Evaluating characteristics of particles’ surface micro-texture of granular materials based on the spectral analysis method

Accurately evaluating particle surface micro-texture levels and distribution at different wavelength scales of granular material is critical to optimise service stability of granular material under long-term dynamic loads. Firstly, generalised regression neural network (GRNN) and empirical mode decomposition (EMD) were adopted to impute missing data points and remove the arc-shaped tendency of granular materials' particle wear raw surface profile. Then, granular materials' particle surface micro-texture levels and their distribution within constant bandwidth narrow band spectrum and octave/fractional octave band spectrum were obtained with spectral analysis method, which could characterise granular materials' particle surface properties at different wavelength scales. Fourteen types of parent rocks of granular materials were tested with a modified micro tribological experiment simulating tribological behaviour among particle contact interfaces under dynamic loads. A contrastive analysis with traditionally used surface mean roughness was performed. The results indicate that the micro-texture levels and their distribution obtained in this study can detect the level of the exact texture scales (i.e. 32 and 2 mu m wavelengths) significantly influencing the kinetic friction coefficient of granular materials, while surface mean roughness can only represent the global surface property. A high correlation was found between normalised micro-texture level (i.e. 32 mu m) and Moh's hardness and coefficient of friction.