Seismic full waveform inversion is a common technique that is used in the investigation of subsurface geology. Its classic implementation involves forward modeling of seismic wavefield based on a certain type of wave equation, which reflects the physics nature of subsurface seismic wavefield propagation. However, obtaining a good inversion result using traditional seismic waveform inversion methods usually comes with a high computational cost. Recently, with the emerging popularity of deep learning techniques in various computer vision tasks, deep neural network (DNN) has demonstrated an impressive ability in dealing with complex nonlinear problems, including seismic velocity inversion. Now, extensive efforts have been made in developing a DNN architecture to tackle the problem of seismic velocity inversion, and promising results have been achieved. However, due to the dependence of a labeled dataset, i.e., the barely accessible true velocity model corresponding to real seismic data, the current supervised deep learning inversion framework may suffer from limitations on generalization. One possible solution to mitigate this issue is to impose the governing physics into this kind of purely data-driven method. Thus, following the procedures of traditional seismic full waveform inversion, we propose a seismic waveform inversion network, namely SWINet, based on wave-equation-based forward modeling network cells. By treating the single-shot observation data and its corresponding shot position as training data pairs, the inverted velocity model can be obtained as the trainable network parameters. Moreover, since the proposed seismic waveform inversion method is performed in a neural-network way, its implementation and inversion effect could benefit from some built-in tools in Pytorch, such as automatic differentiation, Adam optimizer and minibatch strategy, etc. Numerical examples indicate that the SWINet method may possess great potential in resulting a good velocity inversion effect with relatively fast convergence and lower computation cost.

Training a deep learning inversion network usually requires hundreds of thousands of complex velocity models, which is labor-intensive and expensive to acquire. In this work, we develop a new framework to automatically generate various velocity models with common geological structures, such as folding layers, faults and salt bodies. There are three main modules in the proposed framework. The first module generates a folded model with a given number of layers; the other two modules can add faults and salt bodies onto the folded model to form a fault or salt model, respectively. To best simulate the shape of subsurface geological structures while ensuring a good application effect in deep learning inversion, we generate the structural model based on composition of several basic functions with recommended parameter ranges. Then the generated structural model will be assigned with velocity values based on assumptions of underground seismic velocity distribution. To investigate the application effect of the generated 3D models, we conduct a deep learning inversion test. Since currently there is no 3D deep learning inversion algorithms available, the latest 2D inversion network called SeisInvNet is used to test the feasibility of the randomly generated velocity models. Through the experiments, we can see that the 2D inversion results are consistent with the true models from the aspects of velocity values and geological structure shapes, which demonstrates the rationality of the designed 3D models. In the end, we further discuss the feasibility of applying the proposed 3D model dataset to train a 3D inversion network. This work paves the way for the development of 3D deep learning inversion methods.

Severe water inrush currently halts the excavation of the Qinling Water Transportation Tunnel and has even stopped tunneling. Based on the geologic conditions of the construction area, the source of water inrush is abundant groundwater in fractured granite. To accurately predict water inrush, we must know the location, geometry, and boundary information of water-abundant zones before excavation. We have developed a field study of integrated seismic and transient electromagnetic (TEM) applications to map the water-abundant zones ahead of the tunnel face. Before the in-tunnel survey, numerical experiments based on excavation and drilling records verified the feasibility of our scheme. To reveal the fractured zones that may be filled with groundwater, an equitraveltime plane algorithm was used in the seismic data processing. Then, the TEM apparent resistivity was calculated to investigate the water-bearing condition of the fractured zones. To detect the water-abundant zone boundaries, we conducted migration imaging of the TEM pseudowavefield that was extracted from the measured TEM signals. A correlation stacking process to extract the TEM pseudowavefield was used to mitigate the ill-posed effect in the inverse wavefield transformation and obtain boundary information of the water-abundant zones. As expected, the results revealed the depth and geometry of the front and rear boundaries of two water-abundant zones 30 m ahead of the tunnel face, which is consistent with drilling and excavation records after the survey. The results of our case study indicate that the integration of seismic imaging, apparent resistivity imaging, and TEM pseudowavefield migration is an efficient and cost-effective method to provide the location, geometry, and boundary information of water-abundant zones.

Electrical Resistivity Inversion is an inverse problem that intends to infer the resistivity model from apparent resistivity data. The input and output have the format as the image but are with some unique characteristics that differ from the common natural images. The patterns in resistivity data are no longer spatial invariant. In this case, though the vanilla convolutional neural networks could build the mapping between images, their weight sharing kernels may cause ambiguous situations when coping with the resistivity data. To address this problem, in this work, we directly adapt convolutional kernels to fit for the electrical resistivity inversion task. By analyzing the resistivity data, we found the data patterns corresponding to the same abnormal body will vary with the vertical position change, correspondingly we learn extra offset and amplitude parameters for each vertical position the convolutional kernels applied, to make convolutional kernels transformed to capture the varying patterns. Unlike active convolution and deformable convolution, our learned offsets and amplitudes are sharing at each vertical position, which is neither too free nor too rigid. More importantly, it obeys the inherent characteristics of resistivity data. Finally, through adversarial training, the inversion results are more realistic as the resistivity model. From comprehensive comparisons and studies, our adaptive convolutional neural networks (ACN) outperform the baselines consistently, verifying our assumption and the effectiveness of our proposed method.