The inverse problem of electrical resistivity surveys (ERSs) is difficult because its nonlinear and ill-posed nature. For this task, traditional linear inversion methods still face challenges such as suboptimal approximation initial model selection. Inspired by the remarkable mapping ability deep learning approaches, in article, we propose to build from apparent data (input) (output) directly convolutional neural networks (CNNs). However, vertically varying characteristic patterns may cause...

A DNN architecture referred to as GPRInvNet was proposed tackle the challenges of mapping ground-penetrating radar (GPR) B-Scan data complex permittivity maps subsurface structures. The consisted a trace-to-trace encoder and decoder. It specially designed take into account characteristics GPR inversion when faced with data, well addressing spatial alignment issues between time-series maps. displayed ability fuse features from several adjacent traces on enhance each trace, then further...

We propose a new method to tackle the mapping challenge from time-series data spatial image in field of seismic exploration, i.e., reconstructing velocity model directly by deep neural networks (DNNs). The conventional way addressing this ill-posed inversion problem is through iterative algorithms, which suffer poor nonlinear and strong nonuniqueness. Other attempts may either import human intervention errors or underuse data. for DNNs mainly lies weak correspondence, uncertain...

Velocity model inversion is one of the most important tasks in seismic exploration. Full-waveform (FWI) can obtain highest resolution traditional velocity methods, but it heavily depends on initial models and computationally expensive. In recent years, a large number deep-learning (DL)-based methods have been proposed. One critical component those DL-based training set containing different models. We developed method to construct realistic structural for DL network. Our compressional-wave...

Seismic full waveform inversion is a common technique that used in the investigation of subsurface geology. Its classic implementation involves forward modeling seismic wavefield based on certain type wave equation, which reflects physics nature propagation. However, obtaining good result using traditional methods usually comes with high computational cost. Recently, emerging popularity deep learning techniques various computer vision tasks, neural network (DNN) has demonstrated an...

Abstract Deep learning‐based methods have performed well in seismic waveform inversion tasks recent years, while the need for velocity models as labels has somewhat limited their application. Unsupervised learning allows us to train neural network without labels. When inverting from observed data, are often unavailable real data. To address this problem and improve generalization, we introduce a multi‐scale strategy enhance performance of unsupervised learning. The first ‘multi‐scale’ is...

Training a deep learning inversion network usually requires hundreds of thousands complex velocity models, which is labor-intensive and expensive to acquire. In this work, we develop new framework automatically generate various models with common geological structures, such as folding layers, faults salt bodies. There are three main modules in the proposed framework. The first module generates folded model given number layers; other two can add bodies onto form fault or model, respectively....

The high-resolution waveform inversion for seismic velocities is gaining increasing interest as we start to deal with complex structures. Although full (FWI) has been used several years, obtaining velocity models still presents many obstacles, such the high computational cost and limited bandwidth of data. Thus, propose a deep learning (DL)-based algorithm build using low-resolution models, migration images, well-log inputs. well information, specifically, helps enhance resolution...

Seismic velocity inversion plays a vital role in various applied seismology processes. A series of deep learning methods have been developed that rely purely on manually provided labels for supervision; however, their performances depend heavily using large training data sets with corresponding models. Because no physical laws are used the phase, it is usually challenging to generalize trained neural networks new domain. To mitigate these issues, we embedded seismic forward modeling step at...

Excavation under complex geological conditions requires effective and accurate forward-prospecting to detect the unfavorable structure estimate classification of surrounding rock in front tunnel face. In this work, a forward-prediction method for geology is developed based on seismic wave velocity layered tomography. particular, problem strong multi-solution inversion caused by few ray paths narrow space tunnel, regularization proposed. By reducing area each iteration step applying...

ABSTRACT Random noise attenuation utilizing predictive filtering achieves great performance in denoising seismic data. Conventional methods are based on fixed filter operators and neglect the complexity of structures. In this way, denoised data cannot meet requirement balancing signal preservation removal. study, we proposed a structural complexity‐guided method that utilizes an adapted operator to adjust changes complexity. The mainly consists two stages. A slope field information is...

With China's expanding economy, many tunnels are being designed and constructed. However, tunneling in hazardous geologic terrain, with faults, fractures, water-bearing openings, other adverse geological conditions, construction safety is seriously endangered. To ensure the of tunnel construction, a prospecting method was proposed applied at Xiangyun Tunnel Yunnan Province, China. In investigation stage, engineering hydrogeological conditions were analyzed to recognize high-risk sections....

The safety and efficiency of tunnel construction depend on the knowledge complex geologic conditions. Hence, an accurate forward-prospecting technique is required to detect unexpected inhomogeneous structures ahead construction. As method image heterogeneities, elastic reverse time migration (ERTM) introduced field forward prospecting. However, in environment, ERTM images may suffer from interference crosstalk artifacts, which are caused by converted waves surfaces. Therefore, considering...

Seismic forward-prospecting in tunnels is an important step to ensure excavation safety. Nowadays, most advanced imaging techniques seismic exploration involve calculating the solution of elastic wave equation a certain coordinate system. However, considering cylindrical geometry common tunnel body, Cartesian system seemingly has limited applicability forward-prospecting. To accurately simulate signal received tunnels, previous method using decoupled non-conversion extended from coordinates...

Abstract Seismic velocity plays an important role in imaging and identifying underground geology. Conventional seismic inversion methods, like full waveform inversion, directly update the model based on misfit between observed synthetic data. However, is a highly nonlinear process, effect greatly relies initial model. In this paper, we propose novel network‐domain method. Different from existing which use random or fixed numbers as network input, reparameterize low‐dimensional acoustic...

Seismic forward prospecting is essential because it can identify the velocity distribution in front of tunnel face and provide guidance for safe excavation activities. We have developed a convolutional neural network (CNN)-based method to invert forward-prospecting data recorded tunnels accurate rapid estimation seismic distribution. Targeting unusual acquisition setup tunnels, we design two separate encoders extract features from observation on both sidewalls. Subsequently, these are...

Accurate seismic imaging can ensure safe and efficient tunnel construction under complex geologic conditions. As a high-precision migration method, reverse time (RTM) has been introduced into forward prospecting. However, the resolution of traditional RTM results may not meet requirements in environment, which affects interpretation forward-prospecting results. We have developed least-squares method based on decoupled elastic wave equation tunnels. The Born forward-modeling operator its...