New Machine Learning Models Revolutionize Concrete Strength Prediction and Boost Sustainability
The development of new machine learning models is revolutionizing the prediction of concrete strength and enhancing sustainability in the construction industry. Researchers have successfully employed artificial intelligence algorithms to accurately predict the splitting tensile strength (STS) of concrete containing Metakaolin, a more environmentally friendly material than traditional cement binders.
Concrete is the second most consumed material worldwide, and its production consumes a significant amount of energy while contributing to approximately 7% of global carbon dioxide emissions. As the demand for cement continues to rise, finding sustainable alternatives is crucial to reduce environmental impact. Metakaolin, a highly reactive pozzolan, has been identified as a potential solution due to its ability to reduce carbon dioxide emissions by up to 170 kg per ton of cement produced.
Previous studies have demonstrated that incorporating Metakaolin in concrete mixtures improves various mechanical properties, including compressive strength, durability, resistance to weathering and chemical erosion, and permeability. Estimating the mechanical properties of concrete, such as STS, based on mix proportions can significantly save time and costs in the construction process.
With the advancements in artificial intelligence, researchers have explored the application of machine learning models to predict the mechanical properties of concrete. In this study, four models, namely Artificial Neural Network (ANN), support vector regression (SVR), random forest (RF), and Gradient Boosting Decision Tree (GBDT), were utilized to predict the STS of concrete containing Metakaolin. The models were compared using evaluation metrics to determine the best-performing model.
The results of the study revealed that the GBDT model exhibited the highest test performance with an R2 value of 0.967, surpassing the performance of ANN, SVR, and RF models. The GBDT model also had the smallest error metrics, including mean squared error (MSE), root mean squared error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE). This accurate prediction model can serve as a valuable tool for estimating STS in concrete containing Metakaolin, potentially replacing or supplementing traditional laboratory compression tests and saving costs and time.
Furthermore, the study investigated the importance of input variables in predicting concrete strength. By identifying the significant factors affecting STS, researchers can further optimize concrete mixtures for enhanced mechanical properties.
The utilization of these machine learning models not only enhances the prediction of concrete strength but also promotes sustainability in the construction industry. By reducing the reliance on traditional cement binders and incorporating environmentally friendly materials like Metakaolin, carbon dioxide emissions can be significantly reduced. This innovative approach contributes to the global effort of minimizing the environmental impact of construction activities.
In conclusion, the application of machine learning models has revolutionized the prediction of concrete strength, particularly in mixtures containing Metakaolin. The results highlight the superiority of the GBDT model in accurately estimating STS, providing an efficient and cost-effective alternative to traditional laboratory tests. By embracing these intelligent models, the construction industry can improve efficiency, reduce costs, and promote sustainability in concrete production.