RESUMEN
In the domain of medical image segmentation, traditional diffusion probabilistic models are hindered by local inductive biases stemming from convolutional operations, constraining their ability to model long-term dependencies and leading to inaccurate mask generation. Conversely, Transformer offers a remedy by obviating the local inductive biases inherent in convolutional operations, thereby enhancing segmentation precision. Currently, the integration of Transformer and convolution operations mainly occurs in two forms: nesting and stacking. However, both methods address the bias elimination at a relatively large granularity, failing to fully leverage the advantages of both approaches. To address this, this paper proposes a conditional diffusion segmentation model named TransDiffSeg, which combines Transformer with convolution operations from traditional diffusion models in a parallel manner. This approach eliminates the accumulated local inductive bias of convolution operations at a finer granularity within each layer. Additionally, an adaptive feature fusion block is employed to merge conditional semantic features and noise features, enhancing global semantic information and reducing the Transformer's sensitivity to noise features. To validate the impact of granularity in bias elimination on performance and the impact of Transformer in alleviating the accumulated local inductive biases of convolutional operations in diffusion probabilistic models, experiments are conducted on the AMOS22 dataset and BTCV dataset. Experimental results demonstrate that eliminating local inductive bias at a finer granularity significantly improves the segmentation performance of diffusion probabilistic models. Furthermore, the results confirm that the finer the granularity of bias elimination, the better the segmentation performance.
RESUMEN
In order to optimize the efficiency and safety of gas hydrate extraction, it is essential to develop a credible constitutive model for sands containing hydrates. A model incorporating both cementation and damage was constructed to describe the behavior of hydrate-bearing cemented sand. This model is based on the critical state theory and builds upon previous studies. The damage factor Ds is incorporated to consider soil degradation and the reduction in hydrate cementation, as described by plastic shear strain. A computer program was developed to simulate the mechanisms of cementation and damage evolution, as well as the stress-strain curves of hydrate-bearing cemented sand. The results indicate that the model replicates the mechanical behavior of soil cementation and soil deterioration caused by impairment well. By comparing the theoretical curves with the experimental data, the compliance of the model was calculated to be more than 90 percent. The new state-dependent elasto-plastic constitutive model based on cementation and damage of hydrate-bearing cemented sand could provide vital guidance for the construction of deep-buried tunnels, extraction of hydrocarbon compounds, and development of resources.
RESUMEN
Convolutional Neural Networks have been widely applied in medical image segmentation. However, the existence of local inductive bias in convolutional operations restricts the modeling of long-term dependencies. The introduction of Transformer enables the modeling of long-term dependencies and partially eliminates the local inductive bias in convolutional operations, thereby improving the accuracy of tasks such as segmentation and classification. Researchers have proposed various hybrid structures combining Transformer and Convolutional Neural Networks. One strategy is to stack Transformer blocks and convolutional blocks to concentrate on eliminating the accumulated local bias of convolutional operations. Another strategy is to nest convolutional blocks and Transformer blocks to eliminate bias within each nested block. However, due to the granularity of bias elimination operations, these two strategies cannot fully exploit the potential of Transformer. In this paper, a parallel hybrid model is proposed for segmentation, which includes a Transformer branch and a Convolutional Neural Network branch in encoder. After parallel feature extraction, inter-layer information fusion and exchange of complementary information are performed between the two branches, simultaneously extracting local and global features while eliminating the local bias generated by convolutional operations within the current layer. A pure convolutional operation is used in decoder to obtain final segmentation results. To validate the impact of the granularity of bias elimination operations on the effectiveness of local bias elimination, the experiments in this paper were conducted on Flare21 dataset and Amos22 dataset. The average Dice coefficient reached 92.65% on Flare21 dataset, and 91.61% on Amos22 dataset, surpassing comparative methods. The experimental results demonstrate that smaller granularity of bias elimination operations leads to better performance.
Asunto(s)
Redes Neurales de la Computación , Humanos , Abdomen/diagnóstico por imagen , Abdomen/anatomía & histología , Procesamiento de Imagen Asistido por Computador/métodos , Algoritmos , Tomografía Computarizada por Rayos X , Bases de Datos FactualesRESUMEN
In the field of CO2 capture utilization and storage (CCUS), recent advancements in active-source monitoring have significantly enhanced the capabilities of time-lapse acoustical imaging, facilitating continuous capture of detailed physical parameter images from acoustic signals. Central to these advancements is time-lapse full waveform inversion (TLFWI), which is increasingly recognized for its ability to extract high-resolution images from active-source datasets. However, conventional TLFWI methodologies, which are reliant on gradient optimization, face a significant challenge due to the need for complex, explicit formulation of the physical model gradient relative to the misfit function between observed and predicted data over time. Addressing this limitation, our study introduces automatic differentiation (AD) into the TLFWI process, utilizing deep learning frameworks such as PyTorch to automate gradient calculation using the chain rule. This novel approach, AD-TLFWI, not only streamlines the inversion of time-lapse images for CO2 monitoring but also tackles the issue of local minima commonly encountered in deep learning optimizers. The effectiveness of AD-TLFWI was validated using a realistic model from the Frio-II CO2 injection site, where it successfully produced high-resolution images that demonstrate significant changes in velocity due to CO2 injection. This advancement in TLFWI methodology, underpinned by the integration of AD, represents a pivotal development in active-source monitoring systems, enhancing information extraction capabilities and providing potential solutions to complex multiphysics monitoring challenges.
RESUMEN
In the production process of tight gas wells, reservoir fluid distribution and gas-water relative permeability vary with time. However, traditional models fail to handle the time-dependent mechanism and stress sensitivity effect in the reservoir, leading to significant errors in the dynamic analysis results. To address this issue, this article presents a prediction model for fractured well production in tight gas reservoirs. It is based on a three-dimensional embedded discrete fracture model (EDFM), which considers the influences of the time-dependent mechanism and stress-dependent reservoir permeability. Transient flow equations are treated by using the finite volume method to obtain the solution of the model. The accuracy and reliability of the model are verified by comparison with the results of the commercial simulator Eclipse and the field application. Based on the model's solution, this study emphasizes the analysis of the impact of the time-dependent mechanism and reservoir stress sensitivity on gas well productivity. Simulation results show that the time-dependent relative permeability curve can decrease the level of irreducible water saturation and promote the migration of irreducible water, resulting in an increase in water permeability and a decrease in gas permeability. This effect will reduce the period of stable gas production and increase the level of water production. Besides, reservoir stress sensitivity will reduce daily water production and accelerate gas well decline. It is necessary to control the production pressure difference reasonably during the production process to effectively reduce the negative impact of stress sensitivity effects. The results indicate that when the relative permeability curve and the reservoir permeability are constant, the real gas production capacity of the reservoir will be strengthened. The application of field case studies shows that the theoretical model exhibits stronger adaptability, achieves better fitting results, and can guide the compilation and adjustment of development plans for water-bearing tight gas reservoirs. These findings provide insights into understanding the effects of the time-dependent mechanism on gas production rates in tight gas reservoirs. Furthermore, this study offers useful guidance for the prediction of field-scale gas production.
RESUMEN
The organic matter in soft clay tends to affect the properties of cement-stabilized soil. The influence degree of different organic matter varies. In this paper, the influence weights and mechanism of the main organic matter components fulvic acid and humic acid on the mechanical properties of cemented soil were investigated. Impacts of FA/HA (fulvic acid/humic acid) values and curing time on the unconfined compressive strength, deformation characteristics, and microstructure of cemented soil were explored through the unconfined compressive strength test and electrical resistivity test. The results show that with the increase of FA/HA, the unconfined compressive strength of cemented soil gradually decreased and the plastic properties enhanced. The increase in curing time changed the stress-strain relationship of cemented soil, and some specimens showed brittle damage. The initial resistivity and structural property parameters of cemented soil gradually decreased with the increasing FA/HA value and increased with the increase of curing time. It revealed the influence law of FA/HA and curing time change on the microstructure of cemented soil. Based on the experimental results, the quantitative relationship equations between FA/HA and curing time and unconfined compressive strength, failure strain, deformation modulus, and resistivity were established.
RESUMEN
Aiming at the problem of chemical-mechanics-hydro (C-M-H) action encountered by rocks in underground engineering, chemical damage variables, water damage variables, and force damage variables are introduced to define the degree of degradation of rock materials. Stone is selected as the sample for acid corrosion treatment at pH 3, 4, and 7, and a chemical damage factor is defined that coupled the pH value and duration of exposure. Then based on the spatial mobilized plane (SMP) criterion and the Lemaitre strain equivalence hypothesis, this research develops a constitutive model considering rock chemical corrosion-water-confining pressure damage. The proposed damage constitutive model employs the extremum method to ascertain the two Weibull distribution parameters (m and F0) by theoretical derivation and exhibits satisfactory conformity between the theoretical and experimental curves. The damage constitutive model can be consistent in the stress-strain characteristics of the rock triaxial compression process, which verifies the rationality and reliability of the model parameters. The model effectively represents the mechanical properties and damage characteristics of rocks when subjected to the combined influence of water chemistry and confinement. The presented model contributes to a better understanding of tangible rock-engineered structures subjected to chemical corrosion in underwater environments.
RESUMEN
Fluid penetration into the rock during hydraulic fracturing has been an essential issue in studying the mechanism of fracture initiation, especially the seepage force caused by fluid penetration, which has an important effect on the fracture initiation mechanism around a wellbore. However, in previous studies, the effect of seepage force under unsteady seepage on the fracture initiation mechanism was not considered. In this study, a new seepage model that can predict the variations of pore pressure and seepage force with time around a vertical wellbore for hydraulic fracturing was established by using the method of separation of variables and the Bessel function theory. Then, based on the proposed seepage model, a new circumferential stress calculation model considering the time-dependent effect of seepage force was established. The accuracy and applicability of the seepage model and the mechanical model were verified by comparison with numerical, analytical and experimental results. The time-dependent effect of seepage force on fracture initiation under unsteady seepage was analyzed and discussed. The results show that when the wellbore pressure is constant, the circumferential stress induced by seepage force increases over time, and the possibility of fracture initiation also increases. The higher the hydraulic conductivity, the lower the fluid viscosity and the shorter the time required for tensile failure during hydraulic fracturing. In particular, when the tensile strength of rock is lower, the fracture initiation may occur within the rock mass rather than on the wellbore wall. This study is promising to provide a theoretical basis and practical guidance for further research on fracture initiation in the future.
RESUMEN
In order to accurately describe the characteristics of each stage of rock creep behavior under the combined action of acid environment and true triaxial stress, based on damage mechanics, chemical damage is connected with elastic modulus; thus, the damage relations considering creep stress damage and chemical damage are obtained. The elastic body, nonlinear Kelvin body, linear Kelvin body, and viscoelastic-plastic body (Mogi-Coulomb) are connected in series, and the actual situation under the action of true triaxial stress is considered at the same time. Therefore, a damage creep constitutive model considering the coupling of rock acid corrosion and true triaxial stress is established. The parameters of the deduced model are identified and verified with the existing experimental research results. The yield surface equation of rock under true triaxial stress is obtained by data fitting, and the influence of intermediate principal stress on the creep model is discussed. The derived constitutive model can accurately describe the characteristics of each stage of true triaxial creep behavior of rock under acid environment.
RESUMEN
The flat-joint model, which constructs round particles as polygons, can suppress rotation after breakage between particles and simulate more larger compression and tension ratios than the linear parallel-bond model. The flat-joint contact model was chosen for this study to calibrate the rock for 3D experiments. In the unit experiments, the triaxial unit was loaded with flexible boundaries, and the influence of each microscopic parameter on the significance magnitude of the macroscopic parameters (modulus of elasticity E, Poisson's ratio ν, uniaxial compressive strength UCS, crack initiation strength σci, internal friction angle φ and uniaxial tensile strength TS) was analysed by ANOVA (Analysis of Variance) in an orthogonal experimental design. Among them, Eƒ, kƒ has a significant effect on E; Cƒ and kƒ have a significant effect on ν; Cƒ, σƒ and kƒ have a significant effect on UCS; Cƒ; σƒ and Eƒ have a significant effect on TS; Rsd has a significant effect on σci; and φf, Eƒ, kƒ, µƒ, and σƒ have a significant effect on φ. Regressions were then carried out to establish the equations for calculating the macroscopic parameters of the rock material so that the three-dimensional microscopic parameters of the PFC can be quantitatively analysed and calculated. The correctness of the establishment of the macroscopic equations was verified by comparing the numerical and damage patterns of uniaxial compression, Brazilian splitting, and triaxial experiments with those of numerical simulation units in the chamber.
RESUMEN
Rocks are natural materials with a heterogeneous microstructure, and the heterogeneity of the microstructure plays a crucial role in the evolution of microcracks during the compression process. A numerical model of a rock with a heterogeneous structure under compression is developed by digital image processing techniques and the discrete element method. On the grain scale, the damage mechanism and microcrack characteristics of a heterogeneous Biotite granite under compression fracture are investigated. First, the process of constructing a digital image-based heterogeneous grain model is described. The microscopic characteristics of geometric heterogeneity, elastic heterogeneity, and contact heterogeneity are all considered in the numerical model. Then, the model is calibrated according to the macroscopic properties of biotite granite obtained in the laboratory, and the numerically simulated microcrack cracking processes and damage modes are obtained with a high degree of agreement compared to the experiments. Numerical simulations have shown the following: (1) Microcracking occurs first at the weak side of the grain boundaries, and the appearance of intragranular shear cracks indicates that the rock has reached its peak strength. (2) The stress concentration caused by the heterogeneity of the microstructure is an essential factor that causes rock cracks and induces rupture. Intragranular cracks occur successively in quartz, feldspar (plagioclase), and biotite, with far more intragranular cracks in quartz and feldspar (plagioclase) than in biotite. (3) Microcracking in quartz occurs as clusters, fork and fracture features, and in feldspar (plagioclase) it tends to cause penetration microcracking, which usually surrounds or terminates at the biotite. (4) As the confining pressure increases, the tensile break between the grains is suppressed and the number of shear cracks increases. At the macro level, the rock failure mode of the numerical model changes from split damage to shear destruction, which is consistent with the law shown in laboratory experiments.
RESUMEN
In the study of rock mechanics, the variation of rock mechanical characteristics in high-temperature environments is always a major issue. The discrete element method and Voronoi modeling method were used to study the mechanical characteristics and crack evolution of granite specimens subjected to the high temperature and uniaxial compression test in order to study the internal crack evolution process of granite under the influence of high temperatures. Meanwhile, dependable findings were acquired when compared to experimental outcomes. A modified failure criterion was devised, and a Fish function was built to examine the evolution behavior of tensile and shear cracks during uniaxial compression, in order to better understand the evolution process of micro-cracks in granite specimens. Shear contacts occurred first, and the number of shear cracks reached its maximum value earliest, according to the findings. The number of tensile contacts then rapidly grew, whereas the number of shear cracks steadily declined. Furthermore, it was found that when temperature rises, the number of early tensile cracks grows. This study develops a fracture prediction system for rock engineering in high-temperature conditions.
RESUMEN
Based on Lemaitre's strain equivalence hypothesis theory, it is assumed that the strength of acid-etching rock microelements under the coupling effect of temperature and confining pressure follows the Weibull distribution. Under the hypothesis that micro-element damage meets the D-P criterion and based on continuum damage mechanics and statistical theory, chemical damage variables, thermal damage variables and mechanical damage variables were introduced in the construction of damage evolution equations and constitutive models for acid-etching rocks considering the coupled effects of temperature and confining pressure. The required model parameters were obtained by theoretical derivation, and the model was verified based on the triaxial compression test data of granite. Comparing the experimental stress-strain curve with the theoretical stress-strain curve, the results show that they were in good agreement. By selecting reasonable model parameters, the damage statistical constitutive model can accurately reflect the stress-strain curve characteristics of rock in the process of triaxial compression. The comparison between the experimental and theoretical results also verifies the reasonableness and reliability of the model. This model provides a new rock damage statistical constitutive equation for the study of rock mechanics and its application in engineering, and has certain reference significance for rock underground engineering.
RESUMEN
For thermal and loaded rock in engineering structures for some projects, triple-shear Drucker-Prager yield criteria, compaction coefficient K, damage variable correction factor δ, and thermal damage variable DT are introduced in a new thermomechanical (TM) constitutive model for the entire process. The compaction stage of rock in uniaxial compression test and the strain softening of rock caused by thermal attack are considered in this article. The damage evolution of rocks is described by a damage variable and a constitutive equation, which are in agreement with the actual thermal experimental breakage. The uniaxial compressive strength of granite subjected to a TM coupling effect can be predicted properly by this new unified constitutive model. The new TM unified constitutive model considering the compaction stage and post-failure stage is in good agreement with the test curves throughout the entire process. The coupling effect of heat and load in the total damage of rock has obvious nonlinear properties, but the coupling effect significantly weakens the specimens. By using the new TM unified constitutive model, the whole process of changes in rock damage with strain after high temperature can be calculated. Meanwhile, the model well represents the stress-strain curve at the post-failure stage. It is expected that this model can provide references for studying the mechanical response of the rock damage propagation characteristics in the future.