A neural network approach for compressive strength prediction in cement-based materials through the study of pressure-stimulated electrical signals

This paper presents a non-destructive method for predicting the compressive strength of cement-based materials by studying the appearance of weak electrical signals at specimens that are under mechanical stress. A series of lab experiments have been conducted in order to record the pressure-stimulated electrical signals in cement mortar specimens. Selected signal characteristics were correlated with the ultimate compressive strength of each specimen through the use of a neural network, employing a special training algorithm that offers increased predictive abilities. Results showed that the ultimate compressive strength can be successfully predicted without destroying the specimen.

Figure optionsDownload as PowerPoint slideHighlights
► Pressure-stimulated electrical signals were recorded in cement mortar specimens.
► Signal characteristics were correlated with specimen compressive strength.
► Radial Basis Function neural networks were used to produce the correlation.
► A special neural network training algorithm was employed for increased accuracy.
► Compressive strength was predicted successfully in a non-destructive manner.