Forecast flexural strength of pervious concrete by SVR

• Main Authors: Bin Ahmed, Farabi, Mitu, Sadia Mannan, Biswas, Rajib Kumar, Abid Ahsan, Khan, Mim, Sabrina Mostarin, Ahmed, Saif
• Format: Conference Proceeding
• Language: English
• Subjects: Compressive strength; Flexural strength; Kernel radial basis function regression analysis; Power regression analysis; Support vector machine
• ISSN: 2214-7853; 2214-7853
• DOI: 10.1016/j.matpr.2021.01.832

For the development of construction technology, pervious concrete is vastly using in road infrastructure. Pervious concrete has a strong drainage ability that can recharge groundwater, made by the coarse aggregate, water, and cement with little fine aggregate or without fine aggregate. The compressive and flexural strengths of pervious concrete are vital parameters to identify the concrete quality and also necessary in the process of analysis and design of a concrete infrastructure. This research represents a system for forecasting the flexural strength obtained from the compressive strength (0 up to 30 MPa) of concrete cube and cylinder types specimens by support vector regression (SVR) for both 7 and 28 days. Experimental data from various literature have been collected to formulate the model. It appears that the relationship between the compressive strength and flexural strength of pervious concrete is non-linear. The linear models developed with limited lab test found ineffective when they are subjected to wide range of data set for flexural strength prediction. In this study, all compared models and developing model’s effectiveness are examined based on R2, RMSE, MAPE, MSE, and IAE parameters for both 7- and 28-days specimens.