ML-based wavefield separation for shallow water OBN seismic data

The processing efficiency of the widely used time-space (t-x) domain vertical seismic profiling (VSP) wavefield separation methods (such as median and singular value decomposition filtering) via one-dimensional discrete Fourier transform (DFT) depends on the wavefield separation method (or algorithm with processing parameter) and the total number of samples in the input VSP data. Once the wavefield separation method is determined, its processing efficiency is set and cannot be optimized. Although the popular frequency-wavenumber (f-k) domain VSP wavefield separation method via two-dimensional DFT has higher processing efficiency than the t-x domain VSP wavefield separation methods, its processing precision is susceptible to the spatial alias and Gibbs effects. For efficiency and precision improvements, we introduced the discrete cosine transform (DCT) operation into VSP wavefield separation for the first time, and proposed a high-efficiency and precision frequency-space (f-x) domain VSP wavefield separation method via DFT and DCT, in which the wavefield separation efficiency and precision can be optimized using the effective bandwidth cutoff frequency of the input VSP data. Based on the relationship between DFT and DCT, we combined their operations in the proposed method into one step (referred to as DCT–DFT) for further efficiency enhancement, thereby designing a high-efficiency and precision f-x domain VSP wavefield separation method via DCT–DFT. Theory analysis and synthetic and field VSP data examples show that the proposed method is highly efficient and precise, and can be widely used for three-dimensional (3D)-VSP data wavefield separation, especially for large distributed acoustic sensing (DAS)-VSP data.

Seismic data suffer from under-sampling along spatial dimensions due to physical and economic limitations during acquisition, which can adversely impact subsequent imaging processes. Various approaches have been proposed to address this issue, including rank-reduction, transform, and deep learning (DL) methods for reconstructing missing traces in observed seismic data. While many DL-based reconstruction methods employ convolutional neural networks (CNNs) as their core components in supervised or unsupervised settings, the accuracy of CNN-based DL reconstruction methods can still be enhanced due to the inherent localization of convolutional operators. In this study, we present an innovative method, termed Attentive CNN (ACNet), that incorporates a transformer model with a multi-head self-attention mechanism into the CNN framework for unsupervised seismic data reconstruction. ACNet consists of an encoder built with the transformer model and a CNN decoder, enabling an iterative reconstruction process within the proposed unsupervised DL-based approach. We compare the performance of our method with that of the damped rank-reduction (DRR) method using synthetic and field data examples. The results demonstrate that our proposed method exhibits superior signal reconstruction and random noise suppression capabilities compared to the DRR method.

Seismic data interpolation and denoising play vital roles in obtaining complete and clean data in seismic data processing. Seismic data usually misses along various spatial axes and always mix with random noise. In order to obtain complete and clean seismic data, reconstruction technology can interpolate missing data and attenuate random noise. Nonsubsampled contourlet transform is an effective transform to obtain multi-scale and multi-direction sparse domain data for compression sensing interpolation and denoising. However, conventional iterative shrinkage/thresholding cannot handle ill-posed and ill-conditioned equations for solving linear inverse problem. We present a two-step iterative log thresholding method to overcome ill-posed and ill-conditioned problems and improve the convergence rate and solution accuracy, which can interpolate and denoise seismic data simultaneously in the nonsubsampled contourlet transform framework. First, we use nonsubsampled contourlet transform to convert the seismic missing data with random noise to sparse domain. Then, we apply two-step iterative log thresholding algorithm to interpolate and denoise data in sparse domain. The result of each iteration is based on the results of the previous two iterations, which can accelerate convergence rate. In addition, log thresholding can further improve convergence rate and solution accuracy. Finally, we use inverse nonsubsampled contourlet transform to obtain the interpolated and denoised seismic data. The new method can reconstruct the irregularly missing data and attenuate random noise to obtain complete and clean seismic data with high accuracy, which is crucial for seismic imaging and inversion. We demonstrate the applicability and effectiveness of this simultaneous interpolation and denoising technique with successful applications to both synthetic and field data examples.

The physical and cost limitations of seismic data acquisition often result in the spatial undersampling of data, which has a detrimental impact on subsequent data processing. Seismic data reconstruction plays a crucial role in recovering missing traces arising from the acquisition process. In recent years, deep learning (DL) has emerged as an intelligent solution for this purpose. We introduce an innovative approach called the weighted-attentive DL framework for unsupervised 3D seismic data reconstruction, which combines an attentive transformer network (ATNet) with a conventional projection onto convex sets (POCS) algorithm. Our framework follows the plug-and-play concept, requiring only the original subsampled data to achieve robust performance. It accomplishes this by iteratively updating ATNet parameters, wherein ATNet serves as the reconstruction operator during the DL process. This approach allows us to simultaneously recover missing traces and filter out random noise, with a key feature being its enhanced attention to the observed signal compared with convolutional neural networks. In addition, we incorporate the POCS algorithm into our framework to introduce a linear decay weight for the aliasing space containing noise and signal during the reconstruction process. We rigorously evaluate our method against four existing approaches: adaptive POCS, fast dictionary learning sequential generalized K-means, optimally damped rank-reduction, and DenseNet methods. Our comparative experiments, conducted on synthetic and field data examples, clearly demonstrate the substantial improvements achieved by our method in terms of reconstruction and denoising when compared with the four benchmarked methods.

Magnetotelluric (MT) inversion algorithms are an essential tool in exploration geophysics because they provide us with resistivity models of the subsurface. Inverse problems are ill-posed and non-unique. Resistivity models produced by geologically unconstrained MT inversions are typically smooth, which means that the geological interpretations obtained based on the resistivity models will be ambiguous and uncertain. To solve the above problems, geophysicists have made many efforts over the years to introduce any prior information (such as geological structure information and physical property data) into models, which could help to better constrain the inverted model with the inversion algorithms. In the last few years, there has been renewed interest in machine learning techniques. Various machine learning methods have been applied to inversions. Using the fuzzy c-means (FCM) clustering method, geophysicists proposed new inversion algorithms that are capable of building petrophysical information into the inversion. Although the FCM has been used and advanced in previous work, geophysicists concluded that a priori information of the correct number and value of cluster centers is very important for a successful FCM inversion. In fact, it is difficult to obtain the appropriate clustering information in some geophysical survey areas. In this study, we present an effective way to build the petrophysical information for the MT inversion based on FCM clustering algorithm. When the actual petrophysical information is insufficient, considering the characteristics of MT data, we perform a one-dimensional blocky inversion that can clearly identify the interface with distinct electrical property contrast; then, we obtain the number and value of cluster centers from 1D blocky inversion for the further MT inversion with FCM clustering. The algorithm uses guided FCM clustering to improve the model within the iterative minimization during MT inversion. In our MT inversion method, we integrate the geophysical inversion and geological differentiation into a unified scheme, which interact and enhance each other. The resistivity models obtained from the inversion respect the geophysical and petrophysical data and are easy to geologically interpret. We tested the algorithm using two synthetic examples and a field data example.

Multiscale frequency-domain seismic inversion for fracture weakness

Ground-roll attenuation is very challenging because of its high amplitudes and overlapping frequency content with desired signals. A particular challenge is to recover weak reflections underneath strong masking ground-roll. We propose a dual-filter bank setup combined with two convolutional neural networks (CNNs) to realize ground-roll attenuation. The rationale for using a dual-filter bank strategy is that it permits using two CNNs with different input kernel sizes and different complexities to recognize and extract broad-scale (low-wavelength) and narrow-scale (high-wavelength) features separately. We also apply a frequency filter to create a preliminary separation between the signal and the noise. In addition, we use a radial trace transform that focuses desired signal to a smaller area, facilitating separation of the reflections and ground-roll and accelerating training. The network training strategy combines synthetic and field data examples, in addition to noise injection to augment the number of available training samples. Tests on synthetic and field datasets show that the proposed strategy achieves superior ground-roll attenuation compared with standard methods, even in the case of data with irregular spatial spacing or ground-roll characteristics not contained in the training data.

Seismic interpretation is an essential factor in the success of field development projects because it allows geologists and engineers to better understand subsurface formations and identify potential resources; however, noisy seismic data, especially in regions with complex geological structures, frequently hinder this process. Such noise emerging from equipment limitations, environmental interference, and inherent geological complexities can obscure essential geological features, making accurate subsurface imaging challenging and reducing the effectiveness of traditional interpretation methods. This paper presents a machine learning (ML) method to enhance the conditioning of 2D and 3D post-stack seismic data by targeting noise and migration artifacts. These issues are often responsible for compromising the quality of seismic data sets. The method this paper presents leverages convolutional neural networks (CNNs) to identify and correct spatial inconsistencies within the data. With minimal preprocessing, requiring only the conversion of data into a 3D NumPy array format, this approach simplifies the workflow while improving data clarity and accuracy. Early statistical analysis of outliers and data normalization provides the ML model with stable inputs, allowing it to iteratively train on raw data and infer the original signal patterns in an unsupervised manner without the need for labeled data sets. The model outputs a conditioned version of the seismic data, significantly reducing noise while preserving critical geological features. The results demonstrate that ML seismic conditioning improves fault imaging, horizon continuity in areas with low signal-to-noise ratios (SNR), and localized smoothing without compromising resolution. Furthermore, the method presented in this paper retains signal fidelity and relative amplitude strengths, ensuring that geological interpretations remain accurate and reliable. By automating the traditional manual task of data conditioning, the proposed solution optimizes interpreter workflows, allowing them to focus on higher-level geological analysis and decision-making. This approach ultimately increases the accuracy of subsurface interpretations and resource exploration, improving the overall quality and efficiency of seismic interpretation projects. In addition, for 2D seismic, traditional methods of denoising rely on expert judgement using tools such as f-x deconvolution. 2D CNNs offer an efficient alternative, especially when training data can be synthesized and leveraged in supervised learning. This method can help the CNN identify and remove noise while preserving pertinent subsurface structures. This paper presents a 2D CNN architecture containing a model trained with synthetic seismic data to improve SNR of field seismic data; hence, interpretation.

Event detection is an essential component of microseismic data analysis. This process is typically carried out using a short- and long-term-average-ratio (STA/LTA) method, which is simple and computationally efficient but often yields inconsistent results for noisy data sets. We have aimed to optimize the performance of the STA/LTA method by testing different input forms of 3C waveform data and different characteristic functions (CFs), including a proposed [Formula: see text]-mean CF. These tests are evaluated using receiver operating characteristic (ROC) analysis and are compared based on synthetic and field data examples. Our analysis indicates that the STA/LTA method using a [Formula: see text]-mean CF improves the detection sensitivity and yields more robust event detection on noisy data sets than some previous approaches. In addition, microseismic events are detected efficiently on field data examples using the same detection threshold obtained from the ROC analysis on synthetic data examples. We recommend the use of the Youden index based on ROC analysis using a training subset, extracted from the continuous data, to further improve the detection threshold for field microseismic data.

Seismic interpolation via multi-scale HU-Net

Production through wellbores in petroleum industry involves multi-phase flow of oil, gas and water. It is necessary to capture the effects of varying composition of phases to represent wellbore hydraulics accurately. However, a majority of current wellbore simulators prefer a ‘black-oil’ formulation due to its higher speed as compared to a compositional approach. In this work, a neuro-simulation protocol is developed for fast and accurate simulation of wellbore hydraulics in vertical non-isothermal systems. The proposed toolbox is capable of predicting the pressure profile along a production tubing containing flow of pure hydrocarbons (C1 through C45) and water. The tool can operate within a range of flow rates, well depths, pipe diameters, inlet fluid compositions, wellhead pressures and geothermal gradients. The data for training the expert system is generated from a numerical simulator that computes wellbore hydraulics. The numerical simulator is a compositional model, which uses a two-phase equilibrium flash calculation for determining the distribution of hydrocarbon components in oil and gas phases. A two-phase drift flux model accounts for the variation in gas and liquid velocities under various flow regimes. The compositional mass balance and momentum equations are simultaneously solved using a fully implicit scheme, while the temperature profile of the well is taken into consideration in computations. The results provided by both the numerical simulator and the tool described in this paper are compared against field data as well as numerical simulation results reported in the literature. It is observed that the average relative deviation of the results generated in this study from the field data example is 2.2 %. The average relative deviation between the numerical simulator developed and field data example is 2.4 %, showing that the expert system is capable of capturing the related physics and thermodynamics of the wellbore hydraulics phenomena. The average relative deviation between the results generated by the proposed tool and numerical simulation examples from literature data is observed to be between 3.3 and 7.1 %. The average relative deviations between results from the numerical simulator developed in the current work and simulation results reported in literature are found to be between 2.3 and 8.1 %. The variation between predictions provided by various models for wellbore hydraulics in the literature lie within the ranges mentioned above. It is observed that the tool discussed in this paper is a fast, accurate and robust model for predicting wellbore hydraulics for compositional systems. The proposed tool also overcomes the stability issues associated with compositional numerical wellbore simulation models. Such a versatile tool can also be used for generating tubing intake curves for an unrestricted set of scenarios and for quick selection of optimal gas lift design parameters.

Shallow-seismic Rayleigh waves are attractive for geotechnical site investigations. They exhibit a high signal to noise ratio in field data recordings and have a high sensitivity to the S-wave velocity, an important lithological and geotechnical parameter to characterize the very shallow subsurface. In recent years we studied the applicability of the two-dimensional elastic FWI method using numerous synthetic reconstruction tests and two field data examples. Some important challenges are reported here: (1) the accurate correction of the geometrical spreading, (2) the estimation of the source wavelet, (3) the importance of an-elastic attenuation in the forward simulations. We found that Important pre-processing steps for the application of 2-D elastic FWI to shallow-seismic field data are the 3D to 2D correction of geometrical spreading and the estimation of a priori Q-values that must be used as a passive medium parameter during the FWI. Furthermore, a source-wavelet correction filter should be applied during the FWI process. Smooth initial models obtained from the analysis of the first arrivals of body waves are important and seem to be sufficient. Our field data examples indicate that FWI is able to resolve lateral variations of S-wave velocities in the very shallow subsurface.

Summary We have developed a novel random noise attenuation method for seismic data based on inversion in the frequency-space (f-x) domain. The method regards ‘clean’ seismic data in f-x domain as model parameters, and the noise attenuation is equivalent to inverting these parameters from the noisy data. The prediction error filter (PEF) in the f-x prediction method is used as lateral continuity constraints of the inversion. Since the PEF is calculated from the noisy data, it should be recalculated from the denoised seismic data. The process of calculating signal, then getting a new PEF should be iterated a few times. The choice of trade-off parameters also influences the performance of this method. The proposed method overcame the problem of the generation of spurious events and reduction of the signal amplitudes in the conventional f-x prediction method. Tests on synthetic and field data examples demonstrated that, compared with f-x domain prediction, the proposed method has a better performance in noise attenuation and signal-preserving.

In this paper, a comprehensive review on Alzheimer's disease (AD) is carried out, and an exploration of the two machine learning (ML) methods that help to identify the disease in its initial stages. Alzheimer's disease is a neurocognitive disorder occurring in people in their early onset. This disease causes the person to suffer from memory loss, unusual behavior, and language problems. Early detection is essential for developing more advanced treatments for AD. Machine learning (ML), a subfield of Artificial Intelligence (AI), uses various probabilistic and optimization techniques to help computers learn from huge and complicated data sets. To diagnose AD in its early stages, researchers generally use machine learning. The survey provides a broad overview of current research in this field and analyses the classification methods used by researchers working with ADNI data sets. It discusses essential research topics such as the data sets used, the evaluation measures employed, and the machine learning methods used. Our presentation suggests a model that helps better understand current work and highlights the challenges and opportunities for innovative and useful research. The study shows which machine learning method holds best for the ADNI data set. Therefore, the focus is given to two methods: the 18-layer convolutional network and the 3D convolutional network. Hence, CNNs with multi-layered fetch more accurate results as compared to 3D CNN. The work also contributes to the use of the ADNI data set, where the classification of training and testing samples is divided with such a number that brings the highest accuracy achieved with 18-layer CNN. The work concentrates on the early prediction of Alzheimer's disease with machine learning methods. Thus, the accuracy achieved is 98% for 18-layer CNN.

Seismic imaging from sparsely acquired data faces challenges such as low image quality, discontinuities, and migration swing artifacts. Existing convolutional neural network (CNN)-based methods struggle with complex feature distributions and cannot effectively assess uncertainty, making it hard to evaluate the reliability of their processed results. To address these issues, we propose a new method using a generative diffusion model (GDM). Here, in the training phase, we use the imaging results from sparse data as conditional input, combined with noisy versions of dense data imaging results, for the network to predict the added noise. After training, the network can predict the imaging results for test images from sparse data acquisition, using the generative process with conditional control. This GDM not only improves image quality and removes artifacts caused by sparse data but also naturally evaluates uncertainty by leveraging the probabilistic nature of the GDM. To overcome the decline in generation quality and the memory burden of large-scale images, we develop a patch fusion strategy that effectively addresses these issues. Synthetic and field data examples demonstrate that our method significantly enhances imaging quality and provides effective uncertainty quantification.