Empirical mode decomposition approach to simplify the fracture roughness for numerical models

The shear strength of natural, unfilled rock fractures is influenced by surface roughness. The surface curve of a fracture can be viewed as a waveform graph, and in general, it is of the characteristic that high-frequency represents the low amplitude (local variation) and low-frequency represents the high amplitude (general trend). In this work, the signal processing method, Empirical Mode Decomposition (EMD) was employed to decompose the original fracture surface scanned by photogrammetry to several frequency-dependent curves. Low-frequency curves were selected and composed as the element geometry while high-frequency curves were ignored and replaced by parameters related to the roughness in each surface element in Abaqus. The process of push-shear test is simulated using the simplified fracture curve, showing the geometry simplification by EMD can help model the shear failure of rock fractures.