Artificial Intelligence in Geosciences最新文献

The idea of using a deep autoencoder to encode seismic waveform features and then use them in different seismological applications is appealing. In this paper, we designed tests to evaluate this idea of using autoencoders as feature extractors for different seismological applications, such as event discrimination (i.e., earthquake vs. noise waveforms, earthquake vs. explosion waveforms), and phase picking. These tests involve training an autoencoder, either undercomplete or overcomplete, on a large amount of earthquake waveforms, and then using the trained encoder as a feature extractor with subsequent application layers (either a fully connected layer, or a convolutional layer plus a fully connected layer) to make the decision. By comparing the performance of these newly designed models against the baseline models trained from scratch, we conclude that the autoencoder feature extractor approach may only outperform the baseline under certain conditions, such as when the target problems require features that are similar to the autoencoder encoded features, when a relatively small amount of training data is available, and when certain model structures and training strategies are utilized. The model structure that works best in all these tests is an overcomplete autoencoder with a convolutional layer and a fully connected layer to make the estimation.

The identification of high-quality marine shale gas reservoirs has always been a key task in the exploration and development stage. However, due to the serious nonlinear relationship between the logging curve response and high-quality reservoirs, the rapid identification of high-quality reservoirs has always been a problem of low accuracy. This study proposes a combination of the oversampling method and random forest algorithm to improve the identification accuracy of high-quality reservoirs based on logging data. The oversampling method is used to balance the number of samples of different types and the random forest algorithm is used to establish a high-precision and high-quality reservoir identification model. From the perspective of the prediction effect, the reservoir identification method that combines the oversampling method and the random forest algorithm has increased the accuracy of reservoir identification from the 44% seen in other machine learning algorithms to 78%, and the effect is significant. This research can improve the identifiability of high-quality marine shale gas reservoirs, guide the drilling of horizontal wells, and provide tangible help for the precise formulation of marine shale gas development plans.

Derivative and volatility attributes can be usefully calculated from recorded gamma ray (GR) data to enhance lithofacies classification in wellbores penetrating multiple lithologies. Such attributes extract information about the log curve shape that cannot be readily discerned from the recorded well log data. A logged wellbore section for which 8911 data records are available for the three recorded logs (GR, sonic (DT) and bulk density (PB)) is evaluated. That section demonstrates the value of the GR attributes for machine learning (ML) lithofacies predictions. Five feature selection configurations are considered. The 9-var configuration including GR, DT, PB and six GR attributes, and the 7-var configuration of GR and the six GR attributes, provide the most accurate and reproducible lithofacies predictions. The other three feature configurations evaluated do not include the GR attributes but just one to three of the recorded log features. The results of seven ML models and two regression models reveal that K-nearest neighbor (KNN), random forest (RF) and extreme gradient boosting (XGB) are the best performing models. They generate between 14 and 23 misclassification from 8911 data records for the 9-var model. Multi-layer perceptron (MLP) and support vector classification (SVC) do not perform well with the 7-var model which lacks the PB feature displaying the highest correlation with facies class. Annotated confusion matrices reveal that KNN, RF and XGB models can effectively distinguish all facies classes for the 9-var and 7-var configurations (that includes the GR attributes), whereas none of the models can achieve that outcome for the 3-var configuration (that excludes the GR attributes). Accurately distinguishing lithofacies using well-log data in sedimentary sections is an important objective in applied geoscience. The straightforward, GR-attribute method proposed works to improve confidence in ML-lithofacies classifications based on limited recorded well-log data.

Noise suppression is an essential step in many seismic processing workflows. A portion of this noise, particularly in land datasets, presents itself as random noise. In recent years, neural networks have been successfully used to denoise seismic data in a supervised fashion. However, supervised learning always comes with the often unachievable requirement of having noisy-clean data pairs for training. Using blind-spot networks, we redefine the denoising task as a self-supervised procedure where the network uses the surrounding noisy samples to estimate the noise-free value of a central sample. Based on the assumption that noise is statistically independent between samples, the network struggles to predict the noise component of the sample due to its randomicity, whilst the signal component is accurately predicted due to its spatio-temporal coherency. Illustrated on synthetic examples, the blind-spot network is shown to be an efficient denoiser of seismic data contaminated by random noise with minimal damage to the signal; therefore, providing improvements in both the image domain and down-the-line tasks, such as post-stack inversion. To conclude our study, the suggested approach is applied to field data and the results are compared with two commonly used random denoising techniques: FX-deconvolution and sparsity-promoting inversion by Curvelet transform. By demonstrating that blind-spot networks are an efficient suppressor of random noise, we believe this is just the beginning of utilising self-supervised learning in seismic applications.

The purpose of this work is to assess the soil fertility for Tulaipanji rice cultivation in Kaliyaganj C.D. Block using the Analytic Hierarchy Process (AHP) and Machine learning algorithms along with the field survey data and GIS. A total of 40 soil samples from Tulaipanji rice fields (from 0 to 40 ​cm depth) have been randomly collected for the analysis of the soil health condition. For the purpose of assigning ratings to the parameters, ten experts' opinions were taken into account. The final soil fertility map indicates that 18.01% of the land is in excellent health condition to support Tulaipanji cultivation. The artificial neural networks (ANN), support vector machine (SVM), and Bagging models-based suitability analysis was also done using geo-spatial and soil data for Tulaipanji cultivation. Nevertheless, the ANN is the more appropriate model for locational analysis of Tulaipanji cultivation. The ANN-based findings show that areas of 25.8% (77.89 sq. km) are excellent for growing Tulaipanji rice, about 22.01% (66.45 sq. km) are highly suitable, 19.84% (59.90 sq. km) are moderately suitable, 21.19% (63.97 sq. km) are low suitable and 11.16% (33.69 sq. km) are not suitable for Tulaipanji rice cultivation. The receiver operating characteristic (ROC) curve depicts that the applied models have a high degree of accuracy. This endeavour will aid much in the soil fertility and site suitability assessment that will aid local government officials, academics, and the framers, to utilize the lands in a scientific way.

The utilization of urban underground space in a smart city requires an accurate understanding of the underground structure. As an effective technique, Rayleigh wave exploration can accurately obtain information on the subsurface. In particular, Rayleigh wave dispersion curves can be used to determine the near-surface shear-wave velocity structure. This is a typical multiparameter, high-dimensional nonlinear inverse problem because the velocities and thickness of each layer must be inverted simultaneously. Nonlinear methods such as simulated annealing (SA) are commonly used to solve this inverse problem. However, SA controls the iterative process though temperature rather than the error, and the search direction is random; hence, SA always falls into a local optimum when the temperature setting is inaccurate. Specifically, for the inversion of Rayleigh wave dispersion curves, the inversion accuracy will decrease with an increasing number of layers due to the greater number of inversion parameters and large dimension. To solve the above problems, we convert the multiparameter, high-dimensional inverse problem into multiple low-dimensional optimizations to improve the algorithm accuracy by incorporating the principle of block coordinate descent (BCD) into SA. Then, we convert the temperature control conditions in the original SA method into error control conditions. At the same time, we introduce the differential evolution (DE) method to ensure that the iterative error steadily decreases by correcting the iterative error direction in each iteration. Finally, the inversion stability is improved, and the proposed inversion method, the block coordinate descent differential evolution simulated annealing (BCDESA) algorithm, is implemented. The performance of BCDESA is validated by using both synthetic data and field data from western China. The results show that the BCDESA algorithm has stronger global optimization capabilities than SA, and the inversion results have higher stability and accuracy. In addition, synthetic data analysis also shows that BCDESA can avoid the problems of the conventional SA method, which assumes the S-wave velocity structure in advance. The robustness and adaptability of the algorithm are improved, and more accurate shear-wave velocity and thickness information can be extracted from Rayleigh wave dispersion curves.

The critical slip distance in rate and state model for fault friction in the study of potential earthquakes can vary wildly from micrometers to few me-ters depending on the length scale of the critically stressed fault. This makes it incredibly important to construct an inversion framework that provides good estimates of the critical slip distance purely based on the observed ac-celeration at the seismogram. To eventually construct a framework that takes noisy seismogram acceleration data as input and spits out robust estimates of critical slip distance as the output, we first present the performance of the framework for synthetic data. The framework is based on Bayesian inference and Markov chain Monte Carlo methods. The synthetic data is generated by adding noise to the acceleration output of spring-slider-damper idealization of the rate and state model as the forward model.