Neural network approach to predict the compressive strength of concrete utilizing ultrasonic pulse velocity & digital images

Abstract

Concrete compressive strength is the most important property which signifies the quality of concrete. Several nondestructive and semi destructive test can be conducted to evaluate the concrete compressive strength, but there is an issue regarding the direct corelations between compressive strength and different non destructive test results. However, in the present study a image based concrete compressive strength prediction model using machine learning techniques with the help of ultrasonic pulse velocity (UPV) test has been proposed. In the present investigation 3 different concrete mix has been prepared of grade M20, M25 and M30 respectively. Several images at different zoom have been captured using digital microscope after cutting the concrete sample. In addition to that, all the sample have been tested for UPV values followed by destructive compressive strength. The images and corresponding UPV data and compressive strength have been used to predict the compressive strength from the image using the above mentioned methodology. The study clearly reveals that models exhibits better prediction model for estimating compressive strength using the digital microscopic images. The findings from the present investigation corroborate that UPV DATA can be used efficiently to predict cement mortar and concrete compressive strength. Thus, present study demonstrate the applicability of different machine learning technique using UPV values and digital microscopic images as an alternative nondestructive/semi destructive test method for predicting compressive strength of concrete.