Predicting the compressive strength of concrete by spectral bandwidth analysis of nondestructive acoustic measurements

Abstract

In the summer of 2001, a set of 85 soundwaves was produced from a steel ball hitting standard 4-inch concrete cylinders from three locations. The soundwaves were recorded using the software package Cool Edit 2000. These cylinders were then tested with a hydraulic press according to the standard ASTM testing procedures and the compressive strength in pounds per square inch was recorded for each cylinder. This study ran the soundwaves through a set of nine Fast Fourier Transform Filters ranging from 0-110 Hertz to 14080-28160 Hertz labeled A through J (excluding the letter I). A linear regression model was formed for each bandwidth and a combination of all bandwidths with and without location as an independent variable. The study found that the combined bandwidth models had the greatest predictability of compressive strength. The combined bandwidth linear regression model including location was found to be significant at less than 0.001 and had an adjusted R² of 0.931. The combined bandwidth linear regression model including location was found to be significant at less than 0.001 and had an adjusted R² of 0.489.