A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Estimation of rubberized concrete frost resistance using machine learning techniques
https://orcid.org/0000-0002-5875-1486
https://orcid.org/0000-0002-6930-979X
Highlights The two indexes of frost resistance durability of rubberized concrete were analysed. Classical and ensemble models were used to predict the frost resistance durability. XGBoost ensemble learning algorithm outperformed all other models. LOF detection algorithm was used to process the dataset to reduce the model training bias. The feature importance analysis revealed that adding rubber particles improves the frost resistance of concrete.
Abstract Utilizing waste rubber in concrete has effectively reduced global environmental pollution and carbon emission. It is essential to accurately evaluate and predict the evolution of its frost resistance in a complex environment to promote its engineering application. The traditional experimental modeling method is difficult to overcome the complex nonlinearity and environmental sensitivity among concrete components. Therefore, this study employed three classical single ML models, including RG, ANN, and SVR, and three new ensemble learning models (RF, XGBoost, and Stacking) to predict the two indexes of frost resistance of rubberized concrete, i.e., RDEM and MLR. Before modeling, the LOF detection algorithm was used to process the initial dataset to reduce the training bias. In addition, the performance statistics of the above six ML models show that the XGBoost ensemble learning algorithm based on Boosting has the best prediction performance for the frost resistance of rubberized concrete, with more than 95 % of the predictions distributed within the ± 20 % confidence interval (R2 = 0.96, MAE = 0.15, RMSE = 0.27). According to the SHAP and the partial dependence analysis, adding an appropriate amount of rubber can improve the frost resistance of concrete. The recommended RC and RS ranges are 15 to 25 kg/m3 and 0.1 to 0.5 mm, respectively. These findings revealed that ML algorithms are an efficient and robust model for predicting rubberized concrete's frost resistance.
Estimation of rubberized concrete frost resistance using machine learning techniques
https://orcid.org/0000-0002-5875-1486
https://orcid.org/0000-0002-6930-979X
Highlights The two indexes of frost resistance durability of rubberized concrete were analysed. Classical and ensemble models were used to predict the frost resistance durability. XGBoost ensemble learning algorithm outperformed all other models. LOF detection algorithm was used to process the dataset to reduce the model training bias. The feature importance analysis revealed that adding rubber particles improves the frost resistance of concrete.
Abstract Utilizing waste rubber in concrete has effectively reduced global environmental pollution and carbon emission. It is essential to accurately evaluate and predict the evolution of its frost resistance in a complex environment to promote its engineering application. The traditional experimental modeling method is difficult to overcome the complex nonlinearity and environmental sensitivity among concrete components. Therefore, this study employed three classical single ML models, including RG, ANN, and SVR, and three new ensemble learning models (RF, XGBoost, and Stacking) to predict the two indexes of frost resistance of rubberized concrete, i.e., RDEM and MLR. Before modeling, the LOF detection algorithm was used to process the initial dataset to reduce the training bias. In addition, the performance statistics of the above six ML models show that the XGBoost ensemble learning algorithm based on Boosting has the best prediction performance for the frost resistance of rubberized concrete, with more than 95 % of the predictions distributed within the ± 20 % confidence interval (R2 = 0.96, MAE = 0.15, RMSE = 0.27). According to the SHAP and the partial dependence analysis, adding an appropriate amount of rubber can improve the frost resistance of concrete. The recommended RC and RS ranges are 15 to 25 kg/m3 and 0.1 to 0.5 mm, respectively. These findings revealed that ML algorithms are an efficient and robust model for predicting rubberized concrete's frost resistance.
Estimation of rubberized concrete frost resistance using machine learning techniques
https://orcid.org/0000-0002-5875-1486
https://orcid.org/0000-0002-6930-979X
Highlights The two indexes of frost resistance durability of rubberized concrete were analysed. Classical and ensemble models were used to predict the frost resistance durability. XGBoost ensemble learning algorithm outperformed all other models. LOF detection algorithm was used to process the dataset to reduce the model training bias. The feature importance analysis revealed that adding rubber particles improves the frost resistance of concrete.
Abstract Utilizing waste rubber in concrete has effectively reduced global environmental pollution and carbon emission. It is essential to accurately evaluate and predict the evolution of its frost resistance in a complex environment to promote its engineering application. The traditional experimental modeling method is difficult to overcome the complex nonlinearity and environmental sensitivity among concrete components. Therefore, this study employed three classical single ML models, including RG, ANN, and SVR, and three new ensemble learning models (RF, XGBoost, and Stacking) to predict the two indexes of frost resistance of rubberized concrete, i.e., RDEM and MLR. Before modeling, the LOF detection algorithm was used to process the initial dataset to reduce the training bias. In addition, the performance statistics of the above six ML models show that the XGBoost ensemble learning algorithm based on Boosting has the best prediction performance for the frost resistance of rubberized concrete, with more than 95 % of the predictions distributed within the ± 20 % confidence interval (R2 = 0.96, MAE = 0.15, RMSE = 0.27). According to the SHAP and the partial dependence analysis, adding an appropriate amount of rubber can improve the frost resistance of concrete. The recommended RC and RS ranges are 15 to 25 kg/m3 and 0.1 to 0.5 mm, respectively. These findings revealed that ML algorithms are an efficient and robust model for predicting rubberized concrete's frost resistance.
Estimation of rubberized concrete frost resistance using machine learning techniques
Gao, Xifeng (author) / Yang, Jian (author) / Zhu, Han (author) / Xu, Jie (author)
2023-02-16
Article (Journal)
Electronic Resource
English
Experimental Research on Resistance to Frost of Rubberized Concrete
| Tema Archive | 2012
Experimental Research on Resistance to Frost of Rubberized Concrete
| Trans Tech Publications | 2012
Experimental Research on Resistance to Frost of Rubberized Concrete
| British Library Conference Proceedings | 2012
Impact Resistance of Rubberized Self-Compacting Concrete
| Taylor & Francis Verlag | 2015
Wear resistance of rubberized machine parts
| British Library Online Contents | 1998