demand sensitivity evaluation, global variance-based sensitivity, and dependence modelling to investigate fixed-base (FB) and base-isolated (BI) reinforced concrete frames subjected to far-field (FF) and near-fault (NF) ground motions. Incremental dynamic analysis (IDA) was performed using 44 records across four GM categories, generating fragility curves for interstorey drift (MIDR), roof drift (MRDR), base shear (MBS), top-floor acceleration (MTFA), and Maximum Isolator Displacement (MID). Results confirm that isolation effectively reduces drift and acceleration demands, achieving 30–65% reductions compared to FB systems; however, NF pulse-type excitations impose large isolator displacements, with BD often approaching or exceeding design capacity. Scenario- and interval-based sensitivity analyses further identified BD and shear as dominant near-fault demand drivers. Sobol and Morris indices revealed that drifts and shear are primarily governed by spectral velocity content (PGV/PGA), whereas accelerations remain controlled by PGA. Copula dependence models demonstrated that MRDR co-escalates with MIDR in FB frames, while MID drives the joint escalation of roof drift and shear in BI systems. Accordingly, the proposed framework provides an integrated basis for identifying governing fragility parameters and their interactions, offering actionable insight for performance-based seismic design in near-fault environments.

Sustainable construction project management requires balancing competing objectives such as time, cost, and environmental impact under inherent uncertainties. This study proposes a novel fuzzy unified non-dominated sorting genetic algorithm III (Fuzzy U-NSGA-III) to optimize multi-mode project scheduling by addressing the time–cost–environmental impact (TCE) trade-offs. The model integrates fuzzy logic to handle uncertainty in activity durations, costs, and emissions using triangular fuzzy numbers (TFNs), enabling realistic representation of project variability. A mathematical framework is developed for defuzzification and integration into the multi-objective optimization model. The proposed Fuzzy U-NSGA-III is applied to a real-life construction project comprising 21 activities, each with five execution modes, to generate a diverse set of Pareto-optimal solutions. Performance is evaluated through visual trade-off plots, correlation analysis, and comprehensive benchmarking against existing multi-objective algorithms, including MOACO, MOTLBO, MODE, NSGA-III, and Fuzzy-MOPSO. Results show that the proposed method outperforms alternatives in convergence, diversity, and efficiency, achieving superior Hypervolume (HV), spacing (Sp), and non-uniformity of Pareto fronts (NPF), with significantly lower computational time. The model demonstrates strong adaptability, scalability, and practical utility for sustainable decision-making in complex construction environments, offering project managers a robust tool for optimizing conflicting project objectives under uncertainty.

Variation orders (VOs) are largely accountable for construction project dispute, schedule extension, and cost overrun, mostly in the Jordanian construction climate. This work introduces a combined predictive framework that combines Building Information Modeling (BIM), machine learning (ML), and metaheuristic optimization to anticipate and mitigate VO risks. Descriptive statistics investigation of nearby project data sets identified that VOs had caused an average cost increment of 11.4% and schedule extension of 14.1%. Gradient Boosting and LightGBM stood up with higher precision than other tested predictive versions, most effectively with optimization using Particle Swarm Optimization (PSO) and Black Widow Optimization (BWO). VO discriminators like BIM clashes, variation requests during design, and contractor experience were determined with feature importance investigation. The work stipulates a pragmatic, scalable framework for early VO risk detection enhancement and proactive choices. Methodological as well as pragmatic applications are found for its usability with contractors, consultants, and policymakers to mitigate VO effects and improve construction project execution returns. The framework introduced is favorable to digital construction evolution in construction risk management. Moreover, the framework also offers a transferable developmental profile for construction risk management in developing nations with comparable VO issues.

The importance of aerodynamic characteristics of tall structures in structural engineering, particularly in metropolitan areas where wind can be highly variable, is well understood. This research assessed the wind-induced responses of a G + 25 rectangular RCC high-rise building in two directions (0° and 90°) using codal analysis, numerical simulations, wind tunnel testing, and an integration of advanced AI-based modelling methods. The wind tunnel testing confirmed the consistency of the codal and numerical predictions through validation of the aerodynamic pressure distribution and load transfer mechanism in a 1:150 scaled model. The ETABS models allowed for the prediction of structural responses to varying loading conditions, and the codal wind loads were computed based on the standard of IS 875 (Part 3: 2015). An artificial neural network (ANN) and long short-term memory (LSTM) networks were created to predict displacements and drifts at each storey. The LSTM model showed more accurate successive height-wise fluctuations (R² = 0.9985) than the ANN. The SHAP study was conducted to determine the most important factors affecting the model to improve the interpretability of the model, which included wind speed, building height, and pressure coefficient. The increased wind resistance of the 90° design indicated that orientation significantly affects the performance of a structure. The combined approach fills the gap between the prescriptive rules of wind design and empirical verification of performance-based wind design, enabling the development of high-rise buildings in various Indian wind zones.