Compressive Strength Estimation of Concrete Blocks using Fully Connected Deep Neural Network

Compression strength refers to a structure ability to endure loads that cause it to shrink in size. The ultimate load in Newtones scaled by the gross cross section area of the block in square millimeters yields the bearing capacity of a concrete masonry unit. The machine learning technology could extend the interdisciplinary field for the estimation of concrete block compressive strength. With this review, this project aims to estimate the concrete block compressive strength retrieved from the KAGGLE machine-learning repository. The compressive concrete blocks dataset contains 9 features and 1050 concrete block and it is processed with encoding and missing values. The exploratory data analysis is done to analyze the correlation of all the features. The raw dataset is applied to linear regression, Linear Kernel SVR, Polynomial Kernel SVR, Radial Basis Function SVR, Sigmoid Kernel SVR, Decision Tree to analyze the performance of compressive strength estimation of concrete blocks. The proposed model uses Fully Connected Deep Neural Network (FCDNN) where the compressive concrete blocks dataset is applied to estimate the compressive strength estimation of concrete blocks. The performance metrics utilized for analysis are Mean Squared Error, R-Squared error. Experimental results shows that the Fully Connected Deep Neural Network shows Mean Squared Error of 29.5196 and R-Squared error of 0.885 when analyzed with the all regression models.