Title: Neural networks in the NDT identification of the strength of concrete

Author(s): K. Schabowicz

Publication: KILW

Volume: 51

Issue: 3

Appears on pages(s): 371-382

Keywords: non destructive testing, strength

DOI:

Date: 7/1/2005

Abstract:
This paper presents an application of artificial neural networks to concrete compression strength identification based on parameters determined by nondestructive methods. Three ordinary concretes and three high-performance concretes with a compression strength of 24-105 MPa were investigated. The parameters determined by nondestructive methods, i.e. the ultrasonic method, sclerometric methods and the pull-out method, and the age and bulk density of the concretes were used. A neural network with the Levenberg-Marquardt algorithm was chosen from several networks and successfully applied. The paper presents a methodology for the neural identification of the compression strength of concrete and some results of the investigation.