Reconstruction of fracture geometry in material medium by elastic wave

Abstract Over the past few years, intelligent detection and fast and precise positioning for fracture have been hotspots in the field of geological engineering. The propagation of elastic wave in fracture inclusive one-dimensional line segment was obtained through simulation, and then Densely Connected Convolutional Networks (DenseNets) were used to learn waveform. Moreover, the key features of the fracture were obtained automatically from elastic wave to achieve fast, precise, and intelligent detection for fractures in line segment. Furthermore, the Schoenberg linear slippage model was used to describe the contact behavior of closed fractures on two-dimensional plane. The propagation of elastic wave on fracture inclusive plane was obtained by simulation, and the plane is divided into several horizontal and vertical straight lines in which elastic wave was collected. The sampled data were analyzed by the trained neural network model, and geometrical reconstruction was implemented for fractures on the plane based on the detection results of each line and row. Finally, three-dimentional scanning laser Doppler vibrometry was used to experimentally obtain propagation of elastic wave on granite plane under incentive effect, and then experimental data were input into the neural network. This has accurately recovered the geometrical shape of fracture in granite and further verified the precision of the neural network.