Estimation of Foamed Concrete Compressive Strength and Relationship Identification with Input Factors Using Support Vector Machine

Abstract

Foamed concrete (FC) is a building material with many advantages and is widely used in construction to reduce the load on the structure. Compressive strength is an important mechanical characteristic of concrete in general and of FC in particular. Therefore, it is important to quickly and accurately estimate this quantity. This study proposes the application of support vector machine (SVM) model to predict the FC compressive strength and construct a relationship with the input variables for the purpose of optimizing FC design process. A database of 220 test results is collected and used to build and verify the predictive performance of the proposed SVM model. The input factors of the problem are FC density, water/cement ratio, and sand/cement ratio. The results show that SVM is a good predictor of compressive strength of FC with performance evaluation criteria such as root mean square error, RMSE = 3.475 MPa, mean absolute error MAE = 2.816 MPa and coefficient of determination R2 = 0.930. Finally, 2-Dimensional Partial Dependence Plot (PDP) is developed to correlate the three input variables with the compressive strength of FC, which is useful for material engineers in the design phase of FC.