Interrelation on the attributes of recycled aggregate concrete and its strength prediction using gradient boosting
Shamili Syed Rizvon, Karthikeyan Jayakumar
Prediction of concrete strength is essential for proportioning new mixtures and quality assurance of the concrete produced. Researches have been made to predict the above, using statistical approaches like Linear Regression, Machine Learning approaches like Artificial Neural Network. The goal of this paper is to compare the aforementioned approaches with Gradient Boosting algorithms, in this case, extreme gradient boosting (XGB). The data for analysis and experimentation was taken from 7, 28, and 56 days of curing period across several parameters with the readings collected in the laboratory under standard controlled conditions. The resultant data was fed in algorithms developed in Python for predicting the compressive strength of recycled aggregate concrete. Also to the relationship for different cement content (300kg/m3, 350kg/m3, 400kg/m3, 450kg/m3) with varying w/c and different curing ages were developed. From the observations made, 400kg/m3 of cement content achieves 40.16MPa which is 17.1%, 24.5% and 17.2% greater than 300kg/m3, 350kg/m3 and 450kg/m3. In the case of the statistical approach, the XGB algorithm helps us attain more accurate predictions which are 9.5% and 6.4% greater than Linear Regression and Artificial Neural Networks respectively.
Shamili Syed Rizvon, Karthikeyan Jayakumar. Interrelation on the attributes of recycled aggregate concrete and its strength prediction using gradient boosting. International Journal of Advanced Engineering and Technology, Volume 5, Issue 2, 2021, Pages 16-22