A nondestructive method of predicting dynamic properties is introduced to test granular systems composed of residual materials. A low-strain harmonic excitation at randomly selected frequencies produces propagating waves through the granular system. The system wave response is termed the “signature pattern” due to its uniqueness and repeatability. The signature pattern is used to characterize the material by predicting dynamic properties such as shear modulus. Artificial Neural Networks (ANN) are introduced to facilitate pattern recognition and improve characterization.

In this study, crumb rubber modifier (CRM) mixed with 20–30 Ottawa sand was tested under torsional excitation in a resonant column to obtain the material signature pattern. The actual shear modulus determined from the resonant frequency was used to train a neural network to characterize previously unseen signature patterns accordingly. At low strains, signature patterns have been found to be a valuable tool in pattern recognition and prediction of dynamic shear modulus of a granular system composed of residual materials.