Yazar "Tabar, Mehmet Emin" seçeneğine göre listele

Listeleniyor 1 - 3 / 3
  • Küçük Resim Yok
    Öğe
    AI-guided design framework for bond behavior of steel-concrete in steel reinforced concrete composites: From dataset cleaning to feature engineering
    (Elsevier, 2025) Katlav, Metin; Tabar, Mehmet Emin; Turk, Kazim
    This research focuses on establishing an artificial intelligence (AI)-guided design approach for predicting bond behavior of profiled steel-concrete in steel-reinforced concrete (SRC) composites. For that, an extensive literature survey was undertaken, and datasets for three main characteristic bond stresses-bond stress at initial slip (zs), ultimate bond stress (zu), and residual bond stress (zr) -were gathered. In total, it was gathered data points 150 for zs, 251 for zu, and 215 for zr. In addition, the isolation forest algorithm was used to detect and clean the anomalous data in the dataset, resulting in exhaustive and trustworthy data for training the models. As AI models, four popular machine learning algorithms like RF, XGBoost, LightGBM, and CatBoost are adopted. To improve the prediction performance of the models, three cases are established by Shapley additive explanations (SHAP)-based feature engineering. Additionally, SHAP and feature importance analyses were used to examine the impact of each feature on the bond behavior in SRC composites to ensure the explainability of the model. Meanwhile, to enhance the applicability of the study in real-world applications, a graphical user interface (GUI) was designed. According to the results, the CatBoost model proved its superior predictive ability specifically for zs and zr output values; in the test phase, the RMSE values were 0.07 and 0.05, R2 values were 0.904 and 0.934, MAPE were 10.02% and 8.33%, and MAE values were 0.04 and 0.03, respectively. On the other hand, the XGBoost model had the best predictive efficiency in the test phase for the zu output value with RMSE = 0.06, R2 = 0.833, MAPE = 8.32 % and MAE = 0.04. Lastly, based on SHAP and feature importance assessments, the most impactful features on bond behavior were identified as follows: the ratio of side cover to steel section height (cv/ hs), the compressive strength of concrete (fcu), and the ratio of bonded length to steel section height (lb/hs), stirrup ratio (psv), and the yield strength of profiled steel (fy). This knowledge can guide engineers in paying focus to specific features in their design and evaluation processes, resulting in more reliable and optimized outcomes.
  • Küçük Resim Yok
    Öğe
    An intelligent framework for compressive strength prediction of eco-friendly SFR-RCAC: Base and stacked ensemble models combined with experimental verification
    (Elsevier Sci Ltd, 2025) Katlav, Metin; Tabar, Mehmet Emin; Turk, Kazim
    This paper adopts an approach based on base and stacked ensemble models to correctly model the compressive strength (CS), which is a key parameter to provide structural integrity and reliability of steel fiber-reinforced recycled coarse aggregate concrete (SFR-RCAC). To this end, a reliable framework is adopted that includes cleaned 440 instances with 11 input features. Additionally, the impact of the input features on the model is investigated in detail via SHapley additive explanation (SHAP) and partial dependence plots (PDPs) analyses. To facilitate practical implementation, a graphical user interface (GUI) is designed to make the estimation process user-friendly and its reliability is verified by additional experimental tests. Based on the results, all the developed models are capable of predicting the CS of SFR-RCAC with extraordinary accuracy and reliability: the 6 base ensemble models achieved an average R2 = 0.936 and RMSE = 3.55 during the testing phase, while the 35 stacked models recorded R2 = 0.942 and RMSE = 3.38, respectively. Notably, the stacked ensemble model (SM26) with the combination of Extreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM) and Extra Trees Regressor (ETR) showed the best prediction performance in the test phase with the highest R2 (0.948) and the lowest RMSE (3.20) as well as the highest total score (287). Additionally, the error rate between the experimental values and the GUI predictions for the 10 designed mixes remains below +/- 8 %, verifying that the proposed GUI has high accuracy and robust generalization capability. Moreover, based on SHAP and PDP analyses, it is recommended for practical engineering applications to optimize the CS of SFR-RCAC by limiting the recycled coarse aggregate substitution ratio (Rr) to approximately 0.40, maintaining the steel fiber volume fraction (Vf) around 1.0 %, keeping the fiber factor (F) within the range of 0.6-0.8, and adjusting the water-tobinder (W/B) ratio between 0.30 and 0.40. To conclude, this research reveals the outstanding performance of the proposed models and GUI for predicting the CS value of SFR-RCAC and provides a significant contribution to the existing literature in this field. Thus, by promoting the efficient use of recycled coarse aggregates, it reduces the consumption of natural resources and allows the recycling of environmentally hazardous waste in the construction industry.
  • Küçük Resim Yok
    Öğe
    Explainable ensemble algorithms with grey wolf optimization for estimation of the tensile performance of polyethylene fiber-reinforced engineered cementitious composite
    (Elsevier, 2025) Tabar, Mehmet Emin; Katlav, Metin; Turk, Kazim
    This paper extensively examines the applicability of optimized ensemble machine learning (ML) algorithms via grey wolf optimization (GWO) to estimate the tensile performance of polyethylene fiber-reinforced engineered cementitious composites (PE-ECC). A robust and credible dataset is utilized for the establishment of the models based on available studies in the literature: The dataset includes 132 instances of PE-ECC mixes with 11 input features and 2 target output. Moreover, feature importance, Shapley additive explanation (SHAP) and partial dependence (PDP) analyses are implemented to enhance the explainability of the estimation models and to address the black box challenge of ML models. Based on the obtained results, the optimized extreme gradient boosting (XGBoost) and categorical boosting (CatBoost) models with GWO estimated the tensile performance of PE-ECC more effectively and accurately in comparison with other ensemble models. This has been extensively evaluated and proved through various approaches such as performance indicators, Taylor diagram, error analysis, and score analysis. To give a quantitative example, in the testing phase, for the prediction of tensile strain capacity, the GWO-XGBoost model reached the highest accuracy values with R2= 0.785 and RMSE= 1.077, whereas for the GWO-CatBoost model, these performance indicators were 0.764 and 1.129, respectively. In terms of tensile strength prediction, the GWO-XGBoost model achieved a high prediction accuracy with R2= 0.930 and RMSE= 1.004, while for the GWO-CatBoost model, R2 and RMSE were 0.932 and 0.987, respectively. Meanwhile, SHAP and PDP analyses were employed to identify the most influential features on output, and thus providing precious insight for designers to improve the tensile performance of PE-ECC. Additionally, a userfriendly graphical user interface (GUI) was constructed for estimating the tensile performance of PE-ECC and validated with new experimental datasets, illustrating the efficiency of the models. All in all, the importance of this work highlights the superior performance of the advanced GWO-ML models and GUI for estimating the tensile performance of PE-ECC and is thought to be a valuable contribution for further research in this area.