Analysis of the Possibility of Non-Destructive Identification of the Interlayer Bond of Variably Thick Concrete Layers using Artificial Neural Networks

摘要

This paper presents the results of research and analysis of the possibility of nondestructive identification of the interlayer bond of variably thick concrete layers. In previous research the authors showed that it is possible to nondestructively identify the values of the pull-off adhesion of the top layer to the base layer by means of artificial neural networks on the basis of the base layer surface roughness parameters evaluated on the floor surface using three-dimensional optical laser scanning and the parameters evaluated by the acoustic impact echo and impulse-response techniques. However, if one considers the fact that the acoustic parameters determined by the acoustic techniques strongly depend on top layer thickness, the above method of assessment cannot be universally applied to floors differing in their top layer thickness. Since the concrete element which occurs in building practice differs in their top layer thicknesses, the aim of the research, presented in this paper, is to develop a way of identifying pull-off adhesion values by means of artificial intelligence on the basis of parameters independent of top layer thickness. The results of the training and testing of the selected artificial neural networks are presented in this paper. At the end the analysis, the possibility of non-destructive identification of the interlayer bond of variably thick concrete layers has been be presented. Successively the number of parameters included in the database used for the training and testing of artificial neural networks has been be reduced, but leaving each time parameter of the top surface layer thickness.