Asian Journal of Civil Engineering最新文献

Inclined columns are emerging as a fascinating innovation in the field of structural engineering and architecture. Their capacity to effectively transfer loads, particularly in applications like angled facades and bridges, has made them a desirable option in contemporary buildings. Inclined columns structural behaviour and moment resistance capabilities are crucial in modern engineering. Resolving this issue is essential in enhancing inclined column systems, functionality and safety. This study investigates three distinct inclined column configurations that involve a basic inclined column, an inclined column with triangular supports, and an inclined column with banding reinforcement. A comparison was made with the conventional vertical inclined column designs, with an emphasis on their efficiency, applicability for modern architectural and engineering applications, and performance. Utilizing the sophisticated finite element program Abaqus, an in-depth set of analyses was conducted to investigate the complex behaviours of load–displacement interactions, stress–strain correlations, and moment responses. The results indicate that inclined columns with banding reinforcement significantly outperform the others in terms of moment resistance, load-carrying capacity, and overall structural efficiency. In addition to FEM analysis, advanced predictive modelling techniques, including Artificial Neural Networks (ANN) and Response Surface Methodology (RSM), were employed to enhance further the understanding of the columns behaviour along with performance analysis. With an R² value of 0.997, the Artificial Neural Network (ANN) model has remarkable performance and exemplifies an outstanding fit. In contrast, the Response Surface Methodology (RSM) model demonstrates a slightly lower R² value of 0.985, which still indicates a solid fit. These models offer a more precise prediction of the structural performance by capturing the complex, non-linear interactions within the systems.

Choosing the right types and amounts of steel fibers and concrete mixtures is important for steel fiber reinforced concretes (SFRC) because they improve the flexural and compressive strength, making the concrete last longer. The variation in input factors, like the compositions of concrete mixes and steel fibers used, affects the unpredictable value of flexural strength. So, the study uses a set of experimental data to explore how steel fibers and different concrete mix compositions influence the flexural strength of SFRC. This study builds a prediction model for the flexural characteristics of SFRC using a database of 147 experimental results obtained from seventeen research groups. A Random Forest model creates an efficient prediction model and discovers key input features affecting flexural strength. The predictive flexural strength model is used along with Latin hypercube sampling (LHS) and a uniform probability distribution of input variables to create a large dataset. Then, the study used Sobol’s global sensitivity analysis method to investigate how different input factors affect the flexural strength of SFRC. The Sobol Index establishes and discusses the order in which input factors influence flexural strength. It is important information for selecting the optimal compositions of steel fiber-reinforced concrete.

Research has suggested substituting corrugated plates for flat plates to mitigate the adverse effects of buckling in these walls. However, corrugated plates typically result in a reduction in the maximum strength of the system. In this research, the characteristics of a single-layer hybrid system, which combines both flat and corrugated plates in the same plane, have been investigated in comparison to flat and fully corrugated SPSWs. This approach aims to reach the advantages of both plate types. The study employs the finite element method to investigate various configurations, with a focus on different corrugation locations, ratios, and angles. The modeling results indicate that the maximum strength of the hybrid system exceeds that of a fully corrugated wall and, in some cases, is comparable to the strength of a flat steel shear wall. Additionally, configurations with corrugation located in the middle of the hybrid system exhibit higher maximum strength than those with corrugation located elsewhere. The initial stiffness of the proposed hybrid system also exceeds that of traditional flat steel shear walls, with models featuring middle corrugation demonstrating even greater stiffness than fully corrugated shear walls. These findings underscore the efficacy of this new type of steel shear wall system.

Artificial intelligence has progressively revolutionized across different fields over the years, with particularly notable applications in dam engineering for monitoring and behavior prediction. Anomalies in these structures can lead to critical structural failures affecting stability and dam safety. For this reason, this review presents the contributions of artificial intelligence in dam monitoring for various applications, highlighting that many studies have been conducted on dams in China. A bibliometric study is included in this review to analyze the keywords network and the integration of artificial intelligence in civil engineering and dam monitoring by country, based on the last few years. This research focuses on machine learning models used to monitor deformations, cracks, seepage, and other anomalies encountered in structural health monitoring for dams. For training and testing these models, different methods and technologies are used for data collection, particularly sensors installed in dam structures, drone technology for image input, and the Building Information Modeling (BIM) method to facilitate relations among different participants and enable continuous monitoring of dam behavior. These models continue to develop, offering effective monitoring performance, although they require rich databases for optimal results. Additionally, artificial intelligence algorithms can be used for applications other than prediction and monitoring, such as data denoising and optimization. Multiple models can be combined to achieve comprehensive real-time dam monitoring with optimal results and high accuracy.

Graphical abstract

Sustainable retrofitting of infrastructure involves complex trade-offs between project duration, cost, and environmental impact. This study introduces a novel multi-objective optimization framework using the opposition-based learning multi-objective teaching–learning-based optimization (OBL-MOTLBO) algorithm. The proposed time-cost-environmental trade-off (TCET) model aims to minimize retrofitting time (RT), cost (RC), and carbon-equivalent emissions (REI) simultaneously. A case study conducted in the Delhi-NCR region spans eleven retrofitting domains, each with three intervention options. The OBL-MOTLBO algorithm outperforms established methods (e.g., NSGA-II, MOACO) in generating a diverse, converged Pareto front. To support decision-making, the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) is applied, allowing prioritization of solutions under different stakeholder preferences. Among 18 Pareto-optimal solutions, Solution 17 (RT = 98 days, RC = $295,000, REI = 320,000 kg CO₂-eq) ranks highest across all scenarios. The framework offers a robust, scalable method for sustainable retrofit planning, integrating economic and environmental objectives into data-driven decision-making.

As the world moves towards rapid urbanization, there arises a huge need for lightweight, high-strength, and low-cost prefabricated wall panels. Traditional cement-based systems show drawbacks with extremely high porosity along with limited early-age performance and poor microstructural control, especially with the incorporation of supplementary cementitious materials. Most optimization methods deal with strength only, without simultaneous control of perforation, microstructure, and practical constraints such as workability and cost. There is little understanding of microstructure-property relationships in terms of ternary blends modified with silica nanoparticles. The proposed work presents a complete, data-driven, multi-scale modeling framework for designing and optimizing cement-fly ash-lime wall panels augmented with silica nanoparticles. The hybrid machine learning-finite element surrogate modeling (ML-FEM-SM) approach combines the finite element simulation of microstructural stress and porosity evolution with machine learning regression to allow efficient prediction of compressive strength and pore distribution (R² ≈ 0.94, porosity error < 5). This is complemented by MD-MDFMBE where multimodal data fusion entails the integration of FTIR spectra, thermal curing images, and early mechanical data from transformer networks for non-destructive early prediction of strength and shrinkage with ± 1.5 MPa accuracy. Microstructure GAN production (µGAN) synthetic SEM images are of high fidelity for virtual mix validation (SSIM > 0.92). Constrained Multi-Objective Bayesian Optimization (MOBO-C) identified Pareto-optimal mixes under cost and flowability restrictions. Persistent Homology-Based Clustering (PHMC) is now classifying microstructural images into strength-correlated topological clusters (R² ≈ 0.89). The merged framework significantly improves the capabilities of mixing design for pre-casting quality control, deeper microstructure understanding, and performance-driven classification into advanced prefab materials.

Oil refineries present a high risk of accidents due to the use of flammable substances, as well as the operation of equipment at elevated temperatures and pressures. To address this issue, this study aims to analyze data on various potential scenarios for accidents at an oil refinery and develop an artificial neural network (NENKAZ ANN) to predict potential risk areas using real-time manufacturing data. This research employed the following techniques and tools: HAZID (hazard identification) method, hazard modeling software (ALOHA), Matlab R2022b, and the Levenberg-Marquardt algorithm, as well as MSE and R2 metrics. The testing of the developed ANN showed that it achieves high accuracy in predicting the distance traveled by the chemical explosion at oil refineries. The results of the prediction demonstrate that the coefficients of determination during training, validation, and testing are all 0.99. The average absolute error values for calculations using NENKAZ ANN are 190.42, 294.31, and 688.97, respectively. The forecasts generated in the study determined safe air zone dimensions in the period of atmospheric pollution resulting from an oil refinery accident. The developed ANN will be utilized at oil refineries in Kazakhstan, with the potential for implementation at other refineries worldwide. In the field of refinery management, NENKAZ ANN provides a means to evaluate the safe distance from the site of an explosion of a hazardous chemical substance, considering preliminary weather forecasts received from the meteorological service.

The incineration process is adopted when handling urban solid waste (USW). Due to this, ash is generated, known as urban solid waste ash (USWA). USWA can be utilized in bulk in the construction of geotechnical structures. The performance of USWA can be improved by adding admixtures such as cement and fiber. In this study, two different mathematical models were developed to predict the strength behavior, specifically the unconfined compressive strength (UCS) and split tensile strength (STS) of USWA mixed with cement and fiber, based on Artificial Neural Networks (ANNs). For this purpose, data from UCS and STS were used as dependent variables. USWA content, Fiber content (FC), Aspect ratio (AR) of fiber, cement content (CC), and curing period (CP) were considered independent variables. R² value during data validation was 0.9961 and 0.9985 for UCS and STS, respectively.

Accurate prediction of compressive strength is vital for ensuring the structural reliability and quality control of High-Strength Concrete (HSC). This study presents a data-driven modelling framework to predict the compressive strength of HSC incorporating varying proportions of limestone and natural aggregates, under different curing durations and ultimate loading conditions. Four tree-based machine learning models M5P, Reduced Error Pruning Tree (REP Tree), Random Tree (RT), and Random Forest (RF), were applied to a dataset comprising 123 experimental samples. The compressive strength served as the target output. Among the models, the ensemble-based Random Forest model achieved the highest prediction accuracy, with a training phase performance of CC = 0.9998, MAPE = 0.1161, RMSE = 0.2516, rRMSE = 0.23%, and NSEC = 0.9995, while testing metrics remained equally robust with CC = 0.9997, MAPE = 0.2881, RMSE = 0.3758, rRMSE = 0.38%, and NSEC = 0.9994. Sensitivity analysis using the Cosine Amplitude Method (CAM) revealed that ultimate load is the most influential input feature, with a sensitivity coefficient R_i=0.9999, indicating its dominant role in compressive strength development. Model performance was further substantiated through box plots, Taylor diagrams, and residual error visualizations. The findings support the use of Random Forest as a powerful tool for predicting the strength of HSC with blended aggregate systems, offering practical insights for performance-driven concrete design.

This study presents Eco-StrengthNet, a novel framework that unifies concrete mix–design and machine-learning ensemble tuning into a single triobjective optimisation. Leveraging a stacking ensemble of LightGBM, CatBoost, HistGradientBoosting, MLP, and SVR–with an ElasticNet meta-learner–Eco-StrengthNet integrates with NSGA-II to simultaneously maximize 28-day compressive strength and minimize embodied CO(_2) and energy consumption. On a 1 000-sample Sustainable Concrete Mixture dataset, the method achieved RMSE = 10.42 MPa, MAE = 7.36 MPa, and (R^2 = 0.9905) on hold-out test data, outperforming six classical and boosting baselines. The knee solution on the Pareto front yielded RMSE = 5.21 MPa with a 5 % improvement in normalized eco-penalty. A comprehensive explainability toolkit–including permutation importance, SHAP analysis, Sobol sensitivity indices, and 2D PDP surfaces–confirmed that curing age dominates strength prediction, while sustainability metrics exert secondary influence. Finally, a Streamlit GUI enables practitioners to explore mix–performance trade-offs in real time. This study demonstrates that integrated model–mix optimisation advances both performance and environmental sustainability in concrete design.