Characterization of Granular Systems by Digital Signal Processing of Low Strain Wave Response

The characterization and quality assessment of composite materials, particularly those constructed of residual materials such as slags, ashes, and crumb rubber, are difficult because of chemical and physical inhomogeneity and inconsistency at the time of their production. The characterization of granular systems, constructed of a mixture of geological and residual materials, may provide inconclusive information when tested by existing methods. The proposed nondestructive evaluation uses low-strain, low-frequency dynamic excitation as a means to better evaluate such materials. The applied excitation results in a sample response characterized by wave patterns. The wave pattern response is analyzed by digital signal processing and an artificial neural networks (ANN) system to facilitate characterization of the material. Dynamic excitation of representative samples was accomplished using a longitudinal-torsional resonant column. Nondestructive testing was conducted at low strain levels applying a torsional oscillatory motion. The resulting sample response wave forms were recorded. The shear modulus values obtained at the resonant frequency of each sample were used to train an ANN system to characterize sample wave responses measured at random frequencies. The recorded sample response wave forms were analyzed to identify the embedded dominant frequencies, which were unique signatures of the tested materials. These signatures were then submitted to the previously trained ANN to predict the material shear modulus. The samples tested were composed of dry Ottawa sand (0.85 to 0.6 mm), Ottawa sand and crumb rubber modifier, and soda-lime spheres compacted at various densities and tested under different confining pressures. Distinctive patterns, unique to the granular sample composition, were obtained. These are termed the signature patterns. A fast Fourier transform algorithm was used to convert collected data to the frequency domain. ANN analysis was applied to enhance pattern recognition and characterize the samples according to their shear moduli