Predicting Dynamic Modulus of Asphalt Mixtures Based on Sparrow Search Algorithm Optimized Light Gradient Boosting Machine DOI
Ke Zhang, Zhaohui Min, Xiatong Hao

и другие.

Опубликована: Янв. 1, 2023

The dynamic modulus of asphalt mixture is a key parameter to evaluate its viscoelastic and fatigue performance. This can be determined by laboratory measurements or model forecasting. utilization prediction models offers an efficient alternative that avoid time-taking experiments. Therefore, it very important accurately predict the modulus. study aims propose with high accuracy, robustness interpretability considering hyper-parameter optimization. A new hybrid developed combining Sparrow Search Algorithm (SSA) Light Gradient Boosting Machine (LightGBM). input variables are evaluated using Pearson Correlation Coefficient (PCC). accuracy accessed. And relative significance analysis conducted measure effect parameters on prediction. research findings indicate SSA-LightGBM has best precision in compared previous regression machine learning models. binder type for complex shear found most critical feature predicting modulus, followed test temperature, viscosity, performance grade (PG) at low temperature.

Язык: Английский

Metaheuristic optimization based- ensemble learners for the carbonation assessment of recycled aggregate concrete DOI Creative Commons
Emadaldin Mohammadi Golafshani, Ali Behnood, Taehwan Kim

и другие.

Applied Soft Computing, Год журнала: 2024, Номер 159, С. 111661 - 111661

Опубликована: Апрель 23, 2024

This study addresses the enhanced prevalence of carbonation, a process accelerating steel reinforcement corrosion, in recycled aggregate concrete (RAC) compared to natural concrete. Traditional carbonation depth assessment methods RAC are noted for being labor-intensive, costly, and requiring specialized expertise. There's deficiency application machine learning techniques accurately predicting RAC, gap this aims fill. Utilizing extreme gradient boosting (XGBoost) technique, recognized its efficacy ensemble learning, innovates modeling RAC. It emphasizes criticality hyperparameter optimization XGBoost algorithm maximizing model accuracy. To achieve this, three novel metaheuristic algorithms, including reptile search (RSA), Aquila optimizer (AO), arithmetic (AOA), were introduced as global optimizers tunning hyperparameters. The was underpinned by comprehensive database compiled from extensive literature, facilitating development an accurate model. Through rigorous evaluations, sensitivity analyses, Wilcoxon signed-rank test, runtime comparisons, synthesized models demonstrated exceptional accuracy, with coefficients determination exceeding 0.95. XGBoost-AO algorithm, particular, showcased superior performance, XGBoost-RSA providing efficient predictions considering runtime. SHapley Additive exPlanations (SHAP) interpretation highlighted environmental conditions significant influencers. A user-friendly graphical user interface developed, enhancing practical utility findings progression over time. research significantly advances predictive accuracy contributing sustainable management infrastructures emphasizing integration advanced structural engineering advancements.

Язык: Английский

Процитировано

17
Predicting carbonation depth of concrete using a hybrid ensemble model DOI

Zehui Huo,

Ling Wang, Yimiao Huang

и другие.

Journal of Building Engineering, Год журнала: 2023, Номер 76, С. 107320 - 107320

Опубликована: Июль 12, 2023

Язык: Английский

Процитировано

28
A Review of Concrete Carbonation Depth Evaluation Models DOI Open Access
Xinhao Wang, Qiuwei Yang, Xi Peng

и другие.

Coatings, Год журнала: 2024, Номер 14(4), С. 386 - 386

Опубликована: Март 26, 2024

Carbonation is one of the critical issues affecting durability reinforced concrete. Evaluating depth concrete carbonation great significance for ensuring quality and safety construction projects. In recent years, various prediction algorithms have been developed evaluating depth. This article provides a detailed overview existing models According to data processing methods used in model, can be divided into mathematical curve machine learning models. The further following categories: artificial neural network decision tree support vector combined basic idea model directly establish relationship between age by using certain function curves. advantage that only small amount experimental needed fitting, which very convenient engineering applications. limitation it consider influence some factors on concrete, accuracy cannot guaranteed. predict many considered at same time. When there are sufficient data, trained give more accurate results than model. main defect needs lot as training samples, so not A future research direction may combine with evaluate accurately.

Язык: Английский

Процитировано

9
Prediction model for calculation of the limestone powder concrete carbonation depth DOI
Andrija Radović, Vedran Carević, Snežana Marinković

и другие.

Journal of Building Engineering, Год журнала: 2024, Номер 86, С. 108776 - 108776

Опубликована: Фев. 8, 2024

Язык: Английский

Процитировано

8
A systematic literature review of AI-based prediction methods for self-compacting, geopolymer, and other eco-friendly concrete types: Advancing sustainable concrete DOI

Tariq Ali,

Mohamed Hechmi El Ouni,

Muhammad Zeeshan Qureshi

и другие.

Construction and Building Materials, Год журнала: 2024, Номер 440, С. 137370 - 137370

Опубликована: Июль 16, 2024

Язык: Английский

Процитировано

8
A review on properties and multi-objective performance predictions of concrete based on machine learning models DOI

Bowen Ni,

Md Zillur Rahman, Shuaicheng Guo

и другие.

Materials Today Communications, Год журнала: 2025, Номер unknown, С. 112017 - 112017

Опубликована: Фев. 1, 2025

Язык: Английский

Процитировано

1
Review of Prediction Models for Chloride Ion Concentration in Concrete Structures DOI Creative Commons

Jiwei Ma,

Qiuwei Yang, Xinhao Wang

и другие.

Buildings, Год журнала: 2025, Номер 15(1), С. 149 - 149

Опубликована: Янв. 6, 2025

Chloride ion concentration significantly impacts the durability of reinforced concrete, particularly regarding corrosion. Accurately assessing how this varies with age structures is crucial for ensuring their safety and longevity. Recently, several predictive models have emerged to analyze chloride over time, classified into empirical machine learning based on data processing techniques. Empirical directly relate concrete through specific functions. Their primary advantage lies in low requirements, making them convenient engineering use. However, these often fail account multiple influencing factors, which can limit accuracy. Conversely, handle various factors simultaneously, providing a more detailed understanding evolves. When adequately trained sufficient experimental data, generally offer superior prediction accuracy compared mathematical models. The downside that they necessitate larger dataset training, complicate practical application. Future research could focus combining models, leveraging respective strengths achieve precise evaluation relation structural age.

Язык: Английский

Процитировано

0
Enhance the accuracy, robustness and interpretability of asphalt mixture dynamic modulus prediction DOI
Ke Zhang, Zhaohui Min, Xiatong Hao

и другие.

International Journal of Pavement Engineering, Год журнала: 2025, Номер 26(1)

Опубликована: Фев. 7, 2025

Язык: Английский

Процитировано

0
Predictive Model for Erosion Rate of Concrete Under Wind Gravel Flow Based on K-Fold Cross-Validation Combined with Support Vector Machine DOI Creative Commons

Yanhua Zhao,

Kai Zhang, Aojun Guo

и другие.

Buildings, Год журнала: 2025, Номер 15(4), С. 614 - 614

Опубликована: Фев. 17, 2025

In the Gobi region, concrete structures frequently suffer erosion from wind gravel flow. This notably impairs their longevity. Therefore, creating a predictive model for flow-related damage is crucial to proactively address and manage this problem. Traditional theoretical models often fail predict rate of (CER) accurately. issue arises oversimplified assumptions failure account environmental variations complex nonlinear relationships between parameters. Consequently, single traditional inadequate predicting CER under flow conditions in region. To this, study utilized machine learning (ML) more precise prediction evaluation CER. The support vector (SVM) demonstrates superior performance, evidenced by its R2 value nearing one notable reduction RMSE 1.123 1.573 less than long short-term memory network (LSTM) BP neural (BPNN) models, respectively. Ensuring that training set comprises at least 80% total data volume SVM model’s accuracy. Moreover, time identified as most significant factor affecting An enhanced model, derived Bitter Oka framework integrating strength parameters, was formulated. It showed average relative errors 22% 31.6% however, recorded minimal error just −0.5%, markedly surpassing these improved terms Theoretical rely on simplifying assumptions, such linear homogeneous material properties. practice, factors like materials, flow, climate change are non-homogeneous. significantly limits applicability real-world environments. Ultimately, algorithm highly effective developing reliable safeguarding

Язык: Английский

Процитировано

0
Prediction of Ship CO2 Emissions and Fuel Consumption Using Voting-BRL Model DOI Open Access
Yi-Wei Lin, Chuanxu Wang

Sustainability, Год журнала: 2025, Номер 17(4), С. 1726 - 1726

Опубликована: Фев. 19, 2025

The accurate prediction of ship carbon dioxide (CO2) emissions and fuel consumption is critical for enhancing environmental sustainability in the maritime industry. This study introduces a novel ensemble learning approach, Voting-BRL model, which integrates Bayesian Ridge Regression Lasso to improve accuracy robustness. Utilizing four years real-world data from THETIS-MRV platform managed by European Maritime Safety Agency (EMSA), proposed model first employs Analysis Variance (ANOVA) feature selection, effectively reducing dimensionality mitigating noise interference. then combines strengths handling uncertainty correlations with Regression’s capability automatic selection through voting mechanism. Experimental results demonstrate that achieves an R2 0.9981 Root Mean Square Error (RMSE) 8.53, outperforming traditional machine models such as XGBRegressor, attains 0.97 RMSE 45.03. Additionally, ablation studies confirm approach significantly enhances predictive performance leveraging complementary individual models. not only provides superior but also exhibits enhanced generalization capabilities stability, making it reliable tool predicting CO2 consumption. advancement contributes more effective emission management operational efficiency shipping sector, supporting global efforts reduce greenhouse gas emissions.

Язык: Английский

Процитировано

0