1,721,030 research outputs found
Application of Deep Learning in the estimation of CO2 saturation maps
No full textCO2 saturation estimation during the monitoring of a carbon-dioxide storage project using traditional inversion techniques can become an intensive task due to the processing steps and inversion workflows involved. In our work, we propose a deep-learning algorithm based on the U‐Net architecture, to develop a method for automatic prediction of the spatial distribution of CO2 saturation from time-lapse seismic data. We suggest using continuous wavelet transform (CWT) as feature extraction from the recorded shot gathers. CWT could provide more informative and distinguishable images. The method is tested on synthetic time-lapse data generated from the use of Gassmann fluid substitution equation to calculate the post-injection P-wave velocity model used to perform the 2D finite-difference acoustic modelling. The neural network has been trained on pairs of synthetic seismic shot gathers and corresponding CO2 saturation maps, along with pairs of CWT coefficient maps and CO2 saturation maps. The results of this work indicate the prospective utility of CWT in producing more informative images for the prediction of CO2 saturation maps through deep learning algorithm like U-net
Deep attributes: innovative LSTM-based seismic attributes
Full text linkSeismic attributes are derived measures from seismic data that help characterize subsurface geological features and enhance the interpretation of subsurface structures: we propose to exploit the hidden layers of Long-Short Time Memory neural network predictions as possible new reflection seismic attributes. The idea is based on the inference process of a neural network, which in its hidden layers stores information related to different features embedded in the input data and which usually are not considered. Neural network applications typically ignore such intermediate steps because the main interest lies in the final output, which is considered as the exclusive exploitable feature of the process. On the contrary, here we analyse the possibility to exploit the intermediate prediction steps, hereafter referred as “Deep Attributes” because they are produced by a deep learning algorithm, to highlight features and emphasize characteristics embedded in the data but neither recognizable by traditional interpretation, nor evident with classical attributes or multi-attribute approaches. Nowadays, classical signal attributes are numerous and used for different purposes; we here propose an original strategy to calculate attributes previously never exploited, which are potentially complementary or a good alternative to the classical ones. We tested the proposed procedure on synthetic and field 2-D and 3-D reflection seismic data sets to test and demonstrate the stability, affordability and versatility of the entire approach. Furthermore, we evaluated the performance of deep attributes on a 4-D seismic dataset to assess the applicability and effectiveness for time-monitoring purposes and comparing them with the sweetness attribute
Monitoring Elastic Parameters Changes during Underground Hydrogen Storage Using Rock Physics Parametrized FWI
No full textSeasonal storage of hydrogen produced from renewable energy can become one of the key strategies tomeet the high energy demand of today’s society. The most suitable sites for seasonal hydrogen storageare depleted gas fields. This study proposes a workflow based on the integration of full-waveforminversion (FWI), rock physics modeling (RPM) and gas property modeling for monitoring changes inelastic medium parameters due to hydrogen injection. A rock physics model including the Gassmannequation and fluid mixing laws has been implemented, which accurately links rock physical propertiesto elastic properties. The parameterized approach is based on optimizing fluid saturation to reducecrosstalk between model parameters during the inversion process, while simultaneously providing aquantitative estimation of the fluid within the reservoir. The synthetic models show that parameterizedinversion produces higher accuracy and fewer artifacts than conventional FWI. Our results underscorethe importance of an appropriate RPM to reflect real subsurface conditions and proper fluid mixinglaws. Therefore, FWI parameterization provides an efficient technique for monitoring hydrogen storagesites in depleted gas fields to ensure efficient storage
An overview of GPR investigation in the Italian Alps
No full textDifferent applications of Ground Penetrating Radar (GPR) in glaciology are discussed through examples of mapping bedrock or internal features of glaciers or the characterization of snow properties and frozen materials’ physical parameters like electromagnetic velocity and density. A first example focuses on the appli¬cation of GPR to the estimate of snow water equivalent, density at scale of basin, of interest for the analysis of the hydrological behaviour during the snow melt. Examples on the radar survey to map bedrock, investigate the inner features of glaciers and monitor its evolution with time are herein discussed. We focus particularly on the radar survey of the Cevedale glacier, to get information on the thickness of glacier. Another example shows the result of geophysical characterization of iced-bodies in the Canin massif, located in the north-east part of the Italian Alps. Moreover, an example of a 4D GPR data survey is provided, demonstrat¬ing the applicability of GPR as an efficient tool to estimate the seasonal mass balance of a glacier, with a higher overall accuracy than direct methods. Ground Penetrating Radar (GPR) is generally used for locating targets in ice, determining ice and snow thickness, glacialogical studies and crevasse detection. In glaciology, the low electrical conductivity and reduced water content of glaciers allows penetration depths sufficient to detect the bedrock in most glaciers, with frequency ranging from 50 MHz to 200 MHz or even higher. The variations of electromagnetic properties of the inner features causes reflections that are well detectable (Godio and Rege, 2015) and provide extremely valuable data in glaciological studies of the hydrology and structural evolution of glaciers, for mapping and monitoring permafrost evolution (Carturan et al., 2012; Colucci et al., 2014)
Quantitative 3-D GPR analysis to estimate the total volume and water content of a glacier
No full textWe apply an automated picking and inversion algorithm to a 3-D GPR data set acquired on an alpine glacieret, to study its internal stratigraphy, density distribution, total volume, and water content. GPR surveys are particularly useful for glaciological studies, since the transmitted signal can propagate efficiently through the entire glacier volume, while the large number of recorded traces makes any quantitative analysis statistically sound. The applied auto-picking algorithm is designed to accurately and objectively identify the main reflections within a GPR data set, and to characterize them in terms of their peak amplitudes, travel times, and polarities. The inversion algorithm then uses these quantities to recover the subsurface stratigraphy and EM velocity distribution along each GPR profile. In air-ice mixtures, the EM velocity is linked to the density through well-known empirical formulas. Therefore, our inversion algorithm is able to recover the density distribution within a glacier, and combine it with the internal stratigraphy to estimate its water content. By applying this procedure to a 3-D GPR data set, we can obtain an accurate model of an entire glacier, while 4-D surveys can be used to monitor its temporal changes and estimate its annual and seasonal mass balances
Synthetic seismic data generation with deep learning
Full text linkWe study the applicability of deep learning (DL) methods to generate acoustic synthetic data from 1D models of the subsurface. We designed and implemented a Neural Network (NN) and we trained it to generate synthetic seismograms (common shot gathers) from 1-D velocity models on two different datasets: one obtained from published results and the other generated by Finite Differences (FD) numerical simulation. We furthermore compared the results from the proposed model with the published one. Moreover, we tried to to add more flexibility to this methodology by allowing change of wavelet and the acquisition geometry. We obtained good results in terms of both computation efficiency and quality of prediction. The main potentialities of the work are the low computational cost, a high prediction speed and the possibility to solve complex non-linear problems without knowing the physical law behind the phenomenon, which could led great advantages in the solution also of the inverse problem. DL to generate 1-D acoustic synthetic seismograms without solving wave equation Solution to the 1-D problem through custom Recurrent Neural Network Retraining strategy to improve flexibility and applicability Computational complexity analysis
Tecniche lineari e azimutali nelle indagini Ground Penetrating Radar per l’individuazione di sottoservizi in aree urbane
No full textEfficient extraction of seismic reflection with Deep Learning
Full text linkWe propose a procedure for the interpretation of horizons in seismic reflection data based on a Neural Network (NN) approach, which can be at the same time fast, accurate and able to reduce the intrinsic subjectivity of manual or control-points based methods. The training is based on a Long Short Term Memory architecture and is performed on synthetic data obtained from a convolutional model-based scheme, while the extraction step can be applied to any type of field seismic dataset. Synthetic data are contaminated with different types of noise to improve the performance of the NN in a large variety of field conditions. We tested the proposed procedure on 2-D and 3-D synthetic and field seismic datasets. We have successfully applied the procedure also to Ground Penetrating Radar data, verifying its versatility and potential. The proposed algorithm is based on a fully 1-D approach and does not require the input of any interpreter, because the necessary thresholds are automatically estimated. An added benefit is that the prediction has an associated probability, which automatically quantifies the reliability of the results
Polarity assessment of reflection seismic data: a Deep Learning approach
Full text linkWe propose a procedure for the polarity assessment in reflection seismic data based on a Neural Network approach. The algorithm is based on a fully 1D approach, which does not require any input besides the seismic data since the necessary parameters are all automatically estimated. An added benefit is that the prediction has an associated probability, which automatically quantifies the reliability of the results. We tested the proposed procedure on synthetic and real reflection seismic data sets. The algorithm is able to correctly extract the seismic horizons also in case of complex conditions, such as along the flanks of salt domes, and is able to track polarity inversions
Indagini Side Scan Sonar nell’area delle Bocche di Bonifacio. Arcipelago di La Maddalena (Sardegna, Italia)
No full text- …
