High-Fidelity Permeability and Porosity Prediction Using Deep Learning With the Self-Attention Mechanism.

Accurate estimation of reservoir parameters (e.g., permeability and porosity) helps to understand the movement of underground fluids. However, reservoir parameters are usually expensive and time-consuming to obtain through petrophysical experiments of core samples, which makes a fast and reliable prediction method highly demanded. In this article, we propose a deep learning model that combines the 1-D convo- lutional layer and the bidirectional long short-term memory network to predict reservoir permeability and porosity. The mapping relationship between logging data and reservoir parameters is established by training a network with a combination of nonlinear and linear modules. Optimization algorithms, such as layer normalization, recurrent dropout, and early stopping, can help obtain a more accurate training model. Besides, the self-attention mechanism enables the network to better allocate weights to improve the prediction accuracy. The testing results of the well-trained network in blind wells of three different regions show that our proposed method is accurate and robust in the reservoir parameters prediction task.

Self-Attention Fully Convolutional DenseNets for Automatic Salt Segmentation.

3-D salt segmentation is important for many research topics spanning from exploration geophysics to structural geology. In seismic exploration, 3-D salt segmentation is directly related to the velocity modeling building that affects many processing steps, such as seismic migration and full waveform inversion. Manually picking the salt boundary becomes prohibitively time-consuming when the data size is too large. Here, we develop a highly generalized fully convolutional DenseNet for automatic salt segmentation. A squeeze-and-excitation network is used as a self-attention mechanism for guiding the proposed network to extract the most significant information related to the salt signals and discard the others. The proposed framework is a supervised technique and shows robust performance when applied to a new dataset using transfer learning and a small amount of training data. We test the robustness of the proposed framework on the Kaggle TGS salt segmentation dataset. To demonstrate the generalization ability of the framework, we further apply the trained model to an independent dataset synthesized from the 3-D SEAM model. We apply transfer learning to finely tune the trained model from the TGS dataset using only a small percentage of data from the 3-D SEAM dataset and obtain satisfactory results.

Urban DAS Data Processing and Its Preliminary Application to City Traffic Monitoring.

Distributed acoustic sensing (DAS) is an emerging technology for recording vibration signals via the optical fibers buried in subsurface conduits. Its relatively easy-to-deploy and high spatial and temporal sampling characteristics make DAS an appealing tool to record seismic wavefields at higher quantity and quality than traditional geophones. Considering that the usage of optical fibers in the urban environment has drawn relatively less attention aside from its functionality as a telecommunication cable, we examine its ability to record seismic signals and investigate its preliminary application in city traffic monitoring. To solve the problems that DAS signals are prone to a variety of environmental noise and are generally of weak amplitude compared to noise, we propose a fast workflow for real-time DAS data processing, which can enhance the detection of regular car signals and suppress the other components. We conduct a DAS experiment in Hangzhou, China, a typical metropolitan area that can provide us with a rich data library to validate our DAS data-processing workflow. The well-processed data enable us to extract their slope and coherency attributes that can provide an estimate of real traffic situations. The one-minute (with video validations) and 24 h statistics of these attributes show that the speed and volume of car flow are well correlated demonstrates the robustness of the proposed data processing workflow and great potential of DAS for city traffic monitoring with high precision and convenience. However, challenges also exist in view that all the attributes are statistically analyzed based on the behaviors of a large number of cars, which is meaningful but lacking in precision. Therefore, we suggest developing more quantitative processing and analyzing methods to provide precise information on individual cars in future works.

Obtaining free USArray data by multi-dimensional seismic reconstruction.

USArray, a pioneering project for the dense acquisition of earthquake data, provides a semi-uniform sampling of the seismic wavefield beneath its footprint and greatly advances the understanding of the structure and dynamics of Earth. Despite continuing efforts in improving the acquisition design, network irregularity still causes spatial sampling alias and incomplete, noisy data, which imposes major challenges in array-based data analysis and seismic imaging. Here we employ an iterative rank-reduction method to simultaneously reconstruct the missing traces and suppress noise, i.e., obtaining free USArray recordings as well as enhancing the existing data. This method exploits the spatial coherency of three-dimensional data and recovers the missing elements via the principal components of the incomplete data. We examine its merits using simulated and real teleseismic earthquake recordings. The reconstructed P wavefield enhances the spatial coherency and accuracy of tomographic travel time measurements, which demonstrates great potential to benefit seismic investigations based on array techniques.

Unveiling the signals from extremely noisy microseismic data for high-resolution hydraulic fracturing monitoring.

Microseismic method is an essential technique for monitoring the dynamic status of hydraulic fracturing during the development of unconventional reservoirs. However, one of the challenges in microseismic monitoring is that those seismic signals generated from micro seismicity have extremely low amplitude. We develop a methodology to unveil the signals that are smeared in the strong ambient noise and thus facilitate a more accurate arrival-time picking that will ultimately improve the localization accuracy. In the proposed technique, we decompose the recorded data into several morphological multi-scale components. In order to unveil weak signal, we propose an orthogonalization operator which acts as a time-varying weighting in the morphological reconstruction. The orthogonalization operator is obtained using an inversion process. This orthogonalized morphological reconstruction can be interpreted as a projection of the higher-dimensional vector. We first test the proposed technique using a synthetic dataset. Then the proposed technique is applied to a field dataset recorded in a project in China, in which the signals induced from hydraulic fracturing are recorded by twelve three-component (3-C) geophones in a monitoring well. The result demonstrates that the orthogonalized morphological reconstruction can make the extremely weak microseismic signals detectable.